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};
40
41#define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
42
43static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
44static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[AF_INET6 + 1];
45
46static struct gro_cells gro_cells;
47static struct net_device xfrm_napi_dev;
48
49static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
50
51int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
52{
53 int err = 0;
54
55 if (WARN_ON(afinfo->family >= ARRAY_SIZE(xfrm_input_afinfo)))
56 return -EAFNOSUPPORT;
57
58 spin_lock_bh(&xfrm_input_afinfo_lock);
59 if (unlikely(xfrm_input_afinfo[afinfo->family] != NULL))
60 err = -EEXIST;
61 else
62 rcu_assign_pointer(xfrm_input_afinfo[afinfo->family], afinfo);
63 spin_unlock_bh(&xfrm_input_afinfo_lock);
64 return err;
65}
66EXPORT_SYMBOL(xfrm_input_register_afinfo);
67
68int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
69{
70 int err = 0;
71
72 spin_lock_bh(&xfrm_input_afinfo_lock);
73 if (likely(xfrm_input_afinfo[afinfo->family] != NULL)) {
74 if (unlikely(xfrm_input_afinfo[afinfo->family] != afinfo))
75 err = -EINVAL;
76 else
77 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->family], NULL);
78 }
79 spin_unlock_bh(&xfrm_input_afinfo_lock);
80 synchronize_rcu();
81 return err;
82}
83EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
84
85static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(unsigned int family)
86{
87 const struct xfrm_input_afinfo *afinfo;
88
89 if (WARN_ON_ONCE(family >= ARRAY_SIZE(xfrm_input_afinfo)))
90 return NULL;
91
92 rcu_read_lock();
93 afinfo = rcu_dereference(xfrm_input_afinfo[family]);
94 if (unlikely(!afinfo))
95 rcu_read_unlock();
96 return afinfo;
97}
98
99static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
100 int err)
101{
102 int ret;
103 const struct xfrm_input_afinfo *afinfo = xfrm_input_get_afinfo(family);
104
105 if (!afinfo)
106 return -EAFNOSUPPORT;
107
108 ret = afinfo->callback(skb, protocol, err);
109 rcu_read_unlock();
110
111 return ret;
112}
113
114struct sec_path *secpath_set(struct sk_buff *skb)
115{
116 struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
117
118 sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
119 if (!sp)
120 return NULL;
121
122 if (tmp) /* reused existing one (was COW'd if needed) */
123 return sp;
124
125 /* allocated new secpath */
126 memset(sp->ovec, 0, sizeof(sp->ovec));
127 sp->olen = 0;
128 sp->len = 0;
129
130 return sp;
131}
132EXPORT_SYMBOL(secpath_set);
133
134/* Fetch spi and seq from ipsec header */
135
136int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
137{
138 int offset, offset_seq;
139 int hlen;
140
141 switch (nexthdr) {
142 case IPPROTO_AH:
143 hlen = sizeof(struct ip_auth_hdr);
144 offset = offsetof(struct ip_auth_hdr, spi);
145 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
146 break;
147 case IPPROTO_ESP:
148 hlen = sizeof(struct ip_esp_hdr);
149 offset = offsetof(struct ip_esp_hdr, spi);
150 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
151 break;
152 case IPPROTO_COMP:
153 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
154 return -EINVAL;
155 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
156 *seq = 0;
157 return 0;
158 default:
159 return 1;
160 }
161
162 if (!pskb_may_pull(skb, hlen))
163 return -EINVAL;
164
165 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
166 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
167 return 0;
168}
169EXPORT_SYMBOL(xfrm_parse_spi);
170
171static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
172{
173 struct iphdr *iph;
174 int optlen = 0;
175 int err = -EINVAL;
176
177 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
178 struct ip_beet_phdr *ph;
179 int phlen;
180
181 if (!pskb_may_pull(skb, sizeof(*ph)))
182 goto out;
183
184 ph = (struct ip_beet_phdr *)skb->data;
185
186 phlen = sizeof(*ph) + ph->padlen;
187 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
188 if (optlen < 0 || optlen & 3 || optlen > 250)
189 goto out;
190
191 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
192
193 if (!pskb_may_pull(skb, phlen))
194 goto out;
195 __skb_pull(skb, phlen);
196 }
197
198 skb_push(skb, sizeof(*iph));
199 skb_reset_network_header(skb);
200 skb_mac_header_rebuild(skb);
201
202 xfrm4_beet_make_header(skb);
203
204 iph = ip_hdr(skb);
205
206 iph->ihl += optlen / 4;
207 iph->tot_len = htons(skb->len);
208 iph->daddr = x->sel.daddr.a4;
209 iph->saddr = x->sel.saddr.a4;
210 iph->check = 0;
211 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
212 err = 0;
213out:
214 return err;
215}
216
