?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
NFSv4: Fix error handling in nfs4_sp4_select_mode()
[linux-2.6/btrfs-unstable.git] / net / ipv6 / ip6_gre.c
blobcd2cfb04e5d82010a5eb1800a53fc8007479c6f9
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
62 static bool log_ecn_error = true;
63 module_param(log_ecn_error, bool, 0644);
64 MODULE_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
69 static unsigned int ip6gre_net_id __read_mostly;
70 struct 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
78 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
79 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
80 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
81 static int ip6gre_tunnel_init(struct net_device *dev);
82 static void ip6gre_tunnel_setup(struct net_device *dev);
83 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
84 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
85 static 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))
106 static 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
120 static 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
253 static 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
271 static 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
277 static 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
283 static 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
289 static 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
296 static 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
302 static 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
310 static 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
325 static 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
350 static 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
398 failed_free:
399 free_netdev(dev);
400 return NULL;
401 }
402
403 static 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
414 static 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
426 static 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
510 static 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
543 static 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
615 static 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
638 out:
639 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
640 drop:
641 kfree_skb(skb);
642 return 0;
643 }
644
645 static 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
651 static 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
677 static 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
723 static 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
787 static 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
819 static 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 * ip6gre_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
864 static 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
870 static 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
893 static 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
920 tx_err:
921 stats->tx_errors++;
922 stats->tx_dropped++;
923 kfree_skb(skb);
924 return NETDEV_TX_OK;
925 }
926
927 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
928 struct net_device *dev)
929 {
930 struct ip6_tnl *t = netdev_priv(dev);
931 struct dst_entry *dst = skb_dst(skb);
932 struct net_device_stats *stats;
933 bool truncate = false;
934 int encap_limit = -1;
935 __u8 dsfield = false;
936 struct flowi6 fl6;
937 int err = -EINVAL;
938 __u32 mtu;
939 int nhoff;
940 int thoff;
941
942 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
943 goto tx_err;
944
945 if (gre_handle_offloads(skb, false))
946 goto tx_err;
947
948 if (skb->len > dev->mtu + dev->hard_header_len) {
949 pskb_trim(skb, dev->mtu + dev->hard_header_len);
950 truncate = true;
951 }
952
953 nhoff = skb_network_header(skb) - skb_mac_header(skb);
954 if (skb->protocol == htons(ETH_P_IP) &&
955 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
956 truncate = true;
957
958 thoff = skb_transport_header(skb) - skb_mac_header(skb);
959 if (skb->protocol == htons(ETH_P_IPV6) &&
960 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
961 truncate = true;
962
963 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
964 goto tx_err;
965
966 t->parms.o_flags &= ~TUNNEL_KEY;
967 IPCB(skb)->flags = 0;
968
969 /* For collect_md mode, derive fl6 from the tunnel key,
970 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
971 */
972 if (t->parms.collect_md) {
973 struct ip_tunnel_info *tun_info;
974 const struct ip_tunnel_key *key;
975 struct erspan_metadata *md;
976 __be32 tun_id;
977
978 tun_info = skb_tunnel_info(skb);
979 if (unlikely(!tun_info ||
980 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
981 ip_tunnel_info_af(tun_info) != AF_INET6))
982 return -EINVAL;
983
984 key = &tun_info->key;
985 memset(&fl6, 0, sizeof(fl6));
986 fl6.flowi6_proto = IPPROTO_GRE;
987 fl6.daddr = key->u.ipv6.dst;
988 fl6.flowlabel = key->label;
989 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
990
991 dsfield = key->tos;
992 md = ip_tunnel_info_opts(tun_info);
993 if (!md)
994 goto tx_err;
995
996 tun_id = tunnel_id_to_key32(key->tun_id);
997 if (md->version == 1) {
998 erspan_build_header(skb,
999 ntohl(tun_id),
1000 ntohl(md->u.index), truncate,
1001 false);
1002 } else if (md->version == 2) {
1003 erspan_build_header_v2(skb,
1004 ntohl(tun_id),
1005 md->u.md2.dir,
1006 get_hwid(&md->u.md2),
1007 truncate, false);
1008 } else {
1009 goto tx_err;
1010 }
1011 } else {
1012 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1013
1014 switch (skb->protocol) {
1015 case htons(ETH_P_IP):
1016 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1017 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1018 &dsfield, &encap_limit);
