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/addrconf.h>
41 #include <net/xfrm.h>
42 #include <net/net_namespace.h>
43 #include <net/netns/generic.h>
44 #include <linux/etherdevice.h>
45
46 static int xfrmi_dev_init(struct net_device *dev);
47 static void xfrmi_dev_setup(struct net_device *dev);
48 static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
49 static unsigned int xfrmi_net_id __read_mostly;
50
51 struct xfrmi_net {
52 /* lists for storing interfaces in use */
53 struct xfrm_if __rcu *xfrmi[1];
54 };
55
56 #define for_each_xfrmi_rcu(start, xi) \
57 for (xi = rcu_dereference(start); xi; xi = rcu_dereference(xi->next))
58
xfrmi_lookup(struct net * net,struct xfrm_state * x)59 static struct xfrm_if *xfrmi_lookup(struct net *net, struct xfrm_state *x)
60 {
61 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
62 struct xfrm_if *xi;
63
64 for_each_xfrmi_rcu(xfrmn->xfrmi[0], xi) {
65 if (x->if_id == xi->p.if_id &&
66 (xi->dev->flags & IFF_UP))
67 return xi;
68 }
69
70 return NULL;
71 }
72
xfrmi_decode_session(struct sk_buff * skb,unsigned short family)73 static struct xfrm_if *xfrmi_decode_session(struct sk_buff *skb,
74 unsigned short family)
75 {
76 struct xfrmi_net *xfrmn;
77 struct xfrm_if *xi;
78 int ifindex = 0;
79
80 if (!secpath_exists(skb) || !skb->dev)
81 return NULL;
82
83 switch (family) {
84 case AF_INET6:
85 ifindex = inet6_sdif(skb);
86 break;
87 case AF_INET:
88 ifindex = inet_sdif(skb);
89 break;
90 }
91 if (!ifindex)
92 ifindex = skb->dev->ifindex;
93
94 xfrmn = net_generic(xs_net(xfrm_input_state(skb)), xfrmi_net_id);
95
96 for_each_xfrmi_rcu(xfrmn->xfrmi[0], xi) {
97 if (ifindex == xi->dev->ifindex &&
98 (xi->dev->flags & IFF_UP))
99 return xi;
100 }
101
102 return NULL;
103 }
104
xfrmi_link(struct xfrmi_net * xfrmn,struct xfrm_if * xi)105 static void xfrmi_link(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
106 {
107 struct xfrm_if __rcu **xip = &xfrmn->xfrmi[0];
108
109 rcu_assign_pointer(xi->next , rtnl_dereference(*xip));
110 rcu_assign_pointer(*xip, xi);
111 }
112
xfrmi_unlink(struct xfrmi_net * xfrmn,struct xfrm_if * xi)113 static void xfrmi_unlink(struct xfrmi_net *xfrmn, struct xfrm_if *xi)
114 {
115 struct xfrm_if __rcu **xip;
116 struct xfrm_if *iter;
117
118 for (xip = &xfrmn->xfrmi[0];
119 (iter = rtnl_dereference(*xip)) != NULL;
120 xip = &iter->next) {
121 if (xi == iter) {
122 rcu_assign_pointer(*xip, xi->next);
123 break;
124 }
125 }
126 }
127
xfrmi_dev_free(struct net_device * dev)128 static void xfrmi_dev_free(struct net_device *dev)
129 {
130 struct xfrm_if *xi = netdev_priv(dev);
131
132 gro_cells_destroy(&xi->gro_cells);
133 free_percpu(dev->tstats);
134 }
135
xfrmi_create(struct net_device * dev)136 static int xfrmi_create(struct net_device *dev)
137 {
138 struct xfrm_if *xi = netdev_priv(dev);
139 struct net *net = dev_net(dev);
140 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
141 int err;
142
143 dev->rtnl_link_ops = &xfrmi_link_ops;
144 err = register_netdevice(dev);
145 if (err < 0)
146 goto out;
147
148 dev_hold(dev);
149 xfrmi_link(xfrmn, xi);
150
151 return 0;
152
153 out:
154 return err;
155 }
156
xfrmi_locate(struct net * net,struct xfrm_if_parms * p)157 static struct xfrm_if *xfrmi_locate(struct net *net, struct xfrm_if_parms *p)
158 {
159 struct xfrm_if __rcu **xip;
160 struct xfrm_if *xi;
161 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
162
163 for (xip = &xfrmn->xfrmi[0];
164 (xi = rtnl_dereference(*xip)) != NULL;
165 xip = &xi->next)
166 if (xi->p.if_id == p->if_id)
167 return xi;
168
169 return NULL;
170 }
171
xfrmi_dev_uninit(struct net_device * dev)172 static void xfrmi_dev_uninit(struct net_device *dev)
173 {
174 struct xfrm_if *xi = netdev_priv(dev);
175 struct xfrmi_net *xfrmn = net_generic(xi->net, xfrmi_net_id);
176
