dccp: fix use-after-free in dccp_feat_activate_values
[linux/fpc-iii.git] / net / ipv6 / ip6_gre.c
blob630b73be599977599c0021849fc6eb689cfefad7
1 /*
2 * GRE over IPv6 protocol decoder.
4 * Authors: Dmitry Kozlov (xeb@mail.ru)
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.
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
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>
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>
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");
64 #define IP6_GRE_HASH_SIZE_SHIFT 5
65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
67 static unsigned int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
71 struct net_device *fb_tunnel_dev;
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);
81 /* Tunnel hash table */
84 4 hash tables:
86 3: (remote,local)
87 2: (remote,*)
88 1: (*,local)
89 0: (*,*)
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.
95 All keysless packets, if not matched configured keyless tunnels
96 will match fallback tunnel.
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)
102 u32 hash = ipv6_addr_hash(addr);
104 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
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]
112 /* Given src, dst and key, find appropriate for input tunnel. */
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)
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;
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;
135 if (t->dev->type != ARPHRD_IP6GRE &&
136 t->dev->type != dev_type)
137 continue;
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;
147 if (score < cand_score) {
148 cand = t;
149 cand_score = score;
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;
159 if (t->dev->type != ARPHRD_IP6GRE &&
160 t->dev->type != dev_type)
161 continue;
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;
171 if (score < cand_score) {
172 cand = t;
173 cand_score = score;
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;
185 if (t->dev->type != ARPHRD_IP6GRE &&
186 t->dev->type != dev_type)
187 continue;
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;
197 if (score < cand_score) {
198 cand = t;
199 cand_score = score;
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;
208 if (t->dev->type != ARPHRD_IP6GRE &&
209 t->dev->type != dev_type)
210 continue;
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;
220 if (score < cand_score) {
221 cand = t;
222 cand_score = score;
226 if (cand)
227 return cand;
229 dev = ign->fb_tunnel_dev;
230 if (dev->flags & IFF_UP)
231 return netdev_priv(dev);
233 return NULL;
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237 const struct __ip6_tnl_parm *p)
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;
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);
251 return &ign->tunnels[prio][h];
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255 const struct ip6_tnl *t)
257 return __ip6gre_bucket(ign, &t->parms);
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
262 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
264 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265 rcu_assign_pointer(*tp, t);
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
270 struct ip6_tnl __rcu **tp;
271 struct ip6_tnl *iter;
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;
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 const struct __ip6_tnl_parm *parms,
285 int type)
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);
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;
305 return t;
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309 const struct __ip6_tnl_parm *parms, int create)
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);
316 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317 if (t && create)
318 return NULL;
319 if (t || !create)
320 return t;
322 if (parms->name[0])
323 strlcpy(name, parms->name, IFNAMSIZ);
324 else
325 strcpy(name, "ip6gre%d");
327 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328 ip6gre_tunnel_setup);
329 if (!dev)
330 return NULL;
332 dev_net_set(dev, net);
334 nt = netdev_priv(dev);
335 nt->parms = *parms;
336 dev->rtnl_link_ops = &ip6gre_link_ops;
338 nt->dev = dev;
339 nt->net = dev_net(dev);
340 ip6gre_tnl_link_config(nt, 1);
342 if (register_netdevice(dev) < 0)
343 goto failed_free;
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;
349 dev_hold(dev);
350 ip6gre_tunnel_link(ign, nt);
351 return nt;
353 failed_free:
354 free_netdev(dev);
355 return NULL;
358 static void ip6gre_tunnel_uninit(struct net_device *dev)
360 struct ip6_tnl *t = netdev_priv(dev);
361 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
363 ip6gre_tunnel_unlink(ign, t);
364 dst_cache_reset(&t->dst_cache);
365 dev_put(dev);
369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
370 u8 type, u8 code, int offset, __be32 info)
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;
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;
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;
399 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
400 key, greh->protocol);
401 if (!t)
402 return;
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 break;
412 case ICMPV6_TIME_EXCEED:
413 if (code == ICMPV6_EXC_HOPLIMIT) {
414 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
