xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 867e6d38)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * xfrm_state.c
4  *
5  * Changes:
6  *	Mitsuru KANDA @USAGI
7  * 	Kazunori MIYAZAWA @USAGI
8  * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9  * 		IPv6 support
10  * 	YOSHIFUJI Hideaki @USAGI
11  * 		Split up af-specific functions
12  *	Derek Atkins <derek@ihtfp.com>
13  *		Add UDP Encapsulation
14  *
15  */
16 
17 #include <linux/workqueue.h>
18 #include <net/xfrm.h>
19 #include <linux/pfkeyv2.h>
20 #include <linux/ipsec.h>
21 #include <linux/module.h>
22 #include <linux/cache.h>
23 #include <linux/audit.h>
24 #include <linux/uaccess.h>
25 #include <linux/ktime.h>
26 #include <linux/slab.h>
27 #include <linux/interrupt.h>
28 #include <linux/kernel.h>
29 
30 #include <crypto/aead.h>
31 
32 #include "xfrm_hash.h"
33 
34 #define xfrm_state_deref_prot(table, net) \
35 	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
36 
37 static void xfrm_state_gc_task(struct work_struct *work);
38 
39 /* Each xfrm_state may be linked to two tables:
40 
41    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
42    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
43       destination/tunnel endpoint. (output)
44  */
45 
46 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
47 static struct kmem_cache *xfrm_state_cache __ro_after_init;
48 
49 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
50 static HLIST_HEAD(xfrm_state_gc_list);
51 
52 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
53 {
54 	return refcount_inc_not_zero(&x->refcnt);
55 }
56 
57 static inline unsigned int xfrm_dst_hash(struct net *net,
58 					 const xfrm_address_t *daddr,
59 					 const xfrm_address_t *saddr,
60 					 u32 reqid,
61 					 unsigned short family)
62 {
63 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
64 }
65 
66 static inline unsigned int xfrm_src_hash(struct net *net,
67 					 const xfrm_address_t *daddr,
68 					 const xfrm_address_t *saddr,
69 					 unsigned short family)
70 {
71 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
72 }
73 
74 static inline unsigned int
75 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
76 	      __be32 spi, u8 proto, unsigned short family)
77 {
78 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
79 }
80 
81 static void xfrm_hash_transfer(struct hlist_head *list,
82 			       struct hlist_head *ndsttable,
83 			       struct hlist_head *nsrctable,
84 			       struct hlist_head *nspitable,
85 			       unsigned int nhashmask)
86 {
87 	struct hlist_node *tmp;
88 	struct xfrm_state *x;
89 
90 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
91 		unsigned int h;
92 
93 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
94 				    x->props.reqid, x->props.family,
95 				    nhashmask);
96 		hlist_add_head_rcu(&x->bydst, ndsttable + h);
97 
98 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
99 				    x->props.family,
100 				    nhashmask);
101 		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
102 
103 		if (x->id.spi) {
104 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
105 					    x->id.proto, x->props.family,
106 					    nhashmask);
107 			hlist_add_head_rcu(&x->byspi, nspitable + h);
108 		}
109 	}
110 }
111 
112 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
113 {
114 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
115 }
116 
117 static void xfrm_hash_resize(struct work_struct *work)
118 {
119 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
120 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
121 	unsigned long nsize, osize;
122 	unsigned int nhashmask, ohashmask;
123 	int i;
124 
125 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
126 	ndst = xfrm_hash_alloc(nsize);
127 	if (!ndst)
128 		return;
129 	nsrc = xfrm_hash_alloc(nsize);
130 	if (!nsrc) {
131 		xfrm_hash_free(ndst, nsize);
132 		return;
133 	}
134 	nspi = xfrm_hash_alloc(nsize);
135 	if (!nspi) {
136 		xfrm_hash_free(ndst, nsize);
137 		xfrm_hash_free(nsrc, nsize);
138 		return;
139 	}
140 
141 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
142 	write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
143 
144 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
145 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
146 	for (i = net->xfrm.state_hmask; i >= 0; i--)
147 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
148 
149 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
150 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
151 	ohashmask = net->xfrm.state_hmask;
152 
153 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
154 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
155 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
156 	net->xfrm.state_hmask = nhashmask;
157 
158 	write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
159 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
160 
161 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
162 
163 	synchronize_rcu();
164 
165 	xfrm_hash_free(odst, osize);
166 	xfrm_hash_free(osrc, osize);
167 	xfrm_hash_free(ospi, osize);
168 }
169 
170 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
171 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
172 
173 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
174 
175 int __xfrm_state_delete(struct xfrm_state *x);
176 
177 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
178 static bool km_is_alive(const struct km_event *c);
179 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
180 
181 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
182 {
183 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
184 	int err = 0;
185 
186 	if (!afinfo)
187 		return -EAFNOSUPPORT;
188 
189 #define X(afi, T, name) do {			\
190 		WARN_ON((afi)->type_ ## name);	\
191 		(afi)->type_ ## name = (T);	\
192 	} while (0)
193 
194 	switch (type->proto) {
195 	case IPPROTO_COMP:
196 		X(afinfo, type, comp);
197 		break;
198 	case IPPROTO_AH:
199 		X(afinfo, type, ah);
200 		break;
201 	case IPPROTO_ESP:
202 		X(afinfo, type, esp);
203 		break;
204 	case IPPROTO_IPIP:
205 		X(afinfo, type, ipip);
206 		break;
207 	case IPPROTO_DSTOPTS:
208 		X(afinfo, type, dstopts);
209 		break;
210 	case IPPROTO_ROUTING:
211 		X(afinfo, type, routing);
212 		break;
213 	case IPPROTO_IPV6:
214 		X(afinfo, type, ipip6);
215 		break;
216 	default:
217 		WARN_ON(1);
218 		err = -EPROTONOSUPPORT;
219 		break;
220 	}
221 #undef X
222 	rcu_read_unlock();
223 	return err;
224 }
225 EXPORT_SYMBOL(xfrm_register_type);
226 
227 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
228 {
229 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
230 
231 	if (unlikely(afinfo == NULL))
232 		return;
233 
234 #define X(afi, T, name) do {				\
235 		WARN_ON((afi)->type_ ## name != (T));	\
236 		(afi)->type_ ## name = NULL;		\
237 	} while (0)
238 
239 	switch (type->proto) {
240 	case IPPROTO_COMP:
241 		X(afinfo, type, comp);
242 		break;
243 	case IPPROTO_AH:
244 		X(afinfo, type, ah);
245 		break;
246 	case IPPROTO_ESP:
247 		X(afinfo, type, esp);
248 		break;
249 	case IPPROTO_IPIP:
250 		X(afinfo, type, ipip);
251 		break;
252 	case IPPROTO_DSTOPTS:
253 		X(afinfo, type, dstopts);
254 		break;
255 	case IPPROTO_ROUTING:
256 		X(afinfo, type, routing);
257 		break;
258 	case IPPROTO_IPV6:
259 		X(afinfo, type, ipip6);
260 		break;
261 	default:
262 		WARN_ON(1);
263 		break;
264 	}
265 #undef X
266 	rcu_read_unlock();
267 }
268 EXPORT_SYMBOL(xfrm_unregister_type);
269 
270 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
271 {
272 	const struct xfrm_type *type = NULL;
273 	struct xfrm_state_afinfo *afinfo;
274 	int modload_attempted = 0;
275 
276 retry:
277 	afinfo = xfrm_state_get_afinfo(family);
278 	if (unlikely(afinfo == NULL))
279 		return NULL;
280 
281 	switch (proto) {
282 	case IPPROTO_COMP:
283 		type = afinfo->type_comp;
284 		break;
285 	case IPPROTO_AH:
286 		type = afinfo->type_ah;
287 		break;
288 	case IPPROTO_ESP:
289 		type = afinfo->type_esp;
290 		break;
291 	case IPPROTO_IPIP:
292 		type = afinfo->type_ipip;
293 		break;
294 	case IPPROTO_DSTOPTS:
295 		type = afinfo->type_dstopts;
296 		break;
297 	case IPPROTO_ROUTING:
298 		type = afinfo->type_routing;
299 		break;
300 	case IPPROTO_IPV6:
301 		type = afinfo->type_ipip6;
302 		break;
303 	default:
304 		break;
305 	}
306 
307 	if (unlikely(type && !try_module_get(type->owner)))
308 		type = NULL;
309 
310 	rcu_read_unlock();
311 
312 	if (!type && !modload_attempted) {
313 		request_module("xfrm-type-%d-%d", family, proto);
314 		modload_attempted = 1;
315 		goto retry;
316 	}
317 
318 	return type;
319 }
320 
321 static void xfrm_put_type(const struct xfrm_type *type)
322 {
323 	module_put(type->owner);
324 }
325 
326 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
327 			       unsigned short family)
328 {
329 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
330 	int err = 0;
331 
332 	if (unlikely(afinfo == NULL))
