projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
vti6: better validate user provided tunnel names
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / net / ipv6 / sit.c
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
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78
79 static unsigned int sit_net_id __read_mostly;
80 struct 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 */
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 struct net_device *dev, __be32 remote, __be32 local)
95 {
96 unsigned int h0 = HASH(remote);
97 unsigned int h1 = HASH(local);
98 struct ip_tunnel *t;
99 struct sit_net *sitn = net_generic(net, sit_net_id);
100
101 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 if (local == t->parms.iph.saddr &&
103 remote == t->parms.iph.daddr &&
104 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
106 return t;
107 }
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111 (t->dev->flags & IFF_UP))
112 return t;
113 }
114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 if (local == t->parms.iph.saddr &&
116 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
118 return t;
119 }
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if (t && (t->dev->flags & IFF_UP))
122 return t;
123 return NULL;
124 }
125
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 struct ip_tunnel_parm *parms)
128 {
129 __be32 remote = parms->iph.daddr;
130 __be32 local = parms->iph.saddr;
131 unsigned int h = 0;
132 int prio = 0;
133
134 if (remote) {
135 prio |= 2;
136 h ^= HASH(remote);
137 }
138 if (local) {
139 prio |= 1;
140 h ^= HASH(local);
141 }
142 return &sitn->tunnels[prio][h];
143 }
144
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146 struct ip_tunnel *t)
147 {
148 return __ipip6_bucket(sitn, &t->parms);
149 }
150
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152 {
153 struct ip_tunnel __rcu **tp;
154 struct ip_tunnel *iter;
155
156 for (tp = ipip6_bucket(sitn, t);
157 (iter = rtnl_dereference(*tp)) != NULL;
158 tp = &iter->next) {
159 if (t == iter) {
160 rcu_assign_pointer(*tp, t->next);
161 break;
162 }
163 }
164 }
165
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167 {
168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169
170 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 rcu_assign_pointer(*tp, t);
172 }
173
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175 {
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel *t = netdev_priv(dev);
178
179 if (dev == sitn->fb_tunnel_dev) {
180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 t->ip6rd.relay_prefix = 0;
182 t->ip6rd.prefixlen = 16;
183 t->ip6rd.relay_prefixlen = 0;
184 } else {
185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187 }
188 #endif
189 }
190
191 static int ipip6_tunnel_create(struct net_device *dev)
192 {
193 struct ip_tunnel *t = netdev_priv(dev);
194 struct net *net = dev_net(dev);
195 struct sit_net *sitn = net_generic(net, sit_net_id);
196 int err;
197
198 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
199 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
200
201 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
202 dev->priv_flags |= IFF_ISATAP;
203
204 dev->rtnl_link_ops = &sit_link_ops;
205
206 err = register_netdevice(dev);
207 if (err < 0)
208 goto out;
209
210 ipip6_tunnel_clone_6rd(dev, sitn);
211
212 dev_hold(dev);
213
214 ipip6_tunnel_link(sitn, t);
215 return 0;
216
217 out:
218 return err;
219 }
220
221 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
222 struct ip_tunnel_parm *parms, int create)
223 {
224 __be32 remote = parms->iph.daddr;
225 __be32 local = parms->iph.saddr;
226 struct ip_tunnel *t, *nt;
227 struct ip_tunnel __rcu **tp;
228 struct net_device *dev;
229 char name[IFNAMSIZ];
230 struct sit_net *sitn = net_generic(net, sit_net_id);
231
232 for (tp = __ipip6_bucket(sitn, parms);
233 (t = rtnl_dereference(*tp)) != NULL;
234 tp = &t->next) {
235 if (local == t->parms.iph.saddr &&
236 remote == t->parms.iph.daddr &&
237 parms->link == t->parms.link) {
238 if (create)
239 return NULL;
240 else
241 return t;
242 }
243 }
244 if (!create)
245 goto failed;
246
247 if (parms->name[0]) {
248 if (!dev_valid_name(parms->name))
249 goto failed;
250 strlcpy(name, parms->name, IFNAMSIZ);
251 } else {
252 strcpy(name, "sit%d");
253 }
254 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
255 ipip6_tunnel_setup);
256 if (!dev)
257 return NULL;
258
259 dev_net_set(dev, net);
260
261 nt = netdev_priv(dev);
262
263 nt->parms = *parms;
264 if (ipip6_tunnel_create(dev) < 0)
265 goto failed_free;
266
267 return nt;
268
269 failed_free:
270 free_netdev(dev);
271 failed:
272 return NULL;
273 }
274
275 #define for_each_prl_rcu(start) \
276 for (prl = rcu_dereference(start); \
277 prl; \
278 prl = rcu_dereference(prl->next))
279
280 static struct ip_tunnel_prl_entry *
281 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
282 {
283 struct ip_tunnel_prl_entry *prl;
284
285 for_each_prl_rcu(t->prl)
286 if (prl->addr == addr)
287 break;
288 return prl;
289
290 }
291
292 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
293 struct ip_tunnel_prl __user *a)
294 {
295 struct ip_tunnel_prl kprl, *kp;
296 struct ip_tunnel_prl_entry *prl;
297 unsigned int cmax, c = 0, ca, len;
298 int ret = 0;
299
300 if (copy_from_user(&kprl, a, sizeof(kprl)))
301 return -EFAULT;
302 cmax = kprl.datalen / sizeof(kprl);
303 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
304 cmax = 1;
305
306 /* For simple GET or for root users,
307 * we try harder to allocate.
308 */
309 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
310 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
311 NULL;
312
313 rcu_read_lock();
314
315 ca = t->prl_count < cmax ? t->prl_count : cmax;
316
317 if (!kp) {
318 /* We don't try hard to allocate much memory for
319 * non-root users.
320 * For root users, retry allocating enough memory for
321 * the answer.
