Loading...
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
59
60static bool log_ecn_error = true;
61module_param(log_ecn_error, bool, 0644);
62MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63
64#define HASH_SIZE_SHIFT 5
65#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
66
67static int ip6gre_net_id __read_mostly;
68struct ip6gre_net {
69 struct ip6_tnl __rcu *tunnels[4][HASH_SIZE];
70
71 struct net_device *fb_tunnel_dev;
72};
73
74static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75static int ip6gre_tunnel_init(struct net_device *dev);
76static void ip6gre_tunnel_setup(struct net_device *dev);
77static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
78static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
79
80/* Tunnel hash table */
81
82/*
83 4 hash tables:
84
85 3: (remote,local)
86 2: (remote,*)
87 1: (*,local)
88 0: (*,*)
89
90 We require exact key match i.e. if a key is present in packet
91 it will match only tunnel with the same key; if it is not present,
92 it will match only keyless tunnel.
93
94 All keysless packets, if not matched configured keyless tunnels
95 will match fallback tunnel.
96 */
97
98#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(HASH_SIZE - 1))
99static u32 HASH_ADDR(const struct in6_addr *addr)
100{
101 u32 hash = ipv6_addr_hash(addr);
102
103 return hash_32(hash, HASH_SIZE_SHIFT);
104}
105
106#define tunnels_r_l tunnels[3]
107#define tunnels_r tunnels[2]
108#define tunnels_l tunnels[1]
109#define tunnels_wc tunnels[0]
110
111/* Given src, dst and key, find appropriate for input tunnel. */
112
113static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
114 const struct in6_addr *remote, const struct in6_addr *local,
115 __be32 key, __be16 gre_proto)
116{
117 struct net *net = dev_net(dev);
118 int link = dev->ifindex;
119 unsigned int h0 = HASH_ADDR(remote);
120 unsigned int h1 = HASH_KEY(key);
121 struct ip6_tnl *t, *cand = NULL;
122 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
123 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
124 ARPHRD_ETHER : ARPHRD_IP6GRE;
125 int score, cand_score = 4;
126
127 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
128 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
129 !ipv6_addr_equal(remote, &t->parms.raddr) ||
130 key != t->parms.i_key ||
131 !(t->dev->flags & IFF_UP))
132 continue;
133
134 if (t->dev->type != ARPHRD_IP6GRE &&
135 t->dev->type != dev_type)
136 continue;
137
138 score = 0;
139 if (t->parms.link != link)
140 score |= 1;
141 if (t->dev->type != dev_type)
142 score |= 2;
143 if (score == 0)
144 return t;
145
146 if (score < cand_score) {
147 cand = t;
148 cand_score = score;
149 }
150 }
151
152 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
153 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
154 key != t->parms.i_key ||
155 !(t->dev->flags & IFF_UP))
156 continue;
157
158 if (t->dev->type != ARPHRD_IP6GRE &&
159 t->dev->type != dev_type)
160 continue;
161
162 score = 0;
163 if (t->parms.link != link)
164 score |= 1;
165 if (t->dev->type != dev_type)
166 score |= 2;
167 if (score == 0)
168 return t;
169
170 if (score < cand_score) {
171 cand = t;
172 cand_score = score;
173 }
174 }
175
176 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
177 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
178 (!ipv6_addr_equal(local, &t->parms.raddr) ||
179 !ipv6_addr_is_multicast(local))) ||
180 key != t->parms.i_key ||
181 !(t->dev->flags & IFF_UP))
182 continue;
183
184 if (t->dev->type != ARPHRD_IP6GRE &&
185 t->dev->type != dev_type)
186 continue;
187
188 score = 0;
189 if (t->parms.link != link)
190 score |= 1;
191 if (t->dev->type != dev_type)
192 score |= 2;
193 if (score == 0)
194 return t;
195
196 if (score < cand_score) {
197 cand = t;
198 cand_score = score;
199 }
200 }
201
202 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
203 if (t->parms.i_key != key ||
204 !(t->dev->flags & IFF_UP))
205 continue;
206
207 if (t->dev->type != ARPHRD_IP6GRE &&
208 t->dev->type != dev_type)
209 continue;
210
211 score = 0;
212 if (t->parms.link != link)
213 score |= 1;
214 if (t->dev->type != dev_type)
215 score |= 2;
216 if (score == 0)
217 return t;
218
219 if (score < cand_score) {
220 cand = t;
221 cand_score = score;
222 }
223 }
224
225 if (cand != NULL)
226 return cand;
227
228 dev = ign->fb_tunnel_dev;
229 if (dev->flags & IFF_UP)
230 return netdev_priv(dev);
231
232 return NULL;
233}
234
235static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
236 const struct __ip6_tnl_parm *p)
237{
238 const struct in6_addr *remote = &p->raddr;
239 const struct in6_addr *local = &p->laddr;
240 unsigned int h = HASH_KEY(p->i_key);
241 int prio = 0;
242
243 if (!ipv6_addr_any(local))
244 prio |= 1;
245 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
246 prio |= 2;
247 h ^= HASH_ADDR(remote);
248 }
249
250 return &ign->tunnels[prio][h];
251}
252
253static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
254 const struct ip6_tnl *t)
255{
256 return __ip6gre_bucket(ign, &t->parms);
257}
258
259static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
260{
261 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
262
263 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
264 rcu_assign_pointer(*tp, t);
265}
266
267static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
268{
269 struct ip6_tnl __rcu **tp;
270 struct ip6_tnl *iter;
271
272 for (tp = ip6gre_bucket(ign, t);
273 (iter = rtnl_dereference(*tp)) != NULL;
274 tp = &iter->next) {
275 if (t == iter) {
276 rcu_assign_pointer(*tp, t->next);
277 break;
278 }
279 }
280}
281
282static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
283 const struct __ip6_tnl_parm *parms,
284 int type)
285{
286 const struct in6_addr *remote = &parms->raddr;
287 const struct in6_addr *local = &parms->laddr;
288 __be32 key = parms->i_key;
289 int link = parms->link;
290 struct ip6_tnl *t;
291 struct ip6_tnl __rcu **tp;
292 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
293
294 for (tp = __ip6gre_bucket(ign, parms);
295 (t = rtnl_dereference(*tp)) != NULL;
296 tp = &t->next)
297 if (ipv6_addr_equal(local, &t->parms.laddr) &&
298 ipv6_addr_equal(remote, &t->parms.raddr) &&
299 key == t->parms.i_key &&
300 link == t->parms.link &&
301 type == t->dev->type)
302 break;
303
304 return t;
305}
306
307static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
308 const struct __ip6_tnl_parm *parms, int create)
309{
310 struct ip6_tnl *t, *nt;
311 struct net_device *dev;
312 char name[IFNAMSIZ];
313 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
314
315 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
316 if (t || !create)
317 return t;
318
319 if (parms->name[0])
320 strlcpy(name, parms->name, IFNAMSIZ);
321 else
322 strcpy(name, "ip6gre%d");
323
324 dev = alloc_netdev(sizeof(*t), name, ip6gre_tunnel_setup);
325 if (!dev)
326 return NULL;
327
328 dev_net_set(dev, net);
329
330 nt = netdev_priv(dev);
331 nt->parms = *parms;
332 dev->rtnl_link_ops = &ip6gre_link_ops;
333
334 nt->dev = dev;
335 nt->net = dev_net(dev);
336 ip6gre_tnl_link_config(nt, 1);
337
338 if (register_netdevice(dev) < 0)
339 goto failed_free;
340
341 /* Can use a lockless transmit, unless we generate output sequences */
342 if (!(nt->parms.o_flags & GRE_SEQ))
343 dev->features |= NETIF_F_LLTX;
344
345 dev_hold(dev);
346 ip6gre_tunnel_link(ign, nt);
347 return nt;
348
349failed_free:
350 free_netdev(dev);
351 return NULL;
352}
353
354static void ip6gre_tunnel_uninit(struct net_device *dev)
355{
356 struct net *net = dev_net(dev);
357 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
358
359 ip6gre_tunnel_unlink(ign, netdev_priv(dev));
360 dev_put(dev);
361}
362
363
364static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
365 u8 type, u8 code, int offset, __be32 info)
366{
367 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
368 __be16 *p = (__be16 *)(skb->data + offset);
369 int grehlen = offset + 4;
370 struct ip6_tnl *t;
371 __be16 flags;
372
373 flags = p[0];
374 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
375 if (flags&(GRE_VERSION|GRE_ROUTING))
376 return;
377 if (flags&GRE_KEY) {
378 grehlen += 4;
379 if (flags&GRE_CSUM)
380 grehlen += 4;
381 }
382 }
383
384 /* If only 8 bytes returned, keyed message will be dropped here */
385 if (!pskb_may_pull(skb, grehlen))
386 return;
387 ipv6h = (const struct ipv6hdr *)skb->data;
388 p = (__be16 *)(skb->data + offset);
389
390 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
391 flags & GRE_KEY ?
392 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
393 p[1]);
394 if (t == NULL)
395 return;
396
397 switch (type) {
398 __u32 teli;
399 struct ipv6_tlv_tnl_enc_lim *tel;
400 __u32 mtu;
401 case ICMPV6_DEST_UNREACH:
402 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
403 t->parms.name);
404 break;
405 case ICMPV6_TIME_EXCEED:
406 if (code == ICMPV6_EXC_HOPLIMIT) {
407 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
408 t->parms.name);
409 }
410 break;
411 case ICMPV6_PARAMPROB:
412 teli = 0;
413 if (code == ICMPV6_HDR_FIELD)
414 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
415
416 if (teli && teli == info - 2) {
417 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
418 if (tel->encap_limit == 0) {
419 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
420 t->parms.name);
421 }
422 } else {
423 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
424 t->parms.name);
425 }
426 break;
427 case ICMPV6_PKT_TOOBIG:
428 mtu = info - offset;
429 if (mtu < IPV6_MIN_MTU)
430 mtu = IPV6_MIN_MTU;
431 t->dev->mtu = mtu;
432 break;
433 }
434
435 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
436 t->err_count++;
437 else
438 t->err_count = 1;
439 t->err_time = jiffies;
440}
441
442static int ip6gre_rcv(struct sk_buff *skb)
443{
444 const struct ipv6hdr *ipv6h;
445 u8 *h;
446 __be16 flags;
447 __sum16 csum = 0;
448 __be32 key = 0;
449 u32 seqno = 0;
450 struct ip6_tnl *tunnel;
451 int offset = 4;
452 __be16 gre_proto;
453 int err;
454
455 if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
456 goto drop;
457
458 ipv6h = ipv6_hdr(skb);
459 h = skb->data;
460 flags = *(__be16 *)h;
461
462 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
463 /* - Version must be 0.
464 - We do not support routing headers.
465 */
466 if (flags&(GRE_VERSION|GRE_ROUTING))
467 goto drop;
468
469 if (flags&GRE_CSUM) {
470 switch (skb->ip_summed) {
471 case CHECKSUM_COMPLETE:
472 csum = csum_fold(skb->csum);
473 if (!csum)
474 break;
475 /* fall through */
476 case CHECKSUM_NONE:
477 skb->csum = 0;
478 csum = __skb_checksum_complete(skb);
479 skb->ip_summed = CHECKSUM_COMPLETE;
480 }
481 offset += 4;
482 }
483 if (flags&GRE_KEY) {
484 key = *(__be32 *)(h + offset);
485 offset += 4;
486 }
487 if (flags&GRE_SEQ) {
488 seqno = ntohl(*(__be32 *)(h + offset));
489 offset += 4;
490 }
491 }
492
493 gre_proto = *(__be16 *)(h + 2);
494
495 tunnel = ip6gre_tunnel_lookup(skb->dev,
496 &ipv6h->saddr, &ipv6h->daddr, key,
497 gre_proto);
498 if (tunnel) {
499 struct pcpu_sw_netstats *tstats;
500
501 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
502 goto drop;
503
504 if (!ip6_tnl_rcv_ctl(tunnel, &ipv6h->daddr, &ipv6h->saddr)) {
505 tunnel->dev->stats.rx_dropped++;
506 goto drop;
507 }
508
509 skb->protocol = gre_proto;
510 /* WCCP version 1 and 2 protocol decoding.
