1 /*
2 * GRE over IPv6 protocol decoder.
3 *
4 * Authors: Dmitry Kozlov (xeb@mail.ru)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/mroute.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
35 #include <linux/hash.h>
36 #include <linux/if_tunnel.h>
37 #include <linux/ip6_tunnel.h>
38
39 #include <net/sock.h>
40 #include <net/ip.h>
41 #include <net/ip_tunnels.h>
42 #include <net/icmp.h>
43 #include <net/protocol.h>
44 #include <net/addrconf.h>
45 #include <net/arp.h>
46 #include <net/checksum.h>
47 #include <net/dsfield.h>
48 #include <net/inet_ecn.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/rtnetlink.h>
53
54 #include <net/ipv6.h>
55 #include <net/ip6_fib.h>
56 #include <net/ip6_route.h>
57 #include <net/ip6_tunnel.h>
58 #include <net/gre.h>
59
60
61 static bool log_ecn_error = true;
62 module_param(log_ecn_error, bool, 0644);
63 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
64
65 #define HASH_SIZE_SHIFT 5
66 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
67
68 static int ip6gre_net_id __read_mostly;
69 struct ip6gre_net {
70 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
71
72 struct net_device *fb_tunnel_dev;
73 };
74
75 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
76 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
77 static int ip6gre_tunnel_init(struct net_device *dev);
78 static void ip6gre_tunnel_setup(struct net_device *dev);
79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
80 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85 4 hash tables:
86
87 3: (remote,local)
88 2: (remote,*)
89 1: (*,local)
90 0: (*,*)
91
92 We require exact key match i.e. if a key is present in packet
93 it will match only tunnel with the same key; if it is not present,
94 it will match only keyless tunnel.
95
96 All keysless packets, if not matched configured keyless tunnels
97 will match fallback tunnel.
98 */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
HASH_ADDR(const struct in6_addr * addr)101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103 u32 hash = ipv6_addr_hash(addr);
104
105 return hash_32(hash, HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l tunnels[3]
109 #define tunnels_r tunnels[2]
110 #define tunnels_l tunnels[1]
111 #define tunnels_wc tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
ip6gre_tunnel_lookup(struct net_device * dev,const struct in6_addr * remote,const struct in6_addr * local,__be32 key,__be16 gre_proto)115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116 const struct in6_addr *remote, const struct in6_addr *local,
117 __be32 key, __be16 gre_proto)
118 {
119 struct net *net = dev_net(dev);
120 int link = dev->ifindex;
121 unsigned int h0 = HASH_ADDR(remote);
122 unsigned int h1 = HASH_KEY(key);
123 struct ip6_tnl *t, *cand = NULL;
124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
126 ARPHRD_ETHER : ARPHRD_IP6GRE;
127 int score, cand_score = 4;
128
129 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
130 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
131 !ipv6_addr_equal(remote, &t->parms.raddr) ||
132 key != t->parms.i_key ||
133 !(t->dev->flags & IFF_UP))
134 continue;
135
136 if (t->dev->type != ARPHRD_IP6GRE &&
137 t->dev->type != dev_type)
138 continue;
139
140 score = 0;
141 if (t->parms.link != link)
142 score |= 1;
143 if (t->dev->type != dev_type)
144 score |= 2;
145 if (score == 0)
146 return t;
147
148 if (score < cand_score) {
149 cand = t;
150 cand_score = score;
151 }
152 }
153
154 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
155 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
156 key != t->parms.i_key ||
157 !(t->dev->flags & IFF_UP))
158 continue;
159
160 if (t->dev->type != ARPHRD_IP6GRE &&
161 t->dev->type != dev_type)
162 continue;
163
164 score = 0;
165 if (t->parms.link != link)
166 score |= 1;
167 if (t->dev->type != dev_type)
168 score |= 2;
169 if (score == 0)
170 return t;
171
172 if (score < cand_score) {
173 cand = t;
174 cand_score = score;
175 }
176 }
177
178 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
179 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
180 (!ipv6_addr_equal(local, &t->parms.raddr) ||
181 !ipv6_addr_is_multicast(local))) ||
182 key != t->parms.i_key ||
183 !(t->dev->flags & IFF_UP))
184 continue;
185
186 if (t->dev->type != ARPHRD_IP6GRE &&
187 t->dev->type != dev_type)
188 continue;
189
190 score = 0;
191 if (t->parms.link != link)
192 score |= 1;
193 if (t->dev->type != dev_type)
194 score |= 2;
195 if (score == 0)
196 return t;
197
198 if (score < cand_score) {
199 cand = t;
200 cand_score = score;
201 }
202 }
203
204 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
205 if (t->parms.i_key != key ||
206 !(t->dev->flags & IFF_UP))
207 continue;
208
209 if (t->dev->type != ARPHRD_IP6GRE &&
210 t->dev->type != dev_type)
211 continue;
212
213 score = 0;
214 if (t->parms.link != link)
215 score |= 1;
216 if (t->dev->type != dev_type)
217 score |= 2;
218 if (score == 0)
219 return t;
220
221 if (score < cand_score) {
222 cand = t;
223 cand_score = score;
224 }
225 }
226
227 if (cand != NULL)
228 return cand;
229
230 dev = ign->fb_tunnel_dev;
231 if (dev->flags & IFF_UP)
232 return netdev_priv(dev);
233
234 return NULL;
235 }
236
__ip6gre_bucket(struct ip6gre_net * ign,const struct __ip6_tnl_parm * p)237 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
238 const struct __ip6_tnl_parm *p)
239 {
240 const struct in6_addr *remote = &p->raddr;
241 const struct in6_addr *local = &p->laddr;
242 unsigned int h = HASH_KEY(p->i_key);
243 int prio = 0;
244
245 if (!ipv6_addr_any(local))
246 prio |= 1;
247 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
248 prio |= 2;
249 h ^= HASH_ADDR(remote);
250 }
251
252 return &ign->tunnels[prio][h];
253 }
254
ip6gre_bucket(struct ip6gre_net * ign,const struct ip6_tnl * t)255 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
256 const struct ip6_tnl *t)
257 {