322 */
323 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
324 if (!kp) {
325 ret = -ENOMEM;
326 goto out;
327 }
328 }
329
330 c = 0;
331 for_each_prl_rcu(t->prl) {
332 if (c >= cmax)
333 break;
334 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
335 continue;
336 kp[c].addr = prl->addr;
337 kp[c].flags = prl->flags;
338 c++;
339 if (kprl.addr != htonl(INADDR_ANY))
340 break;
341 }
342 out:
343 rcu_read_unlock();
344
345 len = sizeof(*kp) * c;
346 ret = 0;
347 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
348 ret = -EFAULT;
349
350 kfree(kp);
351
352 return ret;
353 }
354
355 static int
356 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
357 {
358 struct ip_tunnel_prl_entry *p;
359 int err = 0;
360
361 if (a->addr == htonl(INADDR_ANY))
362 return -EINVAL;
363
364 ASSERT_RTNL();
365
366 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
367 if (p->addr == a->addr) {
368 if (chg) {
369 p->flags = a->flags;
370 goto out;
371 }
372 err = -EEXIST;
373 goto out;
374 }
375 }
376
377 if (chg) {
378 err = -ENXIO;
379 goto out;
380 }
381
382 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
383 if (!p) {
384 err = -ENOBUFS;
385 goto out;
386 }
387
388 p->next = t->prl;
389 p->addr = a->addr;
390 p->flags = a->flags;
391 t->prl_count++;
392 rcu_assign_pointer(t->prl, p);
393 out:
394 return err;
395 }
396
397 static void prl_list_destroy_rcu(struct rcu_head *head)
398 {
399 struct ip_tunnel_prl_entry *p, *n;
400
401 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
402 do {
403 n = rcu_dereference_protected(p->next, 1);
404 kfree(p);
405 p = n;
406 } while (p);
407 }
408
409 static int
410 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
411 {
412 struct ip_tunnel_prl_entry *x;
413 struct ip_tunnel_prl_entry __rcu **p;
414 int err = 0;
415
416 ASSERT_RTNL();
417
418 if (a && a->addr != htonl(INADDR_ANY)) {
419 for (p = &t->prl;
420 (x = rtnl_dereference(*p)) != NULL;
421 p = &x->next) {
422 if (x->addr == a->addr) {
423 *p = x->next;
424 kfree_rcu(x, rcu_head);
425 t->prl_count--;
426 goto out;
427 }
428 }
429 err = -ENXIO;
430 } else {
431 x = rtnl_dereference(t->prl);
432 if (x) {
433 t->prl_count = 0;
434 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
435 t->prl = NULL;
436 }
437 }
438 out:
439 return err;
440 }
441
442 static int
443 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
444 {
445 struct ip_tunnel_prl_entry *p;
446 int ok = 1;
447
448 rcu_read_lock();
449 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
450 if (p) {
451 if (p->flags & PRL_DEFAULT)
452 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
453 else
454 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
455 } else {
456 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
457
458 if (ipv6_addr_is_isatap(addr6) &&
459 (addr6->s6_addr32[3] == iph->saddr) &&
460 ipv6_chk_prefix(addr6, t->dev))
461 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
462 else
463 ok = 0;
464 }
465 rcu_read_unlock();
466 return ok;
467 }
468
469 static void ipip6_tunnel_uninit(struct net_device *dev)
470 {
471 struct ip_tunnel *tunnel = netdev_priv(dev);
472 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
473
474 if (dev == sitn->fb_tunnel_dev) {
475 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
476 } else {
477 ipip6_tunnel_unlink(sitn, tunnel);
478 ipip6_tunnel_del_prl(tunnel, NULL);
479 }
480 dst_cache_reset(&tunnel->dst_cache);
481 dev_put(dev);
482 }
483
484 static int ipip6_err(struct sk_buff *skb, u32 info)
485 {
486 const struct iphdr *iph = (const struct iphdr *)skb->data;
487 const int type = icmp_hdr(skb)->type;
488 const int code = icmp_hdr(skb)->code;
489 unsigned int data_len = 0;
490 struct ip_tunnel *t;
491 int err;
492
493 switch (type) {
494 default:
495 case ICMP_PARAMETERPROB:
496 return 0;
497
498 case ICMP_DEST_UNREACH:
499 switch (code) {
500 case ICMP_SR_FAILED:
501 /* Impossible event. */
502 return 0;
503 default:
504 /* All others are translated to HOST_UNREACH.
505 rfc2003 contains "deep thoughts" about NET_UNREACH,
506 I believe they are just ether pollution. --ANK
507 */
508 break;
509 }
510 break;
511 case ICMP_TIME_EXCEEDED:
512 if (code != ICMP_EXC_TTL)
513 return 0;
514 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
515 break;
516 case ICMP_REDIRECT:
517 break;
518 }
519
520 err = -ENOENT;
521
522 t = ipip6_tunnel_lookup(dev_net(skb->dev),
523 skb->dev,
524 iph->daddr,
525 iph->saddr);
526 if (!t)
527 goto out;
528
529 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
530 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
531 t->parms.link, 0, iph->protocol, 0);
532 err = 0;
533 goto out;
534 }
535 if (type == ICMP_REDIRECT) {
536 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
537 iph->protocol, 0);
538 err = 0;
539 goto out;
540 }
541
542 err = 0;
543 if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
544 goto out;
545
546 if (t->parms.iph.daddr == 0)
547 goto out;
548
549 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
550 goto out;
551
552 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
553 t->err_count++;
554 else
555 t->err_count = 1;
556 t->err_time = jiffies;
557 out:
558 return err;
559 }
560
561 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
562 const struct in6_addr *v6addr)
563 {
564 __be32 v4embed = 0;
565 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
566 return true;
567 return false;
568 }
569
570 /* Checks if an address matches an address on the tunnel interface.
571 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
572 * Long story:
573 * This function is called after we considered the packet as spoofed
574 * in is_spoofed_6rd.
575 * We may have a router that is doing NAT for proto 41 packets
576 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
577 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
578 * function will return true, dropping the packet.
579 * But, we can still check if is spoofed against the IP
580 * addresses associated with the interface.