511 * - Change protocol to IP
512 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
513 */
514 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
515 skb->protocol = htons(ETH_P_IP);
516 if ((*(h + offset) & 0xF0) != 0x40)
517 offset += 4;
518 }
519
520 skb->mac_header = skb->network_header;
521 __pskb_pull(skb, offset);
522 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
523
524 if (((flags&GRE_CSUM) && csum) ||
525 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
526 tunnel->dev->stats.rx_crc_errors++;
527 tunnel->dev->stats.rx_errors++;
528 goto drop;
529 }
530 if (tunnel->parms.i_flags&GRE_SEQ) {
531 if (!(flags&GRE_SEQ) ||
532 (tunnel->i_seqno &&
533 (s32)(seqno - tunnel->i_seqno) < 0)) {
534 tunnel->dev->stats.rx_fifo_errors++;
535 tunnel->dev->stats.rx_errors++;
536 goto drop;
537 }
538 tunnel->i_seqno = seqno + 1;
539 }
540
541 /* Warning: All skb pointers will be invalidated! */
542 if (tunnel->dev->type == ARPHRD_ETHER) {
543 if (!pskb_may_pull(skb, ETH_HLEN)) {
544 tunnel->dev->stats.rx_length_errors++;
545 tunnel->dev->stats.rx_errors++;
546 goto drop;
547 }
548
549 ipv6h = ipv6_hdr(skb);
550 skb->protocol = eth_type_trans(skb, tunnel->dev);
551 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
552 }
553
554 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
555
556 skb_reset_network_header(skb);
557
558 err = IP6_ECN_decapsulate(ipv6h, skb);
559 if (unlikely(err)) {
560 if (log_ecn_error)
561 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
562 &ipv6h->saddr,
563 ipv6_get_dsfield(ipv6h));
564 if (err > 1) {
565 ++tunnel->dev->stats.rx_frame_errors;
566 ++tunnel->dev->stats.rx_errors;
567 goto drop;
568 }
569 }
570
571 tstats = this_cpu_ptr(tunnel->dev->tstats);
572 u64_stats_update_begin(&tstats->syncp);
573 tstats->rx_packets++;
574 tstats->rx_bytes += skb->len;
575 u64_stats_update_end(&tstats->syncp);
576
577 netif_rx(skb);
578
579 return 0;
580 }
581 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
582
583drop:
584 kfree_skb(skb);
585 return 0;
586}
587
588struct ipv6_tel_txoption {
589 struct ipv6_txoptions ops;
590 __u8 dst_opt[8];
591};
592
593static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
594{
595 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
596
597 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
598 opt->dst_opt[3] = 1;
599 opt->dst_opt[4] = encap_limit;
600 opt->dst_opt[5] = IPV6_TLV_PADN;
601 opt->dst_opt[6] = 1;
602
603 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
604 opt->ops.opt_nflen = 8;
605}
606
607static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
608 struct net_device *dev,
609 __u8 dsfield,
610 struct flowi6 *fl6,
611 int encap_limit,
612 __u32 *pmtu)
613{
614 struct net *net = dev_net(dev);
615 struct ip6_tnl *tunnel = netdev_priv(dev);
616 struct net_device *tdev; /* Device to other host */
617 struct ipv6hdr *ipv6h; /* Our new IP header */
618 unsigned int max_headroom = 0; /* The extra header space needed */
619 int gre_hlen;
620 struct ipv6_tel_txoption opt;
621 int mtu;
622 struct dst_entry *dst = NULL, *ndst = NULL;
623 struct net_device_stats *stats = &tunnel->dev->stats;
624 int err = -1;
625 u8 proto;
626 struct sk_buff *new_skb;
627
628 if (dev->type == ARPHRD_ETHER)
629 IPCB(skb)->flags = 0;
630
631 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) {
632 gre_hlen = 0;
633 ipv6h = (struct ipv6hdr *)skb->data;
634 fl6->daddr = ipv6h->daddr;
635 } else {
636 gre_hlen = tunnel->hlen;
637 fl6->daddr = tunnel->parms.raddr;
638 }
639
640 if (!fl6->flowi6_mark)
641 dst = ip6_tnl_dst_check(tunnel);
642
643 if (!dst) {
644 ndst = ip6_route_output(net, NULL, fl6);
645
646 if (ndst->error)
647 goto tx_err_link_failure;
648 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
649 if (IS_ERR(ndst)) {
650 err = PTR_ERR(ndst);
651 ndst = NULL;
652 goto tx_err_link_failure;
653 }
654 dst = ndst;
655 }
656
657 tdev = dst->dev;
658
659 if (tdev == dev) {
660 stats->collisions++;
661 net_warn_ratelimited("%s: Local routing loop detected!\n",
662 tunnel->parms.name);
663 goto tx_err_dst_release;
664 }
665
666 mtu = dst_mtu(dst) - sizeof(*ipv6h);
667 if (encap_limit >= 0) {
668 max_headroom += 8;
669 mtu -= 8;
670 }
671 if (mtu < IPV6_MIN_MTU)
672 mtu = IPV6_MIN_MTU;
673 if (skb_dst(skb))
674 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
675 if (skb->len > mtu) {
676 *pmtu = mtu;
677 err = -EMSGSIZE;
678 goto tx_err_dst_release;
679 }
680
681 if (tunnel->err_count > 0) {
682 if (time_before(jiffies,
683 tunnel->err_time + IP6TUNNEL_ERR_TIMEO)) {
684 tunnel->err_count--;
685
686 dst_link_failure(skb);
687 } else
688 tunnel->err_count = 0;
689 }
690
691 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
692
693 max_headroom += LL_RESERVED_SPACE(tdev) + gre_hlen + dst->header_len;
694
695 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
696 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
697 new_skb = skb_realloc_headroom(skb, max_headroom);
698 if (max_headroom > dev->needed_headroom)
699 dev->needed_headroom = max_headroom;
700 if (!new_skb)
701 goto tx_err_dst_release;
702
703 if (skb->sk)
704 skb_set_owner_w(new_skb, skb->sk);
705 consume_skb(skb);
706 skb = new_skb;
707 }
708
709 if (fl6->flowi6_mark) {
710 skb_dst_set(skb, dst);
711 ndst = NULL;
712 } else {
713 skb_dst_set_noref(skb, dst);
714 }
715
716 proto = NEXTHDR_GRE;
717 if (encap_limit >= 0) {
718 init_tel_txopt(&opt, encap_limit);
719 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
720 }
721
722 if (likely(!skb->encapsulation)) {
723 skb_reset_inner_headers(skb);
724 skb->encapsulation = 1;
725 }
726
727 skb_push(skb, gre_hlen);
728 skb_reset_network_header(skb);
729 skb_set_transport_header(skb, sizeof(*ipv6h));
730
731 /*
732 * Push down and install the IP header.
733 */
734 ipv6h = ipv6_hdr(skb);
735 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
736 ipv6h->hop_limit = tunnel->parms.hop_limit;
737 ipv6h->nexthdr = proto;
738 ipv6h->saddr = fl6->saddr;
739 ipv6h->daddr = fl6->daddr;
740
741 ((__be16 *)(ipv6h + 1))[0] = tunnel->parms.o_flags;
742 ((__be16 *)(ipv6h + 1))[1] = (dev->type == ARPHRD_ETHER) ?
743 htons(ETH_P_TEB) : skb->protocol;
744
745 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
746 __be32 *ptr = (__be32 *)(((u8 *)ipv6h) + tunnel->hlen - 4);
747
748 if (tunnel->parms.o_flags&GRE_SEQ) {
749 ++tunnel->o_seqno;
750 *ptr = htonl(tunnel->o_seqno);
751 ptr--;
752 }
753 if (tunnel->parms.o_flags&GRE_KEY) {
754 *ptr = tunnel->parms.o_key;
755 ptr--;
756 }
757 if (tunnel->parms.o_flags&GRE_CSUM) {
758 *ptr = 0;
759 *(__sum16 *)ptr = ip_compute_csum((void *)(ipv6h+1),
760 skb->len - sizeof(struct ipv6hdr));
761 }
762 }
763
764 ip6tunnel_xmit(skb, dev);
765 if (ndst)
766 ip6_tnl_dst_store(tunnel, ndst);
767 return 0;
768tx_err_link_failure:
769 stats->tx_carrier_errors++;
770 dst_link_failure(skb);
771tx_err_dst_release:
772 dst_release(ndst);
773 return err;
774}
775
776static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
777{
778 struct ip6_tnl *t = netdev_priv(dev);
779 const struct iphdr *iph = ip_hdr(skb);
780 int encap_limit = -1;
781 struct flowi6 fl6;
782 __u8 dsfield;
783 __u32 mtu;
784 int err;
785
786 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
787 encap_limit = t->parms.encap_limit;
788
789 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
790 fl6.flowi6_proto = IPPROTO_IPIP;
791
792 dsfield = ipv4_get_dsfield(iph);
793
794 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
795 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
796 & IPV6_TCLASS_MASK;
797 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
798 fl6.flowi6_mark = skb->mark;
799
800 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
801 if (err != 0) {
802 /* XXX: send ICMP error even if DF is not set. */
803 if (err == -EMSGSIZE)
804 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
805 htonl(mtu));
806 return -1;
807 }
808
809 return 0;
810}
811
812static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
813{
814 struct ip6_tnl *t = netdev_priv(dev);
815 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
816 int encap_limit = -1;
817 __u16 offset;
818 struct flowi6 fl6;
819 __u8 dsfield;
820 __u32 mtu;
821 int err;
822
823 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
824 return -1;
825
826 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
827 if (offset > 0) {
828 struct ipv6_tlv_tnl_enc_lim *tel;
829 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
830 if (tel->encap_limit == 0) {
831 icmpv6_send(skb, ICMPV6_PARAMPROB,
832 ICMPV6_HDR_FIELD, offset + 2);
833 return -1;
834 }
835 encap_limit = tel->encap_limit - 1;
836 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
837 encap_limit = t->parms.encap_limit;
838
839 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
840 fl6.flowi6_proto = IPPROTO_IPV6;
841
842 dsfield = ipv6_get_dsfield(ipv6h);
843 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
844 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
845 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
846 fl6.flowlabel |= ip6_flowlabel(ipv6h);
847 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
848 fl6.flowi6_mark = skb->mark;
849
850 err = ip6gre_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
851 if (err != 0) {
852 if (err == -EMSGSIZE)
853 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
854 return -1;
855 }
856
857 return 0;
858}
859
860/**
861 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
862 * @t: the outgoing tunnel device
863 * @hdr: IPv6 header from the incoming packet
864 *
865 * Description:
866 * Avoid trivial tunneling loop by checking that tunnel exit-point
867 * doesn't match source of incoming packet.
868 *
869 * Return:
870 * 1 if conflict,
871 * 0 else
872 **/
873
874static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
875 const struct ipv6hdr *hdr)
876{
877 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
878}
879
880static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
881{
882 struct ip6_tnl *t = netdev_priv(dev);
883 int encap_limit = -1;
884 struct flowi6 fl6;
885 __u32 mtu;
886 int err;
887
888 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
889 encap_limit = t->parms.encap_limit;
890
891 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
892 fl6.flowi6_proto = skb->protocol;
893
894 err = ip6gre_xmit2(skb, dev, 0, &fl6, encap_limit, &mtu);
895
896 return err;
897}
898
899static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
900 struct net_device *dev)
901{
902 struct ip6_tnl *t = netdev_priv(dev);
903 struct net_device_stats *stats = &t->dev->stats;
904 int ret;
905
906 if (!ip6_tnl_xmit_ctl(t))
907 goto tx_err;
908
909 switch (skb->protocol) {
910 case htons(ETH_P_IP):
911 ret = ip6gre_xmit_ipv4(skb, dev);
912 break;
913 case htons(ETH_P_IPV6):
914 ret = ip6gre_xmit_ipv6(skb, dev);
915 break;
916 default:
917 ret = ip6gre_xmit_other(skb, dev);
918 break;
919 }
920
921 if (ret < 0)
922 goto tx_err;
923
924 return NETDEV_TX_OK;
925
926tx_err:
927 stats->tx_errors++;
928 stats->tx_dropped++;
929 kfree_skb(skb);
930 return NETDEV_TX_OK;
931}
932
933static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
934{
935 struct net_device *dev = t->dev;
936 struct __ip6_tnl_parm *p = &t->parms;
937 struct flowi6 *fl6 = &t->fl.u.ip6;
938 int addend = sizeof(struct ipv6hdr) + 4;
939
940 if (dev->type != ARPHRD_ETHER) {
941 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
942 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
943 }
944
945 /* Set up flowi template */
946 fl6->saddr = p->laddr;
947 fl6->daddr = p->raddr;
948 fl6->flowi6_oif = p->link;
949 fl6->flowlabel = 0;
950
951 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
952 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
953 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
954 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
955
956 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
957 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
958
959 if (p->flags&IP6_TNL_F_CAP_XMIT &&
960 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
961 dev->flags |= IFF_POINTOPOINT;
962 else
963 dev->flags &= ~IFF_POINTOPOINT;
964
965 dev->iflink = p->link;
966
967 /* Precalculate GRE options length */
968 if (t->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
969 if (t->parms.o_flags&GRE_CSUM)
970 addend += 4;
971 if (t->parms.o_flags&GRE_KEY)
972 addend += 4;
973 if (t->parms.o_flags&GRE_SEQ)
974 addend += 4;
975 }
976 t->hlen = addend;
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 }
1001 ip6_rt_put(rt);
1002 }
1003}
1004
1005static int ip6gre_tnl_change(struct ip6_tnl *t,
1006 const struct __ip6_tnl_parm *p, int set_mtu)
1007{
1008 t->parms.laddr = p->laddr;
1009 t->parms.raddr = p->raddr;
1010 t->parms.flags = p->flags;
1011 t->parms.hop_limit = p->hop_limit;
1012 t->parms.encap_limit = p->encap_limit;
1013 t->parms.flowinfo = p->flowinfo;
1014 t->parms.link = p->link;
1015 t->parms.proto = p->proto;
1016 t->parms.i_key = p->i_key;
1017 t->parms.o_key = p->o_key;
1018 t->parms.i_flags = p->i_flags;
1019 t->parms.o_flags = p->o_flags;
1020 ip6_tnl_dst_reset(t);
1021 ip6gre_tnl_link_config(t, set_mtu);
1022 return 0;
1023}
1024
1025static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1026 const struct ip6_tnl_parm2 *u)
1027{
1028 p->laddr = u->laddr;
1029 p->raddr = u->raddr;
1030 p->flags = u->flags;
1031 p->hop_limit = u->hop_limit;
1032 p->encap_limit = u->encap_limit;
1033 p->flowinfo = u->flowinfo;
1034 p->link = u->link;
1035 p->i_key = u->i_key;
1036 p->o_key = u->o_key;
1037 p->i_flags = u->i_flags;
1038 p->o_flags = u->o_flags;
1039 memcpy(p->name, u->name, sizeof(u->name));
1040}
1041
1042static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1043 const struct __ip6_tnl_parm *p)
1044{
1045 u->proto = IPPROTO_GRE;
1046 u->laddr = p->laddr;
1047 u->raddr = p->raddr;
1048 u->flags = p->flags;
1049 u->hop_limit = p->hop_limit;
1050 u->encap_limit = p->encap_limit;
1051 u->flowinfo = p->flowinfo;
1052 u->link = p->link;
1053 u->i_key = p->i_key;
1054 u->o_key = p->o_key;
1055 u->i_flags = p->i_flags;
1056 u->o_flags = p->o_flags;
1057 memcpy(u->name, p->name, sizeof(u->name));
1058}
1059
1060static int ip6gre_tunnel_ioctl(struct net_device *dev,
1061 struct ifreq *ifr, int cmd)
1062{
1063 int err = 0;
1064 struct ip6_tnl_parm2 p;
1065 struct __ip6_tnl_parm p1;
1066 struct ip6_tnl *t;
1067 struct net *net = dev_net(dev);
1068 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1069
1070 switch (cmd) {
1071 case SIOCGETTUNNEL:
1072 t = NULL;
1073 if (dev == ign->fb_tunnel_dev) {
1074 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1075 err = -EFAULT;
1076 break;
1077 }
1078 ip6gre_tnl_parm_from_user(&p1, &p);
1079 t = ip6gre_tunnel_locate(net, &p1, 0);
1080 }
1081 if (t == NULL)
1082 t = netdev_priv(dev);
1083 memset(&p, 0, sizeof(p));
1084 ip6gre_tnl_parm_to_user(&p, &t->parms);
1085 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1086 err = -EFAULT;
1087 break;
1088
1089 case SIOCADDTUNNEL:
1090 case SIOCCHGTUNNEL:
1091 err = -EPERM;
1092 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1093 goto done;
1094
1095 err = -EFAULT;
1096 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1097 goto done;
1098
1099 err = -EINVAL;
1100 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1101 goto done;
1102
1103 if (!(p.i_flags&GRE_KEY))
1104 p.i_key = 0;
1105 if (!(p.o_flags&GRE_KEY))
1106 p.o_key = 0;
1107
1108 ip6gre_tnl_parm_from_user(&p1, &p);
1109 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1110
1111 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1112 if (t != NULL) {
1113 if (t->dev != dev) {
1114 err = -EEXIST;
1115 break;
1116 }
1117 } else {
1118 t = netdev_priv(dev);
1119
1120 ip6gre_tunnel_unlink(ign, t);
1121 synchronize_net();
1122 ip6gre_tnl_change(t, &p1, 1);
1123 ip6gre_tunnel_link(ign, t);
1124 netdev_state_change(dev);
1125 }
1126 }
1127
1128 if (t) {
1129 err = 0;
1130
1131 memset(&p, 0, sizeof(p));
1132 ip6gre_tnl_parm_to_user(&p, &t->parms);
1133 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1134 err = -EFAULT;
1135 } else
1136 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1137 break;
1138
1139 case SIOCDELTUNNEL:
1140 err = -EPERM;
1141 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1142 goto done;
1143
1144 if (dev == ign->fb_tunnel_dev) {
1145 err = -EFAULT;
1146 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1147 goto done;
1148 err = -ENOENT;
1149 ip6gre_tnl_parm_from_user(&p1, &p);
1150 t = ip6gre_tunnel_locate(net, &p1, 0);
1151 if (t == NULL)
1152 goto done;
1153 err = -EPERM;
1154 if (t == netdev_priv(ign->fb_tunnel_dev))
1155 goto done;
1156 dev = t->dev;
1157 }
1158 unregister_netdevice(dev);
1159 err = 0;
1160 break;
1161
1162 default:
1163 err = -EINVAL;
1164 }
1165
1166done:
1167 return err;
1168}
1169
1170static int ip6gre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1171{
1172 if (new_mtu < 68 ||
1173 new_mtu > 0xFFF8 - dev->hard_header_len)
1174 return -EINVAL;
1175 dev->mtu = new_mtu;
1176 return 0;
1177}
1178
1179static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1180 unsigned short type,
1181 const void *daddr, const void *saddr, unsigned int len)
1182{
1183 struct ip6_tnl *t = netdev_priv(dev);
1184 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
1185 __be16 *p = (__be16 *)(ipv6h+1);
1186
1187 ip6_flow_hdr(ipv6h, 0, t->fl.u.ip6.flowlabel);
1188 ipv6h->hop_limit = t->parms.hop_limit;
1189 ipv6h->nexthdr = NEXTHDR_GRE;
1190 ipv6h->saddr = t->parms.laddr;
1191 ipv6h->daddr = t->parms.raddr;
1192
1193 p[0] = t->parms.o_flags;
1194 p[1] = htons(type);
1195
1196 /*
1197 * Set the source hardware address.