1019 break;
1020 case htons(ETH_P_IPV6):
1021 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
1022 goto tx_err;
1023 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1024 &dsfield, &encap_limit))
1025 goto tx_err;
1026 break;
1027 default:
1028 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1029 break;
1030 }
1031
1032 if (t->parms.erspan_ver == 1)
1033 erspan_build_header(skb, ntohl(t->parms.o_key),
1034 t->parms.index,
1035 truncate, false);
1036 else if (t->parms.erspan_ver == 2)
1037 erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1038 t->parms.dir,
1039 t->parms.hwid,
1040 truncate, false);
1041 else
1042 goto tx_err;
1043
1044 fl6.daddr = t->parms.raddr;
1045 }
1046
1047 /* Push GRE header. */
1048 gre_build_header(skb, 8, TUNNEL_SEQ,
1049 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
1050
1051 /* TooBig packet may have updated dst->dev's mtu */
1052 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1053 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1054
1055 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1056 NEXTHDR_GRE);
1057 if (err != 0) {
1058 /* XXX: send ICMP error even if DF is not set. */
1059 if (err == -EMSGSIZE) {
1060 if (skb->protocol == htons(ETH_P_IP))
1061 icmp_send(skb, ICMP_DEST_UNREACH,
1062 ICMP_FRAG_NEEDED, htonl(mtu));
1063 else
1064 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1065 }
1066
1067 goto tx_err;
1068 }
1069 return NETDEV_TX_OK;
1070
1071 tx_err:
1072 stats = &t->dev->stats;
1073 stats->tx_errors++;
1074 stats->tx_dropped++;
1075 kfree_skb(skb);
1076 return NETDEV_TX_OK;
1077 }
1078
1079 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1080 {
1081 struct net_device *dev = t->dev;
1082 struct __ip6_tnl_parm *p = &t->parms;
1083 struct flowi6 *fl6 = &t->fl.u.ip6;
1084
1085 if (dev->type != ARPHRD_ETHER) {
1086 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1087 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1088 }
1089
1090 /* Set up flowi template */
1091 fl6->saddr = p->laddr;
1092 fl6->daddr = p->raddr;
1093 fl6->flowi6_oif = p->link;
1094 fl6->flowlabel = 0;
1095 fl6->flowi6_proto = IPPROTO_GRE;
1096
1097 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1098 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1099 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1100 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1101
1102 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1103 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1104
1105 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1106 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1107 dev->flags |= IFF_POINTOPOINT;
1108 else
1109 dev->flags &= ~IFF_POINTOPOINT;
1110 }
1111
1112 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1113 int t_hlen)
1114 {
1115 const struct __ip6_tnl_parm *p = &t->parms;
1116 struct net_device *dev = t->dev;
1117
1118 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1119 int strict = (ipv6_addr_type(&p->raddr) &
1120 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1121
1122 struct rt6_info *rt = rt6_lookup(t->net,
1123 &p->raddr, &p->laddr,
1124 p->link, NULL, strict);
1125
1126 if (!rt)
1127 return;
1128
1129 if (rt->dst.dev) {
1130 dev->hard_header_len = rt->dst.dev->hard_header_len +
1131 t_hlen;
1132
1133 if (set_mtu) {
1134 dev->mtu = rt->dst.dev->mtu - t_hlen;
1135 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1136 dev->mtu -= 8;
1137 if (dev->type == ARPHRD_ETHER)
1138 dev->mtu -= ETH_HLEN;
1139
1140 if (dev->mtu < IPV6_MIN_MTU)
1141 dev->mtu = IPV6_MIN_MTU;
1142 }
1143 }
1144 ip6_rt_put(rt);
1145 }
1146 }
1147
1148 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1149 {
1150 int t_hlen;
1151
1152 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1153 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1154
1155 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1156 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1157 return t_hlen;
1158 }
1159
1160 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1161 {
1162 ip6gre_tnl_link_config_common(t);
1163 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1164 }
1165
1166 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1167 const struct __ip6_tnl_parm *p)
1168 {
1169 t->parms.laddr = p->laddr;
1170 t->parms.raddr = p->raddr;
1171 t->parms.flags = p->flags;
1172 t->parms.hop_limit = p->hop_limit;
1173 t->parms.encap_limit = p->encap_limit;
1174 t->parms.flowinfo = p->flowinfo;
1175 t->parms.link = p->link;
1176 t->parms.proto = p->proto;
1177 t->parms.i_key = p->i_key;
1178 t->parms.o_key = p->o_key;
1179 t->parms.i_flags = p->i_flags;
1180 t->parms.o_flags = p->o_flags;
1181 t->parms.fwmark = p->fwmark;
1182 dst_cache_reset(&t->dst_cache);
1183 }
1184
1185 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1186 int set_mtu)
1187 {
1188 ip6gre_tnl_copy_tnl_parm(t, p);
1189 ip6gre_tnl_link_config(t, set_mtu);
1190 return 0;
1191 }
1192
1193 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1194 const struct ip6_tnl_parm2 *u)
1195 {
1196 p->laddr = u->laddr;
1197 p->raddr = u->raddr;
1198 p->flags = u->flags;
1199 p->hop_limit = u->hop_limit;
1200 p->encap_limit = u->encap_limit;
1201 p->flowinfo = u->flowinfo;
1202 p->link = u->link;
1203 p->i_key = u->i_key;
1204 p->o_key = u->o_key;
1205 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1206 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1207 memcpy(p->name, u->name, sizeof(u->name));
1208 }
1209
1210 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1211 const struct __ip6_tnl_parm *p)
1212 {
1213 u->proto = IPPROTO_GRE;
1214 u->laddr = p->laddr;
1215 u->raddr = p->raddr;
1216 u->flags = p->flags;
