Loading...
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
46static int xfrmi_dev_init(struct net_device *dev);
47static void xfrmi_dev_setup(struct net_device *dev);
48static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
49static unsigned int xfrmi_net_id __read_mostly;
50
51struct 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
59static 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
73static 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
105static 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
113static 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
128static 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
136static 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
153out:
154 return err;
155}
156
157static 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
172static 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
181static 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
200static 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
259static int
260xfrmi_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 if (!dst)
272 goto tx_err_link_failure;
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->ignore_df && skb->len > mtu) {
300 skb_dst_update_pmtu(skb, mtu);
301
302 if (skb->protocol == htons(ETH_P_IPV6)) {
303 if (mtu < IPV6_MIN_MTU)
304 mtu = IPV6_MIN_MTU;
305
306 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
307 } else {
308 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
309 htonl(mtu));
310 }
311
312 dst_release(dst);
313 return -EMSGSIZE;
314 }
315
316 xfrmi_scrub_packet(skb, !net_eq(xi->net, dev_net(dev)));
317 skb_dst_set(skb, dst);
318 skb->dev = tdev;
319
320 err = dst_output(xi->net, skb->sk, skb);
321 if (net_xmit_eval(err) == 0) {
322 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
323
324 u64_stats_update_begin(&tstats->syncp);
325 tstats->tx_bytes += length;
326 tstats->tx_packets++;
327 u64_stats_update_end(&tstats->syncp);
328 } else {
329 stats->tx_errors++;
330 stats->tx_aborted_errors++;
331 }
332
333 return 0;
334tx_err_link_failure:
335 stats->tx_carrier_errors++;
336 dst_link_failure(skb);
337tx_err_dst_release:
338 dst_release(dst);
339 return err;
340}
341
342static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
343{
344 struct xfrm_if *xi = netdev_priv(dev);
345 struct net_device_stats *stats = &xi->dev->stats;
346 struct flowi fl;
347 int ret;
348
349 memset(&fl, 0, sizeof(fl));
350
351 switch (skb->protocol) {
352 case htons(ETH_P_IPV6):
353 xfrm_decode_session(skb, &fl, AF_INET6);
354 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
355 break;
356 case htons(ETH_P_IP):
357 xfrm_decode_session(skb, &fl, AF_INET);
358 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
359 break;
360 default:
361 goto tx_err;
362 }
363
364 fl.flowi_oif = xi->p.link;
365
366 ret = xfrmi_xmit2(skb, dev, &fl);
367 if (ret < 0)
368 goto tx_err;
369
370 return NETDEV_TX_OK;
371
372tx_err:
373 stats->tx_errors++;
374 stats->tx_dropped++;
375 kfree_skb(skb);
376 return NETDEV_TX_OK;
377}
378
379static int xfrmi4_err(struct sk_buff *skb, u32 info)
380{
381 const struct iphdr *iph = (const struct iphdr *)skb->data;
382 struct net *net = dev_net(skb->dev);
383 int protocol = iph->protocol;
384 struct ip_comp_hdr *ipch;
385 struct ip_esp_hdr *esph;
386 struct ip_auth_hdr *ah ;
387 struct xfrm_state *x;
388 struct xfrm_if *xi;
389 __be32 spi;
390
391 switch (protocol) {
392 case IPPROTO_ESP:
393 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
394 spi = esph->spi;
395 break;
396 case IPPROTO_AH:
397 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
398 spi = ah->spi;
399 break;
400 case IPPROTO_COMP:
401 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
402 spi = htonl(ntohs(ipch->cpi));
403 break;
404 default:
405 return 0;
406 }
407
408 switch (icmp_hdr(skb)->type) {
409 case ICMP_DEST_UNREACH:
410 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
411 return 0;
412 case ICMP_REDIRECT:
413 break;
414 default:
415 return 0;
416 }
417
418 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
419 spi, protocol, AF_INET);
420 if (!x)
421 return 0;
422
423 xi = xfrmi_lookup(net, x);
424 if (!xi) {
425 xfrm_state_put(x);
426 return -1;
427 }
428
429 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
430 ipv4_update_pmtu(skb, net, info, 0, protocol);
431 else
432 ipv4_redirect(skb, net, 0, protocol);
433 xfrm_state_put(x);
434
435 return 0;
436}
437
438static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
439 u8 type, u8 code, int offset, __be32 info)
440{
441 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
442 struct net *net = dev_net(skb->dev);
443 int protocol = iph->nexthdr;
444 struct ip_comp_hdr *ipch;
445 struct ip_esp_hdr *esph;
446 struct ip_auth_hdr *ah;
447 struct xfrm_state *x;
448 struct xfrm_if *xi;