415 t->parms.name);
417 break;
418 case ICMPV6_PARAMPROB:
419 teli = 0;
420 if (code == ICMPV6_HDR_FIELD)
421 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
423 if (teli && teli == be32_to_cpu(info) - 2) {
424 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
425 if (tel->encap_limit == 0) {
426 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
427 t->parms.name);
429 } else {
430 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
431 t->parms.name);
433 break;
434 case ICMPV6_PKT_TOOBIG:
435 mtu = be32_to_cpu(info) - offset;
436 if (mtu < IPV6_MIN_MTU)
437 mtu = IPV6_MIN_MTU;
438 t->dev->mtu = mtu;
439 break;
442 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
443 t->err_count++;
444 else
445 t->err_count = 1;
446 t->err_time = jiffies;
449 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
451 const struct ipv6hdr *ipv6h;
452 struct ip6_tnl *tunnel;
454 ipv6h = ipv6_hdr(skb);
455 tunnel = ip6gre_tunnel_lookup(skb->dev,
456 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
457 tpi->proto);
458 if (tunnel) {
459 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
461 return PACKET_RCVD;
464 return PACKET_REJECT;
467 static int gre_rcv(struct sk_buff *skb)
469 struct tnl_ptk_info tpi;
470 bool csum_err = false;
471 int hdr_len;
473 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
474 if (hdr_len < 0)
475 goto drop;
477 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
478 goto drop;
480 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
481 return 0;
483 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
484 drop:
485 kfree_skb(skb);
486 return 0;
489 struct ipv6_tel_txoption {
490 struct ipv6_txoptions ops;
491 __u8 dst_opt[8];
494 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
496 return iptunnel_handle_offloads(skb,
497 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
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)
505 struct ip6_tnl *tunnel = netdev_priv(dev);
506 __be16 protocol = (dev->type == ARPHRD_ETHER) ?
507 htons(ETH_P_TEB) : proto;
509 if (dev->type == ARPHRD_ETHER)
510 IPCB(skb)->flags = 0;
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;
517 if (tunnel->parms.o_flags & TUNNEL_SEQ)
518 tunnel->o_seqno++;
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));
524 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
525 NEXTHDR_GRE);
528 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
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;
538 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
540 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
541 encap_limit = t->parms.encap_limit;
543 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
545 dsfield = ipv4_get_dsfield(iph);
547 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
548 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
549 & IPV6_TCLASS_MASK;
550 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
551 fl6.flowi6_mark = skb->mark;
553 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
555 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
556 if (err)
557 return -1;
559 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
560 skb->protocol);
561 if (err != 0) {
562 /* XXX: send ICMP error even if DF is not set. */
563 if (err == -EMSGSIZE)
564 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
565 htonl(mtu));
566 return -1;
569 return 0;
572 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
574 struct ip6_tnl *t = netdev_priv(dev);
575 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
576 int encap_limit = -1;
577 __u16 offset;
578 struct flowi6 fl6;
579 __u8 dsfield;
580 __u32 mtu;
581 int err;
583 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
584 return -1;
586 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
587 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
588 ipv6h = ipv6_hdr(skb);
590 if (offset > 0) {
591 struct ipv6_tlv_tnl_enc_lim *tel;
592 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
593 if (tel->encap_limit == 0) {
594 icmpv6_send(skb, ICMPV6_PARAMPROB,
595 ICMPV6_HDR_FIELD, offset + 2);
596 return -1;
598 encap_limit = tel->encap_limit - 1;
599 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
600 encap_limit = t->parms.encap_limit;
602 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
604 dsfield = ipv6_get_dsfield(ipv6h);
605 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
606 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
607 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
608 fl6.flowlabel |= ip6_flowlabel(ipv6h);
609 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
610 fl6.flowi6_mark = skb->mark;
612 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
614 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
615 return -1;
617 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
618 &mtu, skb->protocol);
619 if (err != 0) {
620 if (err == -EMSGSIZE)
621 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
622 return -1;
625 return 0;
629 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
630 * @t: the outgoing tunnel device
631 * @hdr: IPv6 header from the incoming packet
633 * Description:
634 * Avoid trivial tunneling loop by checking that tunnel exit-point
635 * doesn't match source of incoming packet.
