• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
26 struct xfrm_trans_tasklet {
27 	struct tasklet_struct tasklet;
28 	struct sk_buff_head queue;
29 };
30 
31 struct 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 
44 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
45 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
46 
47 static struct gro_cells gro_cells;
48 static struct net_device xfrm_napi_dev;
49 
50 static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
51 
xfrm_input_register_afinfo(const struct xfrm_input_afinfo * afinfo)52 int 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 }
67 EXPORT_SYMBOL(xfrm_input_register_afinfo);
68 
xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo * afinfo)69 int 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 }
84 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
85 
xfrm_input_get_afinfo(u8 family,bool is_ipip)86 static 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 
xfrm_rcv_cb(struct sk_buff * skb,unsigned int family,u8 protocol,int err)100 static 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 
secpath_set(struct sk_buff * skb)117 struct 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 }
135 EXPORT_SYMBOL(secpath_set);
136 
137 /* Fetch spi and seq from ipsec header */
138 
xfrm_parse_spi(struct sk_buff * skb,u8 nexthdr,__be32 * spi,__be32 * seq)139 int 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 }
172 EXPORT_SYMBOL(xfrm_parse_spi);
173 
xfrm4_remove_beet_encap(struct xfrm_state * x,struct sk_buff * skb)174 static 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;
216 out:
217 	return err;
218 }
219 
ipip_ecn_decapsulate(struct sk_buff * skb)220 static 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 
xfrm4_remove_tunnel_encap(struct xfrm_state * x,struct sk_buff * skb)228 static 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 
254 out:
255 	return err;
256 }
257 
ipip6_ecn_decapsulate(struct sk_buff * skb)258 static 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 
xfrm6_remove_tunnel_encap(struct xfrm_state * x,struct sk_buff * skb)266 static 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(XFRM_MODE_SKB_CB(skb)->tos, ipipv6_hdr(skb));
281 	if (!(x->props.flags & XFRM_STATE_NOECN))
282 		ipip6_ecn_decapsulate(skb);
283 
284 	skb_reset_network_header(skb);
285 	skb_mac_header_rebuild(skb);
286 	if (skb->mac_len)
287 		eth_hdr(skb)->h_proto = skb->protocol;
288 
289 	err = 0;
290 
291 out:
292 	return err;
293 }
294 
xfrm6_remove_beet_encap(struct xfrm_state * x,struct sk_buff * skb)295 static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
296 {
297 	struct ipv6hdr *ip6h;
298 	int size = sizeof(struct ipv6hdr);
299 	int err;
300 
301 	err = skb_cow_head(skb, size + skb->mac_len);
302 	if (err)
303 		goto out;
304 
305 	__skb_push(skb, size);
306 	skb_reset_network_header(skb);
307 	skb_mac_header_rebuild(skb);
308 
309 	xfrm6_beet_make_header(skb);
310 
311 	ip6h = ipv6_hdr(skb);
312 	ip6h->payload_len = htons(skb->len - size);
313 	ip6h->daddr = x->sel.daddr.in6;
314 	ip6h->saddr = x->sel.saddr.in6;
315 	err = 0;
316 out:
317 	return err;
318 }
319 
320 /* Remove encapsulation header.
321  *
322  * The IP header will be moved over the top of the encapsulation
323  * header.
324  *
325  * On entry, the transport header shall point to where the IP header
326  * should be and the network header shall be set to where the IP
327  * header currently is.  skb->data shall point to the start of the
328  * payload.
