projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge 4.14.24 into android-4.14
[GitHub/LineageOS/android_kernel_motorola_exynos9610.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
59
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63
64 #define IP6_GRE_HASH_SIZE_SHIFT 5
65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
66
67 static unsigned int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
70
71 struct net_device *fb_tunnel_dev;
72 };
73
74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
76 static int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80
81 /* Tunnel hash table */
82
83 /*
84 4 hash tables:
85
86 3: (remote,local)
87 2: (remote,*)
88 1: (*,local)
89 0: (*,*)
90
91 We require exact key match i.e. if a key is present in packet
92 it will match only tunnel with the same key; if it is not present,
93 it will match only keyless tunnel.
94
95 All keysless packets, if not matched configured keyless tunnels
96 will match fallback tunnel.
97 */
98
99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
100 static u32 HASH_ADDR(const struct in6_addr *addr)
101 {
102 u32 hash = ipv6_addr_hash(addr);
103
104 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
105 }
106
107 #define tunnels_r_l tunnels[3]
108 #define tunnels_r tunnels[2]
109 #define tunnels_l tunnels[1]
110 #define tunnels_wc tunnels[0]
111
112 /* Given src, dst and key, find appropriate for input tunnel. */
113
114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
115 const struct in6_addr *remote, const struct in6_addr *local,
116 __be32 key, __be16 gre_proto)
117 {
118 struct net *net = dev_net(dev);
119 int link = dev->ifindex;
120 unsigned int h0 = HASH_ADDR(remote);
121 unsigned int h1 = HASH_KEY(key);
122 struct ip6_tnl *t, *cand = NULL;
123 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
124 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
125 ARPHRD_ETHER : ARPHRD_IP6GRE;
126 int score, cand_score = 4;
127
128 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
129 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
130 !ipv6_addr_equal(remote, &t->parms.raddr) ||
131 key != t->parms.i_key ||
132 !(t->dev->flags & IFF_UP))
133 continue;
134
135 if (t->dev->type != ARPHRD_IP6GRE &&
136 t->dev->type != dev_type)
137 continue;
138
139 score = 0;
140 if (t->parms.link != link)
141 score |= 1;
142 if (t->dev->type != dev_type)
143 score |= 2;
144 if (score == 0)
145 return t;
146
147 if (score < cand_score) {
148 cand = t;
149 cand_score = score;
150 }
151 }
152
153 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
154 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
155 key != t->parms.i_key ||
156 !(t->dev->flags & IFF_UP))
157 continue;
158
159 if (t->dev->type != ARPHRD_IP6GRE &&
160 t->dev->type != dev_type)
161 continue;
162
163 score = 0;
164 if (t->parms.link != link)
165 score |= 1;
166 if (t->dev->type != dev_type)
167 score |= 2;
168 if (score == 0)
169 return t;
170
171 if (score < cand_score) {
172 cand = t;
173 cand_score = score;
174 }
175 }
176
177 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
178 if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
179 (!ipv6_addr_equal(local, &t->parms.raddr) ||
180 !ipv6_addr_is_multicast(local))) ||
181 key != t->parms.i_key ||
182 !(t->dev->flags & IFF_UP))
183 continue;
184
185 if (t->dev->type != ARPHRD_IP6GRE &&
186 t->dev->type != dev_type)
187 continue;
188
189 score = 0;
190 if (t->parms.link != link)
191 score |= 1;
192 if (t->dev->type != dev_type)
193 score |= 2;
194 if (score == 0)
195 return t;
196
197 if (score < cand_score) {
198 cand = t;
199 cand_score = score;
200 }
201 }
202
203 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
204 if (t->parms.i_key != key ||
205 !(t->dev->flags & IFF_UP))
206 continue;
207
208 if (t->dev->type != ARPHRD_IP6GRE &&
209 t->dev->type != dev_type)
210 continue;
211
212 score = 0;
213 if (t->parms.link != link)
214 score |= 1;
215 if (t->dev->type != dev_type)
216 score |= 2;
217 if (score == 0)
218 return t;
219
220 if (score < cand_score) {
221 cand = t;
222 cand_score = score;
223 }
224 }
225
226 if (cand)
227 return cand;
228
229 dev = ign->fb_tunnel_dev;
230 if (dev->flags & IFF_UP)
231 return netdev_priv(dev);
232
233 return NULL;
234 }
235
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237 const struct __ip6_tnl_parm *p)
238 {
239 const struct in6_addr *remote = &p->raddr;
240 const struct in6_addr *local = &p->laddr;
241 unsigned int h = HASH_KEY(p->i_key);
242 int prio = 0;
243
244 if (!ipv6_addr_any(local))
245 prio |= 1;
246 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
247 prio |= 2;
248 h ^= HASH_ADDR(remote);
249 }
250
251 return &ign->tunnels[prio][h];
252 }
253
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255 const struct ip6_tnl *t)
256 {
257 return __ip6gre_bucket(ign, &t->parms);
258 }
259
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
261 {
262 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
263
264 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265 rcu_assign_pointer(*tp, t);
