]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv6/ip6_gre.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ip6_gre: only increase err_count for some certain type icmpv6 in ip6gre_err
[mirror_ubuntu-artful-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
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 ip6gre_tnl_link_config(nt, 1);
341
342 if (register_netdevice(dev) < 0)
343 goto failed_free;
344
345 /* Can use a lockless transmit, unless we generate output sequences */
346 if (!(nt->parms.o_flags & TUNNEL_SEQ))
347 dev->features |= NETIF_F_LLTX;
348
349 dev_hold(dev);
350 ip6gre_tunnel_link(ign, nt);
351 return nt;
352
353 failed_free:
354 free_netdev(dev);
355 return NULL;
356 }
357
358 static void ip6gre_tunnel_uninit(struct net_device *dev)
359 {
360 struct ip6_tnl *t = netdev_priv(dev);
361 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
362
363 ip6gre_tunnel_unlink(ign, t);
364 dst_cache_reset(&t->dst_cache);
365 dev_put(dev);
366 }
367
368
369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
370 u8 type, u8 code, int offset, __be32 info)
371 {
372 const struct gre_base_hdr *greh;
373 const struct ipv6hdr *ipv6h;
374 int grehlen = sizeof(*greh);
375 struct ip6_tnl *t;
376 int key_off = 0;
377 __be16 flags;
378 __be32 key;
379
380 if (!pskb_may_pull(skb, offset + grehlen))
381 return;
382 greh = (const struct gre_base_hdr *)(skb->data + offset);
383 flags = greh->flags;
384 if (flags & (GRE_VERSION | GRE_ROUTING))
385 return;
386 if (flags & GRE_CSUM)
387 grehlen += 4;
388 if (flags & GRE_KEY) {
389 key_off = grehlen + offset;
390 grehlen += 4;
391 }
392
393 if (!pskb_may_pull(skb, offset + grehlen))
394 return;
395 ipv6h = (const struct ipv6hdr *)skb->data;
396 greh = (const struct gre_base_hdr *)(skb->data + offset);
397 key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
398
399 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
400 key, greh->protocol);
401 if (!t)
402 return;
403
404 switch (type) {
405 __u32 teli;
406 struct ipv6_tlv_tnl_enc_lim *tel;
407 __u32 mtu;
408 case ICMPV6_DEST_UNREACH:
409 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
410 t->parms.name);
411 if (code != ICMPV6_PORT_UNREACH)
412 break;
413 return;
414 case ICMPV6_TIME_EXCEED:
415 if (code == ICMPV6_EXC_HOPLIMIT) {
416 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
417 t->parms.name);
418 break;
419 }
420 return;
421 case ICMPV6_PARAMPROB:
422 teli = 0;
423 if (code == ICMPV6_HDR_FIELD)
424 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
425
426 if (teli && teli == be32_to_cpu(info) - 2) {
427 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
428 if (tel->encap_limit == 0) {
429 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
430 t->parms.name);
431 }
432 } else {
433 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
434 t->parms.name);
435 }
436 return;
437 case ICMPV6_PKT_TOOBIG:
438 mtu = be32_to_cpu(info) - offset - t->tun_hlen;
439 if (t->dev->type == ARPHRD_ETHER)
440 mtu -= ETH_HLEN;
441 if (mtu < IPV6_MIN_MTU)
442 mtu = IPV6_MIN_MTU;
443 t->dev->mtu = mtu;
444 return;
445 }
446
447 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
448 t->err_count++;
449 else
450 t->err_count = 1;
451 t->err_time = jiffies;
452 }
453
454 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
455 {
456 const struct ipv6hdr *ipv6h;
457 struct ip6_tnl *tunnel;
458
459 ipv6h = ipv6_hdr(skb);
460 tunnel = ip6gre_tunnel_lookup(skb->dev,
461 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
462 tpi->proto);
463 if (tunnel) {
464 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
465
466 return PACKET_RCVD;
467 }
468
469 return PACKET_REJECT;
470 }
471
472 static int gre_rcv(struct sk_buff *skb)
473 {
474 struct tnl_ptk_info tpi;
475 bool csum_err = false;
476 int hdr_len;
477
478 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
479 if (hdr_len < 0)
480 goto drop;
481
482 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
483 goto drop;
484
485 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
486 return 0;
487
488 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
489 drop:
490 kfree_skb(skb);
491 return 0;
492 }
493
494 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
495 {
496 return iptunnel_handle_offloads(skb,
497 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
498 }
499
500 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
501 struct net_device *dev, __u8 dsfield,
502 struct flowi6 *fl6, int encap_limit,
503 __u32 *pmtu, __be16 proto)
504 {
505 struct ip6_tnl *tunnel = netdev_priv(dev);
506 __be16 protocol = (dev->type == ARPHRD_ETHER) ?
