1 // SPDX-License-Identifier: GPL-2.0
3 * XFRM virtual interface
5 * Copyright (C) 2018 secunet Security Networks AG
11 #include <linux/module.h>
12 #include <linux/capability.h>
13 #include <linux/errno.h>
14 #include <linux/types.h>
15 #include <linux/sockios.h>
16 #include <linux/icmp.h>
20 #include <linux/net.h>
21 #include <linux/in6.h>
22 #include <linux/netdevice.h>
23 #include <linux/if_link.h>
24 #include <linux/if_arp.h>
25 #include <linux/icmpv6.h>
26 #include <linux/init.h>
27 #include <linux/route.h>
28 #include <linux/rtnetlink.h>
29 #include <linux/netfilter_ipv6.h>
30 #include <linux/slab.h>
31 #include <linux/hash.h>
33 #include <linux/uaccess.h>
34 #include <linux/atomic.h>
40 #include <net/ip6_route.h>
41 #include <net/ip_tunnels.h>
42 #include <net/addrconf.h>
44 #include <net/net_namespace.h>
45 #include <net/dst_metadata.h>
46 #include <net/netns/generic.h>
47 #include <linux/etherdevice.h>
49 static int xfrmi_dev_init(struct net_device *dev);
50 static void xfrmi_dev_setup(struct net_device *dev);
51 static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
52 static unsigned int xfrmi_net_id __read_mostly;
53 static const struct net_device_ops xfrmi_netdev_ops;
55 #define XFRMI_HASH_BITS 8
56 #define XFRMI_HASH_SIZE BIT(XFRMI_HASH_BITS)
59 /* lists for storing interfaces in use */
60 struct xfrm_if __rcu *xfrmi[XFRMI_HASH_SIZE];
61 struct xfrm_if __rcu *collect_md_xfrmi;
64 static const struct nla_policy xfrm_lwt_policy[LWT_XFRM_MAX + 1] = {
65 [LWT_XFRM_IF_ID] = NLA_POLICY_MIN(NLA_U32, 1),
66 [LWT_XFRM_LINK] = NLA_POLICY_MIN(NLA_U32, 1),
69 static void xfrmi_destroy_state(struct lwtunnel_state *lwt)
73 static int xfrmi_build_state(struct net *net, struct nlattr *nla,
74 unsigned int family, const void *cfg,
75 struct lwtunnel_state **ts,
76 struct netlink_ext_ack *extack)
78 struct nlattr *tb[LWT_XFRM_MAX + 1];
79 struct lwtunnel_state *new_state;
80 struct xfrm_md_info *info;
83 ret = nla_parse_nested(tb, LWT_XFRM_MAX, nla, xfrm_lwt_policy, extack);
87 if (!tb[LWT_XFRM_IF_ID]) {
88 NL_SET_ERR_MSG(extack, "if_id must be set");
92 new_state = lwtunnel_state_alloc(sizeof(*info));
94 NL_SET_ERR_MSG(extack, "failed to create encap info");
98 new_state->type = LWTUNNEL_ENCAP_XFRM;
100 info = lwt_xfrm_info(new_state);
102 info->if_id = nla_get_u32(tb[LWT_XFRM_IF_ID]);
104 if (tb[LWT_XFRM_LINK])
105 info->link = nla_get_u32(tb[LWT_XFRM_LINK]);
111 static int xfrmi_fill_encap_info(struct sk_buff *skb,
112 struct lwtunnel_state *lwt)
114 struct xfrm_md_info *info = lwt_xfrm_info(lwt);
116 if (nla_put_u32(skb, LWT_XFRM_IF_ID, info->if_id) ||
117 (info->link && nla_put_u32(skb, LWT_XFRM_LINK, info->link)))
123 static int xfrmi_encap_nlsize(struct lwtunnel_state *lwtstate)
125 return nla_total_size(sizeof(u32)) + /* LWT_XFRM_IF_ID */
126 nla_total_size(sizeof(u32)); /* LWT_XFRM_LINK */
129 static int xfrmi_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
131 struct xfrm_md_info *a_info = lwt_xfrm_info(a);
132 struct xfrm_md_info *b_info = lwt_xfrm_info(b);
134 return memcmp(a_info, b_info, sizeof(*a_info));
137 static const struct lwtunnel_encap_ops xfrmi_encap_ops = {
138 .build_state = xfrmi_build_state,
139 .destroy_state = xfrmi_destroy_state,
140 .fill_encap = xfrmi_fill_encap_info,
141 .get_encap_size = xfrmi_encap_nlsize,
142 .cmp_encap = xfrmi_encap_cmp,
143 .owner = THIS_MODULE,
146 #define for_each_xfrmi_rcu(start, xi) \
147 for (xi = rcu_dereference(start); xi; xi = rcu_dereference(xi->next))
149 static u32 xfrmi_hash(u32 if_id)
151 return hash_32(if_id, XFRMI_HASH_BITS);
