?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/mm: Add TLB purge to free pmd/pte page interfaces
[linux/fpc-iii.git] / net / ipv4 / ip_tunnel.c
blobe1271e75e107177892870728dbd1434cccd11c08
1 /*
2 * Copyright (c) 2013 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20
21 #include <linux/capability.h>
22 #include <linux/module.h>
23 #include <linux/types.h>
24 #include <linux/kernel.h>
25 #include <linux/slab.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
29 #include <linux/in.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/init.h>
34 #include <linux/in6.h>
35 #include <linux/inetdevice.h>
36 #include <linux/igmp.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/etherdevice.h>
39 #include <linux/if_ether.h>
40 #include <linux/if_vlan.h>
41 #include <linux/rculist.h>
42 #include <linux/err.h>
43
44 #include <net/sock.h>
45 #include <net/ip.h>
46 #include <net/icmp.h>
47 #include <net/protocol.h>
48 #include <net/ip_tunnels.h>
49 #include <net/arp.h>
50 #include <net/checksum.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/xfrm.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 #include <net/rtnetlink.h>
57 #include <net/udp.h>
58 #include <net/dst_metadata.h>
59
60 #if IS_ENABLED(CONFIG_IPV6)
61 #include <net/ipv6.h>
62 #include <net/ip6_fib.h>
63 #include <net/ip6_route.h>
64 #endif
65
66 static unsigned int ip_tunnel_hash(__be32 key, __be32 remote)
67 {
68 return hash_32((__force u32)key ^ (__force u32)remote,
69 IP_TNL_HASH_BITS);
70 }
71
72 static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
73 __be16 flags, __be32 key)
74 {
75 if (p->i_flags & TUNNEL_KEY) {
76 if (flags & TUNNEL_KEY)
77 return key == p->i_key;
78 else
79 /* key expected, none present */
80 return false;
81 } else
82 return !(flags & TUNNEL_KEY);
83 }
84
85 /* Fallback tunnel: no source, no destination, no key, no options
86
87 Tunnel hash table:
88 We require exact key match i.e. if a key is present in packet
89 it will match only tunnel with the same key; if it is not present,
90 it will match only keyless tunnel.
91
92 All keysless packets, if not matched configured keyless tunnels
93 will match fallback tunnel.
94 Given src, dst and key, find appropriate for input tunnel.
95 */
96 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
97 int link, __be16 flags,
98 __be32 remote, __be32 local,
99 __be32 key)
100 {
101 unsigned int hash;
102 struct ip_tunnel *t, *cand = NULL;
103 struct hlist_head *head;
104
105 hash = ip_tunnel_hash(key, remote);
106 head = &itn->tunnels[hash];
107
108 hlist_for_each_entry_rcu(t, head, hash_node) {
109 if (local != t->parms.iph.saddr ||
110 remote != t->parms.iph.daddr ||
111 !(t->dev->flags & IFF_UP))
112 continue;
113
114 if (!ip_tunnel_key_match(&t->parms, flags, key))
115 continue;
116
117 if (t->parms.link == link)
118 return t;
119 else
120 cand = t;
121 }
122
123 hlist_for_each_entry_rcu(t, head, hash_node) {
124 if (remote != t->parms.iph.daddr ||
125 t->parms.iph.saddr != 0 ||
126 !(t->dev->flags & IFF_UP))
127 continue;
128
129 if (!ip_tunnel_key_match(&t->parms, flags, key))
130 continue;
131
132 if (t->parms.link == link)
133 return t;
134 else if (!cand)
135 cand = t;
136 }
137
138 hash = ip_tunnel_hash(key, 0);
139 head = &itn->tunnels[hash];
140
141 hlist_for_each_entry_rcu(t, head, hash_node) {
142 if ((local != t->parms.iph.saddr || t->parms.iph.daddr != 0) &&
143 (local != t->parms.iph.daddr || !ipv4_is_multicast(local)))
144 continue;
145
146 if (!(t->dev->flags & IFF_UP))
147 continue;
148
149 if (!ip_tunnel_key_match(&t->parms, flags, key))
150 continue;
151
152 if (t->parms.link == link)
153 return t;
154 else if (!cand)
155 cand = t;
156 }
157
158 if (flags & TUNNEL_NO_KEY)
159 goto skip_key_lookup;
160
161 hlist_for_each_entry_rcu(t, head, hash_node) {
162 if (t->parms.i_key != key ||
163 t->parms.iph.saddr != 0 ||