1198 */
1199
1200 if (saddr)
1201 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1202 if (daddr)
1203 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1204 if (!ipv6_addr_any(&ipv6h->daddr))
1205 return t->hlen;
1206
1207 return -t->hlen;
1208}
1209
1210static const struct header_ops ip6gre_header_ops = {
1211 .create = ip6gre_header,
1212};
1213
1214static const struct net_device_ops ip6gre_netdev_ops = {
1215 .ndo_init = ip6gre_tunnel_init,
1216 .ndo_uninit = ip6gre_tunnel_uninit,
1217 .ndo_start_xmit = ip6gre_tunnel_xmit,
1218 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1219 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1220 .ndo_get_stats64 = ip_tunnel_get_stats64,
1221};
1222
1223static void ip6gre_dev_free(struct net_device *dev)
1224{
1225 free_percpu(dev->tstats);
1226 free_netdev(dev);
1227}
1228
1229static void ip6gre_tunnel_setup(struct net_device *dev)
1230{
1231 struct ip6_tnl *t;
1232
1233 dev->netdev_ops = &ip6gre_netdev_ops;
1234 dev->destructor = ip6gre_dev_free;
1235
1236 dev->type = ARPHRD_IP6GRE;
1237 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr) + 4;
1238 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr) - 4;
1239 t = netdev_priv(dev);
1240 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1241 dev->mtu -= 8;
1242 dev->flags |= IFF_NOARP;
1243 dev->iflink = 0;
1244 dev->addr_len = sizeof(struct in6_addr);
1245 dev->features |= NETIF_F_NETNS_LOCAL;
1246 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1247}
1248
1249static int ip6gre_tunnel_init(struct net_device *dev)
1250{
1251 struct ip6_tnl *tunnel;
1252 int i;
1253
1254 tunnel = netdev_priv(dev);
1255
1256 tunnel->dev = dev;
1257 tunnel->net = dev_net(dev);
1258 strcpy(tunnel->parms.name, dev->name);
1259
1260 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1261 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1262
1263 if (ipv6_addr_any(&tunnel->parms.raddr))
1264 dev->header_ops = &ip6gre_header_ops;
1265
1266 dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1267 if (!dev->tstats)
1268 return -ENOMEM;
1269
1270 for_each_possible_cpu(i) {
1271 struct pcpu_sw_netstats *ip6gre_tunnel_stats;
1272 ip6gre_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1273 u64_stats_init(&ip6gre_tunnel_stats->syncp);
1274 }
1275
1276
1277 return 0;
1278}
1279
1280static void ip6gre_fb_tunnel_init(struct net_device *dev)
1281{
1282 struct ip6_tnl *tunnel = netdev_priv(dev);
1283
1284 tunnel->dev = dev;
1285 tunnel->net = dev_net(dev);
1286 strcpy(tunnel->parms.name, dev->name);
1287
1288 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1289
1290 dev_hold(dev);
1291}
1292
1293
1294static struct inet6_protocol ip6gre_protocol __read_mostly = {
1295 .handler = ip6gre_rcv,
1296 .err_handler = ip6gre_err,
1297 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1298};
1299
1300static void ip6gre_destroy_tunnels(struct ip6gre_net *ign,
1301 struct list_head *head)
1302{
1303 int prio;
1304
1305 for (prio = 0; prio < 4; prio++) {
1306 int h;
1307 for (h = 0; h < HASH_SIZE; h++) {
1308 struct ip6_tnl *t;
1309
1310 t = rtnl_dereference(ign->tunnels[prio][h]);
1311
1312 while (t != NULL) {
1313 unregister_netdevice_queue(t->dev, head);
1314 t = rtnl_dereference(t->next);
1315 }
1316 }
1317 }
1318}
1319
1320static int __net_init ip6gre_init_net(struct net *net)
1321{
1322 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1323 int err;
1324
1325 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1326 ip6gre_tunnel_setup);
1327 if (!ign->fb_tunnel_dev) {
1328 err = -ENOMEM;
1329 goto err_alloc_dev;
1330 }
1331 dev_net_set(ign->fb_tunnel_dev, net);
1332
1333 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1334 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1335
1336 err = register_netdev(ign->fb_tunnel_dev);
1337 if (err)
1338 goto err_reg_dev;
1339
1340 rcu_assign_pointer(ign->tunnels_wc[0],
1341 netdev_priv(ign->fb_tunnel_dev));
1342 return 0;
1343
1344err_reg_dev:
1345 ip6gre_dev_free(ign->fb_tunnel_dev);
1346err_alloc_dev:
1347 return err;
1348}
1349
1350static void __net_exit ip6gre_exit_net(struct net *net)
1351{
1352 struct ip6gre_net *ign;
1353 LIST_HEAD(list);
1354
1355 ign = net_generic(net, ip6gre_net_id);
1356 rtnl_lock();
1357 ip6gre_destroy_tunnels(ign, &list);
1358 unregister_netdevice_many(&list);
1359 rtnl_unlock();
1360}
1361
1362static struct pernet_operations ip6gre_net_ops = {
1363 .init = ip6gre_init_net,
1364 .exit = ip6gre_exit_net,
1365 .id = &ip6gre_net_id,
1366 .size = sizeof(struct ip6gre_net),
1367};
1368
1369static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1370{
1371 __be16 flags;
1372
1373 if (!data)
1374 return 0;
1375
1376 flags = 0;
1377 if (data[IFLA_GRE_IFLAGS])
1378 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1379 if (data[IFLA_GRE_OFLAGS])
1380 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1381 if (flags & (GRE_VERSION|GRE_ROUTING))
1382 return -EINVAL;
1383
1384 return 0;
1385}
1386
1387static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1388{
1389 struct in6_addr daddr;
1390
1391 if (tb[IFLA_ADDRESS]) {
1392 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1393 return -EINVAL;
1394 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1395 return -EADDRNOTAVAIL;
1396 }
1397
1398 if (!data)
1399 goto out;
1400
1401 if (data[IFLA_GRE_REMOTE]) {
1402 nla_memcpy(&daddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1403 if (ipv6_addr_any(&daddr))
1404 return -EINVAL;
1405 }
1406
1407out:
1408 return ip6gre_tunnel_validate(tb, data);
1409}
1410
1411
1412static void ip6gre_netlink_parms(struct nlattr *data[],
1413 struct __ip6_tnl_parm *parms)
1414{
1415 memset(parms, 0, sizeof(*parms));
1416
1417 if (!data)
1418 return;
1419
1420 if (data[IFLA_GRE_LINK])
1421 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1422
1423 if (data[IFLA_GRE_IFLAGS])
1424 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1425
1426 if (data[IFLA_GRE_OFLAGS])
1427 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1428
1429 if (data[IFLA_GRE_IKEY])
1430 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1431
1432 if (data[IFLA_GRE_OKEY])
1433 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1434
1435 if (data[IFLA_GRE_LOCAL])
1436 nla_memcpy(&parms->laddr, data[IFLA_GRE_LOCAL], sizeof(struct in6_addr));
1437
1438 if (data[IFLA_GRE_REMOTE])
1439 nla_memcpy(&parms->raddr, data[IFLA_GRE_REMOTE], sizeof(struct in6_addr));
1440
1441 if (data[IFLA_GRE_TTL])
1442 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1443
1444 if (data[IFLA_GRE_ENCAP_LIMIT])
1445 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1446
1447 if (data[IFLA_GRE_FLOWINFO])
1448 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1449
1450 if (data[IFLA_GRE_FLAGS])
1451 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1452}
1453
1454static int ip6gre_tap_init(struct net_device *dev)
1455{
1456 struct ip6_tnl *tunnel;
1457
1458 tunnel = netdev_priv(dev);
1459
1460 tunnel->dev = dev;
1461 tunnel->net = dev_net(dev);
1462 strcpy(tunnel->parms.name, dev->name);
1463
1464 ip6gre_tnl_link_config(tunnel, 1);
1465
1466 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1467 if (!dev->tstats)
1468 return -ENOMEM;
1469
1470 return 0;
1471}
1472
1473static const struct net_device_ops ip6gre_tap_netdev_ops = {
1474 .ndo_init = ip6gre_tap_init,
1475 .ndo_uninit = ip6gre_tunnel_uninit,
1476 .ndo_start_xmit = ip6gre_tunnel_xmit,
1477 .ndo_set_mac_address = eth_mac_addr,
1478 .ndo_validate_addr = eth_validate_addr,
1479 .ndo_change_mtu = ip6gre_tunnel_change_mtu,
1480 .ndo_get_stats64 = ip_tunnel_get_stats64,
1481};
1482
1483static void ip6gre_tap_setup(struct net_device *dev)
1484{
1485
1486 ether_setup(dev);
1487
1488 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1489 dev->destructor = ip6gre_dev_free;
1490
1491 dev->iflink = 0;
1492 dev->features |= NETIF_F_NETNS_LOCAL;
1493}
1494
1495static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1496 struct nlattr *tb[], struct nlattr *data[])
1497{
1498 struct ip6_tnl *nt;
1499 struct net *net = dev_net(dev);
1500 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1501 int err;
1502
1503 nt = netdev_priv(dev);
1504 ip6gre_netlink_parms(data, &nt->parms);
1505
1506 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1507 return -EEXIST;
1508
1509 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1510 eth_hw_addr_random(dev);
1511
1512 nt->dev = dev;
1513 nt->net = dev_net(dev);
1514 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1515
1516 /* Can use a lockless transmit, unless we generate output sequences */
1517 if (!(nt->parms.o_flags & GRE_SEQ))
1518 dev->features |= NETIF_F_LLTX;
1519
1520 err = register_netdevice(dev);
1521 if (err)
1522 goto out;
1523
1524 dev_hold(dev);
1525 ip6gre_tunnel_link(ign, nt);
1526
1527out:
1528 return err;
1529}
1530
1531static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1532 struct nlattr *data[])
1533{
1534 struct ip6_tnl *t, *nt;
1535 struct net *net = dev_net(dev);
1536 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1537 struct __ip6_tnl_parm p;
1538
1539 if (dev == ign->fb_tunnel_dev)
1540 return -EINVAL;
1541
1542 nt = netdev_priv(dev);
1543 ip6gre_netlink_parms(data, &p);
1544
1545 t = ip6gre_tunnel_locate(net, &p, 0);
1546
1547 if (t) {
1548 if (t->dev != dev)
1549 return -EEXIST;
1550 } else {
1551 t = nt;
1552
1553 ip6gre_tunnel_unlink(ign, t);
1554 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1555 ip6gre_tunnel_link(ign, t);
1556 netdev_state_change(dev);
1557 }
1558
1559 return 0;
1560}
1561
1562static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1563{
1564 struct net *net = dev_net(dev);
1565 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1566
1567 if (dev != ign->fb_tunnel_dev)
1568 unregister_netdevice_queue(dev, head);
1569}
1570
1571static size_t ip6gre_get_size(const struct net_device *dev)
1572{
1573 return
1574 /* IFLA_GRE_LINK */
1575 nla_total_size(4) +
1576 /* IFLA_GRE_IFLAGS */
1577 nla_total_size(2) +
1578 /* IFLA_GRE_OFLAGS */
1579 nla_total_size(2) +
1580 /* IFLA_GRE_IKEY */
1581 nla_total_size(4) +
1582 /* IFLA_GRE_OKEY */
1583 nla_total_size(4) +
1584 /* IFLA_GRE_LOCAL */
1585 nla_total_size(sizeof(struct in6_addr)) +
1586 /* IFLA_GRE_REMOTE */
1587 nla_total_size(sizeof(struct in6_addr)) +
1588 /* IFLA_GRE_TTL */
1589 nla_total_size(1) +
1590 /* IFLA_GRE_TOS */
1591 nla_total_size(1) +
1592 /* IFLA_GRE_ENCAP_LIMIT */
1593 nla_total_size(1) +
1594 /* IFLA_GRE_FLOWINFO */
1595 nla_total_size(4) +
1596 /* IFLA_GRE_FLAGS */
1597 nla_total_size(4) +
1598 0;
1599}
1600
1601static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1602{
1603 struct ip6_tnl *t = netdev_priv(dev);
1604 struct __ip6_tnl_parm *p = &t->parms;
1605
1606 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1607 nla_put_be16(skb, IFLA_GRE_IFLAGS, p->i_flags) ||
1608 nla_put_be16(skb, IFLA_GRE_OFLAGS, p->o_flags) ||
1609 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1610 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1611 nla_put(skb, IFLA_GRE_LOCAL, sizeof(struct in6_addr), &p->laddr) ||
1612 nla_put(skb, IFLA_GRE_REMOTE, sizeof(struct in6_addr), &p->raddr) ||
1613 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1614 /*nla_put_u8(skb, IFLA_GRE_TOS, t->priority) ||*/
1615 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1616 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1617 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1618 goto nla_put_failure;
1619 return 0;
1620
1621nla_put_failure:
1622 return -EMSGSIZE;
1623}
1624
1625static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1626 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1627 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1628 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1629 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1630 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1631 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1632 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1633 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1634 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1635 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1636 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1637};
1638
1639static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1640 .kind = "ip6gre",
1641 .maxtype = IFLA_GRE_MAX,
1642 .policy = ip6gre_policy,
1643 .priv_size = sizeof(struct ip6_tnl),
1644 .setup = ip6gre_tunnel_setup,
1645 .validate = ip6gre_tunnel_validate,
1646 .newlink = ip6gre_newlink,
1647 .changelink = ip6gre_changelink,
1648 .dellink = ip6gre_dellink,
1649 .get_size = ip6gre_get_size,
1650 .fill_info = ip6gre_fill_info,
1651};
1652
1653static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1654 .kind = "ip6gretap",
1655 .maxtype = IFLA_GRE_MAX,
1656 .policy = ip6gre_policy,
1657 .priv_size = sizeof(struct ip6_tnl),
1658 .setup = ip6gre_tap_setup,
1659 .validate = ip6gre_tap_validate,
1660 .newlink = ip6gre_newlink,
1661 .changelink = ip6gre_changelink,
1662 .get_size = ip6gre_get_size,
1663 .fill_info = ip6gre_fill_info,
1664};
1665
1666/*
1667 * And now the modules code and kernel interface.