637 * Return:
638 * 1 if conflict,
639 * 0 else
642 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
643 const struct ipv6hdr *hdr)
645 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
648 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
650 struct ip6_tnl *t = netdev_priv(dev);
651 int encap_limit = -1;
652 struct flowi6 fl6;
653 __u32 mtu;
654 int err;
656 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
657 encap_limit = t->parms.encap_limit;
659 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
661 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
662 if (err)
663 return err;
665 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
667 return err;
670 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
671 struct net_device *dev)
673 struct ip6_tnl *t = netdev_priv(dev);
674 struct net_device_stats *stats = &t->dev->stats;
675 int ret;
677 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
678 goto tx_err;
680 switch (skb->protocol) {
681 case htons(ETH_P_IP):
682 ret = ip6gre_xmit_ipv4(skb, dev);
683 break;
684 case htons(ETH_P_IPV6):
685 ret = ip6gre_xmit_ipv6(skb, dev);
686 break;
687 default:
688 ret = ip6gre_xmit_other(skb, dev);
689 break;
692 if (ret < 0)
693 goto tx_err;
695 return NETDEV_TX_OK;
697 tx_err:
698 stats->tx_errors++;
699 stats->tx_dropped++;
700 kfree_skb(skb);
701 return NETDEV_TX_OK;
704 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
706 struct net_device *dev = t->dev;
707 struct __ip6_tnl_parm *p = &t->parms;
708 struct flowi6 *fl6 = &t->fl.u.ip6;
709 int t_hlen;
711 if (dev->type != ARPHRD_ETHER) {
712 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
713 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
716 /* Set up flowi template */
717 fl6->saddr = p->laddr;
718 fl6->daddr = p->raddr;
719 fl6->flowi6_oif = p->link;
720 fl6->flowlabel = 0;
721 fl6->flowi6_proto = IPPROTO_GRE;
723 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
724 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
725 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
726 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
728 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
729 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
731 if (p->flags&IP6_TNL_F_CAP_XMIT &&
732 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
733 dev->flags |= IFF_POINTOPOINT;
734 else
735 dev->flags &= ~IFF_POINTOPOINT;
737 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
739 t->hlen = t->encap_hlen + t->tun_hlen;
741 t_hlen = t->hlen + sizeof(struct ipv6hdr);
743 if (p->flags & IP6_TNL_F_CAP_XMIT) {
744 int strict = (ipv6_addr_type(&p->raddr) &
745 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
747 struct rt6_info *rt = rt6_lookup(t->net,
748 &p->raddr, &p->laddr,
749 p->link, strict);
751 if (!rt)
752 return;
754 if (rt->dst.dev) {
755 dev->hard_header_len = rt->dst.dev->hard_header_len +
756 t_hlen;
758 if (set_mtu) {
759 dev->mtu = rt->dst.dev->mtu - t_hlen;
760 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
761 dev->mtu -= 8;
762 if (dev->type == ARPHRD_ETHER)
763 dev->mtu -= ETH_HLEN;
765 if (dev->mtu < IPV6_MIN_MTU)
766 dev->mtu = IPV6_MIN_MTU;
769 ip6_rt_put(rt);
773 static int ip6gre_tnl_change(struct ip6_tnl *t,
774 const struct __ip6_tnl_parm *p, int set_mtu)
776 t->parms.laddr = p->laddr;
777 t->parms.raddr = p->raddr;
778 t->parms.flags = p->flags;
779 t->parms.hop_limit = p->hop_limit;
780 t->parms.encap_limit = p->encap_limit;
781 t->parms.flowinfo = p->flowinfo;
782 t->parms.link = p->link;
783 t->parms.proto = p->proto;
784 t->parms.i_key = p->i_key;
785 t->parms.o_key = p->o_key;
786 t->parms.i_flags = p->i_flags;
787 t->parms.o_flags = p->o_flags;
788 dst_cache_reset(&t->dst_cache);
789 ip6gre_tnl_link_config(t, set_mtu);
790 return 0;
793 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
794 const struct ip6_tnl_parm2 *u)
796 p->laddr = u->laddr;
797 p->raddr = u->raddr;
798 p->flags = u->flags;