1217 u->hop_limit = p->hop_limit;
1218 u->encap_limit = p->encap_limit;
1219 u->flowinfo = p->flowinfo;
1220 u->link = p->link;
1221 u->i_key = p->i_key;
1222 u->o_key = p->o_key;
1223 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1224 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1225 memcpy(u->name, p->name, sizeof(u->name));
1226 }
1227
1228 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1229 struct ifreq *ifr, int cmd)
1230 {
1231 int err = 0;
1232 struct ip6_tnl_parm2 p;
1233 struct __ip6_tnl_parm p1;
1234 struct ip6_tnl *t = netdev_priv(dev);
1235 struct net *net = t->net;
1236 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1237
1238 memset(&p1, 0, sizeof(p1));
1239
1240 switch (cmd) {
1241 case SIOCGETTUNNEL:
1242 if (dev == ign->fb_tunnel_dev) {
1243 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1244 err = -EFAULT;
1245 break;
1246 }
1247 ip6gre_tnl_parm_from_user(&p1, &p);
1248 t = ip6gre_tunnel_locate(net, &p1, 0);
1249 if (!t)
1250 t = netdev_priv(dev);
1251 }
1252 memset(&p, 0, sizeof(p));
1253 ip6gre_tnl_parm_to_user(&p, &t->parms);
1254 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1255 err = -EFAULT;
1256 break;
1257
1258 case SIOCADDTUNNEL:
1259 case SIOCCHGTUNNEL:
1260 err = -EPERM;
1261 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1262 goto done;
1263
1264 err = -EFAULT;
1265 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1266 goto done;
1267
1268 err = -EINVAL;
1269 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1270 goto done;
1271
1272 if (!(p.i_flags&GRE_KEY))
1273 p.i_key = 0;
1274 if (!(p.o_flags&GRE_KEY))
1275 p.o_key = 0;
1276
1277 ip6gre_tnl_parm_from_user(&p1, &p);
1278 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1279
1280 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1281 if (t) {
1282 if (t->dev != dev) {
1283 err = -EEXIST;
1284 break;
1285 }
1286 } else {
1287 t = netdev_priv(dev);
1288
1289 ip6gre_tunnel_unlink(ign, t);
1290 synchronize_net();
1291 ip6gre_tnl_change(t, &p1, 1);
1292 ip6gre_tunnel_link(ign, t);
1293 netdev_state_change(dev);
1294 }
1295 }
1296
1297 if (t) {
1298 err = 0;
1299
1300 memset(&p, 0, sizeof(p));
1301 ip6gre_tnl_parm_to_user(&p, &t->parms);
1302 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1303 err = -EFAULT;
1304 } else
1305 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1306 break;
1307
1308 case SIOCDELTUNNEL:
1309 err = -EPERM;
1310 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1311 goto done;
1312
1313 if (dev == ign->fb_tunnel_dev) {
1314 err = -EFAULT;
1315 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1316 goto done;
1317 err = -ENOENT;
1318 ip6gre_tnl_parm_from_user(&p1, &p);
1319 t = ip6gre_tunnel_locate(net, &p1, 0);
1320 if (!t)
1321 goto done;
1322 err = -EPERM;
1323 if (t == netdev_priv(ign->fb_tunnel_dev))
1324 goto done;
1325 dev = t->dev;
1326 }
1327 unregister_netdevice(dev);
1328 err = 0;
1329 break;
1330
1331 default:
1332 err = -EINVAL;
1333 }
1334
1335 done:
1336 return err;
1337 }
1338
1339 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1340 unsigned short type, const void *daddr,
1341 const void *saddr, unsigned int len)
1342 {
1343 struct ip6_tnl *t = netdev_priv(dev);
1344 struct ipv6hdr *ipv6h;
1345 __be16 *p;
1346
1347 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1348 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1349 t->fl.u.ip6.flowlabel,
1350 true, &t->fl.u.ip6));
1351 ipv6h->hop_limit = t->parms.hop_limit;
1352 ipv6h->nexthdr = NEXTHDR_GRE;
1353 ipv6h->saddr = t->parms.laddr;
1354 ipv6h->daddr = t->parms.raddr;
1355
1356 p = (__be16 *)(ipv6h + 1);
1357 p[0] = t->parms.o_flags;
1358 p[1] = htons(type);
1359
1360 /*
1361 * Set the source hardware address.
1362 */
1363
1364 if (saddr)
1365 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1366 if (daddr)
1367 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1368 if (!ipv6_addr_any(&ipv6h->daddr))
1369 return t->hlen;
1370
1371 return -t->hlen;
1372 }
1373
1374 static const struct header_ops ip6gre_header_ops = {
1375 .create = ip6gre_header,
1376 };
1377
1378 static const struct net_device_ops ip6gre_netdev_ops = {
1379 .ndo_init = ip6gre_tunnel_init,
1380 .ndo_uninit = ip6gre_tunnel_uninit,
1381 .ndo_start_xmit = ip6gre_tunnel_xmit,
1382 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1383 .ndo_change_mtu = ip6_tnl_change_mtu,
1384 .ndo_get_stats64 = ip_tunnel_get_stats64,
1385 .ndo_get_iflink = ip6_tnl_get_iflink,
1386 };
1387
1388 static void ip6gre_dev_free(struct net_device *dev)
1389 {
1390 struct ip6_tnl *t = netdev_priv(dev);
1391
1392 gro_cells_destroy(&t->gro_cells);
1393 dst_cache_destroy(&t->dst_cache);
1394 free_percpu(dev->tstats);
1395 }
1396
1397 static void ip6gre_tunnel_setup(struct net_device *dev)
1398 {
1399 dev->netdev_ops = &ip6gre_netdev_ops;
1400 dev->needs_free_netdev = true;
1401 dev->priv_destructor = ip6gre_dev_free;
1402
1403 dev->type = ARPHRD_IP6GRE;
1404
1405 dev->flags |= IFF_NOARP;
1406 dev->addr_len = sizeof(struct in6_addr);
1407 netif_keep_dst(dev);
1408 /* This perm addr will be used as interface identifier by IPv6 */
1409 dev->addr_assign_type = NET_ADDR_RANDOM;
1410 eth_random_addr(dev->perm_addr);
1411 }
1412
1413 #define GRE6_FEATURES (NETIF_F_SG | \
1414 NETIF_F_FRAGLIST | \
1415 NETIF_F_HIGHDMA | \
1416 NETIF_F_HW_CSUM)
1417
1418 static void ip6gre_tnl_init_features(struct net_device *dev)
1419 {
1420 struct ip6_tnl *nt = netdev_priv(dev);
1421
1422 dev->features |= GRE6_FEATURES;
1423 dev->hw_features |= GRE6_FEATURES;
1424
1425 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1426 /* TCP offload with GRE SEQ is not supported, nor
1427 * can we support 2 levels of outer headers requiring
1428 * an update.
