Merge tag 'mips_fixes_5.2_2' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
[linux-2.6/linux-2.6-arm.git] / net / ipv6 / sit.c
blob80610899a323056e8e7fe1a71363b605b72cb218
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
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
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
35 #include <net/sock.h>
36 #include <net/snmp.h>
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>
56 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 For comments look at net/ipv4/ip_gre.c --ANK
61 #define IP6_SIT_HASH_SIZE 16
62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 static bool log_ecn_error = true;
65 module_param(log_ecn_error, bool, 0644);
66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
72 __be32 *v4dst);
73 static struct rtnl_link_ops sit_link_ops __read_mostly;
75 static unsigned int sit_net_id __read_mostly;
76 struct 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];
83 struct net_device *fb_tunnel_dev;
87 * Must be invoked with rcu_read_lock
89 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
90 struct net_device *dev,
91 __be32 remote, __be32 local,
92 int sifindex)
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;
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;
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;
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;
122 t = rcu_dereference(sitn->tunnels_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
124 return t;
125 return NULL;
128 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
129 struct ip_tunnel_parm *parms)
131 __be32 remote = parms->iph.daddr;
132 __be32 local = parms->iph.saddr;
133 unsigned int h = 0;
134 int prio = 0;
136 if (remote) {
137 prio |= 2;
138 h ^= HASH(remote);
140 if (local) {
141 prio |= 1;
142 h ^= HASH(local);
144 return &sitn->tunnels[prio][h];
147 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
148 struct ip_tunnel *t)
150 return __ipip6_bucket(sitn, &t->parms);
153 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
155 struct ip_tunnel __rcu **tp;
156 struct ip_tunnel *iter;
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;
168 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
170 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
172 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
173 rcu_assign_pointer(*tp, t);
176 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
178 #ifdef CONFIG_IPV6_SIT_6RD
179 struct ip_tunnel *t = netdev_priv(dev);
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));
190 #endif
193 static int ipip6_tunnel_create(struct net_device *dev)
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;
200 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
201 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
203 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 dev->priv_flags |= IFF_ISATAP;
206 dev->rtnl_link_ops = &sit_link_ops;
208 err = register_netdevice(dev);
209 if (err < 0)
210 goto out;
212 ipip6_tunnel_clone_6rd(dev, sitn);
214 dev_hold(dev);
216 ipip6_tunnel_link(sitn, t);
217 return 0;
219 out:
220 return err;
223 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
224 struct ip_tunnel_parm *parms, int create)
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);
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;
246 if (!create)
247 goto failed;
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");
256 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
257 ipip6_tunnel_setup);
258 if (!dev)
259 return NULL;
261 dev_net_set(dev, net);
263 nt = netdev_priv(dev);
265 nt->parms = *parms;
266 if (ipip6_tunnel_create(dev) < 0)
267 goto failed_free;
269 return nt;
271 failed_free:
272 free_netdev(dev);
273 failed:
274 return NULL;
277 #define for_each_prl_rcu(start) \
278 for (prl = rcu_dereference(start); \
279 prl; \
280 prl = rcu_dereference(prl->next))
282 static struct ip_tunnel_prl_entry *
283 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
285 struct ip_tunnel_prl_entry *prl;
287 for_each_prl_rcu(t->prl)
288 if (prl->addr == addr)
289 break;
290 return prl;
294 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
295 struct ip_tunnel_prl __user *a)
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;
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;
308 /* For simple GET or for root users,
309 * we try harder to allocate.
311 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
312 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
313 NULL;
315 rcu_read_lock();
317 ca = t->prl_count < cmax ? t->prl_count : cmax;
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.
325 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
326 if (!kp) {
327 ret = -ENOMEM;
328 goto out;
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;
344 out:
345 rcu_read_unlock();
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;
352 kfree(kp);
354 return ret;
357 static int
358 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
360 struct ip_tunnel_prl_entry *p;
361 int err = 0;
363 if (a->addr == htonl(INADDR_ANY))
364 return -EINVAL;
366 ASSERT_RTNL();
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;
374 err = -EEXIST;
375 goto out;
379 if (chg) {
380 err = -ENXIO;
381 goto out;
384 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
385 if (!p) {
386 err = -ENOBUFS;
387 goto out;
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);
395 out:
396 return err;
399 static void prl_list_destroy_rcu(struct rcu_head *head)
401 struct ip_tunnel_prl_entry *p, *n;
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);
411 static int
412 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
414 struct ip_tunnel_prl_entry *x;
415 struct ip_tunnel_prl_entry __rcu **p;
416 int err = 0;
418 ASSERT_RTNL();
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;
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;
440 out:
441 return err;
444 static int
445 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
447 struct ip_tunnel_prl_entry *p;
448 int ok = 1;
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;
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;
467 rcu_read_unlock();
468 return ok;
471 static void ipip6_tunnel_uninit(struct net_device *dev)
473 struct ip_tunnel *tunnel = netdev_priv(dev);
474 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
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);
482 dst_cache_reset(&tunnel->dst_cache);
483 dev_put(dev);
486 static int ipip6_err(struct sk_buff *skb, u32 info)
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;
496 switch (type) {
497 default:
498 case ICMP_PARAMETERPROB:
499 return 0;
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
511 break;
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;
523 err = -ENOENT;
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;
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;
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;
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;
549 if (t->parms.iph.daddr == 0)
550 goto out;
552 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
553 goto out;
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;
560 out:
561 return err;
564 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
565 const struct in6_addr *v6addr)
567 __be32 v4embed = 0;
568 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
569 return true;
570 return false;
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.