581 */
582 static bool only_dnatted(const struct ip_tunnel *tunnel,
583 const struct in6_addr *v6dst)
584 {
585 int prefix_len;
586
587 #ifdef CONFIG_IPV6_SIT_6RD
588 prefix_len = tunnel->ip6rd.prefixlen + 32
589 - tunnel->ip6rd.relay_prefixlen;
590 #else
591 prefix_len = 48;
592 #endif
593 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
594 }
595
596 /* Returns true if a packet is spoofed */
597 static bool packet_is_spoofed(struct sk_buff *skb,
598 const struct iphdr *iph,
599 struct ip_tunnel *tunnel)
600 {
601 const struct ipv6hdr *ipv6h;
602
603 if (tunnel->dev->priv_flags & IFF_ISATAP) {
604 if (!isatap_chksrc(skb, iph, tunnel))
605 return true;
606
607 return false;
608 }
609
610 if (tunnel->dev->flags & IFF_POINTOPOINT)
611 return false;
612
613 ipv6h = ipv6_hdr(skb);
614
615 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
616 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
617 &iph->saddr, &ipv6h->saddr,
618 &iph->daddr, &ipv6h->daddr);
619 return true;
620 }
621
622 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
623 return false;
624
625 if (only_dnatted(tunnel, &ipv6h->daddr))
626 return false;
627
628 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
629 &iph->saddr, &ipv6h->saddr,
630 &iph->daddr, &ipv6h->daddr);
631 return true;
632 }
633
634 static int ipip6_rcv(struct sk_buff *skb)
635 {
636 const struct iphdr *iph = ip_hdr(skb);
637 struct ip_tunnel *tunnel;
638 int err;
639
640 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
641 iph->saddr, iph->daddr);
642 if (tunnel) {
643 struct pcpu_sw_netstats *tstats;
644
645 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
646 tunnel->parms.iph.protocol != 0)
647 goto out;
648
649 skb->mac_header = skb->network_header;
650 skb_reset_network_header(skb);
651 IPCB(skb)->flags = 0;
652 skb->dev = tunnel->dev;
653
654 if (packet_is_spoofed(skb, iph, tunnel)) {
655 tunnel->dev->stats.rx_errors++;
656 goto out;
657 }
658
659 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
660 !net_eq(tunnel->net, dev_net(tunnel->dev))))
661 goto out;
662
663 err = IP_ECN_decapsulate(iph, skb);
664 if (unlikely(err)) {
665 if (log_ecn_error)
666 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
667 &iph->saddr, iph->tos);
668 if (err > 1) {
669 ++tunnel->dev->stats.rx_frame_errors;
670 ++tunnel->dev->stats.rx_errors;
671 goto out;
672 }
673 }
674
675 tstats = this_cpu_ptr(tunnel->dev->tstats);
676 u64_stats_update_begin(&tstats->syncp);
677 tstats->rx_packets++;
678 tstats->rx_bytes += skb->len;
679 u64_stats_update_end(&tstats->syncp);
680
681 netif_rx(skb);
682
683 return 0;
684 }
685
686 /* no tunnel matched, let upstream know, ipsec may handle it */
687 return 1;
688 out:
689 kfree_skb(skb);
690 return 0;
691 }
692
693 static const struct tnl_ptk_info ipip_tpi = {
694 /* no tunnel info required for ipip. */
695 .proto = htons(ETH_P_IP),
696 };
697
698 #if IS_ENABLED(CONFIG_MPLS)
699 static const struct tnl_ptk_info mplsip_tpi = {
700 /* no tunnel info required for mplsip. */
701 .proto = htons(ETH_P_MPLS_UC),
702 };
703 #endif
704
705 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
706 {
707 const struct iphdr *iph;
708 struct ip_tunnel *tunnel;
709
710 iph = ip_hdr(skb);
711 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
712 iph->saddr, iph->daddr);
713 if (tunnel) {
714 const struct tnl_ptk_info *tpi;
715
716 if (tunnel->parms.iph.protocol != ipproto &&
717 tunnel->parms.iph.protocol != 0)
718 goto drop;
719
720 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
721 goto drop;
722 #if IS_ENABLED(CONFIG_MPLS)
723 if (ipproto == IPPROTO_MPLS)
724 tpi = &mplsip_tpi;
725 else
726 #endif
727 tpi = &ipip_tpi;
728 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
729 goto drop;
730 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
731 }
732
733 return 1;
734
735 drop:
736 kfree_skb(skb);
737 return 0;
738 }
739
740 static int ipip_rcv(struct sk_buff *skb)
741 {
742 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
743 }
744
745 #if IS_ENABLED(CONFIG_MPLS)
746 static int mplsip_rcv(struct sk_buff *skb)
747 {
748 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
749 }
750 #endif
751
752 /*
753 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
754 * stores the embedded IPv4 address in v4dst and returns true.
755 */
756 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
757 __be32 *v4dst)
758 {
759 #ifdef CONFIG_IPV6_SIT_6RD
760 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
761 tunnel->ip6rd.prefixlen)) {
762 unsigned int pbw0, pbi0;
763 int pbi1;
764 u32 d;
765
766 pbw0 = tunnel->ip6rd.prefixlen >> 5;
767 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
768
769 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
770 tunnel->ip6rd.relay_prefixlen;
771
772 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
773 if (pbi1 > 0)
774 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
775 (32 - pbi1);
776
777 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
778 return true;
779 }
780 #else
781 if (v6dst->s6_addr16[0] == htons(0x2002)) {
782 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
783 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
784 return true;
785 }
786 #endif
787 return false;
788 }
789
790 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
791 const struct in6_addr *v6dst)
792 {
793 __be32 dst = 0;
794 check_6rd(tunnel, v6dst, &dst);
795 return dst;
796 }
797
798 /*
799 * This function assumes it is being called from dev_queue_xmit()
800 * and that skb is filled properly by that function.