258 return __ip6gre_bucket(ign, &t->parms);
259 }
260
ip6gre_tunnel_link(struct ip6gre_net * ign,struct ip6_tnl * t)261 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
262 {
263 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
264
265 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
266 rcu_assign_pointer(*tp, t);
267 }
268
ip6gre_tunnel_unlink(struct ip6gre_net * ign,struct ip6_tnl * t)269 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
270 {
271 struct ip6_tnl __rcu **tp;
272 struct ip6_tnl *iter;
273
274 for (tp = ip6gre_bucket(ign, t);
275 (iter = rtnl_dereference(*tp)) != NULL;
276 tp = &iter->next) {
277 if (t == iter) {
278 rcu_assign_pointer(*tp, t->next);
279 break;
280 }
281 }
282 }
283
ip6gre_tunnel_find(struct net * net,const struct __ip6_tnl_parm * parms,int type)284 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
285 const struct __ip6_tnl_parm *parms,
286 int type)
287 {
288 const struct in6_addr *remote = &parms->raddr;
289 const struct in6_addr *local = &parms->laddr;
290 __be32 key = parms->i_key;
291 int link = parms->link;
292 struct ip6_tnl *t;
293 struct ip6_tnl __rcu **tp;
294 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
295
296 for (tp = __ip6gre_bucket(ign, parms);
297 (t = rtnl_dereference(*tp)) != NULL;
298 tp = &t->next)
299 if (ipv6_addr_equal(local, &t->parms.laddr) &&
300 ipv6_addr_equal(remote, &t->parms.raddr) &&
301 key == t->parms.i_key &&
302 link == t->parms.link &&
303 type == t->dev->type)
304 break;
305
306 return t;
307 }
308
ip6gre_tunnel_locate(struct net * net,const struct __ip6_tnl_parm * parms,int create)309 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
310 const struct __ip6_tnl_parm *parms, int create)
311 {
312 struct ip6_tnl *t, *nt;
313 struct net_device *dev;
314 char name[IFNAMSIZ];
315 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
316
317 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
318 if (t && create)
319 return NULL;
320 if (t || !create)
321 return t;
322
323 if (parms->name[0])
324 strlcpy(name, parms->name, IFNAMSIZ);
325 else
326 strcpy(name, "ip6gre%d");
327
328 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
329 ip6gre_tunnel_setup);
330 if (!dev)
331 return NULL;
332
333 dev_net_set(dev, net);
334
335 nt = netdev_priv(dev);
336 nt->parms = *parms;
337 dev->rtnl_link_ops = &ip6gre_link_ops;
338
339 nt->dev = dev;
340 nt->net = dev_net(dev);
341 ip6gre_tnl_link_config(nt, 1);
342
343 if (register_netdevice(dev) < 0)
344 goto failed_free;
345
346 /* Can use a lockless transmit, unless we generate output sequences */
347 if (!(nt->parms.o_flags & GRE_SEQ))
348 dev->features |= NETIF_F_LLTX;
349
350 dev_hold(dev);
351 ip6gre_tunnel_link(ign, nt);
352 return nt;
353
354 failed_free:
355 free_netdev(dev);
356 return NULL;
357 }
358
ip6gre_tunnel_uninit(struct net_device * dev)359 static void ip6gre_tunnel_uninit(struct net_device *dev)
360 {
361 struct ip6_tnl *t = netdev_priv(dev);
362 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
363
364 ip6gre_tunnel_unlink(ign, t);
365 ip6_tnl_dst_reset(netdev_priv(dev));
366 dev_put(dev);
367 }
368
369
ip6gre_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)370 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
371 u8 type, u8 code, int offset, __be32 info)
372 {
373 const struct gre_base_hdr *greh;
374 const struct ipv6hdr *ipv6h;
375 int grehlen = sizeof(*greh);
376 struct ip6_tnl *t;
377 int key_off = 0;
378 __be16 flags;
379 __be32 key;
380
381 if (!pskb_may_pull(skb, offset + grehlen))
382 return;
383 greh = (const struct gre_base_hdr *)(skb->data + offset);
384 flags = greh->flags;
385 if (flags & (GRE_VERSION | GRE_ROUTING))
386 return;
387 if (flags & GRE_CSUM)
388 grehlen += 4;
389 if (flags & GRE_KEY) {
390 key_off = grehlen + offset;
391 grehlen += 4;
392 }
393
394 if (!pskb_may_pull(skb, offset + grehlen))
395 return;
396 ipv6h = (const struct ipv6hdr *)skb->data;
397 greh = (const struct gre_base_hdr *)(skb->data + offset);
398 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
399
400 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
401 key, greh->protocol);
402 if (t == NULL)
403 return;
404
405 switch (type) {
406 __u32 teli;
407 struct ipv6_tlv_tnl_enc_lim *tel;
408 __u32 mtu;
409 case ICMPV6_DEST_UNREACH:
410 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
411 t->parms.name);
412 if (code != ICMPV6_PORT_UNREACH)
413 break;
414 return;
415 case ICMPV6_TIME_EXCEED:
416 if (code == ICMPV6_EXC_HOPLIMIT) {
417 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
418 t->parms.name);
419 break;
420 }
421 return;
422 case ICMPV6_PARAMPROB:
423 teli = 0;
424 if (code == ICMPV6_HDR_FIELD)
425 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
426
427 if (teli && teli == be32_to_cpu(info) - 2) {
428 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
429 if (tel->encap_limit == 0) {
430 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
431 t->parms.name);
432 }
433 } else {
434 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
435 t->parms.name);
436 }
437 return;
438 case ICMPV6_PKT_TOOBIG:
439 mtu = be32_to_cpu(info) - offset;
440 if (mtu < IPV6_MIN_MTU)
441 mtu = IPV6_MIN_MTU;
442 t->dev->mtu = mtu;
443 return;
444 }
445
446 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
447 t->err_count++;
448 else
449 t->err_count = 1;
450 t->err_time = jiffies;
451 }
452
ip6gre_rcv(struct sk_buff * skb)453 static int ip6gre_rcv(struct sk_buff *skb)
454 {
455 const struct ipv6hdr *ipv6h;
456 u8 *h;
457 __be16 flags;
458 __sum16 csum = 0;
459 __be32 key = 0;
460 u32 seqno = 0;
461 struct ip6_tnl *tunnel;
462 int offset = 4;
463 __be16 gre_proto;
464 int err;
465
466 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
467 goto drop;
468
469 ipv6h = ipv6_hdr(skb);
470 h = skb->data;
471 flags = *(__be16 *)h;
472
473 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
474 /* - Version must be 0.
475 - We do not support routing headers.
