1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux NET3: GRE over IP protocol decoder.
5 * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.ru)
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/if_vlan.h>
23 #include <linux/init.h>
24 #include <linux/in6.h>
25 #include <linux/inetdevice.h>
26 #include <linux/igmp.h>
27 #include <linux/netfilter_ipv4.h>
28 #include <linux/etherdevice.h>
29 #include <linux/if_ether.h>
34 #include <net/protocol.h>
35 #include <net/ip_tunnels.h>
37 #include <net/checksum.h>
38 #include <net/dsfield.h>
39 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/rtnetlink.h>
45 #include <net/dst_metadata.h>
46 #include <net/erspan.h>
52 1. The most important issue is detecting local dead loops.
53 They would cause complete host lockup in transmit, which
54 would be "resolved" by stack overflow or, if queueing is enabled,
55 with infinite looping in net_bh.
57 We cannot track such dead loops during route installation,
58 it is infeasible task. The most general solutions would be
59 to keep skb->encapsulation counter (sort of local ttl),
60 and silently drop packet when it expires. It is a good
61 solution, but it supposes maintaining new variable in ALL
62 skb, even if no tunneling is used.
64 Current solution: xmit_recursion breaks dead loops. This is a percpu
65 counter, since when we enter the first ndo_xmit(), cpu migration is
66 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
68 2. Networking dead loops would not kill routers, but would really
69 kill network. IP hop limit plays role of "t->recursion" in this case,
70 if we copy it from packet being encapsulated to upper header.
71 It is very good solution, but it introduces two problems:
73 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
74 do not work over tunnels.
75 - traceroute does not work. I planned to relay ICMP from tunnel,
76 so that this problem would be solved and traceroute output
77 would even more informative. This idea appeared to be wrong:
78 only Linux complies to rfc1812 now (yes, guys, Linux is the only
79 true router now :-)), all routers (at least, in neighbourhood of mine)
80 return only 8 bytes of payload. It is the end.
82 Hence, if we want that OSPF worked or traceroute said something reasonable,
83 we should search for another solution.
85 One of them is to parse packet trying to detect inner encapsulation
86 made by our node. It is difficult or even impossible, especially,
87 taking into account fragmentation. TO be short, ttl is not solution at all.
89 Current solution: The solution was UNEXPECTEDLY SIMPLE.
90 We force DF flag on tunnels with preconfigured hop limit,
91 that is ALL. :-) Well, it does not remove the problem completely,
92 but exponential growth of network traffic is changed to linear
93 (branches, that exceed pmtu are pruned) and tunnel mtu
94 rapidly degrades to value <68, where looping stops.
95 Yes, it is not good if there exists a router in the loop,
96 which does not force DF, even when encapsulating packets have DF set.
97 But it is not our problem! Nobody could accuse us, we made
98 all that we could make. Even if it is your gated who injected
99 fatal route to network, even if it were you who configured
100 fatal static route: you are innocent. :-)
105 static bool log_ecn_error = true;
106 module_param(log_ecn_error, bool, 0644);
107 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
109 static struct rtnl_link_ops ipgre_link_ops __read_mostly;
110 static const struct header_ops ipgre_header_ops;
112 static int ipgre_tunnel_init(struct net_device *dev);
113 static void erspan_build_header(struct sk_buff *skb,
115 bool truncate, bool is_ipv4);
117 static unsigned int ipgre_net_id __read_mostly;
118 static unsigned int gre_tap_net_id __read_mostly;
119 static unsigned int erspan_net_id __read_mostly;
121 static int ipgre_err(struct sk_buff *skb, u32 info,
122 const struct tnl_ptk_info *tpi)
125 /* All the routers (except for Linux) return only
126 8 bytes of packet payload. It means, that precise relaying of
127 ICMP in the real Internet is absolutely infeasible.
129 Moreover, Cisco "wise men" put GRE key to the third word
130 in GRE header. It makes impossible maintaining even soft
131 state for keyed GRE tunnels with enabled checksum. Tell
134 Well, I wonder, rfc1812 was written by Cisco employee,
135 what the hell these idiots break standards established
138 struct net *net = dev_net(skb->dev);
139 struct ip_tunnel_net *itn;
140 const struct iphdr *iph;
141 const int type = icmp_hdr(skb)->type;
142 const int code = icmp_hdr(skb)->code;
143 unsigned int data_len = 0;
146 if (tpi->proto == htons(ETH_P_TEB))
147 itn = net_generic(net, gre_tap_net_id);
148 else if (tpi->proto == htons(ETH_P_ERSPAN) ||
149 tpi->proto == htons(ETH_P_ERSPAN2))
150 itn = net_generic(net, erspan_net_id);
152 itn = net_generic(net, ipgre_net_id);
154 iph = (const struct iphdr *)(icmp_hdr(skb) + 1);
155 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
156 iph->daddr, iph->saddr, tpi->key);
163 case ICMP_PARAMETERPROB:
166 case ICMP_DEST_UNREACH:
169 case ICMP_PORT_UNREACH:
170 /* Impossible event. */
173 /* All others are translated to HOST_UNREACH.