585 static bool only_dnatted(const struct ip_tunnel *tunnel,
586 const struct in6_addr *v6dst)
588 int prefix_len;
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);
599 /* Returns true if a packet is spoofed */
600 static bool packet_is_spoofed(struct sk_buff *skb,
601 const struct iphdr *iph,
602 struct ip_tunnel *tunnel)
604 const struct ipv6hdr *ipv6h;
606 if (tunnel->dev->priv_flags & IFF_ISATAP) {
607 if (!isatap_chksrc(skb, iph, tunnel))
608 return true;
610 return false;
613 if (tunnel->dev->flags & IFF_POINTOPOINT)
614 return false;
616 ipv6h = ipv6_hdr(skb);
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;
625 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
626 return false;
628 if (only_dnatted(tunnel, &ipv6h->daddr))
629 return false;
631 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
632 &iph->saddr, &ipv6h->saddr,
633 &iph->daddr, &ipv6h->daddr);
634 return true;
637 static int ipip6_rcv(struct sk_buff *skb)
639 const struct iphdr *iph = ip_hdr(skb);
640 struct ip_tunnel *tunnel;
641 int sifindex;
642 int err;
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;
650 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
651 tunnel->parms.iph.protocol != 0)
652 goto out;
654 skb->mac_header = skb->network_header;
655 skb_reset_network_header(skb);
656 IPCB(skb)->flags = 0;
657 skb->dev = tunnel->dev;
659 if (packet_is_spoofed(skb, iph, tunnel)) {
660 tunnel->dev->stats.rx_errors++;
661 goto out;
664 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
665 !net_eq(tunnel->net, dev_net(tunnel->dev))))
666 goto out;
668 /* skb can be uncloned in iptunnel_pull_header, so
669 * old iph is no longer valid
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;
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);
690 netif_rx(skb);
692 return 0;
695 /* no tunnel matched, let upstream know, ipsec may handle it */
696 return 1;
697 out:
698 kfree_skb(skb);
699 return 0;
702 static const struct tnl_ptk_info ipip_tpi = {
703 /* no tunnel info required for ipip. */
704 .proto = htons(ETH_P_IP),
707 #if IS_ENABLED(CONFIG_MPLS)
708 static const struct tnl_ptk_info mplsip_tpi = {
709 /* no tunnel info required for mplsip. */
710 .proto = htons(ETH_P_MPLS_UC),
712 #endif
714 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
716 const struct iphdr *iph;
717 struct ip_tunnel *tunnel;
718 int sifindex;
720 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
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;
728 if (tunnel->parms.iph.protocol != ipproto &&
729 tunnel->parms.iph.protocol != 0)
730 goto drop;
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);
745 return 1;
747 drop:
748 kfree_skb(skb);
749 return 0;
752 static int ipip_rcv(struct sk_buff *skb)
754 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
757 #if IS_ENABLED(CONFIG_MPLS)
758 static int mplsip_rcv(struct sk_buff *skb)
760 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
762 #endif
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.
768 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
769 __be32 *v4dst)
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;
778 pbw0 = tunnel->ip6rd.prefixlen >> 5;
779 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
781 d = tunnel->ip6rd.relay_prefixlen < 32 ?
782 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
783 tunnel->ip6rd.relay_prefixlen : 0;
785 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
786 if (pbi1 > 0)
787 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
788 (32 - pbi1);
790 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
791 return true;
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;
799 #endif
800 return false;
803 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
804 const struct in6_addr *v6dst)
806 __be32 dst = 0;
807 check_6rd(tunnel, v6dst, &dst);
808 return dst;
812 * This function assumes it is being called from dev_queue_xmit()
813 * and that skb is filled properly by that function.