154 static struct xfrm_if *xfrmi_lookup(struct net *net, struct xfrm_state *x)
156 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
159 for_each_xfrmi_rcu(xfrmn->xfrmi[xfrmi_hash(x->if_id)], xi) {
160 if (x->if_id == xi->p.if_id &&
161 (xi->dev->flags & IFF_UP))
165 xi = rcu_dereference(xfrmn->collect_md_xfrmi);
166 if (xi && (xi->dev->flags & IFF_UP))
172 static bool xfrmi_decode_session(struct sk_buff *skb,
173 unsigned short family,
174 struct xfrm_if_decode_session_result *res)
176 struct net_device *dev;
180 if (!secpath_exists(skb) || !skb->dev)
185 ifindex = inet6_sdif(skb);
188 ifindex = inet_sdif(skb);
193 struct net *net = xs_net(xfrm_input_state(skb));
195 dev = dev_get_by_index_rcu(net, ifindex);
200 if (!dev || !(dev->flags & IFF_UP))
202 if (dev->netdev_ops != &xfrmi_netdev_ops)
205 xi = netdev_priv(dev);
208 if (xi->p.collect_md)
209 res->if_id = xfrm_input_state(skb)->if_id;
211 res->if_id = xi->p.if_id;
215 static void xfrmi_link(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
217 struct xfrm_if __rcu **xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
219 rcu_assign_pointer(xi->next , rtnl_dereference(*xip));
220 rcu_assign_pointer(*xip, xi);
223 static void xfrmi_unlink(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
225 struct xfrm_if __rcu **xip;
226 struct xfrm_if *iter;
228 for (xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
229 (iter = rtnl_dereference(*xip)) != NULL;
232 rcu_assign_pointer(*xip, xi->next);
238 static void xfrmi_dev_free(struct net_device *dev)
240 struct xfrm_if *xi = netdev_priv(dev);
242 gro_cells_destroy(&xi->gro_cells);
245 static int xfrmi_create(struct net_device *dev)
247 struct xfrm_if *xi = netdev_priv(dev);
248 struct net *net = dev_net(dev);
249 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
252 dev->rtnl_link_ops = &xfrmi_link_ops;
253 err = register_netdevice(dev);
257 if (xi->p.collect_md)
258 rcu_assign_pointer(xfrmn->collect_md_xfrmi, xi);
260 xfrmi_link(xfrmn, xi);
268 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
270 struct xfrm_if __rcu **xip;
272 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
274 for (xip = &xfrmn->xfrmi[xfrmi_hash(p->if_id)];
275 (xi = rtnl_dereference(*xip)) != NULL;
277 if (xi->p.if_id == p->if_id)
283 static void xfrmi_dev_uninit(struct net_device *dev)
285 struct xfrm_if *xi = netdev_priv(dev);
286 struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
288 if (xi->p.collect_md)
289 RCU_INIT_POINTER(xfrmn->collect_md_xfrmi, NULL);
291 xfrmi_unlink(xfrmn, xi);
294 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
296 skb_clear_tstamp(skb);
297 skb->pkt_type = PACKET_HOST;
313 static int xfrmi_input(struct sk_buff *skb, int nexthdr, __be32 spi,
314 int encap_type, unsigned short family)
318 sp = skb_sec_path(skb);
319 if (sp && (sp->len || sp->olen) &&
320 !xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
323 XFRM_SPI_SKB_CB(skb)->family = family;
324 if (family == AF_INET) {
325 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
326 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = NULL;
328 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
329 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = NULL;
332 return xfrm_input(skb, nexthdr, spi, encap_type);
338 static int xfrmi4_rcv(struct sk_buff *skb)
340 return xfrmi_input(skb, ip_hdr(skb)->protocol, 0, 0, AF_INET);
343 static int xfrmi6_rcv(struct sk_buff *skb)
345 return xfrmi_input(skb, skb_network_header(skb)[IP6CB(skb)->nhoff],
349 static int xfrmi4_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