1668 */
1669
1670static int __init ip6gre_init(void)
1671{
1672 int err;
1673
1674 pr_info("GRE over IPv6 tunneling driver\n");
1675
1676 err = register_pernet_device(&ip6gre_net_ops);
1677 if (err < 0)
1678 return err;
1679
1680 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1681 if (err < 0) {
1682 pr_info("%s: can't add protocol\n", __func__);
1683 goto add_proto_failed;
1684 }
1685
1686 err = rtnl_link_register(&ip6gre_link_ops);
1687 if (err < 0)
1688 goto rtnl_link_failed;
1689
1690 err = rtnl_link_register(&ip6gre_tap_ops);
1691 if (err < 0)
1692 goto tap_ops_failed;
1693
1694out:
1695 return err;
1696
1697tap_ops_failed:
1698 rtnl_link_unregister(&ip6gre_link_ops);
1699rtnl_link_failed:
1700 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1701add_proto_failed:
1702 unregister_pernet_device(&ip6gre_net_ops);
1703 goto out;
1704}
1705
1706static void __exit ip6gre_fini(void)
1707{
1708 rtnl_link_unregister(&ip6gre_tap_ops);
1709 rtnl_link_unregister(&ip6gre_link_ops);
1710 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1711 unregister_pernet_device(&ip6gre_net_ops);
1712}
1713
1714module_init(ip6gre_init);
1715module_exit(ip6gre_fini);
1716MODULE_LICENSE("GPL");
1717MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1718MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1719MODULE_ALIAS_RTNL_LINK("ip6gre");
1720MODULE_ALIAS_NETDEV("ip6gre0");
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/init.h>
28#include <linux/in6.h>
29#include <linux/inetdevice.h>
30#include <linux/igmp.h>
31#include <linux/netfilter_ipv4.h>
32#include <linux/etherdevice.h>
33#include <linux/if_ether.h>
34#include <linux/hash.h>
35#include <linux/if_tunnel.h>
36#include <linux/ip6_tunnel.h>
37
38#include <net/sock.h>
39#include <net/ip.h>
40#include <net/ip_tunnels.h>
41#include <net/icmp.h>
42#include <net/protocol.h>
43#include <net/addrconf.h>
44#include <net/arp.h>
45#include <net/checksum.h>
46#include <net/dsfield.h>
47#include <net/inet_ecn.h>
48#include <net/xfrm.h>
49#include <net/net_namespace.h>
50#include <net/netns/generic.h>
51#include <net/rtnetlink.h>
52
53#include <net/ipv6.h>
54#include <net/ip6_fib.h>
55#include <net/ip6_route.h>
56#include <net/ip6_tunnel.h>
57#include <net/gre.h>
58#include <net/erspan.h>
59#include <net/dst_metadata.h>
60
61
62static bool log_ecn_error = true;
63module_param(log_ecn_error, bool, 0644);
64MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
65
66#define IP6_GRE_HASH_SIZE_SHIFT 5
67#define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
68
69static unsigned int ip6gre_net_id __read_mostly;
70struct ip6gre_net {
71 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
72
73 struct ip6_tnl __rcu *collect_md_tun;
74 struct ip6_tnl __rcu *collect_md_tun_erspan;
75 struct net_device *fb_tunnel_dev;
76};
77
78static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
80static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
81static int ip6gre_tunnel_init(struct net_device *dev);
82static void ip6gre_tunnel_setup(struct net_device *dev);
83static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
84static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
85static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
86
87/* Tunnel hash table */
88
89/*
90 4 hash tables:
91
92 3: (remote,local)
93 2: (remote,*)
94 1: (*,local)
95 0: (*,*)
96
97 We require exact key match i.e. if a key is present in packet
98 it will match only tunnel with the same key; if it is not present,
99 it will match only keyless tunnel.
100
101 All keysless packets, if not matched configured keyless tunnels
102 will match fallback tunnel.
103 */
104
105#define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
106static u32 HASH_ADDR(const struct in6_addr *addr)
107{
108 u32 hash = ipv6_addr_hash(addr);
109
110 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
111}
112
113#define tunnels_r_l tunnels[3]
114#define tunnels_r tunnels[2]
115#define tunnels_l tunnels[1]
116#define tunnels_wc tunnels[0]
117
118/* Given src, dst and key, find appropriate for input tunnel. */
119
120static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
121 const struct in6_addr *remote, const struct in6_addr *local,
122 __be32 key, __be16 gre_proto)
123{
124 struct net *net = dev_net(dev);
125 int link = dev->ifindex;
126 unsigned int h0 = HASH_ADDR(remote);
127 unsigned int h1 = HASH_KEY(key);
128 struct ip6_tnl *t, *cand = NULL;
129 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
130 int dev_type = (gre_proto == htons(ETH_P_TEB) ||
131 gre_proto == htons(ETH_P_ERSPAN) ||
132 gre_proto == htons(ETH_P_ERSPAN2)) ?
133 ARPHRD_ETHER : ARPHRD_IP6GRE;
134 int score, cand_score = 4;
135
136 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
137 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138 !ipv6_addr_equal(remote, &t->parms.raddr) ||
139 key != t->parms.i_key ||
140 !(t->dev->flags & IFF_UP))
141 continue;
142
143 if (t->dev->type != ARPHRD_IP6GRE &&
144 t->dev->type != dev_type)
145 continue;
146
147 score = 0;
148 if (t->parms.link != link)
149 score |= 1;
150 if (t->dev->type != dev_type)
151 score |= 2;
152 if (score == 0)
153 return t;
154
155 if (score < cand_score) {
156 cand = t;
157 cand_score = score;
158 }
159 }
160
161 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
162 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
163 key != t->parms.i_key ||
164 !(t->dev->flags & IFF_UP))
165 continue;
166
167 if (t->dev->type != ARPHRD_IP6GRE &&
168 t->dev->type != dev_type)
169 continue;
170
171 score = 0;
172 if (t->parms.link != link)
173 score |= 1;
174 if (t->dev->type != dev_type)
175 score |= 2;
176 if (score == 0)
177 return t;
178
179 if (score < cand_score) {
180 cand = t;
181 cand_score = score;
182 }
183 }
184
185 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
186 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
187 (!ipv6_addr_equal(local, &t->parms.raddr) ||
188 !ipv6_addr_is_multicast(local))) ||
189 key != t->parms.i_key ||
190 !(t->dev->flags & IFF_UP))
191 continue;
192
193 if (t->dev->type != ARPHRD_IP6GRE &&
194 t->dev->type != dev_type)
195 continue;
196
197 score = 0;
198 if (t->parms.link != link)
199 score |= 1;
200 if (t->dev->type != dev_type)
201 score |= 2;
202 if (score == 0)
203 return t;
204
205 if (score < cand_score) {
206 cand = t;
207 cand_score = score;
208 }
209 }
210
211 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
212 if (t->parms.i_key != key ||
213 !(t->dev->flags & IFF_UP))
214 continue;
215
216 if (t->dev->type != ARPHRD_IP6GRE &&
217 t->dev->type != dev_type)
218 continue;
219
220 score = 0;
221 if (t->parms.link != link)
222 score |= 1;
223 if (t->dev->type != dev_type)
224 score |= 2;
225 if (score == 0)
226 return t;
227
228 if (score < cand_score) {
229 cand = t;
230 cand_score = score;
231 }
232 }
233
234 if (cand)
235 return cand;
236
237 if (gre_proto == htons(ETH_P_ERSPAN) ||
238 gre_proto == htons(ETH_P_ERSPAN2))
239 t = rcu_dereference(ign->collect_md_tun_erspan);
240 else
241 t = rcu_dereference(ign->collect_md_tun);
242
243 if (t && t->dev->flags & IFF_UP)
244 return t;
245
246 dev = ign->fb_tunnel_dev;
247 if (dev && dev->flags & IFF_UP)
248 return netdev_priv(dev);
249
250 return NULL;
251}
252
253static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
254 const struct __ip6_tnl_parm *p)
255{
256 const struct in6_addr *remote = &p->raddr;
257 const struct in6_addr *local = &p->laddr;
258 unsigned int h = HASH_KEY(p->i_key);
259 int prio = 0;
260
261 if (!ipv6_addr_any(local))
262 prio |= 1;
263 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
264 prio |= 2;
265 h ^= HASH_ADDR(remote);
266 }
267
268 return &ign->tunnels[prio][h];
269}
270
271static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
272{
273 if (t->parms.collect_md)
274 rcu_assign_pointer(ign->collect_md_tun, t);
275}
276
277static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
278{
279 if (t->parms.collect_md)
280 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
281}
282
283static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
284{
285 if (t->parms.collect_md)
286 rcu_assign_pointer(ign->collect_md_tun, NULL);
287}
288
289static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
290 struct ip6_tnl *t)
291{
292 if (t->parms.collect_md)
293 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
294}
295
296static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
297 const struct ip6_tnl *t)
298{
299 return __ip6gre_bucket(ign, &t->parms);
300}
301
302static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
303{
304 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
305
306 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
307 rcu_assign_pointer(*tp, t);
308}
309
310static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
311{
312 struct ip6_tnl __rcu **tp;
313 struct ip6_tnl *iter;
314
315 for (tp = ip6gre_bucket(ign, t);
316 (iter = rtnl_dereference(*tp)) != NULL;
317 tp = &iter->next) {
318 if (t == iter) {
319 rcu_assign_pointer(*tp, t->next);
320 break;
321 }
322 }
323}
324
325static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
326 const struct __ip6_tnl_parm *parms,
327 int type)
328{
329 const struct in6_addr *remote = &parms->raddr;
330 const struct in6_addr *local = &parms->laddr;
331 __be32 key = parms->i_key;
332 int link = parms->link;
333 struct ip6_tnl *t;
334 struct ip6_tnl __rcu **tp;
335 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
336
337 for (tp = __ip6gre_bucket(ign, parms);
338 (t = rtnl_dereference(*tp)) != NULL;
339 tp = &t->next)
340 if (ipv6_addr_equal(local, &t->parms.laddr) &&
341 ipv6_addr_equal(remote, &t->parms.raddr) &&
342 key == t->parms.i_key &&
343 link == t->parms.link &&
344 type == t->dev->type)
345 break;
346
347 return t;
348}
349
350static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
351 const struct __ip6_tnl_parm *parms, int create)
352{
353 struct ip6_tnl *t, *nt;
354 struct net_device *dev;
355 char name[IFNAMSIZ];
356 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
357
358 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
359 if (t && create)
360 return NULL;
361 if (t || !create)
362 return t;
363
364 if (parms->name[0]) {
365 if (!dev_valid_name(parms->name))
366 return NULL;
367 strlcpy(name, parms->name, IFNAMSIZ);
368 } else {
369 strcpy(name, "ip6gre%d");
370 }
371 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
372 ip6gre_tunnel_setup);
373 if (!dev)
374 return NULL;
375
376 dev_net_set(dev, net);
377
378 nt = netdev_priv(dev);
379 nt->parms = *parms;
380 dev->rtnl_link_ops = &ip6gre_link_ops;
381
382 nt->dev = dev;
383 nt->net = dev_net(dev);
384
385 if (register_netdevice(dev) < 0)
386 goto failed_free;
387
388 ip6gre_tnl_link_config(nt, 1);
389
390 /* Can use a lockless transmit, unless we generate output sequences */
391 if (!(nt->parms.o_flags & TUNNEL_SEQ))
392 dev->features |= NETIF_F_LLTX;
393
394 dev_hold(dev);
395 ip6gre_tunnel_link(ign, nt);
396 return nt;
397
398failed_free:
399 free_netdev(dev);
400 return NULL;
401}
402
403static void ip6erspan_tunnel_uninit(struct net_device *dev)
404{
405 struct ip6_tnl *t = netdev_priv(dev);
406 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
407
408 ip6erspan_tunnel_unlink_md(ign, t);
409 ip6gre_tunnel_unlink(ign, t);
410 dst_cache_reset(&t->dst_cache);
411 dev_put(dev);
412}
413
414static void ip6gre_tunnel_uninit(struct net_device *dev)
415{
416 struct ip6_tnl *t = netdev_priv(dev);
417 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
418
419 ip6gre_tunnel_unlink_md(ign, t);
420 ip6gre_tunnel_unlink(ign, t);
421 dst_cache_reset(&t->dst_cache);
422 dev_put(dev);