816 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
817 struct net_device *dev)
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);
836 if (tos == 1)
837 tos = ipv6_get_dsfield(iph6);
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;
844 if (skb_dst(skb))
845 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
847 if (!neigh) {
848 net_dbg_ratelimited("nexthop == NULL\n");
849 goto tx_error;
852 addr6 = (const struct in6_addr *)&neigh->primary_key;
853 addr_type = ipv6_addr_type(addr6);
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;
861 neigh_release(neigh);
862 if (do_tx_error)
863 goto tx_error;
866 if (!dst)
867 dst = try_6rd(tunnel, &iph6->daddr);
869 if (!dst) {
870 struct neighbour *neigh = NULL;
871 bool do_tx_error = false;
873 if (skb_dst(skb))
874 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
876 if (!neigh) {
877 net_dbg_ratelimited("nexthop == NULL\n");
878 goto tx_error;
881 addr6 = (const struct in6_addr *)&neigh->primary_key;
882 addr_type = ipv6_addr_type(addr6);
884 if (addr_type == IPV6_ADDR_ANY) {
885 addr6 = &ipv6_hdr(skb)->daddr;
886 addr_type = ipv6_addr_type(addr6);
889 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
890 dst = addr6->s6_addr32[3];
891 else
892 do_tx_error = true;
894 neigh_release(neigh);
895 if (do_tx_error)
896 goto tx_error;
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 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
905 if (IS_ERR(rt)) {
906 dev->stats.tx_carrier_errors++;
907 goto tx_error_icmp;
909 if (rt->rt_type != RTN_UNICAST) {
910 ip_rt_put(rt);
911 dev->stats.tx_carrier_errors++;
912 goto tx_error_icmp;
914 tdev = rt->dst.dev;
916 if (tdev == dev) {
917 ip_rt_put(rt);
918 dev->stats.collisions++;
919 goto tx_error;
922 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
923 ip_rt_put(rt);
924 goto tx_error;
927 if (df) {
928 mtu = dst_mtu(&rt->dst) - t_hlen;
930 if (mtu < 68) {
931 dev->stats.collisions++;
932 ip_rt_put(rt);
933 goto tx_error;
936 if (mtu < IPV6_MIN_MTU) {
937 mtu = IPV6_MIN_MTU;
938 df = 0;
941 if (tunnel->parms.iph.daddr)
942 skb_dst_update_pmtu(skb, mtu);
944 if (skb->len > mtu && !skb_is_gso(skb)) {
945 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
946 ip_rt_put(rt);
947 goto tx_error;
951 if (tunnel->err_count > 0) {
952 if (time_before(jiffies,
953 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
954 tunnel->err_count--;
955 dst_link_failure(skb);
956 } else
957 tunnel->err_count = 0;
961 * Okay, now see if we can stuff it in the buffer as-is.
963 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
965 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
966 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
967 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
968 if (!new_skb) {
969 ip_rt_put(rt);
970 dev->stats.tx_dropped++;
971 kfree_skb(skb);
972 return NETDEV_TX_OK;
974 if (skb->sk)
975 skb_set_owner_w(new_skb, skb->sk);
976 dev_kfree_skb(skb);
977 skb = new_skb;
978 iph6 = ipv6_hdr(skb);
980 ttl = tiph->ttl;
981 if (ttl == 0)
982 ttl = iph6->hop_limit;
983 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
985 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
986 ip_rt_put(rt);
987 goto tx_error;
990 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
992 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
993 df, !net_eq(tunnel->net, dev_net(dev)));
994 return NETDEV_TX_OK;
996 tx_error_icmp:
997 dst_link_failure(skb);
998 tx_error:
999 kfree_skb(skb);
1000 dev->stats.tx_errors++;
1001 return NETDEV_TX_OK;
1004 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1005 struct net_device *dev, u8 ipproto)
1007 struct ip_tunnel *tunnel = netdev_priv(dev);
1008 const struct iphdr *tiph = &tunnel->parms.iph;
1010 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1011 goto tx_error;
1013 skb_set_inner_ipproto(skb, ipproto);
1015 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1016 return NETDEV_TX_OK;
1017 tx_error:
1018 kfree_skb(skb);
1019 dev->stats.tx_errors++;
1020 return NETDEV_TX_OK;
1023 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1024 struct net_device *dev)
1026 if (!pskb_inet_may_pull(skb))
1027 goto tx_err;
1029 switch (skb->protocol) {
1030 case htons(ETH_P_IP):
1031 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1032 break;
1033 case htons(ETH_P_IPV6):
1034 ipip6_tunnel_xmit(skb, dev);
1035 break;
1036 #if IS_ENABLED(CONFIG_MPLS)
1037 case htons(ETH_P_MPLS_UC):
1038 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1039 break;
1040 #endif
1041 default:
1042 goto tx_err;
1045 return NETDEV_TX_OK;
1047 tx_err:
1048 dev->stats.tx_errors++;
1049 kfree_skb(skb);
1050 return NETDEV_TX_OK;
1054 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1056 struct net_device *tdev = NULL;
1057 struct ip_tunnel *tunnel;
1058 const struct iphdr *iph;
1059 struct flowi4 fl4;
1061 tunnel = netdev_priv(dev);
1062 iph = &tunnel->parms.iph;
1064 if (iph->daddr) {
1065 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1066 NULL,
1067 iph->daddr, iph->saddr,
