2 * Linux NET3: GRE over IP protocol decoder.
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 <asm/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/mroute.h>
28 #include <linux/init.h>
29 #include <linux/in6.h>
30 #include <linux/inetdevice.h>
31 #include <linux/igmp.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/etherdevice.h>
34 #include <linux/if_ether.h>
39 #include <net/protocol.h>
40 #include <net/ip_tunnels.h>
42 #include <net/checksum.h>
43 #include <net/dsfield.h>
44 #include <net/inet_ecn.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <net/rtnetlink.h>
51 #if IS_ENABLED(CONFIG_IPV6)
53 #include <net/ip6_fib.h>
54 #include <net/ip6_route.h>
61 1. The most important issue is detecting local dead loops.
62 They would cause complete host lockup in transmit, which
63 would be "resolved" by stack overflow or, if queueing is enabled,
64 with infinite looping in net_bh.
66 We cannot track such dead loops during route installation,
67 it is infeasible task. The most general solutions would be
68 to keep skb->encapsulation counter (sort of local ttl),
69 and silently drop packet when it expires. It is a good
70 solution, but it supposes maintaining new variable in ALL
71 skb, even if no tunneling is used.
73 Current solution: xmit_recursion breaks dead loops. This is a percpu
74 counter, since when we enter the first ndo_xmit(), cpu migration is
75 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
77 2. Networking dead loops would not kill routers, but would really
78 kill network. IP hop limit plays role of "t->recursion" in this case,
79 if we copy it from packet being encapsulated to upper header.
80 It is very good solution, but it introduces two problems:
82 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
83 do not work over tunnels.
84 - traceroute does not work. I planned to relay ICMP from tunnel,
85 so that this problem would be solved and traceroute output
86 would even more informative. This idea appeared to be wrong:
87 only Linux complies to rfc1812 now (yes, guys, Linux is the only
88 true router now :-)), all routers (at least, in neighbourhood of mine)
89 return only 8 bytes of payload. It is the end.
91 Hence, if we want that OSPF worked or traceroute said something reasonable,
92 we should search for another solution.
94 One of them is to parse packet trying to detect inner encapsulation
95 made by our node. It is difficult or even impossible, especially,
96 taking into account fragmentation. TO be short, ttl is not solution at all.
98 Current solution: The solution was UNEXPECTEDLY SIMPLE.
99 We force DF flag on tunnels with preconfigured hop limit,
100 that is ALL. :-) Well, it does not remove the problem completely,
101 but exponential growth of network traffic is changed to linear
102 (branches, that exceed pmtu are pruned) and tunnel mtu
103 rapidly degrades to value <68, where looping stops.
104 Yes, it is not good if there exists a router in the loop,
105 which does not force DF, even when encapsulating packets have DF set.
106 But it is not our problem! Nobody could accuse us, we made
107 all that we could make. Even if it is your gated who injected
108 fatal route to network, even if it were you who configured
109 fatal static route: you are innocent. :-)
114 static bool log_ecn_error = true;
115 module_param(log_ecn_error, bool, 0644);
116 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
118 static struct rtnl_link_ops ipgre_link_ops __read_mostly;
119 static int ipgre_tunnel_init(struct net_device *dev);
121 static int ipgre_net_id __read_mostly;
122 static int gre_tap_net_id __read_mostly;
124 static __sum16 check_checksum(struct sk_buff *skb)
128 switch (skb->ip_summed) {
129 case CHECKSUM_COMPLETE:
130 csum = csum_fold(skb->csum);
138 csum = __skb_checksum_complete(skb);
139 skb->ip_summed = CHECKSUM_COMPLETE;
146 static int ip_gre_calc_hlen(__be16 o_flags)
150 if (o_flags&TUNNEL_CSUM)
152 if (o_flags&TUNNEL_KEY)
154 if (o_flags&TUNNEL_SEQ)
159 static int parse_gre_header(struct sk_buff *skb, struct tnl_ptk_info *tpi,
160 bool *csum_err, int *hdr_len)
162 unsigned int ip_hlen = ip_hdrlen(skb);
163 const struct gre_base_hdr *greh;
166 if (unlikely(!pskb_may_pull(skb, sizeof(struct gre_base_hdr))))
169 greh = (struct gre_base_hdr *)(skb_network_header(skb) + ip_hlen);
170 if (unlikely(greh->flags & (GRE_VERSION | GRE_ROUTING)))
173 tpi->flags = gre_flags_to_tnl_flags(greh->flags);
174 *hdr_len = ip_gre_calc_hlen(tpi->flags);
176 if (!pskb_may_pull(skb, *hdr_len))
179 greh = (struct gre_base_hdr *)(skb_network_header(skb) + ip_hlen);
181 tpi->proto = greh->protocol;
183 options = (__be32 *)(greh + 1);
184 if (greh->flags & GRE_CSUM) {
185 if (check_checksum(skb)) {
192 if (greh->flags & GRE_KEY) {
198 if (unlikely(greh->flags & GRE_SEQ)) {
204 /* WCCP version 1 and 2 protocol decoding.