799 p->hop_limit = u->hop_limit;
800 p->encap_limit = u->encap_limit;
801 p->flowinfo = u->flowinfo;
802 p->link = u->link;
803 p->i_key = u->i_key;
804 p->o_key = u->o_key;
805 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
806 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
807 memcpy(p->name, u->name, sizeof(u->name));
810 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
811 const struct __ip6_tnl_parm *p)
813 u->proto = IPPROTO_GRE;
814 u->laddr = p->laddr;
815 u->raddr = p->raddr;
816 u->flags = p->flags;
817 u->hop_limit = p->hop_limit;
818 u->encap_limit = p->encap_limit;
819 u->flowinfo = p->flowinfo;
820 u->link = p->link;
821 u->i_key = p->i_key;
822 u->o_key = p->o_key;
823 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
824 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
825 memcpy(u->name, p->name, sizeof(u->name));
828 static int ip6gre_tunnel_ioctl(struct net_device *dev,
829 struct ifreq *ifr, int cmd)
831 int err = 0;
832 struct ip6_tnl_parm2 p;
833 struct __ip6_tnl_parm p1;
834 struct ip6_tnl *t = netdev_priv(dev);
835 struct net *net = t->net;
836 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
838 memset(&p1, 0, sizeof(p1));
840 switch (cmd) {
841 case SIOCGETTUNNEL:
842 if (dev == ign->fb_tunnel_dev) {
843 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
844 err = -EFAULT;
845 break;
847 ip6gre_tnl_parm_from_user(&p1, &p);
848 t = ip6gre_tunnel_locate(net, &p1, 0);
849 if (!t)
850 t = netdev_priv(dev);
852 memset(&p, 0, sizeof(p));
853 ip6gre_tnl_parm_to_user(&p, &t->parms);
854 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
855 err = -EFAULT;
856 break;
858 case SIOCADDTUNNEL:
859 case SIOCCHGTUNNEL:
860 err = -EPERM;
861 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
862 goto done;
864 err = -EFAULT;
865 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
866 goto done;
868 err = -EINVAL;
869 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
870 goto done;
872 if (!(p.i_flags&GRE_KEY))
873 p.i_key = 0;
874 if (!(p.o_flags&GRE_KEY))
875 p.o_key = 0;
877 ip6gre_tnl_parm_from_user(&p1, &p);
878 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
880 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
881 if (t) {
882 if (t->dev != dev) {
883 err = -EEXIST;
884 break;
886 } else {
887 t = netdev_priv(dev);
889 ip6gre_tunnel_unlink(ign, t);
890 synchronize_net();
891 ip6gre_tnl_change(t, &p1, 1);
892 ip6gre_tunnel_link(ign, t);
893 netdev_state_change(dev);
897 if (t) {
898 err = 0;
900 memset(&p, 0, sizeof(p));
901 ip6gre_tnl_parm_to_user(&p, &t->parms);
902 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
903 err = -EFAULT;
904 } else
905 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
906 break;
908 case SIOCDELTUNNEL:
909 err = -EPERM;
910 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
911 goto done;
913 if (dev == ign->fb_tunnel_dev) {
914 err = -EFAULT;
915 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
916 goto done;
917 err = -ENOENT;
918 ip6gre_tnl_parm_from_user(&p1, &p);
919 t = ip6gre_tunnel_locate(net, &p1, 0);
920 if (!t)
921 goto done;
922 err = -EPERM;
923 if (t == netdev_priv(ign->fb_tunnel_dev))
924 goto done;
925 dev = t->dev;
927 unregister_netdevice(dev);
928 err = 0;
929 break;
931 default:
932 err = -EINVAL;
935 done:
936 return err;
939 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
940 unsigned short type,
941 const void *daddr, const void *saddr, unsigned int len)
943 struct ip6_tnl *t = netdev_priv(dev);
944 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
945 __be16 *p = (__be16 *)(ipv6h+1);
947 ip6_flow_hdr(ipv6h, 0,
948 ip6_make_flowlabel(dev_net(dev), skb,
949 t->fl.u.ip6.flowlabel, true,
950 &t->fl.u.ip6));
951 ipv6h->hop_limit = t->parms.hop_limit;
952 ipv6h->nexthdr = NEXTHDR_GRE;
953 ipv6h->saddr = t->parms.laddr;
954 ipv6h->daddr = t->parms.raddr;
956 p[0] = t->parms.o_flags;
957 p[1] = htons(type);
960 * Set the source hardware address.