177 xfrmi_unlink(xfrmn, xi);
178 dev_put(dev);
179 }
180
xfrmi_scrub_packet(struct sk_buff * skb,bool xnet)181 static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
182 {
183 skb->tstamp = 0;
184 skb->pkt_type = PACKET_HOST;
185 skb->skb_iif = 0;
186 skb->ignore_df = 0;
187 skb_dst_drop(skb);
188 nf_reset_ct(skb);
189 nf_reset_trace(skb);
190
191 if (!xnet)
192 return;
193
194 ipvs_reset(skb);
195 secpath_reset(skb);
196 skb_orphan(skb);
197 skb->mark = 0;
198 }
199
xfrmi_rcv_cb(struct sk_buff * skb,int err)200 static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
201 {
202 const struct xfrm_mode *inner_mode;
203 struct pcpu_sw_netstats *tstats;
204 struct net_device *dev;
205 struct xfrm_state *x;
206 struct xfrm_if *xi;
207 bool xnet;
208
209 if (err && !secpath_exists(skb))
210 return 0;
211
212 x = xfrm_input_state(skb);
213
214 xi = xfrmi_lookup(xs_net(x), x);
215 if (!xi)
216 return 1;
217
218 dev = xi->dev;
219 skb->dev = dev;
220
221 if (err) {
222 dev->stats.rx_errors++;
223 dev->stats.rx_dropped++;
224
225 return 0;
226 }
227
228 xnet = !net_eq(xi->net, dev_net(skb->dev));
229
230 if (xnet) {
231 inner_mode = &x->inner_mode;
232
233 if (x->sel.family == AF_UNSPEC) {
234 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
235 if (inner_mode == NULL) {
236 XFRM_INC_STATS(dev_net(skb->dev),
237 LINUX_MIB_XFRMINSTATEMODEERROR);
238 return -EINVAL;
239 }
240 }
241
242 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
243 inner_mode->family))
244 return -EPERM;
245 }
246
247 xfrmi_scrub_packet(skb, xnet);
248
249 tstats = this_cpu_ptr(dev->tstats);
250
251 u64_stats_update_begin(&tstats->syncp);
252 tstats->rx_packets++;
253 tstats->rx_bytes += skb->len;
254 u64_stats_update_end(&tstats->syncp);
255
256 return 0;
257 }
258
259 static int
xfrmi_xmit2(struct sk_buff * skb,struct net_device * dev,struct flowi * fl)260 xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
261 {
262 struct xfrm_if *xi = netdev_priv(dev);
263 struct net_device_stats *stats = &xi->dev->stats;
264 struct dst_entry *dst = skb_dst(skb);
265 unsigned int length = skb->len;
266 struct net_device *tdev;
267 struct xfrm_state *x;
268 int err = -1;
269 int mtu;
270
271 dst_hold(dst);
272 dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, xi->p.if_id);
273 if (IS_ERR(dst)) {
274 err = PTR_ERR(dst);
275 dst = NULL;
276 goto tx_err_link_failure;
277 }
278
279 x = dst->xfrm;
280 if (!x)
281 goto tx_err_link_failure;
282
283 if (x->if_id != xi->p.if_id)
284 goto tx_err_link_failure;
285
286 tdev = dst->dev;
287
288 if (tdev == dev) {
289 stats->collisions++;
290 net_warn_ratelimited("%s: Local routing loop detected!\n",
291 dev->name);
292 goto tx_err_dst_release;
293 }
294
295 mtu = dst_mtu(dst);
296 if (skb->len > mtu) {
297 skb_dst_update_pmtu_no_confirm(skb, mtu);
298
299 if (skb->protocol == htons(ETH_P_IPV6)) {
300 if (mtu < IPV6_MIN_MTU)
301 mtu = IPV6_MIN_MTU;
302
303 if (skb->len > 1280)
304 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
305 else
306 goto xmit;
307 } else {
308 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
309 goto xmit;
310 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
311 htonl(mtu));
312 }
313
314 dst_release(dst);
315 return -EMSGSIZE;
316 }
317
318 xmit:
319 xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
320 skb_dst_set(skb, dst);
321 skb->dev = tdev;
322
323 err = dst_output(xi->net, skb->sk, skb);
324 if (net_xmit_eval(err) == 0) {
325 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
326
327 u64_stats_update_begin(&tstats->syncp);
328 tstats->tx_bytes += length;
329 tstats->tx_packets++;
330 u64_stats_update_end(&tstats->syncp);
331 } else {
332 stats->tx_errors++;
333 stats->tx_aborted_errors++;
334 }
335
336 return 0;
337 tx_err_link_failure:
338 stats->tx_carrier_errors++;