164 t->parms.iph.daddr != 0 ||
165 !(t->dev->flags & IFF_UP))
166 continue;
167
168 if (t->parms.link == link)
169 return t;
170 else if (!cand)
171 cand = t;
172 }
173
174 skip_key_lookup:
175 if (cand)
176 return cand;
177
178 t = rcu_dereference(itn->collect_md_tun);
179 if (t)
180 return t;
181
182 if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
183 return netdev_priv(itn->fb_tunnel_dev);
184
185 return NULL;
186 }
187 EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
188
189 static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
190 struct ip_tunnel_parm *parms)
191 {
192 unsigned int h;
193 __be32 remote;
194 __be32 i_key = parms->i_key;
195
196 if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
197 remote = parms->iph.daddr;
198 else
199 remote = 0;
200
201 if (!(parms->i_flags & TUNNEL_KEY) && (parms->i_flags & VTI_ISVTI))
202 i_key = 0;
203
204 h = ip_tunnel_hash(i_key, remote);
205 return &itn->tunnels[h];
206 }
207
208 static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
209 {
210 struct hlist_head *head = ip_bucket(itn, &t->parms);
211
212 if (t->collect_md)
213 rcu_assign_pointer(itn->collect_md_tun, t);
214 hlist_add_head_rcu(&t->hash_node, head);
215 }
216
217 static void ip_tunnel_del(struct ip_tunnel_net *itn, struct ip_tunnel *t)
218 {
219 if (t->collect_md)
220 rcu_assign_pointer(itn->collect_md_tun, NULL);
221 hlist_del_init_rcu(&t->hash_node);
222 }
223
224 static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
225 struct ip_tunnel_parm *parms,
226 int type)
227 {
228 __be32 remote = parms->iph.daddr;
229 __be32 local = parms->iph.saddr;
230 __be32 key = parms->i_key;
231 __be16 flags = parms->i_flags;
232 int link = parms->link;
233 struct ip_tunnel *t = NULL;
234 struct hlist_head *head = ip_bucket(itn, parms);
235
236 hlist_for_each_entry_rcu(t, head, hash_node) {
237 if (local == t->parms.iph.saddr &&
238 remote == t->parms.iph.daddr &&
239 link == t->parms.link &&
240 type == t->dev->type &&
241 ip_tunnel_key_match(&t->parms, flags, key))
242 break;
243 }
244 return t;
245 }
246
247 static struct net_device *__ip_tunnel_create(struct net *net,
248 const struct rtnl_link_ops *ops,
249 struct ip_tunnel_parm *parms)
250 {
251 int err;
252 struct ip_tunnel *tunnel;
253 struct net_device *dev;
254 char name[IFNAMSIZ];
255
256 err = -E2BIG;
257 if (parms->name[0]) {
258 if (!dev_valid_name(parms->name))
259 goto failed;
260 strlcpy(name, parms->name, IFNAMSIZ);
261 } else {
262 if (strlen(ops->kind) > (IFNAMSIZ - 3))
263 goto failed;
264 strlcpy(name, ops->kind, IFNAMSIZ);
265 strncat(name, "%d", 2);
266 }
267
268 ASSERT_RTNL();
269 dev = alloc_netdev(ops->priv_size, name, NET_NAME_UNKNOWN, ops->setup);
270 if (!dev) {
271 err = -ENOMEM;
272 goto failed;
273 }
274 dev_net_set(dev, net);
275
276 dev->rtnl_link_ops = ops;
277
278 tunnel = netdev_priv(dev);
279 tunnel->parms = *parms;
280 tunnel->net = net;
281
282 err = register_netdevice(dev);
283 if (err)
284 goto failed_free;
285
286 return dev;
287
288 failed_free:
289 free_netdev(dev);
290 failed:
291 return ERR_PTR(err);
292 }
293
294 static inline void init_tunnel_flow(struct flowi4 *fl4,
295 int proto,
296 __be32 daddr, __be32 saddr,
297 __be32 key, __u8 tos, int oif)
298 {
299 memset(fl4, 0, sizeof(*fl4));
300 fl4->flowi4_oif = oif;
301 fl4->daddr = daddr;
302 fl4->saddr = saddr;
303 fl4->flowi4_tos = tos;
304 fl4->flowi4_proto = proto;
305 fl4->fl4_gre_key = key;
306 }
307
308 static int ip_tunnel_bind_dev(struct net_device *dev)
309 {
310 struct net_device *tdev = NULL;
311 struct ip_tunnel *tunnel = netdev_priv(dev);
312 const struct iphdr *iph;
313 int hlen = LL_MAX_HEADER;
314 int mtu = ETH_DATA_LEN;
315 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
316
317 iph = &tunnel->parms.iph;
318
319 /* Guess output device to choose reasonable mtu and needed_headroom */