963 if (saddr)
964 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
965 if (daddr)
966 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
967 if (!ipv6_addr_any(&ipv6h->daddr))
968 return t->hlen;
970 return -t->hlen;
973 static const struct header_ops ip6gre_header_ops = {
974 .create = ip6gre_header,
977 static const struct net_device_ops ip6gre_netdev_ops = {
978 .ndo_init = ip6gre_tunnel_init,
979 .ndo_uninit = ip6gre_tunnel_uninit,
980 .ndo_start_xmit = ip6gre_tunnel_xmit,
981 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
982 .ndo_change_mtu = ip6_tnl_change_mtu,
983 .ndo_get_stats64 = ip_tunnel_get_stats64,
984 .ndo_get_iflink = ip6_tnl_get_iflink,
987 static void ip6gre_dev_free(struct net_device *dev)
989 struct ip6_tnl *t = netdev_priv(dev);
991 dst_cache_destroy(&t->dst_cache);
992 free_percpu(dev->tstats);
993 free_netdev(dev);
996 static void ip6gre_tunnel_setup(struct net_device *dev)
998 dev->netdev_ops = &ip6gre_netdev_ops;
999 dev->destructor = ip6gre_dev_free;
1001 dev->type = ARPHRD_IP6GRE;
1003 dev->flags |= IFF_NOARP;
1004 dev->addr_len = sizeof(struct in6_addr);
1005 netif_keep_dst(dev);
1008 static int ip6gre_tunnel_init_common(struct net_device *dev)
1010 struct ip6_tnl *tunnel;
1011 int ret;
1012 int t_hlen;
1014 tunnel = netdev_priv(dev);
1016 tunnel->dev = dev;
1017 tunnel->net = dev_net(dev);
1018 strcpy(tunnel->parms.name, dev->name);
1020 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1021 if (!dev->tstats)
1022 return -ENOMEM;
1024 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1025 if (ret) {
1026 free_percpu(dev->tstats);
1027 dev->tstats = NULL;
1028 return ret;
1031 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1032 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1033 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1035 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1036 dev->mtu = ETH_DATA_LEN - t_hlen;
1037 if (dev->type == ARPHRD_ETHER)
1038 dev->mtu -= ETH_HLEN;
1039 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1040 dev->mtu -= 8;
1042 return 0;
1045 static int ip6gre_tunnel_init(struct net_device *dev)
1047 struct ip6_tnl *tunnel;
1048 int ret;
1050 ret = ip6gre_tunnel_init_common(dev);
1051 if (ret)
1052 return ret;
1054 tunnel = netdev_priv(dev);
1056 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1057 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1059 if (ipv6_addr_any(&tunnel->parms.raddr))
1060 dev->header_ops = &ip6gre_header_ops;
1062 return 0;
1065 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1067 struct ip6_tnl *tunnel = netdev_priv(dev);
1069 tunnel->dev = dev;
1070 tunnel->net = dev_net(dev);
1071 strcpy(tunnel->parms.name, dev->name);
1073 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1075 dev_hold(dev);
1079 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1080 .handler = gre_rcv,
1081 .err_handler = ip6gre_err,
1082 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1085 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1087 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1088 struct net_device *dev, *aux;
1089 int prio;
1091 for_each_netdev_safe(net, dev, aux)
1092 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1093 dev->rtnl_link_ops == &ip6gre_tap_ops)
1094 unregister_netdevice_queue(dev, head);
1096 for (prio = 0; prio < 4; prio++) {
1097 int h;
1098 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1099 struct ip6_tnl *t;
1101 t = rtnl_dereference(ign->tunnels[prio][h]);
1103 while (t) {
1104 /* If dev is in the same netns, it has already
1105 * been added to the list by the previous loop.
1107 if (!net_eq(dev_net(t->dev), net))
1108 unregister_netdevice_queue(t->dev,
1109 head);
1110 t = rtnl_dereference(t->next);
1116 static int __net_init ip6gre_init_net(struct net *net)
1118 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1119 int err;
1121 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1122 NET_NAME_UNKNOWN,
1123 ip6gre_tunnel_setup);
1124 if (!ign->fb_tunnel_dev) {
1125 err = -ENOMEM;
1126 goto err_alloc_dev;
1128 dev_net_set(ign->fb_tunnel_dev, net);
1129 /* FB netdevice is special: we have one, and only one per netns.