339 dst_link_failure(skb);
340 tx_err_dst_release:
341 dst_release(dst);
342 return err;
343 }
344
xfrmi_xmit(struct sk_buff * skb,struct net_device * dev)345 static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
346 {
347 struct xfrm_if *xi = netdev_priv(dev);
348 struct net_device_stats *stats = &xi->dev->stats;
349 struct dst_entry *dst = skb_dst(skb);
350 struct flowi fl;
351 int ret;
352
353 memset(&fl, 0, sizeof(fl));
354
355 switch (skb->protocol) {
356 case htons(ETH_P_IPV6):
357 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
358 xfrm_decode_session(skb, &fl, AF_INET6);
359 if (!dst) {
360 fl.u.ip6.flowi6_oif = dev->ifindex;
361 fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
362 dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
363 if (dst->error) {
364 dst_release(dst);
365 stats->tx_carrier_errors++;
366 goto tx_err;
367 }
368 skb_dst_set(skb, dst);
369 }
370 break;
371 case htons(ETH_P_IP):
372 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
373 xfrm_decode_session(skb, &fl, AF_INET);
374 if (!dst) {
375 struct rtable *rt;
376
377 fl.u.ip4.flowi4_oif = dev->ifindex;
378 fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
379 rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
380 if (IS_ERR(rt)) {
381 stats->tx_carrier_errors++;
382 goto tx_err;
383 }
384 skb_dst_set(skb, &rt->dst);
385 }
386 break;
387 default:
388 goto tx_err;
389 }
390
391 fl.flowi_oif = xi->p.link;
392
393 ret = xfrmi_xmit2(skb, dev, &fl);
394 if (ret < 0)
395 goto tx_err;
396
397 return NETDEV_TX_OK;
398
399 tx_err:
400 stats->tx_errors++;
401 stats->tx_dropped++;
402 kfree_skb(skb);
403 return NETDEV_TX_OK;
404 }
405
xfrmi4_err(struct sk_buff * skb,u32 info)406 static int xfrmi4_err(struct sk_buff *skb, u32 info)
407 {
408 const struct iphdr *iph = (const struct iphdr *)skb->data;
409 struct net *net = dev_net(skb->dev);
410 int protocol = iph->protocol;
411 struct ip_comp_hdr *ipch;
412 struct ip_esp_hdr *esph;
413 struct ip_auth_hdr *ah ;
414 struct xfrm_state *x;
415 struct xfrm_if *xi;
416 __be32 spi;
417
418 switch (protocol) {
419 case IPPROTO_ESP:
420 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
421 spi = esph->spi;
422 break;
423 case IPPROTO_AH:
424 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
425 spi = ah->spi;
426 break;
427 case IPPROTO_COMP:
428 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
429 spi = htonl(ntohs(ipch->cpi));
430 break;
431 default:
432 return 0;
433 }
434
435 switch (icmp_hdr(skb)->type) {
436 case ICMP_DEST_UNREACH:
437 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
438 return 0;
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
xfrmi6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)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
xfrmi_change(struct xfrm_if * xi,const struct xfrm_if_parms * p)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
xfrmi_update(struct xfrm_if * xi,struct xfrm_if_parms * p)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
xfrmi_get_stats64(struct net_device * dev,struct rtnl_link_stats64 * s)544 static void xfrmi_get_stats64(struct net_device *dev,
545 struct rtnl_link_stats64 *s)
546 {
547 int cpu;
548
549 for_each_possible_cpu(cpu) {
550 struct pcpu_sw_netstats *stats;
551 struct pcpu_sw_netstats tmp;
552 int start;
553
554 stats = per_cpu_ptr(dev->tstats, cpu);
555 do {
556 start = u64_stats_fetch_begin_irq(&stats->syncp);
557 tmp.rx_packets = stats->rx_packets;
558 tmp.rx_bytes = stats->rx_bytes;
559 tmp.tx_packets = stats->tx_packets;
560 tmp.tx_bytes = stats->tx_bytes;
561 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
562
563 s->rx_packets += tmp.rx_packets;
564 s->rx_bytes += tmp.rx_bytes;
565 s->tx_packets += tmp.tx_packets;
566 s->tx_bytes += tmp.tx_bytes;
567 }
568