320 if (iph->daddr) {
321 struct flowi4 fl4;
322 struct rtable *rt;
323
324 init_tunnel_flow(&fl4, iph->protocol, iph->daddr,
325 iph->saddr, tunnel->parms.o_key,
326 RT_TOS(iph->tos), tunnel->parms.link);
327 rt = ip_route_output_key(tunnel->net, &fl4);
328
329 if (!IS_ERR(rt)) {
330 tdev = rt->dst.dev;
331 ip_rt_put(rt);
332 }
333 if (dev->type != ARPHRD_ETHER)
334 dev->flags |= IFF_POINTOPOINT;
335
336 dst_cache_reset(&tunnel->dst_cache);
337 }
338
339 if (!tdev && tunnel->parms.link)
340 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
341
342 if (tdev) {
343 hlen = tdev->hard_header_len + tdev->needed_headroom;
344 mtu = tdev->mtu;
345 }
346
347 dev->needed_headroom = t_hlen + hlen;
348 mtu -= (dev->hard_header_len + t_hlen);
349
350 if (mtu < IPV4_MIN_MTU)
351 mtu = IPV4_MIN_MTU;
352
353 return mtu;
354 }
355
356 static struct ip_tunnel *ip_tunnel_create(struct net *net,
357 struct ip_tunnel_net *itn,
358 struct ip_tunnel_parm *parms)
359 {
360 struct ip_tunnel *nt;
361 struct net_device *dev;
362
363 BUG_ON(!itn->fb_tunnel_dev);
364 dev = __ip_tunnel_create(net, itn->fb_tunnel_dev->rtnl_link_ops, parms);
365 if (IS_ERR(dev))
366 return ERR_CAST(dev);
367
368 dev->mtu = ip_tunnel_bind_dev(dev);
369
370 nt = netdev_priv(dev);
371 ip_tunnel_add(itn, nt);
372 return nt;
373 }
374
375 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
376 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
377 bool log_ecn_error)
378 {
379 struct pcpu_sw_netstats *tstats;
380 const struct iphdr *iph = ip_hdr(skb);
381 int err;
382
383 #ifdef CONFIG_NET_IPGRE_BROADCAST
384 if (ipv4_is_multicast(iph->daddr)) {
385 tunnel->dev->stats.multicast++;
386 skb->pkt_type = PACKET_BROADCAST;
387 }
388 #endif
389
390 if ((!(tpi->flags&TUNNEL_CSUM) && (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
391 ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
392 tunnel->dev->stats.rx_crc_errors++;
393 tunnel->dev->stats.rx_errors++;
394 goto drop;
395 }
396
397 if (tunnel->parms.i_flags&TUNNEL_SEQ) {
398 if (!(tpi->flags&TUNNEL_SEQ) ||
399 (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
400 tunnel->dev->stats.rx_fifo_errors++;
401 tunnel->dev->stats.rx_errors++;
402 goto drop;
403 }
404 tunnel->i_seqno = ntohl(tpi->seq) + 1;
405 }
406
407 skb_reset_network_header(skb);
408
409 err = IP_ECN_decapsulate(iph, skb);
410 if (unlikely(err)) {
411 if (log_ecn_error)
412 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
413 &iph->saddr, iph->tos);
414 if (err > 1) {
415 ++tunnel->dev->stats.rx_frame_errors;
416 ++tunnel->dev->stats.rx_errors;
417 goto drop;
418 }
419 }
420
421 tstats = this_cpu_ptr(tunnel->dev->tstats);
422 u64_stats_update_begin(&tstats->syncp);
423 tstats->rx_packets++;
424 tstats->rx_bytes += skb->len;
425 u64_stats_update_end(&tstats->syncp);
426
427 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
428
429 if (tunnel->dev->type == ARPHRD_ETHER) {
430 skb->protocol = eth_type_trans(skb, tunnel->dev);
431 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
432 } else {
433 skb->dev = tunnel->dev;
434 }
435
436 if (tun_dst)
437 skb_dst_set(skb, (struct dst_entry *)tun_dst);
438
439 gro_cells_receive(&tunnel->gro_cells, skb);
440 return 0;
441
442 drop:
443 kfree_skb(skb);
444 return 0;
445 }
446 EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
447
448 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *ops,
449 unsigned int num)
450 {
451 if (num >= MAX_IPTUN_ENCAP_OPS)
452 return -ERANGE;
453
454 return !cmpxchg((const struct ip_tunnel_encap_ops **)
455 &iptun_encaps[num],
456 NULL, ops) ? 0 : -1;
457 }
458 EXPORT_SYMBOL(ip_tunnel_encap_add_ops);
459
460 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *ops,
461 unsigned int num)
462 {
463 int ret;
464
465 if (num >= MAX_IPTUN_ENCAP_OPS)
466 return -ERANGE;
467
468 ret = (cmpxchg((const struct ip_tunnel_encap_ops **)