266 }
267
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
269 {
270 struct ip6_tnl __rcu **tp;
271 struct ip6_tnl *iter;
272
273 for (tp = ip6gre_bucket(ign, t);
274 (iter = rtnl_dereference(*tp)) != NULL;
275 tp = &iter->next) {
276 if (t == iter) {
277 rcu_assign_pointer(*tp, t->next);
278 break;
279 }
280 }
281 }
282
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 const struct __ip6_tnl_parm *parms,
285 int type)
286 {
287 const struct in6_addr *remote = &parms->raddr;
288 const struct in6_addr *local = &parms->laddr;
289 __be32 key = parms->i_key;
290 int link = parms->link;
291 struct ip6_tnl *t;
292 struct ip6_tnl __rcu **tp;
293 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
294
295 for (tp = __ip6gre_bucket(ign, parms);
296 (t = rtnl_dereference(*tp)) != NULL;
297 tp = &t->next)
298 if (ipv6_addr_equal(local, &t->parms.laddr) &&
299 ipv6_addr_equal(remote, &t->parms.raddr) &&
300 key == t->parms.i_key &&
301 link == t->parms.link &&
302 type == t->dev->type)
303 break;
304
305 return t;
306 }
307
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309 const struct __ip6_tnl_parm *parms, int create)
310 {
311 struct ip6_tnl *t, *nt;
312 struct net_device *dev;
313 char name[IFNAMSIZ];
314 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315
316 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317 if (t && create)
318 return NULL;
319 if (t || !create)
320 return t;
321
322 if (parms->name[0])
323 strlcpy(name, parms->name, IFNAMSIZ);
324 else
325 strcpy(name, "ip6gre%d");
326
327 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328 ip6gre_tunnel_setup);
329 if (!dev)
330 return NULL;
331
332 dev_net_set(dev, net);
333
334 nt = netdev_priv(dev);
335 nt->parms = *parms;
336 dev->rtnl_link_ops = &ip6gre_link_ops;
337
338 nt->dev = dev;
339 nt->net = dev_net(dev);
340
341 if (register_netdevice(dev) < 0)
342 goto failed_free;
343
344 ip6gre_tnl_link_config(nt, 1);
345
346 /* Can use a lockless transmit, unless we generate output sequences */
347 if (!(nt->parms.o_flags & TUNNEL_SEQ))
348 dev->features |= NETIF_F_LLTX;
349
350 dev_hold(dev);
351 ip6gre_tunnel_link(ign, nt);
352 return nt;
353
354 failed_free:
355 free_netdev(dev);
356 return NULL;
357 }
358
359 static void ip6gre_tunnel_uninit(struct net_device *dev)
360 {
361 struct ip6_tnl *t = netdev_priv(dev);
362 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
363
364 ip6gre_tunnel_unlink(ign, t);
365 dst_cache_reset(&t->dst_cache);
366 dev_put(dev);
367 }
368
369
370 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
371 u8 type, u8 code, int offset, __be32 info)
372 {
373 const struct gre_base_hdr *greh;
374 const struct ipv6hdr *ipv6h;
375 int grehlen = sizeof(*greh);
376 struct ip6_tnl *t;
377 int key_off = 0;
378 __be16 flags;
379 __be32 key;
380
381 if (!pskb_may_pull(skb, offset + grehlen))
382 return;
383 greh = (const struct gre_base_hdr *)(skb->data + offset);
384 flags = greh->flags;
385 if (flags & (GRE_VERSION | GRE_ROUTING))
386 return;
387 if (flags & GRE_CSUM)
388 grehlen += 4;
389 if (flags & GRE_KEY) {
390 key_off = grehlen + offset;
391 grehlen += 4;
392 }
393
394 if (!pskb_may_pull(skb, offset + grehlen))
395 return;
396 ipv6h = (const struct ipv6hdr *)skb->data;
397 greh = (const struct gre_base_hdr *)(skb->data + offset);
398 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
399
400 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
401 key, greh->protocol);
402 if (!t)
403 return;
404
405 switch (type) {
406 __u32 teli;
407 struct ipv6_tlv_tnl_enc_lim *tel;
408 __u32 mtu;
409 case ICMPV6_DEST_UNREACH:
410 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
411 t->parms.name);
412 if (code != ICMPV6_PORT_UNREACH)
413 break;
414 return;
415 case ICMPV6_TIME_EXCEED:
416 if (code == ICMPV6_EXC_HOPLIMIT) {
417 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
418 t->parms.name);
419 break;
420 }
421 return;
422 case ICMPV6_PARAMPROB:
423 teli = 0;
424 if (code == ICMPV6_HDR_FIELD)
425 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
426
427 if (teli && teli == be32_to_cpu(info) - 2) {
428 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
429 if (tel->encap_limit == 0) {
430 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
431 t->parms.name);
432 }
433 } else {
434 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
435 t->parms.name);
436 }
437 return;
438 case ICMPV6_PKT_TOOBIG:
439 mtu = be32_to_cpu(info) - offset - t->tun_hlen;
440 if (t->dev->type == ARPHRD_ETHER)
441 mtu -= ETH_HLEN;
442 if (mtu < IPV6_MIN_MTU)
443 mtu = IPV6_MIN_MTU;
444 t->dev->mtu = mtu;
445 return;
446 }
447
448 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
449 t->err_count++;
450 else
451 t->err_count = 1;
452 t->err_time = jiffies;
453 }
454
455 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
456 {
457 const struct ipv6hdr *ipv6h;
458 struct ip6_tnl *tunnel;
459
460 ipv6h = ipv6_hdr(skb);
461 tunnel = ip6gre_tunnel_lookup(skb->dev,
462 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
463 tpi->proto);
464 if (tunnel) {