449 __be32 spi;
450
451 switch (protocol) {
452 case IPPROTO_ESP:
453 esph = (struct ip_esp_hdr *)(skb->data + offset);
454 spi = esph->spi;
455 break;
456 case IPPROTO_AH:
457 ah = (struct ip_auth_hdr *)(skb->data + offset);
458 spi = ah->spi;
459 break;
460 case IPPROTO_COMP:
461 ipch = (struct ip_comp_hdr *)(skb->data + offset);
462 spi = htonl(ntohs(ipch->cpi));
463 break;
464 default:
465 return 0;
466 }
467
468 if (type != ICMPV6_PKT_TOOBIG &&
469 type != NDISC_REDIRECT)
470 return 0;
471
472 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
473 spi, protocol, AF_INET6);
474 if (!x)
475 return 0;
476
477 xi = xfrmi_lookup(net, x);
478 if (!xi) {
479 xfrm_state_put(x);
480 return -1;
481 }
482
483 if (type == NDISC_REDIRECT)
484 ip6_redirect(skb, net, skb->dev->ifindex, 0,
485 sock_net_uid(net, NULL));
486 else
487 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
488 xfrm_state_put(x);
489
490 return 0;
491}
492
493static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
494{
495 if (xi->p.link != p->link)
496 return -EINVAL;
497
498 xi->p.if_id = p->if_id;
499
500 return 0;
501}
502
503static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
504{
505 struct net *net = xi->net;
506 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
507 int err;
508
509 xfrmi_unlink(xfrmn, xi);
510 synchronize_net();
511 err = xfrmi_change(xi, p);
512 xfrmi_link(xfrmn, xi);
513 netdev_state_change(xi->dev);
514 return err;
515}
516
517static void xfrmi_get_stats64(struct net_device *dev,
518 struct rtnl_link_stats64 *s)
519{
520 int cpu;
521
522 for_each_possible_cpu(cpu) {
523 struct pcpu_sw_netstats *stats;
524 struct pcpu_sw_netstats tmp;
525 int start;
526
527 stats = per_cpu_ptr(dev->tstats, cpu);
528 do {
529 start = u64_stats_fetch_begin_irq(&stats->syncp);
530 tmp.rx_packets = stats->rx_packets;
531 tmp.rx_bytes = stats->rx_bytes;
532 tmp.tx_packets = stats->tx_packets;
533 tmp.tx_bytes = stats->tx_bytes;
534 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
535
536 s->rx_packets += tmp.rx_packets;
537 s->rx_bytes += tmp.rx_bytes;
538 s->tx_packets += tmp.tx_packets;
539 s->tx_bytes += tmp.tx_bytes;
540 }
541
542 s->rx_dropped = dev->stats.rx_dropped;
543 s->tx_dropped = dev->stats.tx_dropped;
544}
545
546static int xfrmi_get_iflink(const struct net_device *dev)
547{
548 struct xfrm_if *xi = netdev_priv(dev);
549
550 return xi->p.link;
551}
552
553
554static const struct net_device_ops xfrmi_netdev_ops = {
555 .ndo_init = xfrmi_dev_init,
556 .ndo_uninit = xfrmi_dev_uninit,
557 .ndo_start_xmit = xfrmi_xmit,
558 .ndo_get_stats64 = xfrmi_get_stats64,
559 .ndo_get_iflink = xfrmi_get_iflink,
560};
561
562static void xfrmi_dev_setup(struct net_device *dev)
563{
564 dev->netdev_ops = &xfrmi_netdev_ops;
565 dev->type = ARPHRD_NONE;
566 dev->hard_header_len = ETH_HLEN;
567 dev->min_header_len = ETH_HLEN;
568 dev->mtu = ETH_DATA_LEN;
569 dev->min_mtu = ETH_MIN_MTU;
570 dev->max_mtu = ETH_DATA_LEN;
571 dev->addr_len = ETH_ALEN;
572 dev->flags = IFF_NOARP;
573 dev->needs_free_netdev = true;
574 dev->priv_destructor = xfrmi_dev_free;
575 netif_keep_dst(dev);
576
577 eth_broadcast_addr(dev->broadcast);
578}
579
580static 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
598 if (phydev) {
599 dev->needed_headroom = phydev->needed_headroom;
600 dev->needed_tailroom = phydev->needed_tailroom;
601
602 if (is_zero_ether_addr(dev->dev_addr))
603 eth_hw_addr_inherit(dev, phydev);
604 if (is_zero_ether_addr(dev->broadcast))
605 memcpy(dev->broadcast, phydev->broadcast,
606 dev->addr_len);
607 } else {
608 eth_hw_addr_random(dev);
609 eth_broadcast_addr(dev->broadcast);
610 }
611
612 return 0;
613}
614
615static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
616 struct netlink_ext_ack *extack)
617{
618 return 0;
619}
620
621static void xfrmi_netlink_parms(struct nlattr *data[],
622 struct xfrm_if_parms *parms)
623{
624 memset(parms, 0, sizeof(*parms));
625
626 if (!data)
627 return;
628
629 if (data[IFLA_XFRM_LINK])
630 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
631
632 if (data[IFLA_XFRM_IF_ID])
633 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
634}
635
636static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
637 struct nlattr *tb[], struct nlattr *data[],
638 struct netlink_ext_ack *extack)
639{
640 struct net *net = dev_net(dev);
641 struct xfrm_if_parms p;
642 struct xfrm_if *xi;
643 int err;
644
645 xfrmi_netlink_parms(data, &p);
646 xi = xfrmi_locate(net, &p);
647 if (xi)
648 return -EEXIST;
649
650 xi = netdev_priv(dev);
651 xi->p = p;
652 xi->net = net;
653 xi->dev = dev;
654
655 err = xfrmi_create(dev);
656 return err;
657}
658
659static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
660{
661 unregister_netdevice_queue(dev, head);
662}
663
664static int xfrmi_changelink(struct net_device *dev, struct nlattr *tb[],
