1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * XFRM virtual interface
4 *
5 * Copyright (C) 2018 secunet Security Networks AG
6 *
7 * Author:
8 * Steffen Klassert <steffen.klassert@secunet.com>
9 */
10
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>
17 #include <linux/if.h>
18 #include <linux/in.h>
19 #include <linux/ip.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>
32
33 #include <linux/uaccess.h>
34 #include <linux/atomic.h>
35
36 #include <net/icmp.h>
37 #include <net/ip.h>
38 #include <net/ipv6.h>
39 #include <net/ip6_route.h>
40 #include <net/ip_tunnels.h>
41 #include <net/addrconf.h>
42 #include <net/xfrm.h>
43 #include <net/net_namespace.h>
44 #include <net/netns/generic.h>
45 #include <linux/etherdevice.h>
46
47 static int xfrmi_dev_init(struct net_device *dev);
48 static void xfrmi_dev_setup(struct net_device *dev);
49 static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
50 static unsigned int xfrmi_net_id __read_mostly;
51 static const struct net_device_ops xfrmi_netdev_ops;
52
53 #define XFRMI_HASH_BITS 8
54 #define XFRMI_HASH_SIZE BIT(XFRMI_HASH_BITS)
55
56 struct xfrmi_net {
57 /* lists for storing interfaces in use */
58 struct xfrm_if __rcu *xfrmi[XFRMI_HASH_SIZE];
59 };
60
61 #define for_each_xfrmi_rcu(start, xi) \
62 for (xi = rcu_dereference(start); xi; xi = rcu_dereference(xi->next))
63
64 static u32 xfrmi_hash(u32 if_id)
65 {
66 return hash_32(if_id, XFRMI_HASH_BITS);
67 }
68
69 static struct xfrm_if *xfrmi_lookup(struct net *net, struct xfrm_state *x)
70 {
71 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
72 struct xfrm_if *xi;
73
74 for_each_xfrmi_rcu(xfrmn->xfrmi[xfrmi_hash(x->if_id)], xi) {
75 if (x->if_id == xi->p.if_id &&
76 (xi->dev->flags & IFF_UP))
77 return xi;
78 }
79
80 return NULL;
81 }
82
83 static struct xfrm_if *xfrmi_decode_session(struct sk_buff *skb,
84 unsigned short family)
85 {
86 struct net_device *dev;
87 int ifindex = 0;
88
89 if (!secpath_exists(skb) || !skb->dev)
90 return NULL;
91
92 switch (family) {
93 case AF_INET6:
94 ifindex = inet6_sdif(skb);
95 break;
96 case AF_INET:
97 ifindex = inet_sdif(skb);
98 break;
99 }
100
101 if (ifindex) {
102 struct net *net = xs_net(xfrm_input_state(skb));
103
104 dev = dev_get_by_index_rcu(net, ifindex);
105 } else {
106 dev = skb->dev;
107 }
108
109 if (!dev || !(dev->flags & IFF_UP))
110 return NULL;
111 if (dev->netdev_ops != &xfrmi_netdev_ops)
112 return NULL;
113
114 return netdev_priv(dev);
115 }
116
117 static void xfrmi_link(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
118 {
119 struct xfrm_if __rcu **xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
120
121 rcu_assign_pointer(xi->next , rtnl_dereference(*xip));
122 rcu_assign_pointer(*xip, xi);
123 }
124
125 static void xfrmi_unlink(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
126 {
127 struct xfrm_if __rcu **xip;
128 struct xfrm_if *iter;
129
130 for (xip = &xfrmn->xfrmi[xfrmi_hash(xi->p.if_id)];
131 (iter = rtnl_dereference(*xip)) != NULL;
132 xip = &iter->next) {
133 if (xi == iter) {
134 rcu_assign_pointer(*xip, xi->next);
135 break;
136 }
137 }
138 }
139
140 static void xfrmi_dev_free(struct net_device *dev)
141 {
142 struct xfrm_if *xi = netdev_priv(dev);
143
144 gro_cells_destroy(&xi->gro_cells);
145 free_percpu(dev->tstats);
146 }
147
148 static int xfrmi_create(struct net_device *dev)
149 {
150 struct xfrm_if *xi = netdev_priv(dev);
151 struct net *net = dev_net(dev);
152 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
153 int err;
154
155 dev->rtnl_link_ops = &xfrmi_link_ops;
156 err = register_netdevice(dev);
157 if (err < 0)
158 goto out;
159
160 xfrmi_link(xfrmn, xi);
161
