Merge tag 'v3.10.106' into update
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / ip6_gre.c
CommitLineData
c12b395a 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>
c5441932 41#include <net/ip_tunnels.h>
c12b395a 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>
a05c8e39 58#include <net/gre.h>
c12b395a 59
60
eccc1bb8 61static bool log_ecn_error = true;
62module_param(log_ecn_error, bool, 0644);
63MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
64
c12b395a 65#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
66#define IPV6_TCLASS_SHIFT 20
67
68#define HASH_SIZE_SHIFT 5
69#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
70
71static int ip6gre_net_id __read_mostly;
72struct ip6gre_net {
73 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
74
75 struct net_device *fb_tunnel_dev;
76};
77
78static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79static int ip6gre_tunnel_init(struct net_device *dev);
80static void ip6gre_tunnel_setup(struct net_device *dev);
81static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
82static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
83
84/* Tunnel hash table */
85
86/*
87 4 hash tables:
88
89 3: (remote,local)
90 2: (remote,*)
91 1: (*,local)
92 0: (*,*)
93
94 We require exact key match i.e. if a key is present in packet
95 it will match only tunnel with the same key; if it is not present,
96 it will match only keyless tunnel.
97
98 All keysless packets, if not matched configured keyless tunnels
99 will match fallback tunnel.
100 */
101
102#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
103static u32 HASH_ADDR(const struct in6_addr *addr)
104{
105 u32 hash = ipv6_addr_hash(addr);
106
107 return hash_32(hash, HASH_SIZE_SHIFT);
108}
109
110#define tunnels_r_l tunnels[3]
111#define tunnels_r tunnels[2]
112#define tunnels_l tunnels[1]
113#define tunnels_wc tunnels[0]
c12b395a 114
c12b395a 115/* Given src, dst and key, find appropriate for input tunnel. */
116
117static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
118 const struct in6_addr *remote, const struct in6_addr *local,
119 __be32 key, __be16 gre_proto)
120{
121 struct net *net = dev_net(dev);
122 int link = dev->ifindex;
123 unsigned int h0 = HASH_ADDR(remote);
124 unsigned int h1 = HASH_KEY(key);
125 struct ip6_tnl *t, *cand = NULL;
126 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
127 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
128 ARPHRD_ETHER : ARPHRD_IP6GRE;
129 int score, cand_score = 4;
130
e086cadc 131 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
c12b395a 132 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
133 !ipv6_addr_equal(remote, &t->parms.raddr) ||
134 key != t->parms.i_key ||
135 !(t->dev->flags & IFF_UP))
136 continue;
137
138 if (t->dev->type != ARPHRD_IP6GRE &&
139 t->dev->type != dev_type)
140 continue;
141
142 score = 0;
143 if (t->parms.link != link)
144 score |= 1;
145 if (t->dev->type != dev_type)
146 score |= 2;
147 if (score == 0)
148 return t;
149
150 if (score < cand_score) {
151 cand = t;
152 cand_score = score;
153 }
154 }
155
e086cadc 156 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
c12b395a 157 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
158 key != t->parms.i_key ||
159 !(t->dev->flags & IFF_UP))
160 continue;
161
162 if (t->dev->type != ARPHRD_IP6GRE &&
163 t->dev->type != dev_type)
164 continue;
165
166 score = 0;
167 if (t->parms.link != link)
168 score |= 1;
169 if (t->dev->type != dev_type)
170 score |= 2;
171 if (score == 0)
172 return t;
173
174 if (score < cand_score) {
175 cand = t;
176 cand_score = score;
177 }
178 }
179
e086cadc 180 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
c12b395a 181 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
182 (!ipv6_addr_equal(local, &t->parms.raddr) ||
183 !ipv6_addr_is_multicast(local))) ||
184 key != t->parms.i_key ||
185 !(t->dev->flags & IFF_UP))
186 continue;
187
188 if (t->dev->type != ARPHRD_IP6GRE &&
189 t->dev->type != dev_type)
190 continue;
191
192 score = 0;
193 if (t->parms.link != link)
194 score |= 1;
195 if (t->dev->type != dev_type)
196 score |= 2;
197 if (score == 0)
198 return t;
199
200 if (score < cand_score) {
201 cand = t;
202 cand_score = score;
203 }
204 }
205
e086cadc 206 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
c12b395a 207 if (t->parms.i_key != key ||
208 !(t->dev->flags & IFF_UP))
209 continue;
210
211 if (t->dev->type != ARPHRD_IP6GRE &&
212 t->dev->type != dev_type)
213 continue;
214
215 score = 0;
216 if (t->parms.link != link)
217 score |= 1;
218 if (t->dev->type != dev_type)
219 score |= 2;
220 if (score == 0)
221 return t;
222
223 if (score < cand_score) {
224 cand = t;
225 cand_score = score;
226 }
227 }
228
229 if (cand != NULL)
230 return cand;
231
232 dev = ign->fb_tunnel_dev;
233 if (dev->flags & IFF_UP)
234 return netdev_priv(dev);
235
236 return NULL;
237}
238
239static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
240 const struct __ip6_tnl_parm *p)
241{
242 const struct in6_addr *remote = &p->raddr;
243 const struct in6_addr *local = &p->laddr;
244 unsigned int h = HASH_KEY(p->i_key);
245 int prio = 0;
246
247 if (!ipv6_addr_any(local))
248 prio |= 1;
249 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
250 prio |= 2;
251 h ^= HASH_ADDR(remote);
252 }
253
254 return &ign->tunnels[prio][h];
255}
256
257static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
258 const struct ip6_tnl *t)
259{