174 rfc2003 contains "deep thoughts" about NET_UNREACH,
175 I believe they are just ether pollution. --ANK
181 case ICMP_TIME_EXCEEDED:
182 if (code != ICMP_EXC_TTL)
184 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
191 #if IS_ENABLED(CONFIG_IPV6)
192 if (tpi->proto == htons(ETH_P_IPV6) &&
193 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4 + tpi->hdr_len,
198 if (t->parms.iph.daddr == 0 ||
199 ipv4_is_multicast(t->parms.iph.daddr))
202 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
205 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
209 t->err_time = jiffies;
214 static void gre_err(struct sk_buff *skb, u32 info)
216 /* All the routers (except for Linux) return only
217 * 8 bytes of packet payload. It means, that precise relaying of
218 * ICMP in the real Internet is absolutely infeasible.
220 * Moreover, Cisco "wise men" put GRE key to the third word
221 * in GRE header. It makes impossible maintaining even soft
223 * GRE tunnels with enabled checksum. Tell them "thank you".
225 * Well, I wonder, rfc1812 was written by Cisco employee,
226 * what the hell these idiots break standards established
230 const struct iphdr *iph = (struct iphdr *)skb->data;
231 const int type = icmp_hdr(skb)->type;
232 const int code = icmp_hdr(skb)->code;
233 struct tnl_ptk_info tpi;
235 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IP),
239 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
240 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
241 skb->dev->ifindex, IPPROTO_GRE);
244 if (type == ICMP_REDIRECT) {
245 ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex,
250 ipgre_err(skb, info, &tpi);
253 static bool is_erspan_type1(int gre_hdr_len)
255 /* Both ERSPAN type I (version 0) and type II (version 1) use
256 * protocol 0x88BE, but the type I has only 4-byte GRE header,
257 * while type II has 8-byte.
259 return gre_hdr_len == 4;
262 static int erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
265 struct net *net = dev_net(skb->dev);
266 struct metadata_dst *tun_dst = NULL;
267 struct erspan_base_hdr *ershdr;
268 struct ip_tunnel_net *itn;
269 struct ip_tunnel *tunnel;
270 const struct iphdr *iph;
271 struct erspan_md2 *md2;
275 itn = net_generic(net, erspan_net_id);
277 if (is_erspan_type1(gre_hdr_len)) {
279 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex,
280 tpi->flags | TUNNEL_NO_KEY,
281 iph->saddr, iph->daddr, 0);
283 ershdr = (struct erspan_base_hdr *)(skb->data + gre_hdr_len);
285 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex,
286 tpi->flags | TUNNEL_KEY,
287 iph->saddr, iph->daddr, tpi->key);
291 if (is_erspan_type1(gre_hdr_len))
294 len = gre_hdr_len + erspan_hdr_len(ver);
296 if (unlikely(!pskb_may_pull(skb, len)))
297 return PACKET_REJECT;
299 if (__iptunnel_pull_header(skb,
305 if (tunnel->collect_md) {
306 struct erspan_metadata *pkt_md, *md;
307 struct ip_tunnel_info *info;
312 tpi->flags |= TUNNEL_KEY;
314 tun_id = key32_to_tunnel_id(tpi->key);
316 tun_dst = ip_tun_rx_dst(skb, flags,
317 tun_id, sizeof(*md));
319 return PACKET_REJECT;
321 /* skb can be uncloned in __iptunnel_pull_header, so
322 * old pkt_md is no longer valid and we need to reset
325 gh = skb_network_header(skb) +
326 skb_network_header_len(skb);
327 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
329 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
332 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
335 info = &tun_dst->u.tun_info;
336 info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
337 info->options_len = sizeof(*md);
340 skb_reset_mac_header(skb);
341 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
344 return PACKET_REJECT;
351 static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
352 struct ip_tunnel_net *itn, int hdr_len, bool raw_proto)
354 struct metadata_dst *tun_dst = NULL;
355 const struct iphdr *iph;
356 struct ip_tunnel *tunnel;
359 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
360 iph->saddr, iph->daddr, tpi->key);
363 const struct iphdr *tnl_params;
365 if (__iptunnel_pull_header(skb, hdr_len, tpi->proto,
366 raw_proto, false) < 0)
369 /* Special case for ipgre_header_parse(), which expects the
370 * mac_header to point to the outer IP header.
372 if (tunnel->dev->header_ops == &ipgre_header_ops)
373 skb_pop_mac_header(skb);
375 skb_reset_mac_header(skb);
377 tnl_params = &tunnel->parms.iph;
378 if (tunnel->collect_md || tnl_params->daddr == 0) {
382 flags = tpi->flags & (TUNNEL_CSUM | TUNNEL_KEY);
383 tun_id = key32_to_tunnel_id(tpi->key);
384 tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0);
386 return PACKET_REJECT;
389 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
399 static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
402 struct net *net = dev_net(skb->dev);
403 struct ip_tunnel_net *itn;
406 if (tpi->proto == htons(ETH_P_TEB))
407 itn = net_generic(net, gre_tap_net_id);
409 itn = net_generic(net, ipgre_net_id);
411 res = __ipgre_rcv(skb, tpi, itn, hdr_len, false);
412 if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) {
413 /* ipgre tunnels in collect metadata mode should receive
414 * also ETH_P_TEB traffic.