665 struct nlattr *data[],
666 struct netlink_ext_ack *extack)
667{
668 struct xfrm_if *xi = netdev_priv(dev);
669 struct net *net = xi->net;
670 struct xfrm_if_parms p;
671
672 xfrmi_netlink_parms(data, &p);
673 xi = xfrmi_locate(net, &p);
674 if (!xi) {
675 xi = netdev_priv(dev);
676 } else {
677 if (xi->dev != dev)
678 return -EEXIST;
679 }
680
681 return xfrmi_update(xi, &p);
682}
683
684static size_t xfrmi_get_size(const struct net_device *dev)
685{
686 return
687 /* IFLA_XFRM_LINK */
688 nla_total_size(4) +
689 /* IFLA_XFRM_IF_ID */
690 nla_total_size(4) +
691 0;
692}
693
694static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
695{
696 struct xfrm_if *xi = netdev_priv(dev);
697 struct xfrm_if_parms *parm = &xi->p;
698
699 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
700 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
701 goto nla_put_failure;
702 return 0;
703
704nla_put_failure:
705 return -EMSGSIZE;
706}
707
708static struct net *xfrmi_get_link_net(const struct net_device *dev)
709{
710 struct xfrm_if *xi = netdev_priv(dev);
711
712 return xi->net;
713}
714
715static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
716 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
717 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
718};
719
720static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
721 .kind = "xfrm",
722 .maxtype = IFLA_XFRM_MAX,
723 .policy = xfrmi_policy,
724 .priv_size = sizeof(struct xfrm_if),
725 .setup = xfrmi_dev_setup,
726 .validate = xfrmi_validate,
727 .newlink = xfrmi_newlink,
728 .dellink = xfrmi_dellink,
729 .changelink = xfrmi_changelink,
730 .get_size = xfrmi_get_size,
731 .fill_info = xfrmi_fill_info,
732 .get_link_net = xfrmi_get_link_net,
733};
734
735static void __net_exit xfrmi_destroy_interfaces(struct xfrmi_net *xfrmn)
736{
737 struct xfrm_if *xi;
738 LIST_HEAD(list);
739
740 xi = rtnl_dereference(xfrmn->xfrmi[0]);
741 if (!xi)
742 return;
743
744 unregister_netdevice_queue(xi->dev, &list);
745 unregister_netdevice_many(&list);
746}
747
748static void __net_exit xfrmi_exit_net(struct net *net)
749{
750 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
751
752 rtnl_lock();
753 xfrmi_destroy_interfaces(xfrmn);
754 rtnl_unlock();
755}
756
757static struct pernet_operations xfrmi_net_ops = {
758 .exit = xfrmi_exit_net,
759 .id = &xfrmi_net_id,
760 .size = sizeof(struct xfrmi_net),
761};
762
763static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
764 .handler = xfrm6_rcv,
765 .cb_handler = xfrmi_rcv_cb,
766 .err_handler = xfrmi6_err,
767 .priority = 10,
768};
769
770static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
771 .handler = xfrm6_rcv,
772 .cb_handler = xfrmi_rcv_cb,
773 .err_handler = xfrmi6_err,
774 .priority = 10,
775};
776
777static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
778 .handler = xfrm6_rcv,
779 .cb_handler = xfrmi_rcv_cb,
780 .err_handler = xfrmi6_err,
781 .priority = 10,
782};
783
784static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
785 .handler = xfrm4_rcv,
786 .input_handler = xfrm_input,
787 .cb_handler = xfrmi_rcv_cb,
788 .err_handler = xfrmi4_err,
789 .priority = 10,
790};
791
792static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
793 .handler = xfrm4_rcv,
794 .input_handler = xfrm_input,
795 .cb_handler = xfrmi_rcv_cb,
796 .err_handler = xfrmi4_err,
797 .priority = 10,
798};
799
800static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
801 .handler = xfrm4_rcv,
802 .input_handler = xfrm_input,
803 .cb_handler = xfrmi_rcv_cb,
804 .err_handler = xfrmi4_err,
805 .priority = 10,
806};
807
808static int __init xfrmi4_init(void)
809{
810 int err;
811
812 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
813 if (err < 0)
814 goto xfrm_proto_esp_failed;
815 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
816 if (err < 0)
817 goto xfrm_proto_ah_failed;
818 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
819 if (err < 0)
820 goto xfrm_proto_comp_failed;
821
822 return 0;
823
824xfrm_proto_comp_failed:
825 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
826xfrm_proto_ah_failed:
827 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
828xfrm_proto_esp_failed:
829 return err;
830}
831
832static void xfrmi4_fini(void)
833{
834 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
835 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
836 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
837}
838
839static int __init xfrmi6_init(void)
840{
841 int err;
842
843 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
844 if (err < 0)
845 goto xfrm_proto_esp_failed;
846 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
847 if (err < 0)
848 goto xfrm_proto_ah_failed;
849 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
850 if (err < 0)
851 goto xfrm_proto_comp_failed;
852
853 return 0;
854