260 return __ip6gre_bucket(ign, &t->parms);
261}
262
263static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
264{
265 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
266
267 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
268 rcu_assign_pointer(*tp, t);
269}
270
271static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
272{
273 struct ip6_tnl __rcu **tp;
274 struct ip6_tnl *iter;
275
276 for (tp = ip6gre_bucket(ign, t);
277 (iter = rtnl_dereference(*tp)) != NULL;
278 tp = &iter->next) {
279 if (t == iter) {
280 rcu_assign_pointer(*tp, t->next);
281 break;
282 }
283 }
284}
285
286static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
287 const struct __ip6_tnl_parm *parms,
288 int type)
289{
290 const struct in6_addr *remote = &parms->raddr;
291 const struct in6_addr *local = &parms->laddr;
292 __be32 key = parms->i_key;
293 int link = parms->link;
294 struct ip6_tnl *t;
295 struct ip6_tnl __rcu **tp;
296 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
297
298 for (tp = __ip6gre_bucket(ign, parms);
299 (t = rtnl_dereference(*tp)) != NULL;
300 tp = &t->next)
301 if (ipv6_addr_equal(local, &t->parms.laddr) &&
302 ipv6_addr_equal(remote, &t->parms.raddr) &&
303 key == t->parms.i_key &&
304 link == t->parms.link &&
305 type == t->dev->type)
306 break;
307
308 return t;
309}
310
311static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
312 const struct __ip6_tnl_parm *parms, int create)
313{
314 struct ip6_tnl *t, *nt;
315 struct net_device *dev;
316 char name[IFNAMSIZ];
317 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
318
319 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
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, ip6gre_tunnel_setup);
329 if (!dev)
330 return NULL;
331
332 dev_net_set(dev, net);
333
334 nt = netdev_priv(dev);
335 nt->parms = *parms;
336 dev->rtnl_link_ops = &ip6gre_link_ops;
337
338 nt->dev = dev;
339 ip6gre_tnl_link_config(nt, 1);
340
341 if (register_netdevice(dev) < 0)
342 goto failed_free;
343
344 /* Can use a lockless transmit, unless we generate output sequences */
345 if (!(nt->parms.o_flags & GRE_SEQ))
346 dev->features |= NETIF_F_LLTX;
347
348 dev_hold(dev);
349 ip6gre_tunnel_link(ign, nt);
350 return nt;
351
352failed_free:
353 free_netdev(dev);
354 return NULL;
355}
356
357static void ip6gre_tunnel_uninit(struct net_device *dev)
358{
359 struct net *net = dev_net(dev);
360 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
361
362 ip6gre_tunnel_unlink(ign, netdev_priv(dev));
6d8c1905 363 ip6_tnl_dst_reset(netdev_priv(dev));
c12b395a 364 dev_put(dev);
365}
366
367
368static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
a05c8e39 369 u8 type, u8 code, int offset, __be32 info)
c12b395a 370{
a05c8e39
ED
371 const struct gre_base_hdr *greh;
372 const struct ipv6hdr *ipv6h;
373 int grehlen = sizeof(*greh);
c12b395a 374 struct ip6_tnl *t;
a05c8e39 375 int key_off = 0;
c12b395a 376 __be16 flags;
a05c8e39 377 __be32 key;
c12b395a 378
a05c8e39
ED
379 if (!pskb_may_pull(skb, offset + grehlen))
380 return;
381 greh = (const struct gre_base_hdr *)(skb->data + offset);
382 flags = greh->flags;
383 if (flags & (GRE_VERSION | GRE_ROUTING))
384 return;
385 if (flags & GRE_CSUM)
386 grehlen += 4;
387 if (flags & GRE_KEY) {
388 key_off = grehlen + offset;
389 grehlen += 4;
c12b395a 390 }
391
a05c8e39 392 if (!pskb_may_pull(skb, offset + grehlen))
c12b395a 393 return;
b87fb39e 394 ipv6h = (const struct ipv6hdr *)skb->data;
a05c8e39
ED
395 greh = (const struct gre_base_hdr *)(skb->data + offset);
396 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
c12b395a 397
c12b395a 398 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
a05c8e39 399 key, greh->protocol);
c12b395a 400 if (t == NULL)
0c5794a6 401 return;
c12b395a 402
403 switch (type) {
404 __u32 teli;
405 struct ipv6_tlv_tnl_enc_lim *tel;
406 __u32 mtu;
407 case ICMPV6_DEST_UNREACH:
408 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
409 t->parms.name);
410 break;
411 case ICMPV6_TIME_EXCEED:
412 if (code == ICMPV6_EXC_HOPLIMIT) {
413 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
414 t->parms.name);
415 }
416 break;
417 case ICMPV6_PARAMPROB:
418 teli = 0;
419 if (code == ICMPV6_HDR_FIELD)
420 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
421
422 if (teli && teli == info - 2) {
423 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
424 if (tel->encap_limit == 0) {
425 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
426 t->parms.name);
427 }
428 } else {
429 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
430 t->parms.name);
431 }
432 break;
433 case ICMPV6_PKT_TOOBIG:
434 mtu = info - offset;
435 if (mtu < IPV6_MIN_MTU)
436 mtu = IPV6_MIN_MTU;
437 t->dev->mtu = mtu;
438 break;
439 }
440
441 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
442 t->err_count++;
443 else
444 t->err_count = 1;
445 t->err_time = jiffies;
c12b395a 446}
447
c12b395a 448static int ip6gre_rcv(struct sk_buff *skb)
449{
450 const struct ipv6hdr *ipv6h;
451 u8 *h;
452 __be16 flags;
453 __sum16 csum = 0;
454 __be32 key = 0;
455 u32 seqno = 0;
456 struct ip6_tnl *tunnel;
457 int offset = 4;
458 __be16 gre_proto;
eccc1bb8 459 int err;
c12b395a 460
461 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
0c5794a6 462 goto drop;
c12b395a 463
464 ipv6h = ipv6_hdr(skb);
465 h = skb->data;
466 flags = *(__be16 *)h;
467
468 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
469 /* - Version must be 0.
470 - We do not support routing headers.