217static void ipip_ecn_decapsulate(struct sk_buff *skb)
218{
219 struct iphdr *inner_iph = ipip_hdr(skb);
220
221 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
222 IP_ECN_set_ce(inner_iph);
223}
224
225static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
226{
227 int err = -EINVAL;
228
229 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPIP)
230 goto out;
231
232 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
233 goto out;
234
235 err = skb_unclone(skb, GFP_ATOMIC);
236 if (err)
237 goto out;
238
239 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
240 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
241 if (!(x->props.flags & XFRM_STATE_NOECN))
242 ipip_ecn_decapsulate(skb);
243
244 skb_reset_network_header(skb);
245 skb_mac_header_rebuild(skb);
246 if (skb->mac_len)
247 eth_hdr(skb)->h_proto = skb->protocol;
248
249 err = 0;
250
251out:
252 return err;
253}
254
255static void ipip6_ecn_decapsulate(struct sk_buff *skb)
256{
257 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
258
259 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
260 IP6_ECN_set_ce(skb, inner_iph);
261}
262
263static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
264{
265 int err = -EINVAL;
266
267 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPV6)
268 goto out;
269 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
270 goto out;
271
272 err = skb_unclone(skb, GFP_ATOMIC);
273 if (err)
274 goto out;
275
276 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
277 ipv6_copy_dscp(ipv6_get_dsfield(ipv6_hdr(skb)),
278 ipipv6_hdr(skb));
279 if (!(x->props.flags & XFRM_STATE_NOECN))
280 ipip6_ecn_decapsulate(skb);
281
282 skb_reset_network_header(skb);
283 skb_mac_header_rebuild(skb);
284 if (skb->mac_len)
285 eth_hdr(skb)->h_proto = skb->protocol;
286
287 err = 0;
288
289out:
290 return err;
291}
292
293static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
294{
295 struct ipv6hdr *ip6h;
296 int size = sizeof(struct ipv6hdr);
297 int err;
298
299 err = skb_cow_head(skb, size + skb->mac_len);
300 if (err)
301 goto out;
302
303 __skb_push(skb, size);
304 skb_reset_network_header(skb);
305 skb_mac_header_rebuild(skb);
306
307 xfrm6_beet_make_header(skb);
308
309 ip6h = ipv6_hdr(skb);
310 ip6h->payload_len = htons(skb->len - size);
311 ip6h->daddr = x->sel.daddr.in6;
312 ip6h->saddr = x->sel.saddr.in6;
313 err = 0;
314out:
315 return err;
316}
317
318/* Remove encapsulation header.
319 *
320 * The IP header will be moved over the top of the encapsulation
321 * header.
322 *
323 * On entry, the transport header shall point to where the IP header
324 * should be and the network header shall be set to where the IP
325 * header currently is. skb->data shall point to the start of the
326 * payload.
327 */
328static int
329xfrm_inner_mode_encap_remove(struct xfrm_state *x,
330 const struct xfrm_mode *inner_mode,
331 struct sk_buff *skb)
332{
333 switch (inner_mode->encap) {
334 case XFRM_MODE_BEET:
335 if (inner_mode->family == AF_INET)
336 return xfrm4_remove_beet_encap(x, skb);
337 if (inner_mode->family == AF_INET6)
338 return xfrm6_remove_beet_encap(x, skb);
339 break;
340 case XFRM_MODE_TUNNEL:
341 if (inner_mode->family == AF_INET)
342 return xfrm4_remove_tunnel_encap(x, skb);
343 if (inner_mode->family == AF_INET6)
344 return xfrm6_remove_tunnel_encap(x, skb);
345 break;
346 }
347
348 WARN_ON_ONCE(1);
349 return -EOPNOTSUPP;
350}
351
352static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
353{
354 const struct xfrm_mode *inner_mode = &x->inner_mode;
355 const struct xfrm_state_afinfo *afinfo;
356 int err = -EAFNOSUPPORT;
357
358 rcu_read_lock();
359 afinfo = xfrm_state_afinfo_get_rcu(x->outer_mode.family);
360 if (likely(afinfo))
361 err = afinfo->extract_input(x, skb);
362 rcu_read_unlock();
363
364 if (err)
365 return err;
366
367 if (x->sel.family == AF_UNSPEC) {
368 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
369 if (!inner_mode)
370 return -EAFNOSUPPORT;
371 }
372
373 switch (inner_mode->family) {
374 case AF_INET:
375 skb->protocol = htons(ETH_P_IP);
376 break;
377 case AF_INET6:
378 skb->protocol = htons(ETH_P_IPV6);
379 break;
380 default:
381 WARN_ON_ONCE(1);
382 break;
383 }
384
385 return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
386}
387
388/* Remove encapsulation header.
389 *
390 * The IP header will be moved over the top of the encapsulation header.
391 *
392 * On entry, skb_transport_header() shall point to where the IP header
393 * should be and skb_network_header() shall be set to where the IP header
394 * currently is. skb->data shall point to the start of the payload.