507 htons(ETH_P_TEB) : proto;
508
509 if (dev->type == ARPHRD_ETHER)
510 IPCB(skb)->flags = 0;
511
512 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
513 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
514 else
515 fl6->daddr = tunnel->parms.raddr;
516
517 if (tunnel->parms.o_flags & TUNNEL_SEQ)
518 tunnel->o_seqno++;
519
520 /* Push GRE header. */
521 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
522 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
523
524 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
525 NEXTHDR_GRE);
526 }
527
528 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
529 {
530 struct ip6_tnl *t = netdev_priv(dev);
531 const struct iphdr *iph = ip_hdr(skb);
532 int encap_limit = -1;
533 struct flowi6 fl6;
534 __u8 dsfield;
535 __u32 mtu;
536 int err;
537
538 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
539
540 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
541 encap_limit = t->parms.encap_limit;
542
543 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
544
545 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
546 dsfield = ipv4_get_dsfield(iph);
547 else
548 dsfield = ip6_tclass(t->parms.flowinfo);
549 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
550 fl6.flowi6_mark = skb->mark;
551 else
552 fl6.flowi6_mark = t->parms.fwmark;
553
554 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
555
556 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
557 if (err)
558 return -1;
559
560 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
561 skb->protocol);
562 if (err != 0) {
563 /* XXX: send ICMP error even if DF is not set. */
564 if (err == -EMSGSIZE)
565 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
566 htonl(mtu));
567 return -1;
568 }
569
570 return 0;
571 }
572
573 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
574 {
575 struct ip6_tnl *t = netdev_priv(dev);
576 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
577 int encap_limit = -1;
578 __u16 offset;
579 struct flowi6 fl6;
580 __u8 dsfield;
581 __u32 mtu;
582 int err;
583
584 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
585 return -1;
586
587 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
588 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
589 ipv6h = ipv6_hdr(skb);
590
591 if (offset > 0) {
592 struct ipv6_tlv_tnl_enc_lim *tel;
593 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
594 if (tel->encap_limit == 0) {
595 icmpv6_send(skb, ICMPV6_PARAMPROB,
596 ICMPV6_HDR_FIELD, offset + 2);
597 return -1;
598 }
599 encap_limit = tel->encap_limit - 1;
600 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
601 encap_limit = t->parms.encap_limit;
602
603 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
604
605 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
606 dsfield = ipv6_get_dsfield(ipv6h);
607 else
608 dsfield = ip6_tclass(t->parms.flowinfo);
609
610 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
611 fl6.flowlabel |= ip6_flowlabel(ipv6h);
612 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
613 fl6.flowi6_mark = skb->mark;
614 else
615 fl6.flowi6_mark = t->parms.fwmark;
616
617 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
618
619 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
620 return -1;
621
622 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
623 &mtu, skb->protocol);
624 if (err != 0) {
625 if (err == -EMSGSIZE)
626 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
627 return -1;
628 }
629
630 return 0;
631 }
632
633 /**
634 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
635 * @t: the outgoing tunnel device
636 * @hdr: IPv6 header from the incoming packet
637 *
638 * Description:
639 * Avoid trivial tunneling loop by checking that tunnel exit-point
640 * doesn't match source of incoming packet.
