]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ipv6/ip6_gre.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[mirror_ubuntu-jammy-kernel.git] / net / ipv6 / ip6_gre.c
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 net_device *fb_tunnel_dev;
75 };
76
77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
78 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
79 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
80 static int ip6gre_tunnel_init(struct net_device *dev);
81 static void ip6gre_tunnel_setup(struct net_device *dev);
82 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
83 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
84
85 /* Tunnel hash table */
86
87 /*
88 4 hash tables:
89
90 3: (remote,local)
91 2: (remote,*)
92 1: (*,local)
93 0: (*,*)
94
95 We require exact key match i.e. if a key is present in packet
96 it will match only tunnel with the same key; if it is not present,
97 it will match only keyless tunnel.
98
99 All keysless packets, if not matched configured keyless tunnels
100 will match fallback tunnel.
101 */
102
103 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
104 static u32 HASH_ADDR(const struct in6_addr *addr)
105 {
106 u32 hash = ipv6_addr_hash(addr);
107
108 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
109 }
110
111 #define tunnels_r_l tunnels[3]
112 #define tunnels_r tunnels[2]
113 #define tunnels_l tunnels[1]
114 #define tunnels_wc tunnels[0]
115
116 /* Given src, dst and key, find appropriate for input tunnel. */
117
118 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
119 const struct in6_addr *remote, const struct in6_addr *local,
120 __be32 key, __be16 gre_proto)
121 {
122 struct net *net = dev_net(dev);
123 int link = dev->ifindex;
124 unsigned int h0 = HASH_ADDR(remote);
125 unsigned int h1 = HASH_KEY(key);
126 struct ip6_tnl *t, *cand = NULL;
127 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
128 int dev_type = (gre_proto == htons(ETH_P_TEB) ||
129 gre_proto == htons(ETH_P_ERSPAN)) ?
130 ARPHRD_ETHER : ARPHRD_IP6GRE;
131 int score, cand_score = 4;
132
133 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
134 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
135 !ipv6_addr_equal(remote, &t->parms.raddr) ||
136 key != t->parms.i_key ||
137 !(t->dev->flags & IFF_UP))
138 continue;
139
140 if (t->dev->type != ARPHRD_IP6GRE &&
141 t->dev->type != dev_type)
142 continue;
143
144 score = 0;
145 if (t->parms.link != link)
146 score |= 1;
147 if (t->dev->type != dev_type)
148 score |= 2;
149 if (score == 0)
150 return t;
151
152 if (score < cand_score) {
153 cand = t;
154 cand_score = score;
155 }
156 }
157
158 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
159 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
160 key != t->parms.i_key ||
161 !(t->dev->flags & IFF_UP))
162 continue;
163
164 if (t->dev->type != ARPHRD_IP6GRE &&
165 t->dev->type != dev_type)
166 continue;
167
168 score = 0;
169 if (t->parms.link != link)
170 score |= 1;
171 if (t->dev->type != dev_type)
172 score |= 2;
173 if (score == 0)
174 return t;
175
176 if (score < cand_score) {
177 cand = t;
178 cand_score = score;
179 }
180 }
181
182 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
183 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
184 (!ipv6_addr_equal(local, &t->parms.raddr) ||
185 !ipv6_addr_is_multicast(local))) ||
186 key != t->parms.i_key ||
187 !(t->dev->flags & IFF_UP))
188 continue;
189
190 if (t->dev->type != ARPHRD_IP6GRE &&
191 t->dev->type != dev_type)
192 continue;
193
194 score = 0;
195 if (t->parms.link != link)
196 score |= 1;
197 if (t->dev->type != dev_type)
198 score |= 2;
199 if (score == 0)
200 return t;
201
202 if (score < cand_score) {
203 cand = t;
204 cand_score = score;
205 }
206 }
207
208 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
209 if (t->parms.i_key != key ||
210 !(t->dev->flags & IFF_UP))
211 continue;
212
213 if (t->dev->type != ARPHRD_IP6GRE &&
214 t->dev->type != dev_type)
215 continue;
216
217 score = 0;
218 if (t->parms.link != link)
219 score |= 1;
220 if (t->dev->type != dev_type)
221 score |= 2;
222 if (score == 0)
223 return t;
224
225 if (score < cand_score) {
226 cand = t;
227 cand_score = score;
228 }
229 }
230
231 if (cand)
232 return cand;
233
234 t = rcu_dereference(ign->collect_md_tun);
235 if (t && t->dev->flags & IFF_UP)
236 return t;
237
238 dev = ign->fb_tunnel_dev;
239 if (dev->flags & IFF_UP)
240 return netdev_priv(dev);
241
242 return NULL;
243 }
244
245 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
246 const struct __ip6_tnl_parm *p)
247 {
248 const struct in6_addr *remote = &p->raddr;
249 const struct in6_addr *local = &p->laddr;
250 unsigned int h = HASH_KEY(p->i_key);
251 int prio = 0;
252
253 if (!ipv6_addr_any(local))
254 prio |= 1;
255 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
256 prio |= 2;
257 h ^= HASH_ADDR(remote);
258 }
259
260 return &ign->tunnels[prio][h];
261 }
262
263 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
264 const struct ip6_tnl *t)
265 {
266 return __ip6gre_bucket(ign, &t->parms);
267 }
268
269 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
270 {
271 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
272
273 if (t->parms.collect_md)
274 rcu_assign_pointer(ign->collect_md_tun, t);
275
276 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
277 rcu_assign_pointer(*tp, t);
278 }
279
280 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
281 {
282 struct ip6_tnl __rcu **tp;
283 struct ip6_tnl *iter;
284
285 if (t->parms.collect_md)
286 rcu_assign_pointer(ign->collect_md_tun, NULL);
287
288 for (tp = ip6gre_bucket(ign, t);
289 (iter = rtnl_dereference(*tp)) != NULL;
290 tp = &iter->next) {
291 if (t == iter) {
292 rcu_assign_pointer(*tp, t->next);
293 break;
294 }
295 }
296 }
297
298 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
299 const struct __ip6_tnl_parm *parms,
300 int type)
301 {
302 const struct in6_addr *remote = &parms->raddr;
303 const struct in6_addr *local = &parms->laddr;
304 __be32 key = parms->i_key;
305 int link = parms->link;
306 struct ip6_tnl *t;
307 struct ip6_tnl __rcu **tp;
308 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
309
310 for (tp = __ip6gre_bucket(ign, parms);
311 (t = rtnl_dereference(*tp)) != NULL;
312 tp = &t->next)
313 if (ipv6_addr_equal(local, &t->parms.laddr) &&
314 ipv6_addr_equal(remote, &t->parms.raddr) &&
315 key == t->parms.i_key &&
316 link == t->parms.link &&
317 type == t->dev->type)
318 break;
319
320 return t;
321 }
322
323 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
324 const struct __ip6_tnl_parm *parms, int create)
325 {
326 struct ip6_tnl *t, *nt;
327 struct net_device *dev;
328 char name[IFNAMSIZ];
329 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
330
331 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
