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>
47 #include <net/inet_dscp.h>
53 1. The most important issue is detecting local dead loops.
54 They would cause complete host lockup in transmit, which
55 would be "resolved" by stack overflow or, if queueing is enabled,
56 with infinite looping in net_bh.
58 We cannot track such dead loops during route installation,
59 it is infeasible task. The most general solutions would be
60 to keep skb->encapsulation counter (sort of local ttl),
61 and silently drop packet when it expires. It is a good
62 solution, but it supposes maintaining new variable in ALL
63 skb, even if no tunneling is used.
65 Current solution: xmit_recursion breaks dead loops. This is a percpu
66 counter, since when we enter the first ndo_xmit(), cpu migration is
67 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
69 2. Networking dead loops would not kill routers, but would really
70 kill network. IP hop limit plays role of "t->recursion" in this case,
71 if we copy it from packet being encapsulated to upper header.
72 It is very good solution, but it introduces two problems:
74 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
75 do not work over tunnels.
76 - traceroute does not work. I planned to relay ICMP from tunnel,
77 so that this problem would be solved and traceroute output
78 would even more informative. This idea appeared to be wrong:
79 only Linux complies to rfc1812 now (yes, guys, Linux is the only
80 true router now :-)), all routers (at least, in neighbourhood of mine)
81 return only 8 bytes of payload. It is the end.
83 Hence, if we want that OSPF worked or traceroute said something reasonable,
84 we should search for another solution.
86 One of them is to parse packet trying to detect inner encapsulation
87 made by our node. It is difficult or even impossible, especially,
88 taking into account fragmentation. TO be short, ttl is not solution at all.
90 Current solution: The solution was UNEXPECTEDLY SIMPLE.
91 We force DF flag on tunnels with preconfigured hop limit,
92 that is ALL. :-) Well, it does not remove the problem completely,
93 but exponential growth of network traffic is changed to linear
94 (branches, that exceed pmtu are pruned) and tunnel mtu
95 rapidly degrades to value <68, where looping stops.
96 Yes, it is not good if there exists a router in the loop,
97 which does not force DF, even when encapsulating packets have DF set.
98 But it is not our problem! Nobody could accuse us, we made
99 all that we could make. Even if it is your gated who injected
100 fatal route to network, even if it were you who configured
101 fatal static route: you are innocent. :-)
106 static bool log_ecn_error = true;
107 module_param(log_ecn_error, bool, 0644);
108 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
110 static struct rtnl_link_ops ipgre_link_ops __read_mostly;
111 static const struct header_ops ipgre_header_ops;
113 static int ipgre_tunnel_init(struct net_device *dev);
114 static void erspan_build_header(struct sk_buff *skb,
116 bool truncate, bool is_ipv4);
118 static unsigned int ipgre_net_id __read_mostly;
119 static unsigned int gre_tap_net_id __read_mostly;
120 static unsigned int erspan_net_id __read_mostly;
122 static int ipgre_err(struct sk_buff *skb, u32 info,
123 const struct tnl_ptk_info *tpi)
126 /* All the routers (except for Linux) return only
127 8 bytes of packet payload. It means, that precise relaying of
128 ICMP in the real Internet is absolutely infeasible.
130 Moreover, Cisco "wise men" put GRE key to the third word
131 in GRE header. It makes impossible maintaining even soft
132 state for keyed GRE tunnels with enabled checksum. Tell
135 Well, I wonder, rfc1812 was written by Cisco employee,
136 what the hell these idiots break standards established
139 struct net *net = dev_net(skb->dev);
140 struct ip_tunnel_net *itn;
141 const struct iphdr *iph;
142 const int type = icmp_hdr(skb)->type;
143 const int code = icmp_hdr(skb)->code;
144 unsigned int data_len = 0;
147 if (tpi->proto == htons(ETH_P_TEB))
148 itn = net_generic(net, gre_tap_net_id);
149 else if (tpi->proto == htons(ETH_P_ERSPAN) ||
150 tpi->proto == htons(ETH_P_ERSPAN2))
151 itn = net_generic(net, erspan_net_id);
153 itn = net_generic(net, ipgre_net_id);
155 iph = (const struct iphdr *)(icmp_hdr(skb) + 1);
156 t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
157 iph->daddr, iph->saddr, tpi->key);
164 case ICMP_PARAMETERPROB:
167 case ICMP_DEST_UNREACH:
170 case ICMP_PORT_UNREACH:
171 /* Impossible event. */
174 /* All others are translated to HOST_UNREACH.
175 rfc2003 contains "deep thoughts" about NET_UNREACH,
176 I believe they are just ether pollution. --ANK
182 case ICMP_TIME_EXCEEDED:
183 if (code != ICMP_EXC_TTL)
185 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
192 #if IS_ENABLED(CONFIG_IPV6)
193 if (tpi->proto == htons(ETH_P_IPV6) &&
194 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4 + tpi->hdr_len,
199 if (t->parms.iph.daddr == 0 ||
200 ipv4_is_multicast(t->parms.iph.daddr))
203 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
206 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
210 t->err_time = jiffies;
215 static void gre_err(struct sk_buff *skb, u32 info)
217 /* All the routers (except for Linux) return only
218 * 8 bytes of packet payload. It means, that precise relaying of
219 * ICMP in the real Internet is absolutely infeasible.
221 * Moreover, Cisco "wise men" put GRE key to the third word
222 * in GRE header. It makes impossible maintaining even soft
224 * GRE tunnels with enabled checksum. Tell them "thank you".
226 * Well, I wonder, rfc1812 was written by Cisco employee,
227 * what the hell these idiots break standards established
231 const struct iphdr *iph = (struct iphdr *)skb->data;
232 const int type = icmp_hdr(skb)->type;
233 const int code = icmp_hdr(skb)->code;
234 struct tnl_ptk_info tpi;
236 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IP),
240 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
241 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
242 skb->dev->ifindex, IPPROTO_GRE);
245 if (type == ICMP_REDIRECT) {
246 ipv4_redirect(skb, dev_net(skb->dev), skb->dev->ifindex,
251 ipgre_err(skb, info, &tpi);
254 static bool is_erspan_type1(int gre_hdr_len)
256 /* Both ERSPAN type I (version 0) and type II (version 1) use
257 * protocol 0x88BE, but the type I has only 4-byte GRE header,
258 * while type II has 8-byte.
