1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * GRE over IPv6 protocol decoder.
4 *
5 * Authors: Dmitry Kozlov (xeb@mail.ru)
6 */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
18 #include <linux/in.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/rtnetlink.h>
47
48 #include <net/ipv6.h>
49 #include <net/ip6_fib.h>
50 #include <net/ip6_route.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/gre.h>
53 #include <net/erspan.h>
54 #include <net/dst_metadata.h>
55
56
57 static bool log_ecn_error = true;
58 module_param(log_ecn_error, bool, 0644);
59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
60
61 #define IP6_GRE_HASH_SIZE_SHIFT 5
62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
63
64 static unsigned int ip6gre_net_id __read_mostly;
65 struct ip6gre_net {
66 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
67
68 struct ip6_tnl __rcu *collect_md_tun;
69 struct ip6_tnl __rcu *collect_md_tun_erspan;
70 struct net_device *fb_tunnel_dev;
71 };
72
73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81
82 /* Tunnel hash table */
83
84 /*
85 4 hash tables:
86
87 3: (remote,local)
88 2: (remote,*)
89 1: (*,local)
90 0: (*,*)
91
92 We require exact key match i.e. if a key is present in packet
93 it will match only tunnel with the same key; if it is not present,
94 it will match only keyless tunnel.
95
96 All keysless packets, if not matched configured keyless tunnels
97 will match fallback tunnel.
98 */
99
100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
101 static u32 HASH_ADDR(const struct in6_addr *addr)
102 {
103 u32 hash = ipv6_addr_hash(addr);
104
105 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
106 }
107
108 #define tunnels_r_l tunnels[3]
109 #define tunnels_r tunnels[2]
110 #define tunnels_l tunnels[1]
111 #define tunnels_wc tunnels[0]
112
113 /* Given src, dst and key, find appropriate for input tunnel. */
114
115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
116 const struct in6_addr *remote, const struct in6_addr *local,
117 __be32 key, __be16 gre_proto)
118 {
119 struct net *net = dev_net(dev);
120 int link = dev->ifindex;
121 unsigned int h0 = HASH_ADDR(remote);
122 unsigned int h1 = HASH_KEY(key);
123 struct ip6_tnl *t, *cand = NULL;
124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
125 int dev_type = (gre_proto == htons(ETH_P_TEB) ||
126 gre_proto == htons(ETH_P_ERSPAN) ||
127 gre_proto == htons(ETH_P_ERSPAN2)) ?
128 ARPHRD_ETHER : ARPHRD_IP6GRE;
129 int score, cand_score = 4;
130 struct net_device *ndev;
131
132 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
133 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
134 !ipv6_addr_equal(remote, &t->parms.raddr) ||
135 key != t->parms.i_key ||
136 !(t->dev->flags & IFF_UP))
137 continue;
138
139 if (t->dev->type != ARPHRD_IP6GRE &&
140 t->dev->type != dev_type)
141 continue;
142
143 score = 0;
144 if (t->parms.link != link)
145 score |= 1;
146 if (t->dev->type != dev_type)
147 score |= 2;
148 if (score == 0)
149 return t;
150
151 if (score < cand_score) {
152 cand = t;
153 cand_score = score;
154 }
155 }
156
157 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
158 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
159 key != t->parms.i_key ||
160 !(t->dev->flags & IFF_UP))
161 continue;
162
163 if (t->dev->type != ARPHRD_IP6GRE &&
164 t->dev->type != dev_type)
165 continue;
166
167 score = 0;
168 if (t->parms.link != link)
169 score |= 1;
170 if (t->dev->type != dev_type)
171 score |= 2;
172 if (score == 0)
173 return t;
174
175 if (score < cand_score) {
176 cand = t;
177 cand_score = score;
178 }
179 }
180
181 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
182 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
183 (!ipv6_addr_equal(local, &t->parms.raddr) ||
184 !ipv6_addr_is_multicast(local))) ||
185 key != t->parms.i_key ||
186 !(t->dev->flags & IFF_UP))
187 continue;
188
189 if (t->dev->type != ARPHRD_IP6GRE &&
190 t->dev->type != dev_type)
191 continue;
192
193 score = 0;
194 if (t->parms.link != link)
195 score |= 1;
196 if (t->dev->type != dev_type)
197 score |= 2;
198 if (score == 0)
199 return t;
200
201 if (score < cand_score) {
202 cand = t;
203 cand_score = score;
204 }
205 }
206
207 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
208 if (t->parms.i_key != key ||
209 !(t->dev->flags & IFF_UP))
210 continue;
211
212 if (t->dev->type != ARPHRD_IP6GRE &&
213 t->dev->type != dev_type)
214 continue;
215
216 score = 0;
217 if (t->parms.link != link)
218 score |= 1;
219 if (t->dev->type != dev_type)
220 score |= 2;
221 if (score == 0)
222 return t;
223
224 if (score < cand_score) {
225 cand = t;
226 cand_score = score;
227 }
228 }
229
230 if (cand)
231 return cand;
232
233 if (gre_proto == htons(ETH_P_ERSPAN) ||
234 gre_proto == htons(ETH_P_ERSPAN2))
235 t = rcu_dereference(ign->collect_md_tun_erspan);
236 else
237 t = rcu_dereference(ign->collect_md_tun);
238
239 if (t && t->dev->flags & IFF_UP)
240 return t;
241
242 ndev = READ_ONCE(ign->fb_tunnel_dev);
243 if (ndev && ndev->flags & IFF_UP)
244 return netdev_priv(ndev);
245
246 return NULL;
247 }
248
249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
250 const struct __ip6_tnl_parm *p)
251 {
252 const struct in6_addr *remote = &p->raddr;
253 const struct in6_addr *local = &p->laddr;
254 unsigned int h = HASH_KEY(p->i_key);
255 int prio = 0;
256
257 if (!ipv6_addr_any(local))
258 prio |= 1;
259 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
260 prio |= 2;
261 h ^= HASH_ADDR(remote);
262 }
263
264 return &ign->tunnels[prio][h];
265 }
266
267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269 if (t->parms.collect_md)
270 rcu_assign_pointer(ign->collect_md_tun, t);
271 }
272
273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
274 {
275 if (t->parms.collect_md)
276 rcu_assign_pointer(ign->collect_md_tun_erspan, t);
277 }
278
279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
280 {
281 if (t->parms.collect_md)
282 rcu_assign_pointer(ign->collect_md_tun, NULL);
283 }
284
285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
286 struct ip6_tnl *t)
287 {
288 if (t->parms.collect_md)
289 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
290 }
291
292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
293 const struct ip6_tnl *t)
294 {
295 return __ip6gre_bucket(ign, &t->parms);
296 }
297
298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
299 {
300 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
301
302 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
303 rcu_assign_pointer(*tp, t);
304 }
305
306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
307 {
308 struct ip6_tnl __rcu **tp;
309 struct ip6_tnl *iter;
310
311 for (tp = ip6gre_bucket(ign, t);
312 (iter = rtnl_dereference(*tp)) != NULL;
313 tp = &iter->next) {
314 if (t == iter) {
315 rcu_assign_pointer(*tp, t->next);
316 break;
317 }
318 }
319 }
320
321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
322 const struct __ip6_tnl_parm *parms,
323 int type)
324 {
325 const struct in6_addr *remote = &parms->raddr;
326 const struct in6_addr *local = &parms->laddr;
327 __be32 key = parms->i_key;
328 int link = parms->link;
329 struct ip6_tnl *t;
330 struct ip6_tnl __rcu **tp;
331 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
332
333 for (tp = __ip6gre_bucket(ign, parms);
334 (t = rtnl_dereference(*tp)) != NULL;
335 tp = &t->next)
336 if (ipv6_addr_equal(local, &t->parms.laddr) &&
337 ipv6_addr_equal(remote, &t->parms.raddr) &&
338 key == t->parms.i_key &&
339 link == t->parms.link &&
340 type == t->dev->type)
341 break;
342
343 return t;
344 }
345
346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
347 const struct __ip6_tnl_parm *parms, int create)
348 {
349 struct ip6_tnl *t, *nt;
350 struct net_device *dev;
351 char name[IFNAMSIZ];
352 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
353
354 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
355 if (t && create)
356 return NULL;
357 if (t || !create)
358 return t;
359
360 if (parms->name[0]) {
361 if (!dev_valid_name(parms->name))
362 return NULL;
363 strlcpy(name, parms->name, IFNAMSIZ);
364 } else {
365 strcpy(name, "ip6gre%d");
366 }