351 return xfrmi_input(skb, nexthdr, spi, encap_type, AF_INET);
354 static int xfrmi6_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
356 return xfrmi_input(skb, nexthdr, spi, encap_type, AF_INET6);
359 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
361 const struct xfrm_mode *inner_mode;
362 struct net_device *dev;
363 struct xfrm_state *x;
368 if (err && !secpath_exists(skb))
371 x = xfrm_input_state(skb);
373 xi = xfrmi_lookup(xs_net(x), x);
377 link = skb->dev->ifindex;
382 DEV_STATS_INC(dev, rx_errors);
383 DEV_STATS_INC(dev, rx_dropped);
388 xnet = !net_eq(xi->net, dev_net(skb->dev));
391 inner_mode = &x->inner_mode;
393 if (x->sel.family == AF_UNSPEC) {
394 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
395 if (inner_mode == NULL) {
396 XFRM_INC_STATS(dev_net(skb->dev),
397 LINUX_MIB_XFRMINSTATEMODEERROR);
402 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
407 xfrmi_scrub_packet(skb, xnet);
408 if (xi->p.collect_md) {
409 struct metadata_dst *md_dst;
411 md_dst = metadata_dst_alloc(0, METADATA_XFRM, GFP_ATOMIC);
415 md_dst->u.xfrm_info.if_id = x->if_id;
416 md_dst->u.xfrm_info.link = link;
417 skb_dst_set(skb, (struct dst_entry *)md_dst);
419 dev_sw_netstats_rx_add(dev, skb->len);
425 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
427 struct xfrm_if *xi = netdev_priv(dev);
428 struct dst_entry *dst = skb_dst(skb);
429 unsigned int length = skb->len;
430 struct net_device *tdev;
431 struct xfrm_state *x;
436 if (xi->p.collect_md) {
437 struct xfrm_md_info *md_info = skb_xfrm_md_info(skb);
439 if (unlikely(!md_info))
442 if_id = md_info->if_id;
443 fl->flowi_oif = md_info->link;
444 if (md_info->dst_orig) {
445 struct dst_entry *tmp_dst = dst;
447 dst = md_info->dst_orig;
448 skb_dst_set(skb, dst);
449 md_info->dst_orig = NULL;
450 dst_release(tmp_dst);
457 dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, if_id);
461 goto tx_err_link_failure;
466 goto tx_err_link_failure;
468 if (x->if_id != if_id)
469 goto tx_err_link_failure;
474 DEV_STATS_INC(dev, collisions);
475 net_warn_ratelimited("%s: Local routing loop detected!\n",
477 goto tx_err_dst_release;
481 if ((!skb_is_gso(skb) && skb->len > mtu) ||
482 (skb_is_gso(skb) && !skb_gso_validate_network_len(skb, mtu))) {
483 skb_dst_update_pmtu_no_confirm(skb, mtu);
485 if (skb->protocol == htons(ETH_P_IPV6)) {
486 if (mtu < IPV6_MIN_MTU)
490 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
494 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
496 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
505 xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
506 skb_dst_set(skb, dst);
509 err = dst_output(xi->net, skb->sk, skb);
510 if (net_xmit_eval(err) == 0) {
511 dev_sw_netstats_tx_add(dev, 1, length);
513 DEV_STATS_INC(dev, tx_errors);
514 DEV_STATS_INC(dev, tx_aborted_errors);
519 DEV_STATS_INC(dev, tx_carrier_errors);
520 dst_link_failure(skb);
526 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
528 struct xfrm_if *xi = netdev_priv(dev);
529 struct dst_entry *dst = skb_dst(skb);
533 memset(&fl, 0, sizeof(fl));
535 switch (skb->protocol) {
536 case htons(ETH_P_IPV6):
537 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
538 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
540 fl.u.ip6.flowi6_oif = dev->ifindex;
541 fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
542 dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
545 DEV_STATS_INC(dev, tx_carrier_errors);
548 skb_dst_set(skb, dst);
551 case htons(ETH_P_IP):
552 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
553 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
