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);
753 len = tunnel->tun_hlen;
754 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
755 len = tunnel->tun_hlen - len;
756 tunnel->hlen = tunnel->hlen + len;
759 dev->hard_header_len += len;
761 dev->needed_headroom += len;
764 dev->mtu = max_t(int, dev->mtu - len, 68);
766 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
767 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
768 tunnel->encap.type == TUNNEL_ENCAP_NONE) {
769 dev->features |= NETIF_F_GSO_SOFTWARE;
770 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
772 dev->features &= ~NETIF_F_GSO_SOFTWARE;
773 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
775 dev->features |= NETIF_F_LLTX;
777 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
778 dev->features &= ~(NETIF_F_LLTX | NETIF_F_GSO_SOFTWARE);
782 static int ipgre_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p,
787 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
788 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
789 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
790 ((p->i_flags | p->o_flags) & (GRE_VERSION | GRE_ROUTING)))
794 p->i_flags = gre_flags_to_tnl_flags(p->i_flags);
795 p->o_flags = gre_flags_to_tnl_flags(p->o_flags);
797 err = ip_tunnel_ctl(dev, p, cmd);
801 if (cmd == SIOCCHGTUNNEL) {
802 struct ip_tunnel *t = netdev_priv(dev);
804 t->parms.i_flags = p->i_flags;
805 t->parms.o_flags = p->o_flags;
807 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
808 ipgre_link_update(dev, true);
811 p->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
812 p->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
816 /* Nice toy. Unfortunately, useless in real life :-)
817 It allows to construct virtual multiprotocol broadcast "LAN"
818 over the Internet, provided multicast routing is tuned.
821 I have no idea was this bicycle invented before me,
822 so that I had to set ARPHRD_IPGRE to a random value.
823 I have an impression, that Cisco could make something similar,
824 but this feature is apparently missing in IOS<=11.2(8).
826 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
827 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
829 ping -t 255 224.66.66.66
831 If nobody answers, mbone does not work.
833 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
834 ip addr add 10.66.66.<somewhat>/24 dev Universe
836 ifconfig Universe add fe80::<Your_real_addr>/10
837 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
840 ftp fec0:6666:6666::193.233.7.65
843 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
845 const void *daddr, const void *saddr, unsigned int len)
847 struct ip_tunnel *t = netdev_priv(dev);
849 struct gre_base_hdr *greh;
851 iph = skb_push(skb, t->hlen + sizeof(*iph));
852 greh = (struct gre_base_hdr *)(iph+1);
853 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
854 greh->protocol = htons(type);
856 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
858 /* Set the source hardware address. */
860 memcpy(&iph->saddr, saddr, 4);
862 memcpy(&iph->daddr, daddr, 4);
864 return t->hlen + sizeof(*iph);
866 return -(t->hlen + sizeof(*iph));
869 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
871 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
872 memcpy(haddr, &iph->saddr, 4);
876 static const struct header_ops ipgre_header_ops = {
877 .create = ipgre_header,
878 .parse = ipgre_header_parse,
881 #ifdef CONFIG_NET_IPGRE_BROADCAST
882 static int ipgre_open(struct net_device *dev)
884 struct ip_tunnel *t = netdev_priv(dev);
886 if (ipv4_is_multicast(t->parms.iph.daddr)) {
890 rt = ip_route_output_gre(t->net, &fl4,
894 RT_TOS(t->parms.iph.tos),
897 return -EADDRNOTAVAIL;
900 if (!__in_dev_get_rtnl(dev))
901 return -EADDRNOTAVAIL;
902 t->mlink = dev->ifindex;
903 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
908 static int ipgre_close(struct net_device *dev)
910 struct ip_tunnel *t = netdev_priv(dev);
912 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
913 struct in_device *in_dev;
914 in_dev = inetdev_by_index(t->net, t->mlink);
916 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
922 static const struct net_device_ops ipgre_netdev_ops = {
923 .ndo_init = ipgre_tunnel_init,
924 .ndo_uninit = ip_tunnel_uninit,
925 #ifdef CONFIG_NET_IPGRE_BROADCAST