332 if (t && create)
333 return NULL;
334 if (t || !create)
335 return t;
336
337 if (parms->name[0])
338 strlcpy(name, parms->name, IFNAMSIZ);
339 else
340 strcpy(name, "ip6gre%d");
341
342 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
343 ip6gre_tunnel_setup);
344 if (!dev)
345 return NULL;
346
347 dev_net_set(dev, net);
348
349 nt = netdev_priv(dev);
350 nt->parms = *parms;
351 dev->rtnl_link_ops = &ip6gre_link_ops;
352
353 nt->dev = dev;
354 nt->net = dev_net(dev);
355
356 if (register_netdevice(dev) < 0)
357 goto failed_free;
358
359 ip6gre_tnl_link_config(nt, 1);
360
361 /* Can use a lockless transmit, unless we generate output sequences */
362 if (!(nt->parms.o_flags & TUNNEL_SEQ))
363 dev->features |= NETIF_F_LLTX;
364
365 dev_hold(dev);
366 ip6gre_tunnel_link(ign, nt);
367 return nt;
368
369 failed_free:
370 free_netdev(dev);
371 return NULL;
372 }
373
374 static void ip6gre_tunnel_uninit(struct net_device *dev)
375 {
376 struct ip6_tnl *t = netdev_priv(dev);
377 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
378
379 ip6gre_tunnel_unlink(ign, t);
380 dst_cache_reset(&t->dst_cache);
381 dev_put(dev);
382 }
383
384
385 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
386 u8 type, u8 code, int offset, __be32 info)
387 {
388 struct net *net = dev_net(skb->dev);
389 const struct gre_base_hdr *greh;
390 const struct ipv6hdr *ipv6h;
391 int grehlen = sizeof(*greh);
392 struct ip6_tnl *t;
393 int key_off = 0;
394 __be16 flags;
395 __be32 key;
396
397 if (!pskb_may_pull(skb, offset + grehlen))
398 return;
399 greh = (const struct gre_base_hdr *)(skb->data + offset);
400 flags = greh->flags;
401 if (flags & (GRE_VERSION | GRE_ROUTING))
402 return;
403 if (flags & GRE_CSUM)
404 grehlen += 4;
405 if (flags & GRE_KEY) {
406 key_off = grehlen + offset;
407 grehlen += 4;
408 }
409
410 if (!pskb_may_pull(skb, offset + grehlen))
411 return;
412 ipv6h = (const struct ipv6hdr *)skb->data;
413 greh = (const struct gre_base_hdr *)(skb->data + offset);
414 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
415
416 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
417 key, greh->protocol);
418 if (!t)
419 return;
420
421 switch (type) {
422 struct ipv6_tlv_tnl_enc_lim *tel;
423 __u32 teli;
424 case ICMPV6_DEST_UNREACH:
425 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
426 t->parms.name);
427 if (code != ICMPV6_PORT_UNREACH)
428 break;
429 return;
430 case ICMPV6_TIME_EXCEED:
431 if (code == ICMPV6_EXC_HOPLIMIT) {
432 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
433 t->parms.name);
434 break;
435 }
436 return;
437 case ICMPV6_PARAMPROB:
438 teli = 0;
439 if (code == ICMPV6_HDR_FIELD)
440 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
441
442 if (teli && teli == be32_to_cpu(info) - 2) {
443 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
444 if (tel->encap_limit == 0) {
445 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
446 t->parms.name);
447 }
448 } else {
449 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
450 t->parms.name);
451 }
452 return;
453 case ICMPV6_PKT_TOOBIG:
454 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
455 return;
456 case NDISC_REDIRECT:
457 ip6_redirect(skb, net, skb->dev->ifindex, 0,
458 sock_net_uid(net, NULL));
459 return;
460 }
461
462 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
463 t->err_count++;
464 else
465 t->err_count = 1;
466 t->err_time = jiffies;
467 }
468
469 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
470 {
471 const struct ipv6hdr *ipv6h;
472 struct ip6_tnl *tunnel;
473
474 ipv6h = ipv6_hdr(skb);
475 tunnel = ip6gre_tunnel_lookup(skb->dev,
476 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
477 tpi->proto);
478 if (tunnel) {
479 if (tunnel->parms.collect_md) {
480 struct metadata_dst *tun_dst;
481 __be64 tun_id;
482 __be16 flags;
483
484 flags = tpi->flags;
485 tun_id = key32_to_tunnel_id(tpi->key);
486
487 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
488 if (!tun_dst)
489 return PACKET_REJECT;
490
491 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
492 } else {
493 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
494 }
495
496 return PACKET_RCVD;
497 }
498
499 return PACKET_REJECT;
500 }
501
502 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
503 struct tnl_ptk_info *tpi)
504 {
505 struct erspan_base_hdr *ershdr;
506 struct erspan_metadata *pkt_md;
507 const struct ipv6hdr *ipv6h;
508 struct ip6_tnl *tunnel;
509 u8 ver;
510
511 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
512 return PACKET_REJECT;
513
514 ipv6h = ipv6_hdr(skb);
515 ershdr = (struct erspan_base_hdr *)skb->data;
516 ver = (ntohs(ershdr->ver_vlan) & VER_MASK) >> VER_OFFSET;
517 tpi->key = cpu_to_be32(ntohs(ershdr->session_id) & ID_MASK);
518
519 tunnel = ip6gre_tunnel_lookup(skb->dev,
520 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
521 tpi->proto);
522 if (tunnel) {
523 int len = erspan_hdr_len(ver);
524
525 if (unlikely(!pskb_may_pull(skb, len)))
526 return PACKET_REJECT;
527
528 ershdr = (struct erspan_base_hdr *)skb->data;
529 pkt_md = (struct erspan_metadata *)(ershdr + 1);
530
531 if (__iptunnel_pull_header(skb, len,
532 htons(ETH_P_TEB),
533 false, false) < 0)
534 return PACKET_REJECT;
535
536 if (tunnel->parms.collect_md) {
537 struct metadata_dst *tun_dst;
538 struct ip_tunnel_info *info;
539 struct erspan_metadata *md;
540 __be64 tun_id;
541 __be16 flags;
542
543 tpi->flags |= TUNNEL_KEY;
544 flags = tpi->flags;
545 tun_id = key32_to_tunnel_id(tpi->key);
546
547 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
548 sizeof(*md));
549 if (!tun_dst)
550 return PACKET_REJECT;
551
552 info = &tun_dst->u.tun_info;
553 md = ip_tunnel_info_opts(info);
554
555 memcpy(md, pkt_md, sizeof(*md));
556 md->version = ver;
557 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
558 info->options_len = sizeof(*md);
559
560 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
561
562 } else {
563 tunnel->parms.erspan_ver = ver;
564
565 if (ver == 1) {
566 tunnel->parms.index = ntohl(pkt_md->u.index);
567 } else {
568 u16 md2_flags;
569 u16 dir, hwid;
570
571 md2_flags = ntohs(pkt_md->u.md2.flags);
572 dir = (md2_flags & DIR_MASK) >> DIR_OFFSET;
573 hwid = (md2_flags & HWID_MASK) >> HWID_OFFSET;
574 tunnel->parms.dir = dir;
575 tunnel->parms.hwid = hwid;
576 }
577
578 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
579 }
580
581 return PACKET_RCVD;
582 }
583
584 return PACKET_REJECT;
585 }
586
587 static int gre_rcv(struct sk_buff *skb)
588 {
589 struct tnl_ptk_info tpi;
590 bool csum_err = false;
591 int hdr_len;
592
593 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
594 if (hdr_len < 0)
595 goto drop;
596
597 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
598 goto drop;
599
600 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
