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 
27 struct xfrm_trans_tasklet {
28 	struct work_struct work;
29 	spinlock_t queue_lock;
30 	struct sk_buff_head queue;
31 };
32 
33 struct 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 
46 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
47 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
48 
49 static struct gro_cells gro_cells;
50 static struct net_device xfrm_napi_dev;
51 
52 static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
53 
xfrm_input_register_afinfo(const struct xfrm_input_afinfo * afinfo)54 int 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 }
69 EXPORT_SYMBOL(xfrm_input_register_afinfo);
70 
xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo * afinfo)71 int 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 }
86 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
87 
xfrm_input_get_afinfo(u8 family,bool is_ipip)88 static 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 
xfrm_rcv_cb(struct sk_buff * skb,unsigned int family,u8 protocol,int err)102 static 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 
secpath_set(struct sk_buff * skb)119 struct 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 }
137 EXPORT_SYMBOL(secpath_set);
138 
139 /* Fetch spi and seq from ipsec header */
140 
xfrm_parse_spi(struct sk_buff * skb,u8 nexthdr,__be32 * spi,__be32 * seq)141 int 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 }
174 EXPORT_SYMBOL(xfrm_parse_spi);
175 
xfrm4_remove_beet_encap(struct xfrm_state * x,struct sk_buff * skb)176 static 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;
218 out:
219 	return err;
220 }
221 
ipip_ecn_decapsulate(struct sk_buff * skb)222 static 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 
xfrm4_remove_tunnel_encap(struct xfrm_state * x,struct sk_buff * skb)230 static 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 
256 out:
257 	return err;
258 }
259 
ipip6_ecn_decapsulate(struct sk_buff * skb)260 static 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 
xfrm6_remove_tunnel_encap(struct xfrm_state * x,struct sk_buff * skb)268 static 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(ipv6_get_dsfield(ipv6_hdr(skb)),
283 			       ipipv6_hdr(skb));
284 	if (!(x->props.flags & XFRM_STATE_NOECN))
285 		ipip6_ecn_decapsulate(skb);
286 
287 	skb_reset_network_header(skb);
288 	skb_mac_header_rebuild(skb);
289 	if (skb->mac_len)
290 		eth_hdr(skb)->h_proto = skb->protocol;
291 
292 	err = 0;
293 
294 out:
295 	return err;
296 }
297 
xfrm6_remove_beet_encap(struct xfrm_state * x,struct sk_buff * skb)298 static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
299 {
300 	struct ipv6hdr *ip6h;
301 	int size = sizeof(struct ipv6hdr);
302 	int err;
303 
304 	err = skb_cow_head(skb, size + skb->mac_len);
305 	if (err)
306 		goto out;
307 
308 	__skb_push(skb, size);
309 	skb_reset_network_header(skb);
310 	skb_mac_header_rebuild(skb);
311 
312 	xfrm6_beet_make_header(skb);
313 
314 	ip6h = ipv6_hdr(skb);
315 	ip6h->payload_len = htons(skb->len - size);
316 	ip6h->daddr = x->sel.daddr.in6;
317 	ip6h->saddr = x->sel.saddr.in6;
318 	err = 0;
319 out:
320 	return err;
321 }
322 
323 /* Remove encapsulation header.
324  *
325  * The IP header will be moved over the top of the encapsulation
326  * header.
327  *
328  * On entry, the transport header shall point to where the IP header
329  * should be and the network header shall be set to where the IP
330  * header currently is.  skb->data shall point to the start of the
331  * payload.