333 		return -EAFNOSUPPORT;
334 
335 	switch (type->proto) {
336 	case IPPROTO_ESP:
337 		WARN_ON(afinfo->type_offload_esp);
338 		afinfo->type_offload_esp = type;
339 		break;
340 	default:
341 		WARN_ON(1);
342 		err = -EPROTONOSUPPORT;
343 		break;
344 	}
345 
346 	rcu_read_unlock();
347 	return err;
348 }
349 EXPORT_SYMBOL(xfrm_register_type_offload);
350 
351 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
352 				  unsigned short family)
353 {
354 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
355 
356 	if (unlikely(afinfo == NULL))
357 		return;
358 
359 	switch (type->proto) {
360 	case IPPROTO_ESP:
361 		WARN_ON(afinfo->type_offload_esp != type);
362 		afinfo->type_offload_esp = NULL;
363 		break;
364 	default:
365 		WARN_ON(1);
366 		break;
367 	}
368 	rcu_read_unlock();
369 }
370 EXPORT_SYMBOL(xfrm_unregister_type_offload);
371 
372 static const struct xfrm_type_offload *
373 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
374 {
375 	const struct xfrm_type_offload *type = NULL;
376 	struct xfrm_state_afinfo *afinfo;
377 
378 retry:
379 	afinfo = xfrm_state_get_afinfo(family);
380 	if (unlikely(afinfo == NULL))
381 		return NULL;
382 
383 	switch (proto) {
384 	case IPPROTO_ESP:
385 		type = afinfo->type_offload_esp;
386 		break;
387 	default:
388 		break;
389 	}
390 
391 	if ((type && !try_module_get(type->owner)))
392 		type = NULL;
393 
394 	rcu_read_unlock();
395 
396 	if (!type && try_load) {
397 		request_module("xfrm-offload-%d-%d", family, proto);
398 		try_load = false;
399 		goto retry;
400 	}
401 
402 	return type;
403 }
404 
405 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
406 {
407 	module_put(type->owner);
408 }
409 
410 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
411 	[XFRM_MODE_BEET] = {
412 		.encap = XFRM_MODE_BEET,
413 		.flags = XFRM_MODE_FLAG_TUNNEL,
414 		.family = AF_INET,
415 	},
416 	[XFRM_MODE_TRANSPORT] = {
417 		.encap = XFRM_MODE_TRANSPORT,
418 		.family = AF_INET,
419 	},
420 	[XFRM_MODE_TUNNEL] = {
421 		.encap = XFRM_MODE_TUNNEL,
422 		.flags = XFRM_MODE_FLAG_TUNNEL,
423 		.family = AF_INET,
424 	},
425 };
426 
427 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
428 	[XFRM_MODE_BEET] = {
429 		.encap = XFRM_MODE_BEET,
430 		.flags = XFRM_MODE_FLAG_TUNNEL,
431 		.family = AF_INET6,
432 	},
433 	[XFRM_MODE_ROUTEOPTIMIZATION] = {
434 		.encap = XFRM_MODE_ROUTEOPTIMIZATION,
435 		.family = AF_INET6,
436 	},
437 	[XFRM_MODE_TRANSPORT] = {
438 		.encap = XFRM_MODE_TRANSPORT,
439 		.family = AF_INET6,
440 	},
441 	[XFRM_MODE_TUNNEL] = {
442 		.encap = XFRM_MODE_TUNNEL,
443 		.flags = XFRM_MODE_FLAG_TUNNEL,
444 		.family = AF_INET6,
445 	},
446 };
447 
448 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
449 {
450 	const struct xfrm_mode *mode;
451 
452 	if (unlikely(encap >= XFRM_MODE_MAX))
453 		return NULL;
454 
455 	switch (family) {
456 	case AF_INET:
457 		mode = &xfrm4_mode_map[encap];
458 		if (mode->family == family)
459 			return mode;
460 		break;
461 	case AF_INET6:
462 		mode = &xfrm6_mode_map[encap];
463 		if (mode->family == family)
464 			return mode;
465 		break;
466 	default:
467 		break;
468 	}
469 
470 	return NULL;
471 }
472 
473 void xfrm_state_free(struct xfrm_state *x)
474 {
475 	kmem_cache_free(xfrm_state_cache, x);
476 }
477 EXPORT_SYMBOL(xfrm_state_free);
478 
479 static void ___xfrm_state_destroy(struct xfrm_state *x)
480 {
481 	hrtimer_cancel(&x->mtimer);
482 	del_timer_sync(&x->rtimer);
483 	kfree(x->aead);
484 	kfree(x->aalg);
485 	kfree(x->ealg);
486 	kfree(x->calg);
487 	kfree(x->encap);
488 	kfree(x->coaddr);
489 	kfree(x->replay_esn);
490 	kfree(x->preplay_esn);
491 	if (x->type_offload)
492 		xfrm_put_type_offload(x->type_offload);
493 	if (x->type) {
494 		x->type->destructor(x);
495 		xfrm_put_type(x->type);
496 	}
497 	if (x->xfrag.page)
498 		put_page(x->xfrag.page);
499 	xfrm_dev_state_free(x);
500 	security_xfrm_state_free(x);
501 	xfrm_state_free(x);
502 }
503 
504 static void xfrm_state_gc_task(struct work_struct *work)
505 {
506 	struct xfrm_state *x;
507 	struct hlist_node *tmp;
508 	struct hlist_head gc_list;
509 
510 	spin_lock_bh(&xfrm_state_gc_lock);
511 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
512 	spin_unlock_bh(&xfrm_state_gc_lock);
513 
514 	synchronize_rcu();
515 
516 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
517 		___xfrm_state_destroy(x);
518 }
519 
520 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
521 {
522 	struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
523 	enum hrtimer_restart ret = HRTIMER_NORESTART;
524 	time64_t now = ktime_get_real_seconds();
525 	time64_t next = TIME64_MAX;
526 	int warn = 0;
527 	int err = 0;
528 
529 	spin_lock(&x->lock);
530 	if (x->km.state == XFRM_STATE_DEAD)
531 		goto out;
532 	if (x->km.state == XFRM_STATE_EXPIRED)
533 		goto expired;
534 	if (x->lft.hard_add_expires_seconds) {
535 		long tmo = x->lft.hard_add_expires_seconds +
536 			x->curlft.add_time - now;
537 		if (tmo <= 0) {
538 			if (x->xflags & XFRM_SOFT_EXPIRE) {
539 				/* enter hard expire without soft expire first?!
540 				 * setting a new date could trigger this.
541 				 * workaround: fix x->curflt.add_time by below:
542 				 */
543 				x->curlft.add_time = now - x->saved_tmo - 1;
544 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
545 			} else
546 				goto expired;
547 		}
548 		if (tmo < next)
549 			next = tmo;
550 	}
551 	if (x->lft.hard_use_expires_seconds) {
552 		long tmo = x->lft.hard_use_expires_seconds +
553 			(x->curlft.use_time ? : now) - now;
554 		if (tmo <= 0)
555 			goto expired;
556 		if (tmo < next)
557 			next = tmo;
558 	}
559 	if (x->km.dying)
560 		goto resched;
561 	if (x->lft.soft_add_expires_seconds) {
562 		long tmo = x->lft.soft_add_expires_seconds +
563 			x->curlft.add_time - now;
564 		if (tmo <= 0) {
565 			warn = 1;
566 			x->xflags &= ~XFRM_SOFT_EXPIRE;
567 		} else if (tmo < next) {
568 			next = tmo;
569 			x->xflags |= XFRM_SOFT_EXPIRE;
570 			x->saved_tmo = tmo;
571 		}
572 	}
573 	if (x->lft.soft_use_expires_seconds) {
574 		long tmo = x->lft.soft_use_expires_seconds +
575 			(x->curlft.use_time ? : now) - now;
576 		if (tmo <= 0)
577 			warn = 1;
578 		else if (tmo < next)
579 			next = tmo;
580 	}
581 
582 	x->km.dying = warn;
583 	if (warn)
584 		km_state_expired(x, 0, 0);
585 resched:
586 	if (next != TIME64_MAX) {
587 		hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
588 		ret = HRTIMER_RESTART;
589 	}
590 
591 	goto out;
592 
593 expired:
594 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
595 		x->km.state = XFRM_STATE_EXPIRED;
596 
597 	err = __xfrm_state_delete(x);
598 	if (!err)
599 		km_state_expired(x, 1, 0);
600 
601 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
602 
603 out:
604 	spin_unlock(&x->lock);
605 	return ret;
606 }
607 
608 static void xfrm_replay_timer_handler(struct timer_list *t);
609 
610 struct xfrm_state *xfrm_state_alloc(struct net *net)
611 {
612 	struct xfrm_state *x;
613 
614 	x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
615 
616 	if (x) {
617 		write_pnet(&x->xs_net, net);
618 		refcount_set(&x->refcnt, 1);
619 		atomic_set(&x->tunnel_users, 0);
620 		INIT_LIST_HEAD(&x->km.all);
621 		INIT_HLIST_NODE(&x->bydst);
622 		INIT_HLIST_NODE(&x->bysrc);
623 		INIT_HLIST_NODE(&x->byspi);
624 		hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
625 		x->mtimer.function = xfrm_timer_handler;
626 		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
627 		x->curlft.add_time = ktime_get_real_seconds();
628 		x->lft.soft_byte_limit = XFRM_INF;
629 		x->lft.soft_packet_limit = XFRM_INF;
630 		x->lft.hard_byte_limit = XFRM_INF;
631 		x->lft.hard_packet_limit = XFRM_INF;
632 		x->replay_maxage = 0;
633 		x->replay_maxdiff = 0;
634 		spin_lock_init(&x->lock);
635 	}
636 	return x;
637 }
638 EXPORT_SYMBOL(xfrm_state_alloc);
639 
640 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
641 {
642 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
643 
644 	if (sync) {
645 		synchronize_rcu();
646 		___xfrm_state_destroy(x);
647 	} else {
648 		spin_lock_bh(&xfrm_state_gc_lock);
649 		hlist_add_head(&x->gclist, &xfrm_state_gc_list);
650 		spin_unlock_bh(&xfrm_state_gc_lock);
651 		schedule_work(&xfrm_state_gc_work);
652 	}
653 }
654 EXPORT_SYMBOL(__xfrm_state_destroy);
655 
656 int __xfrm_state_delete(struct xfrm_state *x)
657 {
658 	struct net *net = xs_net(x);
659 	int err = -ESRCH;
660 
661 	if (x->km.state != XFRM_STATE_DEAD) {
662 		x->km.state = XFRM_STATE_DEAD;
663 		spin_lock(&net->xfrm.xfrm_state_lock);
664 		list_del(&x->km.all);
665 		hlist_del_rcu(&x->bydst);
666 		hlist_del_rcu(&x->bysrc);
667 		if (x->id.spi)
668 			hlist_del_rcu(&x->byspi);
669 		net->xfrm.state_num--;
670 		spin_unlock(&net->xfrm.xfrm_state_lock);
671 
672 		if (x->encap_sk)
673 			sock_put(rcu_dereference_raw(x->encap_sk));
674 
675 		xfrm_dev_state_delete(x);
676 
677 		/* All xfrm_state objects are created by xfrm_state_alloc.