801 */
802
803 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
804 struct net_device *dev)
805 {
806 struct ip_tunnel *tunnel = netdev_priv(dev);
807 const struct iphdr *tiph = &tunnel->parms.iph;
808 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
809 u8 tos = tunnel->parms.iph.tos;
810 __be16 df = tiph->frag_off;
811 struct rtable *rt; /* Route to the other host */
812 struct net_device *tdev; /* Device to other host */
813 unsigned int max_headroom; /* The extra header space needed */
814 __be32 dst = tiph->daddr;
815 struct flowi4 fl4;
816 int mtu;
817 const struct in6_addr *addr6;
818 int addr_type;
819 u8 ttl;
820 u8 protocol = IPPROTO_IPV6;
821 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
822
823 if (tos == 1)
824 tos = ipv6_get_dsfield(iph6);
825
826 /* ISATAP (RFC4214) - must come before 6to4 */
827 if (dev->priv_flags & IFF_ISATAP) {
828 struct neighbour *neigh = NULL;
829 bool do_tx_error = false;
830
831 if (skb_dst(skb))
832 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
833
834 if (!neigh) {
835 net_dbg_ratelimited("nexthop == NULL\n");
836 goto tx_error;
837 }
838
839 addr6 = (const struct in6_addr *)&neigh->primary_key;
840 addr_type = ipv6_addr_type(addr6);
841
842 if ((addr_type & IPV6_ADDR_UNICAST) &&
843 ipv6_addr_is_isatap(addr6))
844 dst = addr6->s6_addr32[3];
845 else
846 do_tx_error = true;
847
848 neigh_release(neigh);
849 if (do_tx_error)
850 goto tx_error;
851 }
852
853 if (!dst)
854 dst = try_6rd(tunnel, &iph6->daddr);
855
856 if (!dst) {
857 struct neighbour *neigh = NULL;
858 bool do_tx_error = false;
859
860 if (skb_dst(skb))
861 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
862
863 if (!neigh) {
864 net_dbg_ratelimited("nexthop == NULL\n");
865 goto tx_error;
866 }
867
868 addr6 = (const struct in6_addr *)&neigh->primary_key;
869 addr_type = ipv6_addr_type(addr6);
870
871 if (addr_type == IPV6_ADDR_ANY) {
872 addr6 = &ipv6_hdr(skb)->daddr;
873 addr_type = ipv6_addr_type(addr6);
874 }
875
876 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
877 dst = addr6->s6_addr32[3];
878 else
879 do_tx_error = true;
880
881 neigh_release(neigh);
882 if (do_tx_error)
883 goto tx_error;
884 }
885
886 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
887 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
888 0, dst, tiph->saddr, 0, 0,
889 sock_net_uid(tunnel->net, NULL));
890 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
891
892 if (IS_ERR(rt)) {
893 dev->stats.tx_carrier_errors++;
894 goto tx_error_icmp;
895 }
896 if (rt->rt_type != RTN_UNICAST) {
897 ip_rt_put(rt);
898 dev->stats.tx_carrier_errors++;
899 goto tx_error_icmp;
900 }
901 tdev = rt->dst.dev;
902
903 if (tdev == dev) {
904 ip_rt_put(rt);
905 dev->stats.collisions++;
906 goto tx_error;
907 }
908
909 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
910 ip_rt_put(rt);
911 goto tx_error;
912 }
913
914 if (df) {
915 mtu = dst_mtu(&rt->dst) - t_hlen;
916
917 if (mtu < 68) {
918 dev->stats.collisions++;
919 ip_rt_put(rt);
920 goto tx_error;
921 }
922
923 if (mtu < IPV6_MIN_MTU) {
924 mtu = IPV6_MIN_MTU;
925 df = 0;
926 }
927
928 if (tunnel->parms.iph.daddr && skb_dst(skb))
929 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
930
931 if (skb->len > mtu && !skb_is_gso(skb)) {
932 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
933 ip_rt_put(rt);
934 goto tx_error;
935 }
936 }
937
938 if (tunnel->err_count > 0) {
939 if (time_before(jiffies,
940 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
941 tunnel->err_count--;
942 dst_link_failure(skb);
943 } else
944 tunnel->err_count = 0;
945 }
946
947 /*
948 * Okay, now see if we can stuff it in the buffer as-is.
949 */
950 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
951
952 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
953 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
954 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
955 if (!new_skb) {
956 ip_rt_put(rt);
957 dev->stats.tx_dropped++;
958 kfree_skb(skb);
959 return NETDEV_TX_OK;
960 }
961 if (skb->sk)
962 skb_set_owner_w(new_skb, skb->sk);
963 dev_kfree_skb(skb);
964 skb = new_skb;
965 iph6 = ipv6_hdr(skb);
966 }
967 ttl = tiph->ttl;
968 if (ttl == 0)
969 ttl = iph6->hop_limit;
970 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
971
972 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
973 ip_rt_put(rt);
974 goto tx_error;
975 }
976
977 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
978
979 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
980 df, !net_eq(tunnel->net, dev_net(dev)));
981 return NETDEV_TX_OK;
982
983 tx_error_icmp:
984 dst_link_failure(skb);
985 tx_error:
986 kfree_skb(skb);
987 dev->stats.tx_errors++;
988 return NETDEV_TX_OK;
989 }
990
991 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
992 struct net_device *dev, u8 ipproto)
993 {
994 struct ip_tunnel *tunnel = netdev_priv(dev);
995 const struct iphdr *tiph = &tunnel->parms.iph;
996
997 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
998 goto tx_error;
999
1000 skb_set_inner_ipproto(skb, ipproto);
1001
1002 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1003 return NETDEV_TX_OK;
1004 tx_error:
1005 kfree_skb(skb);
1006 dev->stats.tx_errors++;
1007 return NETDEV_TX_OK;
1008 }
1009
1010 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1011 struct net_device *dev)
1012 {
1013 switch (skb->protocol) {
1014 case htons(ETH_P_IP):
1015 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1016 break;
1017 case htons(ETH_P_IPV6):
1018 ipip6_tunnel_xmit(skb, dev);
1019 break;
1020 #if IS_ENABLED(CONFIG_MPLS)
1021 case htons(ETH_P_MPLS_UC):
1022 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1023 break;
1024 #endif
1025 default:
1026 goto tx_err;
1027 }
1028
1029 return NETDEV_TX_OK;
1030
1031 tx_err:
1032 dev->stats.tx_errors++;
1033 kfree_skb(skb);
1034 return NETDEV_TX_OK;
1035
1036 }
1037
1038 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1039 {
1040 struct net_device *tdev = NULL;
1041 struct ip_tunnel *tunnel;
1042 const struct iphdr *iph;
1043 struct flowi4 fl4;
1044
1045 tunnel = netdev_priv(dev);
1046 iph = &tunnel->parms.iph;
1047
1048 if (iph->daddr) {
