Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm_input.c
4 *
5 * Changes:
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
8 *
9 */
10
11#include <linux/bottom_half.h>
12#include <linux/cache.h>
13#include <linux/interrupt.h>
14#include <linux/slab.h>
15#include <linux/module.h>
16#include <linux/netdevice.h>
17#include <linux/percpu.h>
18#include <net/dst.h>
19#include <net/ip.h>
20#include <net/xfrm.h>
21#include <net/ip_tunnels.h>
22#include <net/ip6_tunnel.h>
23
24#include "xfrm_inout.h"
25
26struct xfrm_trans_tasklet {
27 struct tasklet_struct tasklet;
28 struct sk_buff_head queue;
29};
30
31struct xfrm_trans_cb {
32 union {
33 struct inet_skb_parm h4;
34#if IS_ENABLED(CONFIG_IPV6)
35 struct inet6_skb_parm h6;
36#endif
37 } header;
38 int (*finish)(struct net *net, struct sock *sk, struct sk_buff *skb);
39 struct net *net;
40};
41
42#define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
43
44static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
45static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
46
47static struct gro_cells gro_cells;
48static struct net_device xfrm_napi_dev;
49
50static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
51
52int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
53{
54 int err = 0;
55
56 if (WARN_ON(afinfo->family > AF_INET6))
57 return -EAFNOSUPPORT;
58
59 spin_lock_bh(&xfrm_input_afinfo_lock);
60 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family]))
61 err = -EEXIST;
62 else
63 rcu_assign_pointer(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], afinfo);
64 spin_unlock_bh(&xfrm_input_afinfo_lock);
65 return err;
66}
67EXPORT_SYMBOL(xfrm_input_register_afinfo);
68
69int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
70{
71 int err = 0;
72
73 spin_lock_bh(&xfrm_input_afinfo_lock);
74 if (likely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family])) {
75 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family] != afinfo))
76 err = -EINVAL;
77 else
78 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], NULL);
79 }
80 spin_unlock_bh(&xfrm_input_afinfo_lock);
81 synchronize_rcu();
82 return err;
83}
84EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
85
86static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(u8 family, bool is_ipip)
87{
88 const struct xfrm_input_afinfo *afinfo;
89
90 if (WARN_ON_ONCE(family > AF_INET6))
91 return NULL;
92
93 rcu_read_lock();
94 afinfo = rcu_dereference(xfrm_input_afinfo[is_ipip][family]);
95 if (unlikely(!afinfo))
96 rcu_read_unlock();
97 return afinfo;
98}
99
100static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
101 int err)
102{
103 bool is_ipip = (protocol == IPPROTO_IPIP || protocol == IPPROTO_IPV6);
104 const struct xfrm_input_afinfo *afinfo;
105 int ret;
106
107 afinfo = xfrm_input_get_afinfo(family, is_ipip);
108 if (!afinfo)
109 return -EAFNOSUPPORT;
110
111 ret = afinfo->callback(skb, protocol, err);
112 rcu_read_unlock();
113
114 return ret;
115}
116
117struct sec_path *secpath_set(struct sk_buff *skb)
118{
119 struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
120
121 sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
122 if (!sp)
123 return NULL;
124
125 if (tmp) /* reused existing one (was COW'd if needed) */
126 return sp;
127
128 /* allocated new secpath */
129 memset(sp->ovec, 0, sizeof(sp->ovec));
130 sp->olen = 0;
131 sp->len = 0;
132
133 return sp;
134}
135EXPORT_SYMBOL(secpath_set);
136
137/* Fetch spi and seq from ipsec header */
138
139int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
140{
141 int offset, offset_seq;
142 int hlen;
143
144 switch (nexthdr) {
145 case IPPROTO_AH:
146 hlen = sizeof(struct ip_auth_hdr);
147 offset = offsetof(struct ip_auth_hdr, spi);
148 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
149 break;
150 case IPPROTO_ESP:
151 hlen = sizeof(struct ip_esp_hdr);
152 offset = offsetof(struct ip_esp_hdr, spi);
153 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
154 break;
155 case IPPROTO_COMP:
156 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
157 return -EINVAL;
158 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
159 *seq = 0;
160 return 0;
161 default:
162 return 1;
163 }
164
165 if (!pskb_may_pull(skb, hlen))
166 return -EINVAL;
167
168 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
169 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
170 return 0;
171}
172EXPORT_SYMBOL(xfrm_parse_spi);
173
174static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
175{
176 struct iphdr *iph;
177 int optlen = 0;
178 int err = -EINVAL;
179
180 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
181 struct ip_beet_phdr *ph;
182 int phlen;
183
184 if (!pskb_may_pull(skb, sizeof(*ph)))
185 goto out;
186
187 ph = (struct ip_beet_phdr *)skb->data;
188
189 phlen = sizeof(*ph) + ph->padlen;
190 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
191 if (optlen < 0 || optlen & 3 || optlen > 250)
192 goto out;
193
194 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
195
196 if (!pskb_may_pull(skb, phlen))