641 *
642 * Return:
643 * 1 if conflict,
644 * 0 else
645 **/
646
647 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
648 const struct ipv6hdr *hdr)
649 {
650 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
651 }
652
653 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
654 {
655 struct ip6_tnl *t = netdev_priv(dev);
656 int encap_limit = -1;
657 struct flowi6 fl6;
658 __u32 mtu;
659 int err;
660
661 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
662 encap_limit = t->parms.encap_limit;
663
664 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
665
666 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
667 if (err)
668 return err;
669
670 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
671
672 return err;
673 }
674
675 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
676 struct net_device *dev)
677 {
678 struct ip6_tnl *t = netdev_priv(dev);
679 struct net_device_stats *stats = &t->dev->stats;
680 int ret;
681
682 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
683 goto tx_err;
684
685 switch (skb->protocol) {
686 case htons(ETH_P_IP):
687 ret = ip6gre_xmit_ipv4(skb, dev);
688 break;
689 case htons(ETH_P_IPV6):
690 ret = ip6gre_xmit_ipv6(skb, dev);
691 break;
692 default:
693 ret = ip6gre_xmit_other(skb, dev);
694 break;
695 }
696
697 if (ret < 0)
698 goto tx_err;
699
700 return NETDEV_TX_OK;
701
702 tx_err:
703 stats->tx_errors++;
704 stats->tx_dropped++;
705 kfree_skb(skb);
706 return NETDEV_TX_OK;
707 }
708
709 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
710 {
711 struct net_device *dev = t->dev;
712 struct __ip6_tnl_parm *p = &t->parms;
713 struct flowi6 *fl6 = &t->fl.u.ip6;
714 int t_hlen;
715
716 if (dev->type != ARPHRD_ETHER) {
717 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
718 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
719 }
720
721 /* Set up flowi template */
722 fl6->saddr = p->laddr;
723 fl6->daddr = p->raddr;
724 fl6->flowi6_oif = p->link;
725 fl6->flowlabel = 0;
726 fl6->flowi6_proto = IPPROTO_GRE;
727
728 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
729 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
730 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
731 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
732
733 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
734 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
735
736 if (p->flags&IP6_TNL_F_CAP_XMIT &&
737 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
738 dev->flags |= IFF_POINTOPOINT;
739 else
740 dev->flags &= ~IFF_POINTOPOINT;
741
742 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
743
744 t->hlen = t->encap_hlen + t->tun_hlen;
745
746 t_hlen = t->hlen + sizeof(struct ipv6hdr);
747
748 if (p->flags & IP6_TNL_F_CAP_XMIT) {
749 int strict = (ipv6_addr_type(&p->raddr) &
750 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
751
752 struct rt6_info *rt = rt6_lookup(t->net,
753 &p->raddr, &p->laddr,
754 p->link, strict);
755
756 if (!rt)
757 return;
758
759 if (rt->dst.dev) {
760 dev->hard_header_len = rt->dst.dev->hard_header_len +
761 t_hlen;
762
763 if (set_mtu) {
764 dev->mtu = rt->dst.dev->mtu - t_hlen;
765 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
766 dev->mtu -= 8;
767 if (dev->type == ARPHRD_ETHER)
768 dev->mtu -= ETH_HLEN;
769
770 if (dev->mtu < IPV6_MIN_MTU)
771 dev->mtu = IPV6_MIN_MTU;
772 }
773 }
774 ip6_rt_put(rt);
775 }
776 }
777
778 static int ip6gre_tnl_change(struct ip6_tnl *t,
779 const struct __ip6_tnl_parm *p, int set_mtu)
780 {
781 t->parms.laddr = p->laddr;
782 t->parms.raddr = p->raddr;
783 t->parms.flags = p->flags;
784 t->parms.hop_limit = p->hop_limit;
785 t->parms.encap_limit = p->encap_limit;
786 t->parms.flowinfo = p->flowinfo;
787 t->parms.link = p->link;
788 t->parms.proto = p->proto;
789 t->parms.i_key = p->i_key;
790 t->parms.o_key = p->o_key;
791 t->parms.i_flags = p->i_flags;
792 t->parms.o_flags = p->o_flags;
793 t->parms.fwmark = p->fwmark;
794 dst_cache_reset(&t->dst_cache);
795 ip6gre_tnl_link_config(t, set_mtu);
796 return 0;
797 }
798
799 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
800 const struct ip6_tnl_parm2 *u)
801 {
802 p->laddr = u->laddr;
803 p->raddr = u->raddr;