395 */
396static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
397{
398 int ihl = skb->data - skb_transport_header(skb);
399
400 if (skb->transport_header != skb->network_header) {
401 memmove(skb_transport_header(skb),
402 skb_network_header(skb), ihl);
403 skb->network_header = skb->transport_header;
404 }
405 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
406 skb_reset_transport_header(skb);
407 return 0;
408}
409
410static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
411{
412#if IS_ENABLED(CONFIG_IPV6)
413 int ihl = skb->data - skb_transport_header(skb);
414
415 if (skb->transport_header != skb->network_header) {
416 memmove(skb_transport_header(skb),
417 skb_network_header(skb), ihl);
418 skb->network_header = skb->transport_header;
419 }
420 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
421 sizeof(struct ipv6hdr));
422 skb_reset_transport_header(skb);
423 return 0;
424#else
425 WARN_ON_ONCE(1);
426 return -EAFNOSUPPORT;
427#endif
428}
429
430static int xfrm_inner_mode_input(struct xfrm_state *x,
431 const struct xfrm_mode *inner_mode,
432 struct sk_buff *skb)
433{
434 switch (inner_mode->encap) {
435 case XFRM_MODE_BEET:
436 case XFRM_MODE_TUNNEL:
437 return xfrm_prepare_input(x, skb);
438 case XFRM_MODE_TRANSPORT:
439 if (inner_mode->family == AF_INET)
440 return xfrm4_transport_input(x, skb);
441 if (inner_mode->family == AF_INET6)
442 return xfrm6_transport_input(x, skb);
443 break;
444 case XFRM_MODE_ROUTEOPTIMIZATION:
445 WARN_ON_ONCE(1);
446 break;
447 default:
448 WARN_ON_ONCE(1);
449 break;
450 }
451
452 return -EOPNOTSUPP;
453}
454
455int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
456{
457 const struct xfrm_state_afinfo *afinfo;
458 struct net *net = dev_net(skb->dev);
459 const struct xfrm_mode *inner_mode;
460 int err;
461 __be32 seq;
462 __be32 seq_hi;
463 struct xfrm_state *x = NULL;
464 xfrm_address_t *daddr;
465 u32 mark = skb->mark;
466 unsigned int family = AF_UNSPEC;
467 int decaps = 0;
468 int async = 0;
469 bool xfrm_gro = false;
470 bool crypto_done = false;
471 struct xfrm_offload *xo = xfrm_offload(skb);
472 struct sec_path *sp;
473
474 if (encap_type < 0) {
475 x = xfrm_input_state(skb);
476
477 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
478 if (x->km.state == XFRM_STATE_ACQ)
479 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
480 else
481 XFRM_INC_STATS(net,
482 LINUX_MIB_XFRMINSTATEINVALID);
483
484 if (encap_type == -1)
485 dev_put(skb->dev);
486 goto drop;
487 }
488
489 family = x->outer_mode.family;
490
491 /* An encap_type of -1 indicates async resumption. */
492 if (encap_type == -1) {
493 async = 1;
494 seq = XFRM_SKB_CB(skb)->seq.input.low;
495 goto resume;
496 }
497
498 /* encap_type < -1 indicates a GRO call. */
499 encap_type = 0;
500 seq = XFRM_SPI_SKB_CB(skb)->seq;
501
502 if (xo && (xo->flags & CRYPTO_DONE)) {
503 crypto_done = true;
504 family = XFRM_SPI_SKB_CB(skb)->family;
505
506 if (!(xo->status & CRYPTO_SUCCESS)) {
507 if (xo->status &
508 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
509 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
510 CRYPTO_TUNNEL_AH_AUTH_FAILED |
511 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
512
513 xfrm_audit_state_icvfail(x, skb,
514 x->type->proto);
515 x->stats.integrity_failed++;
516 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
517 goto drop;
518 }
519
520 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
521 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
522 goto drop;
523 }
524
525 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
526 goto drop;
527 }
528
529 if ((err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
530 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
531 goto drop;
532 }
533 }
534
535 goto lock;
536 }
537
538 family = XFRM_SPI_SKB_CB(skb)->family;
539
540 /* if tunnel is present override skb->mark value with tunnel i_key */
541 switch (family) {
542 case AF_INET:
543 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
544 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
545 break;
546 case AF_INET6:
547 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
548 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
549 break;
550 }
551
552 sp = secpath_set(skb);
553 if (!sp) {
554 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
555 goto drop;
556 }
557
558 seq = 0;
559 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
560 secpath_reset(skb);
561 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
562 goto drop;
563 }
564
565 daddr = (xfrm_address_t *)(skb_network_header(skb) +
566 XFRM_SPI_SKB_CB(skb)->daddroff);
567 do {
568 sp = skb_sec_path(skb);
569
570 if (sp->len == XFRM_MAX_DEPTH) {
571 secpath_reset(skb);
572 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
573 goto drop;
574 }
575
576 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
577 if (x == NULL) {
578 secpath_reset(skb);
579 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