678 		 * The xfrm_state_alloc call gives a reference, and that
679 		 * is what we are dropping here.
680 		 */
681 		xfrm_state_put(x);
682 		err = 0;
683 	}
684 
685 	return err;
686 }
687 EXPORT_SYMBOL(__xfrm_state_delete);
688 
689 int xfrm_state_delete(struct xfrm_state *x)
690 {
691 	int err;
692 
693 	spin_lock_bh(&x->lock);
694 	err = __xfrm_state_delete(x);
695 	spin_unlock_bh(&x->lock);
696 
697 	return err;
698 }
699 EXPORT_SYMBOL(xfrm_state_delete);
700 
701 #ifdef CONFIG_SECURITY_NETWORK_XFRM
702 static inline int
703 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
704 {
705 	int i, err = 0;
706 
707 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
708 		struct xfrm_state *x;
709 
710 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
711 			if (xfrm_id_proto_match(x->id.proto, proto) &&
712 			   (err = security_xfrm_state_delete(x)) != 0) {
713 				xfrm_audit_state_delete(x, 0, task_valid);
714 				return err;
715 			}
716 		}
717 	}
718 
719 	return err;
720 }
721 
722 static inline int
723 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
724 {
725 	int i, err = 0;
726 
727 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
728 		struct xfrm_state *x;
729 		struct xfrm_state_offload *xso;
730 
731 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
732 			xso = &x->xso;
733 
734 			if (xso->dev == dev &&
735 			   (err = security_xfrm_state_delete(x)) != 0) {
736 				xfrm_audit_state_delete(x, 0, task_valid);
737 				return err;
738 			}
739 		}
740 	}
741 
742 	return err;
743 }
744 #else
745 static inline int
746 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
747 {
748 	return 0;
749 }
750 
751 static inline int
752 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
753 {
754 	return 0;
755 }
756 #endif
757 
758 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
759 {
760 	int i, err = 0, cnt = 0;
761 
762 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
763 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
764 	if (err)
765 		goto out;
766 
767 	err = -ESRCH;
768 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
769 		struct xfrm_state *x;
770 restart:
771 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
772 			if (!xfrm_state_kern(x) &&
773 			    xfrm_id_proto_match(x->id.proto, proto)) {
774 				xfrm_state_hold(x);
775 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
776 
777 				err = xfrm_state_delete(x);
778 				xfrm_audit_state_delete(x, err ? 0 : 1,
779 							task_valid);
780 				if (sync)
781 					xfrm_state_put_sync(x);
782 				else
783 					xfrm_state_put(x);
784 				if (!err)
785 					cnt++;
786 
787 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
788 				goto restart;
789 			}
790 		}
791 	}
792 out:
793 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
794 	if (cnt)
795 		err = 0;
796 
797 	return err;
798 }
799 EXPORT_SYMBOL(xfrm_state_flush);
800 
801 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
802 {
803 	int i, err = 0, cnt = 0;
804 
805 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
806 	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
807 	if (err)
808 		goto out;
809 
810 	err = -ESRCH;
811 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
812 		struct xfrm_state *x;
813 		struct xfrm_state_offload *xso;
814 restart:
815 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
816 			xso = &x->xso;
817 
818 			if (!xfrm_state_kern(x) && xso->dev == dev) {
819 				xfrm_state_hold(x);
820 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
821 
822 				err = xfrm_state_delete(x);
823 				xfrm_audit_state_delete(x, err ? 0 : 1,
824 							task_valid);
825 				xfrm_state_put(x);
826 				if (!err)
827 					cnt++;
828 
829 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
830 				goto restart;
831 			}
832 		}
833 	}
834 	if (cnt)
835 		err = 0;
836 
837 out:
838 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
839 	return err;
840 }
841 EXPORT_SYMBOL(xfrm_dev_state_flush);
842 
843 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
844 {
845 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
846 	si->sadcnt = net->xfrm.state_num;
847 	si->sadhcnt = net->xfrm.state_hmask + 1;
848 	si->sadhmcnt = xfrm_state_hashmax;
849 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
850 }
851 EXPORT_SYMBOL(xfrm_sad_getinfo);
852 
853 static void
854 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
855 {
856 	const struct flowi4 *fl4 = &fl->u.ip4;
857 
858 	sel->daddr.a4 = fl4->daddr;
859 	sel->saddr.a4 = fl4->saddr;
860 	sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
861 	sel->dport_mask = htons(0xffff);
862 	sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
863 	sel->sport_mask = htons(0xffff);
864 	sel->family = AF_INET;
865 	sel->prefixlen_d = 32;
866 	sel->prefixlen_s = 32;
867 	sel->proto = fl4->flowi4_proto;
868 	sel->ifindex = fl4->flowi4_oif;
869 }
870 
871 static void
872 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
873 {
874 	const struct flowi6 *fl6 = &fl->u.ip6;
875 
876 	/* Initialize temporary selector matching only to current session. */
877 	*(struct in6_addr *)&sel->daddr = fl6->daddr;
878 	*(struct in6_addr *)&sel->saddr = fl6->saddr;
879 	sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
880 	sel->dport_mask = htons(0xffff);
881 	sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
882 	sel->sport_mask = htons(0xffff);
883 	sel->family = AF_INET6;
884 	sel->prefixlen_d = 128;
885 	sel->prefixlen_s = 128;
886 	sel->proto = fl6->flowi6_proto;
887 	sel->ifindex = fl6->flowi6_oif;
888 }
889 
890 static void
891 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
892 		    const struct xfrm_tmpl *tmpl,
893 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
894 		    unsigned short family)
895 {
896 	switch (family) {
897 	case AF_INET:
898 		__xfrm4_init_tempsel(&x->sel, fl);
899 		break;
900 	case AF_INET6:
901 		__xfrm6_init_tempsel(&x->sel, fl);
902 		break;
903 	}
904 
905 	x->id = tmpl->id;
906 
907 	switch (tmpl->encap_family) {
908 	case AF_INET:
909 		if (x->id.daddr.a4 == 0)
910 			x->id.daddr.a4 = daddr->a4;
911 		x->props.saddr = tmpl->saddr;
912 		if (x->props.saddr.a4 == 0)
913 			x->props.saddr.a4 = saddr->a4;
914 		break;
915 	case AF_INET6:
916 		if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
917 			memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
918 		memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
919 		if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
920 			memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
921 		break;
922 	}
923 
924 	x->props.mode = tmpl->mode;
925 	x->props.reqid = tmpl->reqid;
926 	x->props.family = tmpl->encap_family;
927 }
928 
929 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
930 					      const xfrm_address_t *daddr,
931 					      __be32 spi, u8 proto,
932 					      unsigned short family)
933 {
934 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
935 	struct xfrm_state *x;
936 
937 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
938 		if (x->props.family != family ||
939 		    x->id.spi       != spi ||
940 		    x->id.proto     != proto ||
941 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
942 			continue;
943 
944 		if ((mark & x->mark.m) != x->mark.v)
945 			continue;
946 		if (!xfrm_state_hold_rcu(x))
947 			continue;
948 		return x;
949 	}
950 
951 	return NULL;
952 }
953 
954 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
955 						     const xfrm_address_t *daddr,
956 						     const xfrm_address_t *saddr,
957 						     u8 proto, unsigned short family)
958 {
959 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
960 	struct xfrm_state *x;
961 
962 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
963 		if (x->props.family != family ||
964 		    x->id.proto     != proto ||
965 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
966 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
967 			continue;
968 
969 		if ((mark & x->mark.m) != x->mark.v)
970 			continue;
971 		if (!xfrm_state_hold_rcu(x))
972 			continue;
973 		return x;
974 	}
975 
976 	return NULL;
977 }
978 
979 static inline struct xfrm_state *
980 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
981 {
982 	struct net *net = xs_net(x);
983 	u32 mark = x->mark.v & x->mark.m;
984 
985 	if (use_spi)
986 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
987 					   x->id.spi, x->id.proto, family);
988 	else
989 		return __xfrm_state_lookup_byaddr(net, mark,
990 						  &x->id.daddr,
991 						  &x->props.saddr,
992 						  x->id.proto, family);
993 }
994 
995 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
996 {
997 	if (have_hash_collision &&
998 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
999 	    net->xfrm.state_num > net->xfrm.state_hmask)
1000 		schedule_work(&net->xfrm.state_hash_work);
1001 }
1002 
1003 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1004 			       const struct flowi *fl, unsigned short family,
1005 			       struct xfrm_state **best, int *acq_in_progress,
1006 			       int *error)
1007 {
1008 	/* Resolution logic:
1009 	 * 1. There is a valid state with matching selector. Done.