601 tpi.proto == htons(ETH_P_ERSPAN2))) {
602 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
603 return 0;
604 goto out;
605 }
606
607 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
608 return 0;
609
610 out:
611 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
612 drop:
613 kfree_skb(skb);
614 return 0;
615 }
616
617 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
618 {
619 return iptunnel_handle_offloads(skb,
620 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
621 }
622
623 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
624 struct net_device *dev,
625 struct flowi6 *fl6, __u8 *dsfield,
626 int *encap_limit)
627 {
628 const struct iphdr *iph = ip_hdr(skb);
629 struct ip6_tnl *t = netdev_priv(dev);
630
631 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
632 *encap_limit = t->parms.encap_limit;
633
634 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
635
636 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
637 *dsfield = ipv4_get_dsfield(iph);
638 else
639 *dsfield = ip6_tclass(t->parms.flowinfo);
640
641 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
642 fl6->flowi6_mark = skb->mark;
643 else
644 fl6->flowi6_mark = t->parms.fwmark;
645
646 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
647 }
648
649 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
650 struct net_device *dev,
651 struct flowi6 *fl6, __u8 *dsfield,
652 int *encap_limit)
653 {
654 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
655 struct ip6_tnl *t = netdev_priv(dev);
656 __u16 offset;
657
658 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
659 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
660
661 if (offset > 0) {
662 struct ipv6_tlv_tnl_enc_lim *tel;
663
664 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
665 if (tel->encap_limit == 0) {
666 icmpv6_send(skb, ICMPV6_PARAMPROB,
667 ICMPV6_HDR_FIELD, offset + 2);
668 return -1;
669 }
670 *encap_limit = tel->encap_limit - 1;
671 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
672 *encap_limit = t->parms.encap_limit;
673 }
674
675 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
676
677 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
678 *dsfield = ipv6_get_dsfield(ipv6h);
679 else
680 *dsfield = ip6_tclass(t->parms.flowinfo);
681
682 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
683 fl6->flowlabel |= ip6_flowlabel(ipv6h);
684
685 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
686 fl6->flowi6_mark = skb->mark;
687 else
688 fl6->flowi6_mark = t->parms.fwmark;
689
690 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
691
692 return 0;
693 }
694
695 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
696 struct net_device *dev, __u8 dsfield,
697 struct flowi6 *fl6, int encap_limit,
698 __u32 *pmtu, __be16 proto)
699 {
700 struct ip6_tnl *tunnel = netdev_priv(dev);
701 __be16 protocol;
702
703 if (dev->type == ARPHRD_ETHER)
704 IPCB(skb)->flags = 0;
705
706 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
707 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
708 else
709 fl6->daddr = tunnel->parms.raddr;
710
711 if (tunnel->parms.o_flags & TUNNEL_SEQ)
712 tunnel->o_seqno++;
713
714 /* Push GRE header. */
715 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
716
717 if (tunnel->parms.collect_md) {
718 struct ip_tunnel_info *tun_info;
719 const struct ip_tunnel_key *key;
720 __be16 flags;
721
722 tun_info = skb_tunnel_info(skb);
723 if (unlikely(!tun_info ||
724 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
725 ip_tunnel_info_af(tun_info) != AF_INET6))
726 return -EINVAL;
727
728 key = &tun_info->key;
729 memset(fl6, 0, sizeof(*fl6));
730 fl6->flowi6_proto = IPPROTO_GRE;
731 fl6->daddr = key->u.ipv6.dst;
732 fl6->flowlabel = key->label;
733 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
734
735 dsfield = key->tos;
736 flags = key->tun_flags & (TUNNEL_CSUM | TUNNEL_KEY);
737 tunnel->tun_hlen = gre_calc_hlen(flags);
738
739 gre_build_header(skb, tunnel->tun_hlen,
740 flags, protocol,
741 tunnel_id_to_key32(tun_info->key.tun_id), 0);
742
743 } else {
744 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
745 protocol, tunnel->parms.o_key,
746 htonl(tunnel->o_seqno));
747 }
748
749 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
750 NEXTHDR_GRE);
751 }
752
753 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
754 {
755 struct ip6_tnl *t = netdev_priv(dev);
756 int encap_limit = -1;
757 struct flowi6 fl6;
758 __u8 dsfield = 0;
759 __u32 mtu;
760 int err;
761
762 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
763
764 if (!t->parms.collect_md)
765 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
766 &dsfield, &encap_limit);
767
768 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
769 if (err)
770 return -1;
771
772 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
773 skb->protocol);
774 if (err != 0) {
775 /* XXX: send ICMP error even if DF is not set. */
776 if (err == -EMSGSIZE)
777 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
778 htonl(mtu));
779 return -1;
780 }
781
782 return 0;
783 }
784
785 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
786 {
787 struct ip6_tnl *t = netdev_priv(dev);
788 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
789 int encap_limit = -1;
790 struct flowi6 fl6;
791 __u8 dsfield = 0;
792 __u32 mtu;
793 int err;
794
795 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
796 return -1;
797
798 if (!t->parms.collect_md &&
799 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
800 return -1;
801
802 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
803 return -1;
804
805 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
806 &mtu, skb->protocol);
807 if (err != 0) {
808 if (err == -EMSGSIZE)
809 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
810 return -1;
811 }
812
813 return 0;
814 }
815
816 /**
817 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
818 * @t: the outgoing tunnel device
819 * @hdr: IPv6 header from the incoming packet
820 *
821 * Description:
822 * Avoid trivial tunneling loop by checking that tunnel exit-point
823 * doesn't match source of incoming packet.
824 *
825 * Return:
826 * 1 if conflict,
827 * 0 else
828 **/
829
830 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
831 const struct ipv6hdr *hdr)
832 {
833 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
834 }
835
836 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
837 {
838 struct ip6_tnl *t = netdev_priv(dev);
839 int encap_limit = -1;
840 struct flowi6 fl6;
841 __u32 mtu;
842 int err;
843
844 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
845 encap_limit = t->parms.encap_limit;
846
847 if (!t->parms.collect_md)
848 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
849
850 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
851 if (err)