926 .ndo_open = ipgre_open,
927 .ndo_stop = ipgre_close,
929 .ndo_start_xmit = ipgre_xmit,
930 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
931 .ndo_change_mtu = ip_tunnel_change_mtu,
932 .ndo_get_stats64 = dev_get_tstats64,
933 .ndo_get_iflink = ip_tunnel_get_iflink,
934 .ndo_tunnel_ctl = ipgre_tunnel_ctl,
937 #define GRE_FEATURES (NETIF_F_SG | \
942 static void ipgre_tunnel_setup(struct net_device *dev)
944 dev->netdev_ops = &ipgre_netdev_ops;
945 dev->type = ARPHRD_IPGRE;
946 ip_tunnel_setup(dev, ipgre_net_id);
949 static void __gre_tunnel_init(struct net_device *dev)
951 struct ip_tunnel *tunnel;
953 tunnel = netdev_priv(dev);
954 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
955 tunnel->parms.iph.protocol = IPPROTO_GRE;
957 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
958 dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
960 dev->features |= GRE_FEATURES;
961 dev->hw_features |= GRE_FEATURES;
963 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
964 /* TCP offload with GRE SEQ is not supported, nor
965 * can we support 2 levels of outer headers requiring
968 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
969 (tunnel->encap.type == TUNNEL_ENCAP_NONE)) {
970 dev->features |= NETIF_F_GSO_SOFTWARE;
971 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
974 /* Can use a lockless transmit, unless we generate
977 dev->features |= NETIF_F_LLTX;
981 static int ipgre_tunnel_init(struct net_device *dev)
983 struct ip_tunnel *tunnel = netdev_priv(dev);
984 struct iphdr *iph = &tunnel->parms.iph;
986 __gre_tunnel_init(dev);
988 __dev_addr_set(dev, &iph->saddr, 4);
989 memcpy(dev->broadcast, &iph->daddr, 4);
991 dev->flags = IFF_NOARP;
995 if (iph->daddr && !tunnel->collect_md) {
996 #ifdef CONFIG_NET_IPGRE_BROADCAST
997 if (ipv4_is_multicast(iph->daddr)) {
1000 dev->flags = IFF_BROADCAST;
1001 dev->header_ops = &ipgre_header_ops;
1002 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1003 dev->needed_headroom = 0;
1006 } else if (!tunnel->collect_md) {
1007 dev->header_ops = &ipgre_header_ops;
1008 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1009 dev->needed_headroom = 0;
1012 return ip_tunnel_init(dev);
1015 static const struct gre_protocol ipgre_protocol = {
1017 .err_handler = gre_err,
1020 static int __net_init ipgre_init_net(struct net *net)
1022 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1025 static void __net_exit ipgre_exit_batch_net(struct list_head *list_net)
1027 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops);
1030 static struct pernet_operations ipgre_net_ops = {
1031 .init = ipgre_init_net,
1032 .exit_batch = ipgre_exit_batch_net,
1033 .id = &ipgre_net_id,
1034 .size = sizeof(struct ip_tunnel_net),
1037 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1038 struct netlink_ext_ack *extack)
1046 if (data[IFLA_GRE_IFLAGS])
1047 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1048 if (data[IFLA_GRE_OFLAGS])
1049 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1050 if (flags & (GRE_VERSION|GRE_ROUTING))
1053 if (data[IFLA_GRE_COLLECT_METADATA] &&
1054 data[IFLA_GRE_ENCAP_TYPE] &&
1055 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1061 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1062 struct netlink_ext_ack *extack)
1066 if (tb[IFLA_ADDRESS]) {
1067 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1069 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1070 return -EADDRNOTAVAIL;
1076 if (data[IFLA_GRE_REMOTE]) {
1077 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1083 return ipgre_tunnel_validate(tb, data, extack);
1086 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1087 struct netlink_ext_ack *extack)
1095 ret = ipgre_tap_validate(tb, data, extack);
1099 if (data[IFLA_GRE_ERSPAN_VER] &&
1100 nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1103 /* ERSPAN type II/III should only have GRE sequence and key flag */
1104 if (data[IFLA_GRE_OFLAGS])
1105 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1106 if (data[IFLA_GRE_IFLAGS])
1107 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1108 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1109 flags != (GRE_SEQ | GRE_KEY))
1112 /* ERSPAN Session ID only has 10-bit. Since we reuse
1113 * 32-bit key field as ID, check it's range.