469 &iptun_encaps[num],
470 ops, NULL) == ops) ? 0 : -1;
471
472 synchronize_net();
473
474 return ret;
475 }
476 EXPORT_SYMBOL(ip_tunnel_encap_del_ops);
477
478 int ip_tunnel_encap_setup(struct ip_tunnel *t,
479 struct ip_tunnel_encap *ipencap)
480 {
481 int hlen;
482
483 memset(&t->encap, 0, sizeof(t->encap));
484
485 hlen = ip_encap_hlen(ipencap);
486 if (hlen < 0)
487 return hlen;
488
489 t->encap.type = ipencap->type;
490 t->encap.sport = ipencap->sport;
491 t->encap.dport = ipencap->dport;
492 t->encap.flags = ipencap->flags;
493
494 t->encap_hlen = hlen;
495 t->hlen = t->encap_hlen + t->tun_hlen;
496
497 return 0;
498 }
499 EXPORT_SYMBOL_GPL(ip_tunnel_encap_setup);
500
501 static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
502 struct rtable *rt, __be16 df,
503 const struct iphdr *inner_iph)
504 {
505 struct ip_tunnel *tunnel = netdev_priv(dev);
506 int pkt_size = skb->len - tunnel->hlen - dev->hard_header_len;
507 int mtu;
508
509 if (df)
510 mtu = dst_mtu(&rt->dst) - dev->hard_header_len
511 - sizeof(struct iphdr) - tunnel->hlen;
512 else
513 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
514
515 if (skb_dst(skb))
516 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
517
518 if (skb->protocol == htons(ETH_P_IP)) {
519 if (!skb_is_gso(skb) &&
520 (inner_iph->frag_off & htons(IP_DF)) &&
521 mtu < pkt_size) {
522 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
523 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
524 return -E2BIG;
525 }
526 }
527 #if IS_ENABLED(CONFIG_IPV6)
528 else if (skb->protocol == htons(ETH_P_IPV6)) {
529 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
530
531 if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
532 mtu >= IPV6_MIN_MTU) {
533 if ((tunnel->parms.iph.daddr &&
534 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
535 rt6->rt6i_dst.plen == 128) {
536 rt6->rt6i_flags |= RTF_MODIFIED;
537 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
538 }
539 }
540
541 if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
542 mtu < pkt_size) {
543 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
544 return -E2BIG;
545 }
546 }
547 #endif
548 return 0;
549 }
550
551 void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, u8 proto)
552 {
553 struct ip_tunnel *tunnel = netdev_priv(dev);
554 u32 headroom = sizeof(struct iphdr);
555 struct ip_tunnel_info *tun_info;
556 const struct ip_tunnel_key *key;
557 const struct iphdr *inner_iph;
558 struct rtable *rt;
559 struct flowi4 fl4;
560 __be16 df = 0;
561 u8 tos, ttl;
562
563 tun_info = skb_tunnel_info(skb);
564 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
565 ip_tunnel_info_af(tun_info) != AF_INET))
566 goto tx_error;
567 key = &tun_info->key;
568 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
569 inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
570 tos = key->tos;
571 if (tos == 1) {
572 if (skb->protocol == htons(ETH_P_IP))
573 tos = inner_iph->tos;
574 else if (skb->protocol == htons(ETH_P_IPV6))
575 tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
576 }
577 init_tunnel_flow(&fl4, proto, key->u.ipv4.dst, key->u.ipv4.src, 0,
578 RT_TOS(tos), tunnel->parms.link);
579 if (tunnel->encap.type != TUNNEL_ENCAP_NONE)
580 goto tx_error;
581 rt = ip_route_output_key(tunnel->net, &fl4);
582 if (IS_ERR(rt)) {
583 dev->stats.tx_carrier_errors++;
584 goto tx_error;
585 }
586 if (rt->dst.dev == dev) {
587 ip_rt_put(rt);
588 dev->stats.collisions++;
589 goto tx_error;
590 }
591 tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
592 ttl = key->ttl;
593 if (ttl == 0) {
594 if (skb->protocol == htons(ETH_P_IP))
595 ttl = inner_iph->ttl;
596 else if (skb->protocol == htons(ETH_P_IPV6))
597 ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
598 else
599 ttl = ip4_dst_hoplimit(&rt->dst);
600 }
601 if (key->tun_flags & TUNNEL_DONT_FRAGMENT)