260 return gre_hdr_len == 4;
263 static int erspan_rcv(struct sk_buff *skb, struct tnl_ptk_info *tpi,
266 struct net *net = dev_net(skb->dev);
267 struct metadata_dst *tun_dst = NULL;
268 struct erspan_base_hdr *ershdr;
269 IP_TUNNEL_DECLARE_FLAGS(flags);
270 struct ip_tunnel_net *itn;
271 struct ip_tunnel *tunnel;
272 const struct iphdr *iph;
273 struct erspan_md2 *md2;
277 ip_tunnel_flags_copy(flags, tpi->flags);
279 itn = net_generic(net, erspan_net_id);
281 if (is_erspan_type1(gre_hdr_len)) {
283 __set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
284 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
285 iph->saddr, iph->daddr, 0);
287 if (unlikely(!pskb_may_pull(skb,
288 gre_hdr_len + sizeof(*ershdr))))
289 return PACKET_REJECT;
291 ershdr = (struct erspan_base_hdr *)(skb->data + gre_hdr_len);
294 __set_bit(IP_TUNNEL_KEY_BIT, flags);
295 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
296 iph->saddr, iph->daddr, tpi->key);
300 if (is_erspan_type1(gre_hdr_len))
303 len = gre_hdr_len + erspan_hdr_len(ver);
305 if (unlikely(!pskb_may_pull(skb, len)))
306 return PACKET_REJECT;
308 if (__iptunnel_pull_header(skb,
314 if (tunnel->collect_md) {
315 struct erspan_metadata *pkt_md, *md;
316 struct ip_tunnel_info *info;
320 __set_bit(IP_TUNNEL_KEY_BIT, tpi->flags);
321 ip_tunnel_flags_copy(flags, tpi->flags);
322 tun_id = key32_to_tunnel_id(tpi->key);
324 tun_dst = ip_tun_rx_dst(skb, flags,
325 tun_id, sizeof(*md));
327 return PACKET_REJECT;
329 /* skb can be uncloned in __iptunnel_pull_header, so
330 * old pkt_md is no longer valid and we need to reset
333 gh = skb_network_header(skb) +
334 skb_network_header_len(skb);
335 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
337 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
340 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
343 info = &tun_dst->u.tun_info;
344 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
345 info->key.tun_flags);
346 info->options_len = sizeof(*md);
349 skb_reset_mac_header(skb);
350 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
353 return PACKET_REJECT;
360 static int __ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
361 struct ip_tunnel_net *itn, int hdr_len, bool raw_proto)
363 struct metadata_dst *tun_dst = NULL;
364 const struct iphdr *iph;
365 struct ip_tunnel *tunnel;
368 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi->flags,
369 iph->saddr, iph->daddr, tpi->key);
372 const struct iphdr *tnl_params;
374 if (__iptunnel_pull_header(skb, hdr_len, tpi->proto,
375 raw_proto, false) < 0)
378 /* Special case for ipgre_header_parse(), which expects the
379 * mac_header to point to the outer IP header.
381 if (tunnel->dev->header_ops == &ipgre_header_ops)
382 skb_pop_mac_header(skb);
384 skb_reset_mac_header(skb);
386 tnl_params = &tunnel->parms.iph;
387 if (tunnel->collect_md || tnl_params->daddr == 0) {
388 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
391 __set_bit(IP_TUNNEL_CSUM_BIT, flags);
392 __set_bit(IP_TUNNEL_KEY_BIT, flags);
393 ip_tunnel_flags_and(flags, tpi->flags, flags);
395 tun_id = key32_to_tunnel_id(tpi->key);
396 tun_dst = ip_tun_rx_dst(skb, flags, tun_id, 0);
398 return PACKET_REJECT;
401 ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
411 static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
414 struct net *net = dev_net(skb->dev);
415 struct ip_tunnel_net *itn;
418 if (tpi->proto == htons(ETH_P_TEB))
419 itn = net_generic(net, gre_tap_net_id);
421 itn = net_generic(net, ipgre_net_id);
423 res = __ipgre_rcv(skb, tpi, itn, hdr_len, false);
424 if (res == PACKET_NEXT && tpi->proto == htons(ETH_P_TEB)) {
425 /* ipgre tunnels in collect metadata mode should receive
426 * also ETH_P_TEB traffic.