162 return 0;
163
164 out:
165 return err;
166 }
167
168 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
169 {
170 struct xfrm_if __rcu **xip;
171 struct xfrm_if *xi;
172 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
173
174 for (xip = &xfrmn->xfrmi[xfrmi_hash(p->if_id)];
175 (xi = rtnl_dereference(*xip)) != NULL;
176 xip = &xi->next)
177 if (xi->p.if_id == p->if_id)
178 return xi;
179
180 return NULL;
181 }
182
183 static void xfrmi_dev_uninit(struct net_device *dev)
184 {
185 struct xfrm_if *xi = netdev_priv(dev);
186 struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
187
188 xfrmi_unlink(xfrmn, xi);
189 }
190
191 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
192 {
193 skb_clear_tstamp(skb);
194 skb->pkt_type = PACKET_HOST;
195 skb->skb_iif = 0;
196 skb->ignore_df = 0;
197 skb_dst_drop(skb);
198 nf_reset_ct(skb);
199 nf_reset_trace(skb);
200
201 if (!xnet)
202 return;
203
204 ipvs_reset(skb);
205 secpath_reset(skb);
206 skb_orphan(skb);
207 skb->mark = 0;
208 }
209
210 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
211 {
212 const struct xfrm_mode *inner_mode;
213 struct net_device *dev;
214 struct xfrm_state *x;
215 struct xfrm_if *xi;
216 bool xnet;
217
218 if (err && !secpath_exists(skb))
219 return 0;
220
221 x = xfrm_input_state(skb);
222
223 xi = xfrmi_lookup(xs_net(x), x);
224 if (!xi)
225 return 1;
226
227 dev = xi->dev;
228 skb->dev = dev;
229
230 if (err) {
231 dev->stats.rx_errors++;
232 dev->stats.rx_dropped++;
233
234 return 0;
235 }
236
237 xnet = !net_eq(xi->net, dev_net(skb->dev));
238
239 if (xnet) {
240 inner_mode = &x->inner_mode;
241
242 if (x->sel.family == AF_UNSPEC) {
243 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
244 if (inner_mode == NULL) {
245 XFRM_INC_STATS(dev_net(skb->dev),
246 LINUX_MIB_XFRMINSTATEMODEERROR);
247 return -EINVAL;
248 }
249 }
250
251 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
252 inner_mode->family))
253 return -EPERM;
254 }
255
256 xfrmi_scrub_packet(skb, xnet);
257 dev_sw_netstats_rx_add(dev, skb->len);
258
259 return 0;
260 }
261
262 static int
263 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
264 {
265 struct xfrm_if *xi = netdev_priv(dev);
266 struct net_device_stats *stats = &xi->dev->stats;
267 struct dst_entry *dst = skb_dst(skb);
268 unsigned int length = skb->len;
269 struct net_device *tdev;
270 struct xfrm_state *x;
271 int err = -1;
272 int mtu;
273
274 dst_hold(dst);
275 dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, xi->p.if_id);
276 if (IS_ERR(dst)) {
277 err = PTR_ERR(dst);
278 dst = NULL;
279 goto tx_err_link_failure;
280 }
281
282 x = dst->xfrm;
283 if (!x)
284 goto tx_err_link_failure;
285
286 if (x->if_id != xi->p.if_id)
287 goto tx_err_link_failure;
288
289 tdev = dst->dev;
290
291 if (tdev == dev) {
292 stats->collisions++;
293 net_warn_ratelimited("%s: Local routing loop detected!\n",
294 dev->name);
295 goto tx_err_dst_release;
296 }
297
298 mtu = dst_mtu(dst);
299 if ((!skb_is_gso(skb) && skb->len > mtu) ||
300 (skb_is_gso(skb) && !skb_gso_validate_network_len(skb, mtu))) {
301 skb_dst_update_pmtu_no_confirm(skb, mtu);
302
303 if (skb->protocol == htons(ETH_P_IPV6)) {
304 if (mtu < IPV6_MIN_MTU)
305 mtu = IPV6_MIN_MTU;
306
307 if (skb->len > 1280)
308 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
309 else
310 goto xmit;
311 } else {
312 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
313 goto xmit;
314 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
315 htonl(mtu));
316 }
317
318 dst_release(dst);
319 return -EMSGSIZE;
320 }
321
322 xmit:
323 xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
324 skb_dst_set(skb, dst);
325 skb->dev = tdev;