1068 0, 0,
1069 IPPROTO_IPV6,
1070 RT_TOS(iph->tos),
1071 tunnel->parms.link);
1073 if (!IS_ERR(rt)) {
1074 tdev = rt->dst.dev;
1075 ip_rt_put(rt);
1077 dev->flags |= IFF_POINTOPOINT;
1080 if (!tdev && tunnel->parms.link)
1081 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1083 if (tdev && !netif_is_l3_master(tdev)) {
1084 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1086 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1087 dev->mtu = tdev->mtu - t_hlen;
1088 if (dev->mtu < IPV6_MIN_MTU)
1089 dev->mtu = IPV6_MIN_MTU;
1093 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1094 __u32 fwmark)
1096 struct net *net = t->net;
1097 struct sit_net *sitn = net_generic(net, sit_net_id);
1099 ipip6_tunnel_unlink(sitn, t);
1100 synchronize_net();
1101 t->parms.iph.saddr = p->iph.saddr;
1102 t->parms.iph.daddr = p->iph.daddr;
1103 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1104 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1105 ipip6_tunnel_link(sitn, t);
1106 t->parms.iph.ttl = p->iph.ttl;
1107 t->parms.iph.tos = p->iph.tos;
1108 t->parms.iph.frag_off = p->iph.frag_off;
1109 if (t->parms.link != p->link || t->fwmark != fwmark) {
1110 t->parms.link = p->link;
1111 t->fwmark = fwmark;
1112 ipip6_tunnel_bind_dev(t->dev);
1114 dst_cache_reset(&t->dst_cache);
1115 netdev_state_change(t->dev);
1118 #ifdef CONFIG_IPV6_SIT_6RD
1119 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1120 struct ip_tunnel_6rd *ip6rd)
1122 struct in6_addr prefix;
1123 __be32 relay_prefix;
1125 if (ip6rd->relay_prefixlen > 32 ||
1126 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1127 return -EINVAL;
1129 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1130 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1131 return -EINVAL;
1132 if (ip6rd->relay_prefixlen)
1133 relay_prefix = ip6rd->relay_prefix &
1134 htonl(0xffffffffUL <<
1135 (32 - ip6rd->relay_prefixlen));
1136 else
1137 relay_prefix = 0;
1138 if (relay_prefix != ip6rd->relay_prefix)
1139 return -EINVAL;
1141 t->ip6rd.prefix = prefix;
1142 t->ip6rd.relay_prefix = relay_prefix;
1143 t->ip6rd.prefixlen = ip6rd->prefixlen;
1144 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1145 dst_cache_reset(&t->dst_cache);
1146 netdev_state_change(t->dev);
1147 return 0;
1149 #endif
1151 static bool ipip6_valid_ip_proto(u8 ipproto)
1153 return ipproto == IPPROTO_IPV6 ||
1154 ipproto == IPPROTO_IPIP ||
1155 #if IS_ENABLED(CONFIG_MPLS)
1156 ipproto == IPPROTO_MPLS ||
1157 #endif
1158 ipproto == 0;
1161 static int
1162 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1164 int err = 0;
1165 struct ip_tunnel_parm p;
1166 struct ip_tunnel_prl prl;
1167 struct ip_tunnel *t = netdev_priv(dev);
1168 struct net *net = t->net;
1169 struct sit_net *sitn = net_generic(net, sit_net_id);
1170 #ifdef CONFIG_IPV6_SIT_6RD
1171 struct ip_tunnel_6rd ip6rd;
1172 #endif
1174 switch (cmd) {
1175 case SIOCGETTUNNEL:
1176 #ifdef CONFIG_IPV6_SIT_6RD
1177 case SIOCGET6RD:
1178 #endif
1179 if (dev == sitn->fb_tunnel_dev) {
1180 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1181 err = -EFAULT;
1182 break;
1184 t = ipip6_tunnel_locate(net, &p, 0);
1185 if (!t)
1186 t = netdev_priv(dev);
1189 err = -EFAULT;
1190 if (cmd == SIOCGETTUNNEL) {
1191 memcpy(&p, &t->parms, sizeof(p));
1192 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1193 sizeof(p)))
1194 goto done;
1195 #ifdef CONFIG_IPV6_SIT_6RD
1196 } else {
1197 ip6rd.prefix = t->ip6rd.prefix;
1198 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1199 ip6rd.prefixlen = t->ip6rd.prefixlen;
1200 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1201 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1202 sizeof(ip6rd)))
1203 goto done;
1204 #endif
1206 err = 0;
1207 break;
1209 case SIOCADDTUNNEL:
1210 case SIOCCHGTUNNEL:
1211 err = -EPERM;
1212 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1213 goto done;
1215 err = -EFAULT;
1216 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1217 goto done;
1219 err = -EINVAL;
1220 if (!ipip6_valid_ip_proto(p.iph.protocol))
1221 goto done;
1222 if (p.iph.version != 4 ||
1223 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1224 goto done;
1225 if (p.iph.ttl)
1226 p.iph.frag_off |= htons(IP_DF);
1228 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1230 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1231 if (t) {
1232 if (t->dev != dev) {
1233 err = -EEXIST;
1234 break;
1236 } else {
1237 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1238 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1239 err = -EINVAL;
1240 break;
1242 t = netdev_priv(dev);
1245 ipip6_tunnel_update(t, &p, t->fwmark);
1248 if (t) {
1249 err = 0;
1250 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1251 err = -EFAULT;
1252 } else
1253 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1254 break;
1256 case SIOCDELTUNNEL:
1257 err = -EPERM;
1258 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1259 goto done;
1261 if (dev == sitn->fb_tunnel_dev) {
1262 err = -EFAULT;