367 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
368 ip6gre_tunnel_setup);
369 if (!dev)
370 return NULL;
371
372 dev_net_set(dev, net);
373
374 nt = netdev_priv(dev);
375 nt->parms = *parms;
376 dev->rtnl_link_ops = &ip6gre_link_ops;
377
378 nt->dev = dev;
379 nt->net = dev_net(dev);
380
381 if (register_netdevice(dev) < 0)
382 goto failed_free;
383
384 ip6gre_tnl_link_config(nt, 1);
385
386 /* Can use a lockless transmit, unless we generate output sequences */
387 if (!(nt->parms.o_flags & TUNNEL_SEQ))
388 dev->features |= NETIF_F_LLTX;
389
390 ip6gre_tunnel_link(ign, nt);
391 return nt;
392
393 failed_free:
394 free_netdev(dev);
395 return NULL;
396 }
397
398 static void ip6erspan_tunnel_uninit(struct net_device *dev)
399 {
400 struct ip6_tnl *t = netdev_priv(dev);
401 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
402
403 ip6erspan_tunnel_unlink_md(ign, t);
404 ip6gre_tunnel_unlink(ign, t);
405 dst_cache_reset(&t->dst_cache);
406 dev_put_track(dev, &t->dev_tracker);
407 }
408
409 static void ip6gre_tunnel_uninit(struct net_device *dev)
410 {
411 struct ip6_tnl *t = netdev_priv(dev);
412 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
413
414 ip6gre_tunnel_unlink_md(ign, t);
415 ip6gre_tunnel_unlink(ign, t);
416 if (ign->fb_tunnel_dev == dev)
417 WRITE_ONCE(ign->fb_tunnel_dev, NULL);
418 dst_cache_reset(&t->dst_cache);
419 dev_put_track(dev, &t->dev_tracker);
420 }
421
422
423 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
424 u8 type, u8 code, int offset, __be32 info)
425 {
426 struct net *net = dev_net(skb->dev);
427 const struct ipv6hdr *ipv6h;
428 struct tnl_ptk_info tpi;
429 struct ip6_tnl *t;
430
431 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
432 offset) < 0)
433 return -EINVAL;
434
435 ipv6h = (const struct ipv6hdr *)skb->data;
436 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
437 tpi.key, tpi.proto);
438 if (!t)
439 return -ENOENT;
440
441 switch (type) {
442 case ICMPV6_DEST_UNREACH:
443 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
444 t->parms.name);
445 if (code != ICMPV6_PORT_UNREACH)
446 break;
447 return 0;
448 case ICMPV6_TIME_EXCEED:
449 if (code == ICMPV6_EXC_HOPLIMIT) {
450 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
451 t->parms.name);
452 break;
453 }
454 return 0;
455 case ICMPV6_PARAMPROB: {
456 struct ipv6_tlv_tnl_enc_lim *tel;
457 __u32 teli;
458
459 teli = 0;
460 if (code == ICMPV6_HDR_FIELD)
461 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
462
463 if (teli && teli == be32_to_cpu(info) - 2) {
464 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
465 if (tel->encap_limit == 0) {
466 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
467 t->parms.name);
468 }
469 } else {
470 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
471 t->parms.name);
472 }
473 return 0;
474 }
475 case ICMPV6_PKT_TOOBIG:
476 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
477 return 0;
478 case NDISC_REDIRECT:
479 ip6_redirect(skb, net, skb->dev->ifindex, 0,
480 sock_net_uid(net, NULL));
481 return 0;
482 }
483
484 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
485 t->err_count++;
486 else
487 t->err_count = 1;
488 t->err_time = jiffies;
489
490 return 0;
491 }
492
493 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
494 {
495 const struct ipv6hdr *ipv6h;
496 struct ip6_tnl *tunnel;
497
498 ipv6h = ipv6_hdr(skb);
499 tunnel = ip6gre_tunnel_lookup(skb->dev,
500 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
501 tpi->proto);
502 if (tunnel) {
503 if (tunnel->parms.collect_md) {
504 struct metadata_dst *tun_dst;
505 __be64 tun_id;
506 __be16 flags;
507
508 flags = tpi->flags;
509 tun_id = key32_to_tunnel_id(tpi->key);
510
511 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
512 if (!tun_dst)
513 return PACKET_REJECT;
514
515 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
516 } else {
517 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
518 }
519
520 return PACKET_RCVD;
521 }
522
523 return PACKET_REJECT;
524 }
525
526 static int ip6erspan_rcv(struct sk_buff *skb,
527 struct tnl_ptk_info *tpi,
528 int gre_hdr_len)
529 {
530 struct erspan_base_hdr *ershdr;
531 const struct ipv6hdr *ipv6h;
532 struct erspan_md2 *md2;
533 struct ip6_tnl *tunnel;
534 u8 ver;
535
536 ipv6h = ipv6_hdr(skb);
537 ershdr = (struct erspan_base_hdr *)skb->data;
538 ver = ershdr->ver;
539
540 tunnel = ip6gre_tunnel_lookup(skb->dev,
541 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
542 tpi->proto);
543 if (tunnel) {
544 int len = erspan_hdr_len(ver);
545
546 if (unlikely(!pskb_may_pull(skb, len)))
547 return PACKET_REJECT;
548
549 if (__iptunnel_pull_header(skb, len,
550 htons(ETH_P_TEB),
551 false, false) < 0)
552 return PACKET_REJECT;
553
554 if (tunnel->parms.collect_md) {
555 struct erspan_metadata *pkt_md, *md;
556 struct metadata_dst *tun_dst;
557 struct ip_tunnel_info *info;
558 unsigned char *gh;
559 __be64 tun_id;
560 __be16 flags;
561
562 tpi->flags |= TUNNEL_KEY;
563 flags = tpi->flags;
564 tun_id = key32_to_tunnel_id(tpi->key);
565
566 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
567 sizeof(*md));
568 if (!tun_dst)
569 return PACKET_REJECT;
570
571 /* skb can be uncloned in __iptunnel_pull_header, so
572 * old pkt_md is no longer valid and we need to reset
573 * it
574 */
575 gh = skb_network_header(skb) +
576 skb_network_header_len(skb);
577 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
578 sizeof(*ershdr));
579 info = &tun_dst->u.tun_info;
580 md = ip_tunnel_info_opts(info);
581 md->version = ver;
582 md2 = &md->u.md2;
583 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
584 ERSPAN_V2_MDSIZE);
585 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
586 info->options_len = sizeof(*md);
587
588 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
589
590 } else {
591 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
592 }
593
594 return PACKET_RCVD;
595 }
596
597 return PACKET_REJECT;
598 }
599
600 static int gre_rcv(struct sk_buff *skb)
601 {
602 struct tnl_ptk_info tpi;
603 bool csum_err = false;
604 int hdr_len;
605
606 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
607 if (hdr_len < 0)
608 goto drop;
609
610 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
611 goto drop;
612
613 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
614 tpi.proto == htons(ETH_P_ERSPAN2))) {
615 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
616 return 0;
617 goto out;
618 }
619
620 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
621 return 0;
622
623 out:
624 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
625 drop:
626 kfree_skb(skb);
627 return 0;
628 }
629
630 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
631 {
632 return iptunnel_handle_offloads(skb,
633 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
634 }
635
636 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
637 struct net_device *dev,
638 struct flowi6 *fl6, __u8 *dsfield,
639 int *encap_limit)
640 {
641 const struct iphdr *iph = ip_hdr(skb);
642 struct ip6_tnl *t = netdev_priv(dev);
643
644 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
645 *encap_limit = t->parms.encap_limit;
646
647 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
648
649 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
650 *dsfield = ipv4_get_dsfield(iph);
651 else
652 *dsfield = ip6_tclass(t->parms.flowinfo);
653
654 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
655 fl6->flowi6_mark = skb->mark;
656 else
657 fl6->flowi6_mark = t->parms.fwmark;
658
659 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
660 }
661
662 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
663 struct net_device *dev,
664 struct flowi6 *fl6, __u8 *dsfield,
665 int *encap_limit)
666 {
667 struct ipv6hdr *ipv6h;
668 struct ip6_tnl *t = netdev_priv(dev);
669 __u16 offset;
670
671 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
672 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
673 ipv6h = ipv6_hdr(skb);
674
675 if (offset > 0) {
676 struct ipv6_tlv_tnl_enc_lim *tel;
677
678 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
679 if (tel->encap_limit == 0) {
680 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