580 xfrm_audit_state_notfound(skb, family, spi, seq);
581 goto drop;
582 }
583
584 skb->mark = xfrm_smark_get(skb->mark, x);
585
586 sp->xvec[sp->len++] = x;
587
588 skb_dst_force(skb);
589 if (!skb_dst(skb)) {
590 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
591 goto drop;
592 }
593
594lock:
595 spin_lock(&x->lock);
596
597 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
598 if (x->km.state == XFRM_STATE_ACQ)
599 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
600 else
601 XFRM_INC_STATS(net,
602 LINUX_MIB_XFRMINSTATEINVALID);
603 goto drop_unlock;
604 }
605
606 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
607 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
608 goto drop_unlock;
609 }
610
611 if (x->repl->check(x, skb, seq)) {
612 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
613 goto drop_unlock;
614 }
615
616 if (xfrm_state_check_expire(x)) {
617 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
618 goto drop_unlock;
619 }
620
621 spin_unlock(&x->lock);
622
623 if (xfrm_tunnel_check(skb, x, family)) {
624 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
625 goto drop;
626 }
627
628 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
629
630 XFRM_SKB_CB(skb)->seq.input.low = seq;
631 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
632
633 dev_hold(skb->dev);
634
635 if (crypto_done)
636 nexthdr = x->type_offload->input_tail(x, skb);
637 else
638 nexthdr = x->type->input(x, skb);
639
640 if (nexthdr == -EINPROGRESS)
641 return 0;
642resume:
643 dev_put(skb->dev);
644
645 spin_lock(&x->lock);
646 if (nexthdr <= 0) {
647 if (nexthdr == -EBADMSG) {
648 xfrm_audit_state_icvfail(x, skb,
649 x->type->proto);
650 x->stats.integrity_failed++;
651 }
652 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
653 goto drop_unlock;
654 }
655
656 /* only the first xfrm gets the encap type */
657 encap_type = 0;
658
659 if (async && x->repl->recheck(x, skb, seq)) {
660 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
661 goto drop_unlock;
662 }
663
664 x->repl->advance(x, seq);
665
666 x->curlft.bytes += skb->len;
667 x->curlft.packets++;
668
669 spin_unlock(&x->lock);
670
671 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
672
673 inner_mode = &x->inner_mode;
674
675 if (x->sel.family == AF_UNSPEC) {
676 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
677 if (inner_mode == NULL) {
678 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
679 goto drop;
680 }
681 }
682
683 if (xfrm_inner_mode_input(x, inner_mode, skb)) {
684 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
685 goto drop;
686 }
687
688 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
689 decaps = 1;
690 break;
691 }
692
693 /*
694 * We need the inner address. However, we only get here for
695 * transport mode so the outer address is identical.
696 */
697 daddr = &x->id.daddr;
698 family = x->outer_mode.family;
699
700 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
701 if (err < 0) {
702 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
703 goto drop;
704 }
705 crypto_done = false;
706 } while (!err);
707
708 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
709 if (err)
710 goto drop;
711
712 nf_reset_ct(skb);
713
714 if (decaps) {
715 sp = skb_sec_path(skb);
716 if (sp)
717 sp->olen = 0;
718 skb_dst_drop(skb);
719 gro_cells_receive(&gro_cells, skb);
720 return 0;
721 } else {
722 xo = xfrm_offload(skb);
723 if (xo)
724 xfrm_gro = xo->flags & XFRM_GRO;
725
726 err = -EAFNOSUPPORT;
727 rcu_read_lock();
728 afinfo = xfrm_state_afinfo_get_rcu(x->inner_mode.family);
729 if (likely(afinfo))
730 err = afinfo->transport_finish(skb, xfrm_gro || async);
731 rcu_read_unlock();
732 if (xfrm_gro) {
733 sp = skb_sec_path(skb);
734 if (sp)
735 sp->olen = 0;
736 skb_dst_drop(skb);
737 gro_cells_receive(&gro_cells, skb);
738 return err;
739 }
740
741 return err;
742 }
743
744drop_unlock:
745 spin_unlock(&x->lock);
746drop:
747 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
748 kfree_skb(skb);
749 return 0;
750}
751EXPORT_SYMBOL(xfrm_input);
752
753int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
754{
755 return xfrm_input(skb, nexthdr, 0, -1);
756}
757EXPORT_SYMBOL(xfrm_input_resume);
758
759static void xfrm_trans_reinject(unsigned long data)
760{
761 struct xfrm_trans_tasklet *trans = (void *)data;
762 struct sk_buff_head queue;
763 struct sk_buff *skb;
764
765 __skb_queue_head_init(&queue);
766 skb_queue_splice_init(&trans->queue, &queue);
767
768 while ((skb = __skb_dequeue(&queue)))
769 XFRM_TRANS_SKB_CB(skb)->finish(dev_net(skb->dev), NULL, skb);
770}
771
772int xfrm_trans_queue(struct sk_buff *skb,
773 int (*finish)(struct net *, struct sock *,
774 struct sk_buff *))
775{
776 struct xfrm_trans_tasklet *trans;
777
778 trans = this_cpu_ptr(&xfrm_trans_tasklet);
779
780 if (skb_queue_len(&trans->queue) >= netdev_max_backlog)
781 return -ENOBUFS;
782
783 XFRM_TRANS_SKB_CB(skb)->finish = finish;