569 s->rx_dropped = dev->stats.rx_dropped;
570 s->tx_dropped = dev->stats.tx_dropped;
571 }
572
xfrmi_get_iflink(const struct net_device * dev)573 static int xfrmi_get_iflink(const struct net_device *dev)
574 {
575 struct xfrm_if *xi = netdev_priv(dev);
576
577 return xi->p.link;
578 }
579
580
581 static const struct net_device_ops xfrmi_netdev_ops = {
582 .ndo_init = xfrmi_dev_init,
583 .ndo_uninit = xfrmi_dev_uninit,
584 .ndo_start_xmit = xfrmi_xmit,
585 .ndo_get_stats64 = xfrmi_get_stats64,
586 .ndo_get_iflink = xfrmi_get_iflink,
587 };
588
xfrmi_dev_setup(struct net_device * dev)589 static void xfrmi_dev_setup(struct net_device *dev)
590 {
591 dev->netdev_ops = &xfrmi_netdev_ops;
592 dev->type = ARPHRD_NONE;
593 dev->mtu = ETH_DATA_LEN;
594 dev->min_mtu = ETH_MIN_MTU;
595 dev->max_mtu = IP_MAX_MTU;
596 dev->flags = IFF_NOARP;
597 dev->needs_free_netdev = true;
598 dev->priv_destructor = xfrmi_dev_free;
599 netif_keep_dst(dev);
600
601 eth_broadcast_addr(dev->broadcast);
602 }
603
xfrmi_dev_init(struct net_device * dev)604 static int xfrmi_dev_init(struct net_device *dev)
605 {
606 struct xfrm_if *xi = netdev_priv(dev);
607 struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
608 int err;
609
610 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
611 if (!dev->tstats)
612 return -ENOMEM;
613
614 err = gro_cells_init(&xi->gro_cells, dev);
615 if (err) {
616 free_percpu(dev->tstats);
617 return err;
618 }
619
620 dev->features |= NETIF_F_LLTX;
621
622 if (phydev) {
623 dev->needed_headroom = phydev->needed_headroom;
624 dev->needed_tailroom = phydev->needed_tailroom;
625
626 if (is_zero_ether_addr(dev->dev_addr))
627 eth_hw_addr_inherit(dev, phydev);
628 if (is_zero_ether_addr(dev->broadcast))
629 memcpy(dev->broadcast, phydev->broadcast,
630 dev->addr_len);
631 } else {
632 eth_hw_addr_random(dev);
633 eth_broadcast_addr(dev->broadcast);
634 }
635
636 return 0;
637 }
638
xfrmi_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)639 static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
640 struct netlink_ext_ack *extack)
641 {
642 return 0;
643 }
644
xfrmi_netlink_parms(struct nlattr * data[],struct xfrm_if_parms * parms)645 static void xfrmi_netlink_parms(struct nlattr *data[],
646 struct xfrm_if_parms *parms)
647 {
648 memset(parms, 0, sizeof(*parms));
649
650 if (!data)
651 return;
652
653 if (data[IFLA_XFRM_LINK])
654 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
655
656 if (data[IFLA_XFRM_IF_ID])
657 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
658 }
659
xfrmi_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)660 static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
661 struct nlattr *tb[], struct nlattr *data[],
662 struct netlink_ext_ack *extack)
663 {
664 struct net *net = dev_net(dev);
665 struct xfrm_if_parms p = {};
666 struct xfrm_if *xi;
667 int err;
668
669 xfrmi_netlink_parms(data, &p);
670 if (!p.if_id) {
671 NL_SET_ERR_MSG(extack, "if_id must be non zero");
672 return -EINVAL;
673 }
674
675 xi = xfrmi_locate(net, &p);
676 if (xi)
677 return -EEXIST;
678
679 xi = netdev_priv(dev);
680 xi->p = p;
681 xi->net = net;
682 xi->dev = dev;
683
684 err = xfrmi_create(dev);
685 return err;
686 }
687
xfrmi_dellink(struct net_device * dev,struct list_head * head)688 static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
689 {
690 unregister_netdevice_queue(dev, head);
691 }
692
xfrmi_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)693 static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
694 struct nlattr *data[],
695 struct netlink_ext_ack *extack)
696 {
697 struct xfrm_if *xi = netdev_priv(dev);
698 struct net *net = xi->net;
699 struct xfrm_if_parms p = {};
700
701 xfrmi_netlink_parms(data, &p);
702 if (!p.if_id) {
703 NL_SET_ERR_MSG(extack, "if_id must be non zero");
