1 // SPDX-License-Identifier: GPL-2.0
3 * XFRM virtual interface
5 * Copyright (C) 2018 secunet Security Networks AG
8 * Steffen Klassert <steffen.klassert@secunet.com>
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);
243 free_percpu(dev->tstats);
246 static int xfrmi_create(struct net_device *dev)
248 struct xfrm_if *xi = netdev_priv(dev);
249 struct net *net = dev_net(dev);
250 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
253 dev->rtnl_link_ops = &xfrmi_link_ops;
254 err = register_netdevice(dev);
258 if (xi->p.collect_md)
259 rcu_assign_pointer(xfrmn->collect_md_xfrmi, xi);
261 xfrmi_link(xfrmn, xi);
269 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
271 struct xfrm_if __rcu **xip;
273 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
275 for (xip = &xfrmn->xfrmi[xfrmi_hash(p->if_id)];
276 (xi = rtnl_dereference(*xip)) != NULL;
278 if (xi->p.if_id == p->if_id)
284 static void xfrmi_dev_uninit(struct net_device *dev)
286 struct xfrm_if *xi = netdev_priv(dev);
287 struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
289 if (xi->p.collect_md)
290 RCU_INIT_POINTER(xfrmn->collect_md_xfrmi, NULL);
292 xfrmi_unlink(xfrmn, xi);
295 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
297 skb_clear_tstamp(skb);
298 skb->pkt_type = PACKET_HOST;
314 static int xfrmi_input(struct sk_buff *skb, int nexthdr, __be32 spi,
315 int encap_type, unsigned short family)
319 sp = skb_sec_path(skb);
320 if (sp && (sp->len || sp->olen) &&
321 !xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
324 XFRM_SPI_SKB_CB(skb)->family = family;
325 if (family == AF_INET) {
326 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
327 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = NULL;
329 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
330 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = NULL;
333 return xfrm_input(skb, nexthdr, spi, encap_type);
339 static int xfrmi4_rcv(struct sk_buff *skb)
341 return xfrmi_input(skb, ip_hdr(skb)->protocol, 0, 0, AF_INET);
344 static int xfrmi6_rcv(struct sk_buff *skb)
346 return xfrmi_input(skb, skb_network_header(skb)[IP6CB(skb)->nhoff],
350 static int xfrmi4_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
352 return xfrmi_input(skb, nexthdr, spi, encap_type, AF_INET);
355 static int xfrmi6_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
357 return xfrmi_input(skb, nexthdr, spi, encap_type, AF_INET6);
360 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
362 const struct xfrm_mode *inner_mode;
363 struct net_device *dev;
364 struct xfrm_state *x;
369 if (err && !secpath_exists(skb))
372 x = xfrm_input_state(skb);
374 xi = xfrmi_lookup(xs_net(x), x);
378 link = skb->dev->ifindex;
383 dev->stats.rx_errors++;
384 dev->stats.rx_dropped++;
389 xnet = !net_eq(xi->net, dev_net(skb->dev));
392 inner_mode = &x->inner_mode;
394 if (x->sel.family == AF_UNSPEC) {
395 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
396 if (inner_mode == NULL) {
397 XFRM_INC_STATS(dev_net(skb->dev),
398 LINUX_MIB_XFRMINSTATEMODEERROR);
403 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
408 xfrmi_scrub_packet(skb, xnet);
409 if (xi->p.collect_md) {
410 struct metadata_dst *md_dst;
412 md_dst = metadata_dst_alloc(0, METADATA_XFRM, GFP_ATOMIC);
416 md_dst->u.xfrm_info.if_id = x->if_id;
417 md_dst->u.xfrm_info.link = link;
418 skb_dst_set(skb, (struct dst_entry *)md_dst);
420 dev_sw_netstats_rx_add(dev, skb->len);
426 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
428 struct xfrm_if *xi = netdev_priv(dev);
429 struct net_device_stats *stats = &xi->dev->stats;
430 struct dst_entry *dst = skb_dst(skb);
431 unsigned int length = skb->len;
432 struct net_device *tdev;
433 struct xfrm_state *x;
438 if (xi->p.collect_md) {
439 struct xfrm_md_info *md_info = skb_xfrm_md_info(skb);
441 if (unlikely(!md_info))
444 if_id = md_info->if_id;
445 fl->flowi_oif = md_info->link;
446 if (md_info->dst_orig) {
447 struct dst_entry *tmp_dst = dst;
449 dst = md_info->dst_orig;
