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>
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>
40 #include <net/ip_tunnels.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.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 HASH_SIZE_SHIFT 5
65 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
67 static int ip6gre_net_id __read_mostly;
69 struct ip6_tnl __rcu *tunnels[4][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 */
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))&(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, 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))
135 if (t->dev->type != ARPHRD_IP6GRE &&
136 t->dev->type != dev_type)
140 if (t->parms.link != link)
142 if (t->dev->type != dev_type)
147 if (score < cand_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))
159 if (t->dev->type != ARPHRD_IP6GRE &&
160 t->dev->type != dev_type)
164 if (t->parms.link != link)
166 if (t->dev->type != dev_type)
171 if (score < cand_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))
185 if (t->dev->type != ARPHRD_IP6GRE &&
186 t->dev->type != dev_type)
190 if (t->parms.link != link)
192 if (t->dev->type != dev_type)
197 if (score < cand_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))
208 if (t->dev->type != ARPHRD_IP6GRE &&
209 t->dev->type != dev_type)
213 if (t->parms.link != link)
215 if (t->dev->type != dev_type)
220 if (score < cand_score) {
229 dev = ign->fb_tunnel_dev;
230 if (dev->flags & IFF_UP)
231 return netdev_priv(dev);
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);
244 if (!ipv6_addr_any(local))
246 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
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;
277 rcu_assign_pointer(*tp, t->next);
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 const struct __ip6_tnl_parm *parms,
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;
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;
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)
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;
314 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
316 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
323 strlcpy(name, parms->name, IFNAMSIZ);
325 strcpy(name, "ip6gre%d");
327 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328 ip6gre_tunnel_setup);
332 dev_net_set(dev, net);
334 nt = netdev_priv(dev);
336 dev->rtnl_link_ops = &ip6gre_link_ops;
339 nt->net = dev_net(dev);
340 ip6gre_tnl_link_config(nt, 1);
342 if (register_netdevice(dev) < 0)
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;
350 ip6gre_tunnel_link(ign, nt);
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);
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 ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
373 __be16 *p = (__be16 *)(skb->data + offset);
374 int grehlen = offset + 4;
379 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
380 if (flags&(GRE_VERSION|GRE_ROUTING))
389 /* If only 8 bytes returned, keyed message will be dropped here */
390 if (!pskb_may_pull(skb, grehlen))
392 ipv6h = (const struct ipv6hdr *)skb->data;
393 p = (__be16 *)(skb->data + offset);
395 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
397 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
404 struct ipv6_tlv_tnl_enc_lim *tel;
406 case ICMPV6_DEST_UNREACH:
407 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
410 case ICMPV6_TIME_EXCEED:
411 if (code == ICMPV6_EXC_HOPLIMIT) {
412 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
416 case ICMPV6_PARAMPROB:
418 if (code == ICMPV6_HDR_FIELD)
419 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
421 if (teli && teli == be32_to_cpu(info) - 2) {
422 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
423 if (tel->encap_limit == 0) {
424 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
428 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
432 case ICMPV6_PKT_TOOBIG:
433 mtu = be32_to_cpu(info) - offset;
434 if (mtu < IPV6_MIN_MTU)
440 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
444 t->err_time = jiffies;
447 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
449 const struct ipv6hdr *ipv6h;
450 struct ip6_tnl *tunnel;
452 ipv6h = ipv6_hdr(skb);
453 tunnel = ip6gre_tunnel_lookup(skb->dev,
454 &ipv6h->saddr, &ipv6h->daddr, tpi->key,
457 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
462 return PACKET_REJECT;
465 static int gre_rcv(struct sk_buff *skb)
467 struct tnl_ptk_info tpi;
468 bool csum_err = false;
471 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6));
475 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
478 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
481 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
487 struct ipv6_tel_txoption {
488 struct ipv6_txoptions ops;
492 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
494 return iptunnel_handle_offloads(skb,
495 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
498 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
499 struct net_device *dev, __u8 dsfield,
500 struct flowi6 *fl6, int encap_limit,
501 __u32 *pmtu, __be16 proto)
503 struct ip6_tnl *tunnel = netdev_priv(dev);
504 __be16 protocol = (dev->type == ARPHRD_ETHER) ?
