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);
463 if (tunnel->parms.o_flags & TUNNEL_SEQ)
466 /* Push GRE header. */
467 gre_build_header(skb, tunnel->tun_hlen,
468 tunnel->parms.o_flags, proto, tunnel->parms.o_key,
469 htonl(tunnel->o_seqno));
471 ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
474 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
476 if (csum && skb_checksum_start(skb) < skb->data)
478 return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
481 static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
484 struct ip_tunnel *tunnel = netdev_priv(dev);
485 struct ip_tunnel_info *tun_info;
486 const struct ip_tunnel_key *key;
490 tun_info = skb_tunnel_info(skb);
491 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
492 ip_tunnel_info_af(tun_info) != AF_INET))
495 key = &tun_info->key;
496 tunnel_hlen = gre_calc_hlen(key->tun_flags);
498 if (skb_cow_head(skb, dev->needed_headroom))
501 /* Push Tunnel header. */
502 if (gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM)))
505 flags = tun_info->key.tun_flags &
506 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
507 gre_build_header(skb, tunnel_hlen, flags, proto,
508 tunnel_id_to_key32(tun_info->key.tun_id),
509 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) : 0);
511 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
517 dev->stats.tx_dropped++;
520 static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
522 struct ip_tunnel *tunnel = netdev_priv(dev);
523 struct ip_tunnel_info *tun_info;
524 const struct ip_tunnel_key *key;
525 struct erspan_metadata *md;
526 bool truncate = false;
533 tun_info = skb_tunnel_info(skb);
534 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
535 ip_tunnel_info_af(tun_info) != AF_INET))
538 key = &tun_info->key;
539 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT))
541 if (tun_info->options_len < sizeof(*md))
543 md = ip_tunnel_info_opts(tun_info);
545 /* ERSPAN has fixed 8 byte GRE header */
546 version = md->version;
547 tunnel_hlen = 8 + erspan_hdr_len(version);
549 if (skb_cow_head(skb, dev->needed_headroom))
552 if (gre_handle_offloads(skb, false))
555 if (skb->len > dev->mtu + dev->hard_header_len) {
556 pskb_trim(skb, dev->mtu + dev->hard_header_len);
560 nhoff = skb_network_header(skb) - skb_mac_header(skb);
561 if (skb->protocol == htons(ETH_P_IP) &&
562 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
565 thoff = skb_transport_header(skb) - skb_mac_header(skb);
566 if (skb->protocol == htons(ETH_P_IPV6) &&
567 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff))
571 erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
572 ntohl(md->u.index), truncate, true);
573 proto = htons(ETH_P_ERSPAN);
574 } else if (version == 2) {
575 erspan_build_header_v2(skb,
576 ntohl(tunnel_id_to_key32(key->tun_id)),
578 get_hwid(&md->u.md2),
580 proto = htons(ETH_P_ERSPAN2);
585 gre_build_header(skb, 8, TUNNEL_SEQ,
586 proto, 0, htonl(tunnel->o_seqno++));
588 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
594 dev->stats.tx_dropped++;
597 static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
599 struct ip_tunnel_info *info = skb_tunnel_info(skb);
600 const struct ip_tunnel_key *key;
604 if (ip_tunnel_info_af(info) != AF_INET)
608 ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
609 tunnel_id_to_key32(key->tun_id), key->tos, 0,
610 skb->mark, skb_get_hash(skb));
611 rt = ip_route_output_key(dev_net(dev), &fl4);
616 info->key.u.ipv4.src = fl4.saddr;
620 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
621 struct net_device *dev)
623 struct ip_tunnel *tunnel = netdev_priv(dev);
624 const struct iphdr *tnl_params;
626 if (!pskb_inet_may_pull(skb))
629 if (tunnel->collect_md) {
630 gre_fb_xmit(skb, dev, skb->protocol);
634 if (dev->header_ops) {
635 if (skb_cow_head(skb, 0))
638 tnl_params = (const struct iphdr *)skb->data;
640 /* Pull skb since ip_tunnel_xmit() needs skb->data pointing