471 */
472 if (flags&(GRE_VERSION|GRE_ROUTING))
0c5794a6 473 goto drop;
c12b395a 474
475 if (flags&GRE_CSUM) {
476 switch (skb->ip_summed) {
477 case CHECKSUM_COMPLETE:
478 csum = csum_fold(skb->csum);
479 if (!csum)
480 break;
481 /* fall through */
482 case CHECKSUM_NONE:
483 skb->csum = 0;
484 csum = __skb_checksum_complete(skb);
485 skb->ip_summed = CHECKSUM_COMPLETE;
486 }
487 offset += 4;
488 }
489 if (flags&GRE_KEY) {
490 key = *(__be32 *)(h + offset);
491 offset += 4;
492 }
493 if (flags&GRE_SEQ) {
494 seqno = ntohl(*(__be32 *)(h + offset));
495 offset += 4;
496 }
497 }
498
499 gre_proto = *(__be16 *)(h + 2);
500
c12b395a 501 tunnel = ip6gre_tunnel_lookup(skb->dev,
502 &ipv6h->saddr, &ipv6h->daddr, key,
503 gre_proto);
504 if (tunnel) {
505 struct pcpu_tstats *tstats;
506
507 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
508 goto drop;
509
510 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
511 tunnel->dev->stats.rx_dropped++;
512 goto drop;
513 }
514
515 secpath_reset(skb);
516
517 skb->protocol = gre_proto;
518 /* WCCP version 1 and 2 protocol decoding.
c556ebe8 519 * - Change protocol to IPv6
c12b395a 520 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
521 */
522 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
c556ebe8 523 skb->protocol = htons(ETH_P_IPV6);
c12b395a 524 if ((*(h + offset) & 0xF0) != 0x40)
525 offset += 4;
526 }
527
528 skb->mac_header = skb->network_header;
529 __pskb_pull(skb, offset);
530 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
531 skb->pkt_type = PACKET_HOST;
532
533 if (((flags&GRE_CSUM) && csum) ||
534 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
535 tunnel->dev->stats.rx_crc_errors++;
536 tunnel->dev->stats.rx_errors++;
537 goto drop;
538 }
539 if (tunnel->parms.i_flags&GRE_SEQ) {
540 if (!(flags&GRE_SEQ) ||
541 (tunnel->i_seqno &&
542 (s32)(seqno - tunnel->i_seqno) < 0)) {
543 tunnel->dev->stats.rx_fifo_errors++;
544 tunnel->dev->stats.rx_errors++;
545 goto drop;
546 }
547 tunnel->i_seqno = seqno + 1;
548 }
549
550 /* Warning: All skb pointers will be invalidated! */
551 if (tunnel->dev->type == ARPHRD_ETHER) {
552 if (!pskb_may_pull(skb, ETH_HLEN)) {
553 tunnel->dev->stats.rx_length_errors++;
554 tunnel->dev->stats.rx_errors++;
555 goto drop;
556 }
557
558 ipv6h = ipv6_hdr(skb);
559 skb->protocol = eth_type_trans(skb, tunnel->dev);
560 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
561 }
562
eccc1bb8 563 __skb_tunnel_rx(skb, tunnel->dev);
564
565 skb_reset_network_header(skb);
566
567 err = IP6_ECN_decapsulate(ipv6h, skb);
568 if (unlikely(err)) {
569 if (log_ecn_error)
570 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
571 &ipv6h->saddr,
572 ipv6_get_dsfield(ipv6h));
573 if (err > 1) {
574 ++tunnel->dev->stats.rx_frame_errors;
575 ++tunnel->dev->stats.rx_errors;
576 goto drop;
577 }
578 }
579
c12b395a 580 tstats = this_cpu_ptr(tunnel->dev->tstats);
581 u64_stats_update_begin(&tstats->syncp);
582 tstats->rx_packets++;
583 tstats->rx_bytes += skb->len;
584 u64_stats_update_end(&tstats->syncp);
585
c12b395a 586 netif_rx(skb);
587
c12b395a 588 return 0;
589 }
590 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
591
592drop:
c12b395a 593 kfree_skb(skb);
594 return 0;
595}
596
597struct ipv6_tel_txoption {
598 struct ipv6_txoptions ops;
599 __u8 dst_opt[8];
600};
601
602static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
603{
604 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
605
606 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
607 opt->dst_opt[3] = 1;
608 opt->dst_opt[4] = encap_limit;
609 opt->dst_opt[5] = IPV6_TLV_PADN;
610 opt->dst_opt[6] = 1;
611
612 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
613 opt->ops.opt_nflen = 8;
614}
615
616static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
617 struct net_device *dev,
618 __u8 dsfield,
619 struct flowi6 *fl6,
620 int encap_limit,
621 __u32 *pmtu)
622{
623 struct net *net = dev_net(dev);
624 struct ip6_tnl *tunnel = netdev_priv(dev);
625 struct net_device *tdev; /* Device to other host */
626 struct ipv6hdr *ipv6h; /* Our new IP header */
102ce961 627 unsigned int max_headroom = 0; /* The extra header space needed */
c12b395a 628 int gre_hlen;
629 struct ipv6_tel_txoption opt;
630 int mtu;
631 struct dst_entry *dst = NULL, *ndst = NULL;
632 struct net_device_stats *stats = &tunnel->dev->stats;
633 int err = -1;
634 u8 proto;
c12b395a 635 struct sk_buff *new_skb;
636
637 if (dev->type == ARPHRD_ETHER)
638 IPCB(skb)->flags = 0;
639
640 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
641 gre_hlen = 0;
642 ipv6h = (struct ipv6hdr *)skb->data;
643 fl6->daddr = ipv6h->daddr;
644 } else {
645 gre_hlen = tunnel->hlen;
646 fl6->daddr = tunnel->parms.raddr;
647 }
648
649 if (!fl6->flowi6_mark)
650 dst = ip6_tnl_dst_check(tunnel);
651
652 if (!dst) {
653 ndst = ip6_route_output(net, NULL, fl6);
654
655 if (ndst->error)
656 goto tx_err_link_failure;
657 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
658 if (IS_ERR(ndst)) {
659 err = PTR_ERR(ndst);
660 ndst = NULL;
661 goto tx_err_link_failure;
662 }
663 dst = ndst;
664 }
665
666 tdev = dst->dev;
667
668 if (tdev == dev) {
669 stats->collisions++;
670 net_warn_ratelimited("%s: Local routing loop detected!\n",