197 goto out;
198 __skb_pull(skb, phlen);
199 }
200
201 skb_push(skb, sizeof(*iph));
202 skb_reset_network_header(skb);
203 skb_mac_header_rebuild(skb);
204
205 xfrm4_beet_make_header(skb);
206
207 iph = ip_hdr(skb);
208
209 iph->ihl += optlen / 4;
210 iph->tot_len = htons(skb->len);
211 iph->daddr = x->sel.daddr.a4;
212 iph->saddr = x->sel.saddr.a4;
213 iph->check = 0;
214 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
215 err = 0;
216out:
217 return err;
218}
219
220static void ipip_ecn_decapsulate(struct sk_buff *skb)
221{
222 struct iphdr *inner_iph = ipip_hdr(skb);
223
224 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
225 IP_ECN_set_ce(inner_iph);
226}
227
228static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
229{
230 int err = -EINVAL;
231
232 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPIP)
233 goto out;
234
235 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
236 goto out;
237
238 err = skb_unclone(skb, GFP_ATOMIC);
239 if (err)
240 goto out;
241
242 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
243 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
244 if (!(x->props.flags & XFRM_STATE_NOECN))
245 ipip_ecn_decapsulate(skb);
246
247 skb_reset_network_header(skb);
248 skb_mac_header_rebuild(skb);
249 if (skb->mac_len)
250 eth_hdr(skb)->h_proto = skb->protocol;
251
252 err = 0;
253
254out:
255 return err;
256}
257
258static void ipip6_ecn_decapsulate(struct sk_buff *skb)
259{
260 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
261
262 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
263 IP6_ECN_set_ce(skb, inner_iph);
264}
265
266static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
267{
268 int err = -EINVAL;
269
270 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPV6)
271 goto out;
272 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
273 goto out;
274
275 err = skb_unclone(skb, GFP_ATOMIC);
276 if (err)
277 goto out;
278
279 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
280 ipv6_copy_dscp(ipv6_get_dsfield(ipv6_hdr(skb)),
281 ipipv6_hdr(skb));
282 if (!(x->props.flags & XFRM_STATE_NOECN))
283 ipip6_ecn_decapsulate(skb);
284
285 skb_reset_network_header(skb);
286 skb_mac_header_rebuild(skb);
287 if (skb->mac_len)
288 eth_hdr(skb)->h_proto = skb->protocol;
289
290 err = 0;
291
292out:
293 return err;
294}
295
296static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
297{
298 struct ipv6hdr *ip6h;
299 int size = sizeof(struct ipv6hdr);
300 int err;
301
302 err = skb_cow_head(skb, size + skb->mac_len);
303 if (err)
304 goto out;
305
306 __skb_push(skb, size);
307 skb_reset_network_header(skb);
308 skb_mac_header_rebuild(skb);
309
310 xfrm6_beet_make_header(skb);
311
312 ip6h = ipv6_hdr(skb);
313 ip6h->payload_len = htons(skb->len - size);
314 ip6h->daddr = x->sel.daddr.in6;
315 ip6h->saddr = x->sel.saddr.in6;
316 err = 0;
317out:
318 return err;
319}
320
321/* Remove encapsulation header.
322 *
323 * The IP header will be moved over the top of the encapsulation
324 * header.
325 *
326 * On entry, the transport header shall point to where the IP header
327 * should be and the network header shall be set to where the IP
328 * header currently is. skb->data shall point to the start of the
329 * payload.
330 */
331static int
332xfrm_inner_mode_encap_remove(struct xfrm_state *x,
333 const struct xfrm_mode *inner_mode,
334 struct sk_buff *skb)
335{
336 switch (inner_mode->encap) {
337 case XFRM_MODE_BEET:
338 if (inner_mode->family == AF_INET)
339 return xfrm4_remove_beet_encap(x, skb);
340 if (inner_mode->family == AF_INET6)
341 return xfrm6_remove_beet_encap(x, skb);
342 break;
343 case XFRM_MODE_TUNNEL:
344 if (inner_mode->family == AF_INET)
345 return xfrm4_remove_tunnel_encap(x, skb);
346 if (inner_mode->family == AF_INET6)
347 return xfrm6_remove_tunnel_encap(x, skb);
348 break;
349 }
350
351 WARN_ON_ONCE(1);
352 return -EOPNOTSUPP;
353}
354
355static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
356{
357 const struct xfrm_mode *inner_mode = &x->inner_mode;
358
359 switch (x->outer_mode.family) {
360 case AF_INET:
361 xfrm4_extract_header(skb);
362 break;
363 case AF_INET6:
364 xfrm6_extract_header(skb);
365 break;
366 default:
367 WARN_ON_ONCE(1);
368 return -EAFNOSUPPORT;
369 }
370
371 if (x->sel.family == AF_UNSPEC) {
372 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
373 if (!inner_mode)
374 return -EAFNOSUPPORT;
375 }
376
377 switch (inner_mode->family) {
378 case AF_INET:
379 skb->protocol = htons(ETH_P_IP);
380 break;
381 case AF_INET6:
382 skb->protocol = htons(ETH_P_IPV6);
383 break;
384 default:
385 WARN_ON_ONCE(1);
386 break;
387 }
388
389 return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
390}
391
392/* Remove encapsulation header.
393 *
394 * The IP header will be moved over the top of the encapsulation header.
395 *
396 * On entry, skb_transport_header() shall point to where the IP header
397 * should be and skb_network_header() shall be set to where the IP header
398 * currently is. skb->data shall point to the start of the payload.