1115 if (data[IFLA_GRE_IKEY] &&
1116 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1119 if (data[IFLA_GRE_OKEY] &&
1120 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1126 static int ipgre_netlink_parms(struct net_device *dev,
1127 struct nlattr *data[],
1128 struct nlattr *tb[],
1129 struct ip_tunnel_parm *parms,
1132 struct ip_tunnel *t = netdev_priv(dev);
1134 memset(parms, 0, sizeof(*parms));
1136 parms->iph.protocol = IPPROTO_GRE;
1141 if (data[IFLA_GRE_LINK])
1142 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1144 if (data[IFLA_GRE_IFLAGS])
1145 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
1147 if (data[IFLA_GRE_OFLAGS])
1148 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
1150 if (data[IFLA_GRE_IKEY])
1151 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1153 if (data[IFLA_GRE_OKEY])
1154 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1156 if (data[IFLA_GRE_LOCAL])
1157 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1159 if (data[IFLA_GRE_REMOTE])
1160 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1162 if (data[IFLA_GRE_TTL])
1163 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1165 if (data[IFLA_GRE_TOS])
1166 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1168 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1171 parms->iph.frag_off = htons(IP_DF);
1174 if (data[IFLA_GRE_COLLECT_METADATA]) {
1175 t->collect_md = true;
1176 if (dev->type == ARPHRD_IPGRE)
1177 dev->type = ARPHRD_NONE;
1180 if (data[IFLA_GRE_IGNORE_DF]) {
1181 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1182 && (parms->iph.frag_off & htons(IP_DF)))
1184 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1187 if (data[IFLA_GRE_FWMARK])
1188 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1193 static int erspan_netlink_parms(struct net_device *dev,
1194 struct nlattr *data[],
1195 struct nlattr *tb[],
1196 struct ip_tunnel_parm *parms,
1199 struct ip_tunnel *t = netdev_priv(dev);
1202 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1208 if (data[IFLA_GRE_ERSPAN_VER]) {
1209 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1211 if (t->erspan_ver > 2)
1215 if (t->erspan_ver == 1) {
1216 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1217 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1218 if (t->index & ~INDEX_MASK)
1221 } else if (t->erspan_ver == 2) {
1222 if (data[IFLA_GRE_ERSPAN_DIR]) {
1223 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1224 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1227 if (data[IFLA_GRE_ERSPAN_HWID]) {
1228 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1229 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1237 /* This function returns true when ENCAP attributes are present in the nl msg */
1238 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1239 struct ip_tunnel_encap *ipencap)
1243 memset(ipencap, 0, sizeof(*ipencap));
1248 if (data[IFLA_GRE_ENCAP_TYPE]) {
1250 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1253 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1255 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1258 if (data[IFLA_GRE_ENCAP_SPORT]) {
1260 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1263 if (data[IFLA_GRE_ENCAP_DPORT]) {
1265 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1271 static int gre_tap_init(struct net_device *dev)
1273 __gre_tunnel_init(dev);
1274 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1275 netif_keep_dst(dev);
1277 return ip_tunnel_init(dev);
1280 static const struct net_device_ops gre_tap_netdev_ops = {
1281 .ndo_init = gre_tap_init,
1282 .ndo_uninit = ip_tunnel_uninit,
1283 .ndo_start_xmit = gre_tap_xmit,
1284 .ndo_set_mac_address = eth_mac_addr,
1285 .ndo_validate_addr = eth_validate_addr,
1286 .ndo_change_mtu = ip_tunnel_change_mtu,
1287 .ndo_get_stats64 = dev_get_tstats64,
1288 .ndo_get_iflink = ip_tunnel_get_iflink,
1289 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1292 static int erspan_tunnel_init(struct net_device *dev)
1294 struct ip_tunnel *tunnel = netdev_priv(dev);
1296 if (tunnel->erspan_ver == 0)
1297 tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1299 tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1301 tunnel->parms.iph.protocol = IPPROTO_GRE;
1302 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1303 erspan_hdr_len(tunnel->erspan_ver);
1305 dev->features |= GRE_FEATURES;