804 p->flags = u->flags;
805 p->hop_limit = u->hop_limit;
806 p->encap_limit = u->encap_limit;
807 p->flowinfo = u->flowinfo;
808 p->link = u->link;
809 p->i_key = u->i_key;
810 p->o_key = u->o_key;
811 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
812 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
813 memcpy(p->name, u->name, sizeof(u->name));
814 }
815
816 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
817 const struct __ip6_tnl_parm *p)
818 {
819 u->proto = IPPROTO_GRE;
820 u->laddr = p->laddr;
821 u->raddr = p->raddr;
822 u->flags = p->flags;
823 u->hop_limit = p->hop_limit;
824 u->encap_limit = p->encap_limit;
825 u->flowinfo = p->flowinfo;
826 u->link = p->link;
827 u->i_key = p->i_key;
828 u->o_key = p->o_key;
829 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
830 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
831 memcpy(u->name, p->name, sizeof(u->name));
832 }
833
834 static int ip6gre_tunnel_ioctl(struct net_device *dev,
835 struct ifreq *ifr, int cmd)
836 {
837 int err = 0;
838 struct ip6_tnl_parm2 p;
839 struct __ip6_tnl_parm p1;
840 struct ip6_tnl *t = netdev_priv(dev);
841 struct net *net = t->net;
842 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
843
844 memset(&p1, 0, sizeof(p1));
845
846 switch (cmd) {
847 case SIOCGETTUNNEL:
848 if (dev == ign->fb_tunnel_dev) {
849 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
850 err = -EFAULT;
851 break;
852 }
853 ip6gre_tnl_parm_from_user(&p1, &p);
854 t = ip6gre_tunnel_locate(net, &p1, 0);
855 if (!t)
856 t = netdev_priv(dev);
857 }
858 memset(&p, 0, sizeof(p));
859 ip6gre_tnl_parm_to_user(&p, &t->parms);
860 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
861 err = -EFAULT;
862 break;
863
864 case SIOCADDTUNNEL:
865 case SIOCCHGTUNNEL:
866 err = -EPERM;
867 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
868 goto done;
869
870 err = -EFAULT;
871 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
872 goto done;
873
874 err = -EINVAL;
875 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
876 goto done;
877
878 if (!(p.i_flags&GRE_KEY))
879 p.i_key = 0;
880 if (!(p.o_flags&GRE_KEY))
881 p.o_key = 0;
882
883 ip6gre_tnl_parm_from_user(&p1, &p);
884 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
885
886 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
887 if (t) {
888 if (t->dev != dev) {
889 err = -EEXIST;
890 break;
891 }
892 } else {
893 t = netdev_priv(dev);
894
895 ip6gre_tunnel_unlink(ign, t);
896 synchronize_net();
897 ip6gre_tnl_change(t, &p1, 1);
898 ip6gre_tunnel_link(ign, t);
899 netdev_state_change(dev);
900 }
901 }
902
903 if (t) {
904 err = 0;
905
906 memset(&p, 0, sizeof(p));
907 ip6gre_tnl_parm_to_user(&p, &t->parms);
908 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
909 err = -EFAULT;
910 } else
911 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
912 break;
913
914 case SIOCDELTUNNEL:
915 err = -EPERM;
916 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
917 goto done;
918
919 if (dev == ign->fb_tunnel_dev) {
920 err = -EFAULT;
921 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
922 goto done;
923 err = -ENOENT;
924 ip6gre_tnl_parm_from_user(&p1, &p);
925 t = ip6gre_tunnel_locate(net, &p1, 0);
926 if (!t)
927 goto done;
928 err = -EPERM;
929 if (t == netdev_priv(ign->fb_tunnel_dev))
930 goto done;
931 dev = t->dev;
932 }
933 unregister_netdevice(dev);
934 err = 0;
935 break;
936
937 default:
938 err = -EINVAL;
939 }
940
941 done:
942 return err;
943 }
944
945 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
946 unsigned short type, const void *daddr,
947 const void *saddr, unsigned int len)
948 {
949 struct ip6_tnl *t = netdev_priv(dev);
950 struct ipv6hdr *ipv6h;
951 __be16 *p;
952
953 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
954 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
955 t->fl.u.ip6.flowlabel,
956 true, &t->fl.u.ip6));
957 ipv6h->hop_limit = t->parms.hop_limit;
958 ipv6h->nexthdr = NEXTHDR_GRE;
959 ipv6h->saddr = t->parms.laddr;
960 ipv6h->daddr = t->parms.raddr;
961
962 p = (__be16 *)(ipv6h + 1);
963 p[0] = t->parms.o_flags;
964 p[1] = htons(type);
965
966 /*
967 * Set the source hardware address.