450 skb_dst_set(skb, dst);
451 md_info->dst_orig = NULL;
452 dst_release(tmp_dst);
459 dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, if_id);
463 goto tx_err_link_failure;
468 goto tx_err_link_failure;
470 if (x->if_id != if_id)
471 goto tx_err_link_failure;
477 net_warn_ratelimited("%s: Local routing loop detected!\n",
479 goto tx_err_dst_release;
483 if ((!skb_is_gso(skb) && skb->len > mtu) ||
484 (skb_is_gso(skb) && !skb_gso_validate_network_len(skb, mtu))) {
485 skb_dst_update_pmtu_no_confirm(skb, mtu);
487 if (skb->protocol == htons(ETH_P_IPV6)) {
488 if (mtu < IPV6_MIN_MTU)
492 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
496 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
498 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
507 xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
508 skb_dst_set(skb, dst);
511 err = dst_output(xi->net, skb->sk, skb);
512 if (net_xmit_eval(err) == 0) {
513 dev_sw_netstats_tx_add(dev, 1, length);
516 stats->tx_aborted_errors++;
521 stats->tx_carrier_errors++;
522 dst_link_failure(skb);
528 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
530 struct xfrm_if *xi = netdev_priv(dev);
531 struct net_device_stats *stats = &xi->dev->stats;
532 struct dst_entry *dst = skb_dst(skb);
536 memset(&fl, 0, sizeof(fl));
538 switch (skb->protocol) {
539 case htons(ETH_P_IPV6):
540 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
541 xfrm_decode_session(skb, &fl, AF_INET6);
543 fl.u.ip6.flowi6_oif = dev->ifindex;
544 fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
545 dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
548 stats->tx_carrier_errors++;
551 skb_dst_set(skb, dst);
554 case htons(ETH_P_IP):
555 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
556 xfrm_decode_session(skb, &fl, AF_INET);
560 fl.u.ip4.flowi4_oif = dev->ifindex;
561 fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
562 rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
564 stats->tx_carrier_errors++;
567 skb_dst_set(skb, &rt->dst);
574 fl.flowi_oif = xi->p.link;
576 ret = xfrmi_xmit2(skb, dev, &fl);
589 static int xfrmi4_err(struct sk_buff *skb, u32 info)
591 const struct iphdr *iph = (const struct iphdr *)skb->data;
592 struct net *net = dev_net(skb->dev);
593 int protocol = iph->protocol;
594 struct ip_comp_hdr *ipch;
595 struct ip_esp_hdr *esph;
596 struct ip_auth_hdr *ah ;
597 struct xfrm_state *x;
603 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
607 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
611 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
612 spi = htonl(ntohs(ipch->cpi));
618 switch (icmp_hdr(skb)->type) {
619 case ICMP_DEST_UNREACH:
620 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
629 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
630 spi, protocol, AF_INET);
634 xi = xfrmi_lookup(net, x);
640 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
641 ipv4_update_pmtu(skb, net, info, 0, protocol);
643 ipv4_redirect(skb, net, 0, protocol);
649 static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
650 u8 type, u8 code, int offset, __be32 info)
652 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
653 struct net *net = dev_net(skb->dev);
654 int protocol = iph->nexthdr;
655 struct ip_comp_hdr *ipch;
656 struct ip_esp_hdr *esph;
657 struct ip_auth_hdr *ah;
658 struct xfrm_state *x;
664 esph = (struct ip_esp_hdr *)(skb->data + offset);
668 ah = (struct ip_auth_hdr *)(skb->data + offset);
672 ipch = (struct ip_comp_hdr *)(skb->data + offset);
673 spi = htonl(ntohs(ipch->cpi));
679 if (type != ICMPV6_PKT_TOOBIG &&
680 type != NDISC_REDIRECT)
683 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
684 spi, protocol, AF_INET6);
688 xi = xfrmi_lookup(net, x);
694 if (type == NDISC_REDIRECT)
695 ip6_redirect(skb, net, skb->dev->ifindex, 0,
696 sock_net_uid(net, NULL));
698 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
704 static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
706 if (xi->p.link != p->link)