476 */
477 if (flags&(GRE_VERSION|GRE_ROUTING))
478 goto drop;
479
480 if (flags&GRE_CSUM) {
481 csum = skb_checksum_simple_validate(skb);
482 offset += 4;
483 }
484 if (flags&GRE_KEY) {
485 key = *(__be32 *)(h + offset);
486 offset += 4;
487 }
488 if (flags&GRE_SEQ) {
489 seqno = ntohl(*(__be32 *)(h + offset));
490 offset += 4;
491 }
492 }
493
494 gre_proto = *(__be16 *)(h + 2);
495
496 tunnel = ip6gre_tunnel_lookup(skb->dev,
497 &ipv6h->saddr, &ipv6h->daddr, key,
498 gre_proto);
499 if (tunnel) {
500 struct pcpu_sw_netstats *tstats;
501
502 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
503 goto drop;
504
505 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
506 tunnel->dev->stats.rx_dropped++;
507 goto drop;
508 }
509
510 skb->protocol = gre_proto;
511 /* WCCP version 1 and 2 protocol decoding.
512 * - Change protocol to IPv6
513 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
514 */
515 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
516 skb->protocol = htons(ETH_P_IPV6);
517 if ((*(h + offset) & 0xF0) != 0x40)
518 offset += 4;
519 }
520
521 skb->mac_header = skb->network_header;
522 __pskb_pull(skb, offset);
523 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
524
525 if (((flags&GRE_CSUM) && csum) ||
526 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
527 tunnel->dev->stats.rx_crc_errors++;
528 tunnel->dev->stats.rx_errors++;
529 goto drop;
530 }
531 if (tunnel->parms.i_flags&GRE_SEQ) {
532 if (!(flags&GRE_SEQ) ||
533 (tunnel->i_seqno &&
534 (s32)(seqno - tunnel->i_seqno) < 0)) {
535 tunnel->dev->stats.rx_fifo_errors++;
536 tunnel->dev->stats.rx_errors++;
537 goto drop;
538 }
539 tunnel->i_seqno = seqno + 1;
540 }
541
542 /* Warning: All skb pointers will be invalidated! */
543 if (tunnel->dev->type == ARPHRD_ETHER) {
544 if (!pskb_may_pull(skb, ETH_HLEN)) {
545 tunnel->dev->stats.rx_length_errors++;
546 tunnel->dev->stats.rx_errors++;
547 goto drop;
548 }
549
550 ipv6h = ipv6_hdr(skb);
551 skb->protocol = eth_type_trans(skb, tunnel->dev);
552 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
553 }
554
555 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
556
557 skb_reset_network_header(skb);
558
559 err = IP6_ECN_decapsulate(ipv6h, skb);
560 if (unlikely(err)) {
561 if (log_ecn_error)
562 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
563 &ipv6h->saddr,
564 ipv6_get_dsfield(ipv6h));
565 if (err > 1) {
566 ++tunnel->dev->stats.rx_frame_errors;
567 ++tunnel->dev->stats.rx_errors;
568 goto drop;
569 }
570 }
571
572 tstats = this_cpu_ptr(tunnel->dev->tstats);
573 u64_stats_update_begin(&tstats->syncp);
574 tstats->rx_packets++;
575 tstats->rx_bytes += skb->len;
576 u64_stats_update_end(&tstats->syncp);
577
578 netif_rx(skb);
579
580 return 0;
581 }
582 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
583
584 drop:
585 kfree_skb(skb);
586 return 0;
587 }
588
589 struct ipv6_tel_txoption {
590 struct ipv6_txoptions ops;
591 __u8 dst_opt[8];
592 };
593
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)594 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
595 {
596 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
597
598 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
599 opt->dst_opt[3] = 1;
600 opt->dst_opt[4] = encap_limit;
601 opt->dst_opt[5] = IPV6_TLV_PADN;
602 opt->dst_opt[6] = 1;
603
604 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
605 opt->ops.opt_nflen = 8;
606 }
607
ip6gre_xmit2(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu)608 static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
609 struct net_device *dev,
610 __u8 dsfield,
611 struct flowi6 *fl6,
612 int encap_limit,
613 __u32 *pmtu)
614 {
615 struct ip6_tnl *tunnel = netdev_priv(dev);
616 struct net *net = tunnel->net;
617 struct net_device *tdev; /* Device to other host */
618 struct ipv6hdr *ipv6h; /* Our new IP header */
619 unsigned int max_headroom = 0; /* The extra header space needed */
620 int gre_hlen;
621 struct ipv6_tel_txoption opt;
622 int mtu;
623 struct dst_entry *dst = NULL, *ndst = NULL;
624 struct net_device_stats *stats = &tunnel->dev->stats;
625 int err = -1;
626 u8 proto;
627 struct sk_buff *new_skb;
628 __be16 protocol;
629
630 if (dev->type == ARPHRD_ETHER)
631 IPCB(skb)->flags = 0;
632
633 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
634 gre_hlen = 0;
635 ipv6h = (struct ipv6hdr *)skb->data;
636 fl6->daddr = ipv6h->daddr;
637 } else {
638 gre_hlen = tunnel->hlen;
639 fl6->daddr = tunnel->parms.raddr;
640 }
641
642 if (!fl6->flowi6_mark)
643 dst = ip6_tnl_dst_check(tunnel);
644
645 if (!dst) {
646 ndst = ip6_route_output(net, NULL, fl6);
647
648 if (ndst->error)
649 goto tx_err_link_failure;
650 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
651 if (IS_ERR(ndst)) {
652 err = PTR_ERR(ndst);
653 ndst = NULL;
654 goto tx_err_link_failure;
655 }
656 dst = ndst;
657 }
658
659 tdev = dst->dev;
660
661 if (tdev == dev) {
662 stats->collisions++;
663 net_warn_ratelimited("%s: Local routing loop detected!\n",
664 tunnel->parms.name);
665 goto tx_err_dst_release;
666 }
667
668 mtu = dst_mtu(dst) - sizeof(*ipv6h);
669 if (encap_limit >= 0) {
670 max_headroom += 8;
671 mtu -= 8;
672 }
673 if (mtu < IPV6_MIN_MTU)
674 mtu = IPV6_MIN_MTU;
675 if (skb_dst(skb))
676 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
677 if (skb->len > mtu) {
678 *pmtu = mtu;
679 err = -EMSGSIZE;
680 goto tx_err_dst_release;
681 }
682
683 if (tunnel->err_count > 0) {
684 if (time_before(jiffies,
685 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
686 tunnel->err_count--;
687
688 dst_link_failure(skb);
689 } else
690 tunnel->err_count = 0;
691 }
692
693 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
694
695 max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
696
697 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
698 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
699 new_skb = skb_realloc_headroom(skb, max_headroom);
700 if (max_headroom > dev->needed_headroom)
701 dev->needed_headroom = max_headroom;
702 if (!new_skb)
703 goto tx_err_dst_release;
704
705 if (skb->sk)
706 skb_set_owner_w(new_skb, skb->sk);
707 consume_skb(skb);
708 skb = new_skb;
709 }
710
711 if (fl6->flowi6_mark) {
712 skb_dst_set(skb, dst);
713 ndst = NULL;
714 } else {
715 skb_dst_set_noref(skb, dst);
716 }
717
718 proto = NEXTHDR_GRE;
719 if (encap_limit >= 0) {
720 init_tel_txopt(&opt, encap_limit);
721 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
722 }
723
724 if (likely(!skb->encapsulation)) {
725 skb_reset_inner_headers(skb);
726 skb->encapsulation = 1;
727 }
728
729 skb_push(skb, gre_hlen);
730 skb_reset_network_header(skb);
731 skb_set_transport_header(skb, sizeof(*ipv6h));
732
733 /*
734 * Push down and install the IP header.