428 itn = net_generic(net, ipgre_net_id);
429 res = __ipgre_rcv(skb, tpi, itn, hdr_len, true);
434 static int gre_rcv(struct sk_buff *skb)
436 struct tnl_ptk_info tpi;
437 bool csum_err = false;
440 #ifdef CONFIG_NET_IPGRE_BROADCAST
441 if (ipv4_is_multicast(ip_hdr(skb)->daddr)) {
442 /* Looped back packet, drop it! */
443 if (rt_is_output_route(skb_rtable(skb)))
448 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IP), 0);
452 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
453 tpi.proto == htons(ETH_P_ERSPAN2))) {
454 if (erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
459 if (ipgre_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
463 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
469 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
470 const struct iphdr *tnl_params,
473 struct ip_tunnel *tunnel = netdev_priv(dev);
474 IP_TUNNEL_DECLARE_FLAGS(flags);
476 ip_tunnel_flags_copy(flags, tunnel->parms.o_flags);
478 /* Push GRE header. */
479 gre_build_header(skb, tunnel->tun_hlen,
480 flags, proto, tunnel->parms.o_key,
481 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
482 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
484 ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
487 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
489 return iptunnel_handle_offloads(skb, csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
492 static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev,
495 struct ip_tunnel *tunnel = netdev_priv(dev);
496 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
497 struct ip_tunnel_info *tun_info;
498 const struct ip_tunnel_key *key;
501 tun_info = skb_tunnel_info(skb);
502 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
503 ip_tunnel_info_af(tun_info) != AF_INET))
506 key = &tun_info->key;
507 tunnel_hlen = gre_calc_hlen(key->tun_flags);
509 if (skb_cow_head(skb, dev->needed_headroom))
512 /* Push Tunnel header. */
513 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
514 tunnel->parms.o_flags)))
517 __set_bit(IP_TUNNEL_CSUM_BIT, flags);
518 __set_bit(IP_TUNNEL_KEY_BIT, flags);
519 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
520 ip_tunnel_flags_and(flags, tun_info->key.tun_flags, flags);
522 gre_build_header(skb, tunnel_hlen, flags, proto,
523 tunnel_id_to_key32(tun_info->key.tun_id),
524 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
525 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 0);
527 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
533 DEV_STATS_INC(dev, tx_dropped);
536 static void erspan_fb_xmit(struct sk_buff *skb, struct net_device *dev)
538 struct ip_tunnel *tunnel = netdev_priv(dev);
539 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
540 struct ip_tunnel_info *tun_info;
541 const struct ip_tunnel_key *key;
542 struct erspan_metadata *md;
543 bool truncate = false;
549 tun_info = skb_tunnel_info(skb);
550 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
551 ip_tunnel_info_af(tun_info) != AF_INET))
554 key = &tun_info->key;
555 if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_info->key.tun_flags))
557 if (tun_info->options_len < sizeof(*md))
559 md = ip_tunnel_info_opts(tun_info);
561 /* ERSPAN has fixed 8 byte GRE header */
562 version = md->version;
563 tunnel_hlen = 8 + erspan_hdr_len(version);
565 if (skb_cow_head(skb, dev->needed_headroom))
568 if (gre_handle_offloads(skb, false))
571 if (skb->len > dev->mtu + dev->hard_header_len) {
572 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
577 nhoff = skb_network_offset(skb);
578 if (skb->protocol == htons(ETH_P_IP) &&
579 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
582 if (skb->protocol == htons(ETH_P_IPV6)) {
585 if (skb_transport_header_was_set(skb))
586 thoff = skb_transport_offset(skb);
588 thoff = nhoff + sizeof(struct ipv6hdr);
589 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
594 erspan_build_header(skb, ntohl(tunnel_id_to_key32(key->tun_id)),
595 ntohl(md->u.index), truncate, true);
596 proto = htons(ETH_P_ERSPAN);
597 } else if (version == 2) {
598 erspan_build_header_v2(skb,
599 ntohl(tunnel_id_to_key32(key->tun_id)),
601 get_hwid(&md->u.md2),
603 proto = htons(ETH_P_ERSPAN2);
608 __set_bit(IP_TUNNEL_SEQ_BIT, flags);
609 gre_build_header(skb, 8, flags, proto, 0,
610 htonl(atomic_fetch_inc(&tunnel->o_seqno)));
612 ip_md_tunnel_xmit(skb, dev, IPPROTO_GRE, tunnel_hlen);
618 DEV_STATS_INC(dev, tx_dropped);
621 static int gre_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
623 struct ip_tunnel_info *info = skb_tunnel_info(skb);
624 const struct ip_tunnel_key *key;
628 if (ip_tunnel_info_af(info) != AF_INET)
632 ip_tunnel_init_flow(&fl4, IPPROTO_GRE, key->u.ipv4.dst, key->u.ipv4.src,
633 tunnel_id_to_key32(key->tun_id),
634 key->tos & ~INET_ECN_MASK, dev_net(dev), 0,
635 skb->mark, skb_get_hash(skb), key->flow_flags);
636 rt = ip_route_output_key(dev_net(dev), &fl4);
641 info->key.u.ipv4.src = fl4.saddr;
645 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
646 struct net_device *dev)
648 struct ip_tunnel *tunnel = netdev_priv(dev);
649 const struct iphdr *tnl_params;
651 if (!pskb_inet_may_pull(skb))
654 if (tunnel->collect_md) {
655 gre_fb_xmit(skb, dev, skb->protocol);
659 if (dev->header_ops) {
660 int pull_len = tunnel->hlen + sizeof(struct iphdr);
662 if (skb_cow_head(skb, 0))
665 tnl_params = (const struct iphdr *)skb->data;
667 if (!pskb_network_may_pull(skb, pull_len))
670 /* ip_tunnel_xmit() needs skb->data pointing to gre header. */
671 skb_pull(skb, pull_len);
672 skb_reset_mac_header(skb);
674 if (skb->ip_summed == CHECKSUM_PARTIAL &&
675 skb_checksum_start(skb) < skb->data)
678 if (skb_cow_head(skb, dev->needed_headroom))
681 tnl_params = &tunnel->parms.iph;
684 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
685 tunnel->parms.o_flags)))
688 __gre_xmit(skb, dev, tnl_params, skb->protocol);
693 DEV_STATS_INC(dev, tx_dropped);
697 static netdev_tx_t erspan_xmit(struct sk_buff *skb,
698 struct net_device *dev)
700 struct ip_tunnel *tunnel = netdev_priv(dev);
701 bool truncate = false;
704 if (!pskb_inet_may_pull(skb))
707 if (tunnel->collect_md) {
708 erspan_fb_xmit(skb, dev);
712 if (gre_handle_offloads(skb, false))
715 if (skb_cow_head(skb, dev->needed_headroom))
718 if (skb->len > dev->mtu + dev->hard_header_len) {
719 if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
724 /* Push ERSPAN header */
725 if (tunnel->erspan_ver == 0) {
726 proto = htons(ETH_P_ERSPAN);
727 __clear_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags);
728 } else if (tunnel->erspan_ver == 1) {
729 erspan_build_header(skb, ntohl(tunnel->parms.o_key),