399 */
400static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
401{
402 int ihl = skb->data - skb_transport_header(skb);
403
404 if (skb->transport_header != skb->network_header) {
405 memmove(skb_transport_header(skb),
406 skb_network_header(skb), ihl);
407 skb->network_header = skb->transport_header;
408 }
409 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
410 skb_reset_transport_header(skb);
411 return 0;
412}
413
414static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
415{
416#if IS_ENABLED(CONFIG_IPV6)
417 int ihl = skb->data - skb_transport_header(skb);
418
419 if (skb->transport_header != skb->network_header) {
420 memmove(skb_transport_header(skb),
421 skb_network_header(skb), ihl);
422 skb->network_header = skb->transport_header;
423 }
424 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
425 sizeof(struct ipv6hdr));
426 skb_reset_transport_header(skb);
427 return 0;
428#else
429 WARN_ON_ONCE(1);
430 return -EAFNOSUPPORT;
431#endif
432}
433
434static int xfrm_inner_mode_input(struct xfrm_state *x,
435 const struct xfrm_mode *inner_mode,
436 struct sk_buff *skb)
437{
438 switch (inner_mode->encap) {
439 case XFRM_MODE_BEET:
440 case XFRM_MODE_TUNNEL:
441 return xfrm_prepare_input(x, skb);
442 case XFRM_MODE_TRANSPORT:
443 if (inner_mode->family == AF_INET)
444 return xfrm4_transport_input(x, skb);
445 if (inner_mode->family == AF_INET6)
446 return xfrm6_transport_input(x, skb);
447 break;
448 case XFRM_MODE_ROUTEOPTIMIZATION:
449 WARN_ON_ONCE(1);
450 break;
451 default:
452 WARN_ON_ONCE(1);
453 break;
454 }
455
456 return -EOPNOTSUPP;
457}
458
459int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
460{
461 const struct xfrm_state_afinfo *afinfo;
462 struct net *net = dev_net(skb->dev);
463 const struct xfrm_mode *inner_mode;
464 int err;
465 __be32 seq;
466 __be32 seq_hi;
467 struct xfrm_state *x = NULL;
468 xfrm_address_t *daddr;
469 u32 mark = skb->mark;
470 unsigned int family = AF_UNSPEC;
471 int decaps = 0;
472 int async = 0;
473 bool xfrm_gro = false;
474 bool crypto_done = false;
475 struct xfrm_offload *xo = xfrm_offload(skb);
476 struct sec_path *sp;
477
478 if (encap_type < 0) {
479 x = xfrm_input_state(skb);
480
481 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
482 if (x->km.state == XFRM_STATE_ACQ)
483 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
484 else
485 XFRM_INC_STATS(net,
486 LINUX_MIB_XFRMINSTATEINVALID);
487
488 if (encap_type == -1)
489 dev_put(skb->dev);
490 goto drop;
491 }
492
493 family = x->outer_mode.family;
494
495 /* An encap_type of -1 indicates async resumption. */
496 if (encap_type == -1) {
497 async = 1;
498 seq = XFRM_SKB_CB(skb)->seq.input.low;
499 goto resume;
500 }
501
502 /* encap_type < -1 indicates a GRO call. */
503 encap_type = 0;
504 seq = XFRM_SPI_SKB_CB(skb)->seq;
505
506 if (xo && (xo->flags & CRYPTO_DONE)) {
507 crypto_done = true;
508 family = XFRM_SPI_SKB_CB(skb)->family;
509
510 if (!(xo->status & CRYPTO_SUCCESS)) {
511 if (xo->status &
512 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
513 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
514 CRYPTO_TUNNEL_AH_AUTH_FAILED |
515 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
516
517 xfrm_audit_state_icvfail(x, skb,
518 x->type->proto);
519 x->stats.integrity_failed++;
520 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
521 goto drop;
522 }
523
524 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
525 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
526 goto drop;
527 }
528
529 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
530 goto drop;
531 }
532
533 if ((err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
534 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
535 goto drop;
536 }
537 }
538
539 goto lock;
540 }
541
542 family = XFRM_SPI_SKB_CB(skb)->family;
543
544 /* if tunnel is present override skb->mark value with tunnel i_key */
545 switch (family) {
546 case AF_INET:
547 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
548 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
549 break;
550 case AF_INET6:
551 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
552 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
553 break;
554 }
555
556 sp = secpath_set(skb);
557 if (!sp) {
558 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
559 goto drop;
560 }
561
562 seq = 0;
563 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
564 secpath_reset(skb);
565 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
566 goto drop;
567 }
568
569 daddr = (xfrm_address_t *)(skb_network_header(skb) +
570 XFRM_SPI_SKB_CB(skb)->daddroff);
571 do {
572 sp = skb_sec_path(skb);
573
574 if (sp->len == XFRM_MAX_DEPTH) {
575 secpath_reset(skb);
576 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
577 goto drop;
578 }
579
580 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
581 if (x == NULL) {
582 secpath_reset(skb);
583 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
584 xfrm_audit_state_notfound(skb, family, spi, seq);
585 goto drop;
586 }
587
588 skb->mark = xfrm_smark_get(skb->mark, x);
589
590 sp->xvec[sp->len++] = x;
591
592 skb_dst_force(skb);