735 */
736 ipv6h = ipv6_hdr(skb);
737 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
738 ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
739 ipv6h->hop_limit = tunnel->parms.hop_limit;
740 ipv6h->nexthdr = proto;
741 ipv6h->saddr = fl6->saddr;
742 ipv6h->daddr = fl6->daddr;
743
744 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
745 protocol = (dev->type == ARPHRD_ETHER) ?
746 htons(ETH_P_TEB) : skb->protocol;
747 ((__be16 *)(ipv6h + 1))[1] = protocol;
748
749 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
750 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
751
752 if (tunnel->parms.o_flags&GRE_SEQ) {
753 ++tunnel->o_seqno;
754 *ptr = htonl(tunnel->o_seqno);
755 ptr--;
756 }
757 if (tunnel->parms.o_flags&GRE_KEY) {
758 *ptr = tunnel->parms.o_key;
759 ptr--;
760 }
761 if (tunnel->parms.o_flags&GRE_CSUM) {
762 *ptr = 0;
763 *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
764 skb->len - sizeof(struct ipv6hdr));
765 }
766 }
767
768 skb_set_inner_protocol(skb, protocol);
769
770 ip6tunnel_xmit(skb, dev);
771 if (ndst)
772 ip6_tnl_dst_store(tunnel, ndst);
773 return 0;
774 tx_err_link_failure:
775 stats->tx_carrier_errors++;
776 dst_link_failure(skb);
777 tx_err_dst_release:
778 dst_release(ndst);
779 return err;
780 }
781
ip6gre_xmit_ipv4(struct sk_buff * skb,struct net_device * dev)782 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
783 {
784 struct ip6_tnl *t = netdev_priv(dev);
785 const struct iphdr *iph = ip_hdr(skb);
786 int encap_limit = -1;
787 struct flowi6 fl6;
788 __u8 dsfield;
789 __u32 mtu;
790 int err;
791
792 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
793 encap_limit = t->parms.encap_limit;
794
795 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
796 fl6.flowi6_proto = IPPROTO_GRE;
797
798 dsfield = ipv4_get_dsfield(iph);
799
800 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
801 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
802 & IPV6_TCLASS_MASK;
803 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
804 fl6.flowi6_mark = skb->mark;
805
806 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
807
808 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
809 if (err != 0) {
810 /* XXX: send ICMP error even if DF is not set. */
811 if (err == -EMSGSIZE)
812 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
813 htonl(mtu));
814 return -1;
815 }
816
817 return 0;
818 }
819
ip6gre_xmit_ipv6(struct sk_buff * skb,struct net_device * dev)820 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
821 {
822 struct ip6_tnl *t = netdev_priv(dev);
823 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
824 int encap_limit = -1;
825 __u16 offset;
826 struct flowi6 fl6;
827 __u8 dsfield;
828 __u32 mtu;
829 int err;
830
831 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
832 return -1;
833
834 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
835 if (offset > 0) {
836 struct ipv6_tlv_tnl_enc_lim *tel;
837 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
838 if (tel->encap_limit == 0) {
839 icmpv6_send(skb, ICMPV6_PARAMPROB,
840 ICMPV6_HDR_FIELD, offset + 2);
841 return -1;
842 }
843 encap_limit = tel->encap_limit - 1;
844 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
845 encap_limit = t->parms.encap_limit;
846
847 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
848 fl6.flowi6_proto = IPPROTO_GRE;
849
850 dsfield = ipv6_get_dsfield(ipv6h);
851 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
852 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
853 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
854 fl6.flowlabel |= ip6_flowlabel(ipv6h);
855 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
856 fl6.flowi6_mark = skb->mark;
857
858 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
859
860 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
861 if (err != 0) {
862 if (err == -EMSGSIZE)
863 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
864 return -1;
865 }
866
867 return 0;
868 }
869
870 /**
871 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
872 * @t: the outgoing tunnel device
873 * @hdr: IPv6 header from the incoming packet
874 *
875 * Description:
876 * Avoid trivial tunneling loop by checking that tunnel exit-point
877 * doesn't match source of incoming packet.