1010 	 * 2. Valid state with inappropriate selector. Skip.
1011 	 *
1012 	 * Entering area of "sysdeps".
1013 	 *
1014 	 * 3. If state is not valid, selector is temporary, it selects
1015 	 *    only session which triggered previous resolution. Key
1016 	 *    manager will do something to install a state with proper
1017 	 *    selector.
1018 	 */
1019 	if (x->km.state == XFRM_STATE_VALID) {
1020 		if ((x->sel.family &&
1021 		     (x->sel.family != family ||
1022 		      !xfrm_selector_match(&x->sel, fl, family))) ||
1023 		    !security_xfrm_state_pol_flow_match(x, pol,
1024 							&fl->u.__fl_common))
1025 			return;
1026 
1027 		if (!*best ||
1028 		    (*best)->km.dying > x->km.dying ||
1029 		    ((*best)->km.dying == x->km.dying &&
1030 		     (*best)->curlft.add_time < x->curlft.add_time))
1031 			*best = x;
1032 	} else if (x->km.state == XFRM_STATE_ACQ) {
1033 		*acq_in_progress = 1;
1034 	} else if (x->km.state == XFRM_STATE_ERROR ||
1035 		   x->km.state == XFRM_STATE_EXPIRED) {
1036 		if ((!x->sel.family ||
1037 		     (x->sel.family == family &&
1038 		      xfrm_selector_match(&x->sel, fl, family))) &&
1039 		    security_xfrm_state_pol_flow_match(x, pol,
1040 						       &fl->u.__fl_common))
1041 			*error = -ESRCH;
1042 	}
1043 }
1044 
1045 struct xfrm_state *
1046 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1047 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1048 		struct xfrm_policy *pol, int *err,
1049 		unsigned short family, u32 if_id)
1050 {
1051 	static xfrm_address_t saddr_wildcard = { };
1052 	struct net *net = xp_net(pol);
1053 	unsigned int h, h_wildcard;
1054 	struct xfrm_state *x, *x0, *to_put;
1055 	int acquire_in_progress = 0;
1056 	int error = 0;
1057 	struct xfrm_state *best = NULL;
1058 	u32 mark = pol->mark.v & pol->mark.m;
1059 	unsigned short encap_family = tmpl->encap_family;
1060 	unsigned int sequence;
1061 	struct km_event c;
1062 
1063 	to_put = NULL;
1064 
1065 	sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1066 
1067 	rcu_read_lock();
1068 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1069 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1070 		if (x->props.family == encap_family &&
1071 		    x->props.reqid == tmpl->reqid &&
1072 		    (mark & x->mark.m) == x->mark.v &&
1073 		    x->if_id == if_id &&
1074 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1075 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1076 		    tmpl->mode == x->props.mode &&
1077 		    tmpl->id.proto == x->id.proto &&
1078 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1079 			xfrm_state_look_at(pol, x, fl, family,
1080 					   &best, &acquire_in_progress, &error);
1081 	}
1082 	if (best || acquire_in_progress)
1083 		goto found;
1084 
1085 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1086 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1087 		if (x->props.family == encap_family &&
1088 		    x->props.reqid == tmpl->reqid &&
1089 		    (mark & x->mark.m) == x->mark.v &&
1090 		    x->if_id == if_id &&
1091 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1092 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1093 		    tmpl->mode == x->props.mode &&
1094 		    tmpl->id.proto == x->id.proto &&
1095 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1096 			xfrm_state_look_at(pol, x, fl, family,
1097 					   &best, &acquire_in_progress, &error);
1098 	}
1099 
1100 found:
1101 	x = best;
1102 	if (!x && !error && !acquire_in_progress) {
1103 		if (tmpl->id.spi &&
1104 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1105 					      tmpl->id.proto, encap_family)) != NULL) {
1106 			to_put = x0;
1107 			error = -EEXIST;
1108 			goto out;
1109 		}
1110 
1111 		c.net = net;
1112 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1113 		 * for each and every packet - garbage collection might not
1114 		 * handle the flood.
1115 		 */
1116 		if (!km_is_alive(&c)) {
1117 			error = -ESRCH;
1118 			goto out;
1119 		}
1120 
1121 		x = xfrm_state_alloc(net);
1122 		if (x == NULL) {
1123 			error = -ENOMEM;
1124 			goto out;
1125 		}
1126 		/* Initialize temporary state matching only
1127 		 * to current session. */
1128 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1129 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1130 		x->if_id = if_id;
1131 
1132 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1133 		if (error) {
1134 			x->km.state = XFRM_STATE_DEAD;
1135 			to_put = x;
1136 			x = NULL;
1137 			goto out;
1138 		}
1139 
1140 		if (km_query(x, tmpl, pol) == 0) {
1141 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1142 			x->km.state = XFRM_STATE_ACQ;
1143 			list_add(&x->km.all, &net->xfrm.state_all);
1144 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1145 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1146 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1147 			if (x->id.spi) {
1148 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1149 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1150 			}
1151 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1152 			hrtimer_start(&x->mtimer,
1153 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1154 				      HRTIMER_MODE_REL_SOFT);
1155 			net->xfrm.state_num++;
1156 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1157 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1158 		} else {
1159 			x->km.state = XFRM_STATE_DEAD;
1160 			to_put = x;
1161 			x = NULL;
1162 			error = -ESRCH;
1163 		}
1164 	}
1165 out:
1166 	if (x) {
1167 		if (!xfrm_state_hold_rcu(x)) {
1168 			*err = -EAGAIN;
1169 			x = NULL;
1170 		}
1171 	} else {
1172 		*err = acquire_in_progress ? -EAGAIN : error;
1173 	}
1174 	rcu_read_unlock();
1175 	if (to_put)
1176 		xfrm_state_put(to_put);
1177 
1178 	if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1179 		*err = -EAGAIN;
1180 		if (x) {
1181 			xfrm_state_put(x);
1182 			x = NULL;
1183 		}
1184 	}
1185 
1186 	return x;
1187 }
1188 
1189 struct xfrm_state *
1190 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1191 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1192 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1193 {
1194 	unsigned int h;
1195 	struct xfrm_state *rx = NULL, *x = NULL;
1196 
1197 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1198 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1199 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1200 		if (x->props.family == family &&
1201 		    x->props.reqid == reqid &&
1202 		    (mark & x->mark.m) == x->mark.v &&
1203 		    x->if_id == if_id &&
1204 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1205 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1206 		    mode == x->props.mode &&
1207 		    proto == x->id.proto &&
1208 		    x->km.state == XFRM_STATE_VALID) {
1209 			rx = x;
1210 			break;
1211 		}
1212 	}
1213 
1214 	if (rx)
1215 		xfrm_state_hold(rx);
1216 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1217 
1218 
1219 	return rx;
1220 }
1221 EXPORT_SYMBOL(xfrm_stateonly_find);
1222 
1223 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1224 					      unsigned short family)
1225 {
1226 	struct xfrm_state *x;
1227 	struct xfrm_state_walk *w;
1228 
1229 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1230 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1231 		x = container_of(w, struct xfrm_state, km);
1232 		if (x->props.family != family ||
1233 			x->id.spi != spi)
1234 			continue;
1235 
1236 		xfrm_state_hold(x);
1237 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1238 		return x;
1239 	}
1240 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1241 	return NULL;
1242 }
1243 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1244 
1245 static void __xfrm_state_insert(struct xfrm_state *x)
1246 {
1247 	struct net *net = xs_net(x);
1248 	unsigned int h;
1249 
1250 	list_add(&x->km.all, &net->xfrm.state_all);
1251 
1252 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1253 			  x->props.reqid, x->props.family);
1254 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1255 
1256 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1257 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1258 
1259 	if (x->id.spi) {