1049 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1050 NULL,
1051 iph->daddr, iph->saddr,
1052 0, 0,
1053 IPPROTO_IPV6,
1054 RT_TOS(iph->tos),
1055 tunnel->parms.link);
1056
1057 if (!IS_ERR(rt)) {
1058 tdev = rt->dst.dev;
1059 ip_rt_put(rt);
1060 }
1061 dev->flags |= IFF_POINTOPOINT;
1062 }
1063
1064 if (!tdev && tunnel->parms.link)
1065 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1066
1067 if (tdev) {
1068 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1069
1070 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1071 dev->mtu = tdev->mtu - t_hlen;
1072 if (dev->mtu < IPV6_MIN_MTU)
1073 dev->mtu = IPV6_MIN_MTU;
1074 }
1075 }
1076
1077 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1078 __u32 fwmark)
1079 {
1080 struct net *net = t->net;
1081 struct sit_net *sitn = net_generic(net, sit_net_id);
1082
1083 ipip6_tunnel_unlink(sitn, t);
1084 synchronize_net();
1085 t->parms.iph.saddr = p->iph.saddr;
1086 t->parms.iph.daddr = p->iph.daddr;
1087 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1088 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1089 ipip6_tunnel_link(sitn, t);
1090 t->parms.iph.ttl = p->iph.ttl;
1091 t->parms.iph.tos = p->iph.tos;
1092 t->parms.iph.frag_off = p->iph.frag_off;
1093 if (t->parms.link != p->link || t->fwmark != fwmark) {
1094 t->parms.link = p->link;
1095 t->fwmark = fwmark;
1096 ipip6_tunnel_bind_dev(t->dev);
1097 }
1098 dst_cache_reset(&t->dst_cache);
1099 netdev_state_change(t->dev);
1100 }
1101
1102 #ifdef CONFIG_IPV6_SIT_6RD
1103 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1104 struct ip_tunnel_6rd *ip6rd)
1105 {
1106 struct in6_addr prefix;
1107 __be32 relay_prefix;
1108
1109 if (ip6rd->relay_prefixlen > 32 ||
1110 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1111 return -EINVAL;
1112
1113 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1114 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1115 return -EINVAL;
1116 if (ip6rd->relay_prefixlen)
1117 relay_prefix = ip6rd->relay_prefix &
1118 htonl(0xffffffffUL <<
1119 (32 - ip6rd->relay_prefixlen));
1120 else
1121 relay_prefix = 0;
1122 if (relay_prefix != ip6rd->relay_prefix)
1123 return -EINVAL;
1124
1125 t->ip6rd.prefix = prefix;
1126 t->ip6rd.relay_prefix = relay_prefix;
1127 t->ip6rd.prefixlen = ip6rd->prefixlen;
1128 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1129 dst_cache_reset(&t->dst_cache);
1130 netdev_state_change(t->dev);
1131 return 0;
1132 }
1133 #endif
1134
1135 static bool ipip6_valid_ip_proto(u8 ipproto)
1136 {
1137 return ipproto == IPPROTO_IPV6 ||
1138 ipproto == IPPROTO_IPIP ||
1139 #if IS_ENABLED(CONFIG_MPLS)
1140 ipproto == IPPROTO_MPLS ||
1141 #endif
1142 ipproto == 0;
1143 }
1144
1145 static int
1146 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1147 {
1148 int err = 0;
1149 struct ip_tunnel_parm p;
1150 struct ip_tunnel_prl prl;
1151 struct ip_tunnel *t = netdev_priv(dev);
1152 struct net *net = t->net;
1153 struct sit_net *sitn = net_generic(net, sit_net_id);
1154 #ifdef CONFIG_IPV6_SIT_6RD
1155 struct ip_tunnel_6rd ip6rd;
1156 #endif
1157
1158 switch (cmd) {
1159 case SIOCGETTUNNEL:
1160 #ifdef CONFIG_IPV6_SIT_6RD
1161 case SIOCGET6RD:
1162 #endif
1163 if (dev == sitn->fb_tunnel_dev) {
1164 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1165 err = -EFAULT;
1166 break;
1167 }
1168 t = ipip6_tunnel_locate(net, &p, 0);
1169 if (!t)
1170 t = netdev_priv(dev);
1171 }
1172
1173 err = -EFAULT;
1174 if (cmd == SIOCGETTUNNEL) {
1175 memcpy(&p, &t->parms, sizeof(p));
1176 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1177 sizeof(p)))
1178 goto done;
1179 #ifdef CONFIG_IPV6_SIT_6RD
1180 } else {
1181 ip6rd.prefix = t->ip6rd.prefix;
1182 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1183 ip6rd.prefixlen = t->ip6rd.prefixlen;
1184 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1185 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1186 sizeof(ip6rd)))
1187 goto done;
1188 #endif
1189 }
1190 err = 0;
1191 break;
1192
1193 case SIOCADDTUNNEL:
1194 case SIOCCHGTUNNEL:
1195 err = -EPERM;
1196 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1197 goto done;
1198
1199 err = -EFAULT;
1200 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1201 goto done;
1202
1203 err = -EINVAL;
1204 if (!ipip6_valid_ip_proto(p.iph.protocol))
1205 goto done;
1206 if (p.iph.version != 4 ||
1207 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1208 goto done;
1209 if (p.iph.ttl)
1210 p.iph.frag_off |= htons(IP_DF);
1211
1212 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1213
1214 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1215 if (t) {
1216 if (t->dev != dev) {
1217 err = -EEXIST;
1218 break;
1219 }
1220 } else {
1221 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1222 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1223 err = -EINVAL;
1224 break;
1225 }
1226 t = netdev_priv(dev);
1227 }
1228
1229 ipip6_tunnel_update(t, &p, t->fwmark);
1230 }
1231
1232 if (t) {
1233 err = 0;
1234 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1235 err = -EFAULT;
1236 } else
1237 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1238 break;
1239
1240 case SIOCDELTUNNEL:
1241 err = -EPERM;
1242 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1243 goto done;
1244
1245 if (dev == sitn->fb_tunnel_dev) {
1246 err = -EFAULT;
1247 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1248 goto done;
1249 err = -ENOENT;
1250 t = ipip6_tunnel_locate(net, &p, 0);
1251 if (!t)
1252 goto done;
1253 err = -EPERM;
1254 if (t == netdev_priv(sitn->fb_tunnel_dev))
1255 goto done;
1256 dev = t->dev;
1257 }
1258 unregister_netdevice(dev);
1259 err = 0;
1260 break;
1261
1262 case SIOCGETPRL:
1263 err = -EINVAL;
1264 if (dev == sitn->fb_tunnel_dev)
1265 goto done;
1266 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1267 break;
1268
1269 case SIOCADDPRL:
1270 case SIOCDELPRL:
1271 case SIOCCHGPRL:
1272 err = -EPERM;
1273 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1274 goto done;
1275 err = -EINVAL;
1276 if (dev == sitn->fb_tunnel_dev)
1277 goto done;
1278 err = -EFAULT;
1279 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1280 goto done;
1281
1282 switch (cmd) {
1283 case SIOCDELPRL:
1284 err = ipip6_tunnel_del_prl(t, &prl);
1285 break;
1286 case SIOCADDPRL:
1287 case SIOCCHGPRL:
1288 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1289 break;
1290 }
1291 dst_cache_reset(&t->dst_cache);
1292 netdev_state_change(dev);
1293 break;
1294
1295 #ifdef CONFIG_IPV6_SIT_6RD
1296 case SIOCADD6RD:
1297 case SIOCCHG6RD:
1298 case SIOCDEL6RD:
1299 err = -EPERM;
1300 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1301 goto done;
1302
1303 err = -EFAULT;
1304 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1305 sizeof(ip6rd)))
1306 goto done;
1307
1308 if (cmd != SIOCDEL6RD) {
1309 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1310 if (err < 0)
1311 goto done;
1312 } else
1313 ipip6_tunnel_clone_6rd(dev, sitn);
1314
1315 err = 0;
1316 break;
1317 #endif
1318
1319 default:
1320 err = -EINVAL;
1321 }
1322
1323 done:
1324 return err;
1325 }
1326
1327 static const struct net_device_ops ipip6_netdev_ops = {
1328 .ndo_init = ipip6_tunnel_init,
1329 .ndo_uninit = ipip6_tunnel_uninit,
1330 .ndo_start_xmit = sit_tunnel_xmit,
1331 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1332 .ndo_get_stats64 = ip_tunnel_get_stats64,
1333 .ndo_get_iflink = ip_tunnel_get_iflink,
1334 };
1335
1336 static void ipip6_dev_free(struct net_device *dev)
1337 {
1338 struct ip_tunnel *tunnel = netdev_priv(dev);
1339
1340 dst_cache_destroy(&tunnel->dst_cache);
1341 free_percpu(dev->tstats);
1342 }
1343
1344 #define SIT_FEATURES (NETIF_F_SG | \
1345 NETIF_F_FRAGLIST | \
1346 NETIF_F_HIGHDMA | \
1347 NETIF_F_GSO_SOFTWARE | \
1348 NETIF_F_HW_CSUM)
1349
1350 static void ipip6_tunnel_setup(struct net_device *dev)
1351 {
1352 struct ip_tunnel *tunnel = netdev_priv(dev);
1353 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1354
1355 dev->netdev_ops = &ipip6_netdev_ops;
1356 dev->needs_free_netdev = true;
1357 dev->priv_destructor = ipip6_dev_free;
1358
1359 dev->type = ARPHRD_SIT;
1360 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1361 dev->mtu = ETH_DATA_LEN - t_hlen;
1362 dev->min_mtu = IPV6_MIN_MTU;
1363 dev->max_mtu = 0xFFF8 - t_hlen;
1364 dev->flags = IFF_NOARP;
1365 netif_keep_dst(dev);
1366 dev->addr_len = 4;
1367 dev->features |= NETIF_F_LLTX;
1368 dev->features |= SIT_FEATURES;
1369 dev->hw_features |= SIT_FEATURES;
1370 }
1371
1372 static int ipip6_tunnel_init(struct net_device *dev)
1373 {
1374 struct ip_tunnel *tunnel = netdev_priv(dev);
1375 int err;
1376
1377 tunnel->dev = dev;
1378 tunnel->net = dev_net(dev);
1379 strcpy(tunnel->parms.name, dev->name);
1380
1381 ipip6_tunnel_bind_dev(dev);
1382 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1383 if (!dev->tstats)
1384 return -ENOMEM;
1385
1386 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1387 if (err) {
1388 free_percpu(dev->tstats);
1389 dev->tstats = NULL;
1390 return err;
1391 }
1392
1393 return 0;
1394 }
1395
1396 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1397 {
1398 struct ip_tunnel *tunnel = netdev_priv(dev);
1399 struct iphdr *iph = &tunnel->parms.iph;
1400 struct net *net = dev_net(dev);
1401 struct sit_net *sitn = net_generic(net, sit_net_id);
1402
1403 iph->version = 4;
1404 iph->protocol = IPPROTO_IPV6;
1405 iph->ihl = 5;
1406 iph->ttl = 64;
1407
1408 dev_hold(dev);
1409 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1410 }
1411
1412 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1413 struct netlink_ext_ack *extack)
1414 {
1415 u8 proto;
1416
1417 if (!data || !data[IFLA_IPTUN_PROTO])
1418 return 0;
1419
1420 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1421 if (!ipip6_valid_ip_proto(proto))
1422 return -EINVAL;
1423
1424 return 0;
1425 }
1426
1427 static void ipip6_netlink_parms(struct nlattr *data[],
1428 struct ip_tunnel_parm *parms,
1429 __u32 *fwmark)
1430 {
1431 memset(parms, 0, sizeof(*parms));
1432
1433 parms->iph.version = 4;
1434 parms->iph.protocol = IPPROTO_IPV6;
1435 parms->iph.ihl = 5;
1436 parms->iph.ttl = 64;
1437
1438 if (!data)
1439 return;
1440
1441 if (data[IFLA_IPTUN_LINK])
1442 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1443
1444 if (data[IFLA_IPTUN_LOCAL])
1445 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1446
1447 if (data[IFLA_IPTUN_REMOTE])
1448 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1449
1450 if (data[IFLA_IPTUN_TTL]) {
1451 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1452 if (parms->iph.ttl)
1453 parms->iph.frag_off = htons(IP_DF);
1454 }
1455
1456 if (data[IFLA_IPTUN_TOS])
1457 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1458
1459 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1460 parms->iph.frag_off = htons(IP_DF);
1461
1462 if (data[IFLA_IPTUN_FLAGS])
1463 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1464
1465 if (data[IFLA_IPTUN_PROTO])
1466 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1467
1468 if (data[IFLA_IPTUN_FWMARK])
1469 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1470 }
1471
1472 /* This function returns true when ENCAP attributes are present in the nl msg */
1473 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1474 struct ip_tunnel_encap *ipencap)
1475 {
1476 bool ret = false;
1477
1478 memset(ipencap, 0, sizeof(*ipencap));
1479
1480 if (!data)
1481 return ret;
1482
1483 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1484 ret = true;
1485 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1486 }
1487
1488 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1489 ret = true;
1490 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1491 }
1492
1493 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1494 ret = true;
1495 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1496 }
1497
1498 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1499 ret = true;
1500 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1501 }
1502
1503 return ret;
1504 }
1505
1506 #ifdef CONFIG_IPV6_SIT_6RD
1507 /* This function returns true when 6RD attributes are present in the nl msg */
1508 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1509 struct ip_tunnel_6rd *ip6rd)
1510 {
1511 bool ret = false;
1512 memset(ip6rd, 0, sizeof(*ip6rd));
1513
1514 if (!data)
1515 return ret;
1516
1517 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1518 ret = true;