855xfrm_proto_comp_failed:
856 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
857xfrm_proto_ah_failed:
858 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
859xfrm_proto_esp_failed:
860 return err;
861}
862
863static void xfrmi6_fini(void)
864{
865 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
866 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
867 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
868}
869
870static const struct xfrm_if_cb xfrm_if_cb = {
871 .decode_session = xfrmi_decode_session,
872};
873
874static int __init xfrmi_init(void)
875{
876 const char *msg;
877 int err;
878
879 pr_info("IPsec XFRM device driver\n");
880
881 msg = "tunnel device";
882 err = register_pernet_device(&xfrmi_net_ops);
883 if (err < 0)
884 goto pernet_dev_failed;
885
886 msg = "xfrm4 protocols";
887 err = xfrmi4_init();
888 if (err < 0)
889 goto xfrmi4_failed;
890
891 msg = "xfrm6 protocols";
892 err = xfrmi6_init();
893 if (err < 0)
894 goto xfrmi6_failed;
895
896
897 msg = "netlink interface";
898 err = rtnl_link_register(&xfrmi_link_ops);
899 if (err < 0)
900 goto rtnl_link_failed;
901
902 xfrm_if_register_cb(&xfrm_if_cb);
903
904 return err;
905
906rtnl_link_failed:
907 xfrmi6_fini();
908xfrmi6_failed:
909 xfrmi4_fini();
910xfrmi4_failed:
911 unregister_pernet_device(&xfrmi_net_ops);
912pernet_dev_failed:
913 pr_err("xfrmi init: failed to register %s\n", msg);
914 return err;
915}
916
917static void __exit xfrmi_fini(void)
918{
919 xfrm_if_unregister_cb();
920 rtnl_link_unregister(&xfrmi_link_ops);
921 xfrmi4_fini();
922 xfrmi6_fini();
923 unregister_pernet_device(&xfrmi_net_ops);
924}
925
926module_init(xfrmi_init);
927module_exit(xfrmi_fini);
928MODULE_LICENSE("GPL");
929MODULE_ALIAS_RTNL_LINK("xfrm");
930MODULE_ALIAS_NETDEV("xfrm0");
931MODULE_AUTHOR("Steffen Klassert");
932MODULE_DESCRIPTION("XFRM virtual interface");
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
47static int xfrmi_dev_init(struct net_device *dev);
48static void xfrmi_dev_setup(struct net_device *dev);
49static struct rtnl_link_ops xfrmi_link_ops __read_mostly;
50static unsigned int xfrmi_net_id __read_mostly;
51static 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
56struct 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
64static u32 xfrmi_hash(u32 if_id)
65{
66 return hash_32(if_id, XFRMI_HASH_BITS);
67}
68
69static 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
83static 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
117static 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
125static 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
140static 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
148static 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
164out:
165 return err;
166}
167
168static 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
183static 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
191static void xfrmi_scrub_packet(struct sk_buff *skb, bool xnet)
192{
193 skb->tstamp = 0;
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
210static int xfrmi_rcv_cb(struct sk_buff *skb, int err)
211{
212 const struct xfrm_mode *inner_mode;
213 struct pcpu_sw_netstats *tstats;
214 struct net_device *dev;
215 struct xfrm_state *x;
216 struct xfrm_if *xi;
217 bool xnet;
218
219 if (err && !secpath_exists(skb))
220 return 0;
221
222 x = xfrm_input_state(skb);
223
224 xi = xfrmi_lookup(xs_net(x), x);
225 if (!xi)
226 return 1;
227
228 dev = xi->dev;
229 skb->dev = dev;
230
231 if (err) {
232 dev->stats.rx_errors++;
233 dev->stats.rx_dropped++;
234
235 return 0;
236 }
237
238 xnet = !net_eq(xi->net, dev_net(skb->dev));
239
240 if (xnet) {
241 inner_mode = &x->inner_mode;
242
243 if (x->sel.family == AF_UNSPEC) {
244 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
245 if (inner_mode == NULL) {
246 XFRM_INC_STATS(dev_net(skb->dev),
247 LINUX_MIB_XFRMINSTATEMODEERROR);
248 return -EINVAL;
249 }
250 }
251
252 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb,
253 inner_mode->family))
254 return -EPERM;
255 }
256
257 xfrmi_scrub_packet(skb, xnet);
258
259 tstats = this_cpu_ptr(dev->tstats);
260
261 u64_stats_update_begin(&tstats->syncp);
262 tstats->rx_packets++;
263 tstats->rx_bytes += skb->len;
264 u64_stats_update_end(&tstats->syncp);
265
266 return 0;
267}
268
269static int
270xfrmi_xmit2(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
271{
272 struct xfrm_if *xi = netdev_priv(dev);
273 struct net_device_stats *stats = &xi->dev->stats;
274 struct dst_entry *dst = skb_dst(skb);
275 unsigned int length = skb->len;
276 struct net_device *tdev;
277 struct xfrm_state *x;
278 int err = -1;
279 int mtu;
280
281 dst_hold(dst);
282 dst = xfrm_lookup_with_ifid(xi->net, dst, fl, NULL, 0, xi->p.if_id);
283 if (IS_ERR(dst)) {
284 err = PTR_ERR(dst);
285 dst = NULL;
286 goto tx_err_link_failure;
287 }
288