602 df = htons(IP_DF);
603 else if (skb->protocol == htons(ETH_P_IP))
604 df = inner_iph->frag_off & htons(IP_DF);
605 headroom += LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len;
606 if (headroom > dev->needed_headroom)
607 dev->needed_headroom = headroom;
608
609 if (skb_cow_head(skb, dev->needed_headroom)) {
610 ip_rt_put(rt);
611 goto tx_dropped;
612 }
613 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, proto, tos, ttl,
614 df, !net_eq(tunnel->net, dev_net(dev)));
615 return;
616 tx_error:
617 dev->stats.tx_errors++;
618 goto kfree;
619 tx_dropped:
620 dev->stats.tx_dropped++;
621 kfree:
622 kfree_skb(skb);
623 }
624 EXPORT_SYMBOL_GPL(ip_md_tunnel_xmit);
625
626 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
627 const struct iphdr *tnl_params, u8 protocol)
628 {
629 struct ip_tunnel *tunnel = netdev_priv(dev);
630 const struct iphdr *inner_iph;
631 struct flowi4 fl4;
632 u8 tos, ttl;
633 __be16 df;
634 struct rtable *rt; /* Route to the other host */
635 unsigned int max_headroom; /* The extra header space needed */
636 __be32 dst;
637 bool connected;
638
639 inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
640 connected = (tunnel->parms.iph.daddr != 0);
641
642 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
643
644 dst = tnl_params->daddr;
645 if (dst == 0) {
646 /* NBMA tunnel */
647
648 if (!skb_dst(skb)) {
649 dev->stats.tx_fifo_errors++;
650 goto tx_error;
651 }
652
653 if (skb->protocol == htons(ETH_P_IP)) {
654 rt = skb_rtable(skb);
655 dst = rt_nexthop(rt, inner_iph->daddr);
656 }
657 #if IS_ENABLED(CONFIG_IPV6)
658 else if (skb->protocol == htons(ETH_P_IPV6)) {
659 const struct in6_addr *addr6;
660 struct neighbour *neigh;
661 bool do_tx_error_icmp;
662 int addr_type;
663
664 neigh = dst_neigh_lookup(skb_dst(skb),
665 &ipv6_hdr(skb)->daddr);
666 if (!neigh)
667 goto tx_error;
668
669 addr6 = (const struct in6_addr *)&neigh->primary_key;
670 addr_type = ipv6_addr_type(addr6);
671
672 if (addr_type == IPV6_ADDR_ANY) {
673 addr6 = &ipv6_hdr(skb)->daddr;
674 addr_type = ipv6_addr_type(addr6);
675 }
676
677 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
678 do_tx_error_icmp = true;
679 else {
680 do_tx_error_icmp = false;
681 dst = addr6->s6_addr32[3];
682 }
683 neigh_release(neigh);
684 if (do_tx_error_icmp)
685 goto tx_error_icmp;
686 }
687 #endif
688 else
689 goto tx_error;
690
691 connected = false;
692 }
693
694 tos = tnl_params->tos;
695 if (tos & 0x1) {
696 tos &= ~0x1;
697 if (skb->protocol == htons(ETH_P_IP)) {
698 tos = inner_iph->tos;
699 connected = false;
700 } else if (skb->protocol == htons(ETH_P_IPV6)) {
701 tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
702 connected = false;
703 }
704 }
705
706 init_tunnel_flow(&fl4, protocol, dst, tnl_params->saddr,
707 tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link);
708
709 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0)
710 goto tx_error;
711
712 rt = connected ? dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr) :
713 NULL;
714
715 if (!rt) {
716 rt = ip_route_output_key(tunnel->net, &fl4);
717
718 if (IS_ERR(rt)) {
719 dev->stats.tx_carrier_errors++;
720 goto tx_error;
721 }
722 if (connected)
723 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst,
724 fl4.saddr);
725 }
726
727 if (rt->dst.dev == dev) {
728 ip_rt_put(rt);
729 dev->stats.collisions++;
730 goto tx_error;
731 }
732
733 if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off, inner_iph)) {
734 ip_rt_put(rt);
735 goto tx_error;
736 }
737
738 if (tunnel->err_count > 0) {
739 if (time_before(jiffies,
740 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
741 tunnel->err_count--;
742
743 dst_link_failure(skb);
744 } else
745 tunnel->err_count = 0;
746 }
747
748 tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
749 ttl = tnl_params->ttl;
750 if (ttl == 0) {