505 htons(ETH_P_TEB) : proto;
507 if (dev->type == ARPHRD_ETHER)
508 IPCB(skb)->flags = 0;
510 if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
511 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
513 fl6->daddr = tunnel->parms.raddr;
515 if (tunnel->parms.o_flags & TUNNEL_SEQ)
518 /* Push GRE header. */
519 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
520 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
522 skb_set_inner_protocol(skb, protocol);
524 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
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;
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)
550 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
551 fl6.flowi6_mark = skb->mark;
553 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
557 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
560 /* XXX: send ICMP error even if DF is not set. */
561 if (err == -EMSGSIZE)
562 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
570 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
572 struct ip6_tnl *t = netdev_priv(dev);
573 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
574 int encap_limit = -1;
581 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
584 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
586 struct ipv6_tlv_tnl_enc_lim *tel;
587 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
588 if (tel->encap_limit == 0) {
589 icmpv6_send(skb, ICMPV6_PARAMPROB,
590 ICMPV6_HDR_FIELD, offset + 2);
593 encap_limit = tel->encap_limit - 1;
594 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
595 encap_limit = t->parms.encap_limit;
597 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
599 dsfield = ipv6_get_dsfield(ipv6h);
600 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
601 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
602 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
603 fl6.flowlabel |= ip6_flowlabel(ipv6h);
604 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
605 fl6.flowi6_mark = skb->mark;
607 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
610 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
611 &mtu, skb->protocol);
613 if (err == -EMSGSIZE)
614 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
622 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
623 * @t: the outgoing tunnel device
624 * @hdr: IPv6 header from the incoming packet
627 * Avoid trivial tunneling loop by checking that tunnel exit-point
628 * doesn't match source of incoming packet.
635 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
636 const struct ipv6hdr *hdr)
638 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
641 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
643 struct ip6_tnl *t = netdev_priv(dev);
644 int encap_limit = -1;
649 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
650 encap_limit = t->parms.encap_limit;
652 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
653 fl6.flowi6_proto = skb->protocol;
655 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
659 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
664 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
665 struct net_device *dev)
667 struct ip6_tnl *t = netdev_priv(dev);
668 struct net_device_stats *stats = &t->dev->stats;
671 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
674 switch (skb->protocol) {
675 case htons(ETH_P_IP):
676 ret = ip6gre_xmit_ipv4(skb, dev);
678 case htons(ETH_P_IPV6):
679 ret = ip6gre_xmit_ipv6(skb, dev);
682 ret = ip6gre_xmit_other(skb, dev);
698 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
700 struct net_device *dev = t->dev;
701 struct __ip6_tnl_parm *p = &t->parms;
702 struct flowi6 *fl6 = &t->fl.u.ip6;
705 if (dev->type != ARPHRD_ETHER) {
706 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
707 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
710 /* Set up flowi template */
711 fl6->saddr = p->laddr;
712 fl6->daddr = p->raddr;
713 fl6->flowi6_oif = p->link;
716 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
717 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
718 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
719 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
721 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
722 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
724 if (p->flags&IP6_TNL_F_CAP_XMIT &&