968 */
969
970 if (saddr)
971 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
972 if (daddr)
973 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
974 if (!ipv6_addr_any(&ipv6h->daddr))
975 return t->hlen;
976
977 return -t->hlen;
978 }
979
980 static const struct header_ops ip6gre_header_ops = {
981 .create = ip6gre_header,
982 };
983
984 static const struct net_device_ops ip6gre_netdev_ops = {
985 .ndo_init = ip6gre_tunnel_init,
986 .ndo_uninit = ip6gre_tunnel_uninit,
987 .ndo_start_xmit = ip6gre_tunnel_xmit,
988 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
989 .ndo_change_mtu = ip6_tnl_change_mtu,
990 .ndo_get_stats64 = ip_tunnel_get_stats64,
991 .ndo_get_iflink = ip6_tnl_get_iflink,
992 };
993
994 static void ip6gre_dev_free(struct net_device *dev)
995 {
996 struct ip6_tnl *t = netdev_priv(dev);
997
998 dst_cache_destroy(&t->dst_cache);
999 free_percpu(dev->tstats);
1000 }
1001
1002 static void ip6gre_tunnel_setup(struct net_device *dev)
1003 {
1004 dev->netdev_ops = &ip6gre_netdev_ops;
1005 dev->needs_free_netdev = true;
1006 dev->priv_destructor = ip6gre_dev_free;
1007
1008 dev->type = ARPHRD_IP6GRE;
1009
1010 dev->flags |= IFF_NOARP;
1011 dev->addr_len = sizeof(struct in6_addr);
1012 netif_keep_dst(dev);
1013 /* This perm addr will be used as interface identifier by IPv6 */
1014 dev->addr_assign_type = NET_ADDR_RANDOM;
1015 eth_random_addr(dev->perm_addr);
1016 }
1017
1018 static int ip6gre_tunnel_init_common(struct net_device *dev)
1019 {
1020 struct ip6_tnl *tunnel;
1021 int ret;
1022 int t_hlen;
1023
1024 tunnel = netdev_priv(dev);
1025
1026 tunnel->dev = dev;
1027 tunnel->net = dev_net(dev);
1028 strcpy(tunnel->parms.name, dev->name);
1029
1030 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1031 if (!dev->tstats)
1032 return -ENOMEM;
1033
1034 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1035 if (ret) {
1036 free_percpu(dev->tstats);
1037 dev->tstats = NULL;
1038 return ret;
1039 }
1040
1041 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1042 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1043 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1044
1045 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1046 dev->mtu = ETH_DATA_LEN - t_hlen;
1047 if (dev->type == ARPHRD_ETHER)
1048 dev->mtu -= ETH_HLEN;
1049 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1050 dev->mtu -= 8;
1051
1052 return 0;
1053 }
1054
1055 static int ip6gre_tunnel_init(struct net_device *dev)
1056 {
1057 struct ip6_tnl *tunnel;
1058 int ret;
1059
1060 ret = ip6gre_tunnel_init_common(dev);
1061 if (ret)
1062 return ret;
1063
1064 tunnel = netdev_priv(dev);
1065
1066 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1067 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1068
1069 if (ipv6_addr_any(&tunnel->parms.raddr))
1070 dev->header_ops = &ip6gre_header_ops;
1071
1072 return 0;
1073 }
1074
1075 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1076 {
1077 struct ip6_tnl *tunnel = netdev_priv(dev);
1078
1079 tunnel->dev = dev;
1080 tunnel->net = dev_net(dev);
1081 strcpy(tunnel->parms.name, dev->name);
1082
1083 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1084
1085 dev_hold(dev);
1086 }
1087
1088
1089 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1090 .handler = gre_rcv,
1091 .err_handler = ip6gre_err,
1092 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1093 };
1094
1095 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1096 {
1097 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1098 struct net_device *dev, *aux;
1099 int prio;
1100
1101 for_each_netdev_safe(net, dev, aux)
1102 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1103 dev->rtnl_link_ops == &ip6gre_tap_ops)
1104 unregister_netdevice_queue(dev, head);
1105
1106 for (prio = 0; prio < 4; prio++) {
1107 int h;
1108 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1109 struct ip6_tnl *t;
1110
1111 t = rtnl_dereference(ign->tunnels[prio][h]);
1112
1113 while (t) {
1114 /* If dev is in the same netns, it has already
1115 * been added to the list by the previous loop.
1116 */
1117 if (!net_eq(dev_net(t->dev), net))
1118 unregister_netdevice_queue(t->dev,
1119 head);
1120 t = rtnl_dereference(t->next);
1121 }
1122 }
1123 }
1124 }
1125
1126 static int __net_init ip6gre_init_net(struct net *net)
1127 {
1128 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1129 int err;
1130
1131 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1132 NET_NAME_UNKNOWN,
1133 ip6gre_tunnel_setup);
1134 if (!ign->fb_tunnel_dev) {
1135 err = -ENOMEM;
1136 goto err_alloc_dev;
1137 }
1138 dev_net_set(ign->fb_tunnel_dev, net);
1139 /* FB netdevice is special: we have one, and only one per netns.
1140 * Allowing to move it to another netns is clearly unsafe.