643 skb_pull(skb, tunnel->hlen + sizeof(struct iphdr));
644 skb_reset_mac_header(skb);
646 if (skb_cow_head(skb, dev->needed_headroom))
649 tnl_params = &tunnel->parms.iph;
652 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
655 __gre_xmit(skb, dev, tnl_params, skb->protocol);
660 dev->stats.tx_dropped++;
664 static netdev_tx_t erspan_xmit(struct sk_buff *skb,
665 struct net_device *dev)
667 struct ip_tunnel *tunnel = netdev_priv(dev);
668 bool truncate = false;
671 if (!pskb_inet_may_pull(skb))
674 if (tunnel->collect_md) {
675 erspan_fb_xmit(skb, dev);
679 if (gre_handle_offloads(skb, false))
682 if (skb_cow_head(skb, dev->needed_headroom))
685 if (skb->len > dev->mtu + dev->hard_header_len) {
686 pskb_trim(skb, dev->mtu + dev->hard_header_len);
690 /* Push ERSPAN header */
691 if (tunnel->erspan_ver == 0) {
692 proto = htons(ETH_P_ERSPAN);
693 tunnel->parms.o_flags &= ~TUNNEL_SEQ;
694 } else if (tunnel->erspan_ver == 1) {
695 erspan_build_header(skb, ntohl(tunnel->parms.o_key),
698 proto = htons(ETH_P_ERSPAN);
699 } else if (tunnel->erspan_ver == 2) {
700 erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
701 tunnel->dir, tunnel->hwid,
703 proto = htons(ETH_P_ERSPAN2);
708 tunnel->parms.o_flags &= ~TUNNEL_KEY;
709 __gre_xmit(skb, dev, &tunnel->parms.iph, proto);
714 dev->stats.tx_dropped++;
718 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
719 struct net_device *dev)
721 struct ip_tunnel *tunnel = netdev_priv(dev);
723 if (!pskb_inet_may_pull(skb))
726 if (tunnel->collect_md) {
727 gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
731 if (gre_handle_offloads(skb, !!(tunnel->parms.o_flags & TUNNEL_CSUM)))
734 if (skb_cow_head(skb, dev->needed_headroom))
737 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
742 dev->stats.tx_dropped++;
746 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
748 struct ip_tunnel *tunnel = netdev_priv(dev);
751 len = tunnel->tun_hlen;
752 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
753 len = tunnel->tun_hlen - len;
754 tunnel->hlen = tunnel->hlen + len;
757 dev->hard_header_len += len;
759 dev->needed_headroom += len;
762 dev->mtu = max_t(int, dev->mtu - len, 68);
764 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
765 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
766 tunnel->encap.type == TUNNEL_ENCAP_NONE) {
767 dev->features |= NETIF_F_GSO_SOFTWARE;
768 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
770 dev->features &= ~NETIF_F_GSO_SOFTWARE;
771 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
773 dev->features |= NETIF_F_LLTX;
775 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
776 dev->features &= ~(NETIF_F_LLTX | NETIF_F_GSO_SOFTWARE);
780 static int ipgre_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p,
785 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
786 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
787 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
788 ((p->i_flags | p->o_flags) & (GRE_VERSION | GRE_ROUTING)))
792 p->i_flags = gre_flags_to_tnl_flags(p->i_flags);
793 p->o_flags = gre_flags_to_tnl_flags(p->o_flags);
795 err = ip_tunnel_ctl(dev, p, cmd);
799 if (cmd == SIOCCHGTUNNEL) {
800 struct ip_tunnel *t = netdev_priv(dev);
802 t->parms.i_flags = p->i_flags;
803 t->parms.o_flags = p->o_flags;
805 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
806 ipgre_link_update(dev, true);
809 p->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
810 p->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
814 /* Nice toy. Unfortunately, useless in real life :-)
815 It allows to construct virtual multiprotocol broadcast "LAN"
816 over the Internet, provided multicast routing is tuned.
819 I have no idea was this bicycle invented before me,
820 so that I had to set ARPHRD_IPGRE to a random value.
821 I have an impression, that Cisco could make something similar,
822 but this feature is apparently missing in IOS<=11.2(8).