1306 dev->hw_features |= GRE_FEATURES;
1307 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1308 netif_keep_dst(dev);
1310 return ip_tunnel_init(dev);
1313 static const struct net_device_ops erspan_netdev_ops = {
1314 .ndo_init = erspan_tunnel_init,
1315 .ndo_uninit = ip_tunnel_uninit,
1316 .ndo_start_xmit = erspan_xmit,
1317 .ndo_set_mac_address = eth_mac_addr,
1318 .ndo_validate_addr = eth_validate_addr,
1319 .ndo_change_mtu = ip_tunnel_change_mtu,
1320 .ndo_get_stats64 = dev_get_tstats64,
1321 .ndo_get_iflink = ip_tunnel_get_iflink,
1322 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1325 static void ipgre_tap_setup(struct net_device *dev)
1329 dev->netdev_ops = &gre_tap_netdev_ops;
1330 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1331 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1332 ip_tunnel_setup(dev, gre_tap_net_id);
1336 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1338 struct ip_tunnel_encap ipencap;
1340 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1341 struct ip_tunnel *t = netdev_priv(dev);
1342 int err = ip_tunnel_encap_setup(t, &ipencap);
1351 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1352 struct nlattr *tb[], struct nlattr *data[],
1353 struct netlink_ext_ack *extack)
1355 struct ip_tunnel_parm p;
1359 err = ipgre_newlink_encap_setup(dev, data);
1363 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1366 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1369 static int erspan_newlink(struct net *src_net, struct net_device *dev,
1370 struct nlattr *tb[], struct nlattr *data[],
1371 struct netlink_ext_ack *extack)
1373 struct ip_tunnel_parm p;
1377 err = ipgre_newlink_encap_setup(dev, data);
1381 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1384 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1387 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1388 struct nlattr *data[],
1389 struct netlink_ext_ack *extack)
1391 struct ip_tunnel *t = netdev_priv(dev);
1392 __u32 fwmark = t->fwmark;
1393 struct ip_tunnel_parm p;
1396 err = ipgre_newlink_encap_setup(dev, data);
1400 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1404 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1408 t->parms.i_flags = p.i_flags;
1409 t->parms.o_flags = p.o_flags;
1411 ipgre_link_update(dev, !tb[IFLA_MTU]);
1416 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1417 struct nlattr *data[],
1418 struct netlink_ext_ack *extack)
1420 struct ip_tunnel *t = netdev_priv(dev);
1421 __u32 fwmark = t->fwmark;
1422 struct ip_tunnel_parm p;
1425 err = ipgre_newlink_encap_setup(dev, data);
1429 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1433 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1437 t->parms.i_flags = p.i_flags;
1438 t->parms.o_flags = p.o_flags;
1443 static size_t ipgre_get_size(const struct net_device *dev)
1448 /* IFLA_GRE_IFLAGS */
1450 /* IFLA_GRE_OFLAGS */
1456 /* IFLA_GRE_LOCAL */
1458 /* IFLA_GRE_REMOTE */
1464 /* IFLA_GRE_PMTUDISC */
1466 /* IFLA_GRE_ENCAP_TYPE */
1468 /* IFLA_GRE_ENCAP_FLAGS */
1470 /* IFLA_GRE_ENCAP_SPORT */
1472 /* IFLA_GRE_ENCAP_DPORT */
1474 /* IFLA_GRE_COLLECT_METADATA */
1476 /* IFLA_GRE_IGNORE_DF */
1478 /* IFLA_GRE_FWMARK */
1480 /* IFLA_GRE_ERSPAN_INDEX */
1482 /* IFLA_GRE_ERSPAN_VER */
1484 /* IFLA_GRE_ERSPAN_DIR */
1486 /* IFLA_GRE_ERSPAN_HWID */
1491 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1493 struct ip_tunnel *t = netdev_priv(dev);
1494 struct ip_tunnel_parm *p = &t->parms;
1495 __be16 o_flags = p->o_flags;
1497 if (t->erspan_ver <= 2) {
1498 if (t->erspan_ver != 0 && !t->collect_md)
1499 o_flags |= TUNNEL_KEY;
1501 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1502 goto nla_put_failure;
1504 if (t->erspan_ver == 1) {
1505 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1506 goto nla_put_failure;
1507 } else if (t->erspan_ver == 2) {
1508 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1509 goto nla_put_failure;
1510 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1511 goto nla_put_failure;
1515 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1516 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1517 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1518 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1519 gre_tnl_flags_to_gre_flags(o_flags)) ||