1141 */
1142 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1143
1144
1145 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1146 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1147
1148 err = register_netdev(ign->fb_tunnel_dev);
1149 if (err)
1150 goto err_reg_dev;
1151
1152 rcu_assign_pointer(ign->tunnels_wc[0],
1153 netdev_priv(ign->fb_tunnel_dev));
1154 return 0;
1155
1156 err_reg_dev:
1157 free_netdev(ign->fb_tunnel_dev);
1158 err_alloc_dev:
1159 return err;
1160 }
1161
1162 static void __net_exit ip6gre_exit_net(struct net *net)
1163 {
1164 LIST_HEAD(list);
1165
1166 rtnl_lock();
1167 ip6gre_destroy_tunnels(net, &list);
1168 unregister_netdevice_many(&list);
1169 rtnl_unlock();
1170 }
1171
1172 static struct pernet_operations ip6gre_net_ops = {
1173 .init = ip6gre_init_net,
1174 .exit = ip6gre_exit_net,
1175 .id = &ip6gre_net_id,
1176 .size = sizeof(struct ip6gre_net),
1177 };
1178
1179 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1180 struct netlink_ext_ack *extack)
1181 {
1182 __be16 flags;
1183
1184 if (!data)
1185 return 0;
1186
1187 flags = 0;
1188 if (data[IFLA_GRE_IFLAGS])
1189 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1190 if (data[IFLA_GRE_OFLAGS])
1191 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1192 if (flags & (GRE_VERSION|GRE_ROUTING))
1193 return -EINVAL;
1194
1195 return 0;
1196 }
1197
1198 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1199 struct netlink_ext_ack *extack)
1200 {
1201 struct in6_addr daddr;
1202
1203 if (tb[IFLA_ADDRESS]) {
1204 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1205 return -EINVAL;
1206 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1207 return -EADDRNOTAVAIL;
1208 }
1209
1210 if (!data)
1211 goto out;
1212
1213 if (data[IFLA_GRE_REMOTE]) {
1214 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1215 if (ipv6_addr_any(&daddr))
1216 return -EINVAL;
1217 }
1218
1219 out:
1220 return ip6gre_tunnel_validate(tb, data, extack);
1221 }
1222
1223
1224 static void ip6gre_netlink_parms(struct nlattr *data[],
1225 struct __ip6_tnl_parm *parms)
1226 {
1227 memset(parms, 0, sizeof(*parms));
1228
1229 if (!data)
1230 return;
1231
1232 if (data[IFLA_GRE_LINK])
1233 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1234
1235 if (data[IFLA_GRE_IFLAGS])
1236 parms->i_flags = gre_flags_to_tnl_flags(
1237 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1238
1239 if (data[IFLA_GRE_OFLAGS])
1240 parms->o_flags = gre_flags_to_tnl_flags(
1241 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1242
1243 if (data[IFLA_GRE_IKEY])
1244 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1245
1246 if (data[IFLA_GRE_OKEY])
1247 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1248
1249 if (data[IFLA_GRE_LOCAL])
1250 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1251
1252 if (data[IFLA_GRE_REMOTE])
1253 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1254
1255 if (data[IFLA_GRE_TTL])
1256 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1257
1258 if (data[IFLA_GRE_ENCAP_LIMIT])
1259 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1260
1261 if (data[IFLA_GRE_FLOWINFO])
1262 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1263
1264 if (data[IFLA_GRE_FLAGS])
1265 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1266
1267 if (data[IFLA_GRE_FWMARK])
1268 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1269 }
1270
1271 static int ip6gre_tap_init(struct net_device *dev)
1272 {
1273 struct ip6_tnl *tunnel;
1274 int ret;
1275
1276 ret = ip6gre_tunnel_init_common(dev);
1277 if (ret)
1278 return ret;
1279
1280 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1281
1282 tunnel = netdev_priv(dev);
1283
1284 ip6gre_tnl_link_config(tunnel, 1);
1285
1286 return 0;
1287 }
1288
1289 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1290 .ndo_init = ip6gre_tap_init,
1291 .ndo_uninit = ip6gre_tunnel_uninit,
1292 .ndo_start_xmit = ip6gre_tunnel_xmit,
1293 .ndo_set_mac_address = eth_mac_addr,
1294 .ndo_validate_addr = eth_validate_addr,
1295 .ndo_change_mtu = ip6_tnl_change_mtu,
1296 .ndo_get_stats64 = ip_tunnel_get_stats64,
1297 .ndo_get_iflink = ip6_tnl_get_iflink,
1298 };
1299
1300 #define GRE6_FEATURES (NETIF_F_SG | \
1301 NETIF_F_FRAGLIST | \
1302 NETIF_F_HIGHDMA | \
1303 NETIF_F_HW_CSUM)
1304
1305 static void ip6gre_tap_setup(struct net_device *dev)
1306 {
1307
1308 ether_setup(dev);
1309
1310 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1311 dev->needs_free_netdev = true;
1312 dev->priv_destructor = ip6gre_dev_free;
1313
1314 dev->features |= NETIF_F_NETNS_LOCAL;
1315 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1316 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1317 netif_keep_dst(dev);
1318 }
1319