671 tunnel->parms.name);
672 goto tx_err_dst_release;
673 }
674
675 mtu = dst_mtu(dst) - sizeof(*ipv6h);
676 if (encap_limit >= 0) {
677 max_headroom += 8;
678 mtu -= 8;
679 }
680 if (mtu < IPV6_MIN_MTU)
681 mtu = IPV6_MIN_MTU;
682 if (skb_dst(skb))
683 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
684 if (skb->len > mtu) {
685 *pmtu = mtu;
686 err = -EMSGSIZE;
687 goto tx_err_dst_release;
688 }
689
690 if (tunnel->err_count > 0) {
691 if (time_before(jiffies,
692 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
693 tunnel->err_count--;
694
695 dst_link_failure(skb);
696 } else
697 tunnel->err_count = 0;
698 }
699
102ce961 700 max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
c12b395a 701
702 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
703 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
704 new_skb = skb_realloc_headroom(skb, max_headroom);
705 if (max_headroom > dev->needed_headroom)
706 dev->needed_headroom = max_headroom;
707 if (!new_skb)
708 goto tx_err_dst_release;
709
710 if (skb->sk)
711 skb_set_owner_w(new_skb, skb->sk);
712 consume_skb(skb);
713 skb = new_skb;
714 }
715
716 skb_dst_drop(skb);
717
718 if (fl6->flowi6_mark) {
719 skb_dst_set(skb, dst);
720 ndst = NULL;
721 } else {
722 skb_dst_set_noref(skb, dst);
723 }
724
c12b395a 725 proto = NEXTHDR_GRE;
726 if (encap_limit >= 0) {
727 init_tel_txopt(&opt, encap_limit);
728 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
729 }
730
731 skb_push(skb, gre_hlen);
732 skb_reset_network_header(skb);
ae782bb1 733 skb_set_transport_header(skb, sizeof(*ipv6h));
c12b395a 734
735 /*
736 * Push down and install the IP header.
737 */
738 ipv6h = ipv6_hdr(skb);
3e4e4c1f 739 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
c12b395a 740 ipv6h->hop_limit = tunnel->parms.hop_limit;
741 ipv6h->nexthdr = proto;
742 ipv6h->saddr = fl6->saddr;
743 ipv6h->daddr = fl6->daddr;
744
745 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
746 ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
747 htons(ETH_P_TEB) : skb->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
e8f72ea4 768 ip6tunnel_xmit(skb, dev);
c12b395a 769 if (ndst)
770 ip6_tnl_dst_store(tunnel, ndst);
c12b395a 771 return 0;
772tx_err_link_failure:
773 stats->tx_carrier_errors++;
774 dst_link_failure(skb);
775tx_err_dst_release:
776 dst_release(ndst);
777 return err;
778}
779
780static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
781{
782 struct ip6_tnl *t = netdev_priv(dev);
783 const struct iphdr *iph = ip_hdr(skb);
784 int encap_limit = -1;
785 struct flowi6 fl6;
786 __u8 dsfield;
787 __u32 mtu;
788 int err;
789
790 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
791 encap_limit = t->parms.encap_limit;
792
793 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
ae3b8ed5 794 fl6.flowi6_proto = IPPROTO_GRE;
c12b395a 795
796 dsfield = ipv4_get_dsfield(iph);
797
798 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
799 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
800 & IPV6_TCLASS_MASK;
801 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
802 fl6.flowi6_mark = skb->mark;
803
804 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
805 if (err != 0) {
806 /* XXX: send ICMP error even if DF is not set. */
807 if (err == -EMSGSIZE)
808 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
809 htonl(mtu));
810 return -1;
811 }
812
813 return 0;
814}
815
816static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
817{
818 struct ip6_tnl *t = netdev_priv(dev);
819 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
820 int encap_limit = -1;
821 __u16 offset;
822 struct flowi6 fl6;
823 __u8 dsfield;
824 __u32 mtu;
825 int err;
826
827 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
828 return -1;
829
830 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
831 if (offset > 0) {
832 struct ipv6_tlv_tnl_enc_lim *tel;
833 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
834 if (tel->encap_limit == 0) {
835 icmpv6_send(skb, ICMPV6_PARAMPROB,
836 ICMPV6_HDR_FIELD, offset + 2);
837 return -1;
838 }
839 encap_limit = tel->encap_limit - 1;
840 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
841 encap_limit = t->parms.encap_limit;
842
843 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
ae3b8ed5 844 fl6.flowi6_proto = IPPROTO_GRE;
c12b395a 845
846 dsfield = ipv6_get_dsfield(ipv6h);
847 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
848 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
849 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
850 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
851 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
852 fl6.flowi6_mark = skb->mark;
853
854 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
855 if (err != 0) {
856 if (err == -EMSGSIZE)
857 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
858 return -1;
859 }
860
861 return 0;
862}
863
864/**
865 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
866 * @t: the outgoing tunnel device
867 * @hdr: IPv6 header from the incoming packet
868 *
869 * Description:
870 * Avoid trivial tunneling loop by checking that tunnel exit-point
871 * doesn't match source of incoming packet.