725 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
726 dev->flags |= IFF_POINTOPOINT;
728 dev->flags &= ~IFF_POINTOPOINT;
730 t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
732 t->hlen = t->tun_hlen;
734 t_hlen = t->hlen + sizeof(struct ipv6hdr);
736 if (p->flags & IP6_TNL_F_CAP_XMIT) {
737 int strict = (ipv6_addr_type(&p->raddr) &
738 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
740 struct rt6_info *rt = rt6_lookup(t->net,
741 &p->raddr, &p->laddr,
748 dev->hard_header_len = rt->dst.dev->hard_header_len +
752 dev->mtu = rt->dst.dev->mtu - t_hlen;
753 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
755 if (dev->type == ARPHRD_ETHER)
756 dev->mtu -= ETH_HLEN;
758 if (dev->mtu < IPV6_MIN_MTU)
759 dev->mtu = IPV6_MIN_MTU;
766 static int ip6gre_tnl_change(struct ip6_tnl *t,
767 const struct __ip6_tnl_parm *p, int set_mtu)
769 t->parms.laddr = p->laddr;
770 t->parms.raddr = p->raddr;
771 t->parms.flags = p->flags;
772 t->parms.hop_limit = p->hop_limit;
773 t->parms.encap_limit = p->encap_limit;
774 t->parms.flowinfo = p->flowinfo;
775 t->parms.link = p->link;
776 t->parms.proto = p->proto;
777 t->parms.i_key = p->i_key;
778 t->parms.o_key = p->o_key;
779 t->parms.i_flags = p->i_flags;
780 t->parms.o_flags = p->o_flags;
781 dst_cache_reset(&t->dst_cache);
782 ip6gre_tnl_link_config(t, set_mtu);
786 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
787 const struct ip6_tnl_parm2 *u)
792 p->hop_limit = u->hop_limit;
793 p->encap_limit = u->encap_limit;
794 p->flowinfo = u->flowinfo;
798 p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
799 p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
800 memcpy(p->name, u->name, sizeof(u->name));
803 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
804 const struct __ip6_tnl_parm *p)
806 u->proto = IPPROTO_GRE;
810 u->hop_limit = p->hop_limit;
811 u->encap_limit = p->encap_limit;
812 u->flowinfo = p->flowinfo;
816 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
817 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
818 memcpy(u->name, p->name, sizeof(u->name));
821 static int ip6gre_tunnel_ioctl(struct net_device *dev,
822 struct ifreq *ifr, int cmd)
825 struct ip6_tnl_parm2 p;
826 struct __ip6_tnl_parm p1;
827 struct ip6_tnl *t = netdev_priv(dev);
828 struct net *net = t->net;
829 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
831 memset(&p1, 0, sizeof(p1));
835 if (dev == ign->fb_tunnel_dev) {
836 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
840 ip6gre_tnl_parm_from_user(&p1, &p);
841 t = ip6gre_tunnel_locate(net, &p1, 0);
843 t = netdev_priv(dev);
845 memset(&p, 0, sizeof(p));
846 ip6gre_tnl_parm_to_user(&p, &t->parms);
847 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
854 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
858 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
862 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
865 if (!(p.i_flags&GRE_KEY))
867 if (!(p.o_flags&GRE_KEY))
870 ip6gre_tnl_parm_from_user(&p1, &p);
871 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
873 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
880 t = netdev_priv(dev);
882 ip6gre_tunnel_unlink(ign, t);
884 ip6gre_tnl_change(t, &p1, 1);
885 ip6gre_tunnel_link(ign, t);
886 netdev_state_change(dev);
893 memset(&p, 0, sizeof(p));
894 ip6gre_tnl_parm_to_user(&p, &t->parms);
895 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
898 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
903 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
906 if (dev == ign->fb_tunnel_dev) {
908 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
911 ip6gre_tnl_parm_from_user(&p1, &p);
912 t = ip6gre_tunnel_locate(net, &p1, 0);
916 if (t == netdev_priv(ign->fb_tunnel_dev))
920 unregister_netdevice(dev);
932 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
934 const void *daddr, const void *saddr, unsigned int len)
936 struct ip6_tnl *t = netdev_priv(dev);
937 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
938 __be16 *p = (__be16 *)(ipv6h+1);
940 ip6_flow_hdr(ipv6h, 0,
941 ip6_make_flowlabel(dev_net(dev), skb,
942 t->fl.u.ip6.flowlabel, true,
944 ipv6h->hop_limit = t->parms.hop_limit;
945 ipv6h->nexthdr = NEXTHDR_GRE;
946 ipv6h->saddr = t->parms.laddr;
947 ipv6h->daddr = t->parms.raddr;
949 p[0] = t->parms.o_flags;
953 * Set the source hardware address.
