drm/panthor: Fix race condition when gathering fdinfo group samples
[drm/drm-misc.git] / net / xfrm / xfrm_input.c
blob7e6a71b9d6a3ae2ff1a2742ca620729128823a06
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * xfrm_input.c
5 * Changes:
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
9 */
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 #include <net/hotdata.h>
26 #include "xfrm_inout.h"
28 struct xfrm_trans_tasklet {
29 struct work_struct work;
30 spinlock_t queue_lock;
31 struct sk_buff_head queue;
34 struct xfrm_trans_cb {
35 union {
36 struct inet_skb_parm h4;
37 #if IS_ENABLED(CONFIG_IPV6)
38 struct inet6_skb_parm h6;
39 #endif
40 } header;
41 int (*finish)(struct net *net, struct sock *sk, struct sk_buff *skb);
42 struct net *net;
45 #define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
47 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
48 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
50 static struct gro_cells gro_cells;
51 static struct net_device *xfrm_napi_dev;
53 static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
55 int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
57 int err = 0;
59 if (WARN_ON(afinfo->family > AF_INET6))
60 return -EAFNOSUPPORT;
62 spin_lock_bh(&xfrm_input_afinfo_lock);
63 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family]))
64 err = -EEXIST;
65 else
66 rcu_assign_pointer(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], afinfo);
67 spin_unlock_bh(&xfrm_input_afinfo_lock);
68 return err;
70 EXPORT_SYMBOL(xfrm_input_register_afinfo);
72 int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
74 int err = 0;
76 spin_lock_bh(&xfrm_input_afinfo_lock);
77 if (likely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family])) {
78 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family] != afinfo))
79 err = -EINVAL;
80 else
81 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], NULL);
83 spin_unlock_bh(&xfrm_input_afinfo_lock);
84 synchronize_rcu();
85 return err;
87 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
89 static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(u8 family, bool is_ipip)
91 const struct xfrm_input_afinfo *afinfo;
93 if (WARN_ON_ONCE(family > AF_INET6))
94 return NULL;
96 rcu_read_lock();
97 afinfo = rcu_dereference(xfrm_input_afinfo[is_ipip][family]);
98 if (unlikely(!afinfo))
99 rcu_read_unlock();
100 return afinfo;
103 static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
104 int err)
106 bool is_ipip = (protocol == IPPROTO_IPIP || protocol == IPPROTO_IPV6);
107 const struct xfrm_input_afinfo *afinfo;
108 int ret;
110 afinfo = xfrm_input_get_afinfo(family, is_ipip);
111 if (!afinfo)
112 return -EAFNOSUPPORT;
114 ret = afinfo->callback(skb, protocol, err);
115 rcu_read_unlock();
117 return ret;
120 struct sec_path *secpath_set(struct sk_buff *skb)
122 struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
124 sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
125 if (!sp)
126 return NULL;
128 if (tmp) /* reused existing one (was COW'd if needed) */
129 return sp;
131 /* allocated new secpath */
132 memset(sp->ovec, 0, sizeof(sp->ovec));
133 sp->olen = 0;
134 sp->len = 0;
135 sp->verified_cnt = 0;
137 return sp;
139 EXPORT_SYMBOL(secpath_set);
141 /* Fetch spi and seq from ipsec header */
143 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
145 int offset, offset_seq;
146 int hlen;
148 switch (nexthdr) {
149 case IPPROTO_AH:
150 hlen = sizeof(struct ip_auth_hdr);
151 offset = offsetof(struct ip_auth_hdr, spi);
152 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
153 break;
154 case IPPROTO_ESP:
155 hlen = sizeof(struct ip_esp_hdr);
156 offset = offsetof(struct ip_esp_hdr, spi);
157 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
158 break;
159 case IPPROTO_COMP:
160 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
161 return -EINVAL;
162 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
163 *seq = 0;
164 return 0;
165 default:
166 return 1;
169 if (!pskb_may_pull(skb, hlen))
170 return -EINVAL;
172 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
173 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
174 return 0;
176 EXPORT_SYMBOL(xfrm_parse_spi);
178 static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
180 struct iphdr *iph;
181 int optlen = 0;
182 int err = -EINVAL;
184 skb->protocol = htons(ETH_P_IP);
186 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