784 __skb_queue_tail(&trans->queue, skb);
785 tasklet_schedule(&trans->tasklet);
786 return 0;
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 __skb_queue_head_init(&trans->queue);
805 tasklet_init(&trans->tasklet, xfrm_trans_reinject,
806 (unsigned long)trans);
807 }
808}
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
135 return sp;
136}
137EXPORT_SYMBOL(secpath_set);
138
139/* Fetch spi and seq from ipsec header */
140
141int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
142{
143 int offset, offset_seq;
144 int hlen;
145
146 switch (nexthdr) {
147 case IPPROTO_AH:
148 hlen = sizeof(struct ip_auth_hdr);
149 offset = offsetof(struct ip_auth_hdr, spi);
150 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
151 break;
152 case IPPROTO_ESP:
153 hlen = sizeof(struct ip_esp_hdr);
154 offset = offsetof(struct ip_esp_hdr, spi);
155 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
156 break;
157 case IPPROTO_COMP:
158 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
159 return -EINVAL;
160 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
161 *seq = 0;
162 return 0;
163 default:
164 return 1;
165 }
166
167 if (!pskb_may_pull(skb, hlen))
168 return -EINVAL;
169
170 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
171 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
172 return 0;
173}
174EXPORT_SYMBOL(xfrm_parse_spi);
175
176static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
177{
178 struct iphdr *iph;
179 int optlen = 0;
180 int err = -EINVAL;
181
182 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
183 struct ip_beet_phdr *ph;
184 int phlen;
185
186 if (!pskb_may_pull(skb, sizeof(*ph)))
187 goto out;
188
189 ph = (struct ip_beet_phdr *)skb->data;
190
191 phlen = sizeof(*ph) + ph->padlen;
192 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
193 if (optlen < 0 || optlen & 3 || optlen > 250)
194 goto out;
195
196 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
197
198 if (!pskb_may_pull(skb, phlen))
199 goto out;
200 __skb_pull(skb, phlen);
201 }
202
203 skb_push(skb, sizeof(*iph));
204 skb_reset_network_header(skb);
205 skb_mac_header_rebuild(skb);
206
207 xfrm4_beet_make_header(skb);
208
209 iph = ip_hdr(skb);
210
211 iph->ihl += optlen / 4;
212 iph->tot_len = htons(skb->len);
213 iph->daddr = x->sel.daddr.a4;
214 iph->saddr = x->sel.saddr.a4;
215 iph->check = 0;
216 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
217 err = 0;
218out:
219 return err;
220}
221
222static void ipip_ecn_decapsulate(struct sk_buff *skb)
223{
224 struct iphdr *inner_iph = ipip_hdr(skb);
225
226 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
227 IP_ECN_set_ce(inner_iph);
228}
229
230static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
231{
232 int err = -EINVAL;
233
234 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPIP)
235 goto out;
236
237 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
238 goto out;
239
240 err = skb_unclone(skb, GFP_ATOMIC);
241 if (err)
242 goto out;
243
244 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
245 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
246 if (!(x->props.flags & XFRM_STATE_NOECN))
247 ipip_ecn_decapsulate(skb);
248
249 skb_reset_network_header(skb);
250 skb_mac_header_rebuild(skb);
251 if (skb->mac_len)
252 eth_hdr(skb)->h_proto = skb->protocol;
253
254 err = 0;
255
256out:
257 return err;
258}
259
260static void ipip6_ecn_decapsulate(struct sk_buff *skb)
261{
262 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
263
264 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
265 IP6_ECN_set_ce(skb, inner_iph);
266}
267
268static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
269{
270 int err = -EINVAL;
271
272 if (XFRM_MODE_SKB_CB(skb)->protocol != IPPROTO_IPV6)
273 goto out;
274 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
275 goto out;
276
277 err = skb_unclone(skb, GFP_ATOMIC);
278 if (err)
279 goto out;
280
281 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
282 ipv6_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipipv6_hdr(skb));
283 if (!(x->props.flags & XFRM_STATE_NOECN))
284 ipip6_ecn_decapsulate(skb);
285
286 skb_reset_network_header(skb);
287 skb_mac_header_rebuild(skb);
288 if (skb->mac_len)
289 eth_hdr(skb)->h_proto = skb->protocol;
290
291 err = 0;
292
293out:
294 return err;
295}
296
297static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
298{
299 struct ipv6hdr *ip6h;
300 int size = sizeof(struct ipv6hdr);
301 int err;
302
303 err = skb_cow_head(skb, size + skb->mac_len);
304 if (err)
305 goto out;
306
307 __skb_push(skb, size);
308 skb_reset_network_header(skb);
309 skb_mac_header_rebuild(skb);
310
311 xfrm6_beet_make_header(skb);
312
313 ip6h = ipv6_hdr(skb);
314 ip6h->payload_len = htons(skb->len - size);
315 ip6h->daddr = x->sel.daddr.in6;
316 ip6h->saddr = x->sel.saddr.in6;
317 err = 0;
318out:
319 return err;
320}
321
322/* Remove encapsulation header.