852 return err;
853
854 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
855
856 return err;
857 }
858
859 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
860 struct net_device *dev)
861 {
862 struct ip6_tnl *t = netdev_priv(dev);
863 struct net_device_stats *stats = &t->dev->stats;
864 int ret;
865
866 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
867 goto tx_err;
868
869 switch (skb->protocol) {
870 case htons(ETH_P_IP):
871 ret = ip6gre_xmit_ipv4(skb, dev);
872 break;
873 case htons(ETH_P_IPV6):
874 ret = ip6gre_xmit_ipv6(skb, dev);
875 break;
876 default:
877 ret = ip6gre_xmit_other(skb, dev);
878 break;
879 }
880
881 if (ret < 0)
882 goto tx_err;
883
884 return NETDEV_TX_OK;
885
886 tx_err:
887 stats->tx_errors++;
888 stats->tx_dropped++;
889 kfree_skb(skb);
890 return NETDEV_TX_OK;
891 }
892
893 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
894 struct net_device *dev)
895 {
896 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
897 struct ip6_tnl *t = netdev_priv(dev);
898 struct dst_entry *dst = skb_dst(skb);
899 struct net_device_stats *stats;
900 bool truncate = false;
901 int encap_limit = -1;
902 __u8 dsfield = false;
903 struct flowi6 fl6;
904 int err = -EINVAL;
905 __u32 mtu;
906
907 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
908 goto tx_err;
909
910 if (gre_handle_offloads(skb, false))
911 goto tx_err;
912
913 if (skb->len > dev->mtu + dev->hard_header_len) {
914 pskb_trim(skb, dev->mtu + dev->hard_header_len);
915 truncate = true;
916 }
917
918 t->parms.o_flags &= ~TUNNEL_KEY;
919 IPCB(skb)->flags = 0;
920
921 /* For collect_md mode, derive fl6 from the tunnel key,
922 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
923 */
924 if (t->parms.collect_md) {
925 struct ip_tunnel_info *tun_info;
926 const struct ip_tunnel_key *key;
927 struct erspan_metadata *md;
928
929 tun_info = skb_tunnel_info(skb);
930 if (unlikely(!tun_info ||
931 !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
932 ip_tunnel_info_af(tun_info) != AF_INET6))
933 return -EINVAL;
934
935 key = &tun_info->key;
936 memset(&fl6, 0, sizeof(fl6));
937 fl6.flowi6_proto = IPPROTO_GRE;
938 fl6.daddr = key->u.ipv6.dst;
939 fl6.flowlabel = key->label;
940 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
941
942 dsfield = key->tos;
943 md = ip_tunnel_info_opts(tun_info);
944 if (!md)
945 goto tx_err;
946
947 if (md->version == 1) {
948 erspan_build_header(skb,
949 tunnel_id_to_key32(key->tun_id),
950 ntohl(md->u.index), truncate,
951 false);
952 } else if (md->version == 2) {
953 u16 md2_flags;
954 u16 dir, hwid;
955
956 md2_flags = ntohs(md->u.md2.flags);
957 dir = (md2_flags & DIR_MASK) >> DIR_OFFSET;
958 hwid = (md2_flags & HWID_MASK) >> HWID_OFFSET;
959
960 erspan_build_header_v2(skb,
961 tunnel_id_to_key32(key->tun_id),
962 dir, hwid, truncate,
963 false);
964 }
965 } else {
966 switch (skb->protocol) {
967 case htons(ETH_P_IP):
968 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
969 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
970 &dsfield, &encap_limit);
971 break;
972 case htons(ETH_P_IPV6):
973 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
974 goto tx_err;
975 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
976 &dsfield, &encap_limit))
977 goto tx_err;
978 break;
979 default:
980 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
981 break;
982 }
983
984 if (t->parms.erspan_ver == 1)
985 erspan_build_header(skb, t->parms.o_key,
986 t->parms.index,
987 truncate, false);
988 else
989 erspan_build_header_v2(skb, t->parms.o_key,
990 t->parms.dir,
991 t->parms.hwid,
992 truncate, false);
993 fl6.daddr = t->parms.raddr;
994 }
995
996 /* Push GRE header. */
997 gre_build_header(skb, 8, TUNNEL_SEQ,
998 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
999
1000 /* TooBig packet may have updated dst->dev's mtu */
1001 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
1002 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
1003
1004 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1005 NEXTHDR_GRE);
1006 if (err != 0) {
1007 /* XXX: send ICMP error even if DF is not set. */
1008 if (err == -EMSGSIZE) {
1009 if (skb->protocol == htons(ETH_P_IP))
1010 icmp_send(skb, ICMP_DEST_UNREACH,
1011 ICMP_FRAG_NEEDED, htonl(mtu));
1012 else
1013 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1014 }
1015
1016 goto tx_err;
1017 }
1018 return NETDEV_TX_OK;
1019
1020 tx_err:
1021 stats = &t->dev->stats;
1022 stats->tx_errors++;
1023 stats->tx_dropped++;
1024 kfree_skb(skb);
1025 return NETDEV_TX_OK;
1026 }
1027
1028 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1029 {
1030 struct net_device *dev = t->dev;
1031 struct __ip6_tnl_parm *p = &t->parms;
1032 struct flowi6 *fl6 = &t->fl.u.ip6;
1033 int t_hlen;
1034
1035 if (dev->type != ARPHRD_ETHER) {
1036 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1037 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1038 }
1039
1040 /* Set up flowi template */
1041 fl6->saddr = p->laddr;
1042 fl6->daddr = p->raddr;
1043 fl6->flowi6_oif = p->link;
1044 fl6->flowlabel = 0;
1045 fl6->flowi6_proto = IPPROTO_GRE;
1046
1047 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1048 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1049 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1050 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1051
1052 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1053 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1054
1055 if (p->flags&IP6_TNL_F_CAP_XMIT &&
1056 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1057 dev->flags |= IFF_POINTOPOINT;
1058 else
1059 dev->flags &= ~IFF_POINTOPOINT;
1060
1061 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
1062
1063 t->hlen = t->encap_hlen + t->tun_hlen;
1064
1065 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1066
1067 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1068 int strict = (ipv6_addr_type(&p->raddr) &
1069 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1070
1071 struct rt6_info *rt = rt6_lookup(t->net,
1072 &p->raddr, &p->laddr,
1073 p->link, strict);
1074
1075 if (!rt)
1076 return;
1077
1078 if (rt->dst.dev) {
1079 dev->hard_header_len = rt->dst.dev->hard_header_len +
1080 t_hlen;
1081
1082 if (set_mtu) {
1083 dev->mtu = rt->dst.dev->mtu - t_hlen;
1084 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1085 dev->mtu -= 8;
1086 if (dev->type == ARPHRD_ETHER)
1087 dev->mtu -= ETH_HLEN;
1088
1089 if (dev->mtu < IPV6_MIN_MTU)
1090 dev->mtu = IPV6_MIN_MTU;
1091 }
1092 }
1093 ip6_rt_put(rt);
1094 }
1095 }
1096
1097 static int ip6gre_tnl_change(struct ip6_tnl *t,
1098 const struct __ip6_tnl_parm *p, int set_mtu)
1099 {
1100 t->parms.laddr = p->laddr;
1101 t->parms.raddr = p->raddr;
1102 t->parms.flags = p->flags;
1103 t->parms.hop_limit = p->hop_limit;
1104 t->parms.encap_limit = p->encap_limit;