593 if (!skb_dst(skb)) {
594 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
595 goto drop;
596 }
597
598lock:
599 spin_lock(&x->lock);
600
601 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
602 if (x->km.state == XFRM_STATE_ACQ)
603 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
604 else
605 XFRM_INC_STATS(net,
606 LINUX_MIB_XFRMINSTATEINVALID);
607 goto drop_unlock;
608 }
609
610 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
611 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
612 goto drop_unlock;
613 }
614
615 if (x->repl->check(x, skb, seq)) {
616 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
617 goto drop_unlock;
618 }
619
620 if (xfrm_state_check_expire(x)) {
621 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
622 goto drop_unlock;
623 }
624
625 spin_unlock(&x->lock);
626
627 if (xfrm_tunnel_check(skb, x, family)) {
628 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
629 goto drop;
630 }
631
632 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
633
634 XFRM_SKB_CB(skb)->seq.input.low = seq;
635 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
636
637 dev_hold(skb->dev);
638
639 if (crypto_done)
640 nexthdr = x->type_offload->input_tail(x, skb);
641 else
642 nexthdr = x->type->input(x, skb);
643
644 if (nexthdr == -EINPROGRESS)
645 return 0;
646resume:
647 dev_put(skb->dev);
648
649 spin_lock(&x->lock);
650 if (nexthdr < 0) {
651 if (nexthdr == -EBADMSG) {
652 xfrm_audit_state_icvfail(x, skb,
653 x->type->proto);
654 x->stats.integrity_failed++;
655 }
656 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
657 goto drop_unlock;
658 }
659
660 /* only the first xfrm gets the encap type */
661 encap_type = 0;
662
663 if (async && x->repl->recheck(x, skb, seq)) {
664 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
665 goto drop_unlock;
666 }
667
668 x->repl->advance(x, seq);
669
670 x->curlft.bytes += skb->len;
671 x->curlft.packets++;
672
673 spin_unlock(&x->lock);
674
675 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
676
677 inner_mode = &x->inner_mode;
678
679 if (x->sel.family == AF_UNSPEC) {
680 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
681 if (inner_mode == NULL) {
682 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
683 goto drop;
684 }
685 }
686
687 if (xfrm_inner_mode_input(x, inner_mode, skb)) {
688 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
689 goto drop;
690 }
691
692 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
693 decaps = 1;
694 break;
695 }
696
697 /*
698 * We need the inner address. However, we only get here for
699 * transport mode so the outer address is identical.
700 */
701 daddr = &x->id.daddr;
702 family = x->outer_mode.family;
703
704 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
705 if (err < 0) {
706 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
707 goto drop;
708 }
709 crypto_done = false;
710 } while (!err);
711
712 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
713 if (err)
714 goto drop;
715
716 nf_reset_ct(skb);
717
718 if (decaps) {
719 sp = skb_sec_path(skb);
720 if (sp)
721 sp->olen = 0;
722 skb_dst_drop(skb);
723 gro_cells_receive(&gro_cells, skb);
724 return 0;
725 } else {
726 xo = xfrm_offload(skb);
727 if (xo)
728 xfrm_gro = xo->flags & XFRM_GRO;
729
730 err = -EAFNOSUPPORT;
731 rcu_read_lock();
732 afinfo = xfrm_state_afinfo_get_rcu(x->inner_mode.family);
733 if (likely(afinfo))
734 err = afinfo->transport_finish(skb, xfrm_gro || async);
735 rcu_read_unlock();
736 if (xfrm_gro) {
737 sp = skb_sec_path(skb);
738 if (sp)
739 sp->olen = 0;
740 skb_dst_drop(skb);
741 gro_cells_receive(&gro_cells, skb);
742 return err;
743 }
744
745 return err;
746 }
747
748drop_unlock:
749 spin_unlock(&x->lock);
750drop:
751 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
752 kfree_skb(skb);
753 return 0;
754}
755EXPORT_SYMBOL(xfrm_input);
756
757int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
758{
759 return xfrm_input(skb, nexthdr, 0, -1);
760}
761EXPORT_SYMBOL(xfrm_input_resume);
762
763static void xfrm_trans_reinject(unsigned long data)
764{
765 struct xfrm_trans_tasklet *trans = (void *)data;
766 struct sk_buff_head queue;
767 struct sk_buff *skb;
768
769 __skb_queue_head_init(&queue);
770 skb_queue_splice_init(&trans->queue, &queue);
771
772 while ((skb = __skb_dequeue(&queue)))
773 XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
774 NULL, skb);
775}
776
777int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
778 int (*finish)(struct net *, struct sock *,
779 struct sk_buff *))
780{
781 struct xfrm_trans_tasklet *trans;
782
783 trans = this_cpu_ptr(&xfrm_trans_tasklet);
784
785 if (skb_queue_len(&trans->queue) >= netdev_max_backlog)
786 return -ENOBUFS;
787
788 BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
789
790 XFRM_TRANS_SKB_CB(skb)->finish = finish;
791 XFRM_TRANS_SKB_CB(skb)->net = net;
792 __skb_queue_tail(&trans->queue, skb);
793 tasklet_schedule(&trans->tasklet);
794 return 0;
795}
796EXPORT_SYMBOL(xfrm_trans_queue_net);
797
798int xfrm_trans_queue(struct sk_buff *skb,
799 int (*finish)(struct net *, struct sock *,
800 struct sk_buff *))
801{