423}
424
425
426static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
427 u8 type, u8 code, int offset, __be32 info)
428{
429 struct net *net = dev_net(skb->dev);
430 const struct gre_base_hdr *greh;
431 const struct ipv6hdr *ipv6h;
432 int grehlen = sizeof(*greh);
433 struct ip6_tnl *t;
434 int key_off = 0;
435 __be16 flags;
436 __be32 key;
437
438 if (!pskb_may_pull(skb, offset + grehlen))
439 return;
440 greh = (const struct gre_base_hdr *)(skb->data + offset);
441 flags = greh->flags;
442 if (flags & (GRE_VERSION | GRE_ROUTING))
443 return;
444 if (flags & GRE_CSUM)
445 grehlen += 4;
446 if (flags & GRE_KEY) {
447 key_off = grehlen + offset;
448 grehlen += 4;
449 }
450
451 if (!pskb_may_pull(skb, offset + grehlen))
452 return;
453 ipv6h = (const struct ipv6hdr *)skb->data;
454 greh = (const struct gre_base_hdr *)(skb->data + offset);
455 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
456
457 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
458 key, greh->protocol);
459 if (!t)
460 return;
461
462 switch (type) {
463 struct ipv6_tlv_tnl_enc_lim *tel;
464 __u32 teli;
465 case ICMPV6_DEST_UNREACH:
466 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
467 t->parms.name);
468 if (code != ICMPV6_PORT_UNREACH)
469 break;
470 return;
471 case ICMPV6_TIME_EXCEED:
472 if (code == ICMPV6_EXC_HOPLIMIT) {
473 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
474 t->parms.name);
475 break;
476 }
477 return;
478 case ICMPV6_PARAMPROB:
479 teli = 0;
480 if (code == ICMPV6_HDR_FIELD)
481 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
482
483 if (teli && teli == be32_to_cpu(info) - 2) {
484 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
485 if (tel->encap_limit == 0) {
486 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
487 t->parms.name);
488 }
489 } else {
490 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
491 t->parms.name);
492 }
493 return;
494 case ICMPV6_PKT_TOOBIG:
495 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
496 return;
497 case NDISC_REDIRECT:
498 ip6_redirect(skb, net, skb->dev->ifindex, 0,
499 sock_net_uid(net, NULL));
500 return;
501 }
502
503 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
504 t->err_count++;
505 else
506 t->err_count = 1;
507 t->err_time = jiffies;
508}
509
510static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
511{
512 const struct ipv6hdr *ipv6h;
513 struct ip6_tnl *tunnel;
514
515 ipv6h = ipv6_hdr(skb);
516 tunnel = ip6gre_tunnel_lookup(skb->dev,
517 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
518 tpi->proto);
519 if (tunnel) {
520 if (tunnel->parms.collect_md) {
521 struct metadata_dst *tun_dst;
522 __be64 tun_id;
523 __be16 flags;
524
525 flags = tpi->flags;
526 tun_id = key32_to_tunnel_id(tpi->key);
527
528 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
529 if (!tun_dst)
530 return PACKET_REJECT;
531
532 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
533 } else {
534 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
535 }
536
537 return PACKET_RCVD;
538 }
539
540 return PACKET_REJECT;
541}
542
543static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
544 struct tnl_ptk_info *tpi)
545{
546 struct erspan_base_hdr *ershdr;
547 struct erspan_metadata *pkt_md;
548 const struct ipv6hdr *ipv6h;
549 struct erspan_md2 *md2;
550 struct ip6_tnl *tunnel;
551 u8 ver;
552
553 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
554 return PACKET_REJECT;
555
556 ipv6h = ipv6_hdr(skb);
557 ershdr = (struct erspan_base_hdr *)skb->data;
558 ver = ershdr->ver;
559 tpi->key = cpu_to_be32(get_session_id(ershdr));
560
561 tunnel = ip6gre_tunnel_lookup(skb->dev,
562 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
563 tpi->proto);
564 if (tunnel) {
565 int len = erspan_hdr_len(ver);
566
567 if (unlikely(!pskb_may_pull(skb, len)))
568 return PACKET_REJECT;
569
570 ershdr = (struct erspan_base_hdr *)skb->data;
571 pkt_md = (struct erspan_metadata *)(ershdr + 1);
572
573 if (__iptunnel_pull_header(skb, len,
574 htons(ETH_P_TEB),
575 false, false) < 0)
576 return PACKET_REJECT;
577
578 if (tunnel->parms.collect_md) {
579 struct metadata_dst *tun_dst;
580 struct ip_tunnel_info *info;
581 struct erspan_metadata *md;
582 __be64 tun_id;
583 __be16 flags;
584
585 tpi->flags |= TUNNEL_KEY;
586 flags = tpi->flags;
587 tun_id = key32_to_tunnel_id(tpi->key);
588
589 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
590 sizeof(*md));
591 if (!tun_dst)
592 return PACKET_REJECT;
593
594 info = &tun_dst->u.tun_info;
595 md = ip_tunnel_info_opts(info);
596 md->version = ver;
597 md2 = &md->u.md2;
598 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
599 ERSPAN_V2_MDSIZE);
600 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
601 info->options_len = sizeof(*md);
602
603 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
604
605 } else {
606 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
607 }
608
609 return PACKET_RCVD;
610 }
611
612 return PACKET_REJECT;
613}
614
615static int gre_rcv(struct sk_buff *skb)
616{
617 struct tnl_ptk_info tpi;
618 bool csum_err = false;
619 int hdr_len;
620
621 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
622 if (hdr_len < 0)
623 goto drop;
624
625 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
626 goto drop;
627
628 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
629 tpi.proto == htons(ETH_P_ERSPAN2))) {
630 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
631 return 0;
632 goto out;
633 }
634
635 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
636 return 0;
637
638out:
639 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
640drop:
641 kfree_skb(skb);
642 return 0;
643}
644
645static int gre_handle_offloads(struct sk_buff *skb, bool csum)
646{
647 return iptunnel_handle_offloads(skb,
648 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
649}
650
651static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
652 struct net_device *dev,
653 struct flowi6 *fl6, __u8 *dsfield,
654 int *encap_limit)
655{
656 const struct iphdr *iph = ip_hdr(skb);
657 struct ip6_tnl *t = netdev_priv(dev);
658
659 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
660 *encap_limit = t->parms.encap_limit;
661
662 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
663
664 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
665 *dsfield = ipv4_get_dsfield(iph);
666 else
667 *dsfield = ip6_tclass(t->parms.flowinfo);
668
669 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
670 fl6->flowi6_mark = skb->mark;
671 else
672 fl6->flowi6_mark = t->parms.fwmark;
673
674 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
675}
676
677static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
678 struct net_device *dev,
679 struct flowi6 *fl6, __u8 *dsfield,
680 int *encap_limit)
681{
682 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
683 struct ip6_tnl *t = netdev_priv(dev);
684 __u16 offset;
685
686 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
687 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
688
689 if (offset > 0) {
690 struct ipv6_tlv_tnl_enc_lim *tel;
691
692 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
693 if (tel->encap_limit == 0) {
694 icmpv6_send(skb, ICMPV6_PARAMPROB,
695 ICMPV6_HDR_FIELD, offset + 2);
696 return -1;
697 }
698 *encap_limit = tel->encap_limit - 1;
699 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
700 *encap_limit = t->parms.encap_limit;
701 }
702
703 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
704
705 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
706 *dsfield = ipv6_get_dsfield(ipv6h);
707 else
708 *dsfield = ip6_tclass(t->parms.flowinfo);
709
710 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
711 fl6->flowlabel |= ip6_flowlabel(ipv6h);
712
713 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
714 fl6->flowi6_mark = skb->mark;
715 else
716 fl6->flowi6_mark = t->parms.fwmark;
717
718 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
719
720 return 0;
721}
722
723static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
724 struct net_device *dev, __u8 dsfield,
725 struct flowi6 *fl6, int encap_limit,
726 __u32 *pmtu, __be16 proto)
727{
728 struct ip6_tnl *tunnel = netdev_priv(dev);
729 __be16 protocol;
730
731 if (dev->type == ARPHRD_ETHER)
732 IPCB(skb)->flags = 0;
733
734 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
735 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
736 else
737 fl6->daddr = tunnel->parms.raddr;
738
739 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
740 return -ENOMEM;
741
742 /* Push GRE header. */
743 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
744
745 if (tunnel->parms.collect_md) {
746 struct ip_tunnel_info *tun_info;
747 const struct ip_tunnel_key *key;
748 __be16 flags;
749
750 tun_info = skb_tunnel_info(skb);
751 if (unlikely(!tun_info ||
752 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
753 ip_tunnel_info_af(tun_info) != AF_INET6))
754 return -EINVAL;
755
756 key = &tun_info->key;
757 memset(fl6, 0, sizeof(*fl6));
758 fl6->flowi6_proto = IPPROTO_GRE;
759 fl6->daddr = key->u.ipv6.dst;
760 fl6->flowlabel = key->label;
761 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
762
763 dsfield = key->tos;
764 flags = key->tun_flags &
765 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
766 tunnel->tun_hlen = gre_calc_hlen(flags);
767
768 gre_build_header(skb, tunnel->tun_hlen,
769 flags, protocol,
770 tunnel_id_to_key32(tun_info->key.tun_id),
771 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
772 : 0);
773
774 } else {
775 if (tunnel->parms.o_flags & TUNNEL_SEQ)
776 tunnel->o_seqno++;
777
778 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
779 protocol, tunnel->parms.o_key,
780 htonl(tunnel->o_seqno));
781 }
782
783 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
784 NEXTHDR_GRE);
785}
786
787static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
788{
789 struct ip6_tnl *t = netdev_priv(dev);
790 int encap_limit = -1;
791 struct flowi6 fl6;
792 __u8 dsfield = 0;
793 __u32 mtu;
794 int err;
795
796 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
797
798 if (!t->parms.collect_md)
799 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
800 &dsfield, &encap_limit);
801
802 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
803 if (err)
804 return -1;
805
806 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
807 skb->protocol);
808 if (err != 0) {
809 /* XXX: send ICMP error even if DF is not set. */
810 if (err == -EMSGSIZE)
811 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
812 htonl(mtu));
813 return -1;
814 }
815
816 return 0;
817}
818
819static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
820{
821 struct ip6_tnl *t = netdev_priv(dev);
822 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
823 int encap_limit = -1;
824 struct flowi6 fl6;
825 __u8 dsfield = 0;
826 __u32 mtu;
827 int err;
828
829 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
830 return -1;
831
832 if (!t->parms.collect_md &&
833 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
834 return -1;
835
836 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
837 return -1;
838
839 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
840 &mtu, skb->protocol);
841 if (err != 0) {
842 if (err == -EMSGSIZE)
843 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
844 return -1;
845 }
846
847 return 0;
848}
849
850/**
851 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
852 * @t: the outgoing tunnel device
853 * @hdr: IPv6 header from the incoming packet
854 *
855 * Description:
856 * Avoid trivial tunneling loop by checking that tunnel exit-point
857 * doesn't match source of incoming packet.