824 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
825 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
827 ping -t 255 224.66.66.66
829 If nobody answers, mbone does not work.
831 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
832 ip addr add 10.66.66.<somewhat>/24 dev Universe
834 ifconfig Universe add fe80::<Your_real_addr>/10
835 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
838 ftp fec0:6666:6666::193.233.7.65
841 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
843 const void *daddr, const void *saddr, unsigned int len)
845 struct ip_tunnel *t = netdev_priv(dev);
847 struct gre_base_hdr *greh;
849 iph = skb_push(skb, t->hlen + sizeof(*iph));
850 greh = (struct gre_base_hdr *)(iph+1);
851 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
852 greh->protocol = htons(type);
854 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
856 /* Set the source hardware address. */
858 memcpy(&iph->saddr, saddr, 4);
860 memcpy(&iph->daddr, daddr, 4);
862 return t->hlen + sizeof(*iph);
864 return -(t->hlen + sizeof(*iph));
867 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
869 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
870 memcpy(haddr, &iph->saddr, 4);
874 static const struct header_ops ipgre_header_ops = {
875 .create = ipgre_header,
876 .parse = ipgre_header_parse,
879 #ifdef CONFIG_NET_IPGRE_BROADCAST
880 static int ipgre_open(struct net_device *dev)
882 struct ip_tunnel *t = netdev_priv(dev);
884 if (ipv4_is_multicast(t->parms.iph.daddr)) {
888 rt = ip_route_output_gre(t->net, &fl4,
892 RT_TOS(t->parms.iph.tos),
895 return -EADDRNOTAVAIL;
898 if (!__in_dev_get_rtnl(dev))
899 return -EADDRNOTAVAIL;
900 t->mlink = dev->ifindex;
901 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
906 static int ipgre_close(struct net_device *dev)
908 struct ip_tunnel *t = netdev_priv(dev);
910 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
911 struct in_device *in_dev;
912 in_dev = inetdev_by_index(t->net, t->mlink);
914 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
920 static const struct net_device_ops ipgre_netdev_ops = {
921 .ndo_init = ipgre_tunnel_init,
922 .ndo_uninit = ip_tunnel_uninit,
923 #ifdef CONFIG_NET_IPGRE_BROADCAST
924 .ndo_open = ipgre_open,
925 .ndo_stop = ipgre_close,
927 .ndo_start_xmit = ipgre_xmit,
928 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
929 .ndo_change_mtu = ip_tunnel_change_mtu,
930 .ndo_get_stats64 = dev_get_tstats64,
931 .ndo_get_iflink = ip_tunnel_get_iflink,
932 .ndo_tunnel_ctl = ipgre_tunnel_ctl,
935 #define GRE_FEATURES (NETIF_F_SG | \
940 static void ipgre_tunnel_setup(struct net_device *dev)
942 dev->netdev_ops = &ipgre_netdev_ops;
943 dev->type = ARPHRD_IPGRE;
944 ip_tunnel_setup(dev, ipgre_net_id);
947 static void __gre_tunnel_init(struct net_device *dev)
949 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;
959 dev->hw_features |= GRE_FEATURES;
961 if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
962 /* TCP offload with GRE SEQ is not supported, nor
963 * can we support 2 levels of outer headers requiring
966 if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
967 (tunnel->encap.type == TUNNEL_ENCAP_NONE)) {
968 dev->features |= NETIF_F_GSO_SOFTWARE;
969 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
972 /* Can use a lockless transmit, unless we generate
975 dev->features |= NETIF_F_LLTX;
979 static int ipgre_tunnel_init(struct net_device *dev)
981 struct ip_tunnel *tunnel = netdev_priv(dev);
982 struct iphdr *iph = &tunnel->parms.iph;
984 __gre_tunnel_init(dev);
986 memcpy(dev->dev_addr, &iph->saddr, 4);
987 memcpy(dev->broadcast, &iph->daddr, 4);
989 dev->flags = IFF_NOARP;
993 if (iph->daddr && !tunnel->collect_md) {
994 #ifdef CONFIG_NET_IPGRE_BROADCAST
995 if (ipv4_is_multicast(iph->daddr)) {
998 dev->flags = IFF_BROADCAST;
999 dev->header_ops = &ipgre_header_ops;
1000 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1001 dev->needed_headroom = 0;
1004 } else if (!tunnel->collect_md) {
1005 dev->header_ops = &ipgre_header_ops;
1006 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1007 dev->needed_headroom = 0;
1010 return ip_tunnel_init(dev);
1013 static const struct gre_protocol ipgre_protocol = {
1015 .err_handler = gre_err,
1018 static int __net_init ipgre_init_net(struct net *net)
1020 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1023 static void __net_exit ipgre_exit_batch_net(struct list_head *list_net)
1025 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops);
1028 static struct pernet_operations ipgre_net_ops = {
1029 .init = ipgre_init_net,
1030 .exit_batch = ipgre_exit_batch_net,
1031 .id = &ipgre_net_id,
1032 .size = sizeof(struct ip_tunnel_net),
1035 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1036 struct netlink_ext_ack *extack)
1044 if (data[IFLA_GRE_IFLAGS])
1045 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1046 if (data[IFLA_GRE_OFLAGS])