416 itn = net_generic(net, ipgre_net_id);
417 res = __ipgre_rcv(skb, tpi, itn, hdr_len, true);
422 static int gre_rcv(struct sk_buff *skb)
424 struct tnl_ptk_info tpi;
425 bool csum_err = false;
428 #ifdef CONFIG_NET_IPGRE_BROADCAST
429 if (ipv4_is_multicast(ip_hdr(skb)->daddr)) {
430 /* Looped back packet, drop it! */
431 if (rt_is_output_route(skb_rtable(skb)))
436 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0);
440 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
441 tpi.proto == htons(ETH_P_ERSPAN2))) {
442 if (erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
447 if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
451 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
457 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
458 const struct iphdr *tnl_params,
461 struct ip_tunnel *tunnel = netdev_priv(dev);
462 __be16 flags = tunnel->parms.o_flags;
464 /* Push GRE header. */
465 gre_build_header(skb, tunnel->tun_hlen,
466 flags, proto, tunnel->parms.o_key,
467 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
469 ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
472 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
474 return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
477 static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
480 struct ip_tunnel *tunnel = netdev_priv(dev);
481 struct ip_tunnel_info *tun_info;
482 const struct ip_tunnel_key *key;
486 tun_info = skb_tunnel_info(skb);
487 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
488 ip_tunnel_info_af(tun_info) != AF_INET))
491 key = &tun_info->key;
492 tunnel_hlen = gre_calc_hlen(key->tun_flags);
494 if (skb_cow_head(skb, dev->needed_headroom))
497 /* Push Tunnel header. */
498 if (gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM)))
501 flags = tun_info->key.tun_flags &
502 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
503 gre_build_header(skb, tunnel_hlen, flags, proto,
504 tunnel_id_to_key32(tun_info->key.tun_id),
505 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
507 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
513 dev->stats.tx_dropped++;
516 static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
518 struct ip_tunnel *tunnel = netdev_priv(dev);
519 struct ip_tunnel_info *tun_info;
520 const struct ip_tunnel_key *key;
521 struct erspan_metadata *md;
522 bool truncate = false;
529 tun_info = skb_tunnel_info(skb);
530 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
531 ip_tunnel_info_af(tun_info) != AF_INET))
534 key = &tun_info->key;
535 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
537 if (tun_info->options_len < sizeof(*md))
539 md = ip_tunnel_info_opts(tun_info);
541 /* ERSPAN has fixed 8 byte GRE header */
542 version = md->version;
543 tunnel_hlen = 8 + erspan_hdr_len(version);
545 if (skb_cow_head(skb, dev->needed_headroom))
548 if (gre_handle_offloads(skb, false))
551 if (skb->len > dev->mtu + dev->hard_header_len) {
552 pskb_trim(skb, dev->mtu + dev->hard_header_len);
556 nhoff = skb_network_header(skb) - skb_mac_header(skb);
557 if (skb->protocol == htons(ETH_P_IP) &&
558 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
561 thoff = skb_transport_header(skb) - skb_mac_header(skb);
562 if (skb->protocol == htons(ETH_P_IPV6) &&
563 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
567 erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
568 ntohl(md->u.index), truncate, true);
569 proto = htons(ETH_P_ERSPAN);
570 } else if (version == 2) {
571 erspan_build_header_v2(skb,
572 ntohl(tunnel_id_to_key32(key->tun_id)),
574 get_hwid(&md->u.md2),
576 proto = htons(ETH_P_ERSPAN2);
581 gre_build_header(skb, 8, TUNNEL_SEQ,
582 proto, 0, htonl(atomic_fetch_inc(&tunnel->o_seqno)));
584 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
590 dev->stats.tx_dropped++;
593 static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
595 struct ip_tunnel_info *info = skb_tunnel_info(skb);
596 const struct ip_tunnel_key *key;
600 if (ip_tunnel_info_af(info) != AF_INET)
604 ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
605 tunnel_id_to_key32(key->tun_id),
606 key->tos & ~INET_ECN_MASK, dev_net(dev), 0,
607 skb->mark, skb_get_hash(skb));
608 rt = ip_route_output_key(dev_net(dev), &fl4);
613 info->key.u.ipv4.src = fl4.saddr;