858 *
859 * Return:
860 * 1 if conflict,
861 * 0 else
862 **/
863
864static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
865 const struct ipv6hdr *hdr)
866{
867 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
868}
869
870static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
871{
872 struct ip6_tnl *t = netdev_priv(dev);
873 int encap_limit = -1;
874 struct flowi6 fl6;
875 __u32 mtu;
876 int err;
877
878 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
879 encap_limit = t->parms.encap_limit;
880
881 if (!t->parms.collect_md)
882 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
883
884 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
885 if (err)
886 return err;
887
888 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
889
890 return err;
891}
892
893static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
894 struct net_device *dev)
895{
896 struct ip6_tnl *t = netdev_priv(dev);
897 struct net_device_stats *stats = &t->dev->stats;
898 int ret;
899
900 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
901 goto tx_err;
902
903 switch (skb->protocol) {
904 case htons(ETH_P_IP):
905 ret = ip6gre_xmit_ipv4(skb, dev);
906 break;
907 case htons(ETH_P_IPV6):
908 ret = ip6gre_xmit_ipv6(skb, dev);
909 break;
910 default:
911 ret = ip6gre_xmit_other(skb, dev);
912 break;
913 }
914
915 if (ret < 0)
916 goto tx_err;
917
918 return NETDEV_TX_OK;
919
920tx_err:
921 stats->tx_errors++;
922 stats->tx_dropped++;
923 kfree_skb(skb);
924 return NETDEV_TX_OK;
925}
926
927static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
928 struct net_device *dev)
929{
930 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
931 struct ip6_tnl *t = netdev_priv(dev);
932 struct dst_entry *dst = skb_dst(skb);
933 struct net_device_stats *stats;
934 bool truncate = false;
935 int encap_limit = -1;
936 __u8 dsfield = false;
937 struct flowi6 fl6;
938 int err = -EINVAL;
939 __u32 mtu;
940
941 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
942 goto tx_err;
943
944 if (gre_handle_offloads(skb, false))
945 goto tx_err;
946
947 if (skb->len > dev->mtu + dev->hard_header_len) {
948 pskb_trim(skb, dev->mtu + dev->hard_header_len);
949 truncate = true;
950 }
951
952 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
953 goto tx_err;
954
955 t->parms.o_flags &= ~TUNNEL_KEY;
956 IPCB(skb)->flags = 0;
957
958 /* For collect_md mode, derive fl6 from the tunnel key,
959 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
960 */
961 if (t->parms.collect_md) {
962 struct ip_tunnel_info *tun_info;
963 const struct ip_tunnel_key *key;
964 struct erspan_metadata *md;
965 __be32 tun_id;
966
967 tun_info = skb_tunnel_info(skb);
968 if (unlikely(!tun_info ||
969 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
970 ip_tunnel_info_af(tun_info) != AF_INET6))
971 return -EINVAL;
972
973 key = &tun_info->key;
974 memset(&fl6, 0, sizeof(fl6));
975 fl6.flowi6_proto = IPPROTO_GRE;
976 fl6.daddr = key->u.ipv6.dst;
977 fl6.flowlabel = key->label;
978 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
979
980 dsfield = key->tos;
981 md = ip_tunnel_info_opts(tun_info);
982 if (!md)
983 goto tx_err;
984
985 tun_id = tunnel_id_to_key32(key->tun_id);
986 if (md->version == 1) {
987 erspan_build_header(skb,
988 ntohl(tun_id),
989 ntohl(md->u.index), truncate,
990 false);
991 } else if (md->version == 2) {
992 erspan_build_header_v2(skb,
993 ntohl(tun_id),
994 md->u.md2.dir,
995 get_hwid(&md->u.md2),
996 truncate, false);
997 } else {
998 goto tx_err;
999 }
1000 } else {
1001 switch (skb->protocol) {
1002 case htons(ETH_P_IP):
1003 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1004 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1005 &dsfield, &encap_limit);
1006 break;
1007 case htons(ETH_P_IPV6):
1008 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
1009 goto tx_err;
1010 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1011 &dsfield, &encap_limit))
1012 goto tx_err;
1013 break;
1014 default:
1015 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1016 break;
1017 }
1018
1019 if (t->parms.erspan_ver == 1)
1020 erspan_build_header(skb, ntohl(t->parms.o_key),
1021 t->parms.index,
1022 truncate, false);
1023 else if (t->parms.erspan_ver == 2)
1024 erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1025 t->parms.dir,
1026 t->parms.hwid,
1027 truncate, false);
1028 else
1029 goto tx_err;
1030
1031 fl6.daddr = t->parms.raddr;
1032 }
1033
1034 /* Push GRE header. */
1035 gre_build_header(skb, 8, TUNNEL_SEQ,
1036 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1037
1038 /* TooBig packet may have updated dst->dev's mtu */
1039 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1040 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1041
1042 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1043 NEXTHDR_GRE);
1044 if (err != 0) {
1045 /* XXX: send ICMP error even if DF is not set. */
1046 if (err == -EMSGSIZE) {
1047 if (skb->protocol == htons(ETH_P_IP))
1048 icmp_send(skb, ICMP_DEST_UNREACH,
1049 ICMP_FRAG_NEEDED, htonl(mtu));
1050 else
1051 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1052 }
1053
1054 goto tx_err;
1055 }
1056 return NETDEV_TX_OK;
1057
1058tx_err:
1059 stats = &t->dev->stats;
1060 stats->tx_errors++;
1061 stats->tx_dropped++;
1062 kfree_skb(skb);
1063 return NETDEV_TX_OK;
1064}
1065
1066static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1067{
1068 struct net_device *dev = t->dev;
1069 struct __ip6_tnl_parm *p = &t->parms;
1070 struct flowi6 *fl6 = &t->fl.u.ip6;
1071
1072 if (dev->type != ARPHRD_ETHER) {
1073 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1074 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1075 }
1076
1077 /* Set up flowi template */
1078 fl6->saddr = p->laddr;
1079 fl6->daddr = p->raddr;
1080 fl6->flowi6_oif = p->link;
1081 fl6->flowlabel = 0;
1082 fl6->flowi6_proto = IPPROTO_GRE;
1083
1084 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1085 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1086 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1087 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1088
1089 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1090 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1091
1092 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1093 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1094 dev->flags |= IFF_POINTOPOINT;
1095 else
1096 dev->flags &= ~IFF_POINTOPOINT;
1097}
1098
1099static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1100 int t_hlen)
1101{
1102 const struct __ip6_tnl_parm *p = &t->parms;
1103 struct net_device *dev = t->dev;
1104
1105 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1106 int strict = (ipv6_addr_type(&p->raddr) &
1107 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1108
1109 struct rt6_info *rt = rt6_lookup(t->net,
1110 &p->raddr, &p->laddr,
1111 p->link, NULL, strict);
1112
1113 if (!rt)
1114 return;
1115
1116 if (rt->dst.dev) {
1117 dev->hard_header_len = rt->dst.dev->hard_header_len +
1118 t_hlen;
1119
1120 if (set_mtu) {
1121 dev->mtu = rt->dst.dev->mtu - t_hlen;
1122 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1123 dev->mtu -= 8;
1124 if (dev->type == ARPHRD_ETHER)
1125 dev->mtu -= ETH_HLEN;
1126
1127 if (dev->mtu < IPV6_MIN_MTU)
1128 dev->mtu = IPV6_MIN_MTU;
1129 }
1130 }
1131 ip6_rt_put(rt);
1132 }
1133}
1134
1135static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1136{
1137 int t_hlen;
1138
1139 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1140 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1141
1142 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1143 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1144 return t_hlen;
1145}
1146
1147static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1148{
1149 ip6gre_tnl_link_config_common(t);
1150 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1151}
1152
1153static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1154 const struct __ip6_tnl_parm *p)
1155{
1156 t->parms.laddr = p->laddr;
1157 t->parms.raddr = p->raddr;
1158 t->parms.flags = p->flags;
1159 t->parms.hop_limit = p->hop_limit;
1160 t->parms.encap_limit = p->encap_limit;
1161 t->parms.flowinfo = p->flowinfo;
1162 t->parms.link = p->link;
1163 t->parms.proto = p->proto;
1164 t->parms.i_key = p->i_key;
1165 t->parms.o_key = p->o_key;
1166 t->parms.i_flags = p->i_flags;
1167 t->parms.o_flags = p->o_flags;
1168 t->parms.fwmark = p->fwmark;
1169 dst_cache_reset(&t->dst_cache);
1170}
1171
1172static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1173 int set_mtu)
1174{
1175 ip6gre_tnl_copy_tnl_parm(t, p);
1176 ip6gre_tnl_link_config(t, set_mtu);
1177 return 0;
1178}
1179
1180static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1181 const struct ip6_tnl_parm2 *u)
1182{
1183 p->laddr = u->laddr;
1184 p->raddr = u->raddr;
1185 p->flags = u->flags;
1186 p->hop_limit = u->hop_limit;
1187 p->encap_limit = u->encap_limit;
1188 p->flowinfo = u->flowinfo;
1189 p->link = u->link;
1190 p->i_key = u->i_key;
1191 p->o_key = u->o_key;
1192 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1193 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1194 memcpy(p->name, u->name, sizeof(u->name));
1195}
1196
1197static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1198 const struct __ip6_tnl_parm *p)
1199{
1200 u->proto = IPPROTO_GRE;
1201 u->laddr = p->laddr;
1202 u->raddr = p->raddr;
1203 u->flags = p->flags;
1204 u->hop_limit = p->hop_limit;
1205 u->encap_limit = p->encap_limit;
1206 u->flowinfo = p->flowinfo;
1207 u->link = p->link;
1208 u->i_key = p->i_key;
1209 u->o_key = p->o_key;
1210 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1211 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1212 memcpy(u->name, p->name, sizeof(u->name));
1213}
1214
1215static int ip6gre_tunnel_ioctl(struct net_device *dev,
1216 struct ifreq *ifr, int cmd)
1217{
1218 int err = 0;
1219 struct ip6_tnl_parm2 p;
1220 struct __ip6_tnl_parm p1;
1221 struct ip6_tnl *t = netdev_priv(dev);
1222 struct net *net = t->net;
1223 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1224
1225 memset(&p1, 0, sizeof(p1));
1226
1227 switch (cmd) {
1228 case SIOCGETTUNNEL:
1229 if (dev == ign->fb_tunnel_dev) {
1230 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1231 err = -EFAULT;
1232 break;
1233 }
1234 ip6gre_tnl_parm_from_user(&p1, &p);
1235 t = ip6gre_tunnel_locate(net, &p1, 0);
1236 if (!t)
1237 t = netdev_priv(dev);
1238 }
1239 memset(&p, 0, sizeof(p));
1240 ip6gre_tnl_parm_to_user(&p, &t->parms);
1241 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1242 err = -EFAULT;
1243 break;
1244
1245 case SIOCADDTUNNEL:
1246 case SIOCCHGTUNNEL:
1247 err = -EPERM;
1248 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1249 goto done;
1250
1251 err = -EFAULT;
1252 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1253 goto done;
1254
1255 err = -EINVAL;
1256 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1257 goto done;
1258
1259 if (!(p.i_flags&GRE_KEY))
1260 p.i_key = 0;
1261 if (!(p.o_flags&GRE_KEY))
1262 p.o_key = 0;
1263
1264 ip6gre_tnl_parm_from_user(&p1, &p);
1265 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1266
1267 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1268 if (t) {
1269 if (t->dev != dev) {
1270 err = -EEXIST;
1271 break;
1272 }
1273 } else {
1274 t = netdev_priv(dev);
1275
1276 ip6gre_tunnel_unlink(ign, t);
1277 synchronize_net();
1278 ip6gre_tnl_change(t, &p1, 1);
1279 ip6gre_tunnel_link(ign, t);
1280 netdev_state_change(dev);
1281 }
1282 }
1283
1284 if (t) {
1285 err = 0;
1286
1287 memset(&p, 0, sizeof(p));
1288 ip6gre_tnl_parm_to_user(&p, &t->parms);
1289 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1290 err = -EFAULT;
1291 } else
1292 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1293 break;
1294
1295 case SIOCDELTUNNEL:
1296 err = -EPERM;
1297 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1298 goto done;
1299
1300 if (dev == ign->fb_tunnel_dev) {
1301 err = -EFAULT;
1302 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1303 goto done;
1304 err = -ENOENT;
1305 ip6gre_tnl_parm_from_user(&p1, &p);
1306 t = ip6gre_tunnel_locate(net, &p1, 0);
1307 if (!t)
1308 goto done;
1309 err = -EPERM;
1310 if (t == netdev_priv(ign->fb_tunnel_dev))
1311 goto done;
1312 dev = t->dev;
1313 }
1314 unregister_netdevice(dev);
1315 err = 0;
1316 break;
1317
1318 default:
1319 err = -EINVAL;
1320 }
1321
1322done:
1323 return err;
1324}
1325
1326static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1327 unsigned short type, const void *daddr,
1328 const void *saddr, unsigned int len)
1329{
1330 struct ip6_tnl *t = netdev_priv(dev);
1331 struct ipv6hdr *ipv6h;
1332 __be16 *p;
1333
1334 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1335 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1336 t->fl.u.ip6.flowlabel,
1337 true, &t->fl.u.ip6));
1338 ipv6h->hop_limit = t->parms.hop_limit;
1339 ipv6h->nexthdr = NEXTHDR_GRE;
1340 ipv6h->saddr = t->parms.laddr;
1341 ipv6h->daddr = t->parms.raddr;
1342
1343 p = (__be16 *)(ipv6h + 1);
1344 p[0] = t->parms.o_flags;
1345 p[1] = htons(type);
1346
1347 /*
1348 * Set the source hardware address.