1105 t->parms.flowinfo = p->flowinfo;
1106 t->parms.link = p->link;
1107 t->parms.proto = p->proto;
1108 t->parms.i_key = p->i_key;
1109 t->parms.o_key = p->o_key;
1110 t->parms.i_flags = p->i_flags;
1111 t->parms.o_flags = p->o_flags;
1112 t->parms.fwmark = p->fwmark;
1113 dst_cache_reset(&t->dst_cache);
1114 ip6gre_tnl_link_config(t, set_mtu);
1115 return 0;
1116 }
1117
1118 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1119 const struct ip6_tnl_parm2 *u)
1120 {
1121 p->laddr = u->laddr;
1122 p->raddr = u->raddr;
1123 p->flags = u->flags;
1124 p->hop_limit = u->hop_limit;
1125 p->encap_limit = u->encap_limit;
1126 p->flowinfo = u->flowinfo;
1127 p->link = u->link;
1128 p->i_key = u->i_key;
1129 p->o_key = u->o_key;
1130 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1131 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1132 memcpy(p->name, u->name, sizeof(u->name));
1133 }
1134
1135 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1136 const struct __ip6_tnl_parm *p)
1137 {
1138 u->proto = IPPROTO_GRE;
1139 u->laddr = p->laddr;
1140 u->raddr = p->raddr;
1141 u->flags = p->flags;
1142 u->hop_limit = p->hop_limit;
1143 u->encap_limit = p->encap_limit;
1144 u->flowinfo = p->flowinfo;
1145 u->link = p->link;
1146 u->i_key = p->i_key;
1147 u->o_key = p->o_key;
1148 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1149 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1150 memcpy(u->name, p->name, sizeof(u->name));
1151 }
1152
1153 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1154 struct ifreq *ifr, int cmd)
1155 {
1156 int err = 0;
1157 struct ip6_tnl_parm2 p;
1158 struct __ip6_tnl_parm p1;
1159 struct ip6_tnl *t = netdev_priv(dev);
1160 struct net *net = t->net;
1161 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1162
1163 memset(&p1, 0, sizeof(p1));
1164
1165 switch (cmd) {
1166 case SIOCGETTUNNEL:
1167 if (dev == ign->fb_tunnel_dev) {
1168 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1169 err = -EFAULT;
1170 break;
1171 }
1172 ip6gre_tnl_parm_from_user(&p1, &p);
1173 t = ip6gre_tunnel_locate(net, &p1, 0);
1174 if (!t)
1175 t = netdev_priv(dev);
1176 }
1177 memset(&p, 0, sizeof(p));
1178 ip6gre_tnl_parm_to_user(&p, &t->parms);
1179 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1180 err = -EFAULT;
1181 break;
1182
1183 case SIOCADDTUNNEL:
1184 case SIOCCHGTUNNEL:
1185 err = -EPERM;
1186 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1187 goto done;
1188
1189 err = -EFAULT;
1190 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1191 goto done;
1192
1193 err = -EINVAL;
1194 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1195 goto done;
1196
1197 if (!(p.i_flags&GRE_KEY))
1198 p.i_key = 0;
1199 if (!(p.o_flags&GRE_KEY))
1200 p.o_key = 0;
1201
1202 ip6gre_tnl_parm_from_user(&p1, &p);
1203 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1204
1205 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1206 if (t) {
1207 if (t->dev != dev) {
1208 err = -EEXIST;
1209 break;
1210 }
1211 } else {
1212 t = netdev_priv(dev);
1213
1214 ip6gre_tunnel_unlink(ign, t);
1215 synchronize_net();
1216 ip6gre_tnl_change(t, &p1, 1);
1217 ip6gre_tunnel_link(ign, t);
1218 netdev_state_change(dev);
1219 }
1220 }
1221
1222 if (t) {
1223 err = 0;
1224
1225 memset(&p, 0, sizeof(p));
1226 ip6gre_tnl_parm_to_user(&p, &t->parms);
1227 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1228 err = -EFAULT;
1229 } else
1230 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1231 break;
1232
1233 case SIOCDELTUNNEL:
1234 err = -EPERM;
1235 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1236 goto done;
1237
1238 if (dev == ign->fb_tunnel_dev) {
1239 err = -EFAULT;
1240 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1241 goto done;
1242 err = -ENOENT;
1243 ip6gre_tnl_parm_from_user(&p1, &p);
1244 t = ip6gre_tunnel_locate(net, &p1, 0);
1245 if (!t)
1246 goto done;
1247 err = -EPERM;
1248 if (t == netdev_priv(ign->fb_tunnel_dev))
1249 goto done;
1250 dev = t->dev;
1251 }
1252 unregister_netdevice(dev);
1253 err = 0;
1254 break;
1255
1256 default:
1257 err = -EINVAL;
1258 }
1259
1260 done:
1261 return err;
1262 }
1263
1264 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1265 unsigned short type, const void *daddr,
1266 const void *saddr, unsigned int len)
1267 {
1268 struct ip6_tnl *t = netdev_priv(dev);
1269 struct ipv6hdr *ipv6h;
1270 __be16 *p;
1271
1272 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1273 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1274 t->fl.u.ip6.flowlabel,
1275 true, &t->fl.u.ip6));
1276 ipv6h->hop_limit = t->parms.hop_limit;
1277 ipv6h->nexthdr = NEXTHDR_GRE;
1278 ipv6h->saddr = t->parms.laddr;
1279 ipv6h->daddr = t->parms.raddr;
1280
1281 p = (__be16 *)(ipv6h + 1);
1282 p[0] = t->parms.o_flags;
1283 p[1] = htons(type);
1284
1285 /*
1286 * Set the source hardware address.
1287 */
1288
1289 if (saddr)
1290 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1291 if (daddr)
1292 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1293 if (!ipv6_addr_any(&ipv6h->daddr))
1294 return t->hlen;
1295
1296 return -t->hlen;
1297 }
1298
1299 static const struct header_ops ip6gre_header_ops = {
1300 .create = ip6gre_header,
1301 };
1302
1303 static const struct net_device_ops ip6gre_netdev_ops = {
1304 .ndo_init = ip6gre_tunnel_init,
1305 .ndo_uninit = ip6gre_tunnel_uninit,
1306 .ndo_start_xmit = ip6gre_tunnel_xmit,
1307 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
1308 .ndo_change_mtu = ip6_tnl_change_mtu,
1309 .ndo_get_stats64 = ip_tunnel_get_stats64,
1310 .ndo_get_iflink = ip6_tnl_get_iflink,
1311 };
1312
1313 static void ip6gre_dev_free(struct net_device *dev)
1314 {
1315 struct ip6_tnl *t = netdev_priv(dev);
1316
1317 dst_cache_destroy(&t->dst_cache);
1318 free_percpu(dev->tstats);
1319 }
1320
1321 static void ip6gre_tunnel_setup(struct net_device *dev)
1322 {
1323 dev->netdev_ops = &ip6gre_netdev_ops;
1324 dev->needs_free_netdev = true;
1325 dev->priv_destructor = ip6gre_dev_free;
1326
1327 dev->type = ARPHRD_IP6GRE;
1328
1329 dev->flags |= IFF_NOARP;
1330 dev->addr_len = sizeof(struct in6_addr);
1331 netif_keep_dst(dev);
1332 /* This perm addr will be used as interface identifier by IPv6 */
1333 dev->addr_assign_type = NET_ADDR_RANDOM;
1334 eth_random_addr(dev->perm_addr);
1335 }
1336
1337 #define GRE6_FEATURES (NETIF_F_SG | \
1338 NETIF_F_FRAGLIST | \
1339 NETIF_F_HIGHDMA | \
1340 NETIF_F_HW_CSUM)
1341
1342 static void ip6gre_tnl_init_features(struct net_device *dev)
1343 {
1344 struct ip6_tnl *nt = netdev_priv(dev);
1345
1346 dev->features |= GRE6_FEATURES;
1347 dev->hw_features |= GRE6_FEATURES;
1348
1349 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1350 /* TCP offload with GRE SEQ is not supported, nor
1351 * can we support 2 levels of outer headers requiring
1352 * an update.