1130 * Allowing to move it to another netns is clearly unsafe.
1132 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1135 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1136 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1138 err = register_netdev(ign->fb_tunnel_dev);
1139 if (err)
1140 goto err_reg_dev;
1142 rcu_assign_pointer(ign->tunnels_wc[0],
1143 netdev_priv(ign->fb_tunnel_dev));
1144 return 0;
1146 err_reg_dev:
1147 ip6gre_dev_free(ign->fb_tunnel_dev);
1148 err_alloc_dev:
1149 return err;
1152 static void __net_exit ip6gre_exit_net(struct net *net)
1154 LIST_HEAD(list);
1156 rtnl_lock();
1157 ip6gre_destroy_tunnels(net, &list);
1158 unregister_netdevice_many(&list);
1159 rtnl_unlock();
1162 static struct pernet_operations ip6gre_net_ops = {
1163 .init = ip6gre_init_net,
1164 .exit = ip6gre_exit_net,
1165 .id = &ip6gre_net_id,
1166 .size = sizeof(struct ip6gre_net),
1169 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1171 __be16 flags;
1173 if (!data)
1174 return 0;
1176 flags = 0;
1177 if (data[IFLA_GRE_IFLAGS])
1178 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1179 if (data[IFLA_GRE_OFLAGS])
1180 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1181 if (flags & (GRE_VERSION|GRE_ROUTING))
1182 return -EINVAL;
1184 return 0;
1187 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1189 struct in6_addr daddr;
1191 if (tb[IFLA_ADDRESS]) {
1192 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1193 return -EINVAL;
1194 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1195 return -EADDRNOTAVAIL;
1198 if (!data)
1199 goto out;
1201 if (data[IFLA_GRE_REMOTE]) {
1202 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1203 if (ipv6_addr_any(&daddr))
1204 return -EINVAL;
1207 out:
1208 return ip6gre_tunnel_validate(tb, data);
1212 static void ip6gre_netlink_parms(struct nlattr *data[],
1213 struct __ip6_tnl_parm *parms)
1215 memset(parms, 0, sizeof(*parms));
1217 if (!data)
1218 return;
1220 if (data[IFLA_GRE_LINK])
1221 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1223 if (data[IFLA_GRE_IFLAGS])
1224 parms->i_flags = gre_flags_to_tnl_flags(
1225 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1227 if (data[IFLA_GRE_OFLAGS])
1228 parms->o_flags = gre_flags_to_tnl_flags(
1229 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1231 if (data[IFLA_GRE_IKEY])
1232 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1234 if (data[IFLA_GRE_OKEY])
1235 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1237 if (data[IFLA_GRE_LOCAL])
1238 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1240 if (data[IFLA_GRE_REMOTE])
1241 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1243 if (data[IFLA_GRE_TTL])
1244 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1246 if (data[IFLA_GRE_ENCAP_LIMIT])
1247 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1249 if (data[IFLA_GRE_FLOWINFO])
1250 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1252 if (data[IFLA_GRE_FLAGS])
1253 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1256 static int ip6gre_tap_init(struct net_device *dev)
1258 struct ip6_tnl *tunnel;
1259 int ret;
1261 ret = ip6gre_tunnel_init_common(dev);
1262 if (ret)
1263 return ret;
1265 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1267 tunnel = netdev_priv(dev);
1269 ip6gre_tnl_link_config(tunnel, 1);
1271 return 0;
1274 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1275 .ndo_init = ip6gre_tap_init,
1276 .ndo_uninit = ip6gre_tunnel_uninit,
1277 .ndo_start_xmit = ip6gre_tunnel_xmit,
1278 .ndo_set_mac_address = eth_mac_addr,
1279 .ndo_validate_addr = eth_validate_addr,
1280 .ndo_change_mtu = ip6_tnl_change_mtu,
1281 .ndo_get_stats64 = ip_tunnel_get_stats64,
1282 .ndo_get_iflink = ip6_tnl_get_iflink,
1285 #define GRE6_FEATURES (NETIF_F_SG | \
1286 NETIF_F_FRAGLIST | \
1287 NETIF_F_HIGHDMA | \
1288 NETIF_F_HW_CSUM)
1290 static void ip6gre_tap_setup(struct net_device *dev)
1293 ether_setup(dev);
1295 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1296 dev->destructor = ip6gre_dev_free;
1298 dev->features |= NETIF_F_NETNS_LOCAL;
1299 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1300 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1303 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1304 struct ip_tunnel_encap *ipencap)
1306 bool ret = false;