187 struct ip_beet_phdr *ph;
188 int phlen;
190 if (!pskb_may_pull(skb, sizeof(*ph)))
191 goto out;
193 ph = (struct ip_beet_phdr *)skb->data;
195 phlen = sizeof(*ph) + ph->padlen;
196 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
197 if (optlen < 0 || optlen & 3 || optlen > 250)
198 goto out;
200 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
202 if (!pskb_may_pull(skb, phlen))
203 goto out;
204 __skb_pull(skb, phlen);
207 skb_push(skb, sizeof(*iph));
208 skb_reset_network_header(skb);
209 skb_mac_header_rebuild(skb);
211 xfrm4_beet_make_header(skb);
213 iph = ip_hdr(skb);
215 iph->ihl += optlen / 4;
216 iph->tot_len = htons(skb->len);
217 iph->daddr = x->sel.daddr.a4;
218 iph->saddr = x->sel.saddr.a4;
219 iph->check = 0;
220 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
221 err = 0;
222 out:
223 return err;
226 static void ipip_ecn_decapsulate(struct sk_buff *skb)
228 struct iphdr *inner_iph = ipip_hdr(skb);
230 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
231 IP_ECN_set_ce(inner_iph);
234 static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
236 int err = -EINVAL;
238 skb->protocol = htons(ETH_P_IP);
240 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
241 goto out;
243 err = skb_unclone(skb, GFP_ATOMIC);
244 if (err)
245 goto out;
247 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
248 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
249 if (!(x->props.flags & XFRM_STATE_NOECN))
250 ipip_ecn_decapsulate(skb);
252 skb_reset_network_header(skb);
253 skb_mac_header_rebuild(skb);
254 if (skb->mac_len)
255 eth_hdr(skb)->h_proto = skb->protocol;
257 err = 0;
259 out:
260 return err;
263 static void ipip6_ecn_decapsulate(struct sk_buff *skb)
265 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
267 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
268 IP6_ECN_set_ce(skb, inner_iph);
271 static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
273 int err = -EINVAL;
275 skb->protocol = htons(ETH_P_IPV6);
277 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
278 goto out;
280 err = skb_unclone(skb, GFP_ATOMIC);
281 if (err)
282 goto out;
284 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
285 ipv6_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipipv6_hdr(skb));
286 if (!(x->props.flags & XFRM_STATE_NOECN))
287 ipip6_ecn_decapsulate(skb);
289 skb_reset_network_header(skb);
290 skb_mac_header_rebuild(skb);
291 if (skb->mac_len)
292 eth_hdr(skb)->h_proto = skb->protocol;
294 err = 0;
296 out:
297 return err;
300 static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
302 struct ipv6hdr *ip6h;
303 int size = sizeof(struct ipv6hdr);
304 int err;
306 skb->protocol = htons(ETH_P_IPV6);
308 err = skb_cow_head(skb, size + skb->mac_len);
309 if (err)
310 goto out;
312 __skb_push(skb, size);
313 skb_reset_network_header(skb);
314 skb_mac_header_rebuild(skb);
316 xfrm6_beet_make_header(skb);
318 ip6h = ipv6_hdr(skb);
319 ip6h->payload_len = htons(skb->len - size);
320 ip6h->daddr = x->sel.daddr.in6;
321 ip6h->saddr = x->sel.saddr.in6;
322 err = 0;
323 out:
324 return err;
327 /* Remove encapsulation header.
329 * The IP header will be moved over the top of the encapsulation
330 * header.
332 * On entry, the transport header shall point to where the IP header
333 * should be and the network header shall be set to where the IP
334 * header currently is. skb->data shall point to the start of the
335 * payload.
337 static int
338 xfrm_inner_mode_encap_remove(struct xfrm_state *x,
339 struct sk_buff *skb)
341 switch (x->props.mode) {
342 case XFRM_MODE_BEET:
343 switch (x->sel.family) {
344 case AF_INET:
345 return xfrm4_remove_beet_encap(x, skb);
346 case AF_INET6:
347 return xfrm6_remove_beet_encap(x, skb);
349 break;
350 case XFRM_MODE_TUNNEL:
351 switch (XFRM_MODE_SKB_CB(skb)->protocol) {
352 case IPPROTO_IPIP:
353 return xfrm4_remove_tunnel_encap(x, skb);
354 case IPPROTO_IPV6:
355 return xfrm6_remove_tunnel_encap(x, skb);
356 break;
358 return -EINVAL;
361 WARN_ON_ONCE(1);
362 return -EOPNOTSUPP;
365 static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
367 switch (x->props.family) {
368 case AF_INET:
369 xfrm4_extract_header(skb);
370 break;
371 case AF_INET6:
372 xfrm6_extract_header(skb);
373 break;
374 default:
375 WARN_ON_ONCE(1);
376 return -EAFNOSUPPORT;
379 return xfrm_inner_mode_encap_remove(x, skb);
382 /* Remove encapsulation header.