732 proto = htons(ETH_P_ERSPAN);
733 } else if (tunnel->erspan_ver == 2) {
734 erspan_build_header_v2(skb, ntohl(tunnel->parms.o_key),
735 tunnel->dir, tunnel->hwid,
737 proto = htons(ETH_P_ERSPAN2);
742 __clear_bit(IP_TUNNEL_KEY_BIT, tunnel->parms.o_flags);
743 __gre_xmit(skb, dev, &tunnel->parms.iph, proto);
748 DEV_STATS_INC(dev, tx_dropped);
752 static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
753 struct net_device *dev)
755 struct ip_tunnel *tunnel = netdev_priv(dev);
757 if (!pskb_inet_may_pull(skb))
760 if (tunnel->collect_md) {
761 gre_fb_xmit(skb, dev, htons(ETH_P_TEB));
765 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
766 tunnel->parms.o_flags)))
769 if (skb_cow_head(skb, dev->needed_headroom))
772 __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
777 DEV_STATS_INC(dev, tx_dropped);
781 static void ipgre_link_update(struct net_device *dev, bool set_mtu)
783 struct ip_tunnel *tunnel = netdev_priv(dev);
786 len = tunnel->tun_hlen;
787 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
788 len = tunnel->tun_hlen - len;
789 tunnel->hlen = tunnel->hlen + len;
792 dev->hard_header_len += len;
794 dev->needed_headroom += len;
797 WRITE_ONCE(dev->mtu, max_t(int, dev->mtu - len, 68));
799 if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags) ||
800 (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
801 tunnel->encap.type != TUNNEL_ENCAP_NONE)) {
802 dev->features &= ~NETIF_F_GSO_SOFTWARE;
803 dev->hw_features &= ~NETIF_F_GSO_SOFTWARE;
805 dev->features |= NETIF_F_GSO_SOFTWARE;
806 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
810 static int ipgre_tunnel_ctl(struct net_device *dev,
811 struct ip_tunnel_parm_kern *p,
814 __be16 i_flags, o_flags;
817 if (!ip_tunnel_flags_is_be16_compat(p->i_flags) ||
818 !ip_tunnel_flags_is_be16_compat(p->o_flags))
821 i_flags = ip_tunnel_flags_to_be16(p->i_flags);
822 o_flags = ip_tunnel_flags_to_be16(p->o_flags);
824 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
825 if (p->iph.version != 4 || p->iph.protocol != IPPROTO_GRE ||
826 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)) ||
827 ((i_flags | o_flags) & (GRE_VERSION | GRE_ROUTING)))
831 gre_flags_to_tnl_flags(p->i_flags, i_flags);
832 gre_flags_to_tnl_flags(p->o_flags, o_flags);
834 err = ip_tunnel_ctl(dev, p, cmd);
838 if (cmd == SIOCCHGTUNNEL) {
839 struct ip_tunnel *t = netdev_priv(dev);
841 ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags);
842 ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags);
844 if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
845 ipgre_link_update(dev, true);
848 i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
849 ip_tunnel_flags_from_be16(p->i_flags, i_flags);
850 o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
851 ip_tunnel_flags_from_be16(p->o_flags, o_flags);
856 /* Nice toy. Unfortunately, useless in real life :-)
857 It allows to construct virtual multiprotocol broadcast "LAN"
858 over the Internet, provided multicast routing is tuned.
861 I have no idea was this bicycle invented before me,
862 so that I had to set ARPHRD_IPGRE to a random value.
863 I have an impression, that Cisco could make something similar,
864 but this feature is apparently missing in IOS<=11.2(8).
866 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
867 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
869 ping -t 255 224.66.66.66
871 If nobody answers, mbone does not work.
873 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
874 ip addr add 10.66.66.<somewhat>/24 dev Universe
876 ifconfig Universe add fe80::<Your_real_addr>/10
877 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
880 ftp fec0:6666:6666::193.233.7.65
883 static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
885 const void *daddr, const void *saddr, unsigned int len)
887 struct ip_tunnel *t = netdev_priv(dev);
889 struct gre_base_hdr *greh;
891 iph = skb_push(skb, t->hlen + sizeof(*iph));
892 greh = (struct gre_base_hdr *)(iph+1);
893 greh->flags = gre_tnl_flags_to_gre_flags(t->parms.o_flags);
894 greh->protocol = htons(type);
896 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
898 /* Set the source hardware address. */
900 memcpy(&iph->saddr, saddr, 4);
902 memcpy(&iph->daddr, daddr, 4);
904 return t->hlen + sizeof(*iph);
906 return -(t->hlen + sizeof(*iph));
909 static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
911 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
912 memcpy(haddr, &iph->saddr, 4);
916 static const struct header_ops ipgre_header_ops = {
917 .create = ipgre_header,
918 .parse = ipgre_header_parse,
921 #ifdef CONFIG_NET_IPGRE_BROADCAST
922 static int ipgre_open(struct net_device *dev)
924 struct ip_tunnel *t = netdev_priv(dev);
926 if (ipv4_is_multicast(t->parms.iph.daddr)) {
930 rt = ip_route_output_gre(t->net, &fl4,
934 t->parms.iph.tos & INET_DSCP_MASK,
937 return -EADDRNOTAVAIL;
940 if (!__in_dev_get_rtnl(dev))
941 return -EADDRNOTAVAIL;
942 t->mlink = dev->ifindex;
943 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
948 static int ipgre_close(struct net_device *dev)
950 struct ip_tunnel *t = netdev_priv(dev);
952 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
953 struct in_device *in_dev;
954 in_dev = inetdev_by_index(t->net, t->mlink);
956 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
962 static const struct net_device_ops ipgre_netdev_ops = {
963 .ndo_init = ipgre_tunnel_init,
964 .ndo_uninit = ip_tunnel_uninit,
965 #ifdef CONFIG_NET_IPGRE_BROADCAST
966 .ndo_open = ipgre_open,
967 .ndo_stop = ipgre_close,
969 .ndo_start_xmit = ipgre_xmit,
970 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
971 .ndo_change_mtu = ip_tunnel_change_mtu,
972 .ndo_get_stats64 = dev_get_tstats64,
973 .ndo_get_iflink = ip_tunnel_get_iflink,
974 .ndo_tunnel_ctl = ipgre_tunnel_ctl,
977 #define GRE_FEATURES (NETIF_F_SG | \
982 static void ipgre_tunnel_setup(struct net_device *dev)
984 dev->netdev_ops = &ipgre_netdev_ops;
985 dev->type = ARPHRD_IPGRE;
986 ip_tunnel_setup(dev, ipgre_net_id);
989 static void __gre_tunnel_init(struct net_device *dev)
991 struct ip_tunnel *tunnel;
993 tunnel = netdev_priv(dev);
994 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
995 tunnel->parms.iph.protocol = IPPROTO_GRE;
997 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
998 dev->needed_headroom = tunnel->hlen + sizeof(tunnel->parms.iph);
1000 dev->features |= GRE_FEATURES;
1001 dev->hw_features |= GRE_FEATURES;
1003 /* TCP offload with GRE SEQ is not supported, nor can we support 2
1004 * levels of outer headers requiring an update.