326
327 err = dst_output(xi->net, skb->sk, skb);
328 if (net_xmit_eval(err) == 0) {
329 dev_sw_netstats_tx_add(dev, 1, length);
330 } else {
331 stats->tx_errors++;
332 stats->tx_aborted_errors++;
333 }
334
335 return 0;
336 tx_err_link_failure:
337 stats->tx_carrier_errors++;
338 dst_link_failure(skb);
339 tx_err_dst_release:
340 dst_release(dst);
341 return err;
342 }
343
344 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
345 {
346 struct xfrm_if *xi = netdev_priv(dev);
347 struct net_device_stats *stats = &xi->dev->stats;
348 struct dst_entry *dst = skb_dst(skb);
349 struct flowi fl;
350 int ret;
351
352 memset(&fl, 0, sizeof(fl));
353
354 switch (skb->protocol) {
355 case htons(ETH_P_IPV6):
356 xfrm_decode_session(skb, &fl, AF_INET6);
357 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
358 if (!dst) {
359 fl.u.ip6.flowi6_oif = dev->ifindex;
360 fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
361 dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
362 if (dst->error) {
363 dst_release(dst);
364 stats->tx_carrier_errors++;
365 goto tx_err;
366 }
367 skb_dst_set(skb, dst);
368 }
369 break;
370 case htons(ETH_P_IP):
371 xfrm_decode_session(skb, &fl, AF_INET);
372 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
373 if (!dst) {
374 struct rtable *rt;
375
376 fl.u.ip4.flowi4_oif = dev->ifindex;
377 fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
378 rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
379 if (IS_ERR(rt)) {
380 stats->tx_carrier_errors++;
381 goto tx_err;
382 }
383 skb_dst_set(skb, &rt->dst);
384 }
385 break;
386 default:
387 goto tx_err;
388 }
389
390 fl.flowi_oif = xi->p.link;
391
392 ret = xfrmi_xmit2(skb, dev, &fl);
393 if (ret < 0)
394 goto tx_err;
395
396 return NETDEV_TX_OK;
397
398 tx_err:
399 stats->tx_errors++;
400 stats->tx_dropped++;
401 kfree_skb(skb);
402 return NETDEV_TX_OK;
403 }
404
405 static int xfrmi4_err(struct sk_buff *skb, u32 info)
406 {
407 const struct iphdr *iph = (const struct iphdr *)skb->data;
408 struct net *net = dev_net(skb->dev);
409 int protocol = iph->protocol;
410 struct ip_comp_hdr *ipch;
411 struct ip_esp_hdr *esph;
412 struct ip_auth_hdr *ah ;
413 struct xfrm_state *x;
414 struct xfrm_if *xi;
415 __be32 spi;
416
417 switch (protocol) {
418 case IPPROTO_ESP:
419 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
420 spi = esph->spi;
421 break;
422 case IPPROTO_AH:
423 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
424 spi = ah->spi;
425 break;
426 case IPPROTO_COMP:
427 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
428 spi = htonl(ntohs(ipch->cpi));
429 break;
430 default:
431 return 0;
432 }
433
434 switch (icmp_hdr(skb)->type) {
435 case ICMP_DEST_UNREACH:
436 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
437 return 0;
438 break;
439 case ICMP_REDIRECT:
440 break;
441 default:
442 return 0;
443 }
444
445 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
446 spi, protocol, AF_INET);
447 if (!x)
448 return 0;
449
450 xi = xfrmi_lookup(net, x);
451 if (!xi) {
452 xfrm_state_put(x);
453 return -1;
454 }
455
456 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
457 ipv4_update_pmtu(skb, net, info, 0, protocol);
458 else
459 ipv4_redirect(skb, net, 0, protocol);
460 xfrm_state_put(x);
461
462 return 0;
463 }
464
465 static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
466 u8 type, u8 code, int offset, __be32 info)
467 {
468 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
469 struct net *net = dev_net(skb->dev);
470 int protocol = iph->nexthdr;
471 struct ip_comp_hdr *ipch;
472 struct ip_esp_hdr *esph;
473 struct ip_auth_hdr *ah;
474 struct xfrm_state *x;