1263 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1264 goto done;
1265 err = -ENOENT;
1266 t = ipip6_tunnel_locate(net, &p, 0);
1267 if (!t)
1268 goto done;
1269 err = -EPERM;
1270 if (t == netdev_priv(sitn->fb_tunnel_dev))
1271 goto done;
1272 dev = t->dev;
1274 unregister_netdevice(dev);
1275 err = 0;
1276 break;
1278 case SIOCGETPRL:
1279 err = -EINVAL;
1280 if (dev == sitn->fb_tunnel_dev)
1281 goto done;
1282 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1283 break;
1285 case SIOCADDPRL:
1286 case SIOCDELPRL:
1287 case SIOCCHGPRL:
1288 err = -EPERM;
1289 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1290 goto done;
1291 err = -EINVAL;
1292 if (dev == sitn->fb_tunnel_dev)
1293 goto done;
1294 err = -EFAULT;
1295 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1296 goto done;
1298 switch (cmd) {
1299 case SIOCDELPRL:
1300 err = ipip6_tunnel_del_prl(t, &prl);
1301 break;
1302 case SIOCADDPRL:
1303 case SIOCCHGPRL:
1304 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1305 break;
1307 dst_cache_reset(&t->dst_cache);
1308 netdev_state_change(dev);
1309 break;
1311 #ifdef CONFIG_IPV6_SIT_6RD
1312 case SIOCADD6RD:
1313 case SIOCCHG6RD:
1314 case SIOCDEL6RD:
1315 err = -EPERM;
1316 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1317 goto done;
1319 err = -EFAULT;
1320 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1321 sizeof(ip6rd)))
1322 goto done;
1324 if (cmd != SIOCDEL6RD) {
1325 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1326 if (err < 0)
1327 goto done;
1328 } else
1329 ipip6_tunnel_clone_6rd(dev, sitn);
1331 err = 0;
1332 break;
1333 #endif
1335 default:
1336 err = -EINVAL;
1339 done:
1340 return err;
1343 static const struct net_device_ops ipip6_netdev_ops = {
1344 .ndo_init = ipip6_tunnel_init,
1345 .ndo_uninit = ipip6_tunnel_uninit,
1346 .ndo_start_xmit = sit_tunnel_xmit,
1347 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1348 .ndo_get_stats64 = ip_tunnel_get_stats64,
1349 .ndo_get_iflink = ip_tunnel_get_iflink,
1352 static void ipip6_dev_free(struct net_device *dev)
1354 struct ip_tunnel *tunnel = netdev_priv(dev);
1356 dst_cache_destroy(&tunnel->dst_cache);
1357 free_percpu(dev->tstats);
1360 #define SIT_FEATURES (NETIF_F_SG | \
1361 NETIF_F_FRAGLIST | \
1362 NETIF_F_HIGHDMA | \
1363 NETIF_F_GSO_SOFTWARE | \
1364 NETIF_F_HW_CSUM)
1366 static void ipip6_tunnel_setup(struct net_device *dev)
1368 struct ip_tunnel *tunnel = netdev_priv(dev);
1369 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1371 dev->netdev_ops = &ipip6_netdev_ops;
1372 dev->needs_free_netdev = true;
1373 dev->priv_destructor = ipip6_dev_free;
1375 dev->type = ARPHRD_SIT;
1376 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1377 dev->mtu = ETH_DATA_LEN - t_hlen;
1378 dev->min_mtu = IPV6_MIN_MTU;
1379 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1380 dev->flags = IFF_NOARP;
1381 netif_keep_dst(dev);
1382 dev->addr_len = 4;
1383 dev->features |= NETIF_F_LLTX;
1384 dev->features |= SIT_FEATURES;
1385 dev->hw_features |= SIT_FEATURES;
1388 static int ipip6_tunnel_init(struct net_device *dev)
1390 struct ip_tunnel *tunnel = netdev_priv(dev);
1391 int err;
1393 tunnel->dev = dev;
1394 tunnel->net = dev_net(dev);
1395 strcpy(tunnel->parms.name, dev->name);
1397 ipip6_tunnel_bind_dev(dev);
1398 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1399 if (!dev->tstats)
1400 return -ENOMEM;
1402 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1403 if (err) {
1404 free_percpu(dev->tstats);
1405 dev->tstats = NULL;
1406 return err;
1409 return 0;
1412 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1414 struct ip_tunnel *tunnel = netdev_priv(dev);
1415 struct iphdr *iph = &tunnel->parms.iph;
1416 struct net *net = dev_net(dev);
1417 struct sit_net *sitn = net_generic(net, sit_net_id);
1419 iph->version = 4;
1420 iph->protocol = IPPROTO_IPV6;
1421 iph->ihl = 5;
1422 iph->ttl = 64;
1424 dev_hold(dev);
1425 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1428 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1429 struct netlink_ext_ack *extack)
1431 u8 proto;
1433 if (!data || !data[IFLA_IPTUN_PROTO])
1434 return 0;
1436 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1437 if (!ipip6_valid_ip_proto(proto))
1438 return -EINVAL;
1440 return 0;
1443 static void ipip6_netlink_parms(struct nlattr *data[],
1444 struct ip_tunnel_parm *parms,
1445 __u32 *fwmark)
1447 memset(parms, 0, sizeof(*parms));
1449 parms->iph.version = 4;
1450 parms->iph.protocol = IPPROTO_IPV6;
1451 parms->iph.ihl = 5;
1452 parms->iph.ttl = 64;
1454 if (!data)
1455 return;
1457 if (data[IFLA_IPTUN_LINK])
1458 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1460 if (data[IFLA_IPTUN_LOCAL])
1461 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1463 if (data[IFLA_IPTUN_REMOTE])
1464 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1466 if (data[IFLA_IPTUN_TTL]) {
1467 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1468 if (parms->iph.ttl)
1469 parms->iph.frag_off = htons(IP_DF);
1472 if (data[IFLA_IPTUN_TOS])