205 * - Change protocol to IP
206 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
208 if (greh->flags == 0 && tpi->proto == htons(ETH_P_WCCP)) {
209 tpi->proto = htons(ETH_P_IP);
210 if ((*(u8 *)options & 0xF0) != 0x40) {
212 if (!pskb_may_pull(skb, *hdr_len))
220 static void ipgre_err(struct sk_buff *skb, u32 info)
223 /* All the routers (except for Linux) return only
224 8 bytes of packet payload. It means, that precise relaying of
225 ICMP in the real Internet is absolutely infeasible.
227 Moreover, Cisco "wise men" put GRE key to the third word
228 in GRE header. It makes impossible maintaining even soft
229 state for keyed GRE tunnels with enabled checksum. Tell
232 Well, I wonder, rfc1812 was written by Cisco employee,
233 what the hell these idiots break standards established
236 struct net *net = dev_net(skb->dev);
237 struct ip_tunnel_net *itn;
238 const struct iphdr *iph = (const struct iphdr *)skb->data;
239 const int type = icmp_hdr(skb)->type;
240 const int code = icmp_hdr(skb)->code;
242 struct tnl_ptk_info tpi;
244 bool csum_err = false;
246 if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len)) {
247 if (!csum_err) /* ignore csum errors. */
253 case ICMP_PARAMETERPROB:
256 case ICMP_DEST_UNREACH:
259 case ICMP_PORT_UNREACH:
260 /* Impossible event. */
263 /* All others are translated to HOST_UNREACH.
264 rfc2003 contains "deep thoughts" about NET_UNREACH,
265 I believe they are just ether pollution. --ANK
270 case ICMP_TIME_EXCEEDED:
271 if (code != ICMP_EXC_TTL)
279 if (tpi.proto == htons(ETH_P_TEB))
280 itn = net_generic(net, gre_tap_net_id);
282 itn = net_generic(net, ipgre_net_id);
284 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags,
285 iph->daddr, iph->saddr, tpi.key);
290 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
291 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
292 t->parms.link, 0, IPPROTO_GRE, 0);
295 if (type == ICMP_REDIRECT) {
296 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
300 if (t->parms.iph.daddr == 0 ||
301 ipv4_is_multicast(t->parms.iph.daddr))
304 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
307 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
311 t->err_time = jiffies;
314 static int ipgre_rcv(struct sk_buff *skb)
316 struct net *net = dev_net(skb->dev);
317 struct ip_tunnel_net *itn;
318 const struct iphdr *iph;
319 struct ip_tunnel *tunnel;
320 struct tnl_ptk_info tpi;
322 bool csum_err = false;
324 if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len) < 0)
327 if (tpi.proto == htons(ETH_P_TEB))
328 itn = net_generic(net, gre_tap_net_id);
330 itn = net_generic(net, ipgre_net_id);
333 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags,
334 iph->saddr, iph->daddr, tpi.key);
337 ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
340 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
346 static struct sk_buff *handle_offloads(struct ip_tunnel *tunnel, struct sk_buff *skb)
350 if (skb_is_gso(skb)) {
351 err = skb_unclone(skb, GFP_ATOMIC);
354 skb_shinfo(skb)->gso_type |= SKB_GSO_GRE;
356 } else if (skb->ip_summed == CHECKSUM_PARTIAL &&
357 tunnel->parms.o_flags&TUNNEL_CSUM) {
358 err = skb_checksum_help(skb);