323 *
324 * The IP header will be moved over the top of the encapsulation
325 * header.
326 *
327 * On entry, the transport header shall point to where the IP header
328 * should be and the network header shall be set to where the IP
329 * header currently is. skb->data shall point to the start of the
330 * payload.
331 */
332static int
333xfrm_inner_mode_encap_remove(struct xfrm_state *x,
334 const struct xfrm_mode *inner_mode,
335 struct sk_buff *skb)
336{
337 switch (inner_mode->encap) {
338 case XFRM_MODE_BEET:
339 if (inner_mode->family == AF_INET)
340 return xfrm4_remove_beet_encap(x, skb);
341 if (inner_mode->family == AF_INET6)
342 return xfrm6_remove_beet_encap(x, skb);
343 break;
344 case XFRM_MODE_TUNNEL:
345 if (inner_mode->family == AF_INET)
346 return xfrm4_remove_tunnel_encap(x, skb);
347 if (inner_mode->family == AF_INET6)
348 return xfrm6_remove_tunnel_encap(x, skb);
349 break;
350 }
351
352 WARN_ON_ONCE(1);
353 return -EOPNOTSUPP;
354}
355
356static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
357{
358 const struct xfrm_mode *inner_mode = &x->inner_mode;
359
360 switch (x->outer_mode.family) {
361 case AF_INET:
362 xfrm4_extract_header(skb);
363 break;
364 case AF_INET6:
365 xfrm6_extract_header(skb);
366 break;
367 default:
368 WARN_ON_ONCE(1);
369 return -EAFNOSUPPORT;
370 }
371
372 if (x->sel.family == AF_UNSPEC) {
373 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
374 if (!inner_mode)
375 return -EAFNOSUPPORT;
376 }
377
378 switch (inner_mode->family) {
379 case AF_INET:
380 skb->protocol = htons(ETH_P_IP);
381 break;
382 case AF_INET6:
383 skb->protocol = htons(ETH_P_IPV6);
384 break;
385 default:
386 WARN_ON_ONCE(1);
387 break;
388 }
389
390 return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
391}
392
393/* Remove encapsulation header.
394 *
395 * The IP header will be moved over the top of the encapsulation header.
396 *
397 * On entry, skb_transport_header() shall point to where the IP header
398 * should be and skb_network_header() shall be set to where the IP header
399 * currently is. skb->data shall point to the start of the payload.
400 */
401static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
402{
403 int ihl = skb->data - skb_transport_header(skb);
404
405 if (skb->transport_header != skb->network_header) {
406 memmove(skb_transport_header(skb),
407 skb_network_header(skb), ihl);
408 skb->network_header = skb->transport_header;
409 }
410 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
411 skb_reset_transport_header(skb);
412 return 0;
413}
414
415static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
416{
417#if IS_ENABLED(CONFIG_IPV6)
418 int ihl = skb->data - skb_transport_header(skb);
419
420 if (skb->transport_header != skb->network_header) {
421 memmove(skb_transport_header(skb),
422 skb_network_header(skb), ihl);
423 skb->network_header = skb->transport_header;
424 }
425 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
426 sizeof(struct ipv6hdr));
427 skb_reset_transport_header(skb);
428 return 0;
429#else
430 WARN_ON_ONCE(1);
431 return -EAFNOSUPPORT;
432#endif
433}
434
435static int xfrm_inner_mode_input(struct xfrm_state *x,
436 const struct xfrm_mode *inner_mode,
437 struct sk_buff *skb)
438{
439 switch (inner_mode->encap) {
440 case XFRM_MODE_BEET:
441 case XFRM_MODE_TUNNEL:
442 return xfrm_prepare_input(x, skb);
443 case XFRM_MODE_TRANSPORT:
444 if (inner_mode->family == AF_INET)
445 return xfrm4_transport_input(x, skb);
446 if (inner_mode->family == AF_INET6)
447 return xfrm6_transport_input(x, skb);
448 break;
449 case XFRM_MODE_ROUTEOPTIMIZATION:
450 WARN_ON_ONCE(1);
451 break;
452 default:
453 WARN_ON_ONCE(1);
454 break;
455 }
456
457 return -EOPNOTSUPP;
458}
459
460int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
461{
462 const struct xfrm_state_afinfo *afinfo;
463 struct net *net = dev_net(skb->dev);
464 const struct xfrm_mode *inner_mode;
465 int err;
466 __be32 seq;
467 __be32 seq_hi;
468 struct xfrm_state *x = NULL;
469 xfrm_address_t *daddr;
470 u32 mark = skb->mark;
471 unsigned int family = AF_UNSPEC;
472 int decaps = 0;
473 int async = 0;
474 bool xfrm_gro = false;
475 bool crypto_done = false;
476 struct xfrm_offload *xo = xfrm_offload(skb);
477 struct sec_path *sp;
478
479 if (encap_type < 0) {
480 x = xfrm_input_state(skb);
481
482 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
483 if (x->km.state == XFRM_STATE_ACQ)
484 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
485 else
486 XFRM_INC_STATS(net,
487 LINUX_MIB_XFRMINSTATEINVALID);
488
489 if (encap_type == -1)
490 dev_put(skb->dev);
491 goto drop;
492 }
493
494 family = x->outer_mode.family;
495
496 /* An encap_type of -1 indicates async resumption. */
497 if (encap_type == -1) {
498 async = 1;
499 seq = XFRM_SKB_CB(skb)->seq.input.low;
500 goto resume;
501 }
502
503 /* encap_type < -1 indicates a GRO call. */
504 encap_type = 0;
505 seq = XFRM_SPI_SKB_CB(skb)->seq;
506
507 if (xo && (xo->flags & CRYPTO_DONE)) {