1349 */
1350
1351 if (saddr)
1352 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1353 if (daddr)
1354 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1355 if (!ipv6_addr_any(&ipv6h->daddr))
1356 return t->hlen;
1357
1358 return -t->hlen;
1359}
1360
1361static const struct header_ops ip6gre_header_ops = {
1362 .create = ip6gre_header,
1363};
1364
1365static const struct net_device_ops ip6gre_netdev_ops = {
1366 .ndo_init = ip6gre_tunnel_init,
1367 .ndo_uninit = ip6gre_tunnel_uninit,
1368 .ndo_start_xmit = ip6gre_tunnel_xmit,
1369 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1370 .ndo_change_mtu = ip6_tnl_change_mtu,
1371 .ndo_get_stats64 = ip_tunnel_get_stats64,
1372 .ndo_get_iflink = ip6_tnl_get_iflink,
1373};
1374
1375static void ip6gre_dev_free(struct net_device *dev)
1376{
1377 struct ip6_tnl *t = netdev_priv(dev);
1378
1379 dst_cache_destroy(&t->dst_cache);
1380 free_percpu(dev->tstats);
1381}
1382
1383static void ip6gre_tunnel_setup(struct net_device *dev)
1384{
1385 dev->netdev_ops = &ip6gre_netdev_ops;
1386 dev->needs_free_netdev = true;
1387 dev->priv_destructor = ip6gre_dev_free;
1388
1389 dev->type = ARPHRD_IP6GRE;
1390
1391 dev->flags |= IFF_NOARP;
1392 dev->addr_len = sizeof(struct in6_addr);
1393 netif_keep_dst(dev);
1394 /* This perm addr will be used as interface identifier by IPv6 */
1395 dev->addr_assign_type = NET_ADDR_RANDOM;
1396 eth_random_addr(dev->perm_addr);
1397}
1398
1399#define GRE6_FEATURES (NETIF_F_SG | \
1400 NETIF_F_FRAGLIST | \
1401 NETIF_F_HIGHDMA | \
1402 NETIF_F_HW_CSUM)
1403
1404static void ip6gre_tnl_init_features(struct net_device *dev)
1405{
1406 struct ip6_tnl *nt = netdev_priv(dev);
1407
1408 dev->features |= GRE6_FEATURES;
1409 dev->hw_features |= GRE6_FEATURES;
1410
1411 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1412 /* TCP offload with GRE SEQ is not supported, nor
1413 * can we support 2 levels of outer headers requiring
1414 * an update.
1415 */
1416 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1417 nt->encap.type == TUNNEL_ENCAP_NONE) {
1418 dev->features |= NETIF_F_GSO_SOFTWARE;
1419 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1420 }
1421
1422 /* Can use a lockless transmit, unless we generate
1423 * output sequences
1424 */
1425 dev->features |= NETIF_F_LLTX;
1426 }
1427}
1428
1429static int ip6gre_tunnel_init_common(struct net_device *dev)
1430{
1431 struct ip6_tnl *tunnel;
1432 int ret;
1433 int t_hlen;
1434
1435 tunnel = netdev_priv(dev);
1436
1437 tunnel->dev = dev;
1438 tunnel->net = dev_net(dev);
1439 strcpy(tunnel->parms.name, dev->name);
1440
1441 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1442 if (!dev->tstats)
1443 return -ENOMEM;
1444
1445 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1446 if (ret) {
1447 free_percpu(dev->tstats);
1448 dev->tstats = NULL;
1449 return ret;
1450 }
1451
1452 t_hlen = ip6gre_calc_hlen(tunnel);
1453 dev->mtu = ETH_DATA_LEN - t_hlen;
1454 if (dev->type == ARPHRD_ETHER)
1455 dev->mtu -= ETH_HLEN;
1456 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1457 dev->mtu -= 8;
1458
1459 if (tunnel->parms.collect_md) {
1460 dev->features |= NETIF_F_NETNS_LOCAL;
1461 netif_keep_dst(dev);
1462 }
1463 ip6gre_tnl_init_features(dev);
1464
1465 return 0;
1466}
1467
1468static int ip6gre_tunnel_init(struct net_device *dev)
1469{
1470 struct ip6_tnl *tunnel;
1471 int ret;
1472
1473 ret = ip6gre_tunnel_init_common(dev);
1474 if (ret)
1475 return ret;
1476
1477 tunnel = netdev_priv(dev);
1478
1479 if (tunnel->parms.collect_md)
1480 return 0;
1481
1482 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1483 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1484
1485 if (ipv6_addr_any(&tunnel->parms.raddr))
1486 dev->header_ops = &ip6gre_header_ops;
1487
1488 return 0;
1489}
1490
1491static void ip6gre_fb_tunnel_init(struct net_device *dev)
1492{
1493 struct ip6_tnl *tunnel = netdev_priv(dev);
1494
1495 tunnel->dev = dev;
1496 tunnel->net = dev_net(dev);
1497 strcpy(tunnel->parms.name, dev->name);
1498
1499 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1500
1501 dev_hold(dev);
1502}
1503
1504static struct inet6_protocol ip6gre_protocol __read_mostly = {
1505 .handler = gre_rcv,
1506 .err_handler = ip6gre_err,
1507 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1508};
1509
1510static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1511{
1512 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1513 struct net_device *dev, *aux;
1514 int prio;
1515
1516 for_each_netdev_safe(net, dev, aux)
1517 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1518 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1519 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1520 unregister_netdevice_queue(dev, head);
1521
1522 for (prio = 0; prio < 4; prio++) {
1523 int h;
1524 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1525 struct ip6_tnl *t;
1526
1527 t = rtnl_dereference(ign->tunnels[prio][h]);
1528
1529 while (t) {
1530 /* If dev is in the same netns, it has already
1531 * been added to the list by the previous loop.
1532 */
1533 if (!net_eq(dev_net(t->dev), net))
1534 unregister_netdevice_queue(t->dev,
1535 head);
1536 t = rtnl_dereference(t->next);
1537 }
1538 }
1539 }
1540}
1541
1542static int __net_init ip6gre_init_net(struct net *net)
1543{
1544 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1545 int err;
1546
1547 if (!net_has_fallback_tunnels(net))
1548 return 0;
1549 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1550 NET_NAME_UNKNOWN,
1551 ip6gre_tunnel_setup);
1552 if (!ign->fb_tunnel_dev) {
1553 err = -ENOMEM;
1554 goto err_alloc_dev;
1555 }
1556 dev_net_set(ign->fb_tunnel_dev, net);
1557 /* FB netdevice is special: we have one, and only one per netns.
1558 * Allowing to move it to another netns is clearly unsafe.
1559 */
1560 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1561
1562
1563 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1564 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1565
1566 err = register_netdev(ign->fb_tunnel_dev);
1567 if (err)
1568 goto err_reg_dev;
1569
1570 rcu_assign_pointer(ign->tunnels_wc[0],
1571 netdev_priv(ign->fb_tunnel_dev));
1572 return 0;
1573
1574err_reg_dev:
1575 free_netdev(ign->fb_tunnel_dev);
1576err_alloc_dev:
1577 return err;
1578}
1579
1580static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1581{
1582 struct net *net;
1583 LIST_HEAD(list);
1584
1585 rtnl_lock();
1586 list_for_each_entry(net, net_list, exit_list)
1587 ip6gre_destroy_tunnels(net, &list);
1588 unregister_netdevice_many(&list);
1589 rtnl_unlock();
1590}
1591
1592static struct pernet_operations ip6gre_net_ops = {
1593 .init = ip6gre_init_net,
1594 .exit_batch = ip6gre_exit_batch_net,
1595 .id = &ip6gre_net_id,
1596 .size = sizeof(struct ip6gre_net),
1597};
1598
1599static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1600 struct netlink_ext_ack *extack)
1601{
1602 __be16 flags;
1603
1604 if (!data)
1605 return 0;
1606
1607 flags = 0;
1608 if (data[IFLA_GRE_IFLAGS])
1609 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1610 if (data[IFLA_GRE_OFLAGS])
1611 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1612 if (flags & (GRE_VERSION|GRE_ROUTING))
1613 return -EINVAL;
1614
1615 return 0;
1616}
1617
1618static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1619 struct netlink_ext_ack *extack)
1620{
1621 struct in6_addr daddr;
1622
1623 if (tb[IFLA_ADDRESS]) {
1624 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1625 return -EINVAL;
1626 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1627 return -EADDRNOTAVAIL;
1628 }
1629
1630 if (!data)
1631 goto out;
1632
1633 if (data[IFLA_GRE_REMOTE]) {
1634 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1635 if (ipv6_addr_any(&daddr))
1636 return -EINVAL;
1637 }
1638
1639out:
1640 return ip6gre_tunnel_validate(tb, data, extack);
1641}
1642
1643static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1644 struct netlink_ext_ack *extack)
1645{
1646 __be16 flags = 0;
1647 int ret, ver = 0;
1648
1649 if (!data)
1650 return 0;
1651
1652 ret = ip6gre_tap_validate(tb, data, extack);
1653 if (ret)
1654 return ret;
1655
1656 /* ERSPAN should only have GRE sequence and key flag */
1657 if (data[IFLA_GRE_OFLAGS])
1658 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1659 if (data[IFLA_GRE_IFLAGS])
1660 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1661 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1662 flags != (GRE_SEQ | GRE_KEY))
1663 return -EINVAL;
1664
1665 /* ERSPAN Session ID only has 10-bit. Since we reuse
1666 * 32-bit key field as ID, check it's range.
1667 */
1668 if (data[IFLA_GRE_IKEY] &&
1669 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1670 return -EINVAL;
1671
1672 if (data[IFLA_GRE_OKEY] &&
1673 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1674 return -EINVAL;
1675
1676 if (data[IFLA_GRE_ERSPAN_VER]) {
1677 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1678 if (ver != 1 && ver != 2)
1679 return -EINVAL;
1680 }
1681
1682 if (ver == 1) {
1683 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1684 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1685
1686 if (index & ~INDEX_MASK)
1687 return -EINVAL;
1688 }
1689 } else if (ver == 2) {
1690 if (data[IFLA_GRE_ERSPAN_DIR]) {
1691 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1692
1693 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1694 return -EINVAL;
1695 }
1696
1697 if (data[IFLA_GRE_ERSPAN_HWID]) {
1698 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1699
1700 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1701 return -EINVAL;
1702 }
1703 }
1704
1705 return 0;
1706}
1707
1708static void ip6gre_netlink_parms(struct nlattr *data[],
1709 struct __ip6_tnl_parm *parms)
1710{
1711 memset(parms, 0, sizeof(*parms));
1712
1713 if (!data)
1714 return;
1715
1716 if (data[IFLA_GRE_LINK])
1717 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1718
1719 if (data[IFLA_GRE_IFLAGS])
1720 parms->i_flags = gre_flags_to_tnl_flags(
1721 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1722
1723 if (data[IFLA_GRE_OFLAGS])
1724 parms->o_flags = gre_flags_to_tnl_flags(
1725 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1726
1727 if (data[IFLA_GRE_IKEY])
1728 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1729
1730 if (data[IFLA_GRE_OKEY])
1731 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1732
1733 if (data[IFLA_GRE_LOCAL])
1734 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1735
1736 if (data[IFLA_GRE_REMOTE])
1737 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1738
1739 if (data[IFLA_GRE_TTL])
1740 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1741
1742 if (data[IFLA_GRE_ENCAP_LIMIT])
1743 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1744
1745 if (data[IFLA_GRE_FLOWINFO])
1746 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1747
1748 if (data[IFLA_GRE_FLAGS])
1749 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1750
1751 if (data[IFLA_GRE_FWMARK])
1752 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1753
1754 if (data[IFLA_GRE_COLLECT_METADATA])
1755 parms->collect_md = true;
1756
1757 if (data[IFLA_GRE_ERSPAN_VER])
1758 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1759
1760 if (parms->erspan_ver == 1) {
1761 if (data[IFLA_GRE_ERSPAN_INDEX])
1762 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1763 } else if (parms->erspan_ver == 2) {
1764 if (data[IFLA_GRE_ERSPAN_DIR])
1765 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1766 if (data[IFLA_GRE_ERSPAN_HWID])
1767 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1768 }
1769}
1770
1771static int ip6gre_tap_init(struct net_device *dev)
1772{
1773 int ret;
1774
1775 ret = ip6gre_tunnel_init_common(dev);
1776 if (ret)
1777 return ret;
1778
1779 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1780
1781 return 0;
1782}
1783
1784static const struct net_device_ops ip6gre_tap_netdev_ops = {
1785 .ndo_init = ip6gre_tap_init,
1786 .ndo_uninit = ip6gre_tunnel_uninit,
1787 .ndo_start_xmit = ip6gre_tunnel_xmit,
1788 .ndo_set_mac_address = eth_mac_addr,
1789 .ndo_validate_addr = eth_validate_addr,
1790 .ndo_change_mtu = ip6_tnl_change_mtu,
1791 .ndo_get_stats64 = ip_tunnel_get_stats64,
1792 .ndo_get_iflink = ip6_tnl_get_iflink,
1793};
1794
1795static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1796{
1797 int t_hlen;
1798
1799 tunnel->tun_hlen = 8;
1800 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1801 erspan_hdr_len(tunnel->parms.erspan_ver);
1802
1803 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1804 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1805 return t_hlen;
1806}
1807
1808static int ip6erspan_tap_init(struct net_device *dev)
1809{
1810 struct ip6_tnl *tunnel;
1811 int t_hlen;
1812 int ret;
1813
1814 tunnel = netdev_priv(dev);
1815
1816 tunnel->dev = dev;
1817 tunnel->net = dev_net(dev);
1818 strcpy(tunnel->parms.name, dev->name);
1819
1820 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1821 if (!dev->tstats)
1822 return -ENOMEM;
1823
1824 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1825 if (ret) {
1826 free_percpu(dev->tstats);
1827 dev->tstats = NULL;
1828 return ret;
1829 }
1830
1831 t_hlen = ip6erspan_calc_hlen(tunnel);
1832 dev->mtu = ETH_DATA_LEN - t_hlen;
1833 if (dev->type == ARPHRD_ETHER)
1834 dev->mtu -= ETH_HLEN;
1835 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1836 dev->mtu -= 8;
1837
1838 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1839 ip6erspan_tnl_link_config(tunnel, 1);
1840
1841 return 0;
1842}
1843
1844static const struct net_device_ops ip6erspan_netdev_ops = {
1845 .ndo_init = ip6erspan_tap_init,
1846 .ndo_uninit = ip6erspan_tunnel_uninit,
1847 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1848 .ndo_set_mac_address = eth_mac_addr,
1849 .ndo_validate_addr = eth_validate_addr,
1850 .ndo_change_mtu = ip6_tnl_change_mtu,
1851 .ndo_get_stats64 = ip_tunnel_get_stats64,
1852 .ndo_get_iflink = ip6_tnl_get_iflink,
1853};
1854
1855static void ip6gre_tap_setup(struct net_device *dev)
1856{
1857
1858 ether_setup(dev);
1859
1860 dev->max_mtu = 0;
1861 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1862 dev->needs_free_netdev = true;
1863 dev->priv_destructor = ip6gre_dev_free;
1864
1865 dev->features |= NETIF_F_NETNS_LOCAL;