802 return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
803}
804EXPORT_SYMBOL(xfrm_trans_queue);
805
806void __init xfrm_input_init(void)
807{
808 int err;
809 int i;
810
811 init_dummy_netdev(&xfrm_napi_dev);
812 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
813 if (err)
814 gro_cells.cells = NULL;
815
816 for_each_possible_cpu(i) {
817 struct xfrm_trans_tasklet *trans;
818
819 trans = &per_cpu(xfrm_trans_tasklet, i);
820 __skb_queue_head_init(&trans->queue);
821 tasklet_init(&trans->tasklet, xfrm_trans_reinject,
822 (unsigned long)trans);
823 }
824}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm_input.c
4 *
5 * Changes:
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
8 *
9 */
10
11#include <linux/bottom_half.h>
12#include <linux/cache.h>
13#include <linux/interrupt.h>
14#include <linux/slab.h>
15#include <linux/module.h>
16#include <linux/netdevice.h>
17#include <linux/percpu.h>
18#include <net/dst.h>
19#include <net/ip.h>
20#include <net/xfrm.h>
21#include <net/ip_tunnels.h>
22#include <net/ip6_tunnel.h>
23#include <net/dst_metadata.h>
24
25#include "xfrm_inout.h"
26
27struct xfrm_trans_tasklet {
28 struct work_struct work;
29 spinlock_t queue_lock;
30 struct sk_buff_head queue;
31};
32
33struct xfrm_trans_cb {
34 union {
35 struct inet_skb_parm h4;
36#if IS_ENABLED(CONFIG_IPV6)
37 struct inet6_skb_parm h6;
38#endif
39 } header;
40 int (*finish)(struct net *net, struct sock *sk, struct sk_buff *skb);
41 struct net *net;
42};
43
44#define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
45
46static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
47static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
48
49static struct gro_cells gro_cells;
50static struct net_device xfrm_napi_dev;
51
52static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
53
54int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
55{
56 int err = 0;
57
58 if (WARN_ON(afinfo->family > AF_INET6))
59 return -EAFNOSUPPORT;
60
61 spin_lock_bh(&xfrm_input_afinfo_lock);
62 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family]))
63 err = -EEXIST;
64 else
65 rcu_assign_pointer(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], afinfo);
66 spin_unlock_bh(&xfrm_input_afinfo_lock);
67 return err;
68}
69EXPORT_SYMBOL(xfrm_input_register_afinfo);
70
71int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
72{
73 int err = 0;
74
75 spin_lock_bh(&xfrm_input_afinfo_lock);
76 if (likely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family])) {
77 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family] != afinfo))
78 err = -EINVAL;
79 else
80 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], NULL);
81 }
82 spin_unlock_bh(&xfrm_input_afinfo_lock);
83 synchronize_rcu();
84 return err;
85}
86EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
87
88static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(u8 family, bool is_ipip)
89{
90 const struct xfrm_input_afinfo *afinfo;
91
92 if (WARN_ON_ONCE(family > AF_INET6))
93 return NULL;
94
95 rcu_read_lock();
96 afinfo = rcu_dereference(xfrm_input_afinfo[is_ipip][family]);
97 if (unlikely(!afinfo))
98 rcu_read_unlock();
99 return afinfo;
100}
101
102static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
103 int err)
104{
105 bool is_ipip = (protocol == IPPROTO_IPIP || protocol == IPPROTO_IPV6);
106 const struct xfrm_input_afinfo *afinfo;
107 int ret;
108
109 afinfo = xfrm_input_get_afinfo(family, is_ipip);
110 if (!afinfo)
111 return -EAFNOSUPPORT;
112
113 ret = afinfo->callback(skb, protocol, err);
114 rcu_read_unlock();
115
116 return ret;
117}
118
119struct sec_path *secpath_set(struct sk_buff *skb)
120{
121 struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
122
123 sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
124 if (!sp)
125 return NULL;
126
127 if (tmp) /* reused existing one (was COW'd if needed) */
128 return sp;
129
130 /* allocated new secpath */
131 memset(sp->ovec, 0, sizeof(sp->ovec));
132 sp->olen = 0;
133 sp->len = 0;
134 sp->verified_cnt = 0;
135
136 return sp;
137}
138EXPORT_SYMBOL(secpath_set);
139
140/* Fetch spi and seq from ipsec header */
141
142int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
143{
144 int offset, offset_seq;
145 int hlen;
146
147 switch (nexthdr) {
148 case IPPROTO_AH:
149 hlen = sizeof(struct ip_auth_hdr);
150 offset = offsetof(struct ip_auth_hdr, spi);
151 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
152 break;
153 case IPPROTO_ESP:
154 hlen = sizeof(struct ip_esp_hdr);
155 offset = offsetof(struct ip_esp_hdr, spi);
156 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
157 break;
158 case IPPROTO_COMP:
159 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
160 return -EINVAL;
161 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
162 *seq = 0;
163 return 0;
164 default:
165 return 1;
166 }
167
168 if (!pskb_may_pull(skb, hlen))
169 return -EINVAL;
170
171 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
172 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
173 return 0;
174}
175EXPORT_SYMBOL(xfrm_parse_spi);
176
177static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