681 ICMPV6_HDR_FIELD, offset + 2);
682 return -1;
683 }
684 *encap_limit = tel->encap_limit - 1;
685 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
686 *encap_limit = t->parms.encap_limit;
687 }
688
689 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
690
691 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
692 *dsfield = ipv6_get_dsfield(ipv6h);
693 else
694 *dsfield = ip6_tclass(t->parms.flowinfo);
695
696 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
697 fl6->flowlabel |= ip6_flowlabel(ipv6h);
698
699 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
700 fl6->flowi6_mark = skb->mark;
701 else
702 fl6->flowi6_mark = t->parms.fwmark;
703
704 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
705
706 return 0;
707 }
708
709 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
710 {
711 struct ip_tunnel_info *tun_info;
712
713 tun_info = skb_tunnel_info(skb);
714 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
715 return ERR_PTR(-EINVAL);
716
717 return tun_info;
718 }
719
720 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
721 struct net_device *dev, __u8 dsfield,
722 struct flowi6 *fl6, int encap_limit,
723 __u32 *pmtu, __be16 proto)
724 {
725 struct ip6_tnl *tunnel = netdev_priv(dev);
726 __be16 protocol;
727 __be16 flags;
728
729 if (dev->type == ARPHRD_ETHER)
730 IPCB(skb)->flags = 0;
731
732 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
733 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
734 else
735 fl6->daddr = tunnel->parms.raddr;
736
737 /* Push GRE header. */
738 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
739
740 if (tunnel->parms.collect_md) {
741 struct ip_tunnel_info *tun_info;
742 const struct ip_tunnel_key *key;
743 int tun_hlen;
744
745 tun_info = skb_tunnel_info_txcheck(skb);
746 if (IS_ERR(tun_info) ||
747 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
748 return -EINVAL;
749
750 key = &tun_info->key;
751 memset(fl6, 0, sizeof(*fl6));
752 fl6->flowi6_proto = IPPROTO_GRE;
753 fl6->daddr = key->u.ipv6.dst;
754 fl6->flowlabel = key->label;
755 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
756 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
757
758 dsfield = key->tos;
759 flags = key->tun_flags &
760 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
761 tun_hlen = gre_calc_hlen(flags);
762
763 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
764 return -ENOMEM;
765
766 gre_build_header(skb, tun_hlen,
767 flags, protocol,
768 tunnel_id_to_key32(tun_info->key.tun_id),
769 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
770 : 0);
771
772 } else {
773 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
774 return -ENOMEM;
775
776 flags = tunnel->parms.o_flags;
777
778 gre_build_header(skb, tunnel->tun_hlen, flags,
779 protocol, tunnel->parms.o_key,
780 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno))
781 : 0);
782 }
783
784 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
785 NEXTHDR_GRE);
786 }
787
788 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
789 {
790 struct ip6_tnl *t = netdev_priv(dev);
791 int encap_limit = -1;
792 struct flowi6 fl6;
793 __u8 dsfield = 0;
794 __u32 mtu;
795 int err;
796
797 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
798
799 if (!t->parms.collect_md)
800 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
801 &dsfield, &encap_limit);
802
803 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
804 if (err)
805 return -1;
806
807 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
808 skb->protocol);
809 if (err != 0) {
810 /* XXX: send ICMP error even if DF is not set. */
811 if (err == -EMSGSIZE)
812 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
813 htonl(mtu));
814 return -1;
815 }
816
817 return 0;
818 }
819
820 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
821 {
822 struct ip6_tnl *t = netdev_priv(dev);
823 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
824 int encap_limit = -1;
825 struct flowi6 fl6;
826 __u8 dsfield = 0;
827 __u32 mtu;
828 int err;
829
830 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
831 return -1;
832
833 if (!t->parms.collect_md &&
834 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
835 return -1;
836
837 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
838 return -1;
839
840 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
841 &mtu, skb->protocol);
842 if (err != 0) {
843 if (err == -EMSGSIZE)
844 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
845 return -1;
846 }
847
848 return 0;
849 }
850
851 /**
852 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own
853 * @t: the outgoing tunnel device
854 * @hdr: IPv6 header from the incoming packet
855 *
856 * Description:
857 * Avoid trivial tunneling loop by checking that tunnel exit-point
858 * doesn't match source of incoming packet.
859 *
860 * Return:
861 * 1 if conflict,
862 * 0 else
863 **/
864
865 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
866 const struct ipv6hdr *hdr)
867 {
868 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
869 }
870
871 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
872 {
873 struct ip6_tnl *t = netdev_priv(dev);
874 int encap_limit = -1;
875 struct flowi6 fl6;
876 __u32 mtu;
877 int err;
878
879 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
880 encap_limit = t->parms.encap_limit;
881
882 if (!t->parms.collect_md)
883 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
884
885 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
886 if (err)
887 return err;
888
889 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
890
891 return err;
892 }
893
894 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
895 struct net_device *dev)
896 {
897 struct ip6_tnl *t = netdev_priv(dev);
898 struct net_device_stats *stats = &t->dev->stats;
899 int ret;
900
901 if (!pskb_inet_may_pull(skb))
902 goto tx_err;
903
904 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
905 goto tx_err;
906
907 switch (skb->protocol) {
908 case htons(ETH_P_IP):
909 ret = ip6gre_xmit_ipv4(skb, dev);
910 break;
911 case htons(ETH_P_IPV6):
912 ret = ip6gre_xmit_ipv6(skb, dev);
913 break;
914 default:
915 ret = ip6gre_xmit_other(skb, dev);
916 break;
917 }
918
919 if (ret < 0)
920 goto tx_err;
921
922 return NETDEV_TX_OK;
923
924 tx_err:
925 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb)))
926 stats->tx_errors++;
927 stats->tx_dropped++;
928 kfree_skb(skb);
929 return NETDEV_TX_OK;
930 }
931
932 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
933 struct net_device *dev)
934 {
935 struct ip_tunnel_info *tun_info = NULL;
936 struct ip6_tnl *t = netdev_priv(dev);
937 struct dst_entry *dst = skb_dst(skb);
938 struct net_device_stats *stats;
939 bool truncate = false;
940 int encap_limit = -1;
941 __u8 dsfield = false;
942 struct flowi6 fl6;
943 int err = -EINVAL;
944 __be16 proto;
945 __u32 mtu;
946 int nhoff;
947
948 if (!pskb_inet_may_pull(skb))
949 goto tx_err;
950
951 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
952 goto tx_err;
953
954 if (gre_handle_offloads(skb, false))
955 goto tx_err;
956
957 if (skb->len > dev->mtu + dev->hard_header_len) {
958 pskb_trim(skb, dev->mtu + dev->hard_header_len);
959 truncate = true;
960 }
961
962 nhoff = skb_network_header(skb) - skb_mac_header(skb);
963 if (skb->protocol == htons(ETH_P_IP) &&
964 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
965 truncate = true;
966
967 if (skb->protocol == htons(ETH_P_IPV6)) {
968 int thoff;
969
970 if (skb_transport_header_was_set(skb))
971 thoff = skb_transport_header(skb) - skb_mac_header(skb);
972 else
973 thoff = nhoff + sizeof(struct ipv6hdr);
974 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
975 truncate = true;
976 }
977
978 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
979 goto tx_err;
980
981 t->parms.o_flags &= ~TUNNEL_KEY;
982 IPCB(skb)->flags = 0;
983
984 /* For collect_md mode, derive fl6 from the tunnel key,
985 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
986 */
987 if (t->parms.collect_md) {
988 const struct ip_tunnel_key *key;
989 struct erspan_metadata *md;
990 __be32 tun_id;
991
992 tun_info = skb_tunnel_info_txcheck(skb);
993 if (IS_ERR(tun_info) ||