1260 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1261 				  x->props.family);
1262 
1263 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1264 	}
1265 
1266 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1267 	if (x->replay_maxage)
1268 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1269 
1270 	net->xfrm.state_num++;
1271 
1272 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1273 }
1274 
1275 /* net->xfrm.xfrm_state_lock is held */
1276 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1277 {
1278 	struct net *net = xs_net(xnew);
1279 	unsigned short family = xnew->props.family;
1280 	u32 reqid = xnew->props.reqid;
1281 	struct xfrm_state *x;
1282 	unsigned int h;
1283 	u32 mark = xnew->mark.v & xnew->mark.m;
1284 	u32 if_id = xnew->if_id;
1285 
1286 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1287 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1288 		if (x->props.family	== family &&
1289 		    x->props.reqid	== reqid &&
1290 		    x->if_id		== if_id &&
1291 		    (mark & x->mark.m) == x->mark.v &&
1292 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1293 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1294 			x->genid++;
1295 	}
1296 }
1297 
1298 void xfrm_state_insert(struct xfrm_state *x)
1299 {
1300 	struct net *net = xs_net(x);
1301 
1302 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1303 	__xfrm_state_bump_genids(x);
1304 	__xfrm_state_insert(x);
1305 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1306 }
1307 EXPORT_SYMBOL(xfrm_state_insert);
1308 
1309 /* net->xfrm.xfrm_state_lock is held */
1310 static struct xfrm_state *__find_acq_core(struct net *net,
1311 					  const struct xfrm_mark *m,
1312 					  unsigned short family, u8 mode,
1313 					  u32 reqid, u32 if_id, u8 proto,
1314 					  const xfrm_address_t *daddr,
1315 					  const xfrm_address_t *saddr,
1316 					  int create)
1317 {
1318 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1319 	struct xfrm_state *x;
1320 	u32 mark = m->v & m->m;
1321 
1322 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1323 		if (x->props.reqid  != reqid ||
1324 		    x->props.mode   != mode ||
1325 		    x->props.family != family ||
1326 		    x->km.state     != XFRM_STATE_ACQ ||
1327 		    x->id.spi       != 0 ||
1328 		    x->id.proto	    != proto ||
1329 		    (mark & x->mark.m) != x->mark.v ||
1330 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1331 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1332 			continue;
1333 
1334 		xfrm_state_hold(x);
1335 		return x;
1336 	}
1337 
1338 	if (!create)
1339 		return NULL;
1340 
1341 	x = xfrm_state_alloc(net);
1342 	if (likely(x)) {
1343 		switch (family) {
1344 		case AF_INET:
1345 			x->sel.daddr.a4 = daddr->a4;
1346 			x->sel.saddr.a4 = saddr->a4;
1347 			x->sel.prefixlen_d = 32;
1348 			x->sel.prefixlen_s = 32;
1349 			x->props.saddr.a4 = saddr->a4;
1350 			x->id.daddr.a4 = daddr->a4;
1351 			break;
1352 
1353 		case AF_INET6:
1354 			x->sel.daddr.in6 = daddr->in6;
1355 			x->sel.saddr.in6 = saddr->in6;
1356 			x->sel.prefixlen_d = 128;
1357 			x->sel.prefixlen_s = 128;
1358 			x->props.saddr.in6 = saddr->in6;
1359 			x->id.daddr.in6 = daddr->in6;
1360 			break;
1361 		}
1362 
1363 		x->km.state = XFRM_STATE_ACQ;
1364 		x->id.proto = proto;
1365 		x->props.family = family;
1366 		x->props.mode = mode;
1367 		x->props.reqid = reqid;
1368 		x->if_id = if_id;
1369 		x->mark.v = m->v;
1370 		x->mark.m = m->m;
1371 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1372 		xfrm_state_hold(x);
1373 		hrtimer_start(&x->mtimer,
1374 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1375 			      HRTIMER_MODE_REL_SOFT);
1376 		list_add(&x->km.all, &net->xfrm.state_all);
1377 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1378 		h = xfrm_src_hash(net, daddr, saddr, family);
1379 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1380 
1381 		net->xfrm.state_num++;
1382 
1383 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1384 	}
1385 
1386 	return x;
1387 }
1388 
1389 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1390 
1391 int xfrm_state_add(struct xfrm_state *x)
1392 {
1393 	struct net *net = xs_net(x);
1394 	struct xfrm_state *x1, *to_put;
1395 	int family;
1396 	int err;
1397 	u32 mark = x->mark.v & x->mark.m;
1398 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1399 
1400 	family = x->props.family;
1401 
1402 	to_put = NULL;
1403 
1404 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1405 
1406 	x1 = __xfrm_state_locate(x, use_spi, family);
1407 	if (x1) {
1408 		to_put = x1;
1409 		x1 = NULL;
1410 		err = -EEXIST;
1411 		goto out;
1412 	}
1413 
1414 	if (use_spi && x->km.seq) {
1415 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1416 		if (x1 && ((x1->id.proto != x->id.proto) ||
1417 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1418 			to_put = x1;
1419 			x1 = NULL;
1420 		}
1421 	}
1422 
1423 	if (use_spi && !x1)
1424 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1425 				     x->props.reqid, x->if_id, x->id.proto,
1426 				     &x->id.daddr, &x->props.saddr, 0);
1427 
1428 	__xfrm_state_bump_genids(x);
1429 	__xfrm_state_insert(x);
1430 	err = 0;
1431 
1432 out:
1433 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1434 
1435 	if (x1) {
1436 		xfrm_state_delete(x1);
1437 		xfrm_state_put(x1);
1438 	}
1439 
1440 	if (to_put)
1441 		xfrm_state_put(to_put);
1442 
1443 	return err;
1444 }
1445 EXPORT_SYMBOL(xfrm_state_add);
1446 
1447 #ifdef CONFIG_XFRM_MIGRATE
1448 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1449 {
1450 	struct xfrm_user_sec_ctx *uctx;
1451 	int size = sizeof(*uctx) + security->ctx_len;
1452 	int err;
1453 
1454 	uctx = kmalloc(size, GFP_KERNEL);
1455 	if (!uctx)
1456 		return -ENOMEM;
1457 
1458 	uctx->exttype = XFRMA_SEC_CTX;
1459 	uctx->len = size;
1460 	uctx->ctx_doi = security->ctx_doi;
1461 	uctx->ctx_alg = security->ctx_alg;
1462 	uctx->ctx_len = security->ctx_len;
1463 	memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1464 	err = security_xfrm_state_alloc(x, uctx);
1465 	kfree(uctx);
1466 	if (err)
1467 		return err;
1468 
1469 	return 0;
1470 }
1471 
1472 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1473 					   struct xfrm_encap_tmpl *encap)
1474 {
1475 	struct net *net = xs_net(orig);
1476 	struct xfrm_state *x = xfrm_state_alloc(net);
1477 	if (!x)
1478 		goto out;
1479 
1480 	memcpy(&x->id, &orig->id, sizeof(x->id));
1481 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1482 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1483 	x->props.mode = orig->props.mode;
1484 	x->props.replay_window = orig->props.replay_window;
1485 	x->props.reqid = orig->props.reqid;
1486 	x->props.family = orig->props.family;
1487 	x->props.saddr = orig->props.saddr;
1488 
1489 	if (orig->aalg) {
1490 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1491 		if (!x->aalg)
1492 			goto error;
1493 	}
1494 	x->props.aalgo = orig->props.aalgo;
1495 
1496 	if (orig->aead) {
1497 		x->aead = xfrm_algo_aead_clone(orig->aead);
1498 		x->geniv = orig->geniv;
1499 		if (!x->aead)
1500 			goto error;
1501 	}
1502 	if (orig->ealg) {
1503 		x->ealg = xfrm_algo_clone(orig->ealg);
1504 		if (!x->ealg)
1505 			goto error;
1506 	}
1507 	x->props.ealgo = orig->props.ealgo;
1508 
1509 	if (orig->calg) {
1510 		x->calg = xfrm_algo_clone(orig->calg);
1511 		if (!x->calg)
1512 			goto error;
1513 	}
1514 	x->props.calgo = orig->props.calgo;
1515 
1516 	if (encap || orig->encap) {
1517 		if (encap)
1518 			x->encap = kmemdup(encap, sizeof(*x->encap),
1519 					GFP_KERNEL);
1520 		else
1521 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1522 					GFP_KERNEL);
1523 
1524 		if (!x->encap)
1525 			goto error;
1526 	}
1527 
1528 	if (orig->security)
1529 		if (clone_security(x, orig->security))
1530 			goto error;
1531 
1532 	if (orig->coaddr) {
1533 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1534 				    GFP_KERNEL);
1535 		if (!x->coaddr)
1536 			goto error;
1537 	}
1538 
1539 	if (orig->replay_esn) {
1540 		if (xfrm_replay_clone(x, orig))
1541 			goto error;
1542 	}
1543 
1544 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1545 	memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1546 
1547 	if (xfrm_init_state(x) < 0)