872 *
873 * Return:
874 * 1 if conflict,
875 * 0 else
876 **/
877
878static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
879 const struct ipv6hdr *hdr)
880{
881 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
882}
883
884static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
885{
886 struct ip6_tnl *t = netdev_priv(dev);
887 int encap_limit = -1;
888 struct flowi6 fl6;
889 __u32 mtu;
890 int err;
891
892 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
893 encap_limit = t->parms.encap_limit;
894
895 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
c12b395a 896
897 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
898
899 return err;
900}
901
902static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
903 struct net_device *dev)
904{
905 struct ip6_tnl *t = netdev_priv(dev);
906 struct net_device_stats *stats = &t->dev->stats;
907 int ret;
908
909 if (!ip6_tnl_xmit_ctl(t))
41ab3e31 910 goto tx_err;
c12b395a 911
912 switch (skb->protocol) {
913 case htons(ETH_P_IP):
914 ret = ip6gre_xmit_ipv4(skb, dev);
915 break;
916 case htons(ETH_P_IPV6):
917 ret = ip6gre_xmit_ipv6(skb, dev);
918 break;
919 default:
920 ret = ip6gre_xmit_other(skb, dev);
921 break;
922 }
923
924 if (ret < 0)
925 goto tx_err;
926
927 return NETDEV_TX_OK;
928
929tx_err:
930 stats->tx_errors++;
931 stats->tx_dropped++;
932 kfree_skb(skb);
933 return NETDEV_TX_OK;
934}
935
936static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
937{
938 struct net_device *dev = t->dev;
939 struct __ip6_tnl_parm *p = &t->parms;
940 struct flowi6 *fl6 = &t->fl.u.ip6;
941 int addend = sizeof(struct ipv6hdr) + 4;
942
943 if (dev->type != ARPHRD_ETHER) {
944 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
945 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
946 }
947
948 /* Set up flowi template */
949 fl6->saddr = p->laddr;
950 fl6->daddr = p->raddr;
951 fl6->flowi6_oif = p->link;
952 fl6->flowlabel = 0;
953
954 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
955 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
956 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
957 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
958
959 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
960 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
961
962 if (p->flags&IP6_TNL_F_CAP_XMIT &&
963 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
964 dev->flags |= IFF_POINTOPOINT;
965 else
966 dev->flags &= ~IFF_POINTOPOINT;
967
c12b395a 968 /* Precalculate GRE options length */
969 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
970 if (t->parms.o_flags&GRE_CSUM)
971 addend += 4;
972 if (t->parms.o_flags&GRE_KEY)
973 addend += 4;
974 if (t->parms.o_flags&GRE_SEQ)
975 addend += 4;
976 }
977
978 if (p->flags & IP6_TNL_F_CAP_XMIT) {
979 int strict = (ipv6_addr_type(&p->raddr) &
980 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
981
982 struct rt6_info *rt = rt6_lookup(dev_net(dev),
983 &p->raddr, &p->laddr,
984 p->link, strict);
985
986 if (rt == NULL)
987 return;
988
989 if (rt->dst.dev) {
990 dev->hard_header_len = rt->dst.dev->hard_header_len + addend;
991
992 if (set_mtu) {
993 dev->mtu = rt->dst.dev->mtu - addend;
994 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
995 dev->mtu -= 8;
996
997 if (dev->mtu < IPV6_MIN_MTU)
998 dev->mtu = IPV6_MIN_MTU;
999 }
1000 }
94e187c0 1001 ip6_rt_put(rt);
c12b395a 1002 }
1003
1004 t->hlen = addend;
1005}
1006
1007static int ip6gre_tnl_change(struct ip6_tnl *t,
1008 const struct __ip6_tnl_parm *p, int set_mtu)
1009{
1010 t->parms.laddr = p->laddr;
1011 t->parms.raddr = p->raddr;
1012 t->parms.flags = p->flags;
1013 t->parms.hop_limit = p->hop_limit;
1014 t->parms.encap_limit = p->encap_limit;
1015 t->parms.flowinfo = p->flowinfo;
1016 t->parms.link = p->link;
1017 t->parms.proto = p->proto;
1018 t->parms.i_key = p->i_key;
1019 t->parms.o_key = p->o_key;
1020 t->parms.i_flags = p->i_flags;
1021 t->parms.o_flags = p->o_flags;
1022 ip6_tnl_dst_reset(t);
1023 ip6gre_tnl_link_config(t, set_mtu);
1024 return 0;
1025}
1026
1027static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1028 const struct ip6_tnl_parm2 *u)
1029{
1030 p->laddr = u->laddr;
1031 p->raddr = u->raddr;
1032 p->flags = u->flags;
1033 p->hop_limit = u->hop_limit;
1034 p->encap_limit = u->encap_limit;
1035 p->flowinfo = u->flowinfo;
1036 p->link = u->link;
1037 p->i_key = u->i_key;
1038 p->o_key = u->o_key;
1039 p->i_flags = u->i_flags;
1040 p->o_flags = u->o_flags;
1041 memcpy(p->name, u->name, sizeof(u->name));
1042}
1043
1044static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1045 const struct __ip6_tnl_parm *p)
1046{
1047 u->proto = IPPROTO_GRE;
1048 u->laddr = p->laddr;
1049 u->raddr = p->raddr;
1050 u->flags = p->flags;
1051 u->hop_limit = p->hop_limit;
1052 u->encap_limit = p->encap_limit;
1053 u->flowinfo = p->flowinfo;
1054 u->link = p->link;
1055 u->i_key = p->i_key;
1056 u->o_key = p->o_key;
1057 u->i_flags = p->i_flags;
1058 u->o_flags = p->o_flags;
1059 memcpy(u->name, p->name, sizeof(u->name));
1060}
1061
1062static int ip6gre_tunnel_ioctl(struct net_device *dev,
1063 struct ifreq *ifr, int cmd)
1064{
1065 int err = 0;
1066 struct ip6_tnl_parm2 p;
1067 struct __ip6_tnl_parm p1;
1068 struct ip6_tnl *t;
1069 struct net *net = dev_net(dev);
1070 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1071
1072 switch (cmd) {
1073 case SIOCGETTUNNEL:
1074 t = NULL;
1075 if (dev == ign->fb_tunnel_dev) {
1076 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1077 err = -EFAULT;
1078 break;
1079 }
1080 ip6gre_tnl_parm_from_user(&p1, &p);
1081 t = ip6gre_tunnel_locate(net, &p1, 0);
1082 }
1083 if (t == NULL)
1084 t = netdev_priv(dev);
5dbd5068 1085 memset(&p, 0, sizeof(p));
c12b395a 1086 ip6gre_tnl_parm_to_user(&p, &t->parms);