994 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
995 goto tx_err;
996
997 key = &tun_info->key;
998 memset(&fl6, 0, sizeof(fl6));
999 fl6.flowi6_proto = IPPROTO_GRE;
1000 fl6.daddr = key->u.ipv6.dst;
1001 fl6.flowlabel = key->label;
1002 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1003 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
1004
1005 dsfield = key->tos;
1006 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
1007 goto tx_err;
1008 if (tun_info->options_len < sizeof(*md))
1009 goto tx_err;
1010 md = ip_tunnel_info_opts(tun_info);
1011
1012 tun_id = tunnel_id_to_key32(key->tun_id);
1013 if (md->version == 1) {
1014 erspan_build_header(skb,
1015 ntohl(tun_id),
1016 ntohl(md->u.index), truncate,
1017 false);
1018 } else if (md->version == 2) {
1019 erspan_build_header_v2(skb,
1020 ntohl(tun_id),
1021 md->u.md2.dir,
1022 get_hwid(&md->u.md2),
1023 truncate, false);
1024 } else {
1025 goto tx_err;
1026 }
1027 } else {
1028 switch (skb->protocol) {
1029 case htons(ETH_P_IP):
1030 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1031 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1032 &dsfield, &encap_limit);
1033 break;
1034 case htons(ETH_P_IPV6):
1035 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1036 goto tx_err;
1037 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1038 &dsfield, &encap_limit))
1039 goto tx_err;
1040 break;
1041 default:
1042 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1043 break;
1044 }
1045
1046 if (t->parms.erspan_ver == 1)
1047 erspan_build_header(skb, ntohl(t->parms.o_key),
1048 t->parms.index,
1049 truncate, false);
1050 else if (t->parms.erspan_ver == 2)
1051 erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1052 t->parms.dir,
1053 t->parms.hwid,
1054 truncate, false);
1055 else
1056 goto tx_err;
1057
1058 fl6.daddr = t->parms.raddr;
1059 }
1060
1061 /* Push GRE header. */
1062 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN)
1063 : htons(ETH_P_ERSPAN2);
1064 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno)));
1065
1066 /* TooBig packet may have updated dst->dev's mtu */
1067 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1068 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false);
1069
1070 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1071 NEXTHDR_GRE);
1072 if (err != 0) {
1073 /* XXX: send ICMP error even if DF is not set. */
1074 if (err == -EMSGSIZE) {
1075 if (skb->protocol == htons(ETH_P_IP))
1076 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1077 ICMP_FRAG_NEEDED, htonl(mtu));
1078 else
1079 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1080 }
1081
1082 goto tx_err;
1083 }
1084 return NETDEV_TX_OK;
1085
1086 tx_err:
1087 stats = &t->dev->stats;
1088 if (!IS_ERR(tun_info))
1089 stats->tx_errors++;
1090 stats->tx_dropped++;
1091 kfree_skb(skb);
1092 return NETDEV_TX_OK;
1093 }
1094
1095 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1096 {
1097 struct net_device *dev = t->dev;
1098 struct __ip6_tnl_parm *p = &t->parms;
1099 struct flowi6 *fl6 = &t->fl.u.ip6;
1100
1101 if (dev->type != ARPHRD_ETHER) {
1102 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1103 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1104 }
1105
1106 /* Set up flowi template */
1107 fl6->saddr = p->laddr;
1108 fl6->daddr = p->raddr;
1109 fl6->flowi6_oif = p->link;
1110 fl6->flowlabel = 0;
1111 fl6->flowi6_proto = IPPROTO_GRE;
1112 fl6->fl6_gre_key = t->parms.o_key;
1113
1114 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1115 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1116 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1117 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1118
1119 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1120 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1121
1122 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1123 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1124 dev->flags |= IFF_POINTOPOINT;
1125 else
1126 dev->flags &= ~IFF_POINTOPOINT;
1127 }
1128
1129 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1130 int t_hlen)
1131 {
1132 const struct __ip6_tnl_parm *p = &t->parms;
1133 struct net_device *dev = t->dev;
1134
1135 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1136 int strict = (ipv6_addr_type(&p->raddr) &
1137 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1138
1139 struct rt6_info *rt = rt6_lookup(t->net,
1140 &p->raddr, &p->laddr,
1141 p->link, NULL, strict);
1142
1143 if (!rt)
1144 return;
1145
1146 if (rt->dst.dev) {
1147 unsigned short dst_len = rt->dst.dev->hard_header_len +
1148 t_hlen;
1149
1150 if (t->dev->header_ops)
1151 dev->hard_header_len = dst_len;
1152 else
1153 dev->needed_headroom = dst_len;
1154
1155 if (set_mtu) {
1156 dev->mtu = rt->dst.dev->mtu - t_hlen;
1157 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1158 dev->mtu -= 8;
1159 if (dev->type == ARPHRD_ETHER)
1160 dev->mtu -= ETH_HLEN;
1161
1162 if (dev->mtu < IPV6_MIN_MTU)
1163 dev->mtu = IPV6_MIN_MTU;
1164 }
1165 }
1166 ip6_rt_put(rt);
1167 }
1168 }
1169
1170 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1171 {
1172 int t_hlen;
1173
1174 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1175 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1176
1177 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1178
1179 if (tunnel->dev->header_ops)
1180 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1181 else
1182 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1183
1184 return t_hlen;
1185 }
1186
1187 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1188 {
1189 ip6gre_tnl_link_config_common(t);
1190 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1191 }
1192
1193 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1194 const struct __ip6_tnl_parm *p)
1195 {
1196 t->parms.laddr = p->laddr;
1197 t->parms.raddr = p->raddr;
1198 t->parms.flags = p->flags;
1199 t->parms.hop_limit = p->hop_limit;
1200 t->parms.encap_limit = p->encap_limit;
1201 t->parms.flowinfo = p->flowinfo;
1202 t->parms.link = p->link;
1203 t->parms.proto = p->proto;
1204 t->parms.i_key = p->i_key;
1205 t->parms.o_key = p->o_key;
1206 t->parms.i_flags = p->i_flags;
1207 t->parms.o_flags = p->o_flags;
1208 t->parms.fwmark = p->fwmark;
1209 t->parms.erspan_ver = p->erspan_ver;
1210 t->parms.index = p->index;
1211 t->parms.dir = p->dir;
1212 t->parms.hwid = p->hwid;
1213 dst_cache_reset(&t->dst_cache);
1214 }
1215
1216 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1217 int set_mtu)
1218 {
1219 ip6gre_tnl_copy_tnl_parm(t, p);
1220 ip6gre_tnl_link_config(t, set_mtu);
1221 return 0;
1222 }
1223
1224 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1225 const struct ip6_tnl_parm2 *u)
1226 {
1227 p->laddr = u->laddr;
1228 p->raddr = u->raddr;
1229 p->flags = u->flags;
1230 p->hop_limit = u->hop_limit;
1231 p->encap_limit = u->encap_limit;
1232 p->flowinfo = u->flowinfo;
1233 p->link = u->link;
1234 p->i_key = u->i_key;
1235 p->o_key = u->o_key;
1236 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1237 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1238 memcpy(p->name, u->name, sizeof(u->name));
1239 }
1240
1241 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1242 const struct __ip6_tnl_parm *p)
1243 {
1244 u->proto = IPPROTO_GRE;
1245 u->laddr = p->laddr;
1246 u->raddr = p->raddr;
1247 u->flags = p->flags;
1248 u->hop_limit = p->hop_limit;
1249 u->encap_limit = p->encap_limit;
1250 u->flowinfo = p->flowinfo;
1251 u->link = p->link;
1252 u->i_key = p->i_key;
1253 u->o_key = p->o_key;
1254 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1255 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1256 memcpy(u->name, p->name, sizeof(u->name));
1257 }
1258
1259 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1260 struct ifreq *ifr, void __user *data,
1261 int cmd)
1262 {
1263 int err = 0;
1264 struct ip6_tnl_parm2 p;
1265 struct __ip6_tnl_parm p1;
1266 struct ip6_tnl *t = netdev_priv(dev);
1267 struct net *net = t->net;
1268 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1269
1270 memset(&p1, 0, sizeof(p1));
1271
1272 switch (cmd) {
1273 case SIOCGETTUNNEL:
1274 if (dev == ign->fb_tunnel_dev) {