465 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
466
467 return PACKET_RCVD;
468 }
469
470 return PACKET_REJECT;
471 }
472
473 static int gre_rcv(struct sk_buff *skb)
474 {
475 struct tnl_ptk_info tpi;
476 bool csum_err = false;
477 int hdr_len;
478
479 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
480 if (hdr_len < 0)
481 goto drop;
482
483 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
484 goto drop;
485
486 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
487 return 0;
488
489 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
490 drop:
491 kfree_skb(skb);
492 return 0;
493 }
494
495 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
496 {
497 return iptunnel_handle_offloads(skb,
498 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
499 }
500
501 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
502 struct net_device *dev, __u8 dsfield,
503 struct flowi6 *fl6, int encap_limit,
504 __u32 *pmtu, __be16 proto)
505 {
506 struct ip6_tnl *tunnel = netdev_priv(dev);
507 struct dst_entry *dst = skb_dst(skb);
508 __be16 protocol;
509
510 if (dev->type == ARPHRD_ETHER)
511 IPCB(skb)->flags = 0;
512
513 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
514 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
515 else
516 fl6->daddr = tunnel->parms.raddr;
517
518 if (tunnel->parms.o_flags & TUNNEL_SEQ)
519 tunnel->o_seqno++;
520
521 /* Push GRE header. */
522 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
523 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
524 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
525
526 /* TooBig packet may have updated dst->dev's mtu */
527 if (dst && dst_mtu(dst) > dst->dev->mtu)
528 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
529
530 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
531 NEXTHDR_GRE);
532 }
533
534 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
535 {
536 struct ip6_tnl *t = netdev_priv(dev);
537 const struct iphdr *iph = ip_hdr(skb);
538 int encap_limit = -1;
539 struct flowi6 fl6;
540 __u8 dsfield;
541 __u32 mtu;
542 int err;
543
544 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
545
546 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
547 encap_limit = t->parms.encap_limit;
548
549 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
550
551 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
552 dsfield = ipv4_get_dsfield(iph);
553 else
554 dsfield = ip6_tclass(t->parms.flowinfo);
555 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
556 fl6.flowi6_mark = skb->mark;
557 else
558 fl6.flowi6_mark = t->parms.fwmark;
559
560 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
561
562 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
563 if (err)
564 return -1;
565
566 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
567 skb->protocol);
568 if (err != 0) {
569 /* XXX: send ICMP error even if DF is not set. */
570 if (err == -EMSGSIZE)
571 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
572 htonl(mtu));
573 return -1;
574 }
575
576 return 0;
577 }
578
579 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
580 {
581 struct ip6_tnl *t = netdev_priv(dev);
582 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
583 int encap_limit = -1;
584 __u16 offset;
585 struct flowi6 fl6;
586 __u8 dsfield;
587 __u32 mtu;
588 int err;
589
590 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
591 return -1;
592
593 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
594 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
595 ipv6h = ipv6_hdr(skb);
596
597 if (offset > 0) {
598 struct ipv6_tlv_tnl_enc_lim *tel;
599 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
600 if (tel->encap_limit == 0) {
601 icmpv6_send(skb, ICMPV6_PARAMPROB,
602 ICMPV6_HDR_FIELD, offset + 2);
603 return -1;
604 }
605 encap_limit = tel->encap_limit - 1;
606 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
607 encap_limit = t->parms.encap_limit;
608
609 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
610
611 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
612 dsfield = ipv6_get_dsfield(ipv6h);
613 else
614 dsfield = ip6_tclass(t->parms.flowinfo);
615
616 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
617 fl6.flowlabel |= ip6_flowlabel(ipv6h);
618 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
619 fl6.flowi6_mark = skb->mark;
620 else
621 fl6.flowi6_mark = t->parms.fwmark;
622
623 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
624
625 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
626 return -1;
627
628 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
629 &mtu, skb->protocol);
630 if (err != 0) {
631 if (err == -EMSGSIZE)
632 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
633 return -1;
634 }
635
636 return 0;
637 }
638
639 /**
640 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
641 * @t: the outgoing tunnel device
642 * @hdr: IPv6 header from the incoming packet
643 *
644 * Description:
645 * Avoid trivial tunneling loop by checking that tunnel exit-point
646 * doesn't match source of incoming packet.