557 fl.u.ip4.flowi4_oif = dev->ifindex;
558 fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
559 rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
561 DEV_STATS_INC(dev, tx_carrier_errors);
564 skb_dst_set(skb, &rt->dst);
571 fl.flowi_oif = xi->p.link;
573 ret = xfrmi_xmit2(skb, dev, &fl);
580 DEV_STATS_INC(dev, tx_errors);
581 DEV_STATS_INC(dev, tx_dropped);
586 static int xfrmi4_err(struct sk_buff *skb, u32 info)
588 const struct iphdr *iph = (const struct iphdr *)skb->data;
589 struct net *net = dev_net(skb->dev);
590 int protocol = iph->protocol;
591 struct ip_comp_hdr *ipch;
592 struct ip_esp_hdr *esph;
593 struct ip_auth_hdr *ah ;
594 struct xfrm_state *x;
600 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
604 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
608 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
609 spi = htonl(ntohs(ipch->cpi));
615 switch (icmp_hdr(skb)->type) {
616 case ICMP_DEST_UNREACH:
617 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
626 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
627 spi, protocol, AF_INET);
631 xi = xfrmi_lookup(net, x);
637 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
638 ipv4_update_pmtu(skb, net, info, 0, protocol);
640 ipv4_redirect(skb, net, 0, protocol);
646 static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
647 u8 type, u8 code, int offset, __be32 info)
649 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
650 struct net *net = dev_net(skb->dev);
651 int protocol = iph->nexthdr;
652 struct ip_comp_hdr *ipch;
653 struct ip_esp_hdr *esph;
654 struct ip_auth_hdr *ah;
655 struct xfrm_state *x;
661 esph = (struct ip_esp_hdr *)(skb->data + offset);
665 ah = (struct ip_auth_hdr *)(skb->data + offset);
669 ipch = (struct ip_comp_hdr *)(skb->data + offset);
670 spi = htonl(ntohs(ipch->cpi));
676 if (type != ICMPV6_PKT_TOOBIG &&
677 type != NDISC_REDIRECT)
680 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
681 spi, protocol, AF_INET6);
685 xi = xfrmi_lookup(net, x);
691 if (type == NDISC_REDIRECT)
692 ip6_redirect(skb, net, skb->dev->ifindex, 0,
693 sock_net_uid(net, NULL));
695 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
701 static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
703 if (xi->p.link != p->link)
706 xi->p.if_id = p->if_id;
711 static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
713 struct net *net = xi->net;
714 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
717 xfrmi_unlink(xfrmn, xi);
719 err = xfrmi_change(xi, p);
720 xfrmi_link(xfrmn, xi);
721 netdev_state_change(xi->dev);
725 static int xfrmi_get_iflink(const struct net_device *dev)
727 struct xfrm_if *xi = netdev_priv(dev);
729 return READ_ONCE(xi->p.link);
732 static const struct net_device_ops xfrmi_netdev_ops = {
733 .ndo_init = xfrmi_dev_init,
734 .ndo_uninit = xfrmi_dev_uninit,
735 .ndo_start_xmit = xfrmi_xmit,
736 .ndo_get_stats64 = dev_get_tstats64,
737 .ndo_get_iflink = xfrmi_get_iflink,
740 static void xfrmi_dev_setup(struct net_device *dev)
742 dev->netdev_ops = &xfrmi_netdev_ops;
743 dev->header_ops = &ip_tunnel_header_ops;
744 dev->type = ARPHRD_NONE;
745 dev->mtu = ETH_DATA_LEN;
746 dev->min_mtu = ETH_MIN_MTU;
747 dev->max_mtu = IP_MAX_MTU;
748 dev->flags = IFF_NOARP;
749 dev->needs_free_netdev = true;
750 dev->priv_destructor = xfrmi_dev_free;
751 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
754 eth_broadcast_addr(dev->broadcast);
757 #define XFRMI_FEATURES (NETIF_F_SG | \
759 NETIF_F_GSO_SOFTWARE | \
762 static int xfrmi_dev_init(struct net_device *dev)
764 struct xfrm_if *xi = netdev_priv(dev);