709 xi->p.if_id = p->if_id;
714 static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
716 struct net *net = xi->net;
717 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
720 xfrmi_unlink(xfrmn, xi);
722 err = xfrmi_change(xi, p);
723 xfrmi_link(xfrmn, xi);
724 netdev_state_change(xi->dev);
728 static int xfrmi_get_iflink(const struct net_device *dev)
730 struct xfrm_if *xi = netdev_priv(dev);
735 static const struct net_device_ops xfrmi_netdev_ops = {
736 .ndo_init = xfrmi_dev_init,
737 .ndo_uninit = xfrmi_dev_uninit,
738 .ndo_start_xmit = xfrmi_xmit,
739 .ndo_get_stats64 = dev_get_tstats64,
740 .ndo_get_iflink = xfrmi_get_iflink,
743 static void xfrmi_dev_setup(struct net_device *dev)
745 dev->netdev_ops = &xfrmi_netdev_ops;
746 dev->header_ops = &ip_tunnel_header_ops;
747 dev->type = ARPHRD_NONE;
748 dev->mtu = ETH_DATA_LEN;
749 dev->min_mtu = ETH_MIN_MTU;
750 dev->max_mtu = IP_MAX_MTU;
751 dev->flags = IFF_NOARP;
752 dev->needs_free_netdev = true;
753 dev->priv_destructor = xfrmi_dev_free;
756 eth_broadcast_addr(dev->broadcast);
759 #define XFRMI_FEATURES (NETIF_F_SG | \
761 NETIF_F_GSO_SOFTWARE | \
764 static int xfrmi_dev_init(struct net_device *dev)
766 struct xfrm_if *xi = netdev_priv(dev);
767 struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
770 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
774 err = gro_cells_init(&xi->gro_cells, dev);
776 free_percpu(dev->tstats);
780 dev->features |= NETIF_F_LLTX;
781 dev->features |= XFRMI_FEATURES;
782 dev->hw_features |= XFRMI_FEATURES;
785 dev->needed_headroom = phydev->needed_headroom;
786 dev->needed_tailroom = phydev->needed_tailroom;
788 if (is_zero_ether_addr(dev->dev_addr))
789 eth_hw_addr_inherit(dev, phydev);
790 if (is_zero_ether_addr(dev->broadcast))
791 memcpy(dev->broadcast, phydev->broadcast,
794 eth_hw_addr_random(dev);
795 eth_broadcast_addr(dev->broadcast);
801 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
802 struct netlink_ext_ack *extack)
807 static void xfrmi_netlink_parms(struct nlattr *data[],
808 struct xfrm_if_parms *parms)
810 memset(parms, 0, sizeof(*parms));
815 if (data[IFLA_XFRM_LINK])
816 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
818 if (data[IFLA_XFRM_IF_ID])
819 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
821 if (data[IFLA_XFRM_COLLECT_METADATA])
822 parms->collect_md = true;
825 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
826 struct nlattr *tb[], struct nlattr *data[],
827 struct netlink_ext_ack *extack)
829 struct net *net = dev_net(dev);
830 struct xfrm_if_parms p = {};
834 xfrmi_netlink_parms(data, &p);
836 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
838 if (p.link || p.if_id) {
839 NL_SET_ERR_MSG(extack, "link and if_id must be zero");
843 if (rtnl_dereference(xfrmn->collect_md_xfrmi))
848 NL_SET_ERR_MSG(extack, "if_id must be non zero");
852 xi = xfrmi_locate(net, &p);
857 xi = netdev_priv(dev);
862 err = xfrmi_create(dev);
866 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
868 unregister_netdevice_queue(dev, head);
871 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
872 struct nlattr *data[],
873 struct netlink_ext_ack *extack)
875 struct xfrm_if *xi = netdev_priv(dev);
876 struct net *net = xi->net;
877 struct xfrm_if_parms p = {};
879 xfrmi_netlink_parms(data, &p);
881 NL_SET_ERR_MSG(extack, "if_id must be non zero");
886 NL_SET_ERR_MSG(extack, "collect_md can't be changed");
890 xi = xfrmi_locate(net, &p);
892 xi = netdev_priv(dev);
896 if (xi->p.collect_md) {
897 NL_SET_ERR_MSG(extack,
898 "device can't be changed to collect_md");
903 return xfrmi_update(xi, &p);
906 static size_t xfrmi_get_size(const struct net_device *dev)
911 /* IFLA_XFRM_IF_ID */
913 /* IFLA_XFRM_COLLECT_METADATA */
918 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
920 struct xfrm_if *xi = netdev_priv(dev);
921 struct xfrm_if_parms *parm = &xi->p;
923 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
924 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id) ||