384 * The IP header will be moved over the top of the encapsulation header.
386 * On entry, skb_transport_header() shall point to where the IP header
387 * should be and skb_network_header() shall be set to where the IP header
388 * currently is. skb->data shall point to the start of the payload.
390 static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
392 struct xfrm_offload *xo = xfrm_offload(skb);
393 int ihl = skb->data - skb_transport_header(skb);
395 if (skb->transport_header != skb->network_header) {
396 memmove(skb_transport_header(skb),
397 skb_network_header(skb), ihl);
398 if (xo)
399 xo->orig_mac_len =
400 skb_mac_header_was_set(skb) ? skb_mac_header_len(skb) : 0;
401 skb->network_header = skb->transport_header;
403 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
404 skb_reset_transport_header(skb);
405 return 0;
408 static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
410 #if IS_ENABLED(CONFIG_IPV6)
411 struct xfrm_offload *xo = xfrm_offload(skb);
412 int ihl = skb->data - skb_transport_header(skb);
414 if (skb->transport_header != skb->network_header) {
415 memmove(skb_transport_header(skb),
416 skb_network_header(skb), ihl);
417 if (xo)
418 xo->orig_mac_len =
419 skb_mac_header_was_set(skb) ? skb_mac_header_len(skb) : 0;
420 skb->network_header = skb->transport_header;
422 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
423 sizeof(struct ipv6hdr));
424 skb_reset_transport_header(skb);
425 return 0;
426 #else
427 WARN_ON_ONCE(1);
428 return -EAFNOSUPPORT;
429 #endif
432 static int xfrm_inner_mode_input(struct xfrm_state *x,
433 struct sk_buff *skb)
435 switch (x->props.mode) {
436 case XFRM_MODE_BEET:
437 case XFRM_MODE_TUNNEL:
438 return xfrm_prepare_input(x, skb);
439 case XFRM_MODE_TRANSPORT:
440 if (x->props.family == AF_INET)
441 return xfrm4_transport_input(x, skb);
442 if (x->props.family == AF_INET6)
443 return xfrm6_transport_input(x, skb);
444 break;
445 case XFRM_MODE_ROUTEOPTIMIZATION:
446 WARN_ON_ONCE(1);
447 break;
448 default:
449 if (x->mode_cbs && x->mode_cbs->input)
450 return x->mode_cbs->input(x, skb);
452 WARN_ON_ONCE(1);
453 break;
456 return -EOPNOTSUPP;
459 /* NOTE: encap_type - In addition to the normal (non-negative) values for
460 * encap_type, a negative value of -1 or -2 can be used to resume/restart this
461 * function after a previous invocation early terminated for async operation.