1308 memset(ipencap, 0, sizeof(*ipencap));
1310 if (!data)
1311 return ret;
1313 if (data[IFLA_GRE_ENCAP_TYPE]) {
1314 ret = true;
1315 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1318 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1319 ret = true;
1320 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1323 if (data[IFLA_GRE_ENCAP_SPORT]) {
1324 ret = true;
1325 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1328 if (data[IFLA_GRE_ENCAP_DPORT]) {
1329 ret = true;
1330 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1333 return ret;
1336 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1337 struct nlattr *tb[], struct nlattr *data[])
1339 struct ip6_tnl *nt;
1340 struct net *net = dev_net(dev);
1341 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1342 struct ip_tunnel_encap ipencap;
1343 int err;
1345 nt = netdev_priv(dev);
1347 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1348 int err = ip6_tnl_encap_setup(nt, &ipencap);
1350 if (err < 0)
1351 return err;
1354 ip6gre_netlink_parms(data, &nt->parms);
1356 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1357 return -EEXIST;
1359 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1360 eth_hw_addr_random(dev);
1362 nt->dev = dev;
1363 nt->net = dev_net(dev);
1364 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1366 dev->features |= GRE6_FEATURES;
1367 dev->hw_features |= GRE6_FEATURES;
1369 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1370 /* TCP offload with GRE SEQ is not supported, nor
1371 * can we support 2 levels of outer headers requiring
1372 * an update.
1374 if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1375 (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1376 dev->features |= NETIF_F_GSO_SOFTWARE;
1377 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1380 /* Can use a lockless transmit, unless we generate
1381 * output sequences
1383 dev->features |= NETIF_F_LLTX;
1386 err = register_netdevice(dev);
1387 if (err)
1388 goto out;
1390 dev_hold(dev);
1391 ip6gre_tunnel_link(ign, nt);
1393 out:
1394 return err;
1397 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1398 struct nlattr *data[])
1400 struct ip6_tnl *t, *nt = netdev_priv(dev);
1401 struct net *net = nt->net;
1402 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1403 struct __ip6_tnl_parm p;
1404 struct ip_tunnel_encap ipencap;
1406 if (dev == ign->fb_tunnel_dev)
1407 return -EINVAL;
1409 if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1410 int err = ip6_tnl_encap_setup(nt, &ipencap);
1412 if (err < 0)
1413 return err;
1416 ip6gre_netlink_parms(data, &p);
1418 t = ip6gre_tunnel_locate(net, &p, 0);
1420 if (t) {
1421 if (t->dev != dev)
1422 return -EEXIST;
1423 } else {
1424 t = nt;
1427 ip6gre_tunnel_unlink(ign, t);
1428 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1429 ip6gre_tunnel_link(ign, t);
1430 return 0;
1433 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1435 struct net *net = dev_net(dev);
1436 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1438 if (dev != ign->fb_tunnel_dev)
1439 unregister_netdevice_queue(dev, head);
1442 static size_t ip6gre_get_size(const struct net_device *dev)
1444 return
1445 /* IFLA_GRE_LINK */
1446 nla_total_size(4) +
1447 /* IFLA_GRE_IFLAGS */
1448 nla_total_size(2) +
1449 /* IFLA_GRE_OFLAGS */
1450 nla_total_size(2) +
1451 /* IFLA_GRE_IKEY */
1452 nla_total_size(4) +
1453 /* IFLA_GRE_OKEY */
1454 nla_total_size(4) +
1455 /* IFLA_GRE_LOCAL */
1456 nla_total_size(sizeof(struct in6_addr)) +
1457 /* IFLA_GRE_REMOTE */
1458 nla_total_size(sizeof(struct in6_addr)) +
1459 /* IFLA_GRE_TTL */
1460 nla_total_size(1) +
1461 /* IFLA_GRE_ENCAP_LIMIT */
1462 nla_total_size(1) +
1463 /* IFLA_GRE_FLOWINFO */
1464 nla_total_size(4) +
1465 /* IFLA_GRE_FLAGS */
1466 nla_total_size(4) +
1467 /* IFLA_GRE_ENCAP_TYPE */
1468 nla_total_size(2) +
1469 /* IFLA_GRE_ENCAP_FLAGS */
1470 nla_total_size(2) +
1471 /* IFLA_GRE_ENCAP_SPORT */
1472 nla_total_size(2) +
1473 /* IFLA_GRE_ENCAP_DPORT */
1474 nla_total_size(2) +
1478 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1480 struct ip6_tnl *t = netdev_priv(dev);
1481 struct __ip6_tnl_parm *p = &t->parms;