1275 if (copy_from_user(&p, data, sizeof(p))) {
1276 err = -EFAULT;
1277 break;
1278 }
1279 ip6gre_tnl_parm_from_user(&p1, &p);
1280 t = ip6gre_tunnel_locate(net, &p1, 0);
1281 if (!t)
1282 t = netdev_priv(dev);
1283 }
1284 memset(&p, 0, sizeof(p));
1285 ip6gre_tnl_parm_to_user(&p, &t->parms);
1286 if (copy_to_user(data, &p, sizeof(p)))
1287 err = -EFAULT;
1288 break;
1289
1290 case SIOCADDTUNNEL:
1291 case SIOCCHGTUNNEL:
1292 err = -EPERM;
1293 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1294 goto done;
1295
1296 err = -EFAULT;
1297 if (copy_from_user(&p, data, sizeof(p)))
1298 goto done;
1299
1300 err = -EINVAL;
1301 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1302 goto done;
1303
1304 if (!(p.i_flags&GRE_KEY))
1305 p.i_key = 0;
1306 if (!(p.o_flags&GRE_KEY))
1307 p.o_key = 0;
1308
1309 ip6gre_tnl_parm_from_user(&p1, &p);
1310 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1311
1312 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1313 if (t) {
1314 if (t->dev != dev) {
1315 err = -EEXIST;
1316 break;
1317 }
1318 } else {
1319 t = netdev_priv(dev);
1320
1321 ip6gre_tunnel_unlink(ign, t);
1322 synchronize_net();
1323 ip6gre_tnl_change(t, &p1, 1);
1324 ip6gre_tunnel_link(ign, t);
1325 netdev_state_change(dev);
1326 }
1327 }
1328
1329 if (t) {
1330 err = 0;
1331
1332 memset(&p, 0, sizeof(p));
1333 ip6gre_tnl_parm_to_user(&p, &t->parms);
1334 if (copy_to_user(data, &p, sizeof(p)))
1335 err = -EFAULT;
1336 } else
1337 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1338 break;
1339
1340 case SIOCDELTUNNEL:
1341 err = -EPERM;
1342 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1343 goto done;
1344
1345 if (dev == ign->fb_tunnel_dev) {
1346 err = -EFAULT;
1347 if (copy_from_user(&p, data, sizeof(p)))
1348 goto done;
1349 err = -ENOENT;
1350 ip6gre_tnl_parm_from_user(&p1, &p);
1351 t = ip6gre_tunnel_locate(net, &p1, 0);
1352 if (!t)
1353 goto done;
1354 err = -EPERM;
1355 if (t == netdev_priv(ign->fb_tunnel_dev))
1356 goto done;
1357 dev = t->dev;
1358 }
1359 unregister_netdevice(dev);
1360 err = 0;
1361 break;
1362
1363 default:
1364 err = -EINVAL;
1365 }
1366
1367 done:
1368 return err;
1369 }
1370
1371 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1372 unsigned short type, const void *daddr,
1373 const void *saddr, unsigned int len)
1374 {
1375 struct ip6_tnl *t = netdev_priv(dev);
1376 struct ipv6hdr *ipv6h;
1377 __be16 *p;
1378
1379 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1380 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1381 t->fl.u.ip6.flowlabel,
1382 true, &t->fl.u.ip6));
1383 ipv6h->hop_limit = t->parms.hop_limit;
1384 ipv6h->nexthdr = NEXTHDR_GRE;
1385 ipv6h->saddr = t->parms.laddr;
1386 ipv6h->daddr = t->parms.raddr;
1387
1388 p = (__be16 *)(ipv6h + 1);
1389 p[0] = t->parms.o_flags;
1390 p[1] = htons(type);
1391
1392 /*
1393 * Set the source hardware address.
1394 */
1395
1396 if (saddr)
1397 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1398 if (daddr)
1399 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1400 if (!ipv6_addr_any(&ipv6h->daddr))
1401 return t->hlen;
1402
1403 return -t->hlen;
1404 }
1405
1406 static const struct header_ops ip6gre_header_ops = {
1407 .create = ip6gre_header,
1408 };
1409
1410 static const struct net_device_ops ip6gre_netdev_ops = {
1411 .ndo_init = ip6gre_tunnel_init,
1412 .ndo_uninit = ip6gre_tunnel_uninit,
1413 .ndo_start_xmit = ip6gre_tunnel_xmit,
1414 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate,
1415 .ndo_change_mtu = ip6_tnl_change_mtu,
1416 .ndo_get_stats64 = dev_get_tstats64,
1417 .ndo_get_iflink = ip6_tnl_get_iflink,
1418 };
1419
1420 static void ip6gre_dev_free(struct net_device *dev)
1421 {
1422 struct ip6_tnl *t = netdev_priv(dev);
1423
1424 gro_cells_destroy(&t->gro_cells);
1425 dst_cache_destroy(&t->dst_cache);
1426 free_percpu(dev->tstats);
1427 }
1428
1429 static void ip6gre_tunnel_setup(struct net_device *dev)
1430 {
1431 dev->netdev_ops = &ip6gre_netdev_ops;
1432 dev->needs_free_netdev = true;
1433 dev->priv_destructor = ip6gre_dev_free;
1434
1435 dev->type = ARPHRD_IP6GRE;
1436
1437 dev->flags |= IFF_NOARP;
1438 dev->addr_len = sizeof(struct in6_addr);
1439 netif_keep_dst(dev);
1440 /* This perm addr will be used as interface identifier by IPv6 */
1441 dev->addr_assign_type = NET_ADDR_RANDOM;
1442 eth_random_addr(dev->perm_addr);
1443 }
1444
1445 #define GRE6_FEATURES (NETIF_F_SG | \
1446 NETIF_F_FRAGLIST | \
1447 NETIF_F_HIGHDMA | \
1448 NETIF_F_HW_CSUM)
1449
1450 static void ip6gre_tnl_init_features(struct net_device *dev)
1451 {
1452 struct ip6_tnl *nt = netdev_priv(dev);
1453
1454 dev->features |= GRE6_FEATURES;
1455 dev->hw_features |= GRE6_FEATURES;
1456
1457 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1458 /* TCP offload with GRE SEQ is not supported, nor
1459 * can we support 2 levels of outer headers requiring
1460 * an update.
1461 */
1462 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1463 nt->encap.type == TUNNEL_ENCAP_NONE) {
1464 dev->features |= NETIF_F_GSO_SOFTWARE;
1465 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1466 }
1467
1468 /* Can use a lockless transmit, unless we generate
1469 * output sequences
1470 */
1471 dev->features |= NETIF_F_LLTX;
1472 }
1473 }
1474
1475 static int ip6gre_tunnel_init_common(struct net_device *dev)
1476 {
1477 struct ip6_tnl *tunnel;
1478 int ret;
1479 int t_hlen;
1480
1481 tunnel = netdev_priv(dev);
1482
1483 tunnel->dev = dev;
1484 tunnel->net = dev_net(dev);
1485 strcpy(tunnel->parms.name, dev->name);
1486
1487 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1488 if (!dev->tstats)
1489 return -ENOMEM;
1490
1491 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1492 if (ret)
1493 goto cleanup_alloc_pcpu_stats;
1494
1495 ret = gro_cells_init(&tunnel->gro_cells, dev);
1496 if (ret)
1497 goto cleanup_dst_cache_init;
1498
1499 t_hlen = ip6gre_calc_hlen(tunnel);
1500 dev->mtu = ETH_DATA_LEN - t_hlen;
1501 if (dev->type == ARPHRD_ETHER)
1502 dev->mtu -= ETH_HLEN;
1503 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1504 dev->mtu -= 8;
1505
1506 if (tunnel->parms.collect_md) {
1507 netif_keep_dst(dev);
1508 }
1509 ip6gre_tnl_init_features(dev);
1510
1511 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1512 return 0;
1513
1514 cleanup_dst_cache_init:
1515 dst_cache_destroy(&tunnel->dst_cache);
1516 cleanup_alloc_pcpu_stats:
1517 free_percpu(dev->tstats);
1518 dev->tstats = NULL;
1519 return ret;
1520 }
1521
1522 static int ip6gre_tunnel_init(struct net_device *dev)
1523 {
1524 struct ip6_tnl *tunnel;
1525 int ret;
1526
1527 ret = ip6gre_tunnel_init_common(dev);
1528 if (ret)
1529 return ret;
1530
1531 tunnel = netdev_priv(dev);
1532
1533 if (tunnel->parms.collect_md)
1534 return 0;
1535
1536 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1537 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1538
1539 if (ipv6_addr_any(&tunnel->parms.raddr))
1540 dev->header_ops = &ip6gre_header_ops;
1541
1542 return 0;
1543 }
1544
1545 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1546 {
1547 struct ip6_tnl *tunnel = netdev_priv(dev);
1548
1549 tunnel->dev = dev;
1550 tunnel->net = dev_net(dev);
1551 strcpy(tunnel->parms.name, dev->name);
1552
1553 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1554 }
1555
1556 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1557 .handler = gre_rcv,
1558 .err_handler = ip6gre_err,
1559 .flags = INET6_PROTO_FINAL,
1560 };
1561
1562 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1563 {
1564 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1565 struct net_device *dev, *aux;
1566 int prio;
1567
1568 for_each_netdev_safe(net, dev, aux)
1569 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1570 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1571 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1572 unregister_netdevice_queue(dev, head);
1573
1574 for (prio = 0; prio < 4; prio++) {
1575 int h;
1576 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1577 struct ip6_tnl *t;
1578
1579 t = rtnl_dereference(ign->tunnels[prio][h]);
1580
1581 while (t) {
1582 /* If dev is in the same netns, it has already
1583 * been added to the list by the previous loop.