878 *
879 * Return:
880 * 1 if conflict,
881 * 0 else
882 **/
883
ip6gre_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)884 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
885 const struct ipv6hdr *hdr)
886 {
887 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
888 }
889
ip6gre_xmit_other(struct sk_buff * skb,struct net_device * dev)890 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
891 {
892 struct ip6_tnl *t = netdev_priv(dev);
893 int encap_limit = -1;
894 struct flowi6 fl6;
895 __u32 mtu;
896 int err;
897
898 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
899 encap_limit = t->parms.encap_limit;
900
901 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
902
903 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
904
905 return err;
906 }
907
ip6gre_tunnel_xmit(struct sk_buff * skb,struct net_device * dev)908 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
909 struct net_device *dev)
910 {
911 struct ip6_tnl *t = netdev_priv(dev);
912 struct net_device_stats *stats = &t->dev->stats;
913 int ret;
914
915 if (!ip6_tnl_xmit_ctl(t))
916 goto tx_err;
917
918 switch (skb->protocol) {
919 case htons(ETH_P_IP):
920 ret = ip6gre_xmit_ipv4(skb, dev);
921 break;
922 case htons(ETH_P_IPV6):
923 ret = ip6gre_xmit_ipv6(skb, dev);
924 break;
925 default:
926 ret = ip6gre_xmit_other(skb, dev);
927 break;
928 }
929
930 if (ret < 0)
931 goto tx_err;
932
933 return NETDEV_TX_OK;
934
935 tx_err:
936 stats->tx_errors++;
937 stats->tx_dropped++;
938 kfree_skb(skb);
939 return NETDEV_TX_OK;
940 }
941
ip6gre_tnl_link_config(struct ip6_tnl * t,int set_mtu)942 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
943 {
944 struct net_device *dev = t->dev;
945 struct __ip6_tnl_parm *p = &t->parms;
946 struct flowi6 *fl6 = &t->fl.u.ip6;
947 int addend = sizeof(struct ipv6hdr) + 4;
948
949 if (dev->type != ARPHRD_ETHER) {
950 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
951 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
952 }
953
954 /* Set up flowi template */
955 fl6->saddr = p->laddr;
956 fl6->daddr = p->raddr;
957 fl6->flowi6_oif = p->link;
958 fl6->flowlabel = 0;
959
960 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
961 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
962 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
963 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
964
965 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
966 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
967
968 if (p->flags&IP6_TNL_F_CAP_XMIT &&
969 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
970 dev->flags |= IFF_POINTOPOINT;
971 else
972 dev->flags &= ~IFF_POINTOPOINT;
973
974 /* Precalculate GRE options length */
975 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
976 if (t->parms.o_flags&GRE_CSUM)
977 addend += 4;
978 if (t->parms.o_flags&GRE_KEY)
979 addend += 4;
980 if (t->parms.o_flags&GRE_SEQ)
981 addend += 4;
982 }
983 t->hlen = addend;
984
985 if (p->flags & IP6_TNL_F_CAP_XMIT) {
986 int strict = (ipv6_addr_type(&p->raddr) &
987 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
988
989 struct rt6_info *rt = rt6_lookup(t->net,
990 &p->raddr, &p->laddr,
991 p->link, strict);
992
993 if (rt == NULL)
994 return;
995
996 if (rt->dst.dev) {
997 dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
998
999 if (set_mtu) {
1000 dev->mtu = rt->dst.dev->mtu - addend;
1001 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1002 dev->mtu -= 8;
1003
1004 if (dev->mtu < IPV6_MIN_MTU)
1005 dev->mtu = IPV6_MIN_MTU;
1006 }
1007 }
1008 ip6_rt_put(rt);
1009 }
1010 }
1011
ip6gre_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p,int set_mtu)1012 static int ip6gre_tnl_change(struct ip6_tnl *t,
1013 const struct __ip6_tnl_parm *p, int set_mtu)
1014 {
1015 t->parms.laddr = p->laddr;
1016 t->parms.raddr = p->raddr;
1017 t->parms.flags = p->flags;
1018 t->parms.hop_limit = p->hop_limit;
1019 t->parms.encap_limit = p->encap_limit;
1020 t->parms.flowinfo = p->flowinfo;
1021 t->parms.link = p->link;
1022 t->parms.proto = p->proto;
1023 t->parms.i_key = p->i_key;
1024 t->parms.o_key = p->o_key;
1025 t->parms.i_flags = p->i_flags;
1026 t->parms.o_flags = p->o_flags;
1027 ip6_tnl_dst_reset(t);
1028 ip6gre_tnl_link_config(t, set_mtu);
1029 return 0;
1030 }
1031
ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm2 * u)1032 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1033 const struct ip6_tnl_parm2 *u)
1034 {
1035 p->laddr = u->laddr;
1036 p->raddr = u->raddr;
1037 p->flags = u->flags;
1038 p->hop_limit = u->hop_limit;
1039 p->encap_limit = u->encap_limit;
1040 p->flowinfo = u->flowinfo;
1041 p->link = u->link;
1042 p->i_key = u->i_key;
1043 p->o_key = u->o_key;
1044 p->i_flags = u->i_flags;
1045 p->o_flags = u->o_flags;
1046 memcpy(p->name, u->name, sizeof(u->name));
1047 }
1048
ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 * u,const struct __ip6_tnl_parm * p)1049 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1050 const struct __ip6_tnl_parm *p)
1051 {
1052 u->proto = IPPROTO_GRE;
1053 u->laddr = p->laddr;
1054 u->raddr = p->raddr;
1055 u->flags = p->flags;
1056 u->hop_limit = p->hop_limit;
1057 u->encap_limit = p->encap_limit;
1058 u->flowinfo = p->flowinfo;
1059 u->link = p->link;
1060 u->i_key = p->i_key;
1061 u->o_key = p->o_key;
1062 u->i_flags = p->i_flags;
1063 u->o_flags = p->o_flags;
1064 memcpy(u->name, p->name, sizeof(u->name));
1065 }
1066
ip6gre_tunnel_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)1067 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1068 struct ifreq *ifr, int cmd)
1069 {
1070 int err = 0;
1071 struct ip6_tnl_parm2 p;
1072 struct __ip6_tnl_parm p1;
1073 struct ip6_tnl *t = netdev_priv(dev);
1074 struct net *net = t->net;
1075 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1076
1077 switch (cmd) {
1078 case SIOCGETTUNNEL:
1079 if (dev == ign->fb_tunnel_dev) {
1080 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1081 err = -EFAULT;
1082 break;
1083 }
1084 ip6gre_tnl_parm_from_user(&p1, &p);
1085 t = ip6gre_tunnel_locate(net, &p1, 0);
1086 if (t == NULL)
1087 t = netdev_priv(dev);
1088 }
1089 memset(&p, 0, sizeof(p));
1090 ip6gre_tnl_parm_to_user(&p, &t->parms);
1091 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1092 err = -EFAULT;
1093 break;
1094
1095 case SIOCADDTUNNEL:
1096 case SIOCCHGTUNNEL:
1097 err = -EPERM;
1098 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1099 goto done;