289 x = dst->xfrm;
290 if (!x)
291 goto tx_err_link_failure;
292
293 if (x->if_id != xi->p.if_id)
294 goto tx_err_link_failure;
295
296 tdev = dst->dev;
297
298 if (tdev == dev) {
299 stats->collisions++;
300 net_warn_ratelimited("%s: Local routing loop detected!\n",
301 dev->name);
302 goto tx_err_dst_release;
303 }
304
305 mtu = dst_mtu(dst);
306 if (skb->len > mtu) {
307 skb_dst_update_pmtu_no_confirm(skb, mtu);
308
309 if (skb->protocol == htons(ETH_P_IPV6)) {
310 if (mtu < IPV6_MIN_MTU)
311 mtu = IPV6_MIN_MTU;
312
313 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
314 } else {
315 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
316 htonl(mtu));
317 }
318
319 dst_release(dst);
320 return -EMSGSIZE;
321 }
322
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 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
330
331 u64_stats_update_begin(&tstats->syncp);
332 tstats->tx_bytes += length;
333 tstats->tx_packets++;
334 u64_stats_update_end(&tstats->syncp);
335 } else {
336 stats->tx_errors++;
337 stats->tx_aborted_errors++;
338 }
339
340 return 0;
341tx_err_link_failure:
342 stats->tx_carrier_errors++;
343 dst_link_failure(skb);
344tx_err_dst_release:
345 dst_release(dst);
346 return err;
347}
348
349static netdev_tx_t xfrmi_xmit(struct sk_buff *skb, struct net_device *dev)
350{
351 struct xfrm_if *xi = netdev_priv(dev);
352 struct net_device_stats *stats = &xi->dev->stats;
353 struct dst_entry *dst = skb_dst(skb);
354 struct flowi fl;
355 int ret;
356
357 memset(&fl, 0, sizeof(fl));
358
359 switch (skb->protocol) {
360 case htons(ETH_P_IPV6):
361 xfrm_decode_session(skb, &fl, AF_INET6);
362 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
363 if (!dst) {
364 fl.u.ip6.flowi6_oif = dev->ifindex;
365 fl.u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
366 dst = ip6_route_output(dev_net(dev), NULL, &fl.u.ip6);
367 if (dst->error) {
368 dst_release(dst);
369 stats->tx_carrier_errors++;
370 goto tx_err;
371 }
372 skb_dst_set(skb, dst);
373 }
374 break;
375 case htons(ETH_P_IP):
376 xfrm_decode_session(skb, &fl, AF_INET);
377 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
378 if (!dst) {
379 struct rtable *rt;
380
381 fl.u.ip4.flowi4_oif = dev->ifindex;
382 fl.u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
383 rt = __ip_route_output_key(dev_net(dev), &fl.u.ip4);
384 if (IS_ERR(rt)) {
385 stats->tx_carrier_errors++;
386 goto tx_err;
387 }
388 skb_dst_set(skb, &rt->dst);
389 }
390 break;
391 default:
392 goto tx_err;
393 }
394
395 fl.flowi_oif = xi->p.link;
396
397 ret = xfrmi_xmit2(skb, dev, &fl);
398 if (ret < 0)
399 goto tx_err;
400
401 return NETDEV_TX_OK;
402
403tx_err:
404 stats->tx_errors++;
405 stats->tx_dropped++;
406 kfree_skb(skb);
407 return NETDEV_TX_OK;
408}
409
410static int xfrmi4_err(struct sk_buff *skb, u32 info)
411{
412 const struct iphdr *iph = (const struct iphdr *)skb->data;
413 struct net *net = dev_net(skb->dev);
414 int protocol = iph->protocol;
415 struct ip_comp_hdr *ipch;
416 struct ip_esp_hdr *esph;
417 struct ip_auth_hdr *ah ;
418 struct xfrm_state *x;
419 struct xfrm_if *xi;
420 __be32 spi;
421
422 switch (protocol) {
423 case IPPROTO_ESP:
424 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
425 spi = esph->spi;
426 break;
427 case IPPROTO_AH:
428 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
429 spi = ah->spi;
430 break;
431 case IPPROTO_COMP:
432 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
433 spi = htonl(ntohs(ipch->cpi));
434 break;
435 default:
436 return 0;
437 }
438
439 switch (icmp_hdr(skb)->type) {
440 case ICMP_DEST_UNREACH:
441 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
442 return 0;
443 case ICMP_REDIRECT:
444 break;
445 default:
446 return 0;
447 }
448
449 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
450 spi, protocol, AF_INET);
451 if (!x)
452 return 0;
453
454 xi = xfrmi_lookup(net, x);
455 if (!xi) {
456 xfrm_state_put(x);
457 return -1;
458 }
459
460 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
461 ipv4_update_pmtu(skb, net, info, 0, protocol);
462 else
463 ipv4_redirect(skb, net, 0, protocol);
464 xfrm_state_put(x);
465
466 return 0;
467}
468
469static int xfrmi6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
470 u8 type, u8 code, int offset, __be32 info)
471{
472 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
473 struct net *net = dev_net(skb->dev);
474 int protocol = iph->nexthdr;
475 struct ip_comp_hdr *ipch;
476 struct ip_esp_hdr *esph;
477 struct ip_auth_hdr *ah;
478 struct xfrm_state *x;
479 struct xfrm_if *xi;
480 __be32 spi;
481
482 switch (protocol) {
483 case IPPROTO_ESP:
484 esph = (struct ip_esp_hdr *)(skb->data + offset);
485 spi = esph->spi;
486 break;
487 case IPPROTO_AH:
488 ah = (struct ip_auth_hdr *)(skb->data + offset);
489 spi = ah->spi;
490 break;
491 case IPPROTO_COMP:
492 ipch = (struct ip_comp_hdr *)(skb->data + offset);
493 spi = htonl(ntohs(ipch->cpi));
494 break;
495 default:
496 return 0;
497 }
498
499 if (type != ICMPV6_PKT_TOOBIG &&
500 type != NDISC_REDIRECT)
501 return 0;
502
503 x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
504 spi, protocol, AF_INET6);
505 if (!x)
506 return 0;
507
508 xi = xfrmi_lookup(net, x);
509 if (!xi) {
510 xfrm_state_put(x);
511 return -1;
512 }
513
514 if (type == NDISC_REDIRECT)
515 ip6_redirect(skb, net, skb->dev->ifindex, 0,
516 sock_net_uid(net, NULL));
517 else
518 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
519 xfrm_state_put(x);
520
521 return 0;
522}
523
524static int xfrmi_change(struct xfrm_if *xi, const struct xfrm_if_parms *p)
525{
526 if (xi->p.link != p->link)
527 return -EINVAL;
528
529 xi->p.if_id = p->if_id;
530
531 return 0;
532}
533
534static int xfrmi_update(struct xfrm_if *xi, struct xfrm_if_parms *p)
535{
536 struct net *net = xi->net;
537 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
538 int err;
539
540 xfrmi_unlink(xfrmn, xi);
541 synchronize_net();
542 err = xfrmi_change(xi, p);
543 xfrmi_link(xfrmn, xi);
544 netdev_state_change(xi->dev);
545 return err;
546}
547
548static void xfrmi_get_stats64(struct net_device *dev,
549 struct rtnl_link_stats64 *s)
550{
551 int cpu;
552
553 for_each_possible_cpu(cpu) {
554 struct pcpu_sw_netstats *stats;
555 struct pcpu_sw_netstats tmp;
556 int start;
557
558 stats = per_cpu_ptr(dev->tstats, cpu);
559 do {
560 start = u64_stats_fetch_begin_irq(&stats->syncp);
561 tmp.rx_packets = stats->rx_packets;
562 tmp.rx_bytes = stats->rx_bytes;
563 tmp.tx_packets = stats->tx_packets;
564 tmp.tx_bytes = stats->tx_bytes;
565 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
566
567 s->rx_packets += tmp.rx_packets;
568 s->rx_bytes += tmp.rx_bytes;
569 s->tx_packets += tmp.tx_packets;
570 s->tx_bytes += tmp.tx_bytes;
571 }
572
573 s->rx_dropped = dev->stats.rx_dropped;
574 s->tx_dropped = dev->stats.tx_dropped;
575}
576
577static int xfrmi_get_iflink(const struct net_device *dev)
578{
579 struct xfrm_if *xi = netdev_priv(dev);
580
581 return xi->p.link;
582}
583
584
585static const struct net_device_ops xfrmi_netdev_ops = {
586 .ndo_init = xfrmi_dev_init,
587 .ndo_uninit = xfrmi_dev_uninit,
588 .ndo_start_xmit = xfrmi_xmit,
589 .ndo_get_stats64 = xfrmi_get_stats64,
590 .ndo_get_iflink = xfrmi_get_iflink,
591};
592
593static void xfrmi_dev_setup(struct net_device *dev)
594{
595 dev->netdev_ops = &xfrmi_netdev_ops;
596 dev->header_ops = &ip_tunnel_header_ops;
597 dev->type = ARPHRD_NONE;
598 dev->mtu = ETH_DATA_LEN;
599 dev->min_mtu = ETH_MIN_MTU;
600 dev->max_mtu = IP_MAX_MTU;
601 dev->flags = IFF_NOARP;
602 dev->needs_free_netdev = true;
603 dev->priv_destructor = xfrmi_dev_free;
604 netif_keep_dst(dev);
605
606 eth_broadcast_addr(dev->broadcast);
607}
608
609static int xfrmi_dev_init(struct net_device *dev)
610{
611 struct xfrm_if *xi = netdev_priv(dev);
612 struct net_device *phydev = __dev_get_by_index(xi->net, xi->p.link);
613 int err;
614
615 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
616 if (!dev->tstats)
617 return -ENOMEM;
618
619 err = gro_cells_init(&xi->gro_cells, dev);
620 if (err) {
621 free_percpu(dev->tstats);
622 return err;
623 }
624
625 dev->features |= NETIF_F_LLTX;
626
627 if (phydev) {
628 dev->needed_headroom = phydev->needed_headroom;
629 dev->needed_tailroom = phydev->needed_tailroom;
630
631 if (is_zero_ether_addr(dev->dev_addr))
632 eth_hw_addr_inherit(dev, phydev);
633 if (is_zero_ether_addr(dev->broadcast))
634 memcpy(dev->broadcast, phydev->broadcast,
635 dev->addr_len);
636 } else {
637 eth_hw_addr_random(dev);
638 eth_broadcast_addr(dev->broadcast);
639 }
640
641 return 0;
642}
643
644static int xfrmi_validate(struct nlattr *tb[], struct nlattr *data[],
645 struct netlink_ext_ack *extack)
646{
647 return 0;
648}
649
650static void xfrmi_netlink_parms(struct nlattr *data[],
651 struct xfrm_if_parms *parms)
652{
653 memset(parms, 0, sizeof(*parms));
654
655 if (!data)
656 return;
657
658 if (data[IFLA_XFRM_LINK])
659 parms->link = nla_get_u32(data[IFLA_XFRM_LINK]);
660
661 if (data[IFLA_XFRM_IF_ID])
662 parms->if_id = nla_get_u32(data[IFLA_XFRM_IF_ID]);
663}
664
665static int xfrmi_newlink(struct net *src_net, struct net_device *dev,
666 struct nlattr *tb[], struct nlattr *data[],
667 struct netlink_ext_ack *extack)
668{
669 struct net *net = dev_net(dev);
670 struct xfrm_if_parms p;
671 struct xfrm_if *xi;
672 int err;
673
674 xfrmi_netlink_parms(data, &p);
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
688static void xfrmi_dellink(struct net_device *dev, struct list_head *head)
689{
690 unregister_netdevice_queue(dev, head);
691}
692
693static 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 xi = xfrmi_locate(net, &p);
703 if (!xi) {
704 xi = netdev_priv(dev);
705 } else {