508 crypto_done = true;
509 family = XFRM_SPI_SKB_CB(skb)->family;
510
511 if (!(xo->status & CRYPTO_SUCCESS)) {
512 if (xo->status &
513 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
514 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
515 CRYPTO_TUNNEL_AH_AUTH_FAILED |
516 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
517
518 xfrm_audit_state_icvfail(x, skb,
519 x->type->proto);
520 x->stats.integrity_failed++;
521 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
522 goto drop;
523 }
524
525 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
526 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
527 goto drop;
528 }
529
530 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
531 goto drop;
532 }
533
534 if (xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
535 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
536 goto drop;
537 }
538 }
539
540 goto lock;
541 }
542
543 family = XFRM_SPI_SKB_CB(skb)->family;
544
545 /* if tunnel is present override skb->mark value with tunnel i_key */
546 switch (family) {
547 case AF_INET:
548 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
549 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
550 break;
551 case AF_INET6:
552 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
553 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
554 break;
555 }
556
557 sp = secpath_set(skb);
558 if (!sp) {
559 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
560 goto drop;
561 }
562
563 seq = 0;
564 if (!spi && xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
565 secpath_reset(skb);
566 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
567 goto drop;
568 }
569
570 daddr = (xfrm_address_t *)(skb_network_header(skb) +
571 XFRM_SPI_SKB_CB(skb)->daddroff);
572 do {
573 sp = skb_sec_path(skb);
574
575 if (sp->len == XFRM_MAX_DEPTH) {
576 secpath_reset(skb);
577 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
578 goto drop;
579 }
580
581 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
582 if (x == NULL) {
583 secpath_reset(skb);
584 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
585 xfrm_audit_state_notfound(skb, family, spi, seq);
586 goto drop;
587 }
588
589 skb->mark = xfrm_smark_get(skb->mark, x);
590
591 sp->xvec[sp->len++] = x;
592
593 skb_dst_force(skb);
594 if (!skb_dst(skb)) {
595 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
596 goto drop;
597 }
598
599lock:
600 spin_lock(&x->lock);
601
602 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
603 if (x->km.state == XFRM_STATE_ACQ)
604 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
605 else
606 XFRM_INC_STATS(net,
607 LINUX_MIB_XFRMINSTATEINVALID);
608 goto drop_unlock;
609 }
610
611 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
612 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
613 goto drop_unlock;
614 }
615
616 if (xfrm_replay_check(x, skb, seq)) {
617 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
618 goto drop_unlock;
619 }
620
621 if (xfrm_state_check_expire(x)) {
622 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
623 goto drop_unlock;
624 }
625
626 spin_unlock(&x->lock);
627
628 if (xfrm_tunnel_check(skb, x, family)) {
629 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
630 goto drop;
631 }
632
633 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
634
635 XFRM_SKB_CB(skb)->seq.input.low = seq;
636 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
637
638 dev_hold(skb->dev);
639
640 if (crypto_done)
641 nexthdr = x->type_offload->input_tail(x, skb);
642 else
643 nexthdr = x->type->input(x, skb);
644
645 if (nexthdr == -EINPROGRESS)
646 return 0;
647resume:
648 dev_put(skb->dev);
649
650 spin_lock(&x->lock);
651 if (nexthdr < 0) {
652 if (nexthdr == -EBADMSG) {
653 xfrm_audit_state_icvfail(x, skb,
654 x->type->proto);
655 x->stats.integrity_failed++;
656 }
657 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
658 goto drop_unlock;
659 }
660
661 /* only the first xfrm gets the encap type */
662 encap_type = 0;
663
664 if (xfrm_replay_recheck(x, skb, seq)) {
665 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
666 goto drop_unlock;
667 }
668
669 xfrm_replay_advance(x, seq);
670
671 x->curlft.bytes += skb->len;
672 x->curlft.packets++;
673 x->lastused = ktime_get_real_seconds();
674
675 spin_unlock(&x->lock);
676
677 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
678
679 inner_mode = &x->inner_mode;
680
681 if (x->sel.family == AF_UNSPEC) {
682 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
683 if (inner_mode == NULL) {
684 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
685 goto drop;
686 }
687 }
688
689 if (xfrm_inner_mode_input(x, inner_mode, skb)) {
690 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
691 goto drop;
692 }
693
694 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
695 decaps = 1;
696 break;
697 }
698
699 /*
700 * We need the inner address. However, we only get here for
701 * transport mode so the outer address is identical.