647 *
648 * Return:
649 * 1 if conflict,
650 * 0 else
651 **/
652
653 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
654 const struct ipv6hdr *hdr)
655 {
656 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
657 }
658
659 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
660 {
661 struct ip6_tnl *t = netdev_priv(dev);
662 int encap_limit = -1;
663 struct flowi6 fl6;
664 __u32 mtu;
665 int err;
666
667 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
668 encap_limit = t->parms.encap_limit;
669
670 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
671
672 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
673 if (err)
674 return err;
675
676 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
677
678 return err;
679 }
680
681 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
682 struct net_device *dev)
683 {
684 struct ip6_tnl *t = netdev_priv(dev);
685 struct net_device_stats *stats = &t->dev->stats;
686 int ret;
687
688 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
689 goto tx_err;
690
691 switch (skb->protocol) {
692 case htons(ETH_P_IP):
693 ret = ip6gre_xmit_ipv4(skb, dev);
694 break;
695 case htons(ETH_P_IPV6):
696 ret = ip6gre_xmit_ipv6(skb, dev);
697 break;
698 default:
699 ret = ip6gre_xmit_other(skb, dev);
700 break;
701 }
702
703 if (ret < 0)
704 goto tx_err;
705
706 return NETDEV_TX_OK;
707
708 tx_err:
709 stats->tx_errors++;
710 stats->tx_dropped++;
711 kfree_skb(skb);
712 return NETDEV_TX_OK;
713 }
714
715 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
716 {
717 struct net_device *dev = t->dev;
718 struct __ip6_tnl_parm *p = &t->parms;
719 struct flowi6 *fl6 = &t->fl.u.ip6;
720 int t_hlen;
721
722 if (dev->type != ARPHRD_ETHER) {
723 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
724 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
725 }
726
727 /* Set up flowi template */
728 fl6->saddr = p->laddr;
729 fl6->daddr = p->raddr;
730 fl6->flowi6_oif = p->link;
731 fl6->flowlabel = 0;
732 fl6->flowi6_proto = IPPROTO_GRE;
733
734 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
735 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
736 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
737 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
738
739 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
740 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
741
742 if (p->flags&IP6_TNL_F_CAP_XMIT &&
743 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
744 dev->flags |= IFF_POINTOPOINT;
745 else
746 dev->flags &= ~IFF_POINTOPOINT;
747
748 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
749
750 t->hlen = t->encap_hlen + t->tun_hlen;
751
752 t_hlen = t->hlen + sizeof(struct ipv6hdr);
753
754 if (p->flags & IP6_TNL_F_CAP_XMIT) {
755 int strict = (ipv6_addr_type(&p->raddr) &
756 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
757
758 struct rt6_info *rt = rt6_lookup(t->net,
759 &p->raddr, &p->laddr,
760 p->link, strict);
761
762 if (!rt)
763 return;
764
765 if (rt->dst.dev) {
766 dev->hard_header_len = rt->dst.dev->hard_header_len +
767 t_hlen;
768
769 if (set_mtu) {
770 dev->mtu = rt->dst.dev->mtu - t_hlen;
771 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
772 dev->mtu -= 8;
773 if (dev->type == ARPHRD_ETHER)
774 dev->mtu -= ETH_HLEN;
775
776 if (dev->mtu < IPV6_MIN_MTU)
777 dev->mtu = IPV6_MIN_MTU;
778 }
779 }
780 ip6_rt_put(rt);
781 }
782 }
783
784 static int ip6gre_tnl_change(struct ip6_tnl *t,
785 const struct __ip6_tnl_parm *p, int set_mtu)
786 {
787 t->parms.laddr = p->laddr;
788 t->parms.raddr = p->raddr;
789 t->parms.flags = p->flags;
790 t->parms.hop_limit = p->hop_limit;
791 t->parms.encap_limit = p->encap_limit;
792 t->parms.flowinfo = p->flowinfo;
793 t->parms.link = p->link;
794 t->parms.proto = p->proto;
795 t->parms.i_key = p->i_key;
796 t->parms.o_key = p->o_key;
797 t->parms.i_flags = p->i_flags;
798 t->parms.o_flags = p->o_flags;
799 t->parms.fwmark = p->fwmark;
800 dst_cache_reset(&t->dst_cache);
801 ip6gre_tnl_link_config(t, set_mtu);
802 return 0;
803 }
804
805 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
806 const struct ip6_tnl_parm2 *u)
807 {
808 p->laddr = u->laddr;
809 p->raddr = u->raddr;
810 p->flags = u->flags;
811 p->hop_limit = u->hop_limit;
812 p->encap_limit = u->encap_limit;
813 p->flowinfo = u->flowinfo;
814 p->link = u->link;
815 p->i_key = u->i_key;
816 p->o_key = u->o_key;
817 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
818 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
819 memcpy(p->name, u->name, sizeof(u->name));
820 }
821
822 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
823 const struct __ip6_tnl_parm *p)
824 {
825 u->proto = IPPROTO_GRE;
826 u->laddr = p->laddr;
827 u->raddr = p->raddr;
828 u->flags = p->flags;
829 u->hop_limit = p->hop_limit;
830 u->encap_limit = p->encap_limit;
831 u->flowinfo = p->flowinfo;
832 u->link = p->link;
833 u->i_key = p->i_key;
834 u->o_key = p->o_key;
835 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
836 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
837 memcpy(u->name, p->name, sizeof(u->name));
838 }
839
840 static int ip6gre_tunnel_ioctl(struct net_device *dev,