765 struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
768 err = gro_cells_init(&xi->gro_cells, dev);
772 dev->features |= NETIF_F_LLTX;
773 dev->features |= XFRMI_FEATURES;
774 dev->hw_features |= XFRMI_FEATURES;
777 dev->needed_headroom = phydev->needed_headroom;
778 dev->needed_tailroom = phydev->needed_tailroom;
780 if (is_zero_ether_addr(dev->dev_addr))
781 eth_hw_addr_inherit(dev, phydev);
782 if (is_zero_ether_addr(dev->broadcast))
783 memcpy(dev->broadcast, phydev->broadcast,
786 eth_hw_addr_random(dev);
787 eth_broadcast_addr(dev->broadcast);
793 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
794 struct netlink_ext_ack *extack)
799 static void xfrmi_netlink_parms(struct nlattr *data[],
800 struct xfrm_if_parms *parms)
802 memset(parms, 0, sizeof(*parms));
807 if (data[IFLA_XFRM_LINK])
808 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
810 if (data[IFLA_XFRM_IF_ID])
811 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
813 if (data[IFLA_XFRM_COLLECT_METADATA])
814 parms->collect_md = true;
817 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
818 struct nlattr *tb[], struct nlattr *data[],
819 struct netlink_ext_ack *extack)
821 struct net *net = dev_net(dev);
822 struct xfrm_if_parms p = {};
826 xfrmi_netlink_parms(data, &p);
828 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
830 if (p.link || p.if_id) {
831 NL_SET_ERR_MSG(extack, "link and if_id must be zero");
835 if (rtnl_dereference(xfrmn->collect_md_xfrmi))
840 NL_SET_ERR_MSG(extack, "if_id must be non zero");
844 xi = xfrmi_locate(net, &p);
849 xi = netdev_priv(dev);
854 err = xfrmi_create(dev);
858 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
860 unregister_netdevice_queue(dev, head);
863 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
864 struct nlattr *data[],
865 struct netlink_ext_ack *extack)
867 struct xfrm_if *xi = netdev_priv(dev);
868 struct net *net = xi->net;
869 struct xfrm_if_parms p = {};
871 xfrmi_netlink_parms(data, &p);
873 NL_SET_ERR_MSG(extack, "if_id must be non zero");
878 NL_SET_ERR_MSG(extack, "collect_md can't be changed");
882 xi = xfrmi_locate(net, &p);
884 xi = netdev_priv(dev);
888 if (xi->p.collect_md) {
889 NL_SET_ERR_MSG(extack,
890 "device can't be changed to collect_md");
895 return xfrmi_update(xi, &p);
898 static size_t xfrmi_get_size(const struct net_device *dev)
903 /* IFLA_XFRM_IF_ID */
905 /* IFLA_XFRM_COLLECT_METADATA */
910 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
912 struct xfrm_if *xi = netdev_priv(dev);
913 struct xfrm_if_parms *parm = &xi->p;
915 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
916 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id) ||
917 (xi->p.collect_md && nla_put_flag(skb, IFLA_XFRM_COLLECT_METADATA)))
918 goto nla_put_failure;
925 static struct net *xfrmi_get_link_net(const struct net_device *dev)
927 struct xfrm_if *xi = netdev_priv(dev);
929 return READ_ONCE(xi->net);
932 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
933 [IFLA_XFRM_UNSPEC] = { .strict_start_type = IFLA_XFRM_COLLECT_METADATA },
934 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
935 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
936 [IFLA_XFRM_COLLECT_METADATA] = { .type = NLA_FLAG },
939 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
941 .maxtype = IFLA_XFRM_MAX,
942 .policy = xfrmi_policy,
943 .priv_size = sizeof(struct xfrm_if),
944 .setup = xfrmi_dev_setup,
945 .validate = xfrmi_validate,
946 .newlink = xfrmi_newlink,
947 .dellink = xfrmi_dellink,
948 .changelink = xfrmi_changelink,
949 .get_size = xfrmi_get_size,