1006 if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.o_flags))
1008 if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.o_flags) &&
1009 tunnel->encap.type != TUNNEL_ENCAP_NONE)
1012 dev->features |= NETIF_F_GSO_SOFTWARE;
1013 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1018 static int ipgre_tunnel_init(struct net_device *dev)
1020 struct ip_tunnel *tunnel = netdev_priv(dev);
1021 struct iphdr *iph = &tunnel->parms.iph;
1023 __gre_tunnel_init(dev);
1025 __dev_addr_set(dev, &iph->saddr, 4);
1026 memcpy(dev->broadcast, &iph->daddr, 4);
1028 dev->flags = IFF_NOARP;
1029 netif_keep_dst(dev);
1032 if (iph->daddr && !tunnel->collect_md) {
1033 #ifdef CONFIG_NET_IPGRE_BROADCAST
1034 if (ipv4_is_multicast(iph->daddr)) {
1037 dev->flags = IFF_BROADCAST;
1038 dev->header_ops = &ipgre_header_ops;
1039 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1040 dev->needed_headroom = 0;
1043 } else if (!tunnel->collect_md) {
1044 dev->header_ops = &ipgre_header_ops;
1045 dev->hard_header_len = tunnel->hlen + sizeof(*iph);
1046 dev->needed_headroom = 0;
1049 return ip_tunnel_init(dev);
1052 static const struct gre_protocol ipgre_protocol = {
1054 .err_handler = gre_err,
1057 static int __net_init ipgre_init_net(struct net *net)
1059 return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL);
1062 static void __net_exit ipgre_exit_batch_rtnl(struct list_head *list_net,
1063 struct list_head *dev_to_kill)
1065 ip_tunnel_delete_nets(list_net, ipgre_net_id, &ipgre_link_ops,
1069 static struct pernet_operations ipgre_net_ops = {
1070 .init = ipgre_init_net,
1071 .exit_batch_rtnl = ipgre_exit_batch_rtnl,
1072 .id = &ipgre_net_id,
1073 .size = sizeof(struct ip_tunnel_net),
1076 static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1077 struct netlink_ext_ack *extack)
1085 if (data[IFLA_GRE_IFLAGS])
1086 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1087 if (data[IFLA_GRE_OFLAGS])
1088 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1089 if (flags & (GRE_VERSION|GRE_ROUTING))
1092 if (data[IFLA_GRE_COLLECT_METADATA] &&
1093 data[IFLA_GRE_ENCAP_TYPE] &&
1094 nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]) != TUNNEL_ENCAP_NONE)
1100 static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1101 struct netlink_ext_ack *extack)
1105 if (tb[IFLA_ADDRESS]) {
1106 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1108 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1109 return -EADDRNOTAVAIL;
1115 if (data[IFLA_GRE_REMOTE]) {
1116 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1122 return ipgre_tunnel_validate(tb, data, extack);
1125 static int erspan_validate(struct nlattr *tb[], struct nlattr *data[],
1126 struct netlink_ext_ack *extack)
1134 ret = ipgre_tap_validate(tb, data, extack);
1138 if (data[IFLA_GRE_ERSPAN_VER] &&
1139 nla_get_u8(data[IFLA_GRE_ERSPAN_VER]) == 0)
1142 /* ERSPAN type II/III should only have GRE sequence and key flag */
1143 if (data[IFLA_GRE_OFLAGS])
1144 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1145 if (data[IFLA_GRE_IFLAGS])
1146 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1147 if (!data[IFLA_GRE_COLLECT_METADATA] &&
1148 flags != (GRE_SEQ | GRE_KEY))
1151 /* ERSPAN Session ID only has 10-bit. Since we reuse
1152 * 32-bit key field as ID, check it's range.