1047 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1048 if (flags & (GRE_VERSION|GRE_ROUTING))
1051 if (data[IFLA_GRE_COLLECT_METADATA] &&
1052 data[IFLA_GRE_ENCAP_TYPE] &&
1053 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1059 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1060 struct netlink_ext_ack *extack)
1064 if (tb[IFLA_ADDRESS]) {
1065 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1067 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1068 return -EADDRNOTAVAIL;
1074 if (data[IFLA_GRE_REMOTE]) {
1075 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1081 return ipgre_tunnel_validate(tb, data, extack);
1084 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1085 struct netlink_ext_ack *extack)
1093 ret = ipgre_tap_validate(tb, data, extack);
1097 if (data[IFLA_GRE_ERSPAN_VER] &&
1098 nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1101 /* ERSPAN type II/III should only have GRE sequence and key flag */
1102 if (data[IFLA_GRE_OFLAGS])
1103 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1104 if (data[IFLA_GRE_IFLAGS])
1105 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1106 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1107 flags != (GRE_SEQ | GRE_KEY))
1110 /* ERSPAN Session ID only has 10-bit. Since we reuse
1111 * 32-bit key field as ID, check it's range.
1113 if (data[IFLA_GRE_IKEY] &&
1114 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1117 if (data[IFLA_GRE_OKEY] &&
1118 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1124 static int ipgre_netlink_parms(struct net_device *dev,
1125 struct nlattr *data[],
1126 struct nlattr *tb[],
1127 struct ip_tunnel_parm *parms,
1130 struct ip_tunnel *t = netdev_priv(dev);
1132 memset(parms, 0, sizeof(*parms));
1134 parms->iph.protocol = IPPROTO_GRE;
1139 if (data[IFLA_GRE_LINK])
1140 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1142 if (data[IFLA_GRE_IFLAGS])
1143 parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS]));
1145 if (data[IFLA_GRE_OFLAGS])
1146 parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS]));
1148 if (data[IFLA_GRE_IKEY])
1149 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1151 if (data[IFLA_GRE_OKEY])
1152 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1154 if (data[IFLA_GRE_LOCAL])
1155 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1157 if (data[IFLA_GRE_REMOTE])
1158 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1160 if (data[IFLA_GRE_TTL])
1161 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1163 if (data[IFLA_GRE_TOS])
1164 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1166 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1169 parms->iph.frag_off = htons(IP_DF);
1172 if (data[IFLA_GRE_COLLECT_METADATA]) {
1173 t->collect_md = true;
1174 if (dev->type == ARPHRD_IPGRE)
1175 dev->type = ARPHRD_NONE;
1178 if (data[IFLA_GRE_IGNORE_DF]) {
1179 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1180 && (parms->iph.frag_off & htons(IP_DF)))
1182 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1185 if (data[IFLA_GRE_FWMARK])
1186 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1191 static int erspan_netlink_parms(struct net_device *dev,
1192 struct nlattr *data[],
1193 struct nlattr *tb[],
1194 struct ip_tunnel_parm *parms,
1197 struct ip_tunnel *t = netdev_priv(dev);
1200 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1206 if (data[IFLA_GRE_ERSPAN_VER]) {
1207 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1209 if (t->erspan_ver > 2)
1213 if (t->erspan_ver == 1) {
1214 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1215 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1216 if (t->index & ~INDEX_MASK)
1219 } else if (t->erspan_ver == 2) {
1220 if (data[IFLA_GRE_ERSPAN_DIR]) {
1221 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1222 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1225 if (data[IFLA_GRE_ERSPAN_HWID]) {
1226 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1227 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1235 /* This function returns true when ENCAP attributes are present in the nl msg */
1236 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1237 struct ip_tunnel_encap *ipencap)
1241 memset(ipencap, 0, sizeof(*ipencap));
1246 if (data[IFLA_GRE_ENCAP_TYPE]) {
1248 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1251 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1253 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1256 if (data[IFLA_GRE_ENCAP_SPORT]) {
1258 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1261 if (data[IFLA_GRE_ENCAP_DPORT]) {
1263 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1269 static int gre_tap_init(struct net_device *dev)
1271 __gre_tunnel_init(dev);
1272 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1273 netif_keep_dst(dev);