463 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
465 const struct xfrm_state_afinfo *afinfo;
466 struct net *net = dev_net(skb->dev);
467 int err;
468 __be32 seq;
469 __be32 seq_hi;
470 struct xfrm_state *x = NULL;
471 xfrm_address_t *daddr;
472 u32 mark = skb->mark;
473 unsigned int family = AF_UNSPEC;
474 int decaps = 0;
475 int async = 0;
476 bool xfrm_gro = false;
477 bool crypto_done = false;
478 struct xfrm_offload *xo = xfrm_offload(skb);
479 struct sec_path *sp;
481 if (encap_type < 0 || (xo && (xo->flags & XFRM_GRO || encap_type == 0 ||
482 encap_type == UDP_ENCAP_ESPINUDP))) {
483 x = xfrm_input_state(skb);
485 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
486 if (x->km.state == XFRM_STATE_ACQ)
487 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
488 else
489 XFRM_INC_STATS(net,
490 LINUX_MIB_XFRMINSTATEINVALID);
492 if (encap_type == -1)
493 dev_put(skb->dev);
494 goto drop;
497 family = x->props.family;
499 /* An encap_type of -2 indicates reconstructed inner packet */
500 if (encap_type == -2)
501 goto resume_decapped;
503 /* An encap_type of -1 indicates async resumption. */
504 if (encap_type == -1) {
505 async = 1;
506 seq = XFRM_SKB_CB(skb)->seq.input.low;
507 goto resume;
509 /* GRO call */
510 seq = XFRM_SPI_SKB_CB(skb)->seq;
512 if (xo && (xo->flags & CRYPTO_DONE)) {
513 crypto_done = true;
514 family = XFRM_SPI_SKB_CB(skb)->family;
516 if (!(xo->status & CRYPTO_SUCCESS)) {
517 if (xo->status &
518 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
519 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
520 CRYPTO_TUNNEL_AH_AUTH_FAILED |
521 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
523 xfrm_audit_state_icvfail(x, skb,
524 x->type->proto);
525 x->stats.integrity_failed++;
526 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
527 goto drop;
530 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
531 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
532 goto drop;
535 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
536 goto drop;
539 if (xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
540 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
541 goto drop;
545 goto lock;
548 family = XFRM_SPI_SKB_CB(skb)->family;
550 /* if tunnel is present override skb->mark value with tunnel i_key */
551 switch (family) {
552 case AF_INET:
553 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
554 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
555 break;
556 case AF_INET6:
557 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
558 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
559 break;
562 sp = secpath_set(skb);
563 if (!sp) {
564 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
565 goto drop;
568 seq = 0;
569 if (!spi && xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
570 secpath_reset(skb);
571 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
572 goto drop;
575 daddr = (xfrm_address_t *)(skb_network_header(skb) +
576 XFRM_SPI_SKB_CB(skb)->daddroff);
577 do {
578 sp = skb_sec_path(skb);
580 if (sp->len == XFRM_MAX_DEPTH) {
581 secpath_reset(skb);
582 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
583 goto drop;
586 x = xfrm_input_state_lookup(net, mark, daddr, spi, nexthdr, family);
587 if (x == NULL) {
588 secpath_reset(skb);
589 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
590 xfrm_audit_state_notfound(skb, family, spi, seq);
591 goto drop;
594 if (unlikely(x->dir && x->dir != XFRM_SA_DIR_IN)) {
595 secpath_reset(skb);
596 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEDIRERROR);
597 xfrm_audit_state_notfound(skb, family, spi, seq);
598 xfrm_state_put(x);
599 x = NULL;
600 goto drop;
603 skb->mark = xfrm_smark_get(skb->mark, x);
605 sp->xvec[sp->len++] = x;
607 skb_dst_force(skb);
608 if (!skb_dst(skb)) {
609 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
610 goto drop;
613 lock:
614 spin_lock(&x->lock);
616 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
617 if (x->km.state == XFRM_STATE_ACQ)
618 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
619 else
620 XFRM_INC_STATS(net,
621 LINUX_MIB_XFRMINSTATEINVALID);
622 goto drop_unlock;
625 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
626 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
627 goto drop_unlock;
630 if (xfrm_replay_check(x, skb, seq)) {
631 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
632 goto drop_unlock;
635 if (xfrm_state_check_expire(x)) {
636 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
637 goto drop_unlock;
640 spin_unlock(&x->lock);
642 if (xfrm_tunnel_check(skb, x, family)) {
643 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
644 goto drop;
647 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
649 XFRM_SKB_CB(skb)->seq.input.low = seq;
650 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
652 dev_hold(skb->dev);
654 if (crypto_done)
655 nexthdr = x->type_offload->input_tail(x, skb);
656 else
657 nexthdr = x->type->input(x, skb);
659 if (nexthdr == -EINPROGRESS)
660 return 0;
661 resume:
662 dev_put(skb->dev);
664 spin_lock(&x->lock);
665 if (nexthdr < 0) {
666 if (nexthdr == -EBADMSG) {
667 xfrm_audit_state_icvfail(x, skb,
668 x->type->proto);
669 x->stats.integrity_failed++;
671 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
672 goto drop_unlock;
675 /* only the first xfrm gets the encap type */
676 encap_type = 0;
678 if (xfrm_replay_recheck(x, skb, seq)) {
679 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
680 goto drop_unlock;
683 xfrm_replay_advance(x, seq);
685 x->curlft.bytes += skb->len;
686 x->curlft.packets++;
687 x->lastused = ktime_get_real_seconds();
689 spin_unlock(&x->lock);
691 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
693 err = xfrm_inner_mode_input(x, skb);
694 if (err == -EINPROGRESS)
695 return 0;
696 else if (err) {
697 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
698 goto drop;
700 resume_decapped:
701 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
702 decaps = 1;
703 break;
707 * We need the inner address. However, we only get here for
708 * transport mode so the outer address is identical.