1429 */
1430 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1431 nt->encap.type == TUNNEL_ENCAP_NONE) {
1432 dev->features |= NETIF_F_GSO_SOFTWARE;
1433 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1434 }
1435
1436 /* Can use a lockless transmit, unless we generate
1437 * output sequences
1438 */
1439 dev->features |= NETIF_F_LLTX;
1440 }
1441 }
1442
1443 static int ip6gre_tunnel_init_common(struct net_device *dev)
1444 {
1445 struct ip6_tnl *tunnel;
1446 int ret;
1447 int t_hlen;
1448
1449 tunnel = netdev_priv(dev);
1450
1451 tunnel->dev = dev;
1452 tunnel->net = dev_net(dev);
1453 strcpy(tunnel->parms.name, dev->name);
1454
1455 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1456 if (!dev->tstats)
1457 return -ENOMEM;
1458
1459 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1460 if (ret)
1461 goto cleanup_alloc_pcpu_stats;
1462
1463 ret = gro_cells_init(&tunnel->gro_cells, dev);
1464 if (ret)
1465 goto cleanup_dst_cache_init;
1466
1467 t_hlen = ip6gre_calc_hlen(tunnel);
1468 dev->mtu = ETH_DATA_LEN - t_hlen;
1469 if (dev->type == ARPHRD_ETHER)
1470 dev->mtu -= ETH_HLEN;
1471 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1472 dev->mtu -= 8;
1473
1474 if (tunnel->parms.collect_md) {
1475 dev->features |= NETIF_F_NETNS_LOCAL;
1476 netif_keep_dst(dev);
1477 }
1478 ip6gre_tnl_init_features(dev);
1479
1480 return 0;
1481
1482 cleanup_dst_cache_init:
1483 dst_cache_destroy(&tunnel->dst_cache);
1484 cleanup_alloc_pcpu_stats:
1485 free_percpu(dev->tstats);
1486 dev->tstats = NULL;
1487 return ret;
1488 }
1489
1490 static int ip6gre_tunnel_init(struct net_device *dev)
1491 {
1492 struct ip6_tnl *tunnel;
1493 int ret;
1494
1495 ret = ip6gre_tunnel_init_common(dev);
1496 if (ret)
1497 return ret;
1498
1499 tunnel = netdev_priv(dev);
1500
1501 if (tunnel->parms.collect_md)
1502 return 0;
1503
1504 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1505 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1506
1507 if (ipv6_addr_any(&tunnel->parms.raddr))
1508 dev->header_ops = &ip6gre_header_ops;
1509
1510 return 0;
1511 }
1512
1513 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1514 {
1515 struct ip6_tnl *tunnel = netdev_priv(dev);
1516
1517 tunnel->dev = dev;
1518 tunnel->net = dev_net(dev);
1519 strcpy(tunnel->parms.name, dev->name);
1520
1521 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1522
1523 dev_hold(dev);
1524 }
1525
1526 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1527 .handler = gre_rcv,
1528 .err_handler = ip6gre_err,
1529 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1530 };
1531
1532 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1533 {
1534 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1535 struct net_device *dev, *aux;
1536 int prio;
1537
1538 for_each_netdev_safe(net, dev, aux)
1539 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1540 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1541 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1542 unregister_netdevice_queue(dev, head);
1543
1544 for (prio = 0; prio < 4; prio++) {
1545 int h;
1546 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1547 struct ip6_tnl *t;
1548
1549 t = rtnl_dereference(ign->tunnels[prio][h]);
1550
1551 while (t) {
1552 /* If dev is in the same netns, it has already
1553 * been added to the list by the previous loop.
1554 */
1555 if (!net_eq(dev_net(t->dev), net))
1556 unregister_netdevice_queue(t->dev,
1557 head);
1558 t = rtnl_dereference(t->next);
1559 }
1560 }
1561 }
1562 }
1563
1564 static int __net_init ip6gre_init_net(struct net *net)
1565 {
1566 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1567 int err;
1568
1569 if (!net_has_fallback_tunnels(net))
1570 return 0;
1571 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1572 NET_NAME_UNKNOWN,
1573 ip6gre_tunnel_setup);
1574 if (!ign->fb_tunnel_dev) {
1575 err = -ENOMEM;
1576 goto err_alloc_dev;
1577 }
1578 dev_net_set(ign->fb_tunnel_dev, net);
1579 /* FB netdevice is special: we have one, and only one per netns.
1580 * Allowing to move it to another netns is clearly unsafe.