332  */
333 static int
xfrm_inner_mode_encap_remove(struct xfrm_state * x,const struct xfrm_mode * inner_mode,struct sk_buff * skb)334 xfrm_inner_mode_encap_remove(struct xfrm_state *x,
335 			     const struct xfrm_mode *inner_mode,
336 			     struct sk_buff *skb)
337 {
338 	switch (inner_mode->encap) {
339 	case XFRM_MODE_BEET:
340 		if (inner_mode->family == AF_INET)
341 			return xfrm4_remove_beet_encap(x, skb);
342 		if (inner_mode->family == AF_INET6)
343 			return xfrm6_remove_beet_encap(x, skb);
344 		break;
345 	case XFRM_MODE_TUNNEL:
346 		if (inner_mode->family == AF_INET)
347 			return xfrm4_remove_tunnel_encap(x, skb);
348 		if (inner_mode->family == AF_INET6)
349 			return xfrm6_remove_tunnel_encap(x, skb);
350 		break;
351 	}
352 
353 	WARN_ON_ONCE(1);
354 	return -EOPNOTSUPP;
355 }
356 
xfrm_prepare_input(struct xfrm_state * x,struct sk_buff * skb)357 static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
358 {
359 	const struct xfrm_mode *inner_mode = &x->inner_mode;
360 
361 	switch (x->outer_mode.family) {
362 	case AF_INET:
363 		xfrm4_extract_header(skb);
364 		break;
365 	case AF_INET6:
366 		xfrm6_extract_header(skb);
367 		break;
368 	default:
369 		WARN_ON_ONCE(1);
370 		return -EAFNOSUPPORT;
371 	}
372 
373 	if (x->sel.family == AF_UNSPEC) {
374 		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
375 		if (!inner_mode)
376 			return -EAFNOSUPPORT;
377 	}
378 
379 	switch (inner_mode->family) {
380 	case AF_INET:
381 		skb->protocol = htons(ETH_P_IP);
382 		break;
383 	case AF_INET6:
384 		skb->protocol = htons(ETH_P_IPV6);
385 		break;
386 	default:
387 		WARN_ON_ONCE(1);
388 		break;
389 	}
390 
391 	return xfrm_inner_mode_encap_remove(x, inner_mode, skb);
392 }
393 
394 /* Remove encapsulation header.
395  *
396  * The IP header will be moved over the top of the encapsulation header.
397  *
398  * On entry, skb_transport_header() shall point to where the IP header
399  * should be and skb_network_header() shall be set to where the IP header
400  * currently is.  skb->data shall point to the start of the payload.
401  */
xfrm4_transport_input(struct xfrm_state * x,struct sk_buff * skb)402 static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
403 {
404 	int ihl = skb->data - skb_transport_header(skb);
405 
406 	if (skb->transport_header != skb->network_header) {
407 		memmove(skb_transport_header(skb),
408 			skb_network_header(skb), ihl);
409 		skb->network_header = skb->transport_header;
410 	}
411 	ip_hdr(skb)->tot_len = htons(skb->len + ihl);
412 	skb_reset_transport_header(skb);
413 	return 0;
414 }
415 
xfrm6_transport_input(struct xfrm_state * x,struct sk_buff * skb)416 static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
417 {
418 #if IS_ENABLED(CONFIG_IPV6)
419 	int ihl = skb->data - skb_transport_header(skb);
420 
421 	if (skb->transport_header != skb->network_header) {
422 		memmove(skb_transport_header(skb),
423 			skb_network_header(skb), ihl);
424 		skb->network_header = skb->transport_header;
425 	}
426 	ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
427 					   sizeof(struct ipv6hdr));
428 	skb_reset_transport_header(skb);
429 	return 0;
430 #else
431 	WARN_ON_ONCE(1);
432 	return -EAFNOSUPPORT;
433 #endif
434 }
435 
xfrm_inner_mode_input(struct xfrm_state * x,const struct xfrm_mode * inner_mode,struct sk_buff * skb)436 static int xfrm_inner_mode_input(struct xfrm_state *x,
437 				 const struct xfrm_mode *inner_mode,
438 				 struct sk_buff *skb)
439 {
440 	switch (inner_mode->encap) {
441 	case XFRM_MODE_BEET:
442 	case XFRM_MODE_TUNNEL:
443 		return xfrm_prepare_input(x, skb);
444 	case XFRM_MODE_TRANSPORT:
445 		if (inner_mode->family == AF_INET)
446 			return xfrm4_transport_input(x, skb);