1087 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1088 err = -EFAULT;
1089 break;
1090
1091 case SIOCADDTUNNEL:
1092 case SIOCCHGTUNNEL:
1093 err = -EPERM;
af31f412 1094 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
c12b395a 1095 goto done;
1096
1097 err = -EFAULT;
1098 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1099 goto done;
1100
1101 err = -EINVAL;
1102 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1103 goto done;
1104
1105 if (!(p.i_flags&GRE_KEY))
1106 p.i_key = 0;
1107 if (!(p.o_flags&GRE_KEY))
1108 p.o_key = 0;
1109
1110 ip6gre_tnl_parm_from_user(&p1, &p);
1111 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1112
1113 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1114 if (t != NULL) {
1115 if (t->dev != dev) {
1116 err = -EEXIST;
1117 break;
1118 }
1119 } else {
1120 t = netdev_priv(dev);
1121
1122 ip6gre_tunnel_unlink(ign, t);
1123 synchronize_net();
1124 ip6gre_tnl_change(t, &p1, 1);
1125 ip6gre_tunnel_link(ign, t);
1126 netdev_state_change(dev);
1127 }
1128 }
1129
1130 if (t) {
1131 err = 0;
1132
5dbd5068 1133 memset(&p, 0, sizeof(p));
c12b395a 1134 ip6gre_tnl_parm_to_user(&p, &t->parms);
1135 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1136 err = -EFAULT;
1137 } else
1138 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1139 break;
1140
1141 case SIOCDELTUNNEL:
1142 err = -EPERM;
af31f412 1143 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
c12b395a 1144 goto done;
1145
1146 if (dev == ign->fb_tunnel_dev) {
1147 err = -EFAULT;
1148 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1149 goto done;
1150 err = -ENOENT;
1151 ip6gre_tnl_parm_from_user(&p1, &p);
1152 t = ip6gre_tunnel_locate(net, &p1, 0);
1153 if (t == NULL)
1154 goto done;
1155 err = -EPERM;
1156 if (t == netdev_priv(ign->fb_tunnel_dev))
1157 goto done;
1158 dev = t->dev;
1159 }
1160 unregister_netdevice(dev);
1161 err = 0;
1162 break;
1163
1164 default:
1165 err = -EINVAL;
1166 }
1167
1168done:
1169 return err;
1170}
1171
1172static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1173{
1174 struct ip6_tnl *tunnel = netdev_priv(dev);
1175 if (new_mtu < 68 ||
1176 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
1177 return -EINVAL;
1178 dev->mtu = new_mtu;
1179 return 0;
1180}
1181
1182static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1183 unsigned short type,
1184 const void *daddr, const void *saddr, unsigned int len)
1185{
1186 struct ip6_tnl *t = netdev_priv(dev);
1187 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1188 __be16 *p = (__be16 *)(ipv6h+1);
1189
3e4e4c1f 1190 ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
c12b395a 1191 ipv6h->hop_limit = t->parms.hop_limit;
1192 ipv6h->nexthdr = NEXTHDR_GRE;
1193 ipv6h->saddr = t->parms.laddr;
1194 ipv6h->daddr = t->parms.raddr;
1195
1196 p[0] = t->parms.o_flags;
1197 p[1] = htons(type);
1198
1199 /*
1200 * Set the source hardware address.
1201 */
1202
1203 if (saddr)
1204 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1205 if (daddr)
1206 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1207 if (!ipv6_addr_any(&ipv6h->daddr))
1208 return t->hlen;
1209
1210 return -t->hlen;
1211}
1212
c12b395a 1213static const struct header_ops ip6gre_header_ops = {
1214 .create = ip6gre_header,
c12b395a 1215};
1216
1217static const struct net_device_ops ip6gre_netdev_ops = {
1218 .ndo_init = ip6gre_tunnel_init,
1219 .ndo_uninit = ip6gre_tunnel_uninit,
1220 .ndo_start_xmit = ip6gre_tunnel_xmit,
1221 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1222 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
f61dd388 1223 .ndo_get_stats64 = ip_tunnel_get_stats64,
c12b395a 1224};
1225
1226static void ip6gre_dev_free(struct net_device *dev)
1227{
1228 free_percpu(dev->tstats);
1229 free_netdev(dev);
1230}
1231
1232static void ip6gre_tunnel_setup(struct net_device *dev)
1233{
1234 struct ip6_tnl *t;
1235
1236 dev->netdev_ops = &ip6gre_netdev_ops;
1237 dev->destructor = ip6gre_dev_free;
1238
1239 dev->type = ARPHRD_IP6GRE;
1240 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1241 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1242 t = netdev_priv(dev);
1243 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1244 dev->mtu -= 8;
1245 dev->flags |= IFF_NOARP;
1246 dev->iflink = 0;
1247 dev->addr_len = sizeof(struct in6_addr);
1248 dev->features |= NETIF_F_NETNS_LOCAL;
1249 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1250}
1251
1252static int ip6gre_tunnel_init(struct net_device *dev)
1253{
1254 struct ip6_tnl *tunnel;
1255
1256 tunnel = netdev_priv(dev);
1257
1258 tunnel->dev = dev;
1259 strcpy(tunnel->parms.name, dev->name);
1260
1261 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1262 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1263
1264 if (ipv6_addr_any(&tunnel->parms.raddr))
1265 dev->header_ops = &ip6gre_header_ops;
1266
1267 dev->tstats = alloc_percpu(struct pcpu_tstats);
1268 if (!dev->tstats)
1269 return -ENOMEM;
1270
460ceaa0
SK
1271 dev->iflink = tunnel->parms.link;
1272
c12b395a 1273 return 0;
1274}
1275
1276static void ip6gre_fb_tunnel_init(struct net_device *dev)
1277{
1278 struct ip6_tnl *tunnel = netdev_priv(dev);
1279
1280 tunnel->dev = dev;
1281 strcpy(tunnel->parms.name, dev->name);
1282
1283 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1284
1285 dev_hold(dev);
1286}
1287
c12b395a 1288static struct inet6_protocol ip6gre_protocol __read_mostly = {
1289 .handler = ip6gre_rcv,
1290 .err_handler = ip6gre_err,
1291 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1292};
1293
1294static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1295 struct list_head *head)
1296{
1297 int prio;
1298
1299 for (prio = 0; prio < 4; prio++) {
1300 int h;
1301 for (h = 0; h < HASH_SIZE; h++) {
1302 struct ip6_tnl *t;