1353 */
1354 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1355 nt->encap.type == TUNNEL_ENCAP_NONE) {
1356 dev->features |= NETIF_F_GSO_SOFTWARE;
1357 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1358 }
1359
1360 /* Can use a lockless transmit, unless we generate
1361 * output sequences
1362 */
1363 dev->features |= NETIF_F_LLTX;
1364 }
1365 }
1366
1367 static int ip6gre_tunnel_init_common(struct net_device *dev)
1368 {
1369 struct ip6_tnl *tunnel;
1370 int ret;
1371 int t_hlen;
1372
1373 tunnel = netdev_priv(dev);
1374
1375 tunnel->dev = dev;
1376 tunnel->net = dev_net(dev);
1377 strcpy(tunnel->parms.name, dev->name);
1378
1379 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1380 if (!dev->tstats)
1381 return -ENOMEM;
1382
1383 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1384 if (ret) {
1385 free_percpu(dev->tstats);
1386 dev->tstats = NULL;
1387 return ret;
1388 }
1389
1390 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1391 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1392 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1393
1394 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1395 dev->mtu = ETH_DATA_LEN - t_hlen;
1396 if (dev->type == ARPHRD_ETHER)
1397 dev->mtu -= ETH_HLEN;
1398 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1399 dev->mtu -= 8;
1400
1401 if (tunnel->parms.collect_md) {
1402 dev->features |= NETIF_F_NETNS_LOCAL;
1403 netif_keep_dst(dev);
1404 }
1405 ip6gre_tnl_init_features(dev);
1406
1407 return 0;
1408 }
1409
1410 static int ip6gre_tunnel_init(struct net_device *dev)
1411 {
1412 struct ip6_tnl *tunnel;
1413 int ret;
1414
1415 ret = ip6gre_tunnel_init_common(dev);
1416 if (ret)
1417 return ret;
1418
1419 tunnel = netdev_priv(dev);
1420
1421 if (tunnel->parms.collect_md)
1422 return 0;
1423
1424 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1425 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1426
1427 if (ipv6_addr_any(&tunnel->parms.raddr))
1428 dev->header_ops = &ip6gre_header_ops;
1429
1430 return 0;
1431 }
1432
1433 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1434 {
1435 struct ip6_tnl *tunnel = netdev_priv(dev);
1436
1437 tunnel->dev = dev;
1438 tunnel->net = dev_net(dev);
1439 strcpy(tunnel->parms.name, dev->name);
1440
1441 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1442
1443 dev_hold(dev);
1444 }
1445
1446 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1447 .handler = gre_rcv,
1448 .err_handler = ip6gre_err,
1449 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1450 };
1451
1452 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1453 {
1454 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1455 struct net_device *dev, *aux;
1456 int prio;
1457
1458 for_each_netdev_safe(net, dev, aux)
1459 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1460 dev->rtnl_link_ops == &ip6gre_tap_ops ||
1461 dev->rtnl_link_ops == &ip6erspan_tap_ops)
1462 unregister_netdevice_queue(dev, head);
1463
1464 for (prio = 0; prio < 4; prio++) {
1465 int h;
1466 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1467 struct ip6_tnl *t;
1468
1469 t = rtnl_dereference(ign->tunnels[prio][h]);
1470
1471 while (t) {
1472 /* If dev is in the same netns, it has already
1473 * been added to the list by the previous loop.
1474 */
1475 if (!net_eq(dev_net(t->dev), net))
1476 unregister_netdevice_queue(t->dev,
1477 head);
1478 t = rtnl_dereference(t->next);
1479 }
1480 }
1481 }
1482 }
1483
1484 static int __net_init ip6gre_init_net(struct net *net)
1485 {
1486 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1487 int err;
1488
1489 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1490 NET_NAME_UNKNOWN,
1491 ip6gre_tunnel_setup);
1492 if (!ign->fb_tunnel_dev) {
1493 err = -ENOMEM;
1494 goto err_alloc_dev;
1495 }
1496 dev_net_set(ign->fb_tunnel_dev, net);
1497 /* FB netdevice is special: we have one, and only one per netns.
1498 * Allowing to move it to another netns is clearly unsafe.
1499 */
1500 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1501
1502
1503 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1504 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1505
1506 err = register_netdev(ign->fb_tunnel_dev);
1507 if (err)
1508 goto err_reg_dev;
1509
1510 rcu_assign_pointer(ign->tunnels_wc[0],
1511 netdev_priv(ign->fb_tunnel_dev));
1512 return 0;
1513
1514 err_reg_dev:
1515 free_netdev(ign->fb_tunnel_dev);
1516 err_alloc_dev:
1517 return err;
1518 }
1519
1520 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1521 {
1522 struct net *net;
1523 LIST_HEAD(list);
1524
1525 rtnl_lock();
1526 list_for_each_entry(net, net_list, exit_list)
1527 ip6gre_destroy_tunnels(net, &list);
1528 unregister_netdevice_many(&list);
1529 rtnl_unlock();
1530 }
1531
1532 static struct pernet_operations ip6gre_net_ops = {
1533 .init = ip6gre_init_net,
1534 .exit_batch = ip6gre_exit_batch_net,
1535 .id = &ip6gre_net_id,
1536 .size = sizeof(struct ip6gre_net),
1537 };
1538
1539 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1540 struct netlink_ext_ack *extack)
1541 {
1542 __be16 flags;
1543
1544 if (!data)
1545 return 0;
1546
1547 flags = 0;
1548 if (data[IFLA_GRE_IFLAGS])
1549 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1550 if (data[IFLA_GRE_OFLAGS])
1551 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1552 if (flags & (GRE_VERSION|GRE_ROUTING))
1553 return -EINVAL;
1554
1555 return 0;
1556 }
1557
1558 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1559 struct netlink_ext_ack *extack)
1560 {
1561 struct in6_addr daddr;
1562
1563 if (tb[IFLA_ADDRESS]) {
1564 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1565 return -EINVAL;
1566 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1567 return -EADDRNOTAVAIL;
1568 }
1569
1570 if (!data)
1571 goto out;
1572
1573 if (data[IFLA_GRE_REMOTE]) {
1574 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1575 if (ipv6_addr_any(&daddr))
1576 return -EINVAL;
1577 }
1578
1579 out:
1580 return ip6gre_tunnel_validate(tb, data, extack);
1581 }
1582
1583 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1584 struct netlink_ext_ack *extack)
1585 {
1586 __be16 flags = 0;
1587 int ret, ver = 0;
1588
1589 if (!data)
1590 return 0;
1591
1592 ret = ip6gre_tap_validate(tb, data, extack);
1593 if (ret)
1594 return ret;
1595
1596 /* ERSPAN should only have GRE sequence and key flag */
1597 if (data[IFLA_GRE_OFLAGS])
1598 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1599 if (data[IFLA_GRE_IFLAGS])
1600 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1601 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1602 flags != (GRE_SEQ | GRE_KEY))
1603 return -EINVAL;
1604
1605 /* ERSPAN Session ID only has 10-bit. Since we reuse
1606 * 32-bit key field as ID, check it's range.