178{
179 struct iphdr *iph;
180 int optlen = 0;
181 int err = -EINVAL;
182
183 skb->protocol = htons(ETH_P_IP);
184
185 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
186 struct ip_beet_phdr *ph;
187 int phlen;
188
189 if (!pskb_may_pull(skb, sizeof(*ph)))
190 goto out;
191
192 ph = (struct ip_beet_phdr *)skb->data;
193
194 phlen = sizeof(*ph) + ph->padlen;
195 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
196 if (optlen < 0 || optlen & 3 || optlen > 250)
197 goto out;
198
199 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
200
201 if (!pskb_may_pull(skb, phlen))
202 goto out;
203 __skb_pull(skb, phlen);
204 }
205
206 skb_push(skb, sizeof(*iph));
207 skb_reset_network_header(skb);
208 skb_mac_header_rebuild(skb);
209
210 xfrm4_beet_make_header(skb);
211
212 iph = ip_hdr(skb);
213
214 iph->ihl += optlen / 4;
215 iph->tot_len = htons(skb->len);
216 iph->daddr = x->sel.daddr.a4;
217 iph->saddr = x->sel.saddr.a4;
218 iph->check = 0;
219 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
220 err = 0;
221out:
222 return err;
223}
224
225static void ipip_ecn_decapsulate(struct sk_buff *skb)
226{
227 struct iphdr *inner_iph = ipip_hdr(skb);
228
229 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
230 IP_ECN_set_ce(inner_iph);
231}
232
233static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
234{
235 int err = -EINVAL;
236
237 skb->protocol = htons(ETH_P_IP);
238
239 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
240 goto out;
241
242 err = skb_unclone(skb, GFP_ATOMIC);
243 if (err)
244 goto out;
245
246 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
247 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
248 if (!(x->props.flags & XFRM_STATE_NOECN))
249 ipip_ecn_decapsulate(skb);
250
251 skb_reset_network_header(skb);
252 skb_mac_header_rebuild(skb);
253 if (skb->mac_len)
254 eth_hdr(skb)->h_proto = skb->protocol;
255
256 err = 0;
257
258out:
259 return err;
260}
261
262static void ipip6_ecn_decapsulate(struct sk_buff *skb)
263{
264 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
265
266 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
267 IP6_ECN_set_ce(skb, inner_iph);
268}
269
270static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
271{
272 int err = -EINVAL;
273
274 skb->protocol = htons(ETH_P_IPV6);
275
276 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
277 goto out;
278
279 err = skb_unclone(skb, GFP_ATOMIC);
280 if (err)
281 goto out;
282
283 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
284 ipv6_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipipv6_hdr(skb));
285 if (!(x->props.flags & XFRM_STATE_NOECN))
286 ipip6_ecn_decapsulate(skb);
287
288 skb_reset_network_header(skb);
289 skb_mac_header_rebuild(skb);
290 if (skb->mac_len)
291 eth_hdr(skb)->h_proto = skb->protocol;
292
293 err = 0;
294
295out:
296 return err;
297}
298
299static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
300{
301 struct ipv6hdr *ip6h;
302 int size = sizeof(struct ipv6hdr);
303 int err;
304
305 skb->protocol = htons(ETH_P_IPV6);
306
307 err = skb_cow_head(skb, size + skb->mac_len);
308 if (err)
309 goto out;
310
311 __skb_push(skb, size);
312 skb_reset_network_header(skb);
313 skb_mac_header_rebuild(skb);
314
315 xfrm6_beet_make_header(skb);
316
317 ip6h = ipv6_hdr(skb);
318 ip6h->payload_len = htons(skb->len - size);
319 ip6h->daddr = x->sel.daddr.in6;
320 ip6h->saddr = x->sel.saddr.in6;
321 err = 0;
322out:
323 return err;
324}
325
326/* Remove encapsulation header.
327 *
328 * The IP header will be moved over the top of the encapsulation
329 * header.
330 *
331 * On entry, the transport header shall point to where the IP header
332 * should be and the network header shall be set to where the IP
333 * header currently is. skb->data shall point to the start of the
334 * payload.
335 */
336static int
337xfrm_inner_mode_encap_remove(struct xfrm_state *x,
338 struct sk_buff *skb)
339{
340 switch (x->props.mode) {
341 case XFRM_MODE_BEET:
342 switch (x->sel.family) {
343 case AF_INET:
344 return xfrm4_remove_beet_encap(x, skb);
345 case AF_INET6:
346 return xfrm6_remove_beet_encap(x, skb);
347 }
348 break;
349 case XFRM_MODE_TUNNEL:
350 switch (XFRM_MODE_SKB_CB(skb)->protocol) {
351 case IPPROTO_IPIP:
352 return xfrm4_remove_tunnel_encap(x, skb);
353 case IPPROTO_IPV6:
354 return xfrm6_remove_tunnel_encap(x, skb);
355 break;
356 }
357 return -EINVAL;
358 }
359
360 WARN_ON_ONCE(1);
361 return -EOPNOTSUPP;
362}
363
364static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
365{
366 switch (x->props.family) {
367 case AF_INET:
368 xfrm4_extract_header(skb);
369 break;
370 case AF_INET6:
371 xfrm6_extract_header(skb);
372 break;
373 default:
374 WARN_ON_ONCE(1);
375 return -EAFNOSUPPORT;
376 }
377
378 return xfrm_inner_mode_encap_remove(x, skb);
379}
380
381/* Remove encapsulation header.
382 *
383 * The IP header will be moved over the top of the encapsulation header.
384 *
385 * On entry, skb_transport_header() shall point to where the IP header
386 * should be and skb_network_header() shall be set to where the IP header
387 * currently is. skb->data shall point to the start of the payload.