1581 */
1582 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1583
1584
1585 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1586 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1587
1588 err = register_netdev(ign->fb_tunnel_dev);
1589 if (err)
1590 goto err_reg_dev;
1591
1592 rcu_assign_pointer(ign->tunnels_wc[0],
1593 netdev_priv(ign->fb_tunnel_dev));
1594 return 0;
1595
1596 err_reg_dev:
1597 free_netdev(ign->fb_tunnel_dev);
1598 err_alloc_dev:
1599 return err;
1600 }
1601
1602 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1603 {
1604 struct net *net;
1605 LIST_HEAD(list);
1606
1607 rtnl_lock();
1608 list_for_each_entry(net, net_list, exit_list)
1609 ip6gre_destroy_tunnels(net, &list);
1610 unregister_netdevice_many(&list);
1611 rtnl_unlock();
1612 }
1613
1614 static struct pernet_operations ip6gre_net_ops = {
1615 .init = ip6gre_init_net,
1616 .exit_batch = ip6gre_exit_batch_net,
1617 .id = &ip6gre_net_id,
1618 .size = sizeof(struct ip6gre_net),
1619 };
1620
1621 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1622 struct netlink_ext_ack *extack)
1623 {
1624 __be16 flags;
1625
1626 if (!data)
1627 return 0;
1628
1629 flags = 0;
1630 if (data[IFLA_GRE_IFLAGS])
1631 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1632 if (data[IFLA_GRE_OFLAGS])
1633 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1634 if (flags & (GRE_VERSION|GRE_ROUTING))
1635 return -EINVAL;
1636
1637 return 0;
1638 }
1639
1640 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1641 struct netlink_ext_ack *extack)
1642 {
1643 struct in6_addr daddr;
1644
1645 if (tb[IFLA_ADDRESS]) {
1646 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1647 return -EINVAL;
1648 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1649 return -EADDRNOTAVAIL;
1650 }
1651
1652 if (!data)
1653 goto out;
1654
1655 if (data[IFLA_GRE_REMOTE]) {
1656 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1657 if (ipv6_addr_any(&daddr))
1658 return -EINVAL;
1659 }
1660
1661 out:
1662 return ip6gre_tunnel_validate(tb, data, extack);
1663 }
1664
1665 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1666 struct netlink_ext_ack *extack)
1667 {
1668 __be16 flags = 0;
1669 int ret, ver = 0;
1670
1671 if (!data)
1672 return 0;
1673
1674 ret = ip6gre_tap_validate(tb, data, extack);
1675 if (ret)
1676 return ret;
1677
1678 /* ERSPAN should only have GRE sequence and key flag */
1679 if (data[IFLA_GRE_OFLAGS])
1680 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1681 if (data[IFLA_GRE_IFLAGS])
1682 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1683 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1684 flags != (GRE_SEQ | GRE_KEY))
1685 return -EINVAL;
1686
1687 /* ERSPAN Session ID only has 10-bit. Since we reuse
1688 * 32-bit key field as ID, check it's range.
1689 */
1690 if (data[IFLA_GRE_IKEY] &&
1691 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1692 return -EINVAL;
1693
1694 if (data[IFLA_GRE_OKEY] &&
1695 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1696 return -EINVAL;
1697
1698 if (data[IFLA_GRE_ERSPAN_VER]) {
1699 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1700 if (ver != 1 && ver != 2)
1701 return -EINVAL;
1702 }
1703
1704 if (ver == 1) {
1705 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1706 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1707
1708 if (index & ~INDEX_MASK)
1709 return -EINVAL;
1710 }
1711 } else if (ver == 2) {
1712 if (data[IFLA_GRE_ERSPAN_DIR]) {
1713 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1714
1715 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1716 return -EINVAL;
1717 }
1718
1719 if (data[IFLA_GRE_ERSPAN_HWID]) {
1720 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1721
1722 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1723 return -EINVAL;
1724 }
1725 }
1726
1727 return 0;
1728 }
1729
1730 static void ip6gre_netlink_parms(struct nlattr *data[],
1731 struct __ip6_tnl_parm *parms)
1732 {
1733 memset(parms, 0, sizeof(*parms));
1734
1735 if (!data)
1736 return;
1737
1738 if (data[IFLA_GRE_LINK])
1739 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1740
1741 if (data[IFLA_GRE_IFLAGS])
1742 parms->i_flags = gre_flags_to_tnl_flags(
1743 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1744
1745 if (data[IFLA_GRE_OFLAGS])
1746 parms->o_flags = gre_flags_to_tnl_flags(
1747 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1748
1749 if (data[IFLA_GRE_IKEY])
1750 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1751
1752 if (data[IFLA_GRE_OKEY])
1753 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1754
1755 if (data[IFLA_GRE_LOCAL])
1756 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1757
1758 if (data[IFLA_GRE_REMOTE])
1759 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1760
1761 if (data[IFLA_GRE_TTL])
1762 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1763
1764 if (data[IFLA_GRE_ENCAP_LIMIT])
1765 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1766
1767 if (data[IFLA_GRE_FLOWINFO])
1768 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1769
1770 if (data[IFLA_GRE_FLAGS])
1771 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1772
1773 if (data[IFLA_GRE_FWMARK])
1774 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1775
1776 if (data[IFLA_GRE_COLLECT_METADATA])
1777 parms->collect_md = true;
1778
1779 if (data[IFLA_GRE_ERSPAN_VER])
1780 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1781
1782 if (parms->erspan_ver == 1) {
1783 if (data[IFLA_GRE_ERSPAN_INDEX])
1784 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1785 } else if (parms->erspan_ver == 2) {
1786 if (data[IFLA_GRE_ERSPAN_DIR])
1787 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1788 if (data[IFLA_GRE_ERSPAN_HWID])
1789 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1790 }