361 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
362 skb->ip_summed = CHECKSUM_NONE;
371 static struct sk_buff *gre_build_header(struct sk_buff *skb,
372 const struct tnl_ptk_info *tpi,
375 struct gre_base_hdr *greh;
377 skb_push(skb, hdr_len);
379 greh = (struct gre_base_hdr *)skb->data;
380 greh->flags = tnl_flags_to_gre_flags(tpi->flags);
381 greh->protocol = tpi->proto;
383 if (tpi->flags&(TUNNEL_KEY|TUNNEL_CSUM|TUNNEL_SEQ)) {
384 __be32 *ptr = (__be32 *)(((u8 *)greh) + hdr_len - 4);
386 if (tpi->flags&TUNNEL_SEQ) {
390 if (tpi->flags&TUNNEL_KEY) {
394 if (tpi->flags&TUNNEL_CSUM &&
395 !(skb_shinfo(skb)->gso_type & SKB_GSO_GRE)) {
397 *(__sum16 *)ptr = csum_fold(skb_checksum(skb, 0,
405 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
406 const struct iphdr *tnl_params,
409 struct ip_tunnel *tunnel = netdev_priv(dev);
410 struct tnl_ptk_info tpi;
412 if (likely(!skb->encapsulation)) {
413 skb_reset_inner_headers(skb);
414 skb->encapsulation = 1;
417 tpi.flags = tunnel->parms.o_flags;
419 tpi.key = tunnel->parms.o_key;
420 if (tunnel->parms.o_flags & TUNNEL_SEQ)
422 tpi.seq = htonl(tunnel->o_seqno);
424 /* Push GRE header. */
425 skb = gre_build_header(skb, &tpi, tunnel->hlen);
426 if (unlikely(!skb)) {
427 dev->stats.tx_dropped++;
431 ip_tunnel_xmit(skb, dev, tnl_params);
434 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
435 struct net_device *dev)
437 struct ip_tunnel *tunnel = netdev_priv(dev);
438 const struct iphdr *tnl_params;
440 skb = handle_offloads(tunnel, skb);
444 if (dev->header_ops) {
445 /* Need space for new headers */
446 if (skb_cow_head(skb, dev->needed_headroom -
447 (tunnel->hlen + sizeof(struct iphdr))))
450 tnl_params = (const struct iphdr *)skb->data;
452 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
455 skb_pull(skb, tunnel->hlen + sizeof(struct iphdr));
457 if (skb_cow_head(skb, dev->needed_headroom))
460 tnl_params = &tunnel->parms.iph;
463 __gre_xmit(skb, dev, tnl_params, skb->protocol);
470 dev->stats.tx_dropped++;
474 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
475 struct net_device *dev)
477 struct ip_tunnel *tunnel = netdev_priv(dev);
479 skb = handle_offloads(tunnel, skb);
483 if (skb_cow_head(skb, dev->needed_headroom))
486 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
493 dev->stats.tx_dropped++;
497 static int ipgre_tunnel_ioctl(struct net_device *dev,
498 struct ifreq *ifr, int cmd)
501 struct ip_tunnel_parm p;
503 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
505 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
506 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
507 ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) {
510 p.i_flags = gre_flags_to_tnl_flags(p.i_flags);
511 p.o_flags = gre_flags_to_tnl_flags(p.o_flags);
513 err = ip_tunnel_ioctl(dev, &p, cmd);
517 p.i_flags = tnl_flags_to_gre_flags(p.i_flags);
518 p.o_flags = tnl_flags_to_gre_flags(p.o_flags);
520 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
525 /* Nice toy. Unfortunately, useless in real life :-)
526 It allows to construct virtual multiprotocol broadcast "LAN"
527 over the Internet, provided multicast routing is tuned.