1483 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1484 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1485 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1486 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1487 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1488 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1489 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1490 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1491 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1492 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1493 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1494 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1495 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1496 goto nla_put_failure;
1498 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1499 t->encap.type) ||
1500 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1501 t->encap.sport) ||
1502 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1503 t->encap.dport) ||
1504 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1505 t->encap.flags))
1506 goto nla_put_failure;
1508 return 0;
1510 nla_put_failure:
1511 return -EMSGSIZE;
1514 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1515 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1516 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1517 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1518 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1519 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1520 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1521 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1522 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1523 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1524 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1525 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1526 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1527 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1528 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1529 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1532 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1533 .kind = "ip6gre",
1534 .maxtype = IFLA_GRE_MAX,
1535 .policy = ip6gre_policy,
1536 .priv_size = sizeof(struct ip6_tnl),
1537 .setup = ip6gre_tunnel_setup,
1538 .validate = ip6gre_tunnel_validate,
1539 .newlink = ip6gre_newlink,
1540 .changelink = ip6gre_changelink,
1541 .dellink = ip6gre_dellink,
1542 .get_size = ip6gre_get_size,
1543 .fill_info = ip6gre_fill_info,
1544 .get_link_net = ip6_tnl_get_link_net,
1547 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1548 .kind = "ip6gretap",
1549 .maxtype = IFLA_GRE_MAX,
1550 .policy = ip6gre_policy,
1551 .priv_size = sizeof(struct ip6_tnl),
1552 .setup = ip6gre_tap_setup,
1553 .validate = ip6gre_tap_validate,
1554 .newlink = ip6gre_newlink,
1555 .changelink = ip6gre_changelink,
1556 .get_size = ip6gre_get_size,
1557 .fill_info = ip6gre_fill_info,
1558 .get_link_net = ip6_tnl_get_link_net,
1562 * And now the modules code and kernel interface.
1565 static int __init ip6gre_init(void)
1567 int err;
1569 pr_info("GRE over IPv6 tunneling driver\n");
1571 err = register_pernet_device(&ip6gre_net_ops);
1572 if (err < 0)
1573 return err;
1575 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1576 if (err < 0) {
1577 pr_info("%s: can't add protocol\n", __func__);
1578 goto add_proto_failed;
1581 err = rtnl_link_register(&ip6gre_link_ops);
1582 if (err < 0)
1583 goto rtnl_link_failed;
1585 err = rtnl_link_register(&ip6gre_tap_ops);
1586 if (err < 0)
1587 goto tap_ops_failed;
1589 out:
1590 return err;
1592 tap_ops_failed:
1593 rtnl_link_unregister(&ip6gre_link_ops);
1594 rtnl_link_failed:
1595 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1596 add_proto_failed:
1597 unregister_pernet_device(&ip6gre_net_ops);
1598 goto out;
1601 static void __exit ip6gre_fini(void)
1603 rtnl_link_unregister(&ip6gre_tap_ops);
1604 rtnl_link_unregister(&ip6gre_link_ops);
1605 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1606 unregister_pernet_device(&ip6gre_net_ops);
1609 module_init(ip6gre_init);
1610 module_exit(ip6gre_fini);
1611 MODULE_LICENSE("GPL");
1612 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1613 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1614 MODULE_ALIAS_RTNL_LINK("ip6gre");
1615 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1616 MODULE_ALIAS_NETDEV("ip6gre0");