1791 }
1792
1793 static int ip6gre_tap_init(struct net_device *dev)
1794 {
1795 int ret;
1796
1797 ret = ip6gre_tunnel_init_common(dev);
1798 if (ret)
1799 return ret;
1800
1801 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1802
1803 return 0;
1804 }
1805
1806 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1807 .ndo_init = ip6gre_tap_init,
1808 .ndo_uninit = ip6gre_tunnel_uninit,
1809 .ndo_start_xmit = ip6gre_tunnel_xmit,
1810 .ndo_set_mac_address = eth_mac_addr,
1811 .ndo_validate_addr = eth_validate_addr,
1812 .ndo_change_mtu = ip6_tnl_change_mtu,
1813 .ndo_get_stats64 = ip_tunnel_get_stats64,
1814 .ndo_get_iflink = ip6_tnl_get_iflink,
1815 };
1816
1817 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1818 {
1819 int t_hlen;
1820
1821 tunnel->tun_hlen = 8;
1822 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1823 erspan_hdr_len(tunnel->parms.erspan_ver);
1824
1825 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1826 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1827 return t_hlen;
1828 }
1829
1830 static int ip6erspan_tap_init(struct net_device *dev)
1831 {
1832 struct ip6_tnl *tunnel;
1833 int t_hlen;
1834 int ret;
1835
1836 tunnel = netdev_priv(dev);
1837
1838 tunnel->dev = dev;
1839 tunnel->net = dev_net(dev);
1840 strcpy(tunnel->parms.name, dev->name);
1841
1842 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1843 if (!dev->tstats)
1844 return -ENOMEM;
1845
1846 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1847 if (ret)
1848 goto cleanup_alloc_pcpu_stats;
1849
1850 ret = gro_cells_init(&tunnel->gro_cells, dev);
1851 if (ret)
1852 goto cleanup_dst_cache_init;
1853
1854 t_hlen = ip6erspan_calc_hlen(tunnel);
1855 dev->mtu = ETH_DATA_LEN - t_hlen;
1856 if (dev->type == ARPHRD_ETHER)
1857 dev->mtu -= ETH_HLEN;
1858 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1859 dev->mtu -= 8;
1860
1861 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1862 ip6erspan_tnl_link_config(tunnel, 1);
1863
1864 return 0;
1865
1866 cleanup_dst_cache_init:
1867 dst_cache_destroy(&tunnel->dst_cache);
1868 cleanup_alloc_pcpu_stats:
1869 free_percpu(dev->tstats);
1870 dev->tstats = NULL;
1871 return ret;
1872 }
1873
1874 static const struct net_device_ops ip6erspan_netdev_ops = {
1875 .ndo_init = ip6erspan_tap_init,
1876 .ndo_uninit = ip6erspan_tunnel_uninit,
1877 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1878 .ndo_set_mac_address = eth_mac_addr,
1879 .ndo_validate_addr = eth_validate_addr,
1880 .ndo_change_mtu = ip6_tnl_change_mtu,
1881 .ndo_get_stats64 = ip_tunnel_get_stats64,
1882 .ndo_get_iflink = ip6_tnl_get_iflink,
1883 };
1884
1885 static void ip6gre_tap_setup(struct net_device *dev)
1886 {
1887
1888 ether_setup(dev);
1889
1890 dev->max_mtu = 0;
1891 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1892 dev->needs_free_netdev = true;
1893 dev->priv_destructor = ip6gre_dev_free;
1894
1895 dev->features |= NETIF_F_NETNS_LOCAL;
1896 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1897 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1898 netif_keep_dst(dev);
1899 }
1900
1901 bool is_ip6gretap_dev(const struct net_device *dev)
1902 {
1903 return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1904 }
1905 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1906
1907 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1908 struct ip_tunnel_encap *ipencap)
1909 {
1910 bool ret = false;
1911
1912 memset(ipencap, 0, sizeof(*ipencap));
1913
1914 if (!data)
1915 return ret;
1916
1917 if (data[IFLA_GRE_ENCAP_TYPE]) {
1918 ret = true;
1919 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1920 }
1921
1922 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1923 ret = true;
1924 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1925 }
1926
1927 if (data[IFLA_GRE_ENCAP_SPORT]) {
1928 ret = true;
1929 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1930 }
1931
1932 if (data[IFLA_GRE_ENCAP_DPORT]) {
1933 ret = true;
1934 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1935 }
1936
1937 return ret;
1938 }
1939
1940 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev,
1941 struct nlattr *tb[], struct nlattr *data[],
1942 struct netlink_ext_ack *extack)
1943 {
1944 struct ip6_tnl *nt;
1945 struct ip_tunnel_encap ipencap;
1946 int err;
1947
1948 nt = netdev_priv(dev);
1949
1950 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1951 int err = ip6_tnl_encap_setup(nt, &ipencap);
1952
1953 if (err < 0)
1954 return err;
1955 }
1956
1957 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1958 eth_hw_addr_random(dev);
1959
1960 nt->dev = dev;
1961 nt->net = dev_net(dev);
1962
1963 err = register_netdevice(dev);
1964 if (err)
1965 goto out;
1966
1967 if (tb[IFLA_MTU])
1968 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1969
1970 dev_hold(dev);
1971
1972 out:
1973 return err;
1974 }
1975
1976 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1977 struct nlattr *tb[], struct nlattr *data[],
1978 struct netlink_ext_ack *extack)
1979 {
1980 struct ip6_tnl *nt = netdev_priv(dev);
1981 struct net *net = dev_net(dev);
1982 struct ip6gre_net *ign;
1983 int err;
1984
1985 ip6gre_netlink_parms(data, &nt->parms);
1986 ign = net_generic(net, ip6gre_net_id);
1987
1988 if (nt->parms.collect_md) {
1989 if (rtnl_dereference(ign->collect_md_tun))
1990 return -EEXIST;
1991 } else {
1992 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1993 return -EEXIST;
1994 }
1995
1996 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
1997 if (!err) {
1998 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1999 ip6gre_tunnel_link_md(ign, nt);
2000 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2001 }
2002 return err;
2003 }
2004
2005 static struct ip6_tnl *