1548 		goto error;
1549 
1550 	x->props.flags = orig->props.flags;
1551 	x->props.extra_flags = orig->props.extra_flags;
1552 
1553 	x->if_id = orig->if_id;
1554 	x->tfcpad = orig->tfcpad;
1555 	x->replay_maxdiff = orig->replay_maxdiff;
1556 	x->replay_maxage = orig->replay_maxage;
1557 	memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1558 	x->km.state = orig->km.state;
1559 	x->km.seq = orig->km.seq;
1560 	x->replay = orig->replay;
1561 	x->preplay = orig->preplay;
1562 
1563 	return x;
1564 
1565  error:
1566 	xfrm_state_put(x);
1567 out:
1568 	return NULL;
1569 }
1570 
1571 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1572 {
1573 	unsigned int h;
1574 	struct xfrm_state *x = NULL;
1575 
1576 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1577 
1578 	if (m->reqid) {
1579 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1580 				  m->reqid, m->old_family);
1581 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1582 			if (x->props.mode != m->mode ||
1583 			    x->id.proto != m->proto)
1584 				continue;
1585 			if (m->reqid && x->props.reqid != m->reqid)
1586 				continue;
1587 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1588 					     m->old_family) ||
1589 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1590 					     m->old_family))
1591 				continue;
1592 			xfrm_state_hold(x);
1593 			break;
1594 		}
1595 	} else {
1596 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1597 				  m->old_family);
1598 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1599 			if (x->props.mode != m->mode ||
1600 			    x->id.proto != m->proto)
1601 				continue;
1602 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1603 					     m->old_family) ||
1604 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1605 					     m->old_family))
1606 				continue;
1607 			xfrm_state_hold(x);
1608 			break;
1609 		}
1610 	}
1611 
1612 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1613 
1614 	return x;
1615 }
1616 EXPORT_SYMBOL(xfrm_migrate_state_find);
1617 
1618 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1619 				      struct xfrm_migrate *m,
1620 				      struct xfrm_encap_tmpl *encap)
1621 {
1622 	struct xfrm_state *xc;
1623 
1624 	xc = xfrm_state_clone(x, encap);
1625 	if (!xc)
1626 		return NULL;
1627 
1628 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1629 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1630 
1631 	/* add state */
1632 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1633 		/* a care is needed when the destination address of the
1634 		   state is to be updated as it is a part of triplet */
1635 		xfrm_state_insert(xc);
1636 	} else {
1637 		if (xfrm_state_add(xc) < 0)
1638 			goto error;
1639 	}
1640 
1641 	return xc;
1642 error:
1643 	xfrm_state_put(xc);
1644 	return NULL;
1645 }
1646 EXPORT_SYMBOL(xfrm_state_migrate);
1647 #endif
1648 
1649 int xfrm_state_update(struct xfrm_state *x)
1650 {
1651 	struct xfrm_state *x1, *to_put;
1652 	int err;
1653 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1654 	struct net *net = xs_net(x);
1655 
1656 	to_put = NULL;
1657 
1658 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1659 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1660 
1661 	err = -ESRCH;
1662 	if (!x1)
1663 		goto out;
1664 
1665 	if (xfrm_state_kern(x1)) {
1666 		to_put = x1;
1667 		err = -EEXIST;
1668 		goto out;
1669 	}
1670 
1671 	if (x1->km.state == XFRM_STATE_ACQ) {
1672 		__xfrm_state_insert(x);
1673 		x = NULL;
1674 	}
1675 	err = 0;
1676 
1677 out:
1678 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1679 
1680 	if (to_put)
1681 		xfrm_state_put(to_put);
1682 
1683 	if (err)
1684 		return err;
1685 
1686 	if (!x) {
1687 		xfrm_state_delete(x1);
1688 		xfrm_state_put(x1);
1689 		return 0;
1690 	}
1691 
1692 	err = -EINVAL;
1693 	spin_lock_bh(&x1->lock);
1694 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1695 		if (x->encap && x1->encap &&
1696 		    x->encap->encap_type == x1->encap->encap_type)
1697 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1698 		else if (x->encap || x1->encap)
1699 			goto fail;
1700 
1701 		if (x->coaddr && x1->coaddr) {
1702 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1703 		}
1704 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1705 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1706 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1707 		x1->km.dying = 0;
1708 
1709 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1710 			      HRTIMER_MODE_REL_SOFT);
1711 		if (x1->curlft.use_time)
1712 			xfrm_state_check_expire(x1);
1713 
1714 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1715 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1716 
1717 			if (x->props.smark.m || x->props.smark.v)
1718 				x1->props.smark = x->props.smark;
1719 
1720 			if (x->if_id)
1721 				x1->if_id = x->if_id;
1722 
1723 			__xfrm_state_bump_genids(x1);
1724 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1725 		}
1726 
1727 		err = 0;
1728 		x->km.state = XFRM_STATE_DEAD;
1729 		__xfrm_state_put(x);
1730 	}
1731 
1732 fail:
1733 	spin_unlock_bh(&x1->lock);
1734 
1735 	xfrm_state_put(x1);
1736 
1737 	return err;
1738 }
1739 EXPORT_SYMBOL(xfrm_state_update);
1740 
1741 int xfrm_state_check_expire(struct xfrm_state *x)
1742 {
1743 	if (!x->curlft.use_time)
1744 		x->curlft.use_time = ktime_get_real_seconds();
1745 
1746 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1747 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1748 		x->km.state = XFRM_STATE_EXPIRED;
1749 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1750 		return -EINVAL;
1751 	}
1752 
1753 	if (!x->km.dying &&
1754 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1755 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1756 		x->km.dying = 1;
1757 		km_state_expired(x, 0, 0);
1758 	}
1759 	return 0;
1760 }
1761 EXPORT_SYMBOL(xfrm_state_check_expire);
1762 
1763 struct xfrm_state *
1764 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1765 		  u8 proto, unsigned short family)
1766 {
1767 	struct xfrm_state *x;
1768 
1769 	rcu_read_lock();
1770 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1771 	rcu_read_unlock();
1772 	return x;
1773 }
1774 EXPORT_SYMBOL(xfrm_state_lookup);
1775 
1776 struct xfrm_state *
1777 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1778 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1779 			 u8 proto, unsigned short family)
1780 {
1781 	struct xfrm_state *x;
1782 
1783 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1784 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1785 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1786 	return x;
1787 }
1788 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1789 
1790 struct xfrm_state *
1791 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1792 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1793 	      const xfrm_address_t *saddr, int create, unsigned short family)
1794 {
1795 	struct xfrm_state *x;
1796 
1797 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1798 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1799 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1800 
1801 	return x;
1802 }
1803 EXPORT_SYMBOL(xfrm_find_acq);
1804 
1805 #ifdef CONFIG_XFRM_SUB_POLICY
1806 #if IS_ENABLED(CONFIG_IPV6)
1807 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1808 static void
1809 __xfrm6_sort(void **dst, void **src, int n,
1810 	     int (*cmp)(const void *p), int maxclass)
1811 {
1812 	int count[XFRM_MAX_DEPTH] = { };
1813 	int class[XFRM_MAX_DEPTH];
1814 	int i;
1815 
1816 	for (i = 0; i < n; i++) {
1817 		int c = cmp(src[i]);
1818 
1819 		class[i] = c;
1820 		count[c]++;
1821 	}
1822 
1823 	for (i = 2; i < maxclass; i++)
1824 		count[i] += count[i - 1];
1825 
1826 	for (i = 0; i < n; i++) {
1827 		dst[count[class[i] - 1]++] = src[i];
1828 		src[i] = NULL;
1829 	}
1830 }
1831 
1832 /* Rule for xfrm_state:
1833  *
1834  * rule 1: select IPsec transport except AH
1835  * rule 2: select MIPv6 RO or inbound trigger
1836  * rule 3: select IPsec transport AH
1837  * rule 4: select IPsec tunnel
1838  * rule 5: others
1839  */
1840 static int __xfrm6_state_sort_cmp(const void *p)
1841 {
1842 	const struct xfrm_state *v = p;