841 struct ifreq *ifr, int cmd)
842 {
843 int err = 0;
844 struct ip6_tnl_parm2 p;
845 struct __ip6_tnl_parm p1;
846 struct ip6_tnl *t = netdev_priv(dev);
847 struct net *net = t->net;
848 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
849
850 memset(&p1, 0, sizeof(p1));
851
852 switch (cmd) {
853 case SIOCGETTUNNEL:
854 if (dev == ign->fb_tunnel_dev) {
855 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
856 err = -EFAULT;
857 break;
858 }
859 ip6gre_tnl_parm_from_user(&p1, &p);
860 t = ip6gre_tunnel_locate(net, &p1, 0);
861 if (!t)
862 t = netdev_priv(dev);
863 }
864 memset(&p, 0, sizeof(p));
865 ip6gre_tnl_parm_to_user(&p, &t->parms);
866 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
867 err = -EFAULT;
868 break;
869
870 case SIOCADDTUNNEL:
871 case SIOCCHGTUNNEL:
872 err = -EPERM;
873 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
874 goto done;
875
876 err = -EFAULT;
877 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
878 goto done;
879
880 err = -EINVAL;
881 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
882 goto done;
883
884 if (!(p.i_flags&GRE_KEY))
885 p.i_key = 0;
886 if (!(p.o_flags&GRE_KEY))
887 p.o_key = 0;
888
889 ip6gre_tnl_parm_from_user(&p1, &p);
890 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
891
892 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
893 if (t) {
894 if (t->dev != dev) {
895 err = -EEXIST;
896 break;
897 }
898 } else {
899 t = netdev_priv(dev);
900
901 ip6gre_tunnel_unlink(ign, t);
902 synchronize_net();
903 ip6gre_tnl_change(t, &p1, 1);
904 ip6gre_tunnel_link(ign, t);
905 netdev_state_change(dev);
906 }
907 }
908
909 if (t) {
910 err = 0;
911
912 memset(&p, 0, sizeof(p));
913 ip6gre_tnl_parm_to_user(&p, &t->parms);
914 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
915 err = -EFAULT;
916 } else
917 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
918 break;
919
920 case SIOCDELTUNNEL:
921 err = -EPERM;
922 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
923 goto done;
924
925 if (dev == ign->fb_tunnel_dev) {
926 err = -EFAULT;
927 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
928 goto done;
929 err = -ENOENT;
930 ip6gre_tnl_parm_from_user(&p1, &p);
931 t = ip6gre_tunnel_locate(net, &p1, 0);
932 if (!t)
933 goto done;
934 err = -EPERM;
935 if (t == netdev_priv(ign->fb_tunnel_dev))
936 goto done;
937 dev = t->dev;
938 }
939 unregister_netdevice(dev);
940 err = 0;
941 break;
942
943 default:
944 err = -EINVAL;
945 }
946
947 done:
948 return err;
949 }
950
951 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
952 unsigned short type, const void *daddr,
953 const void *saddr, unsigned int len)
954 {
955 struct ip6_tnl *t = netdev_priv(dev);
956 struct ipv6hdr *ipv6h;
957 __be16 *p;
958
959 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
960 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
961 t->fl.u.ip6.flowlabel,
962 true, &t->fl.u.ip6));
963 ipv6h->hop_limit = t->parms.hop_limit;
964 ipv6h->nexthdr = NEXTHDR_GRE;
965 ipv6h->saddr = t->parms.laddr;
966 ipv6h->daddr = t->parms.raddr;
967
968 p = (__be16 *)(ipv6h + 1);
969 p[0] = t->parms.o_flags;
970 p[1] = htons(type);
971
972 /*
973 * Set the source hardware address.
974 */
975
976 if (saddr)
977 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
978 if (daddr)
979 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
980 if (!ipv6_addr_any(&ipv6h->daddr))
981 return t->hlen;
982
983 return -t->hlen;
984 }
985
986 static const struct header_ops ip6gre_header_ops = {
987 .create = ip6gre_header,
988 };
989
990 static const struct net_device_ops ip6gre_netdev_ops = {
991 .ndo_init = ip6gre_tunnel_init,
992 .ndo_uninit = ip6gre_tunnel_uninit,
993 .ndo_start_xmit = ip6gre_tunnel_xmit,
994 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
995 .ndo_change_mtu = ip6_tnl_change_mtu,
996 .ndo_get_stats64 = ip_tunnel_get_stats64,
997 .ndo_get_iflink = ip6_tnl_get_iflink,
998 };
999
1000 static void ip6gre_dev_free(struct net_device *dev)
1001 {
1002 struct ip6_tnl *t = netdev_priv(dev);
1003
1004 dst_cache_destroy(&t->dst_cache);
1005 free_percpu(dev->tstats);
1006 }
1007
1008 static void ip6gre_tunnel_setup(struct net_device *dev)
1009 {
1010 dev->netdev_ops = &ip6gre_netdev_ops;
1011 dev->needs_free_netdev = true;
1012 dev->priv_destructor = ip6gre_dev_free;
1013
1014 dev->type = ARPHRD_IP6GRE;
1015
1016 dev->flags |= IFF_NOARP;
1017 dev->addr_len = sizeof(struct in6_addr);
1018 netif_keep_dst(dev);
1019 /* This perm addr will be used as interface identifier by IPv6 */
1020 dev->addr_assign_type = NET_ADDR_RANDOM;
1021 eth_random_addr(dev->perm_addr);
1022 }
1023
1024 #define GRE6_FEATURES (NETIF_F_SG | \
1025 NETIF_F_FRAGLIST | \
1026 NETIF_F_HIGHDMA | \
1027 NETIF_F_HW_CSUM)
1028
1029 static void ip6gre_tnl_init_features(struct net_device *dev)
1030 {
1031 struct ip6_tnl *nt = netdev_priv(dev);
1032
1033 dev->features |= GRE6_FEATURES;
1034 dev->hw_features |= GRE6_FEATURES;
1035
1036 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1037 /* TCP offload with GRE SEQ is not supported, nor
1038 * can we support 2 levels of outer headers requiring
1039 * an update.