530 I have no idea was this bicycle invented before me,
531 so that I had to set ARPHRD_IPGRE to a random value.
532 I have an impression, that Cisco could make something similar,
533 but this feature is apparently missing in IOS<=11.2(8).
535 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
536 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
538 ping -t 255 224.66.66.66
540 If nobody answers, mbone does not work.
542 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
543 ip addr add 10.66.66.<somewhat>/24 dev Universe
545 ifconfig Universe add fe80::<Your_real_addr>/10
546 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
549 ftp fec0:6666:6666::193.233.7.65
552 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
554 const void *daddr, const void *saddr, unsigned int len)
556 struct ip_tunnel *t = netdev_priv(dev);
558 struct gre_base_hdr *greh;
560 iph = (struct iphdr *)skb_push(skb, t->hlen + sizeof(*iph));
561 greh = (struct gre_base_hdr *)(iph+1);
562 greh->flags = tnl_flags_to_gre_flags(t->parms.o_flags);
563 greh->protocol = htons(type);
565 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
567 /* Set the source hardware address. */
569 memcpy(&iph->saddr, saddr, 4);
571 memcpy(&iph->daddr, daddr, 4);
575 return -(t->hlen + sizeof(*iph));
578 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
580 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
581 memcpy(haddr, &iph->saddr, 4);
585 static const struct header_ops ipgre_header_ops = {
586 .create = ipgre_header,
587 .parse = ipgre_header_parse,
590 #ifdef CONFIG_NET_IPGRE_BROADCAST
591 static int ipgre_open(struct net_device *dev)
593 struct ip_tunnel *t = netdev_priv(dev);
595 if (ipv4_is_multicast(t->parms.iph.daddr)) {
599 rt = ip_route_output_gre(dev_net(dev), &fl4,
603 RT_TOS(t->parms.iph.tos),
606 return -EADDRNOTAVAIL;
609 if (__in_dev_get_rtnl(dev) == NULL)
610 return -EADDRNOTAVAIL;
611 t->mlink = dev->ifindex;
612 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
617 static int ipgre_close(struct net_device *dev)
619 struct ip_tunnel *t = netdev_priv(dev);
621 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
622 struct in_device *in_dev;
623 in_dev = inetdev_by_index(dev_net(dev), t->mlink);
625 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
631 static const struct net_device_ops ipgre_netdev_ops = {
632 .ndo_init = ipgre_tunnel_init,
633 .ndo_uninit = ip_tunnel_uninit,
634 #ifdef CONFIG_NET_IPGRE_BROADCAST
635 .ndo_open = ipgre_open,
636 .ndo_stop = ipgre_close,
638 .ndo_start_xmit = ipgre_xmit,
639 .ndo_do_ioctl = ipgre_tunnel_ioctl,
640 .ndo_change_mtu = ip_tunnel_change_mtu,
641 .ndo_get_stats64 = ip_tunnel_get_stats64,
644 #define GRE_FEATURES (NETIF_F_SG | \
649 static void ipgre_tunnel_setup(struct net_device *dev)
651 dev->netdev_ops = &ipgre_netdev_ops;
652 ip_tunnel_setup(dev, ipgre_net_id);
655 static void __gre_tunnel_init(struct net_device *dev)
657 struct ip_tunnel *tunnel;
659 tunnel = netdev_priv(dev);
660 tunnel->hlen = ip_gre_calc_hlen(tunnel->parms.o_flags);
661 tunnel->parms.iph.protocol = IPPROTO_GRE;
663 dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
664 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4;
666 dev->features |= NETIF_F_NETNS_LOCAL | GRE_FEATURES;
667 dev->hw_features |= GRE_FEATURES;
669 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
670 /* TCP offload with GRE SEQ is not supported. */
671 dev->features |= NETIF_F_GSO_SOFTWARE;
672 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
673 /* Can use a lockless transmit, unless we generate
676 dev->features |= NETIF_F_LLTX;
680 static int ipgre_tunnel_init(struct net_device *dev)
682 struct ip_tunnel *tunnel = netdev_priv(dev);
683 struct iphdr *iph = &tunnel->parms.iph;
685 __gre_tunnel_init(dev);
687 memcpy(dev->dev_addr, &iph->saddr, 4);
688 memcpy(dev->broadcast, &iph->daddr, 4);