617 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
618 struct net_device *dev)
620 struct ip_tunnel *tunnel = netdev_priv(dev);
621 const struct iphdr *tnl_params;
623 if (!pskb_inet_may_pull(skb))
626 if (tunnel->collect_md) {
627 gre_fb_xmit(skb, dev, skb->protocol);
631 if (dev->header_ops) {
632 const int pull_len = tunnel->hlen + sizeof(struct iphdr);
634 if (skb_cow_head(skb, 0))
637 tnl_params = (const struct iphdr *)skb->data;
639 if (pull_len > skb_transport_offset(skb))
642 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
645 skb_pull(skb, pull_len);
646 skb_reset_mac_header(skb);
648 if (skb_cow_head(skb, dev->needed_headroom))
651 tnl_params = &tunnel->parms.iph;
654 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
657 __gre_xmit(skb, dev, tnl_params, skb->protocol);
662 dev->stats.tx_dropped++;
666 static netdev_tx_t erspan_xmit(struct sk_buff *skb,
667 struct net_device *dev)
669 struct ip_tunnel *tunnel = netdev_priv(dev);
670 bool truncate = false;
673 if (!pskb_inet_may_pull(skb))
676 if (tunnel->collect_md) {
677 erspan_fb_xmit(skb, dev);
681 if (gre_handle_offloads(skb, false))
684 if (skb_cow_head(skb, dev->needed_headroom))
687 if (skb->len > dev->mtu + dev->hard_header_len) {
688 pskb_trim(skb, dev->mtu + dev->hard_header_len);
692 /* Push ERSPAN header */
693 if (tunnel->erspan_ver == 0) {
694 proto = htons(ETH_P_ERSPAN);
695 tunnel->parms.o_flags &= ~TUNNEL_SEQ;
696 } else if (tunnel->erspan_ver == 1) {
697 erspan_build_header(skb, ntohl(tunnel->parms.o_key),
700 proto = htons(ETH_P_ERSPAN);
701 } else if (tunnel->erspan_ver == 2) {
702 erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
703 tunnel->dir, tunnel->hwid,
705 proto = htons(ETH_P_ERSPAN2);
710 tunnel->parms.o_flags &= ~TUNNEL_KEY;
711 __gre_xmit(skb, dev, &tunnel->parms.iph, proto);
716 dev->stats.tx_dropped++;
720 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
721 struct net_device *dev)
723 struct ip_tunnel *tunnel = netdev_priv(dev);
725 if (!pskb_inet_may_pull(skb))
728 if (tunnel->collect_md) {
729 gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
733 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
736 if (skb_cow_head(skb, dev->needed_headroom))
739 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
744 dev->stats.tx_dropped++;
748 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
750 struct ip_tunnel *tunnel = netdev_priv(dev);
754 len = tunnel->tun_hlen;
755 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
756 len = tunnel->tun_hlen - len;
757 tunnel->hlen = tunnel->hlen + len;
760 dev->hard_header_len += len;
762 dev->needed_headroom += len;
765 dev->mtu = max_t(int, dev->mtu - len, 68);
767 flags = tunnel->parms.o_flags;
769 if (flags & TUNNEL_SEQ ||
770 (flags & TUNNEL_CSUM && tunnel->encap.type != TUNNEL_ENCAP_NONE)) {
771 dev->features &= ~NETIF_F_GSO_SOFTWARE;
772 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
774 dev->features |= NETIF_F_GSO_SOFTWARE;
775 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
779 static int ipgre_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p,
784 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
785 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
786 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
787 ((p->i_flags | p->o_flags) & (GRE_VERSION | GRE_ROUTING)))
791 p->i_flags = gre_flags_to_tnl_flags(p->i_flags);
792 p->o_flags = gre_flags_to_tnl_flags(p->o_flags);
794 err = ip_tunnel_ctl(dev, p, cmd);
798 if (cmd == SIOCCHGTUNNEL) {
799 struct ip_tunnel *t = netdev_priv(dev);
801 t->parms.i_flags = p->i_flags;
802 t->parms.o_flags = p->o_flags;
804 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
805 ipgre_link_update(dev, true);
808 p->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
809 p->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
813 /* Nice toy. Unfortunately, useless in real life :-)
814 It allows to construct virtual multiprotocol broadcast "LAN"
815 over the Internet, provided multicast routing is tuned.
818 I have no idea was this bicycle invented before me,
819 so that I had to set ARPHRD_IPGRE to a random value.
820 I have an impression, that Cisco could make something similar,
821 but this feature is apparently missing in IOS<=11.2(8).