1584 */
1585 if (!net_eq(dev_net(t->dev), net))
1586 unregister_netdevice_queue(t->dev,
1587 head);
1588 t = rtnl_dereference(t->next);
1589 }
1590 }
1591 }
1592 }
1593
1594 static int __net_init ip6gre_init_net(struct net *net)
1595 {
1596 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1597 struct net_device *ndev;
1598 int err;
1599
1600 if (!net_has_fallback_tunnels(net))
1601 return 0;
1602 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1603 NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1604 if (!ndev) {
1605 err = -ENOMEM;
1606 goto err_alloc_dev;
1607 }
1608 ign->fb_tunnel_dev = ndev;
1609 dev_net_set(ign->fb_tunnel_dev, net);
1610 /* FB netdevice is special: we have one, and only one per netns.
1611 * Allowing to move it to another netns is clearly unsafe.
1612 */
1613 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1614
1615
1616 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1617 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1618
1619 err = register_netdev(ign->fb_tunnel_dev);
1620 if (err)
1621 goto err_reg_dev;
1622
1623 rcu_assign_pointer(ign->tunnels_wc[0],
1624 netdev_priv(ign->fb_tunnel_dev));
1625 return 0;
1626
1627 err_reg_dev:
1628 free_netdev(ndev);
1629 err_alloc_dev:
1630 return err;
1631 }
1632
1633 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1634 {
1635 struct net *net;
1636 LIST_HEAD(list);
1637
1638 rtnl_lock();
1639 list_for_each_entry(net, net_list, exit_list)
1640 ip6gre_destroy_tunnels(net, &list);
1641 unregister_netdevice_many(&list);
1642 rtnl_unlock();
1643 }
1644
1645 static struct pernet_operations ip6gre_net_ops = {
1646 .init = ip6gre_init_net,
1647 .exit_batch = ip6gre_exit_batch_net,
1648 .id = &ip6gre_net_id,
1649 .size = sizeof(struct ip6gre_net),
1650 };
1651
1652 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1653 struct netlink_ext_ack *extack)
1654 {
1655 __be16 flags;
1656
1657 if (!data)
1658 return 0;
1659
1660 flags = 0;
1661 if (data[IFLA_GRE_IFLAGS])
1662 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1663 if (data[IFLA_GRE_OFLAGS])
1664 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1665 if (flags & (GRE_VERSION|GRE_ROUTING))
1666 return -EINVAL;
1667
1668 return 0;
1669 }
1670
1671 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1672 struct netlink_ext_ack *extack)
1673 {
1674 struct in6_addr daddr;
1675
1676 if (tb[IFLA_ADDRESS]) {
1677 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1678 return -EINVAL;
1679 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1680 return -EADDRNOTAVAIL;
1681 }
1682
1683 if (!data)
1684 goto out;
1685
1686 if (data[IFLA_GRE_REMOTE]) {
1687 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1688 if (ipv6_addr_any(&daddr))
1689 return -EINVAL;
1690 }
1691
1692 out:
1693 return ip6gre_tunnel_validate(tb, data, extack);
1694 }
1695
1696 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1697 struct netlink_ext_ack *extack)
1698 {
1699 __be16 flags = 0;
1700 int ret, ver = 0;
1701
1702 if (!data)
1703 return 0;
1704
1705 ret = ip6gre_tap_validate(tb, data, extack);
1706 if (ret)
1707 return ret;
1708
1709 /* ERSPAN should only have GRE sequence and key flag */
1710 if (data[IFLA_GRE_OFLAGS])
1711 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1712 if (data[IFLA_GRE_IFLAGS])
1713 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1714 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1715 flags != (GRE_SEQ | GRE_KEY))
1716 return -EINVAL;
1717
1718 /* ERSPAN Session ID only has 10-bit. Since we reuse
1719 * 32-bit key field as ID, check it's range.
1720 */
1721 if (data[IFLA_GRE_IKEY] &&
1722 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1723 return -EINVAL;
1724
1725 if (data[IFLA_GRE_OKEY] &&
1726 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1727 return -EINVAL;
1728
1729 if (data[IFLA_GRE_ERSPAN_VER]) {
1730 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1731 if (ver != 1 && ver != 2)
1732 return -EINVAL;
1733 }
1734
1735 if (ver == 1) {
1736 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1737 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1738
1739 if (index & ~INDEX_MASK)
1740 return -EINVAL;
1741 }
1742 } else if (ver == 2) {
1743 if (data[IFLA_GRE_ERSPAN_DIR]) {
1744 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1745
1746 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1747 return -EINVAL;
1748 }
1749
1750 if (data[IFLA_GRE_ERSPAN_HWID]) {
1751 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1752
1753 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1754 return -EINVAL;
1755 }
1756 }
1757
1758 return 0;
1759 }
1760
1761 static void ip6erspan_set_version(struct nlattr *data[],
1762 struct __ip6_tnl_parm *parms)
1763 {
1764 if (!data)
1765 return;
1766
1767 parms->erspan_ver = 1;
1768 if (data[IFLA_GRE_ERSPAN_VER])
1769 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1770
1771 if (parms->erspan_ver == 1) {
1772 if (data[IFLA_GRE_ERSPAN_INDEX])
1773 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1774 } else if (parms->erspan_ver == 2) {
1775 if (data[IFLA_GRE_ERSPAN_DIR])
1776 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1777 if (data[IFLA_GRE_ERSPAN_HWID])
1778 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1779 }
1780 }
1781
1782 static void ip6gre_netlink_parms(struct nlattr *data[],
1783 struct __ip6_tnl_parm *parms)
1784 {
1785 memset(parms, 0, sizeof(*parms));
1786
1787 if (!data)
1788 return;
1789
1790 if (data[IFLA_GRE_LINK])
1791 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1792
1793 if (data[IFLA_GRE_IFLAGS])
1794 parms->i_flags = gre_flags_to_tnl_flags(
1795 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1796
1797 if (data[IFLA_GRE_OFLAGS])
1798 parms->o_flags = gre_flags_to_tnl_flags(
1799 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1800
1801 if (data[IFLA_GRE_IKEY])
1802 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1803
1804 if (data[IFLA_GRE_OKEY])
1805 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1806
1807 if (data[IFLA_GRE_LOCAL])
1808 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1809
1810 if (data[IFLA_GRE_REMOTE])
1811 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1812
1813 if (data[IFLA_GRE_TTL])
1814 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1815
1816 if (data[IFLA_GRE_ENCAP_LIMIT])
1817 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1818
1819 if (data[IFLA_GRE_FLOWINFO])
1820 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1821
1822 if (data[IFLA_GRE_FLAGS])
1823 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1824
1825 if (data[IFLA_GRE_FWMARK])
1826 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1827
1828 if (data[IFLA_GRE_COLLECT_METADATA])
1829 parms->collect_md = true;
1830 }
1831
1832 static int ip6gre_tap_init(struct net_device *dev)
1833 {
1834 int ret;
1835
1836 ret = ip6gre_tunnel_init_common(dev);
1837 if (ret)
1838 return ret;
1839
1840 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1841
1842 return 0;
1843 }
1844
1845 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1846 .ndo_init = ip6gre_tap_init,
1847 .ndo_uninit = ip6gre_tunnel_uninit,
1848 .ndo_start_xmit = ip6gre_tunnel_xmit,
1849 .ndo_set_mac_address = eth_mac_addr,
1850 .ndo_validate_addr = eth_validate_addr,
1851 .ndo_change_mtu = ip6_tnl_change_mtu,
1852 .ndo_get_stats64 = dev_get_tstats64,
1853 .ndo_get_iflink = ip6_tnl_get_iflink,
1854 };
1855
1856 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1857 {
1858 int t_hlen;
1859
1860 tunnel->tun_hlen = 8;
1861 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1862 erspan_hdr_len(tunnel->parms.erspan_ver);
1863
1864 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1865 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1866 return t_hlen;
1867 }
1868
1869 static int ip6erspan_tap_init(struct net_device *dev)
1870 {
1871 struct ip6_tnl *tunnel;
1872 int t_hlen;
1873 int ret;
1874
1875 tunnel = netdev_priv(dev);
1876
1877 tunnel->dev = dev;