1275 return ip_tunnel_init(dev);
1278 static const struct net_device_ops gre_tap_netdev_ops = {
1279 .ndo_init = gre_tap_init,
1280 .ndo_uninit = ip_tunnel_uninit,
1281 .ndo_start_xmit = gre_tap_xmit,
1282 .ndo_set_mac_address = eth_mac_addr,
1283 .ndo_validate_addr = eth_validate_addr,
1284 .ndo_change_mtu = ip_tunnel_change_mtu,
1285 .ndo_get_stats64 = dev_get_tstats64,
1286 .ndo_get_iflink = ip_tunnel_get_iflink,
1287 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1290 static int erspan_tunnel_init(struct net_device *dev)
1292 struct ip_tunnel *tunnel = netdev_priv(dev);
1294 if (tunnel->erspan_ver == 0)
1295 tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1297 tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1299 tunnel->parms.iph.protocol = IPPROTO_GRE;
1300 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1301 erspan_hdr_len(tunnel->erspan_ver);
1303 dev->features |= GRE_FEATURES;
1304 dev->hw_features |= GRE_FEATURES;
1305 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1306 netif_keep_dst(dev);
1308 return ip_tunnel_init(dev);
1311 static const struct net_device_ops erspan_netdev_ops = {
1312 .ndo_init = erspan_tunnel_init,
1313 .ndo_uninit = ip_tunnel_uninit,
1314 .ndo_start_xmit = erspan_xmit,
1315 .ndo_set_mac_address = eth_mac_addr,
1316 .ndo_validate_addr = eth_validate_addr,
1317 .ndo_change_mtu = ip_tunnel_change_mtu,
1318 .ndo_get_stats64 = dev_get_tstats64,
1319 .ndo_get_iflink = ip_tunnel_get_iflink,
1320 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1323 static void ipgre_tap_setup(struct net_device *dev)
1327 dev->netdev_ops = &gre_tap_netdev_ops;
1328 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1329 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1330 ip_tunnel_setup(dev, gre_tap_net_id);
1334 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1336 struct ip_tunnel_encap ipencap;
1338 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1339 struct ip_tunnel *t = netdev_priv(dev);
1340 int err = ip_tunnel_encap_setup(t, &ipencap);
1349 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1350 struct nlattr *tb[], struct nlattr *data[],
1351 struct netlink_ext_ack *extack)
1353 struct ip_tunnel_parm p;
1357 err = ipgre_newlink_encap_setup(dev, data);
1361 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1364 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1367 static int erspan_newlink(struct net *src_net, struct net_device *dev,
1368 struct nlattr *tb[], struct nlattr *data[],
1369 struct netlink_ext_ack *extack)
1371 struct ip_tunnel_parm p;
1375 err = ipgre_newlink_encap_setup(dev, data);
1379 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1382 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1385 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1386 struct nlattr *data[],
1387 struct netlink_ext_ack *extack)
1389 struct ip_tunnel *t = netdev_priv(dev);
1390 __u32 fwmark = t->fwmark;
1391 struct ip_tunnel_parm p;
1394 err = ipgre_newlink_encap_setup(dev, data);
1398 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1402 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1406 t->parms.i_flags = p.i_flags;
1407 t->parms.o_flags = p.o_flags;
1409 ipgre_link_update(dev, !tb[IFLA_MTU]);
1414 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1415 struct nlattr *data[],
1416 struct netlink_ext_ack *extack)
1418 struct ip_tunnel *t = netdev_priv(dev);
1419 __u32 fwmark = t->fwmark;
1420 struct ip_tunnel_parm p;
1423 err = ipgre_newlink_encap_setup(dev, data);
1427 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1431 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1435 t->parms.i_flags = p.i_flags;
1436 t->parms.o_flags = p.o_flags;
1441 static size_t ipgre_get_size(const struct net_device *dev)
1446 /* IFLA_GRE_IFLAGS */
1448 /* IFLA_GRE_OFLAGS */
1454 /* IFLA_GRE_LOCAL */
1456 /* IFLA_GRE_REMOTE */
1462 /* IFLA_GRE_PMTUDISC */
1464 /* IFLA_GRE_ENCAP_TYPE */
1466 /* IFLA_GRE_ENCAP_FLAGS */
1468 /* IFLA_GRE_ENCAP_SPORT */
1470 /* IFLA_GRE_ENCAP_DPORT */
1472 /* IFLA_GRE_COLLECT_METADATA */
1474 /* IFLA_GRE_IGNORE_DF */
1476 /* IFLA_GRE_FWMARK */
1478 /* IFLA_GRE_ERSPAN_INDEX */
1480 /* IFLA_GRE_ERSPAN_VER */
1482 /* IFLA_GRE_ERSPAN_DIR */
1484 /* IFLA_GRE_ERSPAN_HWID */
1489 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1491 struct ip_tunnel *t = netdev_priv(dev);
1492 struct ip_tunnel_parm *p = &t->parms;
1493 __be16 o_flags = p->o_flags;
1495 if (t->erspan_ver <= 2) {
1496 if (t->erspan_ver != 0 && !t->collect_md)
1497 o_flags |= TUNNEL_KEY;
1499 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1500 goto nla_put_failure;
1502 if (t->erspan_ver == 1) {
1503 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1504 goto nla_put_failure;
1505 } else if (t->erspan_ver == 2) {
1506 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1507 goto nla_put_failure;