751 if (skb->protocol == htons(ETH_P_IP))
752 ttl = inner_iph->ttl;
753 #if IS_ENABLED(CONFIG_IPV6)
754 else if (skb->protocol == htons(ETH_P_IPV6))
755 ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
756 #endif
757 else
758 ttl = ip4_dst_hoplimit(&rt->dst);
759 }
760
761 df = tnl_params->frag_off;
762 if (skb->protocol == htons(ETH_P_IP) && !tunnel->ignore_df)
763 df |= (inner_iph->frag_off&htons(IP_DF));
764
765 max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
766 + rt->dst.header_len + ip_encap_hlen(&tunnel->encap);
767 if (max_headroom > dev->needed_headroom)
768 dev->needed_headroom = max_headroom;
769
770 if (skb_cow_head(skb, dev->needed_headroom)) {
771 ip_rt_put(rt);
772 dev->stats.tx_dropped++;
773 kfree_skb(skb);
774 return;
775 }
776
777 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
778 df, !net_eq(tunnel->net, dev_net(dev)));
779 return;
780
781 #if IS_ENABLED(CONFIG_IPV6)
782 tx_error_icmp:
783 dst_link_failure(skb);
784 #endif
785 tx_error:
786 dev->stats.tx_errors++;
787 kfree_skb(skb);
788 }
789 EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
790
791 static void ip_tunnel_update(struct ip_tunnel_net *itn,
792 struct ip_tunnel *t,
793 struct net_device *dev,
794 struct ip_tunnel_parm *p,
795 bool set_mtu)
796 {
797 ip_tunnel_del(itn, t);
798 t->parms.iph.saddr = p->iph.saddr;
799 t->parms.iph.daddr = p->iph.daddr;
800 t->parms.i_key = p->i_key;
801 t->parms.o_key = p->o_key;
802 if (dev->type != ARPHRD_ETHER) {
803 memcpy(dev->dev_addr, &p->iph.saddr, 4);
804 memcpy(dev->broadcast, &p->iph.daddr, 4);
805 }
806 ip_tunnel_add(itn, t);
807
808 t->parms.iph.ttl = p->iph.ttl;
809 t->parms.iph.tos = p->iph.tos;
810 t->parms.iph.frag_off = p->iph.frag_off;
811
812 if (t->parms.link != p->link) {
813 int mtu;
814
815 t->parms.link = p->link;
816 mtu = ip_tunnel_bind_dev(dev);
817 if (set_mtu)
818 dev->mtu = mtu;
819 }
820 dst_cache_reset(&t->dst_cache);
821 netdev_state_change(dev);
822 }
823
824 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
825 {
826 int err = 0;
827 struct ip_tunnel *t = netdev_priv(dev);
828 struct net *net = t->net;
829 struct ip_tunnel_net *itn = net_generic(net, t->ip_tnl_net_id);
830
831 BUG_ON(!itn->fb_tunnel_dev);
832 switch (cmd) {
833 case SIOCGETTUNNEL:
834 if (dev == itn->fb_tunnel_dev) {
835 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
836 if (!t)
837 t = netdev_priv(dev);
838 }
839 memcpy(p, &t->parms, sizeof(*p));
840 break;
841
842 case SIOCADDTUNNEL:
843 case SIOCCHGTUNNEL:
844 err = -EPERM;
845 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
846 goto done;
847 if (p->iph.ttl)
848 p->iph.frag_off |= htons(IP_DF);
849 if (!(p->i_flags & VTI_ISVTI)) {
850 if (!(p->i_flags & TUNNEL_KEY))
851 p->i_key = 0;
852 if (!(p->o_flags & TUNNEL_KEY))
853 p->o_key = 0;
854 }
855
856 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
857
858 if (cmd == SIOCADDTUNNEL) {
859 if (!t) {
860 t = ip_tunnel_create(net, itn, p);
861 err = PTR_ERR_OR_ZERO(t);
862 break;
863 }
864
865 err = -EEXIST;
866 break;
867 }
868 if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
869 if (t) {
870 if (t->dev != dev) {
871 err = -EEXIST;
872 break;
873 }
874 } else {
875 unsigned int nflags = 0;
876
877 if (ipv4_is_multicast(p->iph.daddr))
878 nflags = IFF_BROADCAST;
879 else if (p->iph.daddr)
880 nflags = IFF_POINTOPOINT;
881
882 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
883 err = -EINVAL;
884 break;
885 }
886
887 t = netdev_priv(dev);
888 }
889 }
890
891 if (t) {
892 err = 0;
893 ip_tunnel_update(itn, t, dev, p, true);
894 } else {
895 err = -ENOENT;
896 }
897 break;
898
899 case SIOCDELTUNNEL:
900 err = -EPERM;
901 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
902 goto done;
903
904 if (dev == itn->fb_tunnel_dev) {
905 err = -ENOENT;
906 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
907 if (!t)