1100
1101 err = -EFAULT;
1102 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1103 goto done;
1104
1105 err = -EINVAL;
1106 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1107 goto done;
1108
1109 if (!(p.i_flags&GRE_KEY))
1110 p.i_key = 0;
1111 if (!(p.o_flags&GRE_KEY))
1112 p.o_key = 0;
1113
1114 ip6gre_tnl_parm_from_user(&p1, &p);
1115 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1116
1117 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1118 if (t != NULL) {
1119 if (t->dev != dev) {
1120 err = -EEXIST;
1121 break;
1122 }
1123 } else {
1124 t = netdev_priv(dev);
1125
1126 ip6gre_tunnel_unlink(ign, t);
1127 synchronize_net();
1128 ip6gre_tnl_change(t, &p1, 1);
1129 ip6gre_tunnel_link(ign, t);
1130 netdev_state_change(dev);
1131 }
1132 }
1133
1134 if (t) {
1135 err = 0;
1136
1137 memset(&p, 0, sizeof(p));
1138 ip6gre_tnl_parm_to_user(&p, &t->parms);
1139 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1140 err = -EFAULT;
1141 } else
1142 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1143 break;
1144
1145 case SIOCDELTUNNEL:
1146 err = -EPERM;
1147 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1148 goto done;
1149
1150 if (dev == ign->fb_tunnel_dev) {
1151 err = -EFAULT;
1152 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1153 goto done;
1154 err = -ENOENT;
1155 ip6gre_tnl_parm_from_user(&p1, &p);
1156 t = ip6gre_tunnel_locate(net, &p1, 0);
1157 if (t == NULL)
1158 goto done;
1159 err = -EPERM;
1160 if (t == netdev_priv(ign->fb_tunnel_dev))
1161 goto done;
1162 dev = t->dev;
1163 }
1164 unregister_netdevice(dev);
1165 err = 0;
1166 break;
1167
1168 default:
1169 err = -EINVAL;
1170 }
1171
1172 done:
1173 return err;
1174 }
1175
ip6gre_tunnel_change_mtu(struct net_device * dev,int new_mtu)1176 static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1177 {
1178 if (new_mtu < 68 ||
1179 new_mtu > 0xFFF8 - dev->hard_header_len)
1180 return -EINVAL;
1181 dev->mtu = new_mtu;
1182 return 0;
1183 }
1184
ip6gre_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)1185 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1186 unsigned short type,
1187 const void *daddr, const void *saddr, unsigned int len)
1188 {
1189 struct ip6_tnl *t = netdev_priv(dev);
1190 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1191 __be16 *p = (__be16 *)(ipv6h+1);
1192
1193 ip6_flow_hdr(ipv6h, 0,
1194 ip6_make_flowlabel(dev_net(dev), skb,
1195 t->fl.u.ip6.flowlabel, false));
1196 ipv6h->hop_limit = t->parms.hop_limit;
1197 ipv6h->nexthdr = NEXTHDR_GRE;
1198 ipv6h->saddr = t->parms.laddr;
1199 ipv6h->daddr = t->parms.raddr;
1200
1201 p[0] = t->parms.o_flags;
1202 p[1] = htons(type);
1203
1204 /*
1205 * Set the source hardware address.
1206 */
1207
1208 if (saddr)
1209 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1210 if (daddr)
1211 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1212 if (!ipv6_addr_any(&ipv6h->daddr))
1213 return t->hlen;
1214
1215 return -t->hlen;
1216 }
1217
1218 static const struct header_ops ip6gre_header_ops = {
1219 .create = ip6gre_header,
1220 };
1221
1222 static const struct net_device_ops ip6gre_netdev_ops = {
1223 .ndo_init = ip6gre_tunnel_init,
1224 .ndo_uninit = ip6gre_tunnel_uninit,
1225 .ndo_start_xmit = ip6gre_tunnel_xmit,
1226 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1227 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1228 .ndo_get_stats64 = ip_tunnel_get_stats64,
1229 };
1230
ip6gre_dev_free(struct net_device * dev)1231 static void ip6gre_dev_free(struct net_device *dev)
1232 {
1233 free_percpu(dev->tstats);
1234 free_netdev(dev);
1235 }
1236
ip6gre_tunnel_setup(struct net_device * dev)1237 static void ip6gre_tunnel_setup(struct net_device *dev)
1238 {
1239 struct ip6_tnl *t;
1240
1241 dev->netdev_ops = &ip6gre_netdev_ops;
1242 dev->destructor = ip6gre_dev_free;
1243
1244 dev->type = ARPHRD_IP6GRE;
1245 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1246 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1247 t = netdev_priv(dev);
1248 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1249 dev->mtu -= 8;
1250 dev->flags |= IFF_NOARP;
1251 dev->iflink = 0;
1252 dev->addr_len = sizeof(struct in6_addr);
1253 netif_keep_dst(dev);
1254 }
1255
ip6gre_tunnel_init(struct net_device * dev)1256 static int ip6gre_tunnel_init(struct net_device *dev)
1257 {
1258 struct ip6_tnl *tunnel;
1259 int i;
1260
1261 tunnel = netdev_priv(dev);
1262
1263 tunnel->dev = dev;
1264 tunnel->net = dev_net(dev);
1265 strcpy(tunnel->parms.name, dev->name);
1266
1267 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1268 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1269
1270 if (ipv6_addr_any(&tunnel->parms.raddr))
1271 dev->header_ops = &ip6gre_header_ops;
1272
1273 dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1274 if (!dev->tstats)
1275 return -ENOMEM;
1276
1277 for_each_possible_cpu(i) {
1278 struct pcpu_sw_netstats *ip6gre_tunnel_stats;
1279 ip6gre_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1280 u64_stats_init(&ip6gre_tunnel_stats->syncp);
1281 }
1282
1283 dev->iflink = tunnel->parms.link;
1284
1285 return 0;
1286 }
1287
ip6gre_fb_tunnel_init(struct net_device * dev)1288 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1289 {
1290 struct ip6_tnl *tunnel = netdev_priv(dev);
1291
1292 tunnel->dev = dev;
1293 tunnel->net = dev_net(dev);
1294 strcpy(tunnel->parms.name, dev->name);
1295
1296 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1297
1298 dev_hold(dev);
1299 }
1300
1301
1302 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1303 .handler = ip6gre_rcv,
1304 .err_handler = ip6gre_err,
1305 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1306 };
1307
ip6gre_destroy_tunnels(struct net * net,struct list_head * head)1308 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1309 {
1310 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1311 struct net_device *dev, *aux;
1312 int prio;
1313
1314 for_each_netdev_safe(net, dev, aux)
1315 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1316 dev->rtnl_link_ops == &ip6gre_tap_ops)
1317 unregister_netdevice_queue(dev, head);
1318
1319 for (prio = 0; prio < 4; prio++) {
1320 int h;
1321 for (h = 0; h < HASH_SIZE; h++) {
1322 struct ip6_tnl *t;
1323
1324 t = rtnl_dereference(ign->tunnels[prio][h]);
1325
1326 while (t != NULL) {
1327 /* If dev is in the same netns, it has already
1328 * been added to the list by the previous loop.