1520 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1521 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1522 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1523 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1524 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1525 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1526 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1527 !!(p->iph.frag_off & htons(IP_DF))) ||
1528 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1529 goto nla_put_failure;
1531 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1533 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1535 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1537 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1539 goto nla_put_failure;
1541 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1542 goto nla_put_failure;
1544 if (t->collect_md) {
1545 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1546 goto nla_put_failure;
1555 static void erspan_setup(struct net_device *dev)
1557 struct ip_tunnel *t = netdev_priv(dev);
1561 dev->netdev_ops = &erspan_netdev_ops;
1562 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1563 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1564 ip_tunnel_setup(dev, erspan_net_id);
1568 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1569 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1570 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1571 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1572 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1573 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1574 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
1575 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
1576 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1577 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1578 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1579 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1580 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1581 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1582 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1583 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1584 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1585 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1586 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1587 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1588 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1589 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1592 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1594 .maxtype = IFLA_GRE_MAX,
1595 .policy = ipgre_policy,
1596 .priv_size = sizeof(struct ip_tunnel),
1597 .setup = ipgre_tunnel_setup,
1598 .validate = ipgre_tunnel_validate,
1599 .newlink = ipgre_newlink,
1600 .changelink = ipgre_changelink,
1601 .dellink = ip_tunnel_dellink,
1602 .get_size = ipgre_get_size,
1603 .fill_info = ipgre_fill_info,
1604 .get_link_net = ip_tunnel_get_link_net,
1607 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1609 .maxtype = IFLA_GRE_MAX,
1610 .policy = ipgre_policy,
1611 .priv_size = sizeof(struct ip_tunnel),
1612 .setup = ipgre_tap_setup,
1613 .validate = ipgre_tap_validate,
1614 .newlink = ipgre_newlink,
1615 .changelink = ipgre_changelink,
1616 .dellink = ip_tunnel_dellink,
1617 .get_size = ipgre_get_size,
1618 .fill_info = ipgre_fill_info,
1619 .get_link_net = ip_tunnel_get_link_net,
1622 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1624 .maxtype = IFLA_GRE_MAX,
1625 .policy = ipgre_policy,
1626 .priv_size = sizeof(struct ip_tunnel),
1627 .setup = erspan_setup,
1628 .validate = erspan_validate,
1629 .newlink = erspan_newlink,
1630 .changelink = erspan_changelink,
1631 .dellink = ip_tunnel_dellink,
1632 .get_size = ipgre_get_size,
1633 .fill_info = ipgre_fill_info,
1634 .get_link_net = ip_tunnel_get_link_net,
1637 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1638 u8 name_assign_type)
1640 struct nlattr *tb[IFLA_MAX + 1];
1641 struct net_device *dev;
1642 LIST_HEAD(list_kill);
1643 struct ip_tunnel *t;
1646 memset(&tb, 0, sizeof(tb));
1648 dev = rtnl_create_link(net, name, name_assign_type,
1649 &ipgre_tap_ops, tb, NULL);
1653 /* Configure flow based GRE device. */
1654 t = netdev_priv(dev);
1655 t->collect_md = true;
1657 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1660 return ERR_PTR(err);
1663 /* openvswitch users expect packet sizes to be unrestricted,
1664 * so set the largest MTU we can.