1866 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1867 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1868 netif_keep_dst(dev);
1869}
1870
1871bool is_ip6gretap_dev(const struct net_device *dev)
1872{
1873 return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1874}
1875EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1876
1877static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1878 struct ip_tunnel_encap *ipencap)
1879{
1880 bool ret = false;
1881
1882 memset(ipencap, 0, sizeof(*ipencap));
1883
1884 if (!data)
1885 return ret;
1886
1887 if (data[IFLA_GRE_ENCAP_TYPE]) {
1888 ret = true;
1889 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1890 }
1891
1892 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1893 ret = true;
1894 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1895 }
1896
1897 if (data[IFLA_GRE_ENCAP_SPORT]) {
1898 ret = true;
1899 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1900 }
1901
1902 if (data[IFLA_GRE_ENCAP_DPORT]) {
1903 ret = true;
1904 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1905 }
1906
1907 return ret;
1908}
1909
1910static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1911 struct nlattr *tb[], struct nlattr *data[],
1912 struct netlink_ext_ack *extack)
1913{
1914 struct ip6_tnl *nt;
1915 struct ip_tunnel_encap ipencap;
1916 int err;
1917
1918 nt = netdev_priv(dev);
1919
1920 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1921 int err = ip6_tnl_encap_setup(nt, &ipencap);
1922
1923 if (err < 0)
1924 return err;
1925 }
1926
1927 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1928 eth_hw_addr_random(dev);
1929
1930 nt->dev = dev;
1931 nt->net = dev_net(dev);
1932
1933 err = register_netdevice(dev);
1934 if (err)
1935 goto out;
1936
1937 if (tb[IFLA_MTU])
1938 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1939
1940 dev_hold(dev);
1941
1942out:
1943 return err;
1944}
1945
1946static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1947 struct nlattr *tb[], struct nlattr *data[],
1948 struct netlink_ext_ack *extack)
1949{
1950 struct ip6_tnl *nt = netdev_priv(dev);
1951 struct net *net = dev_net(dev);
1952 struct ip6gre_net *ign;
1953 int err;
1954
1955 ip6gre_netlink_parms(data, &nt->parms);
1956 ign = net_generic(net, ip6gre_net_id);
1957
1958 if (nt->parms.collect_md) {
1959 if (rtnl_dereference(ign->collect_md_tun))
1960 return -EEXIST;
1961 } else {
1962 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1963 return -EEXIST;
1964 }
1965
1966 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1967 if (!err) {
1968 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1969 ip6gre_tunnel_link_md(ign, nt);
1970 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
1971 }
1972 return err;
1973}
1974
1975static struct ip6_tnl *
1976ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
1977 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
1978 struct netlink_ext_ack *extack)
1979{
1980 struct ip6_tnl *t, *nt = netdev_priv(dev);
1981 struct net *net = nt->net;
1982 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1983 struct ip_tunnel_encap ipencap;
1984
1985 if (dev == ign->fb_tunnel_dev)
1986 return ERR_PTR(-EINVAL);
1987
1988 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1989 int err = ip6_tnl_encap_setup(nt, &ipencap);
1990
1991 if (err < 0)
1992 return ERR_PTR(err);
1993 }
1994
1995 ip6gre_netlink_parms(data, p_p);
1996
1997 t = ip6gre_tunnel_locate(net, p_p, 0);
1998
1999 if (t) {
2000 if (t->dev != dev)
2001 return ERR_PTR(-EEXIST);
2002 } else {
2003 t = nt;
2004 }
2005
2006 return t;
2007}
2008
2009static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2010 struct nlattr *data[],
2011 struct netlink_ext_ack *extack)
2012{
2013 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2014 struct __ip6_tnl_parm p;
2015 struct ip6_tnl *t;
2016
2017 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2018 if (IS_ERR(t))
2019 return PTR_ERR(t);
2020
2021 ip6gre_tunnel_unlink_md(ign, t);
2022 ip6gre_tunnel_unlink(ign, t);
2023 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2024 ip6gre_tunnel_link_md(ign, t);
2025 ip6gre_tunnel_link(ign, t);
2026 return 0;
2027}
2028
2029static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2030{
2031 struct net *net = dev_net(dev);
2032 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2033
2034 if (dev != ign->fb_tunnel_dev)
2035 unregister_netdevice_queue(dev, head);
2036}
2037
2038static size_t ip6gre_get_size(const struct net_device *dev)
2039{
2040 return
2041 /* IFLA_GRE_LINK */
2042 nla_total_size(4) +
2043 /* IFLA_GRE_IFLAGS */
2044 nla_total_size(2) +
2045 /* IFLA_GRE_OFLAGS */
2046 nla_total_size(2) +
2047 /* IFLA_GRE_IKEY */
2048 nla_total_size(4) +
2049 /* IFLA_GRE_OKEY */
2050 nla_total_size(4) +
2051 /* IFLA_GRE_LOCAL */
2052 nla_total_size(sizeof(struct in6_addr)) +
2053 /* IFLA_GRE_REMOTE */
2054 nla_total_size(sizeof(struct in6_addr)) +
2055 /* IFLA_GRE_TTL */
2056 nla_total_size(1) +
2057 /* IFLA_GRE_ENCAP_LIMIT */
2058 nla_total_size(1) +
2059 /* IFLA_GRE_FLOWINFO */
2060 nla_total_size(4) +
2061 /* IFLA_GRE_FLAGS */
2062 nla_total_size(4) +
2063 /* IFLA_GRE_ENCAP_TYPE */
2064 nla_total_size(2) +
2065 /* IFLA_GRE_ENCAP_FLAGS */
2066 nla_total_size(2) +
2067 /* IFLA_GRE_ENCAP_SPORT */
2068 nla_total_size(2) +
2069 /* IFLA_GRE_ENCAP_DPORT */
2070 nla_total_size(2) +
2071 /* IFLA_GRE_COLLECT_METADATA */
2072 nla_total_size(0) +
2073 /* IFLA_GRE_FWMARK */
2074 nla_total_size(4) +
2075 /* IFLA_GRE_ERSPAN_INDEX */
2076 nla_total_size(4) +
2077 0;
2078}
2079
2080static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2081{
2082 struct ip6_tnl *t = netdev_priv(dev);
2083 struct __ip6_tnl_parm *p = &t->parms;
2084
2085 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2086 nla_put_be16(skb, IFLA_GRE_IFLAGS,
2087 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2088 nla_put_be16(skb, IFLA_GRE_OFLAGS,
2089 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2090 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2091 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2092 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2093 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2094 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2095 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2096 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2097 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2098 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2099 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2100 goto nla_put_failure;
2101
2102 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2103 t->encap.type) ||
2104 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2105 t->encap.sport) ||
2106 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2107 t->encap.dport) ||
2108 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2109 t->encap.flags))
2110 goto nla_put_failure;
2111
2112 if (p->collect_md) {
2113 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2114 goto nla_put_failure;
2115 }
2116
2117 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2118 goto nla_put_failure;
2119
2120 if (p->erspan_ver == 1) {
2121 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2122 goto nla_put_failure;
2123 } else if (p->erspan_ver == 2) {
2124 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2125 goto nla_put_failure;
2126 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2127 goto nla_put_failure;
2128 }
2129
2130 return 0;
2131
2132nla_put_failure:
2133 return -EMSGSIZE;
2134}
2135
2136static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2137 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2138 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2139 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2140 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2141 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2142 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2143 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2144 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2145 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2146 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2147 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2148 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2149 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2150 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2151 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2152 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2153 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2154 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2155 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2156 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2157 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2158};
2159
2160static void ip6erspan_tap_setup(struct net_device *dev)
2161{
2162 ether_setup(dev);
2163
2164 dev->netdev_ops = &ip6erspan_netdev_ops;
2165 dev->needs_free_netdev = true;
2166 dev->priv_destructor = ip6gre_dev_free;
2167
2168 dev->features |= NETIF_F_NETNS_LOCAL;
2169 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2170 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2171 netif_keep_dst(dev);
2172}
2173
2174static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2175 struct nlattr *tb[], struct nlattr *data[],
2176 struct netlink_ext_ack *extack)
2177{
2178 struct ip6_tnl *nt = netdev_priv(dev);
2179 struct net *net = dev_net(dev);
2180 struct ip6gre_net *ign;
2181 int err;
2182
2183 ip6gre_netlink_parms(data, &nt->parms);
2184 ign = net_generic(net, ip6gre_net_id);
2185
2186 if (nt->parms.collect_md) {
2187 if (rtnl_dereference(ign->collect_md_tun_erspan))
2188 return -EEXIST;
2189 } else {
2190 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2191 return -EEXIST;
2192 }
2193
2194 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2195 if (!err) {
2196 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2197 ip6erspan_tunnel_link_md(ign, nt);
2198 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2199 }
2200 return err;
2201}
2202
2203static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2204{
2205 ip6gre_tnl_link_config_common(t);
2206 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2207}
2208
2209static int ip6erspan_tnl_change(struct ip6_tnl *t,
2210 const struct __ip6_tnl_parm *p, int set_mtu)
2211{
2212 ip6gre_tnl_copy_tnl_parm(t, p);
2213 ip6erspan_tnl_link_config(t, set_mtu);
2214 return 0;
2215}
2216
2217static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2218 struct nlattr *data[],
2219 struct netlink_ext_ack *extack)
2220{
2221 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2222 struct __ip6_tnl_parm p;
2223 struct ip6_tnl *t;
2224
2225 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2226 if (IS_ERR(t))
2227 return PTR_ERR(t);
2228
2229 ip6gre_tunnel_unlink_md(ign, t);
2230 ip6gre_tunnel_unlink(ign, t);
2231 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2232 ip6erspan_tunnel_link_md(ign, t);
2233 ip6gre_tunnel_link(ign, t);
2234 return 0;
2235}
2236
2237static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2238 .kind = "ip6gre",
2239 .maxtype = IFLA_GRE_MAX,
2240 .policy = ip6gre_policy,
2241 .priv_size = sizeof(struct ip6_tnl),
2242 .setup = ip6gre_tunnel_setup,
2243 .validate = ip6gre_tunnel_validate,
2244 .newlink = ip6gre_newlink,
2245 .changelink = ip6gre_changelink,
2246 .dellink = ip6gre_dellink,
2247 .get_size = ip6gre_get_size,
2248 .fill_info = ip6gre_fill_info,
2249 .get_link_net = ip6_tnl_get_link_net,
2250};
2251
2252static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2253 .kind = "ip6gretap",
2254 .maxtype = IFLA_GRE_MAX,
2255 .policy = ip6gre_policy,
2256 .priv_size = sizeof(struct ip6_tnl),
2257 .setup = ip6gre_tap_setup,
2258 .validate = ip6gre_tap_validate,
2259 .newlink = ip6gre_newlink,
2260 .changelink = ip6gre_changelink,
2261 .get_size = ip6gre_get_size,
2262 .fill_info = ip6gre_fill_info,
2263 .get_link_net = ip6_tnl_get_link_net,
2264};
2265
2266static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2267 .kind = "ip6erspan",
2268 .maxtype = IFLA_GRE_MAX,
2269 .policy = ip6gre_policy,
2270 .priv_size = sizeof(struct ip6_tnl),
2271 .setup = ip6erspan_tap_setup,
2272 .validate = ip6erspan_tap_validate,
2273 .newlink = ip6erspan_newlink,
2274 .changelink = ip6erspan_changelink,
2275 .get_size = ip6gre_get_size,
2276 .fill_info = ip6gre_fill_info,
2277 .get_link_net = ip6_tnl_get_link_net,
2278};
2279
2280/*
2281 * And now the modules code and kernel interface.
2282 */
2283
2284static int __init ip6gre_init(void)
2285{
2286 int err;
2287
2288 pr_info("GRE over IPv6 tunneling driver\n");
2289
2290 err = register_pernet_device(&ip6gre_net_ops);
2291 if (err < 0)
2292 return err;
2293
2294 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2295 if (err < 0) {
2296 pr_info("%s: can't add protocol\n", __func__);
2297 goto add_proto_failed;
2298 }
2299
2300 err = rtnl_link_register(&ip6gre_link_ops);
2301 if (err < 0)
2302 goto rtnl_link_failed;
2303
2304 err = rtnl_link_register(&ip6gre_tap_ops);
2305 if (err < 0)
2306 goto tap_ops_failed;
2307
2308 err = rtnl_link_register(&ip6erspan_tap_ops);
2309 if (err < 0)
2310 goto erspan_link_failed;
2311
2312out:
2313 return err;
2314
2315erspan_link_failed:
2316 rtnl_link_unregister(&ip6gre_tap_ops);
2317tap_ops_failed:
2318 rtnl_link_unregister(&ip6gre_link_ops);
2319rtnl_link_failed:
2320 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2321add_proto_failed:
2322 unregister_pernet_device(&ip6gre_net_ops);
2323 goto out;
2324}
2325
2326static void __exit ip6gre_fini(void)
2327{
2328 rtnl_link_unregister(&ip6gre_tap_ops);
2329 rtnl_link_unregister(&ip6gre_link_ops);
2330 rtnl_link_unregister(&ip6erspan_tap_ops);
2331 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2332 unregister_pernet_device(&ip6gre_net_ops);
2333}
2334
2335module_init(ip6gre_init);
2336module_exit(ip6gre_fini);
2337MODULE_LICENSE("GPL");
2338MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2339MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2340MODULE_ALIAS_RTNL_LINK("ip6gre");
2341MODULE_ALIAS_RTNL_LINK("ip6gretap");
2342MODULE_ALIAS_RTNL_LINK("ip6erspan");
2343MODULE_ALIAS_NETDEV("ip6gre0");