1040 */
1041 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1042 nt->encap.type == TUNNEL_ENCAP_NONE) {
1043 dev->features |= NETIF_F_GSO_SOFTWARE;
1044 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1045 }
1046
1047 /* Can use a lockless transmit, unless we generate
1048 * output sequences
1049 */
1050 dev->features |= NETIF_F_LLTX;
1051 }
1052 }
1053
1054 static int ip6gre_tunnel_init_common(struct net_device *dev)
1055 {
1056 struct ip6_tnl *tunnel;
1057 int ret;
1058 int t_hlen;
1059
1060 tunnel = netdev_priv(dev);
1061
1062 tunnel->dev = dev;
1063 tunnel->net = dev_net(dev);
1064 strcpy(tunnel->parms.name, dev->name);
1065
1066 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1067 if (!dev->tstats)
1068 return -ENOMEM;
1069
1070 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1071 if (ret) {
1072 free_percpu(dev->tstats);
1073 dev->tstats = NULL;
1074 return ret;
1075 }
1076
1077 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1078 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1079 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1080
1081 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1082 dev->mtu = ETH_DATA_LEN - t_hlen;
1083 if (dev->type == ARPHRD_ETHER)
1084 dev->mtu -= ETH_HLEN;
1085 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1086 dev->mtu -= 8;
1087
1088 ip6gre_tnl_init_features(dev);
1089
1090 return 0;
1091 }
1092
1093 static int ip6gre_tunnel_init(struct net_device *dev)
1094 {
1095 struct ip6_tnl *tunnel;
1096 int ret;
1097
1098 ret = ip6gre_tunnel_init_common(dev);
1099 if (ret)
1100 return ret;
1101
1102 tunnel = netdev_priv(dev);
1103
1104 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1105 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1106
1107 if (ipv6_addr_any(&tunnel->parms.raddr))
1108 dev->header_ops = &ip6gre_header_ops;
1109
1110 return 0;
1111 }
1112
1113 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1114 {
1115 struct ip6_tnl *tunnel = netdev_priv(dev);
1116
1117 tunnel->dev = dev;
1118 tunnel->net = dev_net(dev);
1119 strcpy(tunnel->parms.name, dev->name);
1120
1121 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1122
1123 dev_hold(dev);
1124 }
1125
1126
1127 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1128 .handler = gre_rcv,
1129 .err_handler = ip6gre_err,
1130 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1131 };
1132
1133 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1134 {
1135 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1136 struct net_device *dev, *aux;
1137 int prio;
1138
1139 for_each_netdev_safe(net, dev, aux)
1140 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1141 dev->rtnl_link_ops == &ip6gre_tap_ops)
1142 unregister_netdevice_queue(dev, head);
1143
1144 for (prio = 0; prio < 4; prio++) {
1145 int h;
1146 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1147 struct ip6_tnl *t;
1148
1149 t = rtnl_dereference(ign->tunnels[prio][h]);
1150
1151 while (t) {
1152 /* If dev is in the same netns, it has already
1153 * been added to the list by the previous loop.
1154 */
1155 if (!net_eq(dev_net(t->dev), net))
1156 unregister_netdevice_queue(t->dev,
1157 head);
1158 t = rtnl_dereference(t->next);
1159 }
1160 }
1161 }
1162 }
1163
1164 static int __net_init ip6gre_init_net(struct net *net)
1165 {
1166 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1167 int err;
1168
1169 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1170 NET_NAME_UNKNOWN,
1171 ip6gre_tunnel_setup);
1172 if (!ign->fb_tunnel_dev) {
1173 err = -ENOMEM;
1174 goto err_alloc_dev;
1175 }
1176 dev_net_set(ign->fb_tunnel_dev, net);
1177 /* FB netdevice is special: we have one, and only one per netns.
1178 * Allowing to move it to another netns is clearly unsafe.
1179 */
1180 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1181
1182
1183 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1184 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1185
1186 err = register_netdev(ign->fb_tunnel_dev);
1187 if (err)
1188 goto err_reg_dev;
1189
1190 rcu_assign_pointer(ign->tunnels_wc[0],
1191 netdev_priv(ign->fb_tunnel_dev));
1192 return 0;
1193
1194 err_reg_dev:
1195 free_netdev(ign->fb_tunnel_dev);
1196 err_alloc_dev:
1197 return err;
1198 }
1199
1200 static void __net_exit ip6gre_exit_net(struct net *net)
1201 {
1202 LIST_HEAD(list);
1203
1204 rtnl_lock();
1205 ip6gre_destroy_tunnels(net, &list);
1206 unregister_netdevice_many(&list);
1207 rtnl_unlock();
1208 }
1209
1210 static struct pernet_operations ip6gre_net_ops = {
1211 .init = ip6gre_init_net,
1212 .exit = ip6gre_exit_net,
1213 .id = &ip6gre_net_id,
1214 .size = sizeof(struct ip6gre_net),
1215 };
1216
1217 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1218 struct netlink_ext_ack *extack)
1219 {
1220 __be16 flags;
1221
1222 if (!data)
1223 return 0;
1224
1225 flags = 0;
1226 if (data[IFLA_GRE_IFLAGS])
1227 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1228 if (data[IFLA_GRE_OFLAGS])
1229 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1230 if (flags & (GRE_VERSION|GRE_ROUTING))
1231 return -EINVAL;
1232
1233 return 0;
1234 }
1235
1236 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1237 struct netlink_ext_ack *extack)
1238 {
1239 struct in6_addr daddr;
1240
1241 if (tb[IFLA_ADDRESS]) {
1242 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1243 return -EINVAL;
1244 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1245 return -EADDRNOTAVAIL;
1246 }
1247
1248 if (!data)
1249 goto out;
1250
1251 if (data[IFLA_GRE_REMOTE]) {
1252 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1253 if (ipv6_addr_any(&daddr))
1254 return -EINVAL;
1255 }
1256
1257 out:
1258 return ip6gre_tunnel_validate(tb, data, extack);
1259 }
1260
1261