1519 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1520 }
1521
1522 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1523 ret = true;
1524 ip6rd->relay_prefix =
1525 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1526 }
1527
1528 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1529 ret = true;
1530 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1531 }
1532
1533 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1534 ret = true;
1535 ip6rd->relay_prefixlen =
1536 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1537 }
1538
1539 return ret;
1540 }
1541 #endif
1542
1543 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1544 struct nlattr *tb[], struct nlattr *data[],
1545 struct netlink_ext_ack *extack)
1546 {
1547 struct net *net = dev_net(dev);
1548 struct ip_tunnel *nt;
1549 struct ip_tunnel_encap ipencap;
1550 #ifdef CONFIG_IPV6_SIT_6RD
1551 struct ip_tunnel_6rd ip6rd;
1552 #endif
1553 int err;
1554
1555 nt = netdev_priv(dev);
1556
1557 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1558 err = ip_tunnel_encap_setup(nt, &ipencap);
1559 if (err < 0)
1560 return err;
1561 }
1562
1563 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1564
1565 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1566 return -EEXIST;
1567
1568 err = ipip6_tunnel_create(dev);
1569 if (err < 0)
1570 return err;
1571
1572 #ifdef CONFIG_IPV6_SIT_6RD
1573 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1574 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1575 #endif
1576
1577 return err;
1578 }
1579
1580 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1581 struct nlattr *data[],
1582 struct netlink_ext_ack *extack)
1583 {
1584 struct ip_tunnel *t = netdev_priv(dev);
1585 struct ip_tunnel_parm p;
1586 struct ip_tunnel_encap ipencap;
1587 struct net *net = t->net;
1588 struct sit_net *sitn = net_generic(net, sit_net_id);
1589 #ifdef CONFIG_IPV6_SIT_6RD
1590 struct ip_tunnel_6rd ip6rd;
1591 #endif
1592 __u32 fwmark = t->fwmark;
1593 int err;
1594
1595 if (dev == sitn->fb_tunnel_dev)
1596 return -EINVAL;
1597
1598 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1599 err = ip_tunnel_encap_setup(t, &ipencap);
1600 if (err < 0)
1601 return err;
1602 }
1603
1604 ipip6_netlink_parms(data, &p, &fwmark);
1605
1606 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1607 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1608 return -EINVAL;
1609
1610 t = ipip6_tunnel_locate(net, &p, 0);
1611
1612 if (t) {
1613 if (t->dev != dev)
1614 return -EEXIST;
1615 } else
1616 t = netdev_priv(dev);
1617
1618 ipip6_tunnel_update(t, &p, fwmark);
1619
1620 #ifdef CONFIG_IPV6_SIT_6RD
1621 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1622 return ipip6_tunnel_update_6rd(t, &ip6rd);
1623 #endif
1624
1625 return 0;
1626 }
1627
1628 static size_t ipip6_get_size(const struct net_device *dev)
1629 {
1630 return
1631 /* IFLA_IPTUN_LINK */
1632 nla_total_size(4) +
1633 /* IFLA_IPTUN_LOCAL */
1634 nla_total_size(4) +
1635 /* IFLA_IPTUN_REMOTE */
1636 nla_total_size(4) +
1637 /* IFLA_IPTUN_TTL */
1638 nla_total_size(1) +
1639 /* IFLA_IPTUN_TOS */
1640 nla_total_size(1) +
1641 /* IFLA_IPTUN_PMTUDISC */
1642 nla_total_size(1) +
1643 /* IFLA_IPTUN_FLAGS */
1644 nla_total_size(2) +
1645 /* IFLA_IPTUN_PROTO */
1646 nla_total_size(1) +
1647 #ifdef CONFIG_IPV6_SIT_6RD
1648 /* IFLA_IPTUN_6RD_PREFIX */
1649 nla_total_size(sizeof(struct in6_addr)) +
1650 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1651 nla_total_size(4) +
1652 /* IFLA_IPTUN_6RD_PREFIXLEN */
1653 nla_total_size(2) +
1654 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1655 nla_total_size(2) +
1656 #endif
1657 /* IFLA_IPTUN_ENCAP_TYPE */
1658 nla_total_size(2) +
1659 /* IFLA_IPTUN_ENCAP_FLAGS */
1660 nla_total_size(2) +
1661 /* IFLA_IPTUN_ENCAP_SPORT */
1662 nla_total_size(2) +
1663 /* IFLA_IPTUN_ENCAP_DPORT */
1664 nla_total_size(2) +
1665 /* IFLA_IPTUN_FWMARK */
1666 nla_total_size(4) +
1667 0;
1668 }
1669
1670 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1671 {
1672 struct ip_tunnel *tunnel = netdev_priv(dev);
1673 struct ip_tunnel_parm *parm = &tunnel->parms;
1674
1675 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1676 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1677 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1678 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1679 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1680 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1681 !!(parm->iph.frag_off & htons(IP_DF))) ||
1682 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1683 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1684 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1685 goto nla_put_failure;
1686
1687 #ifdef CONFIG_IPV6_SIT_6RD
1688 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1689 &tunnel->ip6rd.prefix) ||
1690 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1691 tunnel->ip6rd.relay_prefix) ||
1692 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1693 tunnel->ip6rd.prefixlen) ||
1694 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1695 tunnel->ip6rd.relay_prefixlen))
1696 goto nla_put_failure;
1697 #endif
1698
1699 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1700 tunnel->encap.type) ||
1701 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1702 tunnel->encap.sport) ||
1703 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1704 tunnel->encap.dport) ||
1705 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1706 tunnel->encap.flags))
1707 goto nla_put_failure;
1708
1709 return 0;
1710
1711 nla_put_failure:
1712 return -EMSGSIZE;
1713 }
1714
1715 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1716 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1717 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1718 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1719 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1720 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1721 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1722 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1723 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1724 #ifdef CONFIG_IPV6_SIT_6RD