329  */
330 static int
xfrm_inner_mode_encap_remove(struct xfrm_state * x,const struct xfrm_mode * inner_mode,struct sk_buff * skb)331 xfrm_inner_mode_encap_remove(struct xfrm_state *x,
332 			     const struct xfrm_mode *inner_mode,
333 			     struct sk_buff *skb)
334 {
335 	switch (inner_mode->encap) {
336 	case XFRM_MODE_BEET:
337 		if (inner_mode->family == AF_INET)
338 			return xfrm4_remove_beet_encap(x, skb);
339 		if (inner_mode->family == AF_INET6)
340 			return xfrm6_remove_beet_encap(x, skb);
341 		break;
342 	case XFRM_MODE_TUNNEL:
343 		if (inner_mode->family == AF_INET)
344 			return xfrm4_remove_tunnel_encap(x, skb);
345 		if (inner_mode->family == AF_INET6)
346 			return xfrm6_remove_tunnel_encap(x, skb);
347 		break;
348 	}
349 
350 	WARN_ON_ONCE(1);
351 	return -EOPNOTSUPP;
352 }
353 
xfrm_prepare_input(struct xfrm_state * x,struct sk_buff * skb)354 static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
355 {
356 	const struct xfrm_mode *inner_mode = &x->inner_mode;
357 
358 	switch (x->outer_mode.family) {
359 	case AF_INET:
360 		xfrm4_extract_header(skb);
361 		break;
362 	case AF_INET6:
363 		xfrm6_extract_header(skb);
364 		break;
365 	default:
366 		WARN_ON_ONCE(1);
367 		return -EAFNOSUPPORT;
368 	}
369 
370 	if (x->sel.family == AF_UNSPEC) {
371 		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
372 		if (!inner_mode)
373 			return -EAFNOSUPPORT;
374 	}
375 
376 	switch (inner_mode->family) {
377 	case AF_INET:
378 		skb->protocol = htons(ETH_P_IP);
379 		break;
380 	case AF_INET6:
381 		skb->protocol = htons(ETH_P_IPV6);
382 		break;
383 	default:
384 		WARN_ON_ONCE(1);
385 		break;
386 	}
387 
388 	return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
389 }
390 
391 /* Remove encapsulation header.
392  *
393  * The IP header will be moved over the top of the encapsulation header.
394  *
395  * On entry, skb_transport_header() shall point to where the IP header
396  * should be and skb_network_header() shall be set to where the IP header
397  * currently is.  skb->data shall point to the start of the payload.
398  */
xfrm4_transport_input(struct xfrm_state * x,struct sk_buff * skb)399 static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
400 {
401 	int ihl = skb->data - skb_transport_header(skb);
402 
403 	if (skb->transport_header != skb->network_header) {
404 		memmove(skb_transport_header(skb),
405 			skb_network_header(skb), ihl);
406 		skb->network_header = skb->transport_header;
407 	}
408 	ip_hdr(skb)->tot_len = htons(skb->len + ihl);
409 	skb_reset_transport_header(skb);
410 	return 0;
411 }
412 
xfrm6_transport_input(struct xfrm_state * x,struct sk_buff * skb)413 static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
414 {
415 #if IS_ENABLED(CONFIG_IPV6)
416 	int ihl = skb->data - skb_transport_header(skb);
417 
418 	if (skb->transport_header != skb->network_header) {
419 		memmove(skb_transport_header(skb),
420 			skb_network_header(skb), ihl);
421 		skb->network_header = skb->transport_header;
422 	}
423 	ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
424 					   sizeof(struct ipv6hdr));
425 	skb_reset_transport_header(skb);
426 	return 0;
427 #else
428 	WARN_ON_ONCE(1);
429 	return -EAFNOSUPPORT;
430 #endif
431 }
432 
xfrm_inner_mode_input(struct xfrm_state * x,const struct xfrm_mode * inner_mode,struct sk_buff * skb)433 static int xfrm_inner_mode_input(struct xfrm_state *x,
434 				 const struct xfrm_mode *inner_mode,
435 				 struct sk_buff *skb)
436 {
437 	switch (inner_mode->encap) {
438 	case XFRM_MODE_BEET:
439 	case XFRM_MODE_TUNNEL:
440 		return xfrm_prepare_input(x, skb);