908 goto done;
909 err = -EPERM;
910 if (t == netdev_priv(itn->fb_tunnel_dev))
911 goto done;
912 dev = t->dev;
913 }
914 unregister_netdevice(dev);
915 err = 0;
916 break;
917
918 default:
919 err = -EINVAL;
920 }
921
922 done:
923 return err;
924 }
925 EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
926
927 int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict)
928 {
929 struct ip_tunnel *tunnel = netdev_priv(dev);
930 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
931 int max_mtu = 0xFFF8 - dev->hard_header_len - t_hlen;
932
933 if (new_mtu < 68)
934 return -EINVAL;
935
936 if (new_mtu > max_mtu) {
937 if (strict)
938 return -EINVAL;
939
940 new_mtu = max_mtu;
941 }
942
943 dev->mtu = new_mtu;
944 return 0;
945 }
946 EXPORT_SYMBOL_GPL(__ip_tunnel_change_mtu);
947
948 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
949 {
950 return __ip_tunnel_change_mtu(dev, new_mtu, true);
951 }
952 EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
953
954 static void ip_tunnel_dev_free(struct net_device *dev)
955 {
956 struct ip_tunnel *tunnel = netdev_priv(dev);
957
958 gro_cells_destroy(&tunnel->gro_cells);
959 dst_cache_destroy(&tunnel->dst_cache);
960 free_percpu(dev->tstats);
961 free_netdev(dev);
962 }
963
964 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
965 {
966 struct ip_tunnel *tunnel = netdev_priv(dev);
967 struct ip_tunnel_net *itn;
968
969 itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
970
971 if (itn->fb_tunnel_dev != dev) {
972 ip_tunnel_del(itn, netdev_priv(dev));
973 unregister_netdevice_queue(dev, head);
974 }
975 }
976 EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
977
978 struct net *ip_tunnel_get_link_net(const struct net_device *dev)
979 {
980 struct ip_tunnel *tunnel = netdev_priv(dev);
981
982 return tunnel->net;
983 }
984 EXPORT_SYMBOL(ip_tunnel_get_link_net);
985
986 int ip_tunnel_get_iflink(const struct net_device *dev)
987 {
988 struct ip_tunnel *tunnel = netdev_priv(dev);
989
990 return tunnel->parms.link;
991 }
992 EXPORT_SYMBOL(ip_tunnel_get_iflink);
993
994 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id,
995 struct rtnl_link_ops *ops, char *devname)
996 {
997 struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
998 struct ip_tunnel_parm parms;
999 unsigned int i;
1000
1001 for (i = 0; i < IP_TNL_HASH_SIZE; i++)
1002 INIT_HLIST_HEAD(&itn->tunnels[i]);
1003
1004 if (!ops) {
1005 itn->fb_tunnel_dev = NULL;
1006 return 0;
1007 }
1008
1009 memset(&parms, 0, sizeof(parms));
1010 if (devname)
1011 strlcpy(parms.name, devname, IFNAMSIZ);
1012
1013 rtnl_lock();
1014 itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
1015 /* FB netdevice is special: we have one, and only one per netns.
1016 * Allowing to move it to another netns is clearly unsafe.
1017 */
1018 if (!IS_ERR(itn->fb_tunnel_dev)) {
1019 itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1020 itn->fb_tunnel_dev->mtu = ip_tunnel_bind_dev(itn->fb_tunnel_dev);
1021 ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
1022 }
1023 rtnl_unlock();
1024
1025 return PTR_ERR_OR_ZERO(itn->fb_tunnel_dev);
1026 }
1027 EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
1028
1029 static void ip_tunnel_destroy(struct ip_tunnel_net *itn, struct list_head *head,
1030 struct rtnl_link_ops *ops)
1031 {
1032 struct net *net = dev_net(itn->fb_tunnel_dev);
1033 struct net_device *dev, *aux;
1034 int h;
1035
1036 for_each_netdev_safe(net, dev, aux)
1037 if (dev->rtnl_link_ops == ops)
1038 unregister_netdevice_queue(dev, head);
1039
1040 for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
1041 struct ip_tunnel *t;
1042 struct hlist_node *n;
1043 struct hlist_head *thead = &itn->tunnels[h];
1044
1045 hlist_for_each_entry_safe(t, n, thead, hash_node)
1046 /* If dev is in the same netns, it has already
1047 * been added to the list by the previous loop.