1607 */
1608 if (data[IFLA_GRE_IKEY] &&
1609 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1610 return -EINVAL;
1611
1612 if (data[IFLA_GRE_OKEY] &&
1613 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1614 return -EINVAL;
1615
1616 if (data[IFLA_GRE_ERSPAN_VER]) {
1617 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1618 if (ver != 1 && ver != 2)
1619 return -EINVAL;
1620 }
1621
1622 if (ver == 1) {
1623 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1624 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1625
1626 if (index & ~INDEX_MASK)
1627 return -EINVAL;
1628 }
1629 } else if (ver == 2) {
1630 if (data[IFLA_GRE_ERSPAN_DIR]) {
1631 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1632
1633 if (dir & ~(DIR_MASK >> DIR_OFFSET))
1634 return -EINVAL;
1635 }
1636
1637 if (data[IFLA_GRE_ERSPAN_HWID]) {
1638 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1639
1640 if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1641 return -EINVAL;
1642 }
1643 }
1644
1645 return 0;
1646 }
1647
1648 static void ip6gre_netlink_parms(struct nlattr *data[],
1649 struct __ip6_tnl_parm *parms)
1650 {
1651 memset(parms, 0, sizeof(*parms));
1652
1653 if (!data)
1654 return;
1655
1656 if (data[IFLA_GRE_LINK])
1657 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1658
1659 if (data[IFLA_GRE_IFLAGS])
1660 parms->i_flags = gre_flags_to_tnl_flags(
1661 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1662
1663 if (data[IFLA_GRE_OFLAGS])
1664 parms->o_flags = gre_flags_to_tnl_flags(
1665 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1666
1667 if (data[IFLA_GRE_IKEY])
1668 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1669
1670 if (data[IFLA_GRE_OKEY])
1671 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1672
1673 if (data[IFLA_GRE_LOCAL])
1674 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1675
1676 if (data[IFLA_GRE_REMOTE])
1677 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1678
1679 if (data[IFLA_GRE_TTL])
1680 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1681
1682 if (data[IFLA_GRE_ENCAP_LIMIT])
1683 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1684
1685 if (data[IFLA_GRE_FLOWINFO])
1686 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1687
1688 if (data[IFLA_GRE_FLAGS])
1689 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1690
1691 if (data[IFLA_GRE_FWMARK])
1692 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1693
1694 if (data[IFLA_GRE_COLLECT_METADATA])
1695 parms->collect_md = true;
1696
1697 if (data[IFLA_GRE_ERSPAN_VER])
1698 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1699
1700 if (parms->erspan_ver == 1) {
1701 if (data[IFLA_GRE_ERSPAN_INDEX])
1702 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1703 } else if (parms->erspan_ver == 2) {
1704 if (data[IFLA_GRE_ERSPAN_DIR])
1705 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1706 if (data[IFLA_GRE_ERSPAN_HWID])
1707 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1708 }
1709 }
1710
1711 static int ip6gre_tap_init(struct net_device *dev)
1712 {
1713 int ret;
1714
1715 ret = ip6gre_tunnel_init_common(dev);
1716 if (ret)
1717 return ret;
1718
1719 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1720
1721 return 0;
1722 }
1723
1724 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1725 .ndo_init = ip6gre_tap_init,
1726 .ndo_uninit = ip6gre_tunnel_uninit,
1727 .ndo_start_xmit = ip6gre_tunnel_xmit,
1728 .ndo_set_mac_address = eth_mac_addr,
1729 .ndo_validate_addr = eth_validate_addr,
1730 .ndo_change_mtu = ip6_tnl_change_mtu,
1731 .ndo_get_stats64 = ip_tunnel_get_stats64,
1732 .ndo_get_iflink = ip6_tnl_get_iflink,
1733 };
1734
1735 static int ip6erspan_tap_init(struct net_device *dev)
1736 {
1737 struct ip6_tnl *tunnel;
1738 int t_hlen;
1739 int ret;
1740
1741 tunnel = netdev_priv(dev);
1742
1743 tunnel->dev = dev;
1744 tunnel->net = dev_net(dev);
1745 strcpy(tunnel->parms.name, dev->name);
1746
1747 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1748 if (!dev->tstats)
1749 return -ENOMEM;
1750
1751 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1752 if (ret) {
1753 free_percpu(dev->tstats);
1754 dev->tstats = NULL;
1755 return ret;
1756 }
1757
1758 tunnel->tun_hlen = 8;
1759 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1760 erspan_hdr_len(tunnel->parms.erspan_ver);
1761 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1762
1763 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1764 dev->mtu = ETH_DATA_LEN - t_hlen;
1765 if (dev->type == ARPHRD_ETHER)
1766 dev->mtu -= ETH_HLEN;
1767 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1768 dev->mtu -= 8;
1769
1770 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1771 tunnel = netdev_priv(dev);
1772 ip6gre_tnl_link_config(tunnel, 1);
1773
1774 return 0;
1775 }
1776
1777 static const struct net_device_ops ip6erspan_netdev_ops = {
1778 .ndo_init = ip6erspan_tap_init,
1779 .ndo_uninit = ip6gre_tunnel_uninit,
1780 .ndo_start_xmit = ip6erspan_tunnel_xmit,
1781 .ndo_set_mac_address = eth_mac_addr,
1782 .ndo_validate_addr = eth_validate_addr,
1783 .ndo_change_mtu = ip6_tnl_change_mtu,
1784 .ndo_get_stats64 = ip_tunnel_get_stats64,
1785 .ndo_get_iflink = ip6_tnl_get_iflink,
1786 };
1787
1788 static void ip6gre_tap_setup(struct net_device *dev)
1789 {
1790
1791 ether_setup(dev);
1792
1793 dev->max_mtu = 0;
1794 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1795 dev->needs_free_netdev = true;
1796 dev->priv_destructor = ip6gre_dev_free;
1797
1798 dev->features |= NETIF_F_NETNS_LOCAL;
1799 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1800 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1801 netif_keep_dst(dev);
1802 }
1803
1804 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1805 struct ip_tunnel_encap *ipencap)
1806 {
1807 bool ret = false;
1808
1809 memset(ipencap, 0, sizeof(*ipencap));
1810
1811 if (!data)
1812 return ret;
1813
1814 if (data[IFLA_GRE_ENCAP_TYPE]) {
1815 ret = true;
1816 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1817 }
1818
1819 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1820 ret = true;
1821 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1822 }
1823
1824 if (data[IFLA_GRE_ENCAP_SPORT]) {
1825 ret = true;
1826 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1827 }
1828
1829 if (data[IFLA_GRE_ENCAP_DPORT]) {
1830 ret = true;
1831 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1832 }
1833
1834 return ret;
1835 }
1836
1837 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1838 struct nlattr *tb[], struct nlattr *data[],
1839 struct netlink_ext_ack *extack)
1840 {
1841 struct ip6_tnl *nt;
1842 struct net *net = dev_net(dev);
1843 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1844 struct ip_tunnel_encap ipencap;
1845 int err;
1846
1847 nt = netdev_priv(dev);
1848
1849 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1850 int err = ip6_tnl_encap_setup(nt, &ipencap);
1851
1852 if (err < 0)
1853 return err;
1854 }
1855
1856 ip6gre_netlink_parms(data, &nt->parms);
1857
1858 if (nt->parms.collect_md) {
1859 if (rtnl_dereference(ign->collect_md_tun))
1860 return -EEXIST;
1861 } else {
1862 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1863 return -EEXIST;
1864 }
1865
1866 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1867 eth_hw_addr_random(dev);
1868
1869 nt->dev = dev;
1870 nt->net = dev_net(dev);
1871
1872 err = register_netdevice(dev);
1873 if (err)
1874 goto out;
1875
1876 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1877
1878 if (tb[IFLA_MTU])
1879 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1880
1881 dev_hold(dev);
1882 ip6gre_tunnel_link(ign, nt);
1883
1884 out:
1885 return err;
1886 }
1887
1888 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1889 struct nlattr *data[],
1890 struct netlink_ext_ack *extack)
1891 {
1892 struct ip6_tnl *t, *nt = netdev_priv(dev);
1893 struct net *net = nt->net;
1894 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1895 struct __ip6_tnl_parm p;
1896 struct ip_tunnel_encap ipencap;
1897
1898 if (dev == ign->fb_tunnel_dev)
1899 return -EINVAL;
1900
1901 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1902 int err = ip6_tnl_encap_setup(nt, &ipencap);
1903
1904 if (err < 0)
1905 return err;
1906 }
1907
1908 ip6gre_netlink_parms(data, &p);
1909
1910 t = ip6gre_tunnel_locate(net, &p, 0);
1911
1912 if (t) {
1913 if (t->dev != dev)
1914 return -EEXIST;
1915 } else {
1916 t = nt;
1917 }
1918
1919 ip6gre_tunnel_unlink(ign, t);