1878 tunnel->net = dev_net(dev);
1879 strcpy(tunnel->parms.name, dev->name);
1880
1881 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1882 if (!dev->tstats)
1883 return -ENOMEM;
1884
1885 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1886 if (ret)
1887 goto cleanup_alloc_pcpu_stats;
1888
1889 ret = gro_cells_init(&tunnel->gro_cells, dev);
1890 if (ret)
1891 goto cleanup_dst_cache_init;
1892
1893 t_hlen = ip6erspan_calc_hlen(tunnel);
1894 dev->mtu = ETH_DATA_LEN - t_hlen;
1895 if (dev->type == ARPHRD_ETHER)
1896 dev->mtu -= ETH_HLEN;
1897 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1898 dev->mtu -= 8;
1899
1900 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1901 ip6erspan_tnl_link_config(tunnel, 1);
1902
1903 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1904 return 0;
1905
1906 cleanup_dst_cache_init:
1907 dst_cache_destroy(&tunnel->dst_cache);
1908 cleanup_alloc_pcpu_stats:
1909 free_percpu(dev->tstats);
1910 dev->tstats = NULL;
1911 return ret;
1912 }
1913
1914 static const struct net_device_ops ip6erspan_netdev_ops = {
1915 .ndo_init = ip6erspan_tap_init,
1916 .ndo_uninit = ip6erspan_tunnel_uninit,
1917 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1918 .ndo_set_mac_address = eth_mac_addr,
1919 .ndo_validate_addr = eth_validate_addr,
1920 .ndo_change_mtu = ip6_tnl_change_mtu,
1921 .ndo_get_stats64 = dev_get_tstats64,
1922 .ndo_get_iflink = ip6_tnl_get_iflink,
1923 };
1924
1925 static void ip6gre_tap_setup(struct net_device *dev)
1926 {
1927
1928 ether_setup(dev);
1929
1930 dev->max_mtu = 0;
1931 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1932 dev->needs_free_netdev = true;
1933 dev->priv_destructor = ip6gre_dev_free;
1934
1935 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1936 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1937 netif_keep_dst(dev);
1938 }
1939
1940 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1941 struct ip_tunnel_encap *ipencap)
1942 {
1943 bool ret = false;
1944
1945 memset(ipencap, 0, sizeof(*ipencap));
1946
1947 if (!data)
1948 return ret;
1949
1950 if (data[IFLA_GRE_ENCAP_TYPE]) {
1951 ret = true;
1952 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1953 }
1954
1955 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1956 ret = true;
1957 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1958 }
1959
1960 if (data[IFLA_GRE_ENCAP_SPORT]) {
1961 ret = true;
1962 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1963 }
1964
1965 if (data[IFLA_GRE_ENCAP_DPORT]) {
1966 ret = true;
1967 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1968 }
1969
1970 return ret;
1971 }
1972
1973 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1974 struct nlattr *tb[], struct nlattr *data[],
1975 struct netlink_ext_ack *extack)
1976 {
1977 struct ip6_tnl *nt;
1978 struct ip_tunnel_encap ipencap;
1979 int err;
1980
1981 nt = netdev_priv(dev);
1982
1983 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1984 int err = ip6_tnl_encap_setup(nt, &ipencap);
1985
1986 if (err < 0)
1987 return err;
1988 }
1989
1990 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1991 eth_hw_addr_random(dev);
1992
1993 nt->dev = dev;
1994 nt->net = dev_net(dev);
1995
1996 err = register_netdevice(dev);
1997 if (err)
1998 goto out;
1999
2000 if (tb[IFLA_MTU])
2001 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2002
2003 out:
2004 return err;
2005 }
2006
2007 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
2008 struct nlattr *tb[], struct nlattr *data[],
2009 struct netlink_ext_ack *extack)
2010 {
2011 struct ip6_tnl *nt = netdev_priv(dev);
2012 struct net *net = dev_net(dev);
2013 struct ip6gre_net *ign;
2014 int err;
2015
2016 ip6gre_netlink_parms(data, &nt->parms);
2017 ign = net_generic(net, ip6gre_net_id);
2018
2019 if (nt->parms.collect_md) {
2020 if (rtnl_dereference(ign->collect_md_tun))
2021 return -EEXIST;
2022 } else {
2023 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2024 return -EEXIST;
2025 }
2026
2027 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2028 if (!err) {
2029 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2030 ip6gre_tunnel_link_md(ign, nt);
2031 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2032 }
2033 return err;
2034 }
2035
2036 static struct ip6_tnl *
2037 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2038 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2039 struct netlink_ext_ack *extack)
2040 {
2041 struct ip6_tnl *t, *nt = netdev_priv(dev);
2042 struct net *net = nt->net;
2043 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2044 struct ip_tunnel_encap ipencap;
2045
2046 if (dev == ign->fb_tunnel_dev)
2047 return ERR_PTR(-EINVAL);
2048
2049 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2050 int err = ip6_tnl_encap_setup(nt, &ipencap);
2051
2052 if (err < 0)
2053 return ERR_PTR(err);
2054 }
2055
2056 ip6gre_netlink_parms(data, p_p);
2057
2058 t = ip6gre_tunnel_locate(net, p_p, 0);
2059
2060 if (t) {
2061 if (t->dev != dev)
2062 return ERR_PTR(-EEXIST);
2063 } else {
2064 t = nt;
2065 }
2066
2067 return t;
2068 }
2069
2070 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2071 struct nlattr *data[],
2072 struct netlink_ext_ack *extack)
2073 {
2074 struct ip6_tnl *t = netdev_priv(dev);
2075 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2076 struct __ip6_tnl_parm p;
2077
2078 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2079 if (IS_ERR(t))
2080 return PTR_ERR(t);
2081
2082 ip6gre_tunnel_unlink_md(ign, t);
2083 ip6gre_tunnel_unlink(ign, t);
2084 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2085 ip6gre_tunnel_link_md(ign, t);
2086 ip6gre_tunnel_link(ign, t);
2087 return 0;
2088 }
2089
2090 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2091 {
2092 struct net *net = dev_net(dev);
2093 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2094
2095 if (dev != ign->fb_tunnel_dev)
2096 unregister_netdevice_queue(dev, head);
2097 }
2098
2099 static size_t ip6gre_get_size(const struct net_device *dev)
2100 {
2101 return
2102 /* IFLA_GRE_LINK */
2103 nla_total_size(4) +
2104 /* IFLA_GRE_IFLAGS */
2105 nla_total_size(2) +
2106 /* IFLA_GRE_OFLAGS */
2107 nla_total_size(2) +
2108 /* IFLA_GRE_IKEY */
2109 nla_total_size(4) +
2110 /* IFLA_GRE_OKEY */
2111 nla_total_size(4) +
2112 /* IFLA_GRE_LOCAL */
2113 nla_total_size(sizeof(struct in6_addr)) +
2114 /* IFLA_GRE_REMOTE */
2115 nla_total_size(sizeof(struct in6_addr)) +
2116 /* IFLA_GRE_TTL */
2117 nla_total_size(1) +
2118 /* IFLA_GRE_ENCAP_LIMIT */
2119 nla_total_size(1) +
2120 /* IFLA_GRE_FLOWINFO */
2121 nla_total_size(4) +
2122 /* IFLA_GRE_FLAGS */
2123 nla_total_size(4) +
2124 /* IFLA_GRE_ENCAP_TYPE */
2125 nla_total_size(2) +
2126 /* IFLA_GRE_ENCAP_FLAGS */
2127 nla_total_size(2) +
2128 /* IFLA_GRE_ENCAP_SPORT */
2129 nla_total_size(2) +
2130 /* IFLA_GRE_ENCAP_DPORT */
2131 nla_total_size(2) +
2132 /* IFLA_GRE_COLLECT_METADATA */
2133 nla_total_size(0) +
2134 /* IFLA_GRE_FWMARK */
2135 nla_total_size(4) +
2136 /* IFLA_GRE_ERSPAN_INDEX */
2137 nla_total_size(4) +
2138 0;
2139 }
2140
2141 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2142 {
2143 struct ip6_tnl *t = netdev_priv(dev);
2144 struct __ip6_tnl_parm *p = &t->parms;
2145 __be16 o_flags = p->o_flags;
2146
2147 if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2148 if (!p->collect_md)
2149 o_flags |= TUNNEL_KEY;
2150
2151 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2152 goto nla_put_failure;
2153
2154 if (p->erspan_ver == 1) {
2155 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2156 goto nla_put_failure;
2157 } else {
2158 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2159 goto nla_put_failure;
2160 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2161 goto nla_put_failure;
2162 }
2163 }
2164
2165 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2166 nla_put_be16(skb, IFLA_GRE_IFLAGS,
2167 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2168 nla_put_be16(skb, IFLA_GRE_OFLAGS,
2169 gre_tnl_flags_to_gre_flags(o_flags)) ||