1508 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1509 goto nla_put_failure;
1513 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1514 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1515 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1516 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1517 gre_tnl_flags_to_gre_flags(o_flags)) ||
1518 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1519 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1520 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1521 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1522 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1523 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1524 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1525 !!(p->iph.frag_off & htons(IP_DF))) ||
1526 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1527 goto nla_put_failure;
1529 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1531 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1533 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1535 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1537 goto nla_put_failure;
1539 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1540 goto nla_put_failure;
1542 if (t->collect_md) {
1543 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1544 goto nla_put_failure;
1553 static void erspan_setup(struct net_device *dev)
1555 struct ip_tunnel *t = netdev_priv(dev);
1559 dev->netdev_ops = &erspan_netdev_ops;
1560 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1561 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1562 ip_tunnel_setup(dev, erspan_net_id);
1566 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1567 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1568 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1569 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1570 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1571 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1572 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
1573 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
1574 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1575 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1576 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1577 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1578 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1579 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1580 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1581 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1582 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1583 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1584 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1585 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1586 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1587 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1590 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1592 .maxtype = IFLA_GRE_MAX,
1593 .policy = ipgre_policy,
1594 .priv_size = sizeof(struct ip_tunnel),
1595 .setup = ipgre_tunnel_setup,
1596 .validate = ipgre_tunnel_validate,
1597 .newlink = ipgre_newlink,
1598 .changelink = ipgre_changelink,
1599 .dellink = ip_tunnel_dellink,
1600 .get_size = ipgre_get_size,
1601 .fill_info = ipgre_fill_info,
1602 .get_link_net = ip_tunnel_get_link_net,
1605 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1607 .maxtype = IFLA_GRE_MAX,
1608 .policy = ipgre_policy,
1609 .priv_size = sizeof(struct ip_tunnel),
1610 .setup = ipgre_tap_setup,
1611 .validate = ipgre_tap_validate,
1612 .newlink = ipgre_newlink,
1613 .changelink = ipgre_changelink,
1614 .dellink = ip_tunnel_dellink,
1615 .get_size = ipgre_get_size,
1616 .fill_info = ipgre_fill_info,
1617 .get_link_net = ip_tunnel_get_link_net,
1620 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1622 .maxtype = IFLA_GRE_MAX,
1623 .policy = ipgre_policy,
1624 .priv_size = sizeof(struct ip_tunnel),
1625 .setup = erspan_setup,
1626 .validate = erspan_validate,
1627 .newlink = erspan_newlink,
1628 .changelink = erspan_changelink,
1629 .dellink = ip_tunnel_dellink,
1630 .get_size = ipgre_get_size,
1631 .fill_info = ipgre_fill_info,
1632 .get_link_net = ip_tunnel_get_link_net,
1635 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1636 u8 name_assign_type)
1638 struct nlattr *tb[IFLA_MAX + 1];
1639 struct net_device *dev;
1640 LIST_HEAD(list_kill);
1641 struct ip_tunnel *t;
1644 memset(&tb, 0, sizeof(tb));
1646 dev = rtnl_create_link(net, name, name_assign_type,
1647 &ipgre_tap_ops, tb, NULL);
1651 /* Configure flow based GRE device. */
1652 t = netdev_priv(dev);
1653 t->collect_md = true;
1655 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1658 return ERR_PTR(err);
1661 /* openvswitch users expect packet sizes to be unrestricted,
1662 * so set the largest MTU we can.