1154 if (data[IFLA_GRE_IKEY] &&
1155 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1158 if (data[IFLA_GRE_OKEY] &&
1159 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1165 static int ipgre_netlink_parms(struct net_device *dev,
1166 struct nlattr *data[],
1167 struct nlattr *tb[],
1168 struct ip_tunnel_parm_kern *parms,
1171 struct ip_tunnel *t = netdev_priv(dev);
1173 memset(parms, 0, sizeof(*parms));
1175 parms->iph.protocol = IPPROTO_GRE;
1180 if (data[IFLA_GRE_LINK])
1181 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1183 if (data[IFLA_GRE_IFLAGS])
1184 gre_flags_to_tnl_flags(parms->i_flags,
1185 nla_get_be16(data[IFLA_GRE_IFLAGS]));
1187 if (data[IFLA_GRE_OFLAGS])
1188 gre_flags_to_tnl_flags(parms->o_flags,
1189 nla_get_be16(data[IFLA_GRE_OFLAGS]));
1191 if (data[IFLA_GRE_IKEY])
1192 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1194 if (data[IFLA_GRE_OKEY])
1195 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1197 if (data[IFLA_GRE_LOCAL])
1198 parms->iph.saddr = nla_get_in_addr(data[IFLA_GRE_LOCAL]);
1200 if (data[IFLA_GRE_REMOTE])
1201 parms->iph.daddr = nla_get_in_addr(data[IFLA_GRE_REMOTE]);
1203 if (data[IFLA_GRE_TTL])
1204 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1206 if (data[IFLA_GRE_TOS])
1207 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1209 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) {
1212 parms->iph.frag_off = htons(IP_DF);
1215 if (data[IFLA_GRE_COLLECT_METADATA]) {
1216 t->collect_md = true;
1217 if (dev->type == ARPHRD_IPGRE)
1218 dev->type = ARPHRD_NONE;
1221 if (data[IFLA_GRE_IGNORE_DF]) {
1222 if (nla_get_u8(data[IFLA_GRE_IGNORE_DF])
1223 && (parms->iph.frag_off & htons(IP_DF)))
1225 t->ignore_df = !!nla_get_u8(data[IFLA_GRE_IGNORE_DF]);
1228 if (data[IFLA_GRE_FWMARK])
1229 *fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1234 static int erspan_netlink_parms(struct net_device *dev,
1235 struct nlattr *data[],
1236 struct nlattr *tb[],
1237 struct ip_tunnel_parm_kern *parms,
1240 struct ip_tunnel *t = netdev_priv(dev);
1243 err = ipgre_netlink_parms(dev, data, tb, parms, fwmark);
1249 if (data[IFLA_GRE_ERSPAN_VER]) {
1250 t->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1252 if (t->erspan_ver > 2)
1256 if (t->erspan_ver == 1) {
1257 if (data[IFLA_GRE_ERSPAN_INDEX]) {
1258 t->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1259 if (t->index & ~INDEX_MASK)
1262 } else if (t->erspan_ver == 2) {
1263 if (data[IFLA_GRE_ERSPAN_DIR]) {
1264 t->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1265 if (t->dir & ~(DIR_MASK >> DIR_OFFSET))
1268 if (data[IFLA_GRE_ERSPAN_HWID]) {
1269 t->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1270 if (t->hwid & ~(HWID_MASK >> HWID_OFFSET))
1278 /* This function returns true when ENCAP attributes are present in the nl msg */
1279 static bool ipgre_netlink_encap_parms(struct nlattr *data[],
1280 struct ip_tunnel_encap *ipencap)
1284 memset(ipencap, 0, sizeof(*ipencap));
1289 if (data[IFLA_GRE_ENCAP_TYPE]) {
1291 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1294 if (data[IFLA_GRE_ENCAP_FLAGS]) {
1296 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1299 if (data[IFLA_GRE_ENCAP_SPORT]) {
1301 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1304 if (data[IFLA_GRE_ENCAP_DPORT]) {
1306 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1312 static int gre_tap_init(struct net_device *dev)
1314 __gre_tunnel_init(dev);
1315 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1316 netif_keep_dst(dev);
1318 return ip_tunnel_init(dev);
1321 static const struct net_device_ops gre_tap_netdev_ops = {
1322 .ndo_init = gre_tap_init,
1323 .ndo_uninit = ip_tunnel_uninit,
1324 .ndo_start_xmit = gre_tap_xmit,
1325 .ndo_set_mac_address = eth_mac_addr,
1326 .ndo_validate_addr = eth_validate_addr,
1327 .ndo_change_mtu = ip_tunnel_change_mtu,
1328 .ndo_get_stats64 = dev_get_tstats64,
1329 .ndo_get_iflink = ip_tunnel_get_iflink,
1330 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1333 static int erspan_tunnel_init(struct net_device *dev)
1335 struct ip_tunnel *tunnel = netdev_priv(dev);
1337 if (tunnel->erspan_ver == 0)
1338 tunnel->tun_hlen = 4; /* 4-byte GRE hdr. */
1340 tunnel->tun_hlen = 8; /* 8-byte GRE hdr. */
1342 tunnel->parms.iph.protocol = IPPROTO_GRE;
1343 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1344 erspan_hdr_len(tunnel->erspan_ver);
1346 dev->features |= GRE_FEATURES;
1347 dev->hw_features |= GRE_FEATURES;
1348 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1349 netif_keep_dst(dev);
1351 return ip_tunnel_init(dev);
1354 static const struct net_device_ops erspan_netdev_ops = {
1355 .ndo_init = erspan_tunnel_init,
1356 .ndo_uninit = ip_tunnel_uninit,
1357 .ndo_start_xmit = erspan_xmit,
1358 .ndo_set_mac_address = eth_mac_addr,
1359 .ndo_validate_addr = eth_validate_addr,
1360 .ndo_change_mtu = ip_tunnel_change_mtu,
1361 .ndo_get_stats64 = dev_get_tstats64,
1362 .ndo_get_iflink = ip_tunnel_get_iflink,
1363 .ndo_fill_metadata_dst = gre_fill_metadata_dst,
1366 static void ipgre_tap_setup(struct net_device *dev)
1370 dev->netdev_ops = &gre_tap_netdev_ops;
1371 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1372 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1373 ip_tunnel_setup(dev, gre_tap_net_id);
1377 ipgre_newlink_encap_setup(struct net_device *dev, struct nlattr *data[])
1379 struct ip_tunnel_encap ipencap;
1381 if (ipgre_netlink_encap_parms(data, &ipencap)) {
1382 struct ip_tunnel *t = netdev_priv(dev);
1383 int err = ip_tunnel_encap_setup(t, &ipencap);
1392 static int ipgre_newlink(struct net *src_net, struct net_device *dev,