957 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
959 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
960 if (!ipv6_addr_any(&ipv6h->daddr))
966 static const struct header_ops ip6gre_header_ops = {
967 .create = ip6gre_header,
970 static const struct net_device_ops ip6gre_netdev_ops = {
971 .ndo_init = ip6gre_tunnel_init,
972 .ndo_uninit = ip6gre_tunnel_uninit,
973 .ndo_start_xmit = ip6gre_tunnel_xmit,
974 .ndo_do_ioctl = ip6gre_tunnel_ioctl,
975 .ndo_change_mtu = ip6_tnl_change_mtu,
976 .ndo_get_stats64 = ip_tunnel_get_stats64,
977 .ndo_get_iflink = ip6_tnl_get_iflink,
980 static void ip6gre_dev_free(struct net_device *dev)
982 struct ip6_tnl *t = netdev_priv(dev);
984 dst_cache_destroy(&t->dst_cache);
985 free_percpu(dev->tstats);
989 static void ip6gre_tunnel_setup(struct net_device *dev)
991 dev->netdev_ops = &ip6gre_netdev_ops;
992 dev->destructor = ip6gre_dev_free;
994 dev->type = ARPHRD_IP6GRE;
996 dev->flags |= IFF_NOARP;
997 dev->addr_len = sizeof(struct in6_addr);
1001 static int ip6gre_tunnel_init_common(struct net_device *dev)
1003 struct ip6_tnl *tunnel;
1007 tunnel = netdev_priv(dev);
1010 tunnel->net = dev_net(dev);
1011 strcpy(tunnel->parms.name, dev->name);
1013 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1017 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1019 free_percpu(dev->tstats);
1024 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1026 tunnel->hlen = tunnel->tun_hlen;
1028 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1030 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1031 dev->mtu = ETH_DATA_LEN - t_hlen;
1032 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1038 static int ip6gre_tunnel_init(struct net_device *dev)
1040 struct ip6_tnl *tunnel;
1043 ret = ip6gre_tunnel_init_common(dev);
1047 tunnel = netdev_priv(dev);
1049 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1050 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1052 if (ipv6_addr_any(&tunnel->parms.raddr))
1053 dev->header_ops = &ip6gre_header_ops;
1058 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1060 struct ip6_tnl *tunnel = netdev_priv(dev);
1063 tunnel->net = dev_net(dev);
1064 strcpy(tunnel->parms.name, dev->name);
1066 tunnel->hlen = sizeof(struct ipv6hdr) + 4;
1072 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1074 .err_handler = ip6gre_err,
1075 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1078 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1080 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1081 struct net_device *dev, *aux;
1084 for_each_netdev_safe(net, dev, aux)
1085 if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1086 dev->rtnl_link_ops == &ip6gre_tap_ops)
1087 unregister_netdevice_queue(dev, head);
1089 for (prio = 0; prio < 4; prio++) {
1091 for (h = 0; h < HASH_SIZE; h++) {
1094 t = rtnl_dereference(ign->tunnels[prio][h]);
1097 /* If dev is in the same netns, it has already
1098 * been added to the list by the previous loop.
1100 if (!net_eq(dev_net(t->dev), net))
1101 unregister_netdevice_queue(t->dev,
1103 t = rtnl_dereference(t->next);
1109 static int __net_init ip6gre_init_net(struct net *net)
1111 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1114 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1116 ip6gre_tunnel_setup);
1117 if (!ign->fb_tunnel_dev) {
1121 dev_net_set(ign->fb_tunnel_dev, net);
1122 /* FB netdevice is special: we have one, and only one per netns.
1123 * Allowing to move it to another netns is clearly unsafe.