2006 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2007 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2008 struct netlink_ext_ack *extack)
2009 {
2010 struct ip6_tnl *t, *nt = netdev_priv(dev);
2011 struct net *net = nt->net;
2012 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2013 struct ip_tunnel_encap ipencap;
2014
2015 if (dev == ign->fb_tunnel_dev)
2016 return ERR_PTR(-EINVAL);
2017
2018 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2019 int err = ip6_tnl_encap_setup(nt, &ipencap);
2020
2021 if (err < 0)
2022 return ERR_PTR(err);
2023 }
2024
2025 ip6gre_netlink_parms(data, p_p);
2026
2027 t = ip6gre_tunnel_locate(net, p_p, 0);
2028
2029 if (t) {
2030 if (t->dev != dev)
2031 return ERR_PTR(-EEXIST);
2032 } else {
2033 t = nt;
2034 }
2035
2036 return t;
2037 }
2038
2039 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2040 struct nlattr *data[],
2041 struct netlink_ext_ack *extack)
2042 {
2043 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2044 struct __ip6_tnl_parm p;
2045 struct ip6_tnl *t;
2046
2047 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2048 if (IS_ERR(t))
2049 return PTR_ERR(t);
2050
2051 ip6gre_tunnel_unlink_md(ign, t);
2052 ip6gre_tunnel_unlink(ign, t);
2053 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2054 ip6gre_tunnel_link_md(ign, t);
2055 ip6gre_tunnel_link(ign, t);
2056 return 0;
2057 }
2058
2059 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2060 {
2061 struct net *net = dev_net(dev);
2062 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2063
2064 if (dev != ign->fb_tunnel_dev)
2065 unregister_netdevice_queue(dev, head);
2066 }
2067
2068 static size_t ip6gre_get_size(const struct net_device *dev)
2069 {
2070 return
2071 /* IFLA_GRE_LINK */
2072 nla_total_size(4) +
2073 /* IFLA_GRE_IFLAGS */
2074 nla_total_size(2) +
2075 /* IFLA_GRE_OFLAGS */
2076 nla_total_size(2) +
2077 /* IFLA_GRE_IKEY */
2078 nla_total_size(4) +
2079 /* IFLA_GRE_OKEY */
2080 nla_total_size(4) +
2081 /* IFLA_GRE_LOCAL */
2082 nla_total_size(sizeof(struct in6_addr)) +
2083 /* IFLA_GRE_REMOTE */
2084 nla_total_size(sizeof(struct in6_addr)) +
2085 /* IFLA_GRE_TTL */
2086 nla_total_size(1) +
2087 /* IFLA_GRE_ENCAP_LIMIT */
2088 nla_total_size(1) +
2089 /* IFLA_GRE_FLOWINFO */
2090 nla_total_size(4) +
2091 /* IFLA_GRE_FLAGS */
2092 nla_total_size(4) +
2093 /* IFLA_GRE_ENCAP_TYPE */
2094 nla_total_size(2) +
2095 /* IFLA_GRE_ENCAP_FLAGS */
2096 nla_total_size(2) +
2097 /* IFLA_GRE_ENCAP_SPORT */
2098 nla_total_size(2) +
2099 /* IFLA_GRE_ENCAP_DPORT */
2100 nla_total_size(2) +
2101 /* IFLA_GRE_COLLECT_METADATA */
2102 nla_total_size(0) +
2103 /* IFLA_GRE_FWMARK */
2104 nla_total_size(4) +
2105 /* IFLA_GRE_ERSPAN_INDEX */
2106 nla_total_size(4) +
2107 0;
2108 }
2109
2110 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2111 {
2112 struct ip6_tnl *t = netdev_priv(dev);
2113 struct __ip6_tnl_parm *p = &t->parms;
2114
2115 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2116 nla_put_be16(skb, IFLA_GRE_IFLAGS,
2117 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2118 nla_put_be16(skb, IFLA_GRE_OFLAGS,
2119 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
2120 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2121 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2122 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2123 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2124 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2125 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2126 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2127 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2128 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
2129 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2130 goto nla_put_failure;
2131
2132 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2133 t->encap.type) ||
2134 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2135 t->encap.sport) ||
2136 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2137 t->encap.dport) ||
2138 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2139 t->encap.flags))
2140 goto nla_put_failure;
2141
2142 if (p->collect_md) {
2143 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2144 goto nla_put_failure;
2145 }
2146
2147 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2148 goto nla_put_failure;
2149
2150 if (p->erspan_ver == 1) {
2151 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2152 goto nla_put_failure;
2153 } else if (p->erspan_ver == 2) {
2154 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2155 goto nla_put_failure;
2156 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2157 goto nla_put_failure;
2158 }
2159
2160 return 0;
2161
2162 nla_put_failure:
2163 return -EMSGSIZE;
2164 }
2165
2166 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2167 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2168 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2169 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2170 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2171 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2172 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2173 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2174 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2175 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2176 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2177 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2178 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2179 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2180 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2181 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2182 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2183 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2184 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2185 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2186 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2187 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2188 };