1262 static void ip6gre_netlink_parms(struct nlattr *data[],
1263 struct __ip6_tnl_parm *parms)
1264 {
1265 memset(parms, 0, sizeof(*parms));
1266
1267 if (!data)
1268 return;
1269
1270 if (data[IFLA_GRE_LINK])
1271 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1272
1273 if (data[IFLA_GRE_IFLAGS])
1274 parms->i_flags = gre_flags_to_tnl_flags(
1275 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1276
1277 if (data[IFLA_GRE_OFLAGS])
1278 parms->o_flags = gre_flags_to_tnl_flags(
1279 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1280
1281 if (data[IFLA_GRE_IKEY])
1282 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1283
1284 if (data[IFLA_GRE_OKEY])
1285 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1286
1287 if (data[IFLA_GRE_LOCAL])
1288 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1289
1290 if (data[IFLA_GRE_REMOTE])
1291 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1292
1293 if (data[IFLA_GRE_TTL])
1294 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1295
1296 if (data[IFLA_GRE_ENCAP_LIMIT])
1297 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1298
1299 if (data[IFLA_GRE_FLOWINFO])
1300 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1301
1302 if (data[IFLA_GRE_FLAGS])
1303 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1304
1305 if (data[IFLA_GRE_FWMARK])
1306 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1307 }
1308
1309 static int ip6gre_tap_init(struct net_device *dev)
1310 {
1311 int ret;
1312
1313 ret = ip6gre_tunnel_init_common(dev);
1314 if (ret)
1315 return ret;
1316
1317 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1318
1319 return 0;
1320 }
1321
1322 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1323 .ndo_init = ip6gre_tap_init,
1324 .ndo_uninit = ip6gre_tunnel_uninit,
1325 .ndo_start_xmit = ip6gre_tunnel_xmit,
1326 .ndo_set_mac_address = eth_mac_addr,
1327 .ndo_validate_addr = eth_validate_addr,
1328 .ndo_change_mtu = ip6_tnl_change_mtu,
1329 .ndo_get_stats64 = ip_tunnel_get_stats64,
1330 .ndo_get_iflink = ip6_tnl_get_iflink,
1331 };
1332
1333 static void ip6gre_tap_setup(struct net_device *dev)
1334 {
1335
1336 ether_setup(dev);
1337
1338 dev->max_mtu = 0;
1339 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1340 dev->needs_free_netdev = true;
1341 dev->priv_destructor = ip6gre_dev_free;
1342
1343 dev->features |= NETIF_F_NETNS_LOCAL;
1344 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1345 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1346 netif_keep_dst(dev);
1347 }
1348
1349 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1350 struct ip_tunnel_encap *ipencap)
1351 {
1352 bool ret = false;
1353
1354 memset(ipencap, 0, sizeof(*ipencap));
1355
1356 if (!data)
1357 return ret;
1358
1359 if (data[IFLA_GRE_ENCAP_TYPE]) {
1360 ret = true;
1361 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1362 }
1363
1364 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1365 ret = true;
1366 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1367 }
1368
1369 if (data[IFLA_GRE_ENCAP_SPORT]) {
1370 ret = true;
1371 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1372 }
1373
1374 if (data[IFLA_GRE_ENCAP_DPORT]) {
1375 ret = true;
1376 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1377 }
1378
1379 return ret;
1380 }
1381
1382 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1383 struct nlattr *tb[], struct nlattr *data[],
1384 struct netlink_ext_ack *extack)
1385 {
1386 struct ip6_tnl *nt;
1387 struct net *net = dev_net(dev);
1388 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1389 struct ip_tunnel_encap ipencap;
1390 int err;
1391
1392 nt = netdev_priv(dev);
1393
1394 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1395 int err = ip6_tnl_encap_setup(nt, &ipencap);
1396
1397 if (err < 0)
1398 return err;
1399 }
1400
1401 ip6gre_netlink_parms(data, &nt->parms);
1402
1403 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1404 return -EEXIST;
1405
1406 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1407 eth_hw_addr_random(dev);
1408
1409 nt->dev = dev;
1410 nt->net = dev_net(dev);
1411
1412 err = register_netdevice(dev);
1413 if (err)
1414 goto out;
1415
1416 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1417
1418 if (tb[IFLA_MTU])
1419 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1420
1421 dev_hold(dev);
1422 ip6gre_tunnel_link(ign, nt);
1423
1424 out:
1425 return err;
1426 }
1427
1428 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1429 struct nlattr *data[],
1430 struct netlink_ext_ack *extack)
1431 {
1432 struct ip6_tnl *t, *nt = netdev_priv(dev);
1433 struct net *net = nt->net;
1434 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1435 struct __ip6_tnl_parm p;
1436 struct ip_tunnel_encap ipencap;
1437
1438 if (dev == ign->fb_tunnel_dev)
1439 return -EINVAL;
1440
1441 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1442 int err = ip6_tnl_encap_setup(nt, &ipencap);
1443
1444 if (err < 0)
1445 return err;
1446 }
1447
1448 ip6gre_netlink_parms(data, &p);
1449
1450 t = ip6gre_tunnel_locate(net, &p, 0);
1451
1452 if (t) {
1453 if (t->dev != dev)
1454 return -EEXIST;
1455 } else {
1456 t = nt;
1457 }
1458
1459 ip6gre_tunnel_unlink(ign, t);
1460 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1461 ip6gre_tunnel_link(ign, t);
1462 return 0;
1463 }
1464
1465 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1466 {
1467 struct net *net = dev_net(dev);
1468 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1469
1470 if (dev != ign->fb_tunnel_dev)
1471 unregister_netdevice_queue(dev, head);
1472 }
1473
1474 static size_t ip6gre_get_size(const struct net_device *dev)
1475 {
1476 return
1477 /* IFLA_GRE_LINK */
1478 nla_total_size(4) +
1479 /* IFLA_GRE_IFLAGS */