706 if (xi->dev != dev)
707 return -EEXIST;
708 }
709
710 return xfrmi_update(xi, &p);
711}
712
713static size_t xfrmi_get_size(const struct net_device *dev)
714{
715 return
716 /* IFLA_XFRM_LINK */
717 nla_total_size(4) +
718 /* IFLA_XFRM_IF_ID */
719 nla_total_size(4) +
720 0;
721}
722
723static int xfrmi_fill_info(struct sk_buff *skb, const struct net_device *dev)
724{
725 struct xfrm_if *xi = netdev_priv(dev);
726 struct xfrm_if_parms *parm = &xi->p;
727
728 if (nla_put_u32(skb, IFLA_XFRM_LINK, parm->link) ||
729 nla_put_u32(skb, IFLA_XFRM_IF_ID, parm->if_id))
730 goto nla_put_failure;
731 return 0;
732
733nla_put_failure:
734 return -EMSGSIZE;
735}
736
737static struct net *xfrmi_get_link_net(const struct net_device *dev)
738{
739 struct xfrm_if *xi = netdev_priv(dev);
740
741 return xi->net;
742}
743
744static const struct nla_policy xfrmi_policy[IFLA_XFRM_MAX + 1] = {
745 [IFLA_XFRM_LINK] = { .type = NLA_U32 },
746 [IFLA_XFRM_IF_ID] = { .type = NLA_U32 },
747};
748
749static struct rtnl_link_ops xfrmi_link_ops __read_mostly = {
750 .kind = "xfrm",
751 .maxtype = IFLA_XFRM_MAX,
752 .policy = xfrmi_policy,
753 .priv_size = sizeof(struct xfrm_if),
754 .setup = xfrmi_dev_setup,
755 .validate = xfrmi_validate,
756 .newlink = xfrmi_newlink,
757 .dellink = xfrmi_dellink,
758 .changelink = xfrmi_changelink,
759 .get_size = xfrmi_get_size,
760 .fill_info = xfrmi_fill_info,
761 .get_link_net = xfrmi_get_link_net,
762};
763
764static void __net_exit xfrmi_exit_batch_net(struct list_head *net_exit_list)
765{
766 struct net *net;
767 LIST_HEAD(list);
768
769 rtnl_lock();
770 list_for_each_entry(net, net_exit_list, exit_list) {
771 struct xfrmi_net *xfrmn = net_generic(net, xfrmi_net_id);
772 struct xfrm_if __rcu **xip;
773 struct xfrm_if *xi;
774 int i;
775
776 for (i = 0; i < XFRMI_HASH_SIZE; i++) {
777 for (xip = &xfrmn->xfrmi[i];
778 (xi = rtnl_dereference(*xip)) != NULL;
779 xip = &xi->next)
780 unregister_netdevice_queue(xi->dev, &list);
781 }
782 }
783 unregister_netdevice_many(&list);
784 rtnl_unlock();
785}
786
787static struct pernet_operations xfrmi_net_ops = {
788 .exit_batch = xfrmi_exit_batch_net,
789 .id = &xfrmi_net_id,
790 .size = sizeof(struct xfrmi_net),
791};
792
793static struct xfrm6_protocol xfrmi_esp6_protocol __read_mostly = {
794 .handler = xfrm6_rcv,
795 .input_handler = xfrm_input,
796 .cb_handler = xfrmi_rcv_cb,
797 .err_handler = xfrmi6_err,
798 .priority = 10,
799};
800
801static struct xfrm6_protocol xfrmi_ah6_protocol __read_mostly = {
802 .handler = xfrm6_rcv,
803 .input_handler = xfrm_input,
804 .cb_handler = xfrmi_rcv_cb,
805 .err_handler = xfrmi6_err,
806 .priority = 10,
807};
808
809static struct xfrm6_protocol xfrmi_ipcomp6_protocol __read_mostly = {
810 .handler = xfrm6_rcv,
811 .input_handler = xfrm_input,
812 .cb_handler = xfrmi_rcv_cb,
813 .err_handler = xfrmi6_err,
814 .priority = 10,
815};
816
817#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
818static int xfrmi6_rcv_tunnel(struct sk_buff *skb)
819{
820 const xfrm_address_t *saddr;
821 __be32 spi;
822
823 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
824 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
825
826 return xfrm6_rcv_spi(skb, IPPROTO_IPV6, spi, NULL);
827}
828
829static struct xfrm6_tunnel xfrmi_ipv6_handler __read_mostly = {
830 .handler = xfrmi6_rcv_tunnel,
831 .cb_handler = xfrmi_rcv_cb,
832 .err_handler = xfrmi6_err,
833 .priority = -1,
834};
835
836static struct xfrm6_tunnel xfrmi_ip6ip_handler __read_mostly = {
837 .handler = xfrmi6_rcv_tunnel,
838 .cb_handler = xfrmi_rcv_cb,
839 .err_handler = xfrmi6_err,
840 .priority = -1,
841};
842#endif
843
844static struct xfrm4_protocol xfrmi_esp4_protocol __read_mostly = {
845 .handler = xfrm4_rcv,
846 .input_handler = xfrm_input,
847 .cb_handler = xfrmi_rcv_cb,
848 .err_handler = xfrmi4_err,
849 .priority = 10,
850};
851
852static struct xfrm4_protocol xfrmi_ah4_protocol __read_mostly = {
853 .handler = xfrm4_rcv,
854 .input_handler = xfrm_input,
855 .cb_handler = xfrmi_rcv_cb,
856 .err_handler = xfrmi4_err,
857 .priority = 10,
858};
859
860static struct xfrm4_protocol xfrmi_ipcomp4_protocol __read_mostly = {
861 .handler = xfrm4_rcv,
862 .input_handler = xfrm_input,
863 .cb_handler = xfrmi_rcv_cb,
864 .err_handler = xfrmi4_err,
865 .priority = 10,
866};
867
868#if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
869static int xfrmi4_rcv_tunnel(struct sk_buff *skb)
870{
871 return xfrm4_rcv_spi(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr);
872}
873
874static struct xfrm_tunnel xfrmi_ipip_handler __read_mostly = {
875 .handler = xfrmi4_rcv_tunnel,
876 .cb_handler = xfrmi_rcv_cb,
877 .err_handler = xfrmi4_err,
878 .priority = -1,
879};
880
881static struct xfrm_tunnel xfrmi_ipip6_handler __read_mostly = {
882 .handler = xfrmi4_rcv_tunnel,
883 .cb_handler = xfrmi_rcv_cb,