447 		if (inner_mode->family == AF_INET6)
448 			return xfrm6_transport_input(x, skb);
449 		break;
450 	case XFRM_MODE_ROUTEOPTIMIZATION:
451 		WARN_ON_ONCE(1);
452 		break;
453 	default:
454 		WARN_ON_ONCE(1);
455 		break;
456 	}
457 
458 	return -EOPNOTSUPP;
459 }
460 
xfrm_input(struct sk_buff * skb,int nexthdr,__be32 spi,int encap_type)461 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
462 {
463 	const struct xfrm_state_afinfo *afinfo;
464 	struct net *net = dev_net(skb->dev);
465 	const struct xfrm_mode *inner_mode;
466 	int err;
467 	__be32 seq;
468 	__be32 seq_hi;
469 	struct xfrm_state *x = NULL;
470 	xfrm_address_t *daddr;
471 	u32 mark = skb->mark;
472 	unsigned int family = AF_UNSPEC;
473 	int decaps = 0;
474 	int async = 0;
475 	bool xfrm_gro = false;
476 	bool crypto_done = false;
477 	struct xfrm_offload *xo = xfrm_offload(skb);
478 	struct sec_path *sp;
479 
480 	if (encap_type < 0) {
481 		x = xfrm_input_state(skb);
482 
483 		if (unlikely(x->km.state != XFRM_STATE_VALID)) {
484 			if (x->km.state == XFRM_STATE_ACQ)
485 				XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
486 			else
487 				XFRM_INC_STATS(net,
488 					       LINUX_MIB_XFRMINSTATEINVALID);
489 
490 			if (encap_type == -1)
491 				dev_put(skb->dev);
492 			goto drop;
493 		}
494 
495 		family = x->outer_mode.family;
496 
497 		/* An encap_type of -1 indicates async resumption. */
498 		if (encap_type == -1) {
499 			async = 1;
500 			seq = XFRM_SKB_CB(skb)->seq.input.low;
501 			goto resume;
502 		}
503 
504 		/* encap_type < -1 indicates a GRO call. */
505 		encap_type = 0;
506 		seq = XFRM_SPI_SKB_CB(skb)->seq;
507 
508 		if (xo && (xo->flags & CRYPTO_DONE)) {
509 			crypto_done = true;
510 			family = XFRM_SPI_SKB_CB(skb)->family;
511 
512 			if (!(xo->status & CRYPTO_SUCCESS)) {
513 				if (xo->status &
514 				    (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
515 				     CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
516 				     CRYPTO_TUNNEL_AH_AUTH_FAILED |
517 				     CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
518 
519 					xfrm_audit_state_icvfail(x, skb,
520 								 x->type->proto);
521 					x->stats.integrity_failed++;
522 					XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
523 					goto drop;
524 				}
525 
526 				if (xo->status & CRYPTO_INVALID_PROTOCOL) {
527 					XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
528 					goto drop;
529 				}
530 
531 				XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
532 				goto drop;
533 			}
534 
535 			if (xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
536 				XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
537 				goto drop;
538 			}
539 		}
540 
541 		goto lock;
542 	}
543 
544 	family = XFRM_SPI_SKB_CB(skb)->family;
545 
546 	/* if tunnel is present override skb->mark value with tunnel i_key */
547 	switch (family) {
548 	case AF_INET:
549 		if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
550 			mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
551 		break;
552 	case AF_INET6:
553 		if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
554 			mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
555 		break;
556 	}
557 
558 	sp = secpath_set(skb);
559 	if (!sp) {
560 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
561 		goto drop;
562 	}
563 
564 	seq = 0;
565 	if (!spi && xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
566 		secpath_reset(skb);
567 		XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
568 		goto drop;
569 	}
570 
571 	daddr = (xfrm_address_t *)(skb_network_header(skb) +
572 				   XFRM_SPI_SKB_CB(skb)->daddroff);
573 	do {
574 		sp = skb_sec_path(skb);
575 