441 	case XFRM_MODE_TRANSPORT:
442 		if (inner_mode->family == AF_INET)
443 			return xfrm4_transport_input(x, skb);
444 		if (inner_mode->family == AF_INET6)
445 			return xfrm6_transport_input(x, skb);
446 		break;
447 	case XFRM_MODE_ROUTEOPTIMIZATION:
448 		WARN_ON_ONCE(1);
449 		break;
450 	default:
451 		WARN_ON_ONCE(1);
452 		break;
453 	}
454 
455 	return -EOPNOTSUPP;
456 }
457 
xfrm_input(struct sk_buff * skb,int nexthdr,__be32 spi,int encap_type)458 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
459 {
460 	const struct xfrm_state_afinfo *afinfo;
461 	struct net *net = dev_net(skb->dev);
462 	const struct xfrm_mode *inner_mode;
463 	int err;
464 	__be32 seq;
465 	__be32 seq_hi;
466 	struct xfrm_state *x = NULL;
467 	xfrm_address_t *daddr;
468 	u32 mark = skb->mark;
469 	unsigned int family = AF_UNSPEC;
470 	int decaps = 0;
471 	int async = 0;
472 	bool xfrm_gro = false;
473 	bool crypto_done = false;
474 	struct xfrm_offload *xo = xfrm_offload(skb);
475 	struct sec_path *sp;
476 
477 	if (encap_type < 0) {
478 		x = xfrm_input_state(skb);
479 
480 		if (unlikely(x->km.state != XFRM_STATE_VALID)) {
481 			if (x->km.state == XFRM_STATE_ACQ)
482 				XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
483 			else
484 				XFRM_INC_STATS(net,
485 					       LINUX_MIB_XFRMINSTATEINVALID);
486 
487 			if (encap_type == -1)
488 				dev_put(skb->dev);
489 			goto drop;
490 		}
491 
492 		family = x->outer_mode.family;
493 
494 		/* An encap_type of -1 indicates async resumption. */
495 		if (encap_type == -1) {
496 			async = 1;
497 			seq = XFRM_SKB_CB(skb)->seq.input.low;
498 			goto resume;
499 		}
500 
501 		/* encap_type < -1 indicates a GRO call. */
502 		encap_type = 0;
503 		seq = XFRM_SPI_SKB_CB(skb)->seq;
504 
505 		if (xo && (xo->flags & CRYPTO_DONE)) {
506 			crypto_done = true;
507 			family = XFRM_SPI_SKB_CB(skb)->family;
508 
509 			if (!(xo->status & CRYPTO_SUCCESS)) {
510 				if (xo->status &
511 				    (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
512 				     CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
513 				     CRYPTO_TUNNEL_AH_AUTH_FAILED |
514 				     CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
515 
516 					xfrm_audit_state_icvfail(x, skb,
517 								 x->type->proto);
518 					x->stats.integrity_failed++;
519 					XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
520 					goto drop;
521 				}
522 
523 				if (xo->status & CRYPTO_INVALID_PROTOCOL) {
524 					XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
525 					goto drop;
526 				}
527 
528 				XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
529 				goto drop;
530 			}
531 
532 			if ((err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
533 				XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
534 				goto drop;
535 			}
536 		}
537 
538 		goto lock;
539 	}
540 
541 	family = XFRM_SPI_SKB_CB(skb)->family;
542 
543 	/* if tunnel is present override skb->mark value with tunnel i_key */
544 	switch (family) {
545 	case AF_INET:
546 		if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
547 			mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
548 		break;
549 	case AF_INET6:
550 		if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
551 			mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
552 		break;
553 	}
554 
555 	sp = secpath_set(skb);
556 	if (!sp) {
557 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
558 		goto drop;
559 	}
560 
561 	seq = 0;
562 	if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
563 		secpath_reset(skb);