1473 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1475 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1476 parms->iph.frag_off = htons(IP_DF);
1478 if (data[IFLA_IPTUN_FLAGS])
1479 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1481 if (data[IFLA_IPTUN_PROTO])
1482 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1484 if (data[IFLA_IPTUN_FWMARK])
1485 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1488 /* This function returns true when ENCAP attributes are present in the nl msg */
1489 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1490 struct ip_tunnel_encap *ipencap)
1492 bool ret = false;
1494 memset(ipencap, 0, sizeof(*ipencap));
1496 if (!data)
1497 return ret;
1499 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1500 ret = true;
1501 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1504 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1505 ret = true;
1506 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1509 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1510 ret = true;
1511 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1514 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1515 ret = true;
1516 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1519 return ret;
1522 #ifdef CONFIG_IPV6_SIT_6RD
1523 /* This function returns true when 6RD attributes are present in the nl msg */
1524 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1525 struct ip_tunnel_6rd *ip6rd)
1527 bool ret = false;
1528 memset(ip6rd, 0, sizeof(*ip6rd));
1530 if (!data)
1531 return ret;
1533 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1534 ret = true;
1535 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1538 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1539 ret = true;
1540 ip6rd->relay_prefix =
1541 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1544 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1545 ret = true;
1546 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1549 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1550 ret = true;
1551 ip6rd->relay_prefixlen =
1552 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1555 return ret;
1557 #endif
1559 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1560 struct nlattr *tb[], struct nlattr *data[],
1561 struct netlink_ext_ack *extack)
1563 struct net *net = dev_net(dev);
1564 struct ip_tunnel *nt;
1565 struct ip_tunnel_encap ipencap;
1566 #ifdef CONFIG_IPV6_SIT_6RD
1567 struct ip_tunnel_6rd ip6rd;
1568 #endif
1569 int err;
1571 nt = netdev_priv(dev);
1573 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1574 err = ip_tunnel_encap_setup(nt, &ipencap);
1575 if (err < 0)
1576 return err;
1579 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1581 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1582 return -EEXIST;
1584 err = ipip6_tunnel_create(dev);
1585 if (err < 0)
1586 return err;
1588 if (tb[IFLA_MTU]) {
1589 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1591 if (mtu >= IPV6_MIN_MTU &&
1592 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1593 dev->mtu = mtu;
1596 #ifdef CONFIG_IPV6_SIT_6RD
1597 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1598 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1599 #endif
1601 return err;
1604 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1605 struct nlattr *data[],
1606 struct netlink_ext_ack *extack)
1608 struct ip_tunnel *t = netdev_priv(dev);
1609 struct ip_tunnel_parm p;
1610 struct ip_tunnel_encap ipencap;
1611 struct net *net = t->net;
1612 struct sit_net *sitn = net_generic(net, sit_net_id);
1613 #ifdef CONFIG_IPV6_SIT_6RD
1614 struct ip_tunnel_6rd ip6rd;
1615 #endif
1616 __u32 fwmark = t->fwmark;
1617 int err;
1619 if (dev == sitn->fb_tunnel_dev)
1620 return -EINVAL;
1622 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1623 err = ip_tunnel_encap_setup(t, &ipencap);
1624 if (err < 0)
1625 return err;
1628 ipip6_netlink_parms(data, &p, &fwmark);
1630 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1631 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1632 return -EINVAL;
1634 t = ipip6_tunnel_locate(net, &p, 0);
1636 if (t) {
1637 if (t->dev != dev)
1638 return -EEXIST;
1639 } else
1640 t = netdev_priv(dev);
1642 ipip6_tunnel_update(t, &p, fwmark);
1644 #ifdef CONFIG_IPV6_SIT_6RD
1645 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1646 return ipip6_tunnel_update_6rd(t, &ip6rd);
1647 #endif
1649 return 0;
1652 static size_t ipip6_get_size(const struct net_device *dev)
1654 return
1655 /* IFLA_IPTUN_LINK */
1656 nla_total_size(4) +
1657 /* IFLA_IPTUN_LOCAL */
1658 nla_total_size(4) +
1659 /* IFLA_IPTUN_REMOTE */
1660 nla_total_size(4) +
1661 /* IFLA_IPTUN_TTL */
1662 nla_total_size(1) +
1663 /* IFLA_IPTUN_TOS */
1664 nla_total_size(1) +
1665 /* IFLA_IPTUN_PMTUDISC */
1666 nla_total_size(1) +
1667 /* IFLA_IPTUN_FLAGS */
1668 nla_total_size(2) +
1669 /* IFLA_IPTUN_PROTO */
1670 nla_total_size(1) +
1671 #ifdef CONFIG_IPV6_SIT_6RD
1672 /* IFLA_IPTUN_6RD_PREFIX */
1673 nla_total_size(sizeof(struct in6_addr)) +