704 return -EINVAL;
705 }
706
707 xi = xfrmi_locate(net, &p);
708 if (!xi) {
709 xi = netdev_priv(dev);
710 } else {
711 if (xi->dev != dev)
712 return -EEXIST;
713 }
714
715 return xfrmi_update(xi, &p);
716 }
717
xfrmi_get_size(const struct net_device * dev)718 static size_t xfrmi_get_size(const struct net_device *dev)
719 {
720 return
721 /* IFLA_XFRM_LINK */
722 nla_total_size(4) +
723 /* IFLA_XFRM_IF_ID */
724 nla_total_size(4) +
725 0;
726 }
727
xfrmi_fill_info(struct sk_buff * skb,const struct net_device * dev)728 static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
729 {
730 struct xfrm_if *xi = netdev_priv(dev);
731 struct xfrm_if_parms *parm = &xi->p;
732
733 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
734 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
735 goto nla_put_failure;
736 return 0;
737
738 nla_put_failure:
739 return -EMSGSIZE;
740 }
741
xfrmi_get_link_net(const struct net_device * dev)742 static struct net *xfrmi_get_link_net(const struct net_device *dev)
743 {
744 struct xfrm_if *xi = netdev_priv(dev);
745
746 return xi->net;
747 }
748
749 static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
750 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
751 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
752 };
753
754 static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
755 .kind = "xfrm",
756 .maxtype = IFLA_XFRM_MAX,
757 .policy = xfrmi_policy,
758 .priv_size = sizeof(struct xfrm_if),
759 .setup = xfrmi_dev_setup,
760 .validate = xfrmi_validate,
761 .newlink = xfrmi_newlink,
762 .dellink = xfrmi_dellink,
763 .changelink = xfrmi_changelink,
764 .get_size = xfrmi_get_size,
765 .fill_info = xfrmi_fill_info,
766 .get_link_net = xfrmi_get_link_net,
767 };
768
xfrmi_destroy_interfaces(struct xfrmi_net * xfrmn)769 static void __net_exit xfrmi_destroy_interfaces(struct xfrmi_net *xfrmn)
770 {
771 struct xfrm_if *xi;
772 LIST_HEAD(list);
773
774 xi = rtnl_dereference(xfrmn->xfrmi[0]);
775 if (!xi)
776 return;
777
778 unregister_netdevice_queue(xi->dev, &list);
779 unregister_netdevice_many(&list);
780 }
781
xfrmi_exit_net(struct net * net)782 static void __net_exit xfrmi_exit_net(struct net *net)
783 {
784 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
785
786 rtnl_lock();
787 xfrmi_destroy_interfaces(xfrmn);
788 rtnl_unlock();
789 }
790
xfrmi_exit_batch_net(struct list_head * net_exit_list)791 static void __net_exit xfrmi_exit_batch_net(struct list_head *net_exit_list)
792 {
793 struct net *net;
794 LIST_HEAD(list);
795
796 rtnl_lock();
797 list_for_each_entry(net, net_exit_list, exit_list) {
798 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
799 struct xfrm_if __rcu **xip;
800 struct xfrm_if *xi;
801
802 for (xip = &xfrmn->xfrmi[0];
803 (xi = rtnl_dereference(*xip)) != NULL;
804 xip = &xi->next)
805 unregister_netdevice_queue(xi->dev, &list);
806 }
807 unregister_netdevice_many(&list);
808 rtnl_unlock();
809 }
810
811 static struct pernet_operations xfrmi_net_ops = {
812 .exit_batch = xfrmi_exit_batch_net,
813 .exit = xfrmi_exit_net,
814 .id = &xfrmi_net_id,
815 .size = sizeof(struct xfrmi_net),
816 };
817
818 static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
819 .handler = xfrm6_rcv,
820 .cb_handler = xfrmi_rcv_cb,
821 .err_handler = xfrmi6_err,
822 .priority = 10,
823 };
824
825 static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
826 .handler = xfrm6_rcv,
827 .cb_handler = xfrmi_rcv_cb,
828 .err_handler = xfrmi6_err,
829 .priority = 10,
830 };
831
832 static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
833 .handler = xfrm6_rcv,
834 .cb_handler = xfrmi_rcv_cb,
835 .err_handler = xfrmi6_err,
836 .priority = 10,
837 };
838
839 static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
840 .handler = xfrm4_rcv,
841 .input_handler = xfrm_input,