1843 
1844 	switch (v->props.mode) {
1845 	case XFRM_MODE_TRANSPORT:
1846 		if (v->id.proto != IPPROTO_AH)
1847 			return 1;
1848 		else
1849 			return 3;
1850 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1851 	case XFRM_MODE_ROUTEOPTIMIZATION:
1852 	case XFRM_MODE_IN_TRIGGER:
1853 		return 2;
1854 #endif
1855 	case XFRM_MODE_TUNNEL:
1856 	case XFRM_MODE_BEET:
1857 		return 4;
1858 	}
1859 	return 5;
1860 }
1861 
1862 /* Rule for xfrm_tmpl:
1863  *
1864  * rule 1: select IPsec transport
1865  * rule 2: select MIPv6 RO or inbound trigger
1866  * rule 3: select IPsec tunnel
1867  * rule 4: others
1868  */
1869 static int __xfrm6_tmpl_sort_cmp(const void *p)
1870 {
1871 	const struct xfrm_tmpl *v = p;
1872 
1873 	switch (v->mode) {
1874 	case XFRM_MODE_TRANSPORT:
1875 		return 1;
1876 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1877 	case XFRM_MODE_ROUTEOPTIMIZATION:
1878 	case XFRM_MODE_IN_TRIGGER:
1879 		return 2;
1880 #endif
1881 	case XFRM_MODE_TUNNEL:
1882 	case XFRM_MODE_BEET:
1883 		return 3;
1884 	}
1885 	return 4;
1886 }
1887 #else
1888 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1889 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1890 
1891 static inline void
1892 __xfrm6_sort(void **dst, void **src, int n,
1893 	     int (*cmp)(const void *p), int maxclass)
1894 {
1895 	int i;
1896 
1897 	for (i = 0; i < n; i++)
1898 		dst[i] = src[i];
1899 }
1900 #endif /* CONFIG_IPV6 */
1901 
1902 void
1903 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1904 	       unsigned short family)
1905 {
1906 	int i;
1907 
1908 	if (family == AF_INET6)
1909 		__xfrm6_sort((void **)dst, (void **)src, n,
1910 			     __xfrm6_tmpl_sort_cmp, 5);
1911 	else
1912 		for (i = 0; i < n; i++)
1913 			dst[i] = src[i];
1914 }
1915 
1916 void
1917 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1918 		unsigned short family)
1919 {
1920 	int i;
1921 
1922 	if (family == AF_INET6)
1923 		__xfrm6_sort((void **)dst, (void **)src, n,
1924 			     __xfrm6_state_sort_cmp, 6);
1925 	else
1926 		for (i = 0; i < n; i++)
1927 			dst[i] = src[i];
1928 }
1929 #endif
1930 
1931 /* Silly enough, but I'm lazy to build resolution list */
1932 
1933 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1934 {
1935 	int i;
1936 
1937 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1938 		struct xfrm_state *x;
1939 
1940 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1941 			if (x->km.seq == seq &&
1942 			    (mark & x->mark.m) == x->mark.v &&
1943 			    x->km.state == XFRM_STATE_ACQ) {
1944 				xfrm_state_hold(x);
1945 				return x;
1946 			}
1947 		}
1948 	}
1949 	return NULL;
1950 }
1951 
1952 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1953 {
1954 	struct xfrm_state *x;
1955 
1956 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1957 	x = __xfrm_find_acq_byseq(net, mark, seq);
1958 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1959 	return x;
1960 }
1961 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1962 
1963 u32 xfrm_get_acqseq(void)
1964 {
1965 	u32 res;
1966 	static atomic_t acqseq;
1967 
1968 	do {
1969 		res = atomic_inc_return(&acqseq);
1970 	} while (!res);
1971 
1972 	return res;
1973 }
1974 EXPORT_SYMBOL(xfrm_get_acqseq);
1975 
1976 int verify_spi_info(u8 proto, u32 min, u32 max)
1977 {
1978 	switch (proto) {
1979 	case IPPROTO_AH:
1980 	case IPPROTO_ESP:
1981 		break;
1982 
1983 	case IPPROTO_COMP:
1984 		/* IPCOMP spi is 16-bits. */
1985 		if (max >= 0x10000)
1986 			return -EINVAL;
1987 		break;
1988 
1989 	default:
1990 		return -EINVAL;
1991 	}
1992 
1993 	if (min > max)
1994 		return -EINVAL;
1995 
1996 	return 0;
1997 }
1998 EXPORT_SYMBOL(verify_spi_info);
1999 
2000 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2001 {
2002 	struct net *net = xs_net(x);
2003 	unsigned int h;
2004 	struct xfrm_state *x0;
2005 	int err = -ENOENT;
2006 	__be32 minspi = htonl(low);
2007 	__be32 maxspi = htonl(high);
2008 	__be32 newspi = 0;
2009 	u32 mark = x->mark.v & x->mark.m;
2010 
2011 	spin_lock_bh(&x->lock);
2012 	if (x->km.state == XFRM_STATE_DEAD)
2013 		goto unlock;
2014 
2015 	err = 0;
2016 	if (x->id.spi)
2017 		goto unlock;
2018 
2019 	err = -ENOENT;
2020 
2021 	if (minspi == maxspi) {
2022 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2023 		if (x0) {
2024 			xfrm_state_put(x0);
2025 			goto unlock;
2026 		}
2027 		newspi = minspi;
2028 	} else {
2029 		u32 spi = 0;
2030 		for (h = 0; h < high-low+1; h++) {
2031 			spi = low + prandom_u32()%(high-low+1);
2032 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2033 			if (x0 == NULL) {
2034 				newspi = htonl(spi);
2035 				break;
2036 			}
2037 			xfrm_state_put(x0);
2038 		}
2039 	}
2040 	if (newspi) {
2041 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2042 		x->id.spi = newspi;
2043 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2044 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2045 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2046 
2047 		err = 0;
2048 	}
2049 
2050 unlock:
2051 	spin_unlock_bh(&x->lock);
2052 
2053 	return err;
2054 }
2055 EXPORT_SYMBOL(xfrm_alloc_spi);
2056 
2057 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2058 				      struct xfrm_address_filter *filter)
2059 {
2060 	if (filter) {
2061 		if ((filter->family == AF_INET ||
2062 		     filter->family == AF_INET6) &&
2063 		    x->props.family != filter->family)
2064 			return false;
2065 
2066 		return addr_match(&x->props.saddr, &filter->saddr,
2067 				  filter->splen) &&
2068 		       addr_match(&x->id.daddr, &filter->daddr,
2069 				  filter->dplen);
2070 	}
2071 	return true;
2072 }
2073 
2074 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2075 		    int (*func)(struct xfrm_state *, int, void*),
2076 		    void *data)
2077 {
2078 	struct xfrm_state *state;
2079 	struct xfrm_state_walk *x;
2080 	int err = 0;
2081 
2082 	if (walk->seq != 0 && list_empty(&walk->all))
2083 		return 0;
2084 
2085 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2086 	if (list_empty(&walk->all))
2087 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2088 	else
2089 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2090 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2091 		if (x->state == XFRM_STATE_DEAD)
2092 			continue;
2093 		state = container_of(x, struct xfrm_state, km);
2094 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2095 			continue;
2096 		if (!__xfrm_state_filter_match(state, walk->filter))
2097 			continue;
2098 		err = func(state, walk->seq, data);
2099 		if (err) {
2100 			list_move_tail(&walk->all, &x->all);
2101 			goto out;
2102 		}
2103 		walk->seq++;
2104 	}
2105 	if (walk->seq == 0) {
2106 		err = -ENOENT;
2107 		goto out;
2108 	}
2109 	list_del_init(&walk->all);
2110 out:
2111 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2112 	return err;
2113 }
2114 EXPORT_SYMBOL(xfrm_state_walk);
2115 
2116 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2117 			  struct xfrm_address_filter *filter)
2118 {
2119 	INIT_LIST_HEAD(&walk->all);
2120 	walk->proto = proto;
2121 	walk->state = XFRM_STATE_DEAD;
2122 	walk->seq = 0;
2123 	walk->filter = filter;
2124 }
2125 EXPORT_SYMBOL(xfrm_state_walk_init);
2126 
2127 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2128 {
2129 	kfree(walk->filter);
2130 
2131 	if (list_empty(&walk->all))
2132 		return;
2133 
2134 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2135 	list_del(&walk->all);
2136 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2137 }
2138 EXPORT_SYMBOL(xfrm_state_walk_done);
2139 
2140 static void xfrm_replay_timer_handler(struct timer_list *t)
2141 {
2142 	struct xfrm_state *x = from_timer(x, t, rtimer);
2143 
2144 	spin_lock(&x->lock);
2145 
2146 	if (x->km.state == XFRM_STATE_VALID) {
2147 		if (xfrm_aevent_is_on(xs_net(x)))
2148 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2149 		else
2150 			x->xflags |= XFRM_TIME_DEFER;
2151 	}
2152 
2153 	spin_unlock(&x->lock);
2154 }
2155 
2156 static LIST_HEAD(xfrm_km_list);
2157 
2158 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2159 {
2160 	struct xfrm_mgr *km;
2161 
2162 	rcu_read_lock();
2163 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2164 		if (km->notify_policy)
2165 			km->notify_policy(xp, dir, c);