1320 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1321 struct ip_tunnel_encap *ipencap)
1322 {
1323 bool ret = false;
1324
1325 memset(ipencap, 0, sizeof(*ipencap));
1326
1327 if (!data)
1328 return ret;
1329
1330 if (data[IFLA_GRE_ENCAP_TYPE]) {
1331 ret = true;
1332 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1333 }
1334
1335 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1336 ret = true;
1337 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1338 }
1339
1340 if (data[IFLA_GRE_ENCAP_SPORT]) {
1341 ret = true;
1342 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1343 }
1344
1345 if (data[IFLA_GRE_ENCAP_DPORT]) {
1346 ret = true;
1347 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1348 }
1349
1350 return ret;
1351 }
1352
1353 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1354 struct nlattr *tb[], struct nlattr *data[],
1355 struct netlink_ext_ack *extack)
1356 {
1357 struct ip6_tnl *nt;
1358 struct net *net = dev_net(dev);
1359 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1360 struct ip_tunnel_encap ipencap;
1361 int err;
1362
1363 nt = netdev_priv(dev);
1364
1365 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1366 int err = ip6_tnl_encap_setup(nt, &ipencap);
1367
1368 if (err < 0)
1369 return err;
1370 }
1371
1372 ip6gre_netlink_parms(data, &nt->parms);
1373
1374 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1375 return -EEXIST;
1376
1377 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1378 eth_hw_addr_random(dev);
1379
1380 nt->dev = dev;
1381 nt->net = dev_net(dev);
1382 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1383
1384 dev->features |= GRE6_FEATURES;
1385 dev->hw_features |= GRE6_FEATURES;
1386
1387 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1388 /* TCP offload with GRE SEQ is not supported, nor
1389 * can we support 2 levels of outer headers requiring
1390 * an update.
1391 */
1392 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1393 (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1394 dev->features |= NETIF_F_GSO_SOFTWARE;
1395 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1396 }
1397
1398 /* Can use a lockless transmit, unless we generate
1399 * output sequences
1400 */
1401 dev->features |= NETIF_F_LLTX;
1402 }
1403
1404 err = register_netdevice(dev);
1405 if (err)
1406 goto out;
1407
1408 dev_hold(dev);
1409 ip6gre_tunnel_link(ign, nt);
1410
1411 out:
1412 return err;
1413 }
1414
1415 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1416 struct nlattr *data[],
1417 struct netlink_ext_ack *extack)
1418 {
1419 struct ip6_tnl *t, *nt = netdev_priv(dev);
1420 struct net *net = nt->net;
1421 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1422 struct __ip6_tnl_parm p;
1423 struct ip_tunnel_encap ipencap;
1424
1425 if (dev == ign->fb_tunnel_dev)
1426 return -EINVAL;
1427
1428 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1429 int err = ip6_tnl_encap_setup(nt, &ipencap);
1430
1431 if (err < 0)
1432 return err;
1433 }
1434
1435 ip6gre_netlink_parms(data, &p);
1436
1437 t = ip6gre_tunnel_locate(net, &p, 0);
1438
1439 if (t) {
1440 if (t->dev != dev)
1441 return -EEXIST;
1442 } else {
1443 t = nt;
1444 }
1445
1446 ip6gre_tunnel_unlink(ign, t);
1447 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1448 ip6gre_tunnel_link(ign, t);
1449 return 0;
1450 }
1451
1452 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1453 {
1454 struct net *net = dev_net(dev);
1455 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1456
1457 if (dev != ign->fb_tunnel_dev)
1458 unregister_netdevice_queue(dev, head);
1459 }
1460
1461 static size_t ip6gre_get_size(const struct net_device *dev)
1462 {
1463 return
1464 /* IFLA_GRE_LINK */
1465 nla_total_size(4) +
1466 /* IFLA_GRE_IFLAGS */
1467 nla_total_size(2) +
1468 /* IFLA_GRE_OFLAGS */
1469 nla_total_size(2) +
1470 /* IFLA_GRE_IKEY */
1471 nla_total_size(4) +
1472 /* IFLA_GRE_OKEY */
1473 nla_total_size(4) +
1474 /* IFLA_GRE_LOCAL */
1475 nla_total_size(sizeof(struct in6_addr)) +
1476 /* IFLA_GRE_REMOTE */
1477 nla_total_size(sizeof(struct in6_addr)) +
1478 /* IFLA_GRE_TTL */
1479 nla_total_size(1) +
1480 /* IFLA_GRE_ENCAP_LIMIT */
1481 nla_total_size(1) +
1482 /* IFLA_GRE_FLOWINFO */
1483 nla_total_size(4) +
1484 /* IFLA_GRE_FLAGS */
1485 nla_total_size(4) +
1486 /* IFLA_GRE_ENCAP_TYPE */
1487 nla_total_size(2) +
1488 /* IFLA_GRE_ENCAP_FLAGS */
1489 nla_total_size(2) +
1490 /* IFLA_GRE_ENCAP_SPORT */
1491 nla_total_size(2) +
1492 /* IFLA_GRE_ENCAP_DPORT */
1493 nla_total_size(2) +
1494 /* IFLA_GRE_FWMARK */
1495 nla_total_size(4) +
1496 0;
1497 }
1498
1499 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1500 {
1501 struct ip6_tnl *t = netdev_priv(dev);
1502 struct __ip6_tnl_parm *p = &t->parms;
1503