690 dev->type = ARPHRD_IPGRE;
691 dev->flags = IFF_NOARP;
692 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
696 #ifdef CONFIG_NET_IPGRE_BROADCAST
697 if (ipv4_is_multicast(iph->daddr)) {
700 dev->flags = IFF_BROADCAST;
701 dev->header_ops = &ipgre_header_ops;
705 dev->header_ops = &ipgre_header_ops;
707 return ip_tunnel_init(dev);
710 static const struct gre_protocol ipgre_protocol = {
711 .handler = ipgre_rcv,
712 .err_handler = ipgre_err,
715 static int __net_init ipgre_init_net(struct net *net)
717 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
720 static void __net_exit ipgre_exit_net(struct net *net)
722 struct ip_tunnel_net *itn = net_generic(net, ipgre_net_id);
723 ip_tunnel_delete_net(itn);
726 static struct pernet_operations ipgre_net_ops = {
727 .init = ipgre_init_net,
728 .exit = ipgre_exit_net,
730 .size = sizeof(struct ip_tunnel_net),
733 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
741 if (data[IFLA_GRE_IFLAGS])
742 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
743 if (data[IFLA_GRE_OFLAGS])
744 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
745 if (flags & (GRE_VERSION|GRE_ROUTING))
751 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
755 if (tb[IFLA_ADDRESS]) {
756 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
758 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
759 return -EADDRNOTAVAIL;
765 if (data[IFLA_GRE_REMOTE]) {
766 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
772 return ipgre_tunnel_validate(tb, data);
775 static void ipgre_netlink_parms(struct nlattr *data[], struct nlattr *tb[],
776 struct ip_tunnel_parm *parms)
778 memset(parms, 0, sizeof(*parms));
780 parms->iph.protocol = IPPROTO_GRE;
785 if (data[IFLA_GRE_LINK])
786 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
788 if (data[IFLA_GRE_IFLAGS])
789 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
791 if (data[IFLA_GRE_OFLAGS])
792 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
794 if (data[IFLA_GRE_IKEY])
795 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
797 if (data[IFLA_GRE_OKEY])
798 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
800 if (data[IFLA_GRE_LOCAL])
801 parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]);
803 if (data[IFLA_GRE_REMOTE])
804 parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]);
806 if (data[IFLA_GRE_TTL])
807 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
809 if (data[IFLA_GRE_TOS])
810 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
812 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
813 parms->iph.frag_off = htons(IP_DF);
816 static int gre_tap_init(struct net_device *dev)
818 __gre_tunnel_init(dev);
820 return ip_tunnel_init(dev);
823 static const struct net_device_ops gre_tap_netdev_ops = {
824 .ndo_init = gre_tap_init,
825 .ndo_uninit = ip_tunnel_uninit,
826 .ndo_start_xmit = gre_tap_xmit,
827 .ndo_set_mac_address = eth_mac_addr,
828 .ndo_validate_addr = eth_validate_addr,
829 .ndo_change_mtu = ip_tunnel_change_mtu,
830 .ndo_get_stats64 = ip_tunnel_get_stats64,
833 static void ipgre_tap_setup(struct net_device *dev)
836 dev->netdev_ops = &gre_tap_netdev_ops;
837 ip_tunnel_setup(dev, gre_tap_net_id);
840 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
841 struct nlattr *tb[], struct nlattr *data[])
843 struct ip_tunnel_parm p;
845 ipgre_netlink_parms(data, tb, &p);
846 return ip_tunnel_newlink(dev, tb, &p);
849 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
850 struct nlattr *data[])
852 struct ip_tunnel_parm p;
854 ipgre_netlink_parms(data, tb, &p);
855 return ip_tunnel_changelink(dev, tb, &p);
858 static size_t ipgre_get_size(const struct net_device *dev)
863 /* IFLA_GRE_IFLAGS */
865 /* IFLA_GRE_OFLAGS */
873 /* IFLA_GRE_REMOTE */
879 /* IFLA_GRE_PMTUDISC */
884 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