564 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
565 		goto drop;
566 	}
567 
568 	daddr = (xfrm_address_t *)(skb_network_header(skb) +
569 				   XFRM_SPI_SKB_CB(skb)->daddroff);
570 	do {
571 		sp = skb_sec_path(skb);
572 
573 		if (sp->len == XFRM_MAX_DEPTH) {
574 			secpath_reset(skb);
575 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
576 			goto drop;
577 		}
578 
579 		x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
580 		if (x == NULL) {
581 			secpath_reset(skb);
582 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
583 			xfrm_audit_state_notfound(skb, family, spi, seq);
584 			goto drop;
585 		}
586 
587 		skb->mark = xfrm_smark_get(skb->mark, x);
588 
589 		sp->xvec[sp->len++] = x;
590 
591 		skb_dst_force(skb);
592 		if (!skb_dst(skb)) {
593 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
594 			goto drop;
595 		}
596 
597 lock:
598 		spin_lock(&x->lock);
599 
600 		if (unlikely(x->km.state != XFRM_STATE_VALID)) {
601 			if (x->km.state == XFRM_STATE_ACQ)
602 				XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
603 			else
604 				XFRM_INC_STATS(net,
605 					       LINUX_MIB_XFRMINSTATEINVALID);
606 			goto drop_unlock;
607 		}
608 
609 		if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
610 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
611 			goto drop_unlock;
612 		}
613 
614 		if (x->repl->check(x, skb, seq)) {
615 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
616 			goto drop_unlock;
617 		}
618 
619 		if (xfrm_state_check_expire(x)) {
620 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
621 			goto drop_unlock;
622 		}
623 
624 		spin_unlock(&x->lock);
625 
626 		if (xfrm_tunnel_check(skb, x, family)) {
627 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
628 			goto drop;
629 		}
630 
631 		seq_hi = htonl(xfrm_replay_seqhi(x, seq));
632 
633 		XFRM_SKB_CB(skb)->seq.input.low = seq;
634 		XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
635 
636 		dev_hold(skb->dev);
637 
638 		if (crypto_done)
639 			nexthdr = x->type_offload->input_tail(x, skb);
640 		else
641 			nexthdr = x->type->input(x, skb);
642 
643 		if (nexthdr == -EINPROGRESS)
644 			return 0;
645 resume:
646 		dev_put(skb->dev);
647 
648 		spin_lock(&x->lock);
649 		if (nexthdr < 0) {
650 			if (nexthdr == -EBADMSG) {
651 				xfrm_audit_state_icvfail(x, skb,
652 							 x->type->proto);
653 				x->stats.integrity_failed++;
654 			}
655 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
656 			goto drop_unlock;
657 		}
658 
659 		/* only the first xfrm gets the encap type */
660 		encap_type = 0;
661 
662 		if (x->repl->recheck(x, skb, seq)) {
663 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
664 			goto drop_unlock;
665 		}
666 
667 		x->repl->advance(x, seq);
668 
669 		x->curlft.bytes += skb->len;
670 		x->curlft.packets++;
671 
672 		spin_unlock(&x->lock);
673 
674 		XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
675 
676 		inner_mode = &x->inner_mode;
677 
678 		if (x->sel.family == AF_UNSPEC) {
679 			inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
680 			if (inner_mode == NULL) {
681 				XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
682 				goto drop;
683 			}
684 		}
685 
686 		if (xfrm_inner_mode_input(x, inner_mode, skb)) {
687 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
688 			goto drop;
689 		}
690 
691 		if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
692 			decaps = 1;
693 			break;
694 		}
695 
696 		/*
697 		 * We need the inner address.  However, we only get here for
698 		 * transport mode so the outer address is identical.