1674 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1675 nla_total_size(4) +
1676 /* IFLA_IPTUN_6RD_PREFIXLEN */
1677 nla_total_size(2) +
1678 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1679 nla_total_size(2) +
1680 #endif
1681 /* IFLA_IPTUN_ENCAP_TYPE */
1682 nla_total_size(2) +
1683 /* IFLA_IPTUN_ENCAP_FLAGS */
1684 nla_total_size(2) +
1685 /* IFLA_IPTUN_ENCAP_SPORT */
1686 nla_total_size(2) +
1687 /* IFLA_IPTUN_ENCAP_DPORT */
1688 nla_total_size(2) +
1689 /* IFLA_IPTUN_FWMARK */
1690 nla_total_size(4) +
1694 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1696 struct ip_tunnel *tunnel = netdev_priv(dev);
1697 struct ip_tunnel_parm *parm = &tunnel->parms;
1699 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1700 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1701 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1702 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1703 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1704 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1705 !!(parm->iph.frag_off & htons(IP_DF))) ||
1706 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1707 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1708 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1709 goto nla_put_failure;
1711 #ifdef CONFIG_IPV6_SIT_6RD
1712 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1713 &tunnel->ip6rd.prefix) ||
1714 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1715 tunnel->ip6rd.relay_prefix) ||
1716 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1717 tunnel->ip6rd.prefixlen) ||
1718 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1719 tunnel->ip6rd.relay_prefixlen))
1720 goto nla_put_failure;
1721 #endif
1723 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1724 tunnel->encap.type) ||
1725 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1726 tunnel->encap.sport) ||
1727 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1728 tunnel->encap.dport) ||
1729 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1730 tunnel->encap.flags))
1731 goto nla_put_failure;
1733 return 0;
1735 nla_put_failure:
1736 return -EMSGSIZE;
1739 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1740 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1741 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1742 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1743 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1744 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1745 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1746 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1747 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1748 #ifdef CONFIG_IPV6_SIT_6RD
1749 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1750 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1751 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1752 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1753 #endif
1754 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1755 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1756 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1757 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1758 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1761 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1763 struct net *net = dev_net(dev);
1764 struct sit_net *sitn = net_generic(net, sit_net_id);
1766 if (dev != sitn->fb_tunnel_dev)
1767 unregister_netdevice_queue(dev, head);
1770 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1771 .kind = "sit",
1772 .maxtype = IFLA_IPTUN_MAX,
1773 .policy = ipip6_policy,
1774 .priv_size = sizeof(struct ip_tunnel),
1775 .setup = ipip6_tunnel_setup,
1776 .validate = ipip6_validate,
1777 .newlink = ipip6_newlink,
1778 .changelink = ipip6_changelink,
1779 .get_size = ipip6_get_size,
1780 .fill_info = ipip6_fill_info,
1781 .dellink = ipip6_dellink,
1782 .get_link_net = ip_tunnel_get_link_net,
1785 static struct xfrm_tunnel sit_handler __read_mostly = {
1786 .handler = ipip6_rcv,
1787 .err_handler = ipip6_err,
1788 .priority = 1,
1791 static struct xfrm_tunnel ipip_handler __read_mostly = {
1792 .handler = ipip_rcv,
1793 .err_handler = ipip6_err,
1794 .priority = 2,
1797 #if IS_ENABLED(CONFIG_MPLS)
1798 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1799 .handler = mplsip_rcv,
1800 .err_handler = ipip6_err,
1801 .priority = 2,
1803 #endif
1805 static void __net_exit sit_destroy_tunnels(struct net *net,
1806 struct list_head *head)
1808 struct sit_net *sitn = net_generic(net, sit_net_id);
1809 struct net_device *dev, *aux;
1810 int prio;
1812 for_each_netdev_safe(net, dev, aux)
1813 if (dev->rtnl_link_ops == &sit_link_ops)
1814 unregister_netdevice_queue(dev, head);
1816 for (prio = 1; prio < 4; prio++) {
1817 int h;
1818 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1819 struct ip_tunnel *t;
1821 t = rtnl_dereference(sitn->tunnels[prio][h]);
1822 while (t) {
1823 /* If dev is in the same netns, it has already
1824 * been added to the list by the previous loop.