1393 struct nlattr *tb[], struct nlattr *data[],
1394 struct netlink_ext_ack *extack)
1396 struct ip_tunnel_parm_kern p;
1400 err = ipgre_newlink_encap_setup(dev, data);
1404 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1407 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1410 static int erspan_newlink(struct net *src_net, struct net_device *dev,
1411 struct nlattr *tb[], struct nlattr *data[],
1412 struct netlink_ext_ack *extack)
1414 struct ip_tunnel_parm_kern p;
1418 err = ipgre_newlink_encap_setup(dev, data);
1422 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1425 return ip_tunnel_newlink(dev, tb, &p, fwmark);
1428 static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1429 struct nlattr *data[],
1430 struct netlink_ext_ack *extack)
1432 struct ip_tunnel *t = netdev_priv(dev);
1433 struct ip_tunnel_parm_kern p;
1434 __u32 fwmark = t->fwmark;
1437 err = ipgre_newlink_encap_setup(dev, data);
1441 err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
1445 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1449 ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1450 ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1452 ipgre_link_update(dev, !tb[IFLA_MTU]);
1457 static int erspan_changelink(struct net_device *dev, struct nlattr *tb[],
1458 struct nlattr *data[],
1459 struct netlink_ext_ack *extack)
1461 struct ip_tunnel *t = netdev_priv(dev);
1462 struct ip_tunnel_parm_kern p;
1463 __u32 fwmark = t->fwmark;
1466 err = ipgre_newlink_encap_setup(dev, data);
1470 err = erspan_netlink_parms(dev, data, tb, &p, &fwmark);
1474 err = ip_tunnel_changelink(dev, tb, &p, fwmark);
1478 ip_tunnel_flags_copy(t->parms.i_flags, p.i_flags);
1479 ip_tunnel_flags_copy(t->parms.o_flags, p.o_flags);
1484 static size_t ipgre_get_size(const struct net_device *dev)
1489 /* IFLA_GRE_IFLAGS */
1491 /* IFLA_GRE_OFLAGS */
1497 /* IFLA_GRE_LOCAL */
1499 /* IFLA_GRE_REMOTE */
1505 /* IFLA_GRE_PMTUDISC */
1507 /* IFLA_GRE_ENCAP_TYPE */
1509 /* IFLA_GRE_ENCAP_FLAGS */
1511 /* IFLA_GRE_ENCAP_SPORT */
1513 /* IFLA_GRE_ENCAP_DPORT */
1515 /* IFLA_GRE_COLLECT_METADATA */
1517 /* IFLA_GRE_IGNORE_DF */
1519 /* IFLA_GRE_FWMARK */
1521 /* IFLA_GRE_ERSPAN_INDEX */
1523 /* IFLA_GRE_ERSPAN_VER */
1525 /* IFLA_GRE_ERSPAN_DIR */
1527 /* IFLA_GRE_ERSPAN_HWID */
1532 static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1534 struct ip_tunnel *t = netdev_priv(dev);
1535 struct ip_tunnel_parm_kern *p = &t->parms;
1536 IP_TUNNEL_DECLARE_FLAGS(o_flags);
1538 ip_tunnel_flags_copy(o_flags, p->o_flags);
1540 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1541 nla_put_be16(skb, IFLA_GRE_IFLAGS,
1542 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1543 nla_put_be16(skb, IFLA_GRE_OFLAGS,
1544 gre_tnl_flags_to_gre_flags(o_flags)) ||
1545 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1546 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1547 nla_put_in_addr(skb, IFLA_GRE_LOCAL, p->iph.saddr) ||
1548 nla_put_in_addr(skb, IFLA_GRE_REMOTE, p->iph.daddr) ||
1549 nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) ||
1550 nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) ||
1551 nla_put_u8(skb, IFLA_GRE_PMTUDISC,
1552 !!(p->iph.frag_off & htons(IP_DF))) ||
1553 nla_put_u32(skb, IFLA_GRE_FWMARK, t->fwmark))
1554 goto nla_put_failure;
1556 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1558 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1560 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1562 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1564 goto nla_put_failure;
1566 if (nla_put_u8(skb, IFLA_GRE_IGNORE_DF, t->ignore_df))
1567 goto nla_put_failure;
1569 if (t->collect_md) {
1570 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1571 goto nla_put_failure;
1580 static int erspan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1582 struct ip_tunnel *t = netdev_priv(dev);
1584 if (t->erspan_ver <= 2) {
1585 if (t->erspan_ver != 0 && !t->collect_md)
1586 __set_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags);
1588 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, t->erspan_ver))
1589 goto nla_put_failure;
1591 if (t->erspan_ver == 1) {
1592 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, t->index))
1593 goto nla_put_failure;
1594 } else if (t->erspan_ver == 2) {
1595 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, t->dir))
1596 goto nla_put_failure;
1597 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, t->hwid))
1598 goto nla_put_failure;
1602 return ipgre_fill_info(skb, dev);
1608 static void erspan_setup(struct net_device *dev)
1610 struct ip_tunnel *t = netdev_priv(dev);
1614 dev->netdev_ops = &erspan_netdev_ops;
1615 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1616 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1617 ip_tunnel_setup(dev, erspan_net_id);
1621 static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1622 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1623 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1624 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1625 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1626 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
1627 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
1628 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
1629 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1630 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1631 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1632 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 },