475 struct xfrm_if *xi;
476 __be32 spi;
477
478 switch (protocol) {
479 case IPPROTO_ESP:
480 esph = (struct ip_esp_hdr *)(skb->data + offset);
481 spi = esph->spi;
482 break;
483 case IPPROTO_AH:
484 ah = (struct ip_auth_hdr *)(skb->data + offset);
485 spi = ah->spi;
486 break;
487 case IPPROTO_COMP:
488 ipch = (struct ip_comp_hdr *)(skb->data + offset);
489 spi = htonl(ntohs(ipch->cpi));
490 break;
491 default:
492 return 0;
493 }
494
495 if (type != ICMPV6_PKT_TOOBIG &&
496 type != NDISC_REDIRECT)
497 return 0;
498
499 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
500 spi, protocol, AF_INET6);
501 if (!x)
502 return 0;
503
504 xi = xfrmi_lookup(net, x);
505 if (!xi) {
506 xfrm_state_put(x);
507 return -1;
508 }
509
510 if (type == NDISC_REDIRECT)
511 ip6_redirect(skb, net, skb->dev->ifindex, 0,
512 sock_net_uid(net, NULL));
513 else
514 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
515 xfrm_state_put(x);
516
517 return 0;
518 }
519
520 static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
521 {
522 if (xi->p.link != p->link)
523 return -EINVAL;
524
525 xi->p.if_id = p->if_id;
526
527 return 0;
528 }
529
530 static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
531 {
532 struct net *net = xi->net;
533 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
534 int err;
535
536 xfrmi_unlink(xfrmn, xi);
537 synchronize_net();
538 err = xfrmi_change(xi, p);
539 xfrmi_link(xfrmn, xi);
540 netdev_state_change(xi->dev);
541 return err;
542 }
543
544 static int xfrmi_get_iflink(const struct net_device *dev)
545 {
546 struct xfrm_if *xi = netdev_priv(dev);
547
548 return xi->p.link;
549 }
550
551 static const struct net_device_ops xfrmi_netdev_ops = {
552 .ndo_init = xfrmi_dev_init,
553 .ndo_uninit = xfrmi_dev_uninit,
554 .ndo_start_xmit = xfrmi_xmit,
555 .ndo_get_stats64 = dev_get_tstats64,
556 .ndo_get_iflink = xfrmi_get_iflink,
557 };
558
559 static void xfrmi_dev_setup(struct net_device *dev)
560 {
561 dev->netdev_ops = &xfrmi_netdev_ops;
562 dev->header_ops = &ip_tunnel_header_ops;
563 dev->type = ARPHRD_NONE;
564 dev->mtu = ETH_DATA_LEN;
565 dev->min_mtu = ETH_MIN_MTU;
566 dev->max_mtu = IP_MAX_MTU;
567 dev->flags = IFF_NOARP;
568 dev->needs_free_netdev = true;
569 dev->priv_destructor = xfrmi_dev_free;
570 netif_keep_dst(dev);
571
572 eth_broadcast_addr(dev->broadcast);
573 }
574
575 #define XFRMI_FEATURES (NETIF_F_SG | \
576 NETIF_F_FRAGLIST | \
577 NETIF_F_GSO_SOFTWARE | \
578 NETIF_F_HW_CSUM)
579
580 static int xfrmi_dev_init(struct net_device *dev)
581 {
582 struct xfrm_if *xi = netdev_priv(dev);
583 struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
584 int err;
585
586 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
587 if (!dev->tstats)
588 return -ENOMEM;
589
590 err = gro_cells_init(&xi->gro_cells, dev);
591 if (err) {
592 free_percpu(dev->tstats);
593 return err;
594 }
595
596 dev->features |= NETIF_F_LLTX;
597 dev->features |= XFRMI_FEATURES;
598 dev->hw_features |= XFRMI_FEATURES;
599
600 if (phydev) {
601 dev->needed_headroom = phydev->needed_headroom;
602 dev->needed_tailroom = phydev->needed_tailroom;
603
604 if (is_zero_ether_addr(dev->dev_addr))
605 eth_hw_addr_inherit(dev, phydev);
606 if (is_zero_ether_addr(dev->broadcast))
607 memcpy(dev->broadcast, phydev->broadcast,
608 dev->addr_len);
609 } else {
610 eth_hw_addr_random(dev);
611 eth_broadcast_addr(dev->broadcast);
612 }
613
614 return 0;
615 }
616
617 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
618 struct netlink_ext_ack *extack)
619 {
620 return 0;
621 }
622
623 static void xfrmi_netlink_parms(struct nlattr *data[],
624 struct xfrm_if_parms *parms)
625 {
626 memset(parms, 0, sizeof(*parms));