1329 */
1330 if (!net_eq(dev_net(t->dev), net))
1331 unregister_netdevice_queue(t->dev,
1332 head);
1333 t = rtnl_dereference(t->next);
1334 }
1335 }
1336 }
1337 }
1338
ip6gre_init_net(struct net * net)1339 static int __net_init ip6gre_init_net(struct net *net)
1340 {
1341 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1342 int err;
1343
1344 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1345 NET_NAME_UNKNOWN,
1346 ip6gre_tunnel_setup);
1347 if (!ign->fb_tunnel_dev) {
1348 err = -ENOMEM;
1349 goto err_alloc_dev;
1350 }
1351 dev_net_set(ign->fb_tunnel_dev, net);
1352 /* FB netdevice is special: we have one, and only one per netns.
1353 * Allowing to move it to another netns is clearly unsafe.
1354 */
1355 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1356
1357
1358 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1359 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1360
1361 err = register_netdev(ign->fb_tunnel_dev);
1362 if (err)
1363 goto err_reg_dev;
1364
1365 rcu_assign_pointer(ign->tunnels_wc[0],
1366 netdev_priv(ign->fb_tunnel_dev));
1367 return 0;
1368
1369 err_reg_dev:
1370 ip6gre_dev_free(ign->fb_tunnel_dev);
1371 err_alloc_dev:
1372 return err;
1373 }
1374
ip6gre_exit_net(struct net * net)1375 static void __net_exit ip6gre_exit_net(struct net *net)
1376 {
1377 LIST_HEAD(list);
1378
1379 rtnl_lock();
1380 ip6gre_destroy_tunnels(net, &list);
1381 unregister_netdevice_many(&list);
1382 rtnl_unlock();
1383 }
1384
1385 static struct pernet_operations ip6gre_net_ops = {
1386 .init = ip6gre_init_net,
1387 .exit = ip6gre_exit_net,
1388 .id = &ip6gre_net_id,
1389 .size = sizeof(struct ip6gre_net),
1390 };
1391
ip6gre_tunnel_validate(struct nlattr * tb[],struct nlattr * data[])1392 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1393 {
1394 __be16 flags;
1395
1396 if (!data)
1397 return 0;
1398
1399 flags = 0;
1400 if (data[IFLA_GRE_IFLAGS])
1401 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1402 if (data[IFLA_GRE_OFLAGS])
1403 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1404 if (flags & (GRE_VERSION|GRE_ROUTING))
1405 return -EINVAL;
1406
1407 return 0;
1408 }
1409
ip6gre_tap_validate(struct nlattr * tb[],struct nlattr * data[])1410 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1411 {
1412 struct in6_addr daddr;
1413
1414 if (tb[IFLA_ADDRESS]) {
1415 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1416 return -EINVAL;
1417 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1418 return -EADDRNOTAVAIL;
1419 }
1420
1421 if (!data)
1422 goto out;
1423
1424 if (data[IFLA_GRE_REMOTE]) {
1425 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1426 if (ipv6_addr_any(&daddr))
1427 return -EINVAL;
1428 }
1429
1430 out:
1431 return ip6gre_tunnel_validate(tb, data);
1432 }
1433
1434
ip6gre_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1435 static void ip6gre_netlink_parms(struct nlattr *data[],
1436 struct __ip6_tnl_parm *parms)
1437 {
1438 memset(parms, 0, sizeof(*parms));
1439
1440 if (!data)
1441 return;
1442
1443 if (data[IFLA_GRE_LINK])
1444 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1445
1446 if (data[IFLA_GRE_IFLAGS])
1447 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1448
1449 if (data[IFLA_GRE_OFLAGS])
1450 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1451
1452 if (data[IFLA_GRE_IKEY])
1453 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1454
1455 if (data[IFLA_GRE_OKEY])
1456 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1457
1458 if (data[IFLA_GRE_LOCAL])
1459 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1460
1461 if (data[IFLA_GRE_REMOTE])
1462 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1463
1464 if (data[IFLA_GRE_TTL])
1465 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1466
1467 if (data[IFLA_GRE_ENCAP_LIMIT])
1468 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1469
1470 if (data[IFLA_GRE_FLOWINFO])
1471 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1472
1473 if (data[IFLA_GRE_FLAGS])
1474 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1475 }
1476
ip6gre_tap_init(struct net_device * dev)1477 static int ip6gre_tap_init(struct net_device *dev)
1478 {
1479 struct ip6_tnl *tunnel;
1480
1481 tunnel = netdev_priv(dev);
1482
1483 tunnel->dev = dev;
1484 tunnel->net = dev_net(dev);
1485 strcpy(tunnel->parms.name, dev->name);
1486
1487 ip6gre_tnl_link_config(tunnel, 1);
1488
1489 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1490 if (!dev->tstats)
1491 return -ENOMEM;
1492
1493 dev->iflink = tunnel->parms.link;
1494
1495 return 0;
1496 }
1497
1498 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1499 .ndo_init = ip6gre_tap_init,
1500 .ndo_uninit = ip6gre_tunnel_uninit,
1501 .ndo_start_xmit = ip6gre_tunnel_xmit,
1502 .ndo_set_mac_address = eth_mac_addr,
1503 .ndo_validate_addr = eth_validate_addr,
1504 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1505 .ndo_get_stats64 = ip_tunnel_get_stats64,
1506 };
1507
ip6gre_tap_setup(struct net_device * dev)1508 static void ip6gre_tap_setup(struct net_device *dev)
1509 {
1510
1511 ether_setup(dev);
1512
1513 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1514 dev->destructor = ip6gre_dev_free;
1515
1516 dev->iflink = 0;
1517 dev->features |= NETIF_F_NETNS_LOCAL;
1518 }
1519
ip6gre_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])1520 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1521 struct nlattr *tb[], struct nlattr *data[])
1522 {
1523 struct ip6_tnl *nt;
1524 struct net *net = dev_net(dev);
1525 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1526 int err;
1527
1528 nt = netdev_priv(dev);
1529 ip6gre_netlink_parms(data, &nt->parms);
1530
1531 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1532 return -EEXIST;
1533
1534 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1535 eth_hw_addr_random(dev);
1536
1537 nt->dev = dev;
1538 nt->net = dev_net(dev);
1539 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1540
1541 /* Can use a lockless transmit, unless we generate output sequences */
1542 if (!(nt->parms.o_flags & GRE_SEQ))
1543 dev->features |= NETIF_F_LLTX;
1544
1545 err = register_netdevice(dev);