1633 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 },
1634 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 },
1635 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 },
1636 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
1637 [IFLA_GRE_IGNORE_DF] = { .type = NLA_U8 },
1638 [IFLA_GRE_FWMARK] = { .type = NLA_U32 },
1639 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
1640 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 },
1641 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 },
1642 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 },
1645 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1647 .maxtype = IFLA_GRE_MAX,
1648 .policy = ipgre_policy,
1649 .priv_size = sizeof(struct ip_tunnel),
1650 .setup = ipgre_tunnel_setup,
1651 .validate = ipgre_tunnel_validate,
1652 .newlink = ipgre_newlink,
1653 .changelink = ipgre_changelink,
1654 .dellink = ip_tunnel_dellink,
1655 .get_size = ipgre_get_size,
1656 .fill_info = ipgre_fill_info,
1657 .get_link_net = ip_tunnel_get_link_net,
1660 static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1662 .maxtype = IFLA_GRE_MAX,
1663 .policy = ipgre_policy,
1664 .priv_size = sizeof(struct ip_tunnel),
1665 .setup = ipgre_tap_setup,
1666 .validate = ipgre_tap_validate,
1667 .newlink = ipgre_newlink,
1668 .changelink = ipgre_changelink,
1669 .dellink = ip_tunnel_dellink,
1670 .get_size = ipgre_get_size,
1671 .fill_info = ipgre_fill_info,
1672 .get_link_net = ip_tunnel_get_link_net,
1675 static struct rtnl_link_ops erspan_link_ops __read_mostly = {
1677 .maxtype = IFLA_GRE_MAX,
1678 .policy = ipgre_policy,
1679 .priv_size = sizeof(struct ip_tunnel),
1680 .setup = erspan_setup,
1681 .validate = erspan_validate,
1682 .newlink = erspan_newlink,
1683 .changelink = erspan_changelink,
1684 .dellink = ip_tunnel_dellink,
1685 .get_size = ipgre_get_size,
1686 .fill_info = erspan_fill_info,
1687 .get_link_net = ip_tunnel_get_link_net,
1690 struct net_device *gretap_fb_dev_create(struct net *net, const char *name,
1691 u8 name_assign_type)
1693 struct nlattr *tb[IFLA_MAX + 1];
1694 struct net_device *dev;
1695 LIST_HEAD(list_kill);
1696 struct ip_tunnel *t;
1699 memset(&tb, 0, sizeof(tb));
1701 dev = rtnl_create_link(net, name, name_assign_type,
1702 &ipgre_tap_ops, tb, NULL);
1706 /* Configure flow based GRE device. */
1707 t = netdev_priv(dev);
1708 t->collect_md = true;
1710 err = ipgre_newlink(net, dev, tb, NULL, NULL);
1713 return ERR_PTR(err);
1716 /* openvswitch users expect packet sizes to be unrestricted,
1717 * so set the largest MTU we can.
1719 err = __ip_tunnel_change_mtu(dev, IP_MAX_MTU, false);
1723 err = rtnl_configure_link(dev, NULL, 0, NULL);
1729 ip_tunnel_dellink(dev, &list_kill);
1730 unregister_netdevice_many(&list_kill);
1731 return ERR_PTR(err);
1733 EXPORT_SYMBOL_GPL(gretap_fb_dev_create);
1735 static int __net_init ipgre_tap_init_net(struct net *net)
1737 return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
1740 static void __net_exit ipgre_tap_exit_batch_rtnl(struct list_head *list_net,
1741 struct list_head *dev_to_kill)
1743 ip_tunnel_delete_nets(list_net, gre_tap_net_id, &ipgre_tap_ops,
1747 static struct pernet_operations ipgre_tap_net_ops = {
1748 .init = ipgre_tap_init_net,
1749 .exit_batch_rtnl = ipgre_tap_exit_batch_rtnl,
1750 .id = &gre_tap_net_id,
1751 .size = sizeof(struct ip_tunnel_net),
1754 static int __net_init erspan_init_net(struct net *net)
1756 return ip_tunnel_init_net(net, erspan_net_id,
1757 &erspan_link_ops, "erspan0");
1760 static void __net_exit erspan_exit_batch_rtnl(struct list_head *net_list,
1761 struct list_head *dev_to_kill)
1763 ip_tunnel_delete_nets(net_list, erspan_net_id, &erspan_link_ops,
1767 static struct pernet_operations erspan_net_ops = {
1768 .init = erspan_init_net,
1769 .exit_batch_rtnl = erspan_exit_batch_rtnl,
1770 .id = &erspan_net_id,
1771 .size = sizeof(struct ip_tunnel_net),
1774 static int __init ipgre_init(void)
1778 pr_info("GRE over IPv4 tunneling driver\n");
1780 err = register_pernet_device(&ipgre_net_ops);
1784 err = register_pernet_device(&ipgre_tap_net_ops);
1786 goto pnet_tap_failed;
1788 err = register_pernet_device(&erspan_net_ops);
1790 goto pnet_erspan_failed;
1792 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
1794 pr_info("%s: can't add protocol\n", __func__);
1795 goto add_proto_failed;
1798 err = rtnl_link_register(&ipgre_link_ops);
1800 goto rtnl_link_failed;
1802 err = rtnl_link_register(&ipgre_tap_ops);
1804 goto tap_ops_failed;
1806 err = rtnl_link_register(&erspan_link_ops);
1808 goto erspan_link_failed;
1813 rtnl_link_unregister(&ipgre_tap_ops);
1815 rtnl_link_unregister(&ipgre_link_ops);
1817 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1819 unregister_pernet_device(&erspan_net_ops);
1821 unregister_pernet_device(&ipgre_tap_net_ops);
1823 unregister_pernet_device(&ipgre_net_ops);
1827 static void __exit ipgre_fini(void)
1829 rtnl_link_unregister(&ipgre_tap_ops);
1830 rtnl_link_unregister(&ipgre_link_ops);
1831 rtnl_link_unregister(&erspan_link_ops);
1832 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
1833 unregister_pernet_device(&ipgre_tap_net_ops);
1834 unregister_pernet_device(&ipgre_net_ops);
1835 unregister_pernet_device(&erspan_net_ops);
1838 module_init(ipgre_init);
1839 module_exit(ipgre_fini);
1840 MODULE_DESCRIPTION("IPv4 GRE tunnels over IP library");
1841 MODULE_LICENSE("GPL");
1842 MODULE_ALIAS_RTNL_LINK("gre");
1843 MODULE_ALIAS_RTNL_LINK("gretap");
1844 MODULE_ALIAS_RTNL_LINK("erspan");
1845 MODULE_ALIAS_NETDEV("gre0");
1846 MODULE_ALIAS_NETDEV("gretap0");
1847 MODULE_ALIAS_NETDEV("erspan0");