1664 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1668 err = rtnl_configure_link(dev, NULL);
1674 ip_tunnel_dellink(dev, &list_kill);
1675 unregister_netdevice_many(&list_kill);
1676 return ERR_PTR(err);
1678 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1680 static int __net_init ipgre_tap_init_net(struct net *net)
1682 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1685 static void __net_exit ipgre_tap_exit_batch_net(struct list_head *list_net)
1687 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops);
1690 static struct pernet_operations ipgre_tap_net_ops = {
1691 .init = ipgre_tap_init_net,
1692 .exit_batch = ipgre_tap_exit_batch_net,
1693 .id = &gre_tap_net_id,
1694 .size = sizeof(struct ip_tunnel_net),
1697 static int __net_init erspan_init_net(struct net *net)
1699 return ip_tunnel_init_net(net, erspan_net_id,
1700 &erspan_link_ops, "erspan0");
1703 static void __net_exit erspan_exit_batch_net(struct list_head *net_list)
1705 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops);
1708 static struct pernet_operations erspan_net_ops = {
1709 .init = erspan_init_net,
1710 .exit_batch = erspan_exit_batch_net,
1711 .id = &erspan_net_id,
1712 .size = sizeof(struct ip_tunnel_net),
1715 static int __init ipgre_init(void)
1719 pr_info("GRE over IPv4 tunneling driver\n");
1721 err = register_pernet_device(&ipgre_net_ops);
1725 err = register_pernet_device(&ipgre_tap_net_ops);
1727 goto pnet_tap_failed;
1729 err = register_pernet_device(&erspan_net_ops);
1731 goto pnet_erspan_failed;
1733 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1735 pr_info("%s: can't add protocol\n", __func__);
1736 goto add_proto_failed;
1739 err = rtnl_link_register(&ipgre_link_ops);
1741 goto rtnl_link_failed;
1743 err = rtnl_link_register(&ipgre_tap_ops);
1745 goto tap_ops_failed;
1747 err = rtnl_link_register(&erspan_link_ops);
1749 goto erspan_link_failed;
1754 rtnl_link_unregister(&ipgre_tap_ops);
1756 rtnl_link_unregister(&ipgre_link_ops);
1758 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1760 unregister_pernet_device(&erspan_net_ops);
1762 unregister_pernet_device(&ipgre_tap_net_ops);
1764 unregister_pernet_device(&ipgre_net_ops);
1768 static void __exit ipgre_fini(void)
1770 rtnl_link_unregister(&ipgre_tap_ops);
1771 rtnl_link_unregister(&ipgre_link_ops);
1772 rtnl_link_unregister(&erspan_link_ops);
1773 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1774 unregister_pernet_device(&ipgre_tap_net_ops);
1775 unregister_pernet_device(&ipgre_net_ops);
1776 unregister_pernet_device(&erspan_net_ops);
1779 module_init(ipgre_init);
1780 module_exit(ipgre_fini);
1781 MODULE_LICENSE("GPL");
1782 MODULE_ALIAS_RTNL_LINK("gre");
1783 MODULE_ALIAS_RTNL_LINK("gretap");
1784 MODULE_ALIAS_RTNL_LINK("erspan");
1785 MODULE_ALIAS_NETDEV("gre0");
1786 MODULE_ALIAS_NETDEV("gretap0");
1787 MODULE_ALIAS_NETDEV("erspan0");