1303
1304 t = rtnl_dereference(ign->tunnels[prio][h]);
1305
1306 while (t != NULL) {
1307 unregister_netdevice_queue(t->dev, head);
1308 t = rtnl_dereference(t->next);
1309 }
1310 }
1311 }
1312}
1313
1314static int __net_init ip6gre_init_net(struct net *net)
1315{
1316 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1317 int err;
1318
1319 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1320 ip6gre_tunnel_setup);
1321 if (!ign->fb_tunnel_dev) {
1322 err = -ENOMEM;
1323 goto err_alloc_dev;
1324 }
1325 dev_net_set(ign->fb_tunnel_dev, net);
1326
1327 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1328 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1329
1330 err = register_netdev(ign->fb_tunnel_dev);
1331 if (err)
1332 goto err_reg_dev;
1333
1334 rcu_assign_pointer(ign->tunnels_wc[0],
1335 netdev_priv(ign->fb_tunnel_dev));
1336 return 0;
1337
1338err_reg_dev:
1339 ip6gre_dev_free(ign->fb_tunnel_dev);
1340err_alloc_dev:
1341 return err;
1342}
1343
1344static void __net_exit ip6gre_exit_net(struct net *net)
1345{
1346 struct ip6gre_net *ign;
1347 LIST_HEAD(list);
1348
1349 ign = net_generic(net, ip6gre_net_id);
1350 rtnl_lock();
1351 ip6gre_destroy_tunnels(ign, &list);
1352 unregister_netdevice_many(&list);
1353 rtnl_unlock();
1354}
1355
1356static struct pernet_operations ip6gre_net_ops = {
1357 .init = ip6gre_init_net,
1358 .exit = ip6gre_exit_net,
1359 .id = &ip6gre_net_id,
1360 .size = sizeof(struct ip6gre_net),
1361};
1362
1363static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1364{
1365 __be16 flags;
1366
1367 if (!data)
1368 return 0;
1369
1370 flags = 0;
1371 if (data[IFLA_GRE_IFLAGS])
1372 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1373 if (data[IFLA_GRE_OFLAGS])
1374 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1375 if (flags & (GRE_VERSION|GRE_ROUTING))
1376 return -EINVAL;
1377
1378 return 0;
1379}
1380
1381static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1382{
1383 struct in6_addr daddr;
1384
1385 if (tb[IFLA_ADDRESS]) {
1386 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1387 return -EINVAL;
1388 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1389 return -EADDRNOTAVAIL;
1390 }
1391
1392 if (!data)
1393 goto out;
1394
1395 if (data[IFLA_GRE_REMOTE]) {
1396 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1397 if (ipv6_addr_any(&daddr))
1398 return -EINVAL;
1399 }
1400
1401out:
1402 return ip6gre_tunnel_validate(tb, data);
1403}
1404
1405
1406static void ip6gre_netlink_parms(struct nlattr *data[],
1407 struct __ip6_tnl_parm *parms)
1408{
1409 memset(parms, 0, sizeof(*parms));
1410
1411 if (!data)
1412 return;
1413
1414 if (data[IFLA_GRE_LINK])
1415 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1416
1417 if (data[IFLA_GRE_IFLAGS])
1418 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1419
1420 if (data[IFLA_GRE_OFLAGS])
1421 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1422
1423 if (data[IFLA_GRE_IKEY])
1424 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1425
1426 if (data[IFLA_GRE_OKEY])
1427 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1428
1429 if (data[IFLA_GRE_LOCAL])
1430 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1431
1432 if (data[IFLA_GRE_REMOTE])
1433 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1434
1435 if (data[IFLA_GRE_TTL])
1436 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1437
1438 if (data[IFLA_GRE_ENCAP_LIMIT])
1439 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1440
1441 if (data[IFLA_GRE_FLOWINFO])
1442 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1443
1444 if (data[IFLA_GRE_FLAGS])
1445 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1446}
1447
1448static int ip6gre_tap_init(struct net_device *dev)
1449{
1450 struct ip6_tnl *tunnel;
1451
1452 tunnel = netdev_priv(dev);
1453
1454 tunnel->dev = dev;
1455 strcpy(tunnel->parms.name, dev->name);
1456
1457 ip6gre_tnl_link_config(tunnel, 1);
1458
1459 dev->tstats = alloc_percpu(struct pcpu_tstats);
1460 if (!dev->tstats)
1461 return -ENOMEM;
1462
460ceaa0
SK
1463 dev->iflink = tunnel->parms.link;
1464
c12b395a 1465 return 0;
1466}
1467
1468static const struct net_device_ops ip6gre_tap_netdev_ops = {
1469 .ndo_init = ip6gre_tap_init,
1470 .ndo_uninit = ip6gre_tunnel_uninit,
1471 .ndo_start_xmit = ip6gre_tunnel_xmit,
1472 .ndo_set_mac_address = eth_mac_addr,
1473 .ndo_validate_addr = eth_validate_addr,
1474 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
f61dd388 1475 .ndo_get_stats64 = ip_tunnel_get_stats64,
c12b395a 1476};
1477
1478static void ip6gre_tap_setup(struct net_device *dev)
1479{
1480
1481 ether_setup(dev);
1482
1483 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1484 dev->destructor = ip6gre_dev_free;
1485
1486 dev->iflink = 0;
1487 dev->features |= NETIF_F_NETNS_LOCAL;
1488}
1489
1490static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1491 struct nlattr *tb[], struct nlattr *data[])
1492{
1493 struct ip6_tnl *nt;
1494 struct net *net = dev_net(dev);
1495 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1496 int err;
1497
1498 nt = netdev_priv(dev);
1499 ip6gre_netlink_parms(data, &nt->parms);
1500
1501 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1502 return -EEXIST;
1503
1504 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1505 eth_hw_addr_random(dev);
1506
1507 nt->dev = dev;
1508 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1509
1510 /* Can use a lockless transmit, unless we generate output sequences */
1511 if (!(nt->parms.o_flags & GRE_SEQ))
1512 dev->features |= NETIF_F_LLTX;
1513
1514 err = register_netdevice(dev);
1515 if (err)
1516 goto out;
1517
1518 dev_hold(dev);
1519 ip6gre_tunnel_link(ign, nt);
1520
1521out:
1522 return err;
1523}
1524
1525static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1526 struct nlattr *data[])