2170 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2171 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2172 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2173 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2174 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2175 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2176 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2177 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2178 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2179 goto nla_put_failure;
2180
2181 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2182 t->encap.type) ||
2183 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2184 t->encap.sport) ||
2185 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2186 t->encap.dport) ||
2187 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2188 t->encap.flags))
2189 goto nla_put_failure;
2190
2191 if (p->collect_md) {
2192 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2193 goto nla_put_failure;
2194 }
2195
2196 return 0;
2197
2198 nla_put_failure:
2199 return -EMSGSIZE;
2200 }
2201
2202 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2203 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2204 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2205 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2206 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2207 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2208 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) },
2209 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) },
2210 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2211 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2212 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2213 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2214 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2215 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2216 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2217 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2218 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2219 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2220 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2221 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2222 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2223 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2224 };
2225
2226 static void ip6erspan_tap_setup(struct net_device *dev)
2227 {
2228 ether_setup(dev);
2229
2230 dev->max_mtu = 0;
2231 dev->netdev_ops = &ip6erspan_netdev_ops;
2232 dev->needs_free_netdev = true;
2233 dev->priv_destructor = ip6gre_dev_free;
2234
2235 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2236 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2237 netif_keep_dst(dev);
2238 }
2239
2240 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2241 struct nlattr *tb[], struct nlattr *data[],
2242 struct netlink_ext_ack *extack)
2243 {
2244 struct ip6_tnl *nt = netdev_priv(dev);
2245 struct net *net = dev_net(dev);
2246 struct ip6gre_net *ign;
2247 int err;
2248
2249 ip6gre_netlink_parms(data, &nt->parms);
2250 ip6erspan_set_version(data, &nt->parms);
2251 ign = net_generic(net, ip6gre_net_id);
2252
2253 if (nt->parms.collect_md) {
2254 if (rtnl_dereference(ign->collect_md_tun_erspan))
2255 return -EEXIST;
2256 } else {
2257 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2258 return -EEXIST;
2259 }
2260
2261 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2262 if (!err) {
2263 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2264 ip6erspan_tunnel_link_md(ign, nt);
2265 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2266 }
2267 return err;
2268 }
2269
2270 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2271 {
2272 ip6gre_tnl_link_config_common(t);
2273 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2274 }
2275
2276 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2277 const struct __ip6_tnl_parm *p, int set_mtu)
2278 {
2279 ip6gre_tnl_copy_tnl_parm(t, p);
2280 ip6erspan_tnl_link_config(t, set_mtu);
2281 return 0;
2282 }
2283
2284 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2285 struct nlattr *data[],
2286 struct netlink_ext_ack *extack)
2287 {
2288 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2289 struct __ip6_tnl_parm p;
2290 struct ip6_tnl *t;
2291
2292 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2293 if (IS_ERR(t))
2294 return PTR_ERR(t);
2295
2296 ip6erspan_set_version(data, &p);
2297 ip6gre_tunnel_unlink_md(ign, t);
2298 ip6gre_tunnel_unlink(ign, t);
2299 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2300 ip6erspan_tunnel_link_md(ign, t);
2301 ip6gre_tunnel_link(ign, t);
2302 return 0;
2303 }
2304
2305 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2306 .kind = "ip6gre",
2307 .maxtype = IFLA_GRE_MAX,
2308 .policy = ip6gre_policy,
2309 .priv_size = sizeof(struct ip6_tnl),
2310 .setup = ip6gre_tunnel_setup,
2311 .validate = ip6gre_tunnel_validate,
2312 .newlink = ip6gre_newlink,
2313 .changelink = ip6gre_changelink,
2314 .dellink = ip6gre_dellink,
2315 .get_size = ip6gre_get_size,
2316 .fill_info = ip6gre_fill_info,
2317 .get_link_net = ip6_tnl_get_link_net,
2318 };
2319
2320 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2321 .kind = "ip6gretap",
2322 .maxtype = IFLA_GRE_MAX,
2323 .policy = ip6gre_policy,
2324 .priv_size = sizeof(struct ip6_tnl),
2325 .setup = ip6gre_tap_setup,
2326 .validate = ip6gre_tap_validate,
2327 .newlink = ip6gre_newlink,
2328 .changelink = ip6gre_changelink,
2329 .get_size = ip6gre_get_size,
2330 .fill_info = ip6gre_fill_info,
2331 .get_link_net = ip6_tnl_get_link_net,
2332 };
2333
2334 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2335 .kind = "ip6erspan",
2336 .maxtype = IFLA_GRE_MAX,
2337 .policy = ip6gre_policy,
2338 .priv_size = sizeof(struct ip6_tnl),
2339 .setup = ip6erspan_tap_setup,
2340 .validate = ip6erspan_tap_validate,
2341 .newlink = ip6erspan_newlink,
2342 .changelink = ip6erspan_changelink,
2343 .get_size = ip6gre_get_size,
2344 .fill_info = ip6gre_fill_info,
2345 .get_link_net = ip6_tnl_get_link_net,
2346 };
2347
2348 /*
2349 * And now the modules code and kernel interface.
2350 */
2351
2352 static int __init ip6gre_init(void)
2353 {
2354 int err;
2355
2356 pr_info("GRE over IPv6 tunneling driver\n");
2357
2358 err = register_pernet_device(&ip6gre_net_ops);
2359 if (err < 0)
2360 return err;
2361
2362 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2363 if (err < 0) {
2364 pr_info("%s: can't add protocol\n", __func__);
2365 goto add_proto_failed;
2366 }
2367
2368 err = rtnl_link_register(&ip6gre_link_ops);
2369 if (err < 0)
2370 goto rtnl_link_failed;
2371
2372 err = rtnl_link_register(&ip6gre_tap_ops);
2373 if (err < 0)
2374 goto tap_ops_failed;
2375
2376 err = rtnl_link_register(&ip6erspan_tap_ops);
2377 if (err < 0)
2378 goto erspan_link_failed;
2379
2380 out:
2381 return err;
2382
2383 erspan_link_failed:
2384 rtnl_link_unregister(&ip6gre_tap_ops);
2385 tap_ops_failed:
2386 rtnl_link_unregister(&ip6gre_link_ops);
2387 rtnl_link_failed:
2388 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2389 add_proto_failed:
2390 unregister_pernet_device(&ip6gre_net_ops);
2391 goto out;
2392 }
2393
2394 static void __exit ip6gre_fini(void)
2395 {
2396 rtnl_link_unregister(&ip6gre_tap_ops);
2397 rtnl_link_unregister(&ip6gre_link_ops);
2398 rtnl_link_unregister(&ip6erspan_tap_ops);
2399 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2400 unregister_pernet_device(&ip6gre_net_ops);
2401 }
2402
2403 module_init(ip6gre_init);
2404 module_exit(ip6gre_fini);
2405 MODULE_LICENSE("GPL");
2406 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2407 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2408 MODULE_ALIAS_RTNL_LINK("ip6gre");
2409 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2410 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2411 MODULE_ALIAS_NETDEV("ip6gre0");
2412
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.