1504 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1505 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1506 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1507 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1508 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1509 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1510 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1511 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1512 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1513 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1514 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1515 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1516 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1517 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1518 goto nla_put_failure;
1519
1520 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1521 t->encap.type) ||
1522 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1523 t->encap.sport) ||
1524 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1525 t->encap.dport) ||
1526 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1527 t->encap.flags))
1528 goto nla_put_failure;
1529
1530 return 0;
1531
1532 nla_put_failure:
1533 return -EMSGSIZE;
1534 }
1535
1536 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1537 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1538 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1539 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1540 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1541 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1542 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1543 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1544 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1545 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1546 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1547 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1548 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1549 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1550 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1551 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1552 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1553 };
1554
1555 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1556 .kind = "ip6gre",
1557 .maxtype = IFLA_GRE_MAX,
1558 .policy = ip6gre_policy,
1559 .priv_size = sizeof(struct ip6_tnl),
1560 .setup = ip6gre_tunnel_setup,
1561 .validate = ip6gre_tunnel_validate,
1562 .newlink = ip6gre_newlink,
1563 .changelink = ip6gre_changelink,
1564 .dellink = ip6gre_dellink,
1565 .get_size = ip6gre_get_size,
1566 .fill_info = ip6gre_fill_info,
1567 .get_link_net = ip6_tnl_get_link_net,
1568 };
1569
1570 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1571 .kind = "ip6gretap",
1572 .maxtype = IFLA_GRE_MAX,
1573 .policy = ip6gre_policy,
1574 .priv_size = sizeof(struct ip6_tnl),
1575 .setup = ip6gre_tap_setup,
1576 .validate = ip6gre_tap_validate,
1577 .newlink = ip6gre_newlink,
1578 .changelink = ip6gre_changelink,
1579 .get_size = ip6gre_get_size,
1580 .fill_info = ip6gre_fill_info,
1581 .get_link_net = ip6_tnl_get_link_net,
1582 };
1583
1584 /*
1585 * And now the modules code and kernel interface.
1586 */
1587
1588 static int __init ip6gre_init(void)
1589 {
1590 int err;
1591
1592 pr_info("GRE over IPv6 tunneling driver\n");
1593
1594 err = register_pernet_device(&ip6gre_net_ops);
1595 if (err < 0)
1596 return err;
1597
1598 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1599 if (err < 0) {
1600 pr_info("%s: can't add protocol\n", __func__);
1601 goto add_proto_failed;
1602 }
1603
1604 err = rtnl_link_register(&ip6gre_link_ops);
1605 if (err < 0)
1606 goto rtnl_link_failed;
1607
1608 err = rtnl_link_register(&ip6gre_tap_ops);
1609 if (err < 0)
1610 goto tap_ops_failed;
1611
1612 out:
1613 return err;
1614
1615 tap_ops_failed:
1616 rtnl_link_unregister(&ip6gre_link_ops);
1617 rtnl_link_failed:
1618 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1619 add_proto_failed:
1620 unregister_pernet_device(&ip6gre_net_ops);
1621 goto out;
1622 }
1623
1624 static void __exit ip6gre_fini(void)
1625 {
1626 rtnl_link_unregister(&ip6gre_tap_ops);
1627 rtnl_link_unregister(&ip6gre_link_ops);
1628 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1629 unregister_pernet_device(&ip6gre_net_ops);
1630 }
1631
1632 module_init(ip6gre_init);
1633 module_exit(ip6gre_fini);
1634 MODULE_LICENSE("GPL");
1635 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1636 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1637 MODULE_ALIAS_RTNL_LINK("ip6gre");
1638 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1639 MODULE_ALIAS_NETDEV("ip6gre0");
Ubuntu Artful kernel mirror