1666 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1670 err = rtnl_configure_link(dev, NULL);
1676 ip_tunnel_dellink(dev, &list_kill);
1677 unregister_netdevice_many(&list_kill);
1678 return ERR_PTR(err);
1680 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1682 static int __net_init ipgre_tap_init_net(struct net *net)
1684 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1687 static void __net_exit ipgre_tap_exit_batch_net(struct list_head *list_net)
1689 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops);
1692 static struct pernet_operations ipgre_tap_net_ops = {
1693 .init = ipgre_tap_init_net,
1694 .exit_batch = ipgre_tap_exit_batch_net,
1695 .id = &gre_tap_net_id,
1696 .size = sizeof(struct ip_tunnel_net),
1699 static int __net_init erspan_init_net(struct net *net)
1701 return ip_tunnel_init_net(net, erspan_net_id,
1702 &erspan_link_ops, "erspan0");
1705 static void __net_exit erspan_exit_batch_net(struct list_head *net_list)
1707 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops);
1710 static struct pernet_operations erspan_net_ops = {
1711 .init = erspan_init_net,
1712 .exit_batch = erspan_exit_batch_net,
1713 .id = &erspan_net_id,
1714 .size = sizeof(struct ip_tunnel_net),
1717 static int __init ipgre_init(void)
1721 pr_info("GRE over IPv4 tunneling driver\n");
1723 err = register_pernet_device(&ipgre_net_ops);
1727 err = register_pernet_device(&ipgre_tap_net_ops);
1729 goto pnet_tap_failed;
1731 err = register_pernet_device(&erspan_net_ops);
1733 goto pnet_erspan_failed;
1735 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1737 pr_info("%s: can't add protocol\n", __func__);
1738 goto add_proto_failed;
1741 err = rtnl_link_register(&ipgre_link_ops);
1743 goto rtnl_link_failed;
1745 err = rtnl_link_register(&ipgre_tap_ops);
1747 goto tap_ops_failed;
1749 err = rtnl_link_register(&erspan_link_ops);
1751 goto erspan_link_failed;
1756 rtnl_link_unregister(&ipgre_tap_ops);
1758 rtnl_link_unregister(&ipgre_link_ops);
1760 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1762 unregister_pernet_device(&erspan_net_ops);
1764 unregister_pernet_device(&ipgre_tap_net_ops);
1766 unregister_pernet_device(&ipgre_net_ops);
1770 static void __exit ipgre_fini(void)
1772 rtnl_link_unregister(&ipgre_tap_ops);
1773 rtnl_link_unregister(&ipgre_link_ops);
1774 rtnl_link_unregister(&erspan_link_ops);
1775 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1776 unregister_pernet_device(&ipgre_tap_net_ops);
1777 unregister_pernet_device(&ipgre_net_ops);
1778 unregister_pernet_device(&erspan_net_ops);
1781 module_init(ipgre_init);
1782 module_exit(ipgre_fini);
1783 MODULE_LICENSE("GPL");
1784 MODULE_ALIAS_RTNL_LINK("gre");
1785 MODULE_ALIAS_RTNL_LINK("gretap");
1786 MODULE_ALIAS_RTNL_LINK("erspan");
1787 MODULE_ALIAS_NETDEV("gre0");
1788 MODULE_ALIAS_NETDEV("gretap0");
1789 MODULE_ALIAS_NETDEV("erspan0");