702 */
703 daddr = &x->id.daddr;
704 family = x->outer_mode.family;
705
706 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
707 if (err < 0) {
708 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
709 goto drop;
710 }
711 crypto_done = false;
712 } while (!err);
713
714 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
715 if (err)
716 goto drop;
717
718 nf_reset_ct(skb);
719
720 if (decaps) {
721 sp = skb_sec_path(skb);
722 if (sp)
723 sp->olen = 0;
724 if (skb_valid_dst(skb))
725 skb_dst_drop(skb);
726 gro_cells_receive(&gro_cells, skb);
727 return 0;
728 } else {
729 xo = xfrm_offload(skb);
730 if (xo)
731 xfrm_gro = xo->flags & XFRM_GRO;
732
733 err = -EAFNOSUPPORT;
734 rcu_read_lock();
735 afinfo = xfrm_state_afinfo_get_rcu(x->inner_mode.family);
736 if (likely(afinfo))
737 err = afinfo->transport_finish(skb, xfrm_gro || async);
738 rcu_read_unlock();
739 if (xfrm_gro) {
740 sp = skb_sec_path(skb);
741 if (sp)
742 sp->olen = 0;
743 if (skb_valid_dst(skb))
744 skb_dst_drop(skb);
745 gro_cells_receive(&gro_cells, skb);
746 return err;
747 }
748
749 return err;
750 }
751
752drop_unlock:
753 spin_unlock(&x->lock);
754drop:
755 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
756 kfree_skb(skb);
757 return 0;
758}
759EXPORT_SYMBOL(xfrm_input);
760
761int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
762{
763 return xfrm_input(skb, nexthdr, 0, -1);
764}
765EXPORT_SYMBOL(xfrm_input_resume);
766
767static void xfrm_trans_reinject(struct work_struct *work)
768{
769 struct xfrm_trans_tasklet *trans = container_of(work, struct xfrm_trans_tasklet, work);
770 struct sk_buff_head queue;
771 struct sk_buff *skb;
772
773 __skb_queue_head_init(&queue);
774 spin_lock_bh(&trans->queue_lock);
775 skb_queue_splice_init(&trans->queue, &queue);
776 spin_unlock_bh(&trans->queue_lock);
777
778 local_bh_disable();
779 while ((skb = __skb_dequeue(&queue)))
780 XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
781 NULL, skb);
782 local_bh_enable();
783}
784
785int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
786 int (*finish)(struct net *, struct sock *,
787 struct sk_buff *))
788{
789 struct xfrm_trans_tasklet *trans;
790
791 trans = this_cpu_ptr(&xfrm_trans_tasklet);
792
793 if (skb_queue_len(&trans->queue) >= READ_ONCE(netdev_max_backlog))
794 return -ENOBUFS;
795
796 BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
797
798 XFRM_TRANS_SKB_CB(skb)->finish = finish;
799 XFRM_TRANS_SKB_CB(skb)->net = net;
800 spin_lock_bh(&trans->queue_lock);
801 __skb_queue_tail(&trans->queue, skb);
802 spin_unlock_bh(&trans->queue_lock);
803 schedule_work(&trans->work);
804 return 0;
805}
806EXPORT_SYMBOL(xfrm_trans_queue_net);
807
808int xfrm_trans_queue(struct sk_buff *skb,
809 int (*finish)(struct net *, struct sock *,
810 struct sk_buff *))
811{
812 return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
813}
814EXPORT_SYMBOL(xfrm_trans_queue);
815
816void __init xfrm_input_init(void)
817{
818 int err;
819 int i;
820
821 init_dummy_netdev(&xfrm_napi_dev);
822 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
823 if (err)
824 gro_cells.cells = NULL;
825
826 for_each_possible_cpu(i) {
827 struct xfrm_trans_tasklet *trans;
828
829 trans = &per_cpu(xfrm_trans_tasklet, i);
830 spin_lock_init(&trans->queue_lock);
831 __skb_queue_head_init(&trans->queue);
832 INIT_WORK(&trans->work, xfrm_trans_reinject);
833 }
834}