1826 if (!net_eq(dev_net(t->dev), net))
1827 unregister_netdevice_queue(t->dev,
1828 head);
1829 t = rtnl_dereference(t->next);
1835 static int __net_init sit_init_net(struct net *net)
1837 struct sit_net *sitn = net_generic(net, sit_net_id);
1838 struct ip_tunnel *t;
1839 int err;
1841 sitn->tunnels[0] = sitn->tunnels_wc;
1842 sitn->tunnels[1] = sitn->tunnels_l;
1843 sitn->tunnels[2] = sitn->tunnels_r;
1844 sitn->tunnels[3] = sitn->tunnels_r_l;
1846 if (!net_has_fallback_tunnels(net))
1847 return 0;
1849 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1850 NET_NAME_UNKNOWN,
1851 ipip6_tunnel_setup);
1852 if (!sitn->fb_tunnel_dev) {
1853 err = -ENOMEM;
1854 goto err_alloc_dev;
1856 dev_net_set(sitn->fb_tunnel_dev, net);
1857 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1858 /* FB netdevice is special: we have one, and only one per netns.
1859 * Allowing to move it to another netns is clearly unsafe.
1861 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1863 err = register_netdev(sitn->fb_tunnel_dev);
1864 if (err)
1865 goto err_reg_dev;
1867 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1868 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1870 t = netdev_priv(sitn->fb_tunnel_dev);
1872 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1873 return 0;
1875 err_reg_dev:
1876 ipip6_dev_free(sitn->fb_tunnel_dev);
1877 free_netdev(sitn->fb_tunnel_dev);
1878 err_alloc_dev:
1879 return err;
1882 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1884 LIST_HEAD(list);
1885 struct net *net;
1887 rtnl_lock();
1888 list_for_each_entry(net, net_list, exit_list)
1889 sit_destroy_tunnels(net, &list);
1891 unregister_netdevice_many(&list);
1892 rtnl_unlock();
1895 static struct pernet_operations sit_net_ops = {
1896 .init = sit_init_net,
1897 .exit_batch = sit_exit_batch_net,
1898 .id = &sit_net_id,
1899 .size = sizeof(struct sit_net),
1902 static void __exit sit_cleanup(void)
1904 rtnl_link_unregister(&sit_link_ops);
1905 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1906 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1907 #if IS_ENABLED(CONFIG_MPLS)
1908 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1909 #endif
1911 unregister_pernet_device(&sit_net_ops);
1912 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1915 static int __init sit_init(void)
1917 int err;
1919 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1921 err = register_pernet_device(&sit_net_ops);
1922 if (err < 0)
1923 return err;
1924 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1925 if (err < 0) {
1926 pr_info("%s: can't register ip6ip4\n", __func__);
1927 goto xfrm_tunnel_failed;
1929 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1930 if (err < 0) {
1931 pr_info("%s: can't register ip4ip4\n", __func__);
1932 goto xfrm_tunnel4_failed;
1934 #if IS_ENABLED(CONFIG_MPLS)
1935 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1936 if (err < 0) {
1937 pr_info("%s: can't register mplsip\n", __func__);
1938 goto xfrm_tunnel_mpls_failed;
1940 #endif
1941 err = rtnl_link_register(&sit_link_ops);
1942 if (err < 0)
1943 goto rtnl_link_failed;
1945 out:
1946 return err;
1948 rtnl_link_failed:
1949 #if IS_ENABLED(CONFIG_MPLS)
1950 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1951 xfrm_tunnel_mpls_failed:
1952 #endif
1953 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1954 xfrm_tunnel4_failed:
1955 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1956 xfrm_tunnel_failed:
1957 unregister_pernet_device(&sit_net_ops);
1958 goto out;
1961 module_init(sit_init);
1962 module_exit(sit_cleanup);
1963 MODULE_LICENSE("GPL");
1964 MODULE_ALIAS_RTNL_LINK("sit");
1965 MODULE_ALIAS_NETDEV("sit0");