1527{
1528 struct ip6_tnl *t, *nt;
1529 struct net *net = dev_net(dev);
1530 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1531 struct __ip6_tnl_parm p;
1532
1533 if (dev == ign->fb_tunnel_dev)
1534 return -EINVAL;
1535
1536 nt = netdev_priv(dev);
1537 ip6gre_netlink_parms(data, &p);
1538
1539 t = ip6gre_tunnel_locate(net, &p, 0);
1540
1541 if (t) {
1542 if (t->dev != dev)
1543 return -EEXIST;
1544 } else {
1545 t = nt;
c12b395a 1546 }
1547
6089a803
ND
1548 ip6gre_tunnel_unlink(ign, t);
1549 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1550 ip6gre_tunnel_link(ign, t);
c12b395a 1551 return 0;
1552}
1553
278d20a5
ND
1554static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1555{
1556 struct net *net = dev_net(dev);
1557 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1558
1559 if (dev != ign->fb_tunnel_dev)
1560 unregister_netdevice_queue(dev, head);
1561}
1562
c12b395a 1563static size_t ip6gre_get_size(const struct net_device *dev)
1564{
1565 return
1566 /* IFLA_GRE_LINK */
1567 nla_total_size(4) +
1568 /* IFLA_GRE_IFLAGS */
1569 nla_total_size(2) +
1570 /* IFLA_GRE_OFLAGS */
1571 nla_total_size(2) +
1572 /* IFLA_GRE_IKEY */
1573 nla_total_size(4) +
1574 /* IFLA_GRE_OKEY */
1575 nla_total_size(4) +
1576 /* IFLA_GRE_LOCAL */
a3754133 1577 nla_total_size(sizeof(struct in6_addr)) +
c12b395a 1578 /* IFLA_GRE_REMOTE */
a3754133 1579 nla_total_size(sizeof(struct in6_addr)) +
c12b395a 1580 /* IFLA_GRE_TTL */
1581 nla_total_size(1) +
1582 /* IFLA_GRE_TOS */
1583 nla_total_size(1) +
1584 /* IFLA_GRE_ENCAP_LIMIT */
1585 nla_total_size(1) +
1586 /* IFLA_GRE_FLOWINFO */
1587 nla_total_size(4) +
1588 /* IFLA_GRE_FLAGS */
1589 nla_total_size(4) +
1590 0;
1591}
1592
1593static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1594{
1595 struct ip6_tnl *t = netdev_priv(dev);
1596 struct __ip6_tnl_parm *p = &t->parms;
1597
1598 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1599 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1600 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1601 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1602 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
a3754133
ND
1603 nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1604 nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
c12b395a 1605 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1606 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1607 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1608 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1609 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1610 goto nla_put_failure;
1611 return 0;
1612
1613nla_put_failure:
1614 return -EMSGSIZE;
1615}
1616
1617static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1618 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1619 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1620 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1621 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1622 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1623 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1624 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1625 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1626 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1627 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1628 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1629};
1630
1631static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1632 .kind = "ip6gre",
1633 .maxtype = IFLA_GRE_MAX,
1634 .policy = ip6gre_policy,
1635 .priv_size = sizeof(struct ip6_tnl),
1636 .setup = ip6gre_tunnel_setup,
1637 .validate = ip6gre_tunnel_validate,
1638 .newlink = ip6gre_newlink,
1639 .changelink = ip6gre_changelink,
278d20a5 1640 .dellink = ip6gre_dellink,
c12b395a 1641 .get_size = ip6gre_get_size,
1642 .fill_info = ip6gre_fill_info,
1643};
1644
1645static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1646 .kind = "ip6gretap",
1647 .maxtype = IFLA_GRE_MAX,
1648 .policy = ip6gre_policy,
1649 .priv_size = sizeof(struct ip6_tnl),
1650 .setup = ip6gre_tap_setup,
1651 .validate = ip6gre_tap_validate,
1652 .newlink = ip6gre_newlink,
1653 .changelink = ip6gre_changelink,
1654 .get_size = ip6gre_get_size,
1655 .fill_info = ip6gre_fill_info,
1656};
1657
1658/*
1659 * And now the modules code and kernel interface.
1660 */
1661
1662static int __init ip6gre_init(void)
1663{
1664 int err;
1665
1666 pr_info("GRE over IPv6 tunneling driver\n");
1667
1668 err = register_pernet_device(&ip6gre_net_ops);
1669 if (err < 0)
1670 return err;
1671
1672 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1673 if (err < 0) {
1674 pr_info("%s: can't add protocol\n", __func__);
1675 goto add_proto_failed;
1676 }
1677
1678 err = rtnl_link_register(&ip6gre_link_ops);
1679 if (err < 0)
1680 goto rtnl_link_failed;
1681
1682 err = rtnl_link_register(&ip6gre_tap_ops);
1683 if (err < 0)
1684 goto tap_ops_failed;
1685
1686out:
1687 return err;
1688
1689tap_ops_failed:
1690 rtnl_link_unregister(&ip6gre_link_ops);
1691rtnl_link_failed:
1692 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1693add_proto_failed:
1694 unregister_pernet_device(&ip6gre_net_ops);
1695 goto out;
1696}
1697
1698static void __exit ip6gre_fini(void)
1699{
1700 rtnl_link_unregister(&ip6gre_tap_ops);
1701 rtnl_link_unregister(&ip6gre_link_ops);
1702 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1703 unregister_pernet_device(&ip6gre_net_ops);
1704}
1705
1706module_init(ip6gre_init);
1707module_exit(ip6gre_fini);
1708MODULE_LICENSE("GPL");
1709MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1710MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1711MODULE_ALIAS_RTNL_LINK("ip6gre");
1712MODULE_ALIAS_NETDEV("ip6gre0");