1125 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1128 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1129 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1131 err = register_netdev(ign->fb_tunnel_dev);
1135 rcu_assign_pointer(ign->tunnels_wc[0],
1136 netdev_priv(ign->fb_tunnel_dev));
1140 ip6gre_dev_free(ign->fb_tunnel_dev);
1145 static void __net_exit ip6gre_exit_net(struct net *net)
1150 ip6gre_destroy_tunnels(net, &list);
1151 unregister_netdevice_many(&list);
1155 static struct pernet_operations ip6gre_net_ops = {
1156 .init = ip6gre_init_net,
1157 .exit = ip6gre_exit_net,
1158 .id = &ip6gre_net_id,
1159 .size = sizeof(struct ip6gre_net),
1162 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1170 if (data[IFLA_GRE_IFLAGS])
1171 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1172 if (data[IFLA_GRE_OFLAGS])
1173 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1174 if (flags & (GRE_VERSION|GRE_ROUTING))
1180 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1182 struct in6_addr daddr;
1184 if (tb[IFLA_ADDRESS]) {
1185 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1187 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1188 return -EADDRNOTAVAIL;
1194 if (data[IFLA_GRE_REMOTE]) {
1195 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1196 if (ipv6_addr_any(&daddr))
1201 return ip6gre_tunnel_validate(tb, data);
1205 static void ip6gre_netlink_parms(struct nlattr *data[],
1206 struct __ip6_tnl_parm *parms)
1208 memset(parms, 0, sizeof(*parms));
1213 if (data[IFLA_GRE_LINK])
1214 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1216 if (data[IFLA_GRE_IFLAGS])
1217 parms->i_flags = gre_flags_to_tnl_flags(
1218 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1220 if (data[IFLA_GRE_OFLAGS])
1221 parms->o_flags = gre_flags_to_tnl_flags(
1222 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1224 if (data[IFLA_GRE_IKEY])
1225 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1227 if (data[IFLA_GRE_OKEY])
1228 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1230 if (data[IFLA_GRE_LOCAL])
1231 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1233 if (data[IFLA_GRE_REMOTE])
1234 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1236 if (data[IFLA_GRE_TTL])
1237 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1239 if (data[IFLA_GRE_ENCAP_LIMIT])
1240 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1242 if (data[IFLA_GRE_FLOWINFO])
1243 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]);
1245 if (data[IFLA_GRE_FLAGS])
1246 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1249 static int ip6gre_tap_init(struct net_device *dev)
1251 struct ip6_tnl *tunnel;
1254 ret = ip6gre_tunnel_init_common(dev);
1258 tunnel = netdev_priv(dev);
1260 ip6gre_tnl_link_config(tunnel, 1);
1265 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1266 .ndo_init = ip6gre_tap_init,
1267 .ndo_uninit = ip6gre_tunnel_uninit,
1268 .ndo_start_xmit = ip6gre_tunnel_xmit,
1269 .ndo_set_mac_address = eth_mac_addr,
1270 .ndo_validate_addr = eth_validate_addr,
1271 .ndo_change_mtu = ip6_tnl_change_mtu,
1272 .ndo_get_stats64 = ip_tunnel_get_stats64,
1273 .ndo_get_iflink = ip6_tnl_get_iflink,
1276 #define GRE6_FEATURES (NETIF_F_SG | \
1277 NETIF_F_FRAGLIST | \
1281 static void ip6gre_tap_setup(struct net_device *dev)
1286 dev->netdev_ops = &ip6gre_tap_netdev_ops;
1287 dev->destructor = ip6gre_dev_free;
1289 dev->features |= NETIF_F_NETNS_LOCAL;
1290 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1293 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1294 struct nlattr *tb[], struct nlattr *data[])
1297 struct net *net = dev_net(dev);
1298 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1301 nt = netdev_priv(dev);
1302 ip6gre_netlink_parms(data, &nt->parms);
1304 if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1307 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1308 eth_hw_addr_random(dev);
1311 nt->net = dev_net(dev);
1312 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1314 dev->features |= GRE6_FEATURES;
1315 dev->hw_features |= GRE6_FEATURES;
1317 if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1318 /* TCP segmentation offload is not supported when we
1319 * generate output sequences.