388 */
389static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
390{
391 int ihl = skb->data - skb_transport_header(skb);
392
393 if (skb->transport_header != skb->network_header) {
394 memmove(skb_transport_header(skb),
395 skb_network_header(skb), ihl);
396 skb->network_header = skb->transport_header;
397 }
398 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
399 skb_reset_transport_header(skb);
400 return 0;
401}
402
403static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
404{
405#if IS_ENABLED(CONFIG_IPV6)
406 int ihl = skb->data - skb_transport_header(skb);
407
408 if (skb->transport_header != skb->network_header) {
409 memmove(skb_transport_header(skb),
410 skb_network_header(skb), ihl);
411 skb->network_header = skb->transport_header;
412 }
413 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
414 sizeof(struct ipv6hdr));
415 skb_reset_transport_header(skb);
416 return 0;
417#else
418 WARN_ON_ONCE(1);
419 return -EAFNOSUPPORT;
420#endif
421}
422
423static int xfrm_inner_mode_input(struct xfrm_state *x,
424 struct sk_buff *skb)
425{
426 switch (x->props.mode) {
427 case XFRM_MODE_BEET:
428 case XFRM_MODE_TUNNEL:
429 return xfrm_prepare_input(x, skb);
430 case XFRM_MODE_TRANSPORT:
431 if (x->props.family == AF_INET)
432 return xfrm4_transport_input(x, skb);
433 if (x->props.family == AF_INET6)
434 return xfrm6_transport_input(x, skb);
435 break;
436 case XFRM_MODE_ROUTEOPTIMIZATION:
437 WARN_ON_ONCE(1);
438 break;
439 default:
440 WARN_ON_ONCE(1);
441 break;
442 }
443
444 return -EOPNOTSUPP;
445}
446
447int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
448{
449 const struct xfrm_state_afinfo *afinfo;
450 struct net *net = dev_net(skb->dev);
451 int err;
452 __be32 seq;
453 __be32 seq_hi;
454 struct xfrm_state *x = NULL;
455 xfrm_address_t *daddr;
456 u32 mark = skb->mark;
457 unsigned int family = AF_UNSPEC;
458 int decaps = 0;
459 int async = 0;
460 bool xfrm_gro = false;
461 bool crypto_done = false;
462 struct xfrm_offload *xo = xfrm_offload(skb);
463 struct sec_path *sp;
464
465 if (encap_type < 0 || (xo && xo->flags & XFRM_GRO)) {
466 x = xfrm_input_state(skb);
467
468 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
469 if (x->km.state == XFRM_STATE_ACQ)
470 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
471 else
472 XFRM_INC_STATS(net,
473 LINUX_MIB_XFRMINSTATEINVALID);
474
475 if (encap_type == -1)
476 dev_put(skb->dev);
477 goto drop;
478 }
479
480 family = x->props.family;
481
482 /* An encap_type of -1 indicates async resumption. */
483 if (encap_type == -1) {
484 async = 1;
485 seq = XFRM_SKB_CB(skb)->seq.input.low;
486 goto resume;
487 }
488 /* GRO call */
489 seq = XFRM_SPI_SKB_CB(skb)->seq;
490
491 if (xo && (xo->flags & CRYPTO_DONE)) {
492 crypto_done = true;
493 family = XFRM_SPI_SKB_CB(skb)->family;
494
495 if (!(xo->status & CRYPTO_SUCCESS)) {
496 if (xo->status &
497 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
498 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
499 CRYPTO_TUNNEL_AH_AUTH_FAILED |
500 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
501
502 xfrm_audit_state_icvfail(x, skb,
503 x->type->proto);
504 x->stats.integrity_failed++;
505 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
506 goto drop;
507 }
508
509 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
510 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
511 goto drop;
512 }
513
514 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
515 goto drop;
516 }
517
518 if (xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
519 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
520 goto drop;
521 }
522 }
523
524 goto lock;
525 }
526
527 family = XFRM_SPI_SKB_CB(skb)->family;
528
529 /* if tunnel is present override skb->mark value with tunnel i_key */
530 switch (family) {
531 case AF_INET:
532 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
533 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
534 break;
535 case AF_INET6:
536 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
537 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
538 break;
539 }
540
541 sp = secpath_set(skb);
542 if (!sp) {
543 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
544 goto drop;
545 }
546
547 seq = 0;
548 if (!spi && xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
549 secpath_reset(skb);
550 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
551 goto drop;
552 }
553
554 daddr = (xfrm_address_t *)(skb_network_header(skb) +
555 XFRM_SPI_SKB_CB(skb)->daddroff);
556 do {
557 sp = skb_sec_path(skb);
558
559 if (sp->len == XFRM_MAX_DEPTH) {
560 secpath_reset(skb);
561 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
562 goto drop;
563 }
564
565 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
566 if (x == NULL) {
567 secpath_reset(skb);
568 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
569 xfrm_audit_state_notfound(skb, family, spi, seq);
570 goto drop;
571 }
572
573 skb->mark = xfrm_smark_get(skb->mark, x);
574
575 sp->xvec[sp->len++] = x;
576
577 skb_dst_force(skb);
578 if (!skb_dst(skb)) {
579 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
580 goto drop;
581 }
582
583lock:
584 spin_lock(&x->lock);
585
586 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
587 if (x->km.state == XFRM_STATE_ACQ)
588 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
589 else
590 XFRM_INC_STATS(net,
591 LINUX_MIB_XFRMINSTATEINVALID);
592 goto drop_unlock;
593 }
594
595 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
596 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
597 goto drop_unlock;
598 }
599
600 if (xfrm_replay_check(x, skb, seq)) {
601 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
602 goto drop_unlock;
603 }
604
605 if (xfrm_state_check_expire(x)) {
606 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
607 goto drop_unlock;
608 }
609
610 spin_unlock(&x->lock);
611
612 if (xfrm_tunnel_check(skb, x, family)) {
613 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
614 goto drop;
615 }
616
617 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
618
619 XFRM_SKB_CB(skb)->seq.input.low = seq;
620 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
621
622 dev_hold(skb->dev);
623
624 if (crypto_done)
625 nexthdr = x->type_offload->input_tail(x, skb);
626 else
627 nexthdr = x->type->input(x, skb);
628
629 if (nexthdr == -EINPROGRESS)
630 return 0;
631resume:
632 dev_put(skb->dev);
633
634 spin_lock(&x->lock);
635 if (nexthdr < 0) {
636 if (nexthdr == -EBADMSG) {
637 xfrm_audit_state_icvfail(x, skb,
638 x->type->proto);
639 x->stats.integrity_failed++;
640 }
641 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
642 goto drop_unlock;
643 }
644
645 /* only the first xfrm gets the encap type */
646 encap_type = 0;
647
648 if (xfrm_replay_recheck(x, skb, seq)) {
649 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
650 goto drop_unlock;
651 }
652
653 xfrm_replay_advance(x, seq);
654
655 x->curlft.bytes += skb->len;
656 x->curlft.packets++;
657 x->lastused = ktime_get_real_seconds();
658
659 spin_unlock(&x->lock);
660
661 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
662
663 if (xfrm_inner_mode_input(x, skb)) {
664 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
665 goto drop;
666 }
667
668 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
669 decaps = 1;
670 break;
671 }
672
673 /*
674 * We need the inner address. However, we only get here for
675 * transport mode so the outer address is identical.