576 		if (sp->len == XFRM_MAX_DEPTH) {
577 			secpath_reset(skb);
578 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
579 			goto drop;
580 		}
581 
582 		x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
583 		if (x == NULL) {
584 			secpath_reset(skb);
585 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
586 			xfrm_audit_state_notfound(skb, family, spi, seq);
587 			goto drop;
588 		}
589 
590 		skb->mark = xfrm_smark_get(skb->mark, x);
591 
592 		sp->xvec[sp->len++] = x;
593 
594 		skb_dst_force(skb);
595 		if (!skb_dst(skb)) {
596 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
597 			goto drop;
598 		}
599 
600 lock:
601 		spin_lock(&x->lock);
602 
603 		if (unlikely(x->km.state != XFRM_STATE_VALID)) {
604 			if (x->km.state == XFRM_STATE_ACQ)
605 				XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
606 			else
607 				XFRM_INC_STATS(net,
608 					       LINUX_MIB_XFRMINSTATEINVALID);
609 			goto drop_unlock;
610 		}
611 
612 		if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
613 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
614 			goto drop_unlock;
615 		}
616 
617 		if (xfrm_replay_check(x, skb, seq)) {
618 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
619 			goto drop_unlock;
620 		}
621 
622 		if (xfrm_state_check_expire(x)) {
623 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
624 			goto drop_unlock;
625 		}
626 
627 		spin_unlock(&x->lock);
628 
629 		if (xfrm_tunnel_check(skb, x, family)) {
630 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
631 			goto drop;
632 		}
633 
634 		seq_hi = htonl(xfrm_replay_seqhi(x, seq));
635 
636 		XFRM_SKB_CB(skb)->seq.input.low = seq;
637 		XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
638 
639 		dev_hold(skb->dev);
640 
641 		if (crypto_done)
642 			nexthdr = x->type_offload->input_tail(x, skb);
643 		else
644 			nexthdr = x->type->input(x, skb);
645 
646 		if (nexthdr == -EINPROGRESS)
647 			return 0;
648 resume:
649 		dev_put(skb->dev);
650 
651 		spin_lock(&x->lock);
652 		if (nexthdr < 0) {
653 			if (nexthdr == -EBADMSG) {
654 				xfrm_audit_state_icvfail(x, skb,
655 							 x->type->proto);
656 				x->stats.integrity_failed++;
657 			}
658 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
659 			goto drop_unlock;
660 		}
661 
662 		/* only the first xfrm gets the encap type */
663 		encap_type = 0;
664 
665 		if (xfrm_replay_recheck(x, skb, seq)) {
666 			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
667 			goto drop_unlock;
668 		}
669 
670 		xfrm_replay_advance(x, seq);
671 
672 		x->curlft.bytes += skb->len;
673 		x->curlft.packets++;
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 
752 drop_unlock:
753 	spin_unlock(&x->lock);
754 drop:
755 	xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
756 	kfree_skb(skb);
757 	return 0;
758 }
759 EXPORT_SYMBOL(xfrm_input);
760 
xfrm_input_resume(struct sk_buff * skb,int nexthdr)761 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
762 {
763 	return xfrm_input(skb, nexthdr, 0, -1);
764 }
765 EXPORT_SYMBOL(xfrm_input_resume);
766 
xfrm_trans_reinject(struct work_struct * work)767 static 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 
xfrm_trans_queue_net(struct net * net,struct sk_buff * skb,int (* finish)(struct net *,struct sock *,struct sk_buff *))785 int 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 }
806 EXPORT_SYMBOL(xfrm_trans_queue_net);
807 
xfrm_trans_queue(struct sk_buff * skb,int (* finish)(struct net *,struct sock *,struct sk_buff *))808 int 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 }
814 EXPORT_SYMBOL(xfrm_trans_queue);
815 
xfrm_input_init(void)816 void __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 }
835