886 struct ip_tunnel *t = netdev_priv(dev);
887 struct ip_tunnel_parm *p = &t->parms;
889 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
890 nla_put_be16(skb, IFLA_GRE_IFLAGS, tnl_flags_to_gre_flags(p->i_flags)) ||
891 nla_put_be16(skb, IFLA_GRE_OFLAGS, tnl_flags_to_gre_flags(p->o_flags)) ||
892 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
893 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
894 nla_put_be32(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
895 nla_put_be32(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
896 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
897 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
898 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
899 !!(p->iph.frag_off & htons(IP_DF))))
900 goto nla_put_failure;
907 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
908 [IFLA_GRE_LINK] = { .type = NLA_U32 },
909 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
910 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
911 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
912 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
913 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
914 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
915 [IFLA_GRE_TTL] = { .type = NLA_U8 },
916 [IFLA_GRE_TOS] = { .type = NLA_U8 },
917 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
920 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
922 .maxtype = IFLA_GRE_MAX,
923 .policy = ipgre_policy,
924 .priv_size = sizeof(struct ip_tunnel),
925 .setup = ipgre_tunnel_setup,
926 .validate = ipgre_tunnel_validate,
927 .newlink = ipgre_newlink,
928 .changelink = ipgre_changelink,
929 .dellink = ip_tunnel_dellink,
930 .get_size = ipgre_get_size,
931 .fill_info = ipgre_fill_info,
934 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
936 .maxtype = IFLA_GRE_MAX,
937 .policy = ipgre_policy,
938 .priv_size = sizeof(struct ip_tunnel),
939 .setup = ipgre_tap_setup,
940 .validate = ipgre_tap_validate,
941 .newlink = ipgre_newlink,
942 .changelink = ipgre_changelink,
943 .dellink = ip_tunnel_dellink,
944 .get_size = ipgre_get_size,
945 .fill_info = ipgre_fill_info,
948 static int __net_init ipgre_tap_init_net(struct net *net)
950 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, NULL);
953 static void __net_exit ipgre_tap_exit_net(struct net *net)
955 struct ip_tunnel_net *itn = net_generic(net, gre_tap_net_id);
956 ip_tunnel_delete_net(itn);
959 static struct pernet_operations ipgre_tap_net_ops = {
960 .init = ipgre_tap_init_net,
961 .exit = ipgre_tap_exit_net,
962 .id = &gre_tap_net_id,
963 .size = sizeof(struct ip_tunnel_net),
966 static int __init ipgre_init(void)
970 pr_info("GRE over IPv4 tunneling driver\n");
972 err = register_pernet_device(&ipgre_net_ops);
976 err = register_pernet_device(&ipgre_tap_net_ops);
980 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
982 pr_info("%s: can't add protocol\n", __func__);
983 goto add_proto_failed;
986 err = rtnl_link_register(&ipgre_link_ops);
988 goto rtnl_link_failed;
990 err = rtnl_link_register(&ipgre_tap_ops);
997 rtnl_link_unregister(&ipgre_link_ops);
999 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1001 unregister_pernet_device(&ipgre_tap_net_ops);
1003 unregister_pernet_device(&ipgre_net_ops);
1007 static void __exit ipgre_fini(void)
1009 rtnl_link_unregister(&ipgre_tap_ops);
1010 rtnl_link_unregister(&ipgre_link_ops);
1011 if (gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO) < 0)
1012 pr_info("%s: can't remove protocol\n", __func__);
1013 unregister_pernet_device(&ipgre_tap_net_ops);
1014 unregister_pernet_device(&ipgre_net_ops);
1017 module_init(ipgre_init);
1018 module_exit(ipgre_fini);
1019 MODULE_LICENSE("GPL");
1020 MODULE_ALIAS_RTNL_LINK("gre");
1021 MODULE_ALIAS_RTNL_LINK("gretap");
1022 MODULE_ALIAS_NETDEV("gre0");
1023 MODULE_ALIAS_NETDEV("gretap0");