627
628 if (!data)
629 return;
630
631 if (data[IFLA_XFRM_LINK])
632 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
633
634 if (data[IFLA_XFRM_IF_ID])
635 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
636 }
637
638 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
639 struct nlattr *tb[], struct nlattr *data[],
640 struct netlink_ext_ack *extack)
641 {
642 struct net *net = dev_net(dev);
643 struct xfrm_if_parms p = {};
644 struct xfrm_if *xi;
645 int err;
646
647 xfrmi_netlink_parms(data, &p);
648 if (!p.if_id) {
649 NL_SET_ERR_MSG(extack, "if_id must be non zero");
650 return -EINVAL;
651 }
652
653 xi = xfrmi_locate(net, &p);
654 if (xi)
655 return -EEXIST;
656
657 xi = netdev_priv(dev);
658 xi->p = p;
659 xi->net = net;
660 xi->dev = dev;
661
662 err = xfrmi_create(dev);
663 return err;
664 }
665
666 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
667 {
668 unregister_netdevice_queue(dev, head);
669 }
670
671 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
672 struct nlattr *data[],
673 struct netlink_ext_ack *extack)
674 {
675 struct xfrm_if *xi = netdev_priv(dev);
676 struct net *net = xi->net;
677 struct xfrm_if_parms p = {};
678
679 xfrmi_netlink_parms(data, &p);
680 if (!p.if_id) {
681 NL_SET_ERR_MSG(extack, "if_id must be non zero");
682 return -EINVAL;
683 }
684
685 xi = xfrmi_locate(net, &p);
686 if (!xi) {
687 xi = netdev_priv(dev);
688 } else {
689 if (xi->dev != dev)
690 return -EEXIST;
691 }
692
693 return xfrmi_update(xi, &p);
694 }
695
696 static size_t xfrmi_get_size(const struct net_device *dev)
697 {
698 return
699 /* IFLA_XFRM_LINK */
700 nla_total_size(4) +
701 /* IFLA_XFRM_IF_ID */
702 nla_total_size(4) +
703 0;
704 }
705
706 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
707 {
708 struct xfrm_if *xi = netdev_priv(dev);
709 struct xfrm_if_parms *parm = &xi->p;
710
711 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
712 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
713 goto nla_put_failure;
714 return 0;
715
716 nla_put_failure:
717 return -EMSGSIZE;
718 }
719
720 static struct net *xfrmi_get_link_net(const struct net_device *dev)
721 {
722 struct xfrm_if *xi = netdev_priv(dev);
723
724 return xi->net;
725 }
726
727 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
728 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
729 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
730 };
731
732 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
733 .kind = "xfrm",
734 .maxtype = IFLA_XFRM_MAX,
735 .policy = xfrmi_policy,
736 .priv_size = sizeof(struct xfrm_if),
737 .setup = xfrmi_dev_setup,
738 .validate = xfrmi_validate,
739 .newlink = xfrmi_newlink,
740 .dellink = xfrmi_dellink,
741 .changelink = xfrmi_changelink,
742 .get_size = xfrmi_get_size,
743 .fill_info = xfrmi_fill_info,
744 .get_link_net = xfrmi_get_link_net,
745 };
746
747 static void __net_exit xfrmi_exit_batch_net(struct list_head *net_exit_list)
748 {
749 struct net *net;
750 LIST_HEAD(list);
751
752 rtnl_lock();
753 list_for_each_entry(net, net_exit_list, exit_list) {
754 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
755 struct xfrm_if __rcu **xip;
756 struct xfrm_if *xi;
757 int i;
758
759 for (i = 0; i < XFRMI_HASH_SIZE; i++) {
760 for (xip = &xfrmn->xfrmi[i];
761 (xi = rtnl_dereference(*xip)) != NULL;
762 xip = &xi->next)
763 unregister_netdevice_queue(xi->dev, &list);
764 }
765 }
766 unregister_netdevice_many(&list);
767 rtnl_unlock();
768 }
769
770 static struct pernet_operations xfrmi_net_ops = {
771 .exit_batch = xfrmi_exit_batch_net,
772 .id = &xfrmi_net_id,
773 .size = sizeof(struct xfrmi_net),
774 };
775