925 (xi->p.collect_md && nla_put_flag(skb, IFLA_XFRM_COLLECT_METADATA)))
926 goto nla_put_failure;
933 static struct net *xfrmi_get_link_net(const struct net_device *dev)
935 struct xfrm_if *xi = netdev_priv(dev);
940 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
941 [IFLA_XFRM_UNSPEC] = { .strict_start_type = IFLA_XFRM_COLLECT_METADATA },
942 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
943 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
944 [IFLA_XFRM_COLLECT_METADATA] = { .type = NLA_FLAG },
947 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
949 .maxtype = IFLA_XFRM_MAX,
950 .policy = xfrmi_policy,
951 .priv_size = sizeof(struct xfrm_if),
952 .setup = xfrmi_dev_setup,
953 .validate = xfrmi_validate,
954 .newlink = xfrmi_newlink,
955 .dellink = xfrmi_dellink,
956 .changelink = xfrmi_changelink,
957 .get_size = xfrmi_get_size,
958 .fill_info = xfrmi_fill_info,
959 .get_link_net = xfrmi_get_link_net,
962 static void __net_exit xfrmi_exit_batch_net(struct list_head *net_exit_list)
968 list_for_each_entry(net, net_exit_list, exit_list) {
969 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
970 struct xfrm_if __rcu **xip;
974 for (i = 0; i < XFRMI_HASH_SIZE; i++) {
975 for (xip = &xfrmn->xfrmi[i];
976 (xi = rtnl_dereference(*xip)) != NULL;
978 unregister_netdevice_queue(xi->dev, &list);
980 xi = rtnl_dereference(xfrmn->collect_md_xfrmi);
982 unregister_netdevice_queue(xi->dev, &list);
984 unregister_netdevice_many(&list);
988 static struct pernet_operations xfrmi_net_ops = {
989 .exit_batch = xfrmi_exit_batch_net,
991 .size = sizeof(struct xfrmi_net),
994 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
995 .handler = xfrmi6_rcv,
996 .input_handler = xfrmi6_input,
997 .cb_handler = xfrmi_rcv_cb,
998 .err_handler = xfrmi6_err,
1002 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
1003 .handler = xfrm6_rcv,
1004 .input_handler = xfrm_input,
1005 .cb_handler = xfrmi_rcv_cb,
1006 .err_handler = xfrmi6_err,
1010 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
1011 .handler = xfrm6_rcv,
1012 .input_handler = xfrm_input,
1013 .cb_handler = xfrmi_rcv_cb,
1014 .err_handler = xfrmi6_err,
1018 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1019 static int xfrmi6_rcv_tunnel(struct sk_buff *skb)
1021 const xfrm_address_t *saddr;
1024 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1025 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1027 return xfrm6_rcv_spi(skb, IPPROTO_IPV6, spi, NULL);
1030 static struct xfrm6_tunnel xfrmi_ipv6_handler __read_mostly = {
1031 .handler = xfrmi6_rcv_tunnel,
1032 .cb_handler = xfrmi_rcv_cb,
1033 .err_handler = xfrmi6_err,
1037 static struct xfrm6_tunnel xfrmi_ip6ip_handler __read_mostly = {
1038 .handler = xfrmi6_rcv_tunnel,
1039 .cb_handler = xfrmi_rcv_cb,
1040 .err_handler = xfrmi6_err,
1045 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
1046 .handler = xfrmi4_rcv,
1047 .input_handler = xfrmi4_input,
1048 .cb_handler = xfrmi_rcv_cb,
1049 .err_handler = xfrmi4_err,
1053 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
1054 .handler = xfrm4_rcv,
1055 .input_handler = xfrm_input,
1056 .cb_handler = xfrmi_rcv_cb,
1057 .err_handler = xfrmi4_err,
1061 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
1062 .handler = xfrm4_rcv,
1063 .input_handler = xfrm_input,
1064 .cb_handler = xfrmi_rcv_cb,
1065 .err_handler = xfrmi4_err,
1069 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1070 static int xfrmi4_rcv_tunnel(struct sk_buff *skb)
1072 return xfrm4_rcv_spi(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr);
1075 static struct xfrm_tunnel xfrmi_ipip_handler __read_mostly = {
1076 .handler = xfrmi4_rcv_tunnel,
1077 .cb_handler = xfrmi_rcv_cb,
1078 .err_handler = xfrmi4_err,
1082 static struct xfrm_tunnel xfrmi_ipip6_handler __read_mostly = {
1083 .handler = xfrmi4_rcv_tunnel,
1084 .cb_handler = xfrmi_rcv_cb,
1085 .err_handler = xfrmi4_err,
1090 static int __init xfrmi4_init(void)
1094 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