823 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
824 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
826 ping -t 255 224.66.66.66
828 If nobody answers, mbone does not work.
830 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
831 ip addr add 10.66.66.<somewhat>/24 dev Universe
833 ifconfig Universe add fe80::<Your_real_addr>/10
834 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
837 ftp fec0:6666:6666::193.233.7.65
840 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
842 const void *daddr, const void *saddr, unsigned int len)
844 struct ip_tunnel *t = netdev_priv(dev);
846 struct gre_base_hdr *greh;
848 iph = skb_push(skb, t->hlen + sizeof(*iph));
849 greh = (struct gre_base_hdr *)(iph+1);
850 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
851 greh->protocol = htons(type);
853 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
855 /* Set the source hardware address. */
857 memcpy(&iph->saddr, saddr, 4);
859 memcpy(&iph->daddr, daddr, 4);
861 return t->hlen + sizeof(*iph);
863 return -(t->hlen + sizeof(*iph));
866 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
868 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
869 memcpy(haddr, &iph->saddr, 4);
873 static const struct header_ops ipgre_header_ops = {
874 .create = ipgre_header,
875 .parse = ipgre_header_parse,
878 #ifdef CONFIG_NET_IPGRE_BROADCAST
879 static int ipgre_open(struct net_device *dev)
881 struct ip_tunnel *t = netdev_priv(dev);
883 if (ipv4_is_multicast(t->parms.iph.daddr)) {
887 rt = ip_route_output_gre(t->net, &fl4,
891 RT_TOS(t->parms.iph.tos),
894 return -EADDRNOTAVAIL;
897 if (!__in_dev_get_rtnl(dev))
898 return -EADDRNOTAVAIL;
899 t->mlink = dev->ifindex;
900 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
905 static int ipgre_close(struct net_device *dev)
907 struct ip_tunnel *t = netdev_priv(dev);
909 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
910 struct in_device *in_dev;
911 in_dev = inetdev_by_index(t->net, t->mlink);
913 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
919 static const struct net_device_ops ipgre_netdev_ops = {
920 .ndo_init = ipgre_tunnel_init,
921 .ndo_uninit = ip_tunnel_uninit,
922 #ifdef CONFIG_NET_IPGRE_BROADCAST
923 .ndo_open = ipgre_open,
924 .ndo_stop = ipgre_close,
926 .ndo_start_xmit = ipgre_xmit,
927 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
928 .ndo_change_mtu = ip_tunnel_change_mtu,
929 .ndo_get_stats64 = dev_get_tstats64,
930 .ndo_get_iflink = ip_tunnel_get_iflink,
931 .ndo_tunnel_ctl = ipgre_tunnel_ctl,
934 #define GRE_FEATURES (NETIF_F_SG | \
939 static void ipgre_tunnel_setup(struct net_device *dev)
941 dev->netdev_ops = &ipgre_netdev_ops;
942 dev->type = ARPHRD_IPGRE;
943 ip_tunnel_setup(dev, ipgre_net_id);
946 static void __gre_tunnel_init(struct net_device *dev)
948 struct ip_tunnel *tunnel;
951 tunnel = netdev_priv(dev);
952 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
953 tunnel->parms.iph.protocol = IPPROTO_GRE;
955 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
956 dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
958 dev->features |= GRE_FEATURES | NETIF_F_LLTX;
959 dev->hw_features |= GRE_FEATURES;
961 flags = tunnel->parms.o_flags;
963 /* TCP offload with GRE SEQ is not supported, nor can we support 2
964 * levels of outer headers requiring an update.
966 if (flags & TUNNEL_SEQ)
968 if (flags & TUNNEL_CSUM && tunnel->encap.type != TUNNEL_ENCAP_NONE)
971 dev->features |= NETIF_F_GSO_SOFTWARE;
972 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
975 static int ipgre_tunnel_init(struct net_device *dev)
977 struct ip_tunnel *tunnel = netdev_priv(dev);
978 struct iphdr *iph = &tunnel->parms.iph;
980 __gre_tunnel_init(dev);
982 __dev_addr_set(dev, &iph->saddr, 4);
983 memcpy(dev->broadcast, &iph->daddr, 4);
985 dev->flags = IFF_NOARP;
989 if (iph->daddr && !tunnel->collect_md) {
990 #ifdef CONFIG_NET_IPGRE_BROADCAST
991 if (ipv4_is_multicast(iph->daddr)) {
994 dev->flags = IFF_BROADCAST;
995 dev->header_ops = &ipgre_header_ops;
996 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
997 dev->needed_headroom = 0;
1000 } else if (!tunnel->collect_md) {
1001 dev->header_ops = &ipgre_header_ops;
1002 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1003 dev->needed_headroom = 0;
1006 return ip_tunnel_init(dev);
1009 static const struct gre_protocol ipgre_protocol = {
1011 .err_handler = gre_err,
1014 static int __net_init ipgre_init_net(struct net *net)
1016 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1019 static void __net_exit ipgre_exit_batch_net(struct list_head *list_net)
1021 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops);
1024 static struct pernet_operations ipgre_net_ops = {
1025 .init = ipgre_init_net,
1026 .exit_batch = ipgre_exit_batch_net,
1027 .id = &ipgre_net_id,
1028 .size = sizeof(struct ip_tunnel_net),
1031 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1032 struct netlink_ext_ack *extack)
1040 if (data[IFLA_GRE_IFLAGS])
1041 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1042 if (data[IFLA_GRE_OFLAGS])
1043 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1044 if (flags & (GRE_VERSION|GRE_ROUTING))
1047 if (data[IFLA_GRE_COLLECT_METADATA] &&
1048 data[IFLA_GRE_ENCAP_TYPE] &&
1049 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1055 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1056 struct netlink_ext_ack *extack)
1060 if (tb[IFLA_ADDRESS]) {
1061 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1063 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1064 return -EADDRNOTAVAIL;
1070 if (data[IFLA_GRE_REMOTE]) {
1071 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1077 return ipgre_tunnel_validate(tb, data, extack);
1080 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1081 struct netlink_ext_ack *extack)
1089 ret = ipgre_tap_validate(tb, data, extack);
1093 if (data[IFLA_GRE_ERSPAN_VER] &&
1094 nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1097 /* ERSPAN type II/III should only have GRE sequence and key flag */
1098 if (data[IFLA_GRE_OFLAGS])
1099 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1100 if (data[IFLA_GRE_IFLAGS])
1101 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1102 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1103 flags != (GRE_SEQ | GRE_KEY))
1106 /* ERSPAN Session ID only has 10-bit. Since we reuse
1107 * 32-bit key field as ID, check it's range.