950 .fill_info = xfrmi_fill_info,
951 .get_link_net = xfrmi_get_link_net,
954 static void __net_exit xfrmi_exit_batch_rtnl(struct list_head *net_exit_list,
955 struct list_head *dev_to_kill)
960 list_for_each_entry(net, net_exit_list, exit_list) {
961 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
962 struct xfrm_if __rcu **xip;
966 for (i = 0; i < XFRMI_HASH_SIZE; i++) {
967 for (xip = &xfrmn->xfrmi[i];
968 (xi = rtnl_dereference(*xip)) != NULL;
970 unregister_netdevice_queue(xi->dev, dev_to_kill);
972 xi = rtnl_dereference(xfrmn->collect_md_xfrmi);
974 unregister_netdevice_queue(xi->dev, dev_to_kill);
978 static struct pernet_operations xfrmi_net_ops = {
979 .exit_batch_rtnl = xfrmi_exit_batch_rtnl,
981 .size = sizeof(struct xfrmi_net),
984 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
985 .handler = xfrmi6_rcv,
986 .input_handler = xfrmi6_input,
987 .cb_handler = xfrmi_rcv_cb,
988 .err_handler = xfrmi6_err,
992 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
993 .handler = xfrm6_rcv,
994 .input_handler = xfrm_input,
995 .cb_handler = xfrmi_rcv_cb,
996 .err_handler = xfrmi6_err,
1000 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
1001 .handler = xfrm6_rcv,
1002 .input_handler = xfrm_input,
1003 .cb_handler = xfrmi_rcv_cb,
1004 .err_handler = xfrmi6_err,
1008 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1009 static int xfrmi6_rcv_tunnel(struct sk_buff *skb)
1011 const xfrm_address_t *saddr;
1014 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1015 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1017 return xfrm6_rcv_spi(skb, IPPROTO_IPV6, spi, NULL);
1020 static struct xfrm6_tunnel xfrmi_ipv6_handler __read_mostly = {
1021 .handler = xfrmi6_rcv_tunnel,
1022 .cb_handler = xfrmi_rcv_cb,
1023 .err_handler = xfrmi6_err,
1027 static struct xfrm6_tunnel xfrmi_ip6ip_handler __read_mostly = {
1028 .handler = xfrmi6_rcv_tunnel,
1029 .cb_handler = xfrmi_rcv_cb,
1030 .err_handler = xfrmi6_err,
1035 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
1036 .handler = xfrmi4_rcv,
1037 .input_handler = xfrmi4_input,
1038 .cb_handler = xfrmi_rcv_cb,
1039 .err_handler = xfrmi4_err,
1043 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
1044 .handler = xfrm4_rcv,
1045 .input_handler = xfrm_input,
1046 .cb_handler = xfrmi_rcv_cb,
1047 .err_handler = xfrmi4_err,
1051 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
1052 .handler = xfrm4_rcv,
1053 .input_handler = xfrm_input,
1054 .cb_handler = xfrmi_rcv_cb,
1055 .err_handler = xfrmi4_err,
1059 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1060 static int xfrmi4_rcv_tunnel(struct sk_buff *skb)
1062 return xfrm4_rcv_spi(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr);
1065 static struct xfrm_tunnel xfrmi_ipip_handler __read_mostly = {
1066 .handler = xfrmi4_rcv_tunnel,
1067 .cb_handler = xfrmi_rcv_cb,
1068 .err_handler = xfrmi4_err,
1072 static struct xfrm_tunnel xfrmi_ipip6_handler __read_mostly = {
1073 .handler = xfrmi4_rcv_tunnel,
1074 .cb_handler = xfrmi_rcv_cb,
1075 .err_handler = xfrmi4_err,
1080 static int __init xfrmi4_init(void)
1084 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
1086 goto xfrm_proto_esp_failed;
1087 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
1089 goto xfrm_proto_ah_failed;
1090 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1092 goto xfrm_proto_comp_failed;
1093 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1094 err = xfrm4_tunnel_register(&xfrmi_ipip_handler, AF_INET);
1096 goto xfrm_tunnel_ipip_failed;
1097 err = xfrm4_tunnel_register(&xfrmi_ipip6_handler, AF_INET6);