699 		 */
700 		daddr = &x->id.daddr;
701 		family = x->outer_mode.family;
702 
703 		err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
704 		if (err < 0) {
705 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
706 			goto drop;
707 		}
708 		crypto_done = false;
709 	} while (!err);
710 
711 	err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
712 	if (err)
713 		goto drop;
714 
715 	nf_reset_ct(skb);
716 
717 	if (decaps) {
718 		sp = skb_sec_path(skb);
719 		if (sp)
720 			sp->olen = 0;
721 		skb_dst_drop(skb);
722 		gro_cells_receive(&gro_cells, skb);
723 		return 0;
724 	} else {
725 		xo = xfrm_offload(skb);
726 		if (xo)
727 			xfrm_gro = xo->flags & XFRM_GRO;
728 
729 		err = -EAFNOSUPPORT;
730 		rcu_read_lock();
731 		afinfo = xfrm_state_afinfo_get_rcu(x->inner_mode.family);
732 		if (likely(afinfo))
733 			err = afinfo->transport_finish(skb, xfrm_gro || async);
734 		rcu_read_unlock();
735 		if (xfrm_gro) {
736 			sp = skb_sec_path(skb);
737 			if (sp)
738 				sp->olen = 0;
739 			skb_dst_drop(skb);
740 			gro_cells_receive(&gro_cells, skb);
741 			return err;
742 		}
743 
744 		return err;
745 	}
746 
747 drop_unlock:
748 	spin_unlock(&x->lock);
749 drop:
750 	xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
751 	kfree_skb(skb);
752 	return 0;
753 }
754 EXPORT_SYMBOL(xfrm_input);
755 
xfrm_input_resume(struct sk_buff * skb,int nexthdr)756 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
757 {
758 	return xfrm_input(skb, nexthdr, 0, -1);
759 }
760 EXPORT_SYMBOL(xfrm_input_resume);
761 
xfrm_trans_reinject(unsigned long data)762 static void xfrm_trans_reinject(unsigned long data)
763 {
764 	struct xfrm_trans_tasklet *trans = (void *)data;
765 	struct sk_buff_head queue;
766 	struct sk_buff *skb;
767 
768 	__skb_queue_head_init(&queue);
769 	skb_queue_splice_init(&trans->queue, &queue);
770 
771 	while ((skb = __skb_dequeue(&queue)))
772 		XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
773 					       NULL, skb);
774 }
775 
xfrm_trans_queue_net(struct net * net,struct sk_buff * skb,int (* finish)(struct net *,struct sock *,struct sk_buff *))776 int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
777 			 int (*finish)(struct net *, struct sock *,
778 				       struct sk_buff *))
779 {
780 	struct xfrm_trans_tasklet *trans;
781 
782 	trans = this_cpu_ptr(&xfrm_trans_tasklet);
783 
784 	if (skb_queue_len(&trans->queue) >= READ_ONCE(netdev_max_backlog))
785 		return -ENOBUFS;
786 
787 	BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
788 
789 	XFRM_TRANS_SKB_CB(skb)->finish = finish;
790 	XFRM_TRANS_SKB_CB(skb)->net = net;
791 	__skb_queue_tail(&trans->queue, skb);
792 	tasklet_schedule(&trans->tasklet);
793 	return 0;
794 }
795 EXPORT_SYMBOL(xfrm_trans_queue_net);
796 
xfrm_trans_queue(struct sk_buff * skb,int (* finish)(struct net *,struct sock *,struct sk_buff *))797 int xfrm_trans_queue(struct sk_buff *skb,
798 		     int (*finish)(struct net *, struct sock *,
799 				   struct sk_buff *))
800 {
801 	return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
802 }
803 EXPORT_SYMBOL(xfrm_trans_queue);
804 
xfrm_input_init(void)805 void __init xfrm_input_init(void)
806 {
807 	int err;
808 	int i;
809 
810 	init_dummy_netdev(&xfrm_napi_dev);
811 	err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
812 	if (err)
813 		gro_cells.cells = NULL;
814 
815 	for_each_possible_cpu(i) {
816 		struct xfrm_trans_tasklet *trans;
817 
818 		trans = &per_cpu(xfrm_trans_tasklet, i);
819 		__skb_queue_head_init(&trans->queue);
820 		tasklet_init(&trans->tasklet, xfrm_trans_reinject,
821 			     (unsigned long)trans);
822 	}
823 }
824