1109 if (data[IFLA_GRE_IKEY] &&
1110 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1113 if (data[IFLA_GRE_OKEY] &&
1114 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1120 static int ipgre_netlink_parms(struct net_device *dev,
1121 struct nlattr *data[],
1122 struct nlattr *tb[],
1123 struct ip_tunnel_parm *parms,
1126 struct ip_tunnel *t = netdev_priv(dev);
1128 memset(parms, 0, sizeof(*parms));
1130 parms->iph.protocol = IPPROTO_GRE;
1135 if (data[IFLA_GRE_LINK])
1136 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1138 if (data[IFLA_GRE_IFLAGS])
1139 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
1141 if (data[IFLA_GRE_OFLAGS])
1142 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
1144 if (data[IFLA_GRE_IKEY])
1145 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1147 if (data[IFLA_GRE_OKEY])
1148 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1150 if (data[IFLA_GRE_LOCAL])
1151 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1153 if (data[IFLA_GRE_REMOTE])
1154 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1156 if (data[IFLA_GRE_TTL])
1157 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1159 if (data[IFLA_GRE_TOS])
1160 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1162 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1165 parms->iph.frag_off = htons(IP_DF);
1168 if (data[IFLA_GRE_COLLECT_METADATA]) {
1169 t->collect_md = true;
1170 if (dev->type == ARPHRD_IPGRE)
1171 dev->type = ARPHRD_NONE;
1174 if (data[IFLA_GRE_IGNORE_DF]) {
1175 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1176 && (parms->iph.frag_off & htons(IP_DF)))
1178 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1181 if (data[IFLA_GRE_FWMARK])
1182 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1187 static int erspan_netlink_parms(struct net_device *dev,
1188 struct nlattr *data[],
1189 struct nlattr *tb[],
1190 struct ip_tunnel_parm *parms,
1193 struct ip_tunnel *t = netdev_priv(dev);
1196 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1202 if (data[IFLA_GRE_ERSPAN_VER]) {
1203 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1205 if (t->erspan_ver > 2)
1209 if (t->erspan_ver == 1) {
1210 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1211 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1212 if (t->index & ~INDEX_MASK)
1215 } else if (t->erspan_ver == 2) {
1216 if (data[IFLA_GRE_ERSPAN_DIR]) {
1217 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1218 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1221 if (data[IFLA_GRE_ERSPAN_HWID]) {
1222 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1223 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1231 /* This function returns true when ENCAP attributes are present in the nl msg */
1232 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1233 struct ip_tunnel_encap *ipencap)
1237 memset(ipencap, 0, sizeof(*ipencap));
1242 if (data[IFLA_GRE_ENCAP_TYPE]) {
1244 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1247 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1249 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1252 if (data[IFLA_GRE_ENCAP_SPORT]) {
1254 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1257 if (data[IFLA_GRE_ENCAP_DPORT]) {
1259 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1265 static int gre_tap_init(struct net_device *dev)
1267 __gre_tunnel_init(dev);
1268 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1269 netif_keep_dst(dev);
1271 return ip_tunnel_init(dev);
1274 static const struct net_device_ops gre_tap_netdev_ops = {
1275 .ndo_init = gre_tap_init,
1276 .ndo_uninit = ip_tunnel_uninit,
1277 .ndo_start_xmit = gre_tap_xmit,
1278 .ndo_set_mac_address = eth_mac_addr,
1279 .ndo_validate_addr = eth_validate_addr,
1280 .ndo_change_mtu = ip_tunnel_change_mtu,
1281 .ndo_get_stats64 = dev_get_tstats64,
1282 .ndo_get_iflink = ip_tunnel_get_iflink,
1283 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1286 static int erspan_tunnel_init(struct net_device *dev)
1288 struct ip_tunnel *tunnel = netdev_priv(dev);
1290 if (tunnel->erspan_ver == 0)
1291 tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1293 tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1295 tunnel->parms.iph.protocol = IPPROTO_GRE;
1296 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1297 erspan_hdr_len(tunnel->erspan_ver);
1299 dev->features |= GRE_FEATURES;
1300 dev->hw_features |= GRE_FEATURES;
1301 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1302 netif_keep_dst(dev);
1304 return ip_tunnel_init(dev);
1307 static const struct net_device_ops erspan_netdev_ops = {
1308 .ndo_init = erspan_tunnel_init,
1309 .ndo_uninit = ip_tunnel_uninit,
1310 .ndo_start_xmit = erspan_xmit,
1311 .ndo_set_mac_address = eth_mac_addr,
1312 .ndo_validate_addr = eth_validate_addr,
1313 .ndo_change_mtu = ip_tunnel_change_mtu,
1314 .ndo_get_stats64 = dev_get_tstats64,
1315 .ndo_get_iflink = ip_tunnel_get_iflink,
1316 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1319 static void ipgre_tap_setup(struct net_device *dev)
1323 dev->netdev_ops = &gre_tap_netdev_ops;
1324 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1325 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1326 ip_tunnel_setup(dev, gre_tap_net_id);
1330 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1332 struct ip_tunnel_encap ipencap;