1480 nla_total_size(2) +
1481 /* IFLA_GRE_OFLAGS */
1482 nla_total_size(2) +
1483 /* IFLA_GRE_IKEY */
1484 nla_total_size(4) +
1485 /* IFLA_GRE_OKEY */
1486 nla_total_size(4) +
1487 /* IFLA_GRE_LOCAL */
1488 nla_total_size(sizeof(struct in6_addr)) +
1489 /* IFLA_GRE_REMOTE */
1490 nla_total_size(sizeof(struct in6_addr)) +
1491 /* IFLA_GRE_TTL */
1492 nla_total_size(1) +
1493 /* IFLA_GRE_ENCAP_LIMIT */
1494 nla_total_size(1) +
1495 /* IFLA_GRE_FLOWINFO */
1496 nla_total_size(4) +
1497 /* IFLA_GRE_FLAGS */
1498 nla_total_size(4) +
1499 /* IFLA_GRE_ENCAP_TYPE */
1500 nla_total_size(2) +
1501 /* IFLA_GRE_ENCAP_FLAGS */
1502 nla_total_size(2) +
1503 /* IFLA_GRE_ENCAP_SPORT */
1504 nla_total_size(2) +
1505 /* IFLA_GRE_ENCAP_DPORT */
1506 nla_total_size(2) +
1507 /* IFLA_GRE_FWMARK */
1508 nla_total_size(4) +
1509 0;
1510 }
1511
1512 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1513 {
1514 struct ip6_tnl *t = netdev_priv(dev);
1515 struct __ip6_tnl_parm *p = &t->parms;
1516
1517 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1518 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1519 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1520 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1521 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1522 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1523 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1524 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1525 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1526 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1527 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1528 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1529 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1530 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1531 goto nla_put_failure;
1532
1533 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1534 t->encap.type) ||
1535 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1536 t->encap.sport) ||
1537 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1538 t->encap.dport) ||
1539 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1540 t->encap.flags))
1541 goto nla_put_failure;
1542
1543 return 0;
1544
1545 nla_put_failure:
1546 return -EMSGSIZE;
1547 }
1548
1549 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1550 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1551 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1552 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1553 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1554 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1555 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1556 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1557 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1558 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1559 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1560 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1561 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1562 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1563 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1564 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1565 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1566 };
1567
1568 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1569 .kind = "ip6gre",
1570 .maxtype = IFLA_GRE_MAX,
1571 .policy = ip6gre_policy,
1572 .priv_size = sizeof(struct ip6_tnl),
1573 .setup = ip6gre_tunnel_setup,
1574 .validate = ip6gre_tunnel_validate,
1575 .newlink = ip6gre_newlink,
1576 .changelink = ip6gre_changelink,
1577 .dellink = ip6gre_dellink,
1578 .get_size = ip6gre_get_size,
1579 .fill_info = ip6gre_fill_info,
1580 .get_link_net = ip6_tnl_get_link_net,
1581 };
1582
1583 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1584 .kind = "ip6gretap",
1585 .maxtype = IFLA_GRE_MAX,
1586 .policy = ip6gre_policy,
1587 .priv_size = sizeof(struct ip6_tnl),
1588 .setup = ip6gre_tap_setup,
1589 .validate = ip6gre_tap_validate,
1590 .newlink = ip6gre_newlink,
1591 .changelink = ip6gre_changelink,
1592 .get_size = ip6gre_get_size,
1593 .fill_info = ip6gre_fill_info,
1594 .get_link_net = ip6_tnl_get_link_net,
1595 };
1596
1597 /*
1598 * And now the modules code and kernel interface.
1599 */
1600
1601 static int __init ip6gre_init(void)
1602 {
1603 int err;
1604
1605 pr_info("GRE over IPv6 tunneling driver\n");
1606
1607 err = register_pernet_device(&ip6gre_net_ops);
1608 if (err < 0)
1609 return err;
1610
1611 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1612 if (err < 0) {
1613 pr_info("%s: can't add protocol\n", __func__);
1614 goto add_proto_failed;
1615 }
1616
1617 err = rtnl_link_register(&ip6gre_link_ops);
1618 if (err < 0)
1619 goto rtnl_link_failed;
1620
1621 err = rtnl_link_register(&ip6gre_tap_ops);
1622 if (err < 0)
1623 goto tap_ops_failed;
1624
1625 out:
1626 return err;
1627
1628 tap_ops_failed:
1629 rtnl_link_unregister(&ip6gre_link_ops);
1630 rtnl_link_failed:
1631 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1632 add_proto_failed:
1633 unregister_pernet_device(&ip6gre_net_ops);
1634 goto out;
1635 }
1636
1637 static void __exit ip6gre_fini(void)
1638 {
1639 rtnl_link_unregister(&ip6gre_tap_ops);
1640 rtnl_link_unregister(&ip6gre_link_ops);
1641 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1642 unregister_pernet_device(&ip6gre_net_ops);
1643 }
1644
1645 module_init(ip6gre_init);
1646 module_exit(ip6gre_fini);
1647 MODULE_LICENSE("GPL");
1648 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1649 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1650 MODULE_ALIAS_RTNL_LINK("ip6gre");
1651 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1652 MODULE_ALIAS_NETDEV("ip6gre0");