1099 goto xfrm_tunnel_ipip6_failed;
1104 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1105 xfrm_tunnel_ipip6_failed:
1106 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
1107 xfrm_tunnel_ipip_failed:
1108 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1110 xfrm_proto_comp_failed:
1111 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
1112 xfrm_proto_ah_failed:
1113 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
1114 xfrm_proto_esp_failed:
1118 static void xfrmi4_fini(void)
1120 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1121 xfrm4_tunnel_deregister(&xfrmi_ipip6_handler, AF_INET6);
1122 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
1124 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1125 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
1126 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
1129 static int __init xfrmi6_init(void)
1133 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
1135 goto xfrm_proto_esp_failed;
1136 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
1138 goto xfrm_proto_ah_failed;
1139 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1141 goto xfrm_proto_comp_failed;
1142 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1143 err = xfrm6_tunnel_register(&xfrmi_ipv6_handler, AF_INET6);
1145 goto xfrm_tunnel_ipv6_failed;
1146 err = xfrm6_tunnel_register(&xfrmi_ip6ip_handler, AF_INET);
1148 goto xfrm_tunnel_ip6ip_failed;
1153 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1154 xfrm_tunnel_ip6ip_failed:
1155 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
1156 xfrm_tunnel_ipv6_failed:
1157 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1159 xfrm_proto_comp_failed:
1160 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
1161 xfrm_proto_ah_failed:
1162 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
1163 xfrm_proto_esp_failed:
1167 static void xfrmi6_fini(void)
1169 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1170 xfrm6_tunnel_deregister(&xfrmi_ip6ip_handler, AF_INET);
1171 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
1173 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1174 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
1175 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
1178 static const struct xfrm_if_cb xfrm_if_cb = {
1179 .decode_session = xfrmi_decode_session,
1182 static int __init xfrmi_init(void)
1187 pr_info("IPsec XFRM device driver\n");
1189 msg = "tunnel device";
1190 err = register_pernet_device(&xfrmi_net_ops);
1192 goto pernet_dev_failed;
1194 msg = "xfrm4 protocols";
1195 err = xfrmi4_init();
1199 msg = "xfrm6 protocols";
1200 err = xfrmi6_init();
1205 msg = "netlink interface";
1206 err = rtnl_link_register(&xfrmi_link_ops);
1208 goto rtnl_link_failed;
1210 err = register_xfrm_interface_bpf();
1214 lwtunnel_encap_add_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
1216 xfrm_if_register_cb(&xfrm_if_cb);
1221 rtnl_link_unregister(&xfrmi_link_ops);
1227 unregister_pernet_device(&xfrmi_net_ops);
1229 pr_err("xfrmi init: failed to register %s\n", msg);
1233 static void __exit xfrmi_fini(void)
1235 xfrm_if_unregister_cb();
1236 lwtunnel_encap_del_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
1237 rtnl_link_unregister(&xfrmi_link_ops);
1240 unregister_pernet_device(&xfrmi_net_ops);
1243 module_init(xfrmi_init);
1244 module_exit(xfrmi_fini);
1245 MODULE_LICENSE("GPL");
1246 MODULE_ALIAS_RTNL_LINK("xfrm");
1247 MODULE_ALIAS_NETDEV("xfrm0");
1248 MODULE_AUTHOR("Steffen Klassert");
1249 MODULE_DESCRIPTION("XFRM virtual interface");