1334 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1335 struct ip_tunnel *t = netdev_priv(dev);
1336 int err = ip_tunnel_encap_setup(t, &ipencap);
1345 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1346 struct nlattr *tb[], struct nlattr *data[],
1347 struct netlink_ext_ack *extack)
1349 struct ip_tunnel_parm p;
1353 err = ipgre_newlink_encap_setup(dev, data);
1357 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1360 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1363 static int erspan_newlink(struct net *src_net, struct net_device *dev,
1364 struct nlattr *tb[], struct nlattr *data[],
1365 struct netlink_ext_ack *extack)
1367 struct ip_tunnel_parm p;
1371 err = ipgre_newlink_encap_setup(dev, data);
1375 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1378 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1381 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1382 struct nlattr *data[],
1383 struct netlink_ext_ack *extack)
1385 struct ip_tunnel *t = netdev_priv(dev);
1386 __u32 fwmark = t->fwmark;
1387 struct ip_tunnel_parm p;
1390 err = ipgre_newlink_encap_setup(dev, data);
1394 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1398 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1402 t->parms.i_flags = p.i_flags;
1403 t->parms.o_flags = p.o_flags;
1405 ipgre_link_update(dev, !tb[IFLA_MTU]);
1410 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1411 struct nlattr *data[],
1412 struct netlink_ext_ack *extack)
1414 struct ip_tunnel *t = netdev_priv(dev);
1415 __u32 fwmark = t->fwmark;
1416 struct ip_tunnel_parm p;
1419 err = ipgre_newlink_encap_setup(dev, data);
1423 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1427 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1431 t->parms.i_flags = p.i_flags;
1432 t->parms.o_flags = p.o_flags;
1437 static size_t ipgre_get_size(const struct net_device *dev)
1442 /* IFLA_GRE_IFLAGS */
1444 /* IFLA_GRE_OFLAGS */
1450 /* IFLA_GRE_LOCAL */
1452 /* IFLA_GRE_REMOTE */
1458 /* IFLA_GRE_PMTUDISC */
1460 /* IFLA_GRE_ENCAP_TYPE */
1462 /* IFLA_GRE_ENCAP_FLAGS */
1464 /* IFLA_GRE_ENCAP_SPORT */
1466 /* IFLA_GRE_ENCAP_DPORT */
1468 /* IFLA_GRE_COLLECT_METADATA */
1470 /* IFLA_GRE_IGNORE_DF */
1472 /* IFLA_GRE_FWMARK */
1474 /* IFLA_GRE_ERSPAN_INDEX */
1476 /* IFLA_GRE_ERSPAN_VER */
1478 /* IFLA_GRE_ERSPAN_DIR */
1480 /* IFLA_GRE_ERSPAN_HWID */
1485 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1487 struct ip_tunnel *t = netdev_priv(dev);
1488 struct ip_tunnel_parm *p = &t->parms;
1489 __be16 o_flags = p->o_flags;
1491 if (t->erspan_ver <= 2) {
1492 if (t->erspan_ver != 0 && !t->collect_md)
1493 o_flags |= TUNNEL_KEY;
1495 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1496 goto nla_put_failure;
1498 if (t->erspan_ver == 1) {
1499 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1500 goto nla_put_failure;
1501 } else if (t->erspan_ver == 2) {
1502 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1503 goto nla_put_failure;
1504 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1505 goto nla_put_failure;
1509 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1510 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1511 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1512 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1513 gre_tnl_flags_to_gre_flags(o_flags)) ||
1514 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1515 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1516 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1517 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1518 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1519 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1520 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1521 !!(p->iph.frag_off & htons(IP_DF))) ||
1522 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1523 goto nla_put_failure;
1525 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1527 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1529 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1531 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1533 goto nla_put_failure;
1535 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1536 goto nla_put_failure;
1538 if (t->collect_md) {
1539 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1540 goto nla_put_failure;
1549 static void erspan_setup(struct net_device *dev)
1551 struct ip_tunnel *t = netdev_priv(dev);
1555 dev->netdev_ops = &erspan_netdev_ops;
1556 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1557 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1558 ip_tunnel_setup(dev, erspan_net_id);
1562 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1563 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1564 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1565 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1566 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1567 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1568 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
1569 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