676 */
677 daddr = &x->id.daddr;
678 family = x->props.family;
679
680 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
681 if (err < 0) {
682 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
683 goto drop;
684 }
685 crypto_done = false;
686 } while (!err);
687
688 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
689 if (err)
690 goto drop;
691
692 nf_reset_ct(skb);
693
694 if (decaps) {
695 sp = skb_sec_path(skb);
696 if (sp)
697 sp->olen = 0;
698 if (skb_valid_dst(skb))
699 skb_dst_drop(skb);
700 gro_cells_receive(&gro_cells, skb);
701 return 0;
702 } else {
703 xo = xfrm_offload(skb);
704 if (xo)
705 xfrm_gro = xo->flags & XFRM_GRO;
706
707 err = -EAFNOSUPPORT;
708 rcu_read_lock();
709 afinfo = xfrm_state_afinfo_get_rcu(x->props.family);
710 if (likely(afinfo))
711 err = afinfo->transport_finish(skb, xfrm_gro || async);
712 rcu_read_unlock();
713 if (xfrm_gro) {
714 sp = skb_sec_path(skb);
715 if (sp)
716 sp->olen = 0;
717 if (skb_valid_dst(skb))
718 skb_dst_drop(skb);
719 gro_cells_receive(&gro_cells, skb);
720 return err;
721 }
722
723 return err;
724 }
725
726drop_unlock:
727 spin_unlock(&x->lock);
728drop:
729 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
730 kfree_skb(skb);
731 return 0;
732}
733EXPORT_SYMBOL(xfrm_input);
734
735int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
736{
737 return xfrm_input(skb, nexthdr, 0, -1);
738}
739EXPORT_SYMBOL(xfrm_input_resume);
740
741static void xfrm_trans_reinject(struct work_struct *work)
742{
743 struct xfrm_trans_tasklet *trans = container_of(work, struct xfrm_trans_tasklet, work);
744 struct sk_buff_head queue;
745 struct sk_buff *skb;
746
747 __skb_queue_head_init(&queue);
748 spin_lock_bh(&trans->queue_lock);
749 skb_queue_splice_init(&trans->queue, &queue);
750 spin_unlock_bh(&trans->queue_lock);
751
752 local_bh_disable();
753 while ((skb = __skb_dequeue(&queue)))
754 XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
755 NULL, skb);
756 local_bh_enable();
757}
758
759int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
760 int (*finish)(struct net *, struct sock *,
761 struct sk_buff *))
762{
763 struct xfrm_trans_tasklet *trans;
764
765 trans = this_cpu_ptr(&xfrm_trans_tasklet);
766
767 if (skb_queue_len(&trans->queue) >= READ_ONCE(netdev_max_backlog))
768 return -ENOBUFS;
769
770 BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
771
772 XFRM_TRANS_SKB_CB(skb)->finish = finish;
773 XFRM_TRANS_SKB_CB(skb)->net = net;
774 spin_lock_bh(&trans->queue_lock);
775 __skb_queue_tail(&trans->queue, skb);
776 spin_unlock_bh(&trans->queue_lock);
777 schedule_work(&trans->work);
778 return 0;
779}
780EXPORT_SYMBOL(xfrm_trans_queue_net);
781
782int xfrm_trans_queue(struct sk_buff *skb,
783 int (*finish)(struct net *, struct sock *,
784 struct sk_buff *))
785{
786 return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
787}
788EXPORT_SYMBOL(xfrm_trans_queue);
789
790void __init xfrm_input_init(void)
791{
792 int err;
793 int i;
794
795 init_dummy_netdev(&xfrm_napi_dev);
796 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
797 if (err)
798 gro_cells.cells = NULL;
799
800 for_each_possible_cpu(i) {
801 struct xfrm_trans_tasklet *trans;
802
803 trans = &per_cpu(xfrm_trans_tasklet, i);
804 spin_lock_init(&trans->queue_lock);
805 __skb_queue_head_init(&trans->queue);
806 INIT_WORK(&trans->work, xfrm_trans_reinject);
807 }
808}