710 daddr = &x->id.daddr;
711 family = x->props.family;
713 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
714 if (err < 0) {
715 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
716 goto drop;
718 crypto_done = false;
719 } while (!err);
721 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
722 if (err)
723 goto drop;
725 nf_reset_ct(skb);
727 if (decaps) {
728 sp = skb_sec_path(skb);
729 if (sp)
730 sp->olen = 0;
731 if (skb_valid_dst(skb))
732 skb_dst_drop(skb);
733 gro_cells_receive(&gro_cells, skb);
734 return 0;
735 } else {
736 xo = xfrm_offload(skb);
737 if (xo)
738 xfrm_gro = xo->flags & XFRM_GRO;
740 err = -EAFNOSUPPORT;
741 rcu_read_lock();
742 afinfo = xfrm_state_afinfo_get_rcu(x->props.family);
743 if (likely(afinfo))
744 err = afinfo->transport_finish(skb, xfrm_gro || async);
745 rcu_read_unlock();
746 if (xfrm_gro) {
747 sp = skb_sec_path(skb);
748 if (sp)
749 sp->olen = 0;
750 if (skb_valid_dst(skb))
751 skb_dst_drop(skb);
752 gro_cells_receive(&gro_cells, skb);
753 return err;
756 return err;
759 drop_unlock:
760 spin_unlock(&x->lock);
761 drop:
762 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
763 kfree_skb(skb);
764 return 0;
766 EXPORT_SYMBOL(xfrm_input);
768 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
770 return xfrm_input(skb, nexthdr, 0, -1);
772 EXPORT_SYMBOL(xfrm_input_resume);
774 static void xfrm_trans_reinject(struct work_struct *work)
776 struct xfrm_trans_tasklet *trans = container_of(work, struct xfrm_trans_tasklet, work);
777 struct sk_buff_head queue;
778 struct sk_buff *skb;
780 __skb_queue_head_init(&queue);
781 spin_lock_bh(&trans->queue_lock);
782 skb_queue_splice_init(&trans->queue, &queue);
783 spin_unlock_bh(&trans->queue_lock);
785 local_bh_disable();
786 while ((skb = __skb_dequeue(&queue)))
787 XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
788 NULL, skb);
789 local_bh_enable();
792 int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
793 int (*finish)(struct net *, struct sock *,
794 struct sk_buff *))
796 struct xfrm_trans_tasklet *trans;
798 trans = this_cpu_ptr(&xfrm_trans_tasklet);
800 if (skb_queue_len(&trans->queue) >= READ_ONCE(net_hotdata.max_backlog))
801 return -ENOBUFS;
803 BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
805 XFRM_TRANS_SKB_CB(skb)->finish = finish;
806 XFRM_TRANS_SKB_CB(skb)->net = net;
807 spin_lock_bh(&trans->queue_lock);
808 __skb_queue_tail(&trans->queue, skb);
809 spin_unlock_bh(&trans->queue_lock);
810 schedule_work(&trans->work);
811 return 0;
813 EXPORT_SYMBOL(xfrm_trans_queue_net);
815 int xfrm_trans_queue(struct sk_buff *skb,
816 int (*finish)(struct net *, struct sock *,
817 struct sk_buff *))
819 return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
821 EXPORT_SYMBOL(xfrm_trans_queue);
823 void __init xfrm_input_init(void)
825 int err;
826 int i;
828 xfrm_napi_dev = alloc_netdev_dummy(0);
829 if (!xfrm_napi_dev)
830 panic("Failed to allocate XFRM dummy netdev\n");
832 err = gro_cells_init(&gro_cells, xfrm_napi_dev);
833 if (err)
834 gro_cells.cells = NULL;
836 for_each_possible_cpu(i) {
837 struct xfrm_trans_tasklet *trans;
839 trans = &per_cpu(xfrm_trans_tasklet, i);
840 spin_lock_init(&trans->queue_lock);
841 __skb_queue_head_init(&trans->queue);
842 INIT_WORK(&trans->work, xfrm_trans_reinject);