1096 goto xfrm_proto_esp_failed;
1097 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
1099 goto xfrm_proto_ah_failed;
1100 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1102 goto xfrm_proto_comp_failed;
1103 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1104 err = xfrm4_tunnel_register(&xfrmi_ipip_handler, AF_INET);
1106 goto xfrm_tunnel_ipip_failed;
1107 err = xfrm4_tunnel_register(&xfrmi_ipip6_handler, AF_INET6);
1109 goto xfrm_tunnel_ipip6_failed;
1114 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1115 xfrm_tunnel_ipip6_failed:
1116 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
1117 xfrm_tunnel_ipip_failed:
1118 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1120 xfrm_proto_comp_failed:
1121 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
1122 xfrm_proto_ah_failed:
1123 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
1124 xfrm_proto_esp_failed:
1128 static void xfrmi4_fini(void)
1130 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
1131 xfrm4_tunnel_deregister(&xfrmi_ipip6_handler, AF_INET6);
1132 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
1134 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
1135 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
1136 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
1139 static int __init xfrmi6_init(void)
1143 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
1145 goto xfrm_proto_esp_failed;
1146 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
1148 goto xfrm_proto_ah_failed;
1149 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1151 goto xfrm_proto_comp_failed;
1152 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1153 err = xfrm6_tunnel_register(&xfrmi_ipv6_handler, AF_INET6);
1155 goto xfrm_tunnel_ipv6_failed;
1156 err = xfrm6_tunnel_register(&xfrmi_ip6ip_handler, AF_INET);
1158 goto xfrm_tunnel_ip6ip_failed;
1163 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1164 xfrm_tunnel_ip6ip_failed:
1165 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
1166 xfrm_tunnel_ipv6_failed:
1167 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1169 xfrm_proto_comp_failed:
1170 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
1171 xfrm_proto_ah_failed:
1172 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
1173 xfrm_proto_esp_failed:
1177 static void xfrmi6_fini(void)
1179 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1180 xfrm6_tunnel_deregister(&xfrmi_ip6ip_handler, AF_INET);
1181 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
1183 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
1184 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
1185 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
1188 static const struct xfrm_if_cb xfrm_if_cb = {
1189 .decode_session = xfrmi_decode_session,
1192 static int __init xfrmi_init(void)
1197 pr_info("IPsec XFRM device driver\n");
1199 msg = "tunnel device";
1200 err = register_pernet_device(&xfrmi_net_ops);
1202 goto pernet_dev_failed;
1204 msg = "xfrm4 protocols";
1205 err = xfrmi4_init();
1209 msg = "xfrm6 protocols";
1210 err = xfrmi6_init();
1215 msg = "netlink interface";
1216 err = rtnl_link_register(&xfrmi_link_ops);
1218 goto rtnl_link_failed;
1220 err = register_xfrm_interface_bpf();
1224 lwtunnel_encap_add_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
1226 xfrm_if_register_cb(&xfrm_if_cb);
1231 rtnl_link_unregister(&xfrmi_link_ops);
1237 unregister_pernet_device(&xfrmi_net_ops);
1239 pr_err("xfrmi init: failed to register %s\n", msg);
1243 static void __exit xfrmi_fini(void)
1245 xfrm_if_unregister_cb();
1246 lwtunnel_encap_del_ops(&xfrmi_encap_ops, LWTUNNEL_ENCAP_XFRM);
1247 rtnl_link_unregister(&xfrmi_link_ops);
1250 unregister_pernet_device(&xfrmi_net_ops);
1253 module_init(xfrmi_init);
1254 module_exit(xfrmi_fini);
1255 MODULE_LICENSE("GPL");
1256 MODULE_ALIAS_RTNL_LINK("xfrm");
1257 MODULE_ALIAS_NETDEV("xfrm0");
1258 MODULE_AUTHOR("Steffen Klassert");
1259 MODULE_DESCRIPTION("XFRM virtual interface");