1321 dev->features |= NETIF_F_GSO_SOFTWARE;
1322 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1324 /* Can use a lockless transmit, unless we generate
1327 dev->features |= NETIF_F_LLTX;
1330 err = register_netdevice(dev);
1335 ip6gre_tunnel_link(ign, nt);
1341 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1342 struct nlattr *data[])
1344 struct ip6_tnl *t, *nt = netdev_priv(dev);
1345 struct net *net = nt->net;
1346 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1347 struct __ip6_tnl_parm p;
1349 if (dev == ign->fb_tunnel_dev)
1352 ip6gre_netlink_parms(data, &p);
1354 t = ip6gre_tunnel_locate(net, &p, 0);
1363 ip6gre_tunnel_unlink(ign, t);
1364 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1365 ip6gre_tunnel_link(ign, t);
1369 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1371 struct net *net = dev_net(dev);
1372 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1374 if (dev != ign->fb_tunnel_dev)
1375 unregister_netdevice_queue(dev, head);
1378 static size_t ip6gre_get_size(const struct net_device *dev)
1383 /* IFLA_GRE_IFLAGS */
1385 /* IFLA_GRE_OFLAGS */
1391 /* IFLA_GRE_LOCAL */
1392 nla_total_size(sizeof(struct in6_addr)) +
1393 /* IFLA_GRE_REMOTE */
1394 nla_total_size(sizeof(struct in6_addr)) +
1397 /* IFLA_GRE_ENCAP_LIMIT */
1399 /* IFLA_GRE_FLOWINFO */
1401 /* IFLA_GRE_FLAGS */
1406 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1408 struct ip6_tnl *t = netdev_priv(dev);
1409 struct __ip6_tnl_parm *p = &t->parms;
1411 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1412 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1413 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1414 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1415 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1416 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1417 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1418 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1419 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1420 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1421 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1422 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1423 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1424 goto nla_put_failure;
1431 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1432 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1433 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1434 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1435 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1436 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1437 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1438 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1439 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1440 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1441 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 },
1442 [IFLA_GRE_FLAGS] = { .type = NLA_U32 },
1445 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1447 .maxtype = IFLA_GRE_MAX,
1448 .policy = ip6gre_policy,
1449 .priv_size = sizeof(struct ip6_tnl),
1450 .setup = ip6gre_tunnel_setup,
1451 .validate = ip6gre_tunnel_validate,
1452 .newlink = ip6gre_newlink,
1453 .changelink = ip6gre_changelink,
1454 .dellink = ip6gre_dellink,
1455 .get_size = ip6gre_get_size,
1456 .fill_info = ip6gre_fill_info,
1457 .get_link_net = ip6_tnl_get_link_net,
1460 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1461 .kind = "ip6gretap",
1462 .maxtype = IFLA_GRE_MAX,
1463 .policy = ip6gre_policy,
1464 .priv_size = sizeof(struct ip6_tnl),
1465 .setup = ip6gre_tap_setup,
1466 .validate = ip6gre_tap_validate,
1467 .newlink = ip6gre_newlink,
1468 .changelink = ip6gre_changelink,
1469 .get_size = ip6gre_get_size,
1470 .fill_info = ip6gre_fill_info,
1471 .get_link_net = ip6_tnl_get_link_net,
1475 * And now the modules code and kernel interface.
1478 static int __init ip6gre_init(void)
1482 pr_info("GRE over IPv6 tunneling driver\n");
1484 err = register_pernet_device(&ip6gre_net_ops);
1488 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1490 pr_info("%s: can't add protocol\n", __func__);
1491 goto add_proto_failed;
1494 err = rtnl_link_register(&ip6gre_link_ops);
1496 goto rtnl_link_failed;
1498 err = rtnl_link_register(&ip6gre_tap_ops);
1500 goto tap_ops_failed;
1506 rtnl_link_unregister(&ip6gre_link_ops);
1508 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1510 unregister_pernet_device(&ip6gre_net_ops);
1514 static void __exit ip6gre_fini(void)
1516 rtnl_link_unregister(&ip6gre_tap_ops);
1517 rtnl_link_unregister(&ip6gre_link_ops);
1518 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1519 unregister_pernet_device(&ip6gre_net_ops);
1522 module_init(ip6gre_init);
1523 module_exit(ip6gre_fini);
1524 MODULE_LICENSE("GPL");
1525 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1526 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1527 MODULE_ALIAS_RTNL_LINK("ip6gre");
1528 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1529 MODULE_ALIAS_NETDEV("ip6gre0");