1725 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1726 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1727 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1728 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1729 #endif
1730 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1731 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1732 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1733 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1734 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1735 };
1736
1737 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1738 {
1739 struct net *net = dev_net(dev);
1740 struct sit_net *sitn = net_generic(net, sit_net_id);
1741
1742 if (dev != sitn->fb_tunnel_dev)
1743 unregister_netdevice_queue(dev, head);
1744 }
1745
1746 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1747 .kind = "sit",
1748 .maxtype = IFLA_IPTUN_MAX,
1749 .policy = ipip6_policy,
1750 .priv_size = sizeof(struct ip_tunnel),
1751 .setup = ipip6_tunnel_setup,
1752 .validate = ipip6_validate,
1753 .newlink = ipip6_newlink,
1754 .changelink = ipip6_changelink,
1755 .get_size = ipip6_get_size,
1756 .fill_info = ipip6_fill_info,
1757 .dellink = ipip6_dellink,
1758 .get_link_net = ip_tunnel_get_link_net,
1759 };
1760
1761 static struct xfrm_tunnel sit_handler __read_mostly = {
1762 .handler = ipip6_rcv,
1763 .err_handler = ipip6_err,
1764 .priority = 1,
1765 };
1766
1767 static struct xfrm_tunnel ipip_handler __read_mostly = {
1768 .handler = ipip_rcv,
1769 .err_handler = ipip6_err,
1770 .priority = 2,
1771 };
1772
1773 #if IS_ENABLED(CONFIG_MPLS)
1774 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1775 .handler = mplsip_rcv,
1776 .err_handler = ipip6_err,
1777 .priority = 2,
1778 };
1779 #endif
1780
1781 static void __net_exit sit_destroy_tunnels(struct net *net,
1782 struct list_head *head)
1783 {
1784 struct sit_net *sitn = net_generic(net, sit_net_id);
1785 struct net_device *dev, *aux;
1786 int prio;
1787
1788 for_each_netdev_safe(net, dev, aux)
1789 if (dev->rtnl_link_ops == &sit_link_ops)
1790 unregister_netdevice_queue(dev, head);
1791
1792 for (prio = 1; prio < 4; prio++) {
1793 int h;
1794 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1795 struct ip_tunnel *t;
1796
1797 t = rtnl_dereference(sitn->tunnels[prio][h]);
1798 while (t) {
1799 /* If dev is in the same netns, it has already
1800 * been added to the list by the previous loop.
1801 */
1802 if (!net_eq(dev_net(t->dev), net))
1803 unregister_netdevice_queue(t->dev,
1804 head);
1805 t = rtnl_dereference(t->next);
1806 }
1807 }
1808 }
1809 }
1810
1811 static int __net_init sit_init_net(struct net *net)
1812 {
1813 struct sit_net *sitn = net_generic(net, sit_net_id);
1814 struct ip_tunnel *t;
1815 int err;
1816
1817 sitn->tunnels[0] = sitn->tunnels_wc;
1818 sitn->tunnels[1] = sitn->tunnels_l;
1819 sitn->tunnels[2] = sitn->tunnels_r;
1820 sitn->tunnels[3] = sitn->tunnels_r_l;
1821
1822 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1823 NET_NAME_UNKNOWN,
1824 ipip6_tunnel_setup);
1825 if (!sitn->fb_tunnel_dev) {
1826 err = -ENOMEM;
1827 goto err_alloc_dev;
1828 }
1829 dev_net_set(sitn->fb_tunnel_dev, net);
1830 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1831 /* FB netdevice is special: we have one, and only one per netns.
1832 * Allowing to move it to another netns is clearly unsafe.
1833 */
1834 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1835
1836 err = register_netdev(sitn->fb_tunnel_dev);
1837 if (err)
1838 goto err_reg_dev;
1839
1840 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1841 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1842
1843 t = netdev_priv(sitn->fb_tunnel_dev);
1844
1845 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1846 return 0;
1847
1848 err_reg_dev:
1849 ipip6_dev_free(sitn->fb_tunnel_dev);
1850 err_alloc_dev:
1851 return err;
1852 }
1853
1854 static void __net_exit sit_exit_net(struct net *net)
1855 {
1856 LIST_HEAD(list);
1857
1858 rtnl_lock();
1859 sit_destroy_tunnels(net, &list);
1860 unregister_netdevice_many(&list);
1861 rtnl_unlock();
1862 }
1863
1864 static struct pernet_operations sit_net_ops = {
1865 .init = sit_init_net,
1866 .exit = sit_exit_net,
1867 .id = &sit_net_id,
1868 .size = sizeof(struct sit_net),
1869 };
1870
1871 static void __exit sit_cleanup(void)
1872 {
1873 rtnl_link_unregister(&sit_link_ops);
1874 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1875 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1876 #if IS_ENABLED(CONFIG_MPLS)
1877 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1878 #endif
1879
1880 unregister_pernet_device(&sit_net_ops);
1881 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1882 }
1883
1884 static int __init sit_init(void)
1885 {
1886 int err;
1887
1888 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1889
1890 err = register_pernet_device(&sit_net_ops);
1891 if (err < 0)
1892 return err;
1893 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1894 if (err < 0) {
1895 pr_info("%s: can't register ip6ip4\n", __func__);
1896 goto xfrm_tunnel_failed;
1897 }
1898 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1899 if (err < 0) {
1900 pr_info("%s: can't register ip4ip4\n", __func__);
1901 goto xfrm_tunnel4_failed;
1902 }
1903 #if IS_ENABLED(CONFIG_MPLS)
1904 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1905 if (err < 0) {
1906 pr_info("%s: can't register mplsip\n", __func__);
1907 goto xfrm_tunnel_mpls_failed;
1908 }
1909 #endif
1910 err = rtnl_link_register(&sit_link_ops);
1911 if (err < 0)
1912 goto rtnl_link_failed;
1913
1914 out:
1915 return err;
1916
1917 rtnl_link_failed:
1918 #if IS_ENABLED(CONFIG_MPLS)
1919 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1920 xfrm_tunnel_mpls_failed:
1921 #endif
1922 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1923 xfrm_tunnel4_failed:
1924 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1925 xfrm_tunnel_failed:
1926 unregister_pernet_device(&sit_net_ops);
1927 goto out;
1928 }
1929
1930 module_init(sit_init);
1931 module_exit(sit_cleanup);
1932 MODULE_LICENSE("GPL");
1933 MODULE_ALIAS_RTNL_LINK("sit");
1934 MODULE_ALIAS_NETDEV("sit0");