1920 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1921 ip6gre_tunnel_link(ign, t);
1922 return 0;
1923 }
1924
1925 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1926 {
1927 struct net *net = dev_net(dev);
1928 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1929
1930 if (dev != ign->fb_tunnel_dev)
1931 unregister_netdevice_queue(dev, head);
1932 }
1933
1934 static size_t ip6gre_get_size(const struct net_device *dev)
1935 {
1936 return
1937 /* IFLA_GRE_LINK */
1938 nla_total_size(4) +
1939 /* IFLA_GRE_IFLAGS */
1940 nla_total_size(2) +
1941 /* IFLA_GRE_OFLAGS */
1942 nla_total_size(2) +
1943 /* IFLA_GRE_IKEY */
1944 nla_total_size(4) +
1945 /* IFLA_GRE_OKEY */
1946 nla_total_size(4) +
1947 /* IFLA_GRE_LOCAL */
1948 nla_total_size(sizeof(struct in6_addr)) +
1949 /* IFLA_GRE_REMOTE */
1950 nla_total_size(sizeof(struct in6_addr)) +
1951 /* IFLA_GRE_TTL */
1952 nla_total_size(1) +
1953 /* IFLA_GRE_ENCAP_LIMIT */
1954 nla_total_size(1) +
1955 /* IFLA_GRE_FLOWINFO */
1956 nla_total_size(4) +
1957 /* IFLA_GRE_FLAGS */
1958 nla_total_size(4) +
1959 /* IFLA_GRE_ENCAP_TYPE */
1960 nla_total_size(2) +
1961 /* IFLA_GRE_ENCAP_FLAGS */
1962 nla_total_size(2) +
1963 /* IFLA_GRE_ENCAP_SPORT */
1964 nla_total_size(2) +
1965 /* IFLA_GRE_ENCAP_DPORT */
1966 nla_total_size(2) +
1967 /* IFLA_GRE_COLLECT_METADATA */
1968 nla_total_size(0) +
1969 /* IFLA_GRE_FWMARK */
1970 nla_total_size(4) +
1971 /* IFLA_GRE_ERSPAN_INDEX */
1972 nla_total_size(4) +
1973 0;
1974 }
1975
1976 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1977 {
1978 struct ip6_tnl *t = netdev_priv(dev);
1979 struct __ip6_tnl_parm *p = &t->parms;
1980
1981 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1982 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1983 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1984 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1985 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1986 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1987 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1988 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1989 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1990 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1991 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1992 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1993 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1994 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
1995 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
1996 goto nla_put_failure;
1997
1998 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1999 t->encap.type) ||
2000 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2001 t->encap.sport) ||
2002 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2003 t->encap.dport) ||
2004 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2005 t->encap.flags))
2006 goto nla_put_failure;
2007
2008 if (p->collect_md) {
2009 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2010 goto nla_put_failure;
2011 }
2012
2013 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2014 goto nla_put_failure;
2015
2016 if (p->erspan_ver == 1) {
2017 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2018 goto nla_put_failure;
2019 } else if (p->erspan_ver == 2) {
2020 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2021 goto nla_put_failure;
2022 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2023 goto nla_put_failure;
2024 }
2025
2026 return 0;
2027
2028 nla_put_failure:
2029 return -EMSGSIZE;
2030 }
2031
2032 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2033 [IFLA_GRE_LINK] = { .type = NLA_U32 },
2034 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
2035 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
2036 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
2037 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
2038 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2039 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2040 [IFLA_GRE_TTL] = { .type = NLA_U8 },
2041 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2042 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
2043 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
2044 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
2045 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
2046 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
2047 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
2048 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2049 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
2050 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2051 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
2052 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
2053 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
2054 };
2055
2056 static void ip6erspan_tap_setup(struct net_device *dev)
2057 {
2058 ether_setup(dev);
2059
2060 dev->netdev_ops = &ip6erspan_netdev_ops;
2061 dev->needs_free_netdev = true;
2062 dev->priv_destructor = ip6gre_dev_free;
2063
2064 dev->features |= NETIF_F_NETNS_LOCAL;
2065 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2066 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2067 netif_keep_dst(dev);
2068 }
2069
2070 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2071 .kind = "ip6gre",
2072 .maxtype = IFLA_GRE_MAX,
2073 .policy = ip6gre_policy,
2074 .priv_size = sizeof(struct ip6_tnl),
2075 .setup = ip6gre_tunnel_setup,
2076 .validate = ip6gre_tunnel_validate,
2077 .newlink = ip6gre_newlink,
2078 .changelink = ip6gre_changelink,
2079 .dellink = ip6gre_dellink,
2080 .get_size = ip6gre_get_size,
2081 .fill_info = ip6gre_fill_info,
2082 .get_link_net = ip6_tnl_get_link_net,
2083 };
2084
2085 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2086 .kind = "ip6gretap",
2087 .maxtype = IFLA_GRE_MAX,
2088 .policy = ip6gre_policy,
2089 .priv_size = sizeof(struct ip6_tnl),
2090 .setup = ip6gre_tap_setup,
2091 .validate = ip6gre_tap_validate,
2092 .newlink = ip6gre_newlink,
2093 .changelink = ip6gre_changelink,
2094 .get_size = ip6gre_get_size,
2095 .fill_info = ip6gre_fill_info,
2096 .get_link_net = ip6_tnl_get_link_net,
2097 };
2098
2099 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2100 .kind = "ip6erspan",
2101 .maxtype = IFLA_GRE_MAX,
2102 .policy = ip6gre_policy,
2103 .priv_size = sizeof(struct ip6_tnl),
2104 .setup = ip6erspan_tap_setup,
2105 .validate = ip6erspan_tap_validate,
2106 .newlink = ip6gre_newlink,
2107 .changelink = ip6gre_changelink,
2108 .get_size = ip6gre_get_size,
2109 .fill_info = ip6gre_fill_info,
2110 .get_link_net = ip6_tnl_get_link_net,
2111 };
2112
2113 /*
2114 * And now the modules code and kernel interface.
2115 */
2116
2117 static int __init ip6gre_init(void)
2118 {
2119 int err;
2120
2121 pr_info("GRE over IPv6 tunneling driver\n");
2122
2123 err = register_pernet_device(&ip6gre_net_ops);
2124 if (err < 0)
2125 return err;
2126
2127 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2128 if (err < 0) {
2129 pr_info("%s: can't add protocol\n", __func__);
2130 goto add_proto_failed;
2131 }
2132
2133 err = rtnl_link_register(&ip6gre_link_ops);
2134 if (err < 0)
2135 goto rtnl_link_failed;
2136
2137 err = rtnl_link_register(&ip6gre_tap_ops);
2138 if (err < 0)
2139 goto tap_ops_failed;
2140
2141 err = rtnl_link_register(&ip6erspan_tap_ops);
2142 if (err < 0)
2143 goto erspan_link_failed;
2144
2145 out:
2146 return err;
2147
2148 erspan_link_failed:
2149 rtnl_link_unregister(&ip6gre_tap_ops);
2150 tap_ops_failed:
2151 rtnl_link_unregister(&ip6gre_link_ops);
2152 rtnl_link_failed:
2153 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2154 add_proto_failed:
2155 unregister_pernet_device(&ip6gre_net_ops);
2156 goto out;
2157 }
2158
2159 static void __exit ip6gre_fini(void)
2160 {
2161 rtnl_link_unregister(&ip6gre_tap_ops);
2162 rtnl_link_unregister(&ip6gre_link_ops);
2163 rtnl_link_unregister(&ip6erspan_tap_ops);
2164 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2165 unregister_pernet_device(&ip6gre_net_ops);
2166 }
2167
2168 module_init(ip6gre_init);
2169 module_exit(ip6gre_fini);
2170 MODULE_LICENSE("GPL");
2171 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2172 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2173 MODULE_ALIAS_RTNL_LINK("ip6gre");
2174 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2175 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2176 MODULE_ALIAS_NETDEV("ip6gre0");
Ubuntu Jammy Linux kernel mirror