1048 */
1049 if (!net_eq(dev_net(t->dev), net))
1050 unregister_netdevice_queue(t->dev, head);
1051 }
1052 }
1053
1054 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops)
1055 {
1056 LIST_HEAD(list);
1057
1058 rtnl_lock();
1059 ip_tunnel_destroy(itn, &list, ops);
1060 unregister_netdevice_many(&list);
1061 rtnl_unlock();
1062 }
1063 EXPORT_SYMBOL_GPL(ip_tunnel_delete_net);
1064
1065 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
1066 struct ip_tunnel_parm *p)
1067 {
1068 struct ip_tunnel *nt;
1069 struct net *net = dev_net(dev);
1070 struct ip_tunnel_net *itn;
1071 int mtu;
1072 int err;
1073
1074 nt = netdev_priv(dev);
1075 itn = net_generic(net, nt->ip_tnl_net_id);
1076
1077 if (nt->collect_md) {
1078 if (rtnl_dereference(itn->collect_md_tun))
1079 return -EEXIST;
1080 } else {
1081 if (ip_tunnel_find(itn, p, dev->type))
1082 return -EEXIST;
1083 }
1084
1085 nt->net = net;
1086 nt->parms = *p;
1087 err = register_netdevice(dev);
1088 if (err)
1089 goto out;
1090
1091 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1092 eth_hw_addr_random(dev);
1093
1094 mtu = ip_tunnel_bind_dev(dev);
1095 if (!tb[IFLA_MTU])
1096 dev->mtu = mtu;
1097
1098 ip_tunnel_add(itn, nt);
1099 out:
1100 return err;
1101 }
1102 EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
1103
1104 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
1105 struct ip_tunnel_parm *p)
1106 {
1107 struct ip_tunnel *t;
1108 struct ip_tunnel *tunnel = netdev_priv(dev);
1109 struct net *net = tunnel->net;
1110 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
1111
1112 if (dev == itn->fb_tunnel_dev)
1113 return -EINVAL;
1114
1115 t = ip_tunnel_find(itn, p, dev->type);
1116
1117 if (t) {
1118 if (t->dev != dev)
1119 return -EEXIST;
1120 } else {
1121 t = tunnel;
1122
1123 if (dev->type != ARPHRD_ETHER) {
1124 unsigned int nflags = 0;
1125
1126 if (ipv4_is_multicast(p->iph.daddr))
1127 nflags = IFF_BROADCAST;
1128 else if (p->iph.daddr)
1129 nflags = IFF_POINTOPOINT;
1130
1131 if ((dev->flags ^ nflags) &
1132 (IFF_POINTOPOINT | IFF_BROADCAST))
1133 return -EINVAL;
1134 }
1135 }
1136
1137 ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU]);
1138 return 0;
1139 }
1140 EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1141
1142 int ip_tunnel_init(struct net_device *dev)
1143 {
1144 struct ip_tunnel *tunnel = netdev_priv(dev);
1145 struct iphdr *iph = &tunnel->parms.iph;
1146 int err;
1147
1148 dev->destructor = ip_tunnel_dev_free;
1149 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1150 if (!dev->tstats)
1151 return -ENOMEM;
1152
1153 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1154 if (err) {
1155 free_percpu(dev->tstats);
1156 return err;
1157 }
1158
1159 err = gro_cells_init(&tunnel->gro_cells, dev);
1160 if (err) {
1161 dst_cache_destroy(&tunnel->dst_cache);
1162 free_percpu(dev->tstats);
1163 return err;
1164 }
1165
1166 tunnel->dev = dev;
1167 tunnel->net = dev_net(dev);
1168 strcpy(tunnel->parms.name, dev->name);
1169 iph->version = 4;
1170 iph->ihl = 5;
1171
1172 if (tunnel->collect_md) {
1173 dev->features |= NETIF_F_NETNS_LOCAL;
1174 netif_keep_dst(dev);
1175 }
1176 return 0;
1177 }
1178 EXPORT_SYMBOL_GPL(ip_tunnel_init);
1179
1180 void ip_tunnel_uninit(struct net_device *dev)
1181 {
1182 struct ip_tunnel *tunnel = netdev_priv(dev);
1183 struct net *net = tunnel->net;
1184 struct ip_tunnel_net *itn;
1185
1186 itn = net_generic(net, tunnel->ip_tnl_net_id);
1187 /* fb_tunnel_dev will be unregisted in net-exit call. */
1188 if (itn->fb_tunnel_dev != dev)
1189 ip_tunnel_del(itn, netdev_priv(dev));
1190
1191 dst_cache_reset(&tunnel->dst_cache);
1192 }
1193 EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1194
1195 /* Do least required initialization, rest of init is done in tunnel_init call */
1196 void ip_tunnel_setup(struct net_device *dev, int net_id)
1197 {
1198 struct ip_tunnel *tunnel = netdev_priv(dev);
1199 tunnel->ip_tnl_net_id = net_id;
1200 }
1201 EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1202
1203 MODULE_LICENSE("GPL");
Linux with added FPC-III support