1546 if (err)
1547 goto out;
1548
1549 dev_hold(dev);
1550 ip6gre_tunnel_link(ign, nt);
1551
1552 out:
1553 return err;
1554 }
1555
ip6gre_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])1556 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1557 struct nlattr *data[])
1558 {
1559 struct ip6_tnl *t, *nt = netdev_priv(dev);
1560 struct net *net = nt->net;
1561 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1562 struct __ip6_tnl_parm p;
1563
1564 if (dev == ign->fb_tunnel_dev)
1565 return -EINVAL;
1566
1567 ip6gre_netlink_parms(data, &p);
1568
1569 t = ip6gre_tunnel_locate(net, &p, 0);
1570
1571 if (t) {
1572 if (t->dev != dev)
1573 return -EEXIST;
1574 } else {
1575 t = nt;
1576 }
1577
1578 ip6gre_tunnel_unlink(ign, t);
1579 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1580 ip6gre_tunnel_link(ign, t);
1581 return 0;
1582 }
1583
ip6gre_dellink(struct net_device * dev,struct list_head * head)1584 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1585 {
1586 struct net *net = dev_net(dev);
1587 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1588
1589 if (dev != ign->fb_tunnel_dev)
1590 unregister_netdevice_queue(dev, head);
1591 }
1592
ip6gre_get_size(const struct net_device * dev)1593 static size_t ip6gre_get_size(const struct net_device *dev)
1594 {
1595 return
1596 /* IFLA_GRE_LINK */
1597 nla_total_size(4) +
1598 /* IFLA_GRE_IFLAGS */
1599 nla_total_size(2) +
1600 /* IFLA_GRE_OFLAGS */
1601 nla_total_size(2) +
1602 /* IFLA_GRE_IKEY */
1603 nla_total_size(4) +
1604 /* IFLA_GRE_OKEY */
1605 nla_total_size(4) +
1606 /* IFLA_GRE_LOCAL */
1607 nla_total_size(sizeof(struct in6_addr)) +
1608 /* IFLA_GRE_REMOTE */
1609 nla_total_size(sizeof(struct in6_addr)) +
1610 /* IFLA_GRE_TTL */
1611 nla_total_size(1) +
1612 /* IFLA_GRE_TOS */
1613 nla_total_size(1) +
1614 /* IFLA_GRE_ENCAP_LIMIT */
1615 nla_total_size(1) +
1616 /* IFLA_GRE_FLOWINFO */
1617 nla_total_size(4) +
1618 /* IFLA_GRE_FLAGS */
1619 nla_total_size(4) +
1620 0;
1621 }
1622
ip6gre_fill_info(struct sk_buff * skb,const struct net_device * dev)1623 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1624 {
1625 struct ip6_tnl *t = netdev_priv(dev);
1626 struct __ip6_tnl_parm *p = &t->parms;
1627
1628 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1629 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1630 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1631 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1632 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1633 nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1634 nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1635 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1636 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1637 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1638 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1639 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1640 goto nla_put_failure;
1641 return 0;
1642
1643 nla_put_failure:
1644 return -EMSGSIZE;
1645 }
1646
1647 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1648 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1649 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1650 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1651 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1652 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1653 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1654 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1655 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1656 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1657 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1658 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1659 };
1660
1661 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1662 .kind = "ip6gre",
1663 .maxtype = IFLA_GRE_MAX,
1664 .policy = ip6gre_policy,
1665 .priv_size = sizeof(struct ip6_tnl),
1666 .setup = ip6gre_tunnel_setup,
1667 .validate = ip6gre_tunnel_validate,
1668 .newlink = ip6gre_newlink,
1669 .changelink = ip6gre_changelink,
1670 .dellink = ip6gre_dellink,
1671 .get_size = ip6gre_get_size,
1672 .fill_info = ip6gre_fill_info,
1673 };
1674
1675 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1676 .kind = "ip6gretap",
1677 .maxtype = IFLA_GRE_MAX,
1678 .policy = ip6gre_policy,
1679 .priv_size = sizeof(struct ip6_tnl),
1680 .setup = ip6gre_tap_setup,
1681 .validate = ip6gre_tap_validate,
1682 .newlink = ip6gre_newlink,
1683 .changelink = ip6gre_changelink,
1684 .get_size = ip6gre_get_size,
1685 .fill_info = ip6gre_fill_info,
1686 };
1687
1688 /*
1689 * And now the modules code and kernel interface.
1690 */
1691
ip6gre_init(void)1692 static int __init ip6gre_init(void)
1693 {
1694 int err;
1695
1696 pr_info("GRE over IPv6 tunneling driver\n");
1697
1698 err = register_pernet_device(&ip6gre_net_ops);
1699 if (err < 0)
1700 return err;
1701
1702 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1703 if (err < 0) {
1704 pr_info("%s: can't add protocol\n", __func__);
1705 goto add_proto_failed;
1706 }
1707
1708 err = rtnl_link_register(&ip6gre_link_ops);
1709 if (err < 0)
1710 goto rtnl_link_failed;
1711
1712 err = rtnl_link_register(&ip6gre_tap_ops);
1713 if (err < 0)
1714 goto tap_ops_failed;
1715
1716 out:
1717 return err;
1718
1719 tap_ops_failed:
1720 rtnl_link_unregister(&ip6gre_link_ops);
1721 rtnl_link_failed:
1722 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1723 add_proto_failed:
1724 unregister_pernet_device(&ip6gre_net_ops);
1725 goto out;
1726 }
1727
ip6gre_fini(void)1728 static void __exit ip6gre_fini(void)
1729 {
1730 rtnl_link_unregister(&ip6gre_tap_ops);
1731 rtnl_link_unregister(&ip6gre_link_ops);
1732 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1733 unregister_pernet_device(&ip6gre_net_ops);
1734 }
1735
1736 module_init(ip6gre_init);
1737 module_exit(ip6gre_fini);
1738 MODULE_LICENSE("GPL");
1739 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1740 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1741 MODULE_ALIAS_RTNL_LINK("ip6gre");
1742 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1743 MODULE_ALIAS_NETDEV("ip6gre0");
1744