842 .cb_handler = xfrmi_rcv_cb,
843 .err_handler = xfrmi4_err,
844 .priority = 10,
845 };
846
847 static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
848 .handler = xfrm4_rcv,
849 .input_handler = xfrm_input,
850 .cb_handler = xfrmi_rcv_cb,
851 .err_handler = xfrmi4_err,
852 .priority = 10,
853 };
854
855 static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
856 .handler = xfrm4_rcv,
857 .input_handler = xfrm_input,
858 .cb_handler = xfrmi_rcv_cb,
859 .err_handler = xfrmi4_err,
860 .priority = 10,
861 };
862
xfrmi4_init(void)863 static int __init xfrmi4_init(void)
864 {
865 int err;
866
867 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
868 if (err < 0)
869 goto xfrm_proto_esp_failed;
870 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
871 if (err < 0)
872 goto xfrm_proto_ah_failed;
873 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
874 if (err < 0)
875 goto xfrm_proto_comp_failed;
876
877 return 0;
878
879 xfrm_proto_comp_failed:
880 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
881 xfrm_proto_ah_failed:
882 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
883 xfrm_proto_esp_failed:
884 return err;
885 }
886
xfrmi4_fini(void)887 static void xfrmi4_fini(void)
888 {
889 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
890 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
891 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
892 }
893
xfrmi6_init(void)894 static int __init xfrmi6_init(void)
895 {
896 int err;
897
898 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
899 if (err < 0)
900 goto xfrm_proto_esp_failed;
901 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
902 if (err < 0)
903 goto xfrm_proto_ah_failed;
904 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
905 if (err < 0)
906 goto xfrm_proto_comp_failed;
907
908 return 0;
909
910 xfrm_proto_comp_failed:
911 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
912 xfrm_proto_ah_failed:
913 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
914 xfrm_proto_esp_failed:
915 return err;
916 }
917
xfrmi6_fini(void)918 static void xfrmi6_fini(void)
919 {
920 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
921 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
922 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
923 }
924
925 static const struct xfrm_if_cb xfrm_if_cb = {
926 .decode_session = xfrmi_decode_session,
927 };
928
xfrmi_init(void)929 static int __init xfrmi_init(void)
930 {
931 const char *msg;
932 int err;
933
934 pr_info("IPsec XFRM device driver\n");
935
936 msg = "tunnel device";
937 err = register_pernet_device(&xfrmi_net_ops);
938 if (err < 0)
939 goto pernet_dev_failed;
940
941 msg = "xfrm4 protocols";
942 err = xfrmi4_init();
943 if (err < 0)
944 goto xfrmi4_failed;
945
946 msg = "xfrm6 protocols";
947 err = xfrmi6_init();
948 if (err < 0)
949 goto xfrmi6_failed;
950
951
952 msg = "netlink interface";
953 err = rtnl_link_register(&xfrmi_link_ops);
954 if (err < 0)
955 goto rtnl_link_failed;
956
957 xfrm_if_register_cb(&xfrm_if_cb);
958
959 return err;
960
961 rtnl_link_failed:
962 xfrmi6_fini();
963 xfrmi6_failed:
964 xfrmi4_fini();
965 xfrmi4_failed:
966 unregister_pernet_device(&xfrmi_net_ops);
967 pernet_dev_failed:
968 pr_err("xfrmi init: failed to register %s\n", msg);
969 return err;
970 }
971
xfrmi_fini(void)972 static void __exit xfrmi_fini(void)
973 {
974 xfrm_if_unregister_cb();
975 rtnl_link_unregister(&xfrmi_link_ops);
976 xfrmi4_fini();
977 xfrmi6_fini();
978 unregister_pernet_device(&xfrmi_net_ops);
979 }
980
981 module_init(xfrmi_init);
982 module_exit(xfrmi_fini);
983 MODULE_LICENSE("GPL");
984 MODULE_ALIAS_RTNL_LINK("xfrm");
985 MODULE_ALIAS_NETDEV("xfrm0");
986 MODULE_AUTHOR("Steffen Klassert");
987 MODULE_DESCRIPTION("XFRM virtual interface");
988