1570 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1571 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1572 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1573 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1574 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1575 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1576 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1577 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1578 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1579 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1580 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1581 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1582 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1583 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1586 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1588 .maxtype = IFLA_GRE_MAX,
1589 .policy = ipgre_policy,
1590 .priv_size = sizeof(struct ip_tunnel),
1591 .setup = ipgre_tunnel_setup,
1592 .validate = ipgre_tunnel_validate,
1593 .newlink = ipgre_newlink,
1594 .changelink = ipgre_changelink,
1595 .dellink = ip_tunnel_dellink,
1596 .get_size = ipgre_get_size,
1597 .fill_info = ipgre_fill_info,
1598 .get_link_net = ip_tunnel_get_link_net,
1601 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1603 .maxtype = IFLA_GRE_MAX,
1604 .policy = ipgre_policy,
1605 .priv_size = sizeof(struct ip_tunnel),
1606 .setup = ipgre_tap_setup,
1607 .validate = ipgre_tap_validate,
1608 .newlink = ipgre_newlink,
1609 .changelink = ipgre_changelink,
1610 .dellink = ip_tunnel_dellink,
1611 .get_size = ipgre_get_size,
1612 .fill_info = ipgre_fill_info,
1613 .get_link_net = ip_tunnel_get_link_net,
1616 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1618 .maxtype = IFLA_GRE_MAX,
1619 .policy = ipgre_policy,
1620 .priv_size = sizeof(struct ip_tunnel),
1621 .setup = erspan_setup,
1622 .validate = erspan_validate,
1623 .newlink = erspan_newlink,
1624 .changelink = erspan_changelink,
1625 .dellink = ip_tunnel_dellink,
1626 .get_size = ipgre_get_size,
1627 .fill_info = ipgre_fill_info,
1628 .get_link_net = ip_tunnel_get_link_net,
1631 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1632 u8 name_assign_type)
1634 struct nlattr *tb[IFLA_MAX + 1];
1635 struct net_device *dev;
1636 LIST_HEAD(list_kill);
1637 struct ip_tunnel *t;
1640 memset(&tb, 0, sizeof(tb));
1642 dev = rtnl_create_link(net, name, name_assign_type,
1643 &ipgre_tap_ops, tb, NULL);
1647 /* Configure flow based GRE device. */
1648 t = netdev_priv(dev);
1649 t->collect_md = true;
1651 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1654 return ERR_PTR(err);
1657 /* openvswitch users expect packet sizes to be unrestricted,
1658 * so set the largest MTU we can.
1660 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1664 err = rtnl_configure_link(dev, NULL);
1670 ip_tunnel_dellink(dev, &list_kill);
1671 unregister_netdevice_many(&list_kill);
1672 return ERR_PTR(err);
1674 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1676 static int __net_init ipgre_tap_init_net(struct net *net)
1678 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1681 static void __net_exit ipgre_tap_exit_batch_net(struct list_head *list_net)
1683 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops);
1686 static struct pernet_operations ipgre_tap_net_ops = {
1687 .init = ipgre_tap_init_net,
1688 .exit_batch = ipgre_tap_exit_batch_net,
1689 .id = &gre_tap_net_id,
1690 .size = sizeof(struct ip_tunnel_net),
1693 static int __net_init erspan_init_net(struct net *net)
1695 return ip_tunnel_init_net(net, erspan_net_id,
1696 &erspan_link_ops, "erspan0");
1699 static void __net_exit erspan_exit_batch_net(struct list_head *net_list)
1701 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops);
1704 static struct pernet_operations erspan_net_ops = {
1705 .init = erspan_init_net,
1706 .exit_batch = erspan_exit_batch_net,
1707 .id = &erspan_net_id,
1708 .size = sizeof(struct ip_tunnel_net),
1711 static int __init ipgre_init(void)
1715 pr_info("GRE over IPv4 tunneling driver\n");
1717 err = register_pernet_device(&ipgre_net_ops);
1721 err = register_pernet_device(&ipgre_tap_net_ops);
1723 goto pnet_tap_failed;
1725 err = register_pernet_device(&erspan_net_ops);
1727 goto pnet_erspan_failed;
1729 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1731 pr_info("%s: can't add protocol\n", __func__);
1732 goto add_proto_failed;
1735 err = rtnl_link_register(&ipgre_link_ops);
1737 goto rtnl_link_failed;
1739 err = rtnl_link_register(&ipgre_tap_ops);
1741 goto tap_ops_failed;
1743 err = rtnl_link_register(&erspan_link_ops);
1745 goto erspan_link_failed;
1750 rtnl_link_unregister(&ipgre_tap_ops);
1752 rtnl_link_unregister(&ipgre_link_ops);
1754 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1756 unregister_pernet_device(&erspan_net_ops);
1758 unregister_pernet_device(&ipgre_tap_net_ops);
1760 unregister_pernet_device(&ipgre_net_ops);
1764 static void __exit ipgre_fini(void)
1766 rtnl_link_unregister(&ipgre_tap_ops);
1767 rtnl_link_unregister(&ipgre_link_ops);
1768 rtnl_link_unregister(&erspan_link_ops);
1769 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1770 unregister_pernet_device(&ipgre_tap_net_ops);
1771 unregister_pernet_device(&ipgre_net_ops);
1772 unregister_pernet_device(&erspan_net_ops);
1775 module_init(ipgre_init);
1776 module_exit(ipgre_fini);
1777 MODULE_LICENSE("GPL");
1778 MODULE_ALIAS_RTNL_LINK("gre");
1779 MODULE_ALIAS_RTNL_LINK("gretap");
1780 MODULE_ALIAS_RTNL_LINK("erspan");
1781 MODULE_ALIAS_NETDEV("gre0");
1782 MODULE_ALIAS_NETDEV("gretap0");
1783 MODULE_ALIAS_NETDEV("erspan0");