2189
2190 static void ip6erspan_tap_setup(struct net_device *dev)
2191 {
2192 ether_setup(dev);
2193
2194 dev->netdev_ops = &ip6erspan_netdev_ops;
2195 dev->needs_free_netdev = true;
2196 dev->priv_destructor = ip6gre_dev_free;
2197
2198 dev->features |= NETIF_F_NETNS_LOCAL;
2199 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2200 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2201 netif_keep_dst(dev);
2202 }
2203
2204 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev,
2205 struct nlattr *tb[], struct nlattr *data[],
2206 struct netlink_ext_ack *extack)
2207 {
2208 struct ip6_tnl *nt = netdev_priv(dev);
2209 struct net *net = dev_net(dev);
2210 struct ip6gre_net *ign;
2211 int err;
2212
2213 ip6gre_netlink_parms(data, &nt->parms);
2214 ign = net_generic(net, ip6gre_net_id);
2215
2216 if (nt->parms.collect_md) {
2217 if (rtnl_dereference(ign->collect_md_tun_erspan))
2218 return -EEXIST;
2219 } else {
2220 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2221 return -EEXIST;
2222 }
2223
2224 err = ip6gre_newlink_common(src_net, dev, tb, data, extack);
2225 if (!err) {
2226 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2227 ip6erspan_tunnel_link_md(ign, nt);
2228 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2229 }
2230 return err;
2231 }
2232
2233 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2234 {
2235 ip6gre_tnl_link_config_common(t);
2236 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2237 }
2238
2239 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2240 const struct __ip6_tnl_parm *p, int set_mtu)
2241 {
2242 ip6gre_tnl_copy_tnl_parm(t, p);
2243 ip6erspan_tnl_link_config(t, set_mtu);
2244 return 0;
2245 }
2246
2247 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2248 struct nlattr *data[],
2249 struct netlink_ext_ack *extack)
2250 {
2251 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id);
2252 struct __ip6_tnl_parm p;
2253 struct ip6_tnl *t;
2254
2255 t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2256 if (IS_ERR(t))
2257 return PTR_ERR(t);
2258
2259 ip6gre_tunnel_unlink_md(ign, t);
2260 ip6gre_tunnel_unlink(ign, t);
2261 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2262 ip6erspan_tunnel_link_md(ign, t);
2263 ip6gre_tunnel_link(ign, t);
2264 return 0;
2265 }
2266
2267 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2268 .kind = "ip6gre",
2269 .maxtype = IFLA_GRE_MAX,
2270 .policy = ip6gre_policy,
2271 .priv_size = sizeof(struct ip6_tnl),
2272 .setup = ip6gre_tunnel_setup,
2273 .validate = ip6gre_tunnel_validate,
2274 .newlink = ip6gre_newlink,
2275 .changelink = ip6gre_changelink,
2276 .dellink = ip6gre_dellink,
2277 .get_size = ip6gre_get_size,
2278 .fill_info = ip6gre_fill_info,
2279 .get_link_net = ip6_tnl_get_link_net,
2280 };
2281
2282 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2283 .kind = "ip6gretap",
2284 .maxtype = IFLA_GRE_MAX,
2285 .policy = ip6gre_policy,
2286 .priv_size = sizeof(struct ip6_tnl),
2287 .setup = ip6gre_tap_setup,
2288 .validate = ip6gre_tap_validate,
2289 .newlink = ip6gre_newlink,
2290 .changelink = ip6gre_changelink,
2291 .get_size = ip6gre_get_size,
2292 .fill_info = ip6gre_fill_info,
2293 .get_link_net = ip6_tnl_get_link_net,
2294 };
2295
2296 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2297 .kind = "ip6erspan",
2298 .maxtype = IFLA_GRE_MAX,
2299 .policy = ip6gre_policy,
2300 .priv_size = sizeof(struct ip6_tnl),
2301 .setup = ip6erspan_tap_setup,
2302 .validate = ip6erspan_tap_validate,
2303 .newlink = ip6erspan_newlink,
2304 .changelink = ip6erspan_changelink,
2305 .get_size = ip6gre_get_size,
2306 .fill_info = ip6gre_fill_info,
2307 .get_link_net = ip6_tnl_get_link_net,
2308 };
2309
2310 /*
2311 * And now the modules code and kernel interface.
2312 */
2313
2314 static int __init ip6gre_init(void)
2315 {
2316 int err;
2317
2318 pr_info("GRE over IPv6 tunneling driver\n");
2319
2320 err = register_pernet_device(&ip6gre_net_ops);
2321 if (err < 0)
2322 return err;
2323
2324 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2325 if (err < 0) {
2326 pr_info("%s: can't add protocol\n", __func__);
2327 goto add_proto_failed;
2328 }
2329
2330 err = rtnl_link_register(&ip6gre_link_ops);
2331 if (err < 0)
2332 goto rtnl_link_failed;
2333
2334 err = rtnl_link_register(&ip6gre_tap_ops);
2335 if (err < 0)
2336 goto tap_ops_failed;
2337
2338 err = rtnl_link_register(&ip6erspan_tap_ops);
2339 if (err < 0)
2340 goto erspan_link_failed;
2341
2342 out:
2343 return err;
2344
2345 erspan_link_failed:
2346 rtnl_link_unregister(&ip6gre_tap_ops);
2347 tap_ops_failed:
2348 rtnl_link_unregister(&ip6gre_link_ops);
2349 rtnl_link_failed:
2350 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2351 add_proto_failed:
2352 unregister_pernet_device(&ip6gre_net_ops);
2353 goto out;
2354 }
2355
2356 static void __exit ip6gre_fini(void)
2357 {
2358 rtnl_link_unregister(&ip6gre_tap_ops);
2359 rtnl_link_unregister(&ip6gre_link_ops);
2360 rtnl_link_unregister(&ip6erspan_tap_ops);
2361 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2362 unregister_pernet_device(&ip6gre_net_ops);
2363 }
2364
2365 module_init(ip6gre_init);
2366 module_exit(ip6gre_fini);
2367 MODULE_LICENSE("GPL");
2368 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2369 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2370 MODULE_ALIAS_RTNL_LINK("ip6gre");
2371 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2372 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2373 MODULE_ALIAS_NETDEV("ip6gre0");
(deprecated) Btrfs devel tree