884 .err_handler = xfrmi4_err,
885 .priority = -1,
886};
887#endif
888
889static int __init xfrmi4_init(void)
890{
891 int err;
892
893 err = xfrm4_protocol_register(&xfrmi_esp4_protocol, IPPROTO_ESP);
894 if (err < 0)
895 goto xfrm_proto_esp_failed;
896 err = xfrm4_protocol_register(&xfrmi_ah4_protocol, IPPROTO_AH);
897 if (err < 0)
898 goto xfrm_proto_ah_failed;
899 err = xfrm4_protocol_register(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
900 if (err < 0)
901 goto xfrm_proto_comp_failed;
902#if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
903 err = xfrm4_tunnel_register(&xfrmi_ipip_handler, AF_INET);
904 if (err < 0)
905 goto xfrm_tunnel_ipip_failed;
906 err = xfrm4_tunnel_register(&xfrmi_ipip6_handler, AF_INET6);
907 if (err < 0)
908 goto xfrm_tunnel_ipip6_failed;
909#endif
910
911 return 0;
912
913#if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
914xfrm_tunnel_ipip6_failed:
915 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
916xfrm_tunnel_ipip_failed:
917 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
918#endif
919xfrm_proto_comp_failed:
920 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
921xfrm_proto_ah_failed:
922 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
923xfrm_proto_esp_failed:
924 return err;
925}
926
927static void xfrmi4_fini(void)
928{
929#if IS_REACHABLE(CONFIG_INET_XFRM_TUNNEL)
930 xfrm4_tunnel_deregister(&xfrmi_ipip6_handler, AF_INET6);
931 xfrm4_tunnel_deregister(&xfrmi_ipip_handler, AF_INET);
932#endif
933 xfrm4_protocol_deregister(&xfrmi_ipcomp4_protocol, IPPROTO_COMP);
934 xfrm4_protocol_deregister(&xfrmi_ah4_protocol, IPPROTO_AH);
935 xfrm4_protocol_deregister(&xfrmi_esp4_protocol, IPPROTO_ESP);
936}
937
938static int __init xfrmi6_init(void)
939{
940 int err;
941
942 err = xfrm6_protocol_register(&xfrmi_esp6_protocol, IPPROTO_ESP);
943 if (err < 0)
944 goto xfrm_proto_esp_failed;
945 err = xfrm6_protocol_register(&xfrmi_ah6_protocol, IPPROTO_AH);
946 if (err < 0)
947 goto xfrm_proto_ah_failed;
948 err = xfrm6_protocol_register(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
949 if (err < 0)
950 goto xfrm_proto_comp_failed;
951#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
952 err = xfrm6_tunnel_register(&xfrmi_ipv6_handler, AF_INET6);
953 if (err < 0)
954 goto xfrm_tunnel_ipv6_failed;
955 err = xfrm6_tunnel_register(&xfrmi_ip6ip_handler, AF_INET);
956 if (err < 0)
957 goto xfrm_tunnel_ip6ip_failed;
958#endif
959
960 return 0;
961
962#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
963xfrm_tunnel_ip6ip_failed:
964 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
965xfrm_tunnel_ipv6_failed:
966 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
967#endif
968xfrm_proto_comp_failed:
969 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
970xfrm_proto_ah_failed:
971 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
972xfrm_proto_esp_failed:
973 return err;
974}
975
976static void xfrmi6_fini(void)
977{
978#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
979 xfrm6_tunnel_deregister(&xfrmi_ip6ip_handler, AF_INET);
980 xfrm6_tunnel_deregister(&xfrmi_ipv6_handler, AF_INET6);
981#endif
982 xfrm6_protocol_deregister(&xfrmi_ipcomp6_protocol, IPPROTO_COMP);
983 xfrm6_protocol_deregister(&xfrmi_ah6_protocol, IPPROTO_AH);
984 xfrm6_protocol_deregister(&xfrmi_esp6_protocol, IPPROTO_ESP);
985}
986
987static const struct xfrm_if_cb xfrm_if_cb = {
988 .decode_session = xfrmi_decode_session,
989};
990
991static int __init xfrmi_init(void)
992{
993 const char *msg;
994 int err;
995
996 pr_info("IPsec XFRM device driver\n");
997
998 msg = "tunnel device";
999 err = register_pernet_device(&xfrmi_net_ops);
1000 if (err < 0)
1001 goto pernet_dev_failed;
1002
1003 msg = "xfrm4 protocols";
1004 err = xfrmi4_init();
1005 if (err < 0)
1006 goto xfrmi4_failed;
1007
1008 msg = "xfrm6 protocols";
1009 err = xfrmi6_init();
1010 if (err < 0)
1011 goto xfrmi6_failed;
1012
1013
1014 msg = "netlink interface";
1015 err = rtnl_link_register(&xfrmi_link_ops);
1016 if (err < 0)
1017 goto rtnl_link_failed;
1018
1019 xfrm_if_register_cb(&xfrm_if_cb);
1020
1021 return err;
1022
1023rtnl_link_failed:
1024 xfrmi6_fini();
1025xfrmi6_failed:
1026 xfrmi4_fini();
1027xfrmi4_failed:
1028 unregister_pernet_device(&xfrmi_net_ops);
1029pernet_dev_failed:
1030 pr_err("xfrmi init: failed to register %s\n", msg);
1031 return err;
1032}
1033
1034static void __exit xfrmi_fini(void)
1035{
1036 xfrm_if_unregister_cb();
1037 rtnl_link_unregister(&xfrmi_link_ops);
1038 xfrmi4_fini();
1039 xfrmi6_fini();
1040 unregister_pernet_device(&xfrmi_net_ops);
1041}
1042
1043module_init(xfrmi_init);
1044module_exit(xfrmi_fini);
1045MODULE_LICENSE("GPL");
1046MODULE_ALIAS_RTNL_LINK("xfrm");
1047MODULE_ALIAS_NETDEV("xfrm0");
1048MODULE_AUTHOR("Steffen Klassert");
1049MODULE_DESCRIPTION("XFRM virtual interface");