776 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
777 .handler = xfrm6_rcv,
778 .input_handler = xfrm_input,
779 .cb_handler = xfrmi_rcv_cb,
780 .err_handler = xfrmi6_err,
781 .priority = 10,
782 };
783
784 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
785 .handler = xfrm6_rcv,
786 .input_handler = xfrm_input,
787 .cb_handler = xfrmi_rcv_cb,
788 .err_handler = xfrmi6_err,
789 .priority = 10,
790 };
791
792 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
793 .handler = xfrm6_rcv,
794 .input_handler = xfrm_input,
795 .cb_handler = xfrmi_rcv_cb,
796 .err_handler = xfrmi6_err,
797 .priority = 10,
798 };
799
800 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
801 static int xfrmi6_rcv_tunnel(struct sk_buff *skb)
802 {
803 const xfrm_address_t *saddr;
804 __be32 spi;
805
806 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
807 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
808
809 return xfrm6_rcv_spi(skb, IPPROTO_IPV6, spi, NULL);
810 }
811
812 static struct xfrm6_tunnel xfrmi_ipv6_handler __read_mostly = {
813 .handler = xfrmi6_rcv_tunnel,
814 .cb_handler = xfrmi_rcv_cb,
815 .err_handler = xfrmi6_err,
816 .priority = 2,
817 };
818
819 static struct xfrm6_tunnel xfrmi_ip6ip_handler __read_mostly = {
820 .handler = xfrmi6_rcv_tunnel,
821 .cb_handler = xfrmi_rcv_cb,
822 .err_handler = xfrmi6_err,
823 .priority = 2,
824 };
825 #endif
826
827 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
828 .handler = xfrm4_rcv,
829 .input_handler = xfrm_input,
830 .cb_handler = xfrmi_rcv_cb,
831 .err_handler = xfrmi4_err,
832 .priority = 10,
833 };
834
835 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
836 .handler = xfrm4_rcv,
837 .input_handler = xfrm_input,
838 .cb_handler = xfrmi_rcv_cb,
839 .err_handler = xfrmi4_err,
840 .priority = 10,
841 };
842
843 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
844 .handler = xfrm4_rcv,
845 .input_handler = xfrm_input,
846 .cb_handler = xfrmi_rcv_cb,
847 .err_handler = xfrmi4_err,
848 .priority = 10,
849 };
850
851 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
852 static int xfrmi4_rcv_tunnel(struct sk_buff *skb)
853 {
854 return xfrm4_rcv_spi(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr);
855 }
856
857 static struct xfrm_tunnel xfrmi_ipip_handler __read_mostly = {
858 .handler = xfrmi4_rcv_tunnel,
859 .cb_handler = xfrmi_rcv_cb,
860 .err_handler = xfrmi4_err,
861 .priority = 3,
862 };
863
864 static struct xfrm_tunnel xfrmi_ipip6_handler __read_mostly = {
865 .handler = xfrmi4_rcv_tunnel,
866 .cb_handler = xfrmi_rcv_cb,
867 .err_handler = xfrmi4_err,
868 .priority = 2,
869 };
870 #endif
871
872 static int __init xfrmi4_init(void)
873 {
874 int err;
875
876 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
877 if (err < 0)
878 goto xfrm_proto_esp_failed;
879 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
880 if (err < 0)
881 goto xfrm_proto_ah_failed;
882 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
883 if (err < 0)
884 goto xfrm_proto_comp_failed;
885 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
886 err = xfrm4_tunnel_register(&xfrmi_ipip_handler, AF_INET);
887 if (err < 0)
888 goto xfrm_tunnel_ipip_failed;
889 err = xfrm4_tunnel_register(&xfrmi_ipip6_handler, AF_INET6);
890 if (err < 0)
891 goto xfrm_tunnel_ipip6_failed;
892 #endif
893
894 return 0;
895
896 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
897 xfrm_tunnel_ipip6_failed:
898 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
899 xfrm_tunnel_ipip_failed:
900 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
901 #endif
902 xfrm_proto_comp_failed:
903 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
904 xfrm_proto_ah_failed:
905 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
906 xfrm_proto_esp_failed:
907 return err;
908 }
909
910 static void xfrmi4_fini(void)
911 {
912 #if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
913 xfrm4_tunnel_deregister(&xfrmi_ipip6_handler, AF_INET6);
914 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
915 #endif
916 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
917 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
918 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
919 }
920
921 static int __init xfrmi6_init(void)
922 {
923 int err;
924
925 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
926 if (err < 0)
927 goto xfrm_proto_esp_failed;
928 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
929 if (err < 0)
930 goto xfrm_proto_ah_failed;
931 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
932 if (err < 0)
933 goto xfrm_proto_comp_failed;
934 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
935 err = xfrm6_tunnel_register(&xfrmi_ipv6_handler, AF_INET6);
936 if (err < 0)
937 goto xfrm_tunnel_ipv6_failed;
938 err = xfrm6_tunnel_register(&xfrmi_ip6ip_handler, AF_INET);
939 if (err < 0)
940 goto xfrm_tunnel_ip6ip_failed;
941 #endif
942
943 return 0;
944
945 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
946 xfrm_tunnel_ip6ip_failed:
947 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
948 xfrm_tunnel_ipv6_failed:
949 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
950 #endif
951 xfrm_proto_comp_failed:
952 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
953 xfrm_proto_ah_failed:
954 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
955 xfrm_proto_esp_failed:
956 return err;
957 }
958
959 static void xfrmi6_fini(void)
960 {
961 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
962 xfrm6_tunnel_deregister(&xfrmi_ip6ip_handler, AF_INET);
963 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
964 #endif
965 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
966 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
967 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
968 }
969
970 static const struct xfrm_if_cb xfrm_if_cb = {
971 .decode_session = xfrmi_decode_session,
972 };
973
974 static int __init xfrmi_init(void)
975 {
976 const char *msg;
977 int err;
978
979 pr_info("IPsec XFRM device driver\n");
980
981 msg = "tunnel device";
982 err = register_pernet_device(&xfrmi_net_ops);
983 if (err < 0)
984 goto pernet_dev_failed;
985
986 msg = "xfrm4 protocols";
987 err = xfrmi4_init();
988 if (err < 0)
989 goto xfrmi4_failed;
990
991 msg = "xfrm6 protocols";
992 err = xfrmi6_init();
993 if (err < 0)
994 goto xfrmi6_failed;
995
996
997 msg = "netlink interface";
998 err = rtnl_link_register(&xfrmi_link_ops);
999 if (err < 0)
1000 goto rtnl_link_failed;
1001
1002 xfrm_if_register_cb(&xfrm_if_cb);
1003
1004 return err;
1005
1006 rtnl_link_failed:
1007 xfrmi6_fini();
1008 xfrmi6_failed:
1009 xfrmi4_fini();
1010 xfrmi4_failed:
1011 unregister_pernet_device(&xfrmi_net_ops);
1012 pernet_dev_failed:
1013 pr_err("xfrmi init: failed to register %s\n", msg);
1014 return err;
1015 }
1016
1017 static void __exit xfrmi_fini(void)
1018 {
1019 xfrm_if_unregister_cb();
1020 rtnl_link_unregister(&xfrmi_link_ops);
1021 xfrmi4_fini();
1022 xfrmi6_fini();
1023 unregister_pernet_device(&xfrmi_net_ops);
1024 }
1025
1026 module_init(xfrmi_init);
1027 module_exit(xfrmi_fini);
1028 MODULE_LICENSE("GPL");
1029 MODULE_ALIAS_RTNL_LINK("xfrm");
1030 MODULE_ALIAS_NETDEV("xfrm0");
1031 MODULE_AUTHOR("Steffen Klassert");
1032 MODULE_DESCRIPTION("XFRM virtual interface");
1033
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.