2166 	rcu_read_unlock();
2167 }
2168 
2169 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2170 {
2171 	struct xfrm_mgr *km;
2172 	rcu_read_lock();
2173 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2174 		if (km->notify)
2175 			km->notify(x, c);
2176 	rcu_read_unlock();
2177 }
2178 
2179 EXPORT_SYMBOL(km_policy_notify);
2180 EXPORT_SYMBOL(km_state_notify);
2181 
2182 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2183 {
2184 	struct km_event c;
2185 
2186 	c.data.hard = hard;
2187 	c.portid = portid;
2188 	c.event = XFRM_MSG_EXPIRE;
2189 	km_state_notify(x, &c);
2190 }
2191 
2192 EXPORT_SYMBOL(km_state_expired);
2193 /*
2194  * We send to all registered managers regardless of failure
2195  * We are happy with one success
2196 */
2197 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2198 {
2199 	int err = -EINVAL, acqret;
2200 	struct xfrm_mgr *km;
2201 
2202 	rcu_read_lock();
2203 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2204 		acqret = km->acquire(x, t, pol);
2205 		if (!acqret)
2206 			err = acqret;
2207 	}
2208 	rcu_read_unlock();
2209 	return err;
2210 }
2211 EXPORT_SYMBOL(km_query);
2212 
2213 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2214 {
2215 	int err = -EINVAL;
2216 	struct xfrm_mgr *km;
2217 
2218 	rcu_read_lock();
2219 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2220 		if (km->new_mapping)
2221 			err = km->new_mapping(x, ipaddr, sport);
2222 		if (!err)
2223 			break;
2224 	}
2225 	rcu_read_unlock();
2226 	return err;
2227 }
2228 EXPORT_SYMBOL(km_new_mapping);
2229 
2230 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2231 {
2232 	struct km_event c;
2233 
2234 	c.data.hard = hard;
2235 	c.portid = portid;
2236 	c.event = XFRM_MSG_POLEXPIRE;
2237 	km_policy_notify(pol, dir, &c);
2238 }
2239 EXPORT_SYMBOL(km_policy_expired);
2240 
2241 #ifdef CONFIG_XFRM_MIGRATE
2242 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2243 	       const struct xfrm_migrate *m, int num_migrate,
2244 	       const struct xfrm_kmaddress *k,
2245 	       const struct xfrm_encap_tmpl *encap)
2246 {
2247 	int err = -EINVAL;
2248 	int ret;
2249 	struct xfrm_mgr *km;
2250 
2251 	rcu_read_lock();
2252 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2253 		if (km->migrate) {
2254 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2255 					  encap);
2256 			if (!ret)
2257 				err = ret;
2258 		}
2259 	}
2260 	rcu_read_unlock();
2261 	return err;
2262 }
2263 EXPORT_SYMBOL(km_migrate);
2264 #endif
2265 
2266 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2267 {
2268 	int err = -EINVAL;
2269 	int ret;
2270 	struct xfrm_mgr *km;
2271 
2272 	rcu_read_lock();
2273 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2274 		if (km->report) {
2275 			ret = km->report(net, proto, sel, addr);
2276 			if (!ret)
2277 				err = ret;
2278 		}
2279 	}
2280 	rcu_read_unlock();
2281 	return err;
2282 }
2283 EXPORT_SYMBOL(km_report);
2284 
2285 static bool km_is_alive(const struct km_event *c)
2286 {
2287 	struct xfrm_mgr *km;
2288 	bool is_alive = false;
2289 
2290 	rcu_read_lock();
2291 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2292 		if (km->is_alive && km->is_alive(c)) {
2293 			is_alive = true;
2294 			break;
2295 		}
2296 	}
2297 	rcu_read_unlock();
2298 
2299 	return is_alive;
2300 }
2301 
2302 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2303 static DEFINE_SPINLOCK(xfrm_translator_lock);
2304 static struct xfrm_translator __rcu *xfrm_translator;
2305 
2306 struct xfrm_translator *xfrm_get_translator(void)
2307 {
2308 	struct xfrm_translator *xtr;
2309 
2310 	rcu_read_lock();
2311 	xtr = rcu_dereference(xfrm_translator);
2312 	if (unlikely(!xtr))
2313 		goto out;
2314 	if (!try_module_get(xtr->owner))
2315 		xtr = NULL;
2316 out:
2317 	rcu_read_unlock();
2318 	return xtr;
2319 }
2320 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2321 
2322 void xfrm_put_translator(struct xfrm_translator *xtr)
2323 {
2324 	module_put(xtr->owner);
2325 }
2326 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2327 
2328 int xfrm_register_translator(struct xfrm_translator *xtr)
2329 {
2330 	int err = 0;
2331 
2332 	spin_lock_bh(&xfrm_translator_lock);
2333 	if (unlikely(xfrm_translator != NULL))
2334 		err = -EEXIST;
2335 	else
2336 		rcu_assign_pointer(xfrm_translator, xtr);
2337 	spin_unlock_bh(&xfrm_translator_lock);
2338 
2339 	return err;
2340 }
2341 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2342 
2343 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2344 {
2345 	int err = 0;
2346 
2347 	spin_lock_bh(&xfrm_translator_lock);
2348 	if (likely(xfrm_translator != NULL)) {
2349 		if (rcu_access_pointer(xfrm_translator) != xtr)
2350 			err = -EINVAL;
2351 		else
2352 			RCU_INIT_POINTER(xfrm_translator, NULL);
2353 	}
2354 	spin_unlock_bh(&xfrm_translator_lock);
2355 	synchronize_rcu();
2356 
2357 	return err;
2358 }
2359 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2360 #endif
2361 
2362 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2363 {
2364 	int err;
2365 	u8 *data;
2366 	struct xfrm_mgr *km;
2367 	struct xfrm_policy *pol = NULL;
2368 
2369 	if (sockptr_is_null(optval) && !optlen) {
2370 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2371 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2372 		__sk_dst_reset(sk);
2373 		return 0;
2374 	}
2375 
2376 	if (optlen <= 0 || optlen > PAGE_SIZE)
2377 		return -EMSGSIZE;
2378 
2379 	data = memdup_sockptr(optval, optlen);
2380 	if (IS_ERR(data))
2381 		return PTR_ERR(data);
2382 
2383 	if (in_compat_syscall()) {
2384 		struct xfrm_translator *xtr = xfrm_get_translator();
2385 
2386 		if (!xtr) {
2387 			kfree(data);
2388 			return -EOPNOTSUPP;
2389 		}
2390 
2391 		err = xtr->xlate_user_policy_sockptr(&data, optlen);
2392 		xfrm_put_translator(xtr);
2393 		if (err) {
2394 			kfree(data);
2395 			return err;
2396 		}
2397 	}
2398 
2399 	err = -EINVAL;
2400 	rcu_read_lock();
2401 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2402 		pol = km->compile_policy(sk, optname, data,
2403 					 optlen, &err);
2404 		if (err >= 0)
2405 			break;
2406 	}
2407 	rcu_read_unlock();
2408 
2409 	if (err >= 0) {
2410 		xfrm_sk_policy_insert(sk, err, pol);
2411 		xfrm_pol_put(pol);
2412 		__sk_dst_reset(sk);
2413 		err = 0;
2414 	}
2415 
2416 	kfree(data);
2417 	return err;
2418 }
2419 EXPORT_SYMBOL(xfrm_user_policy);
2420 
2421 static DEFINE_SPINLOCK(xfrm_km_lock);
2422 
2423 int xfrm_register_km(struct xfrm_mgr *km)
2424 {
2425 	spin_lock_bh(&xfrm_km_lock);
2426 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2427 	spin_unlock_bh(&xfrm_km_lock);
2428 	return 0;
2429 }
2430 EXPORT_SYMBOL(xfrm_register_km);
2431 
2432 int xfrm_unregister_km(struct xfrm_mgr *km)
2433 {
2434 	spin_lock_bh(&xfrm_km_lock);
2435 	list_del_rcu(&km->list);
2436 	spin_unlock_bh(&xfrm_km_lock);
2437 	synchronize_rcu();
2438 	return 0;
2439 }
2440 EXPORT_SYMBOL(xfrm_unregister_km);
2441 
2442 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2443 {
2444 	int err = 0;
2445 
2446 	if (WARN_ON(afinfo->family >= NPROTO))
2447 		return -EAFNOSUPPORT;
2448 
2449 	spin_lock_bh(&xfrm_state_afinfo_lock);
2450 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2451 		err = -EEXIST;
2452 	else
2453 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2454 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2455 	return err;
2456 }
2457 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2458 
2459 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2460 {
2461 	int err = 0, family = afinfo->family;
2462 
2463 	if (WARN_ON(family >= NPROTO))
2464 		return -EAFNOSUPPORT;
2465 
2466 	spin_lock_bh(&xfrm_state_afinfo_lock);
2467 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2468 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2469 			err = -EINVAL;
2470 		else
2471 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2472 	}
2473 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2474 	synchronize_rcu();
2475 	return err;
2476 }
2477 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2478 
2479 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2480 {
2481 	if (unlikely(family >= NPROTO))
2482 		return NULL;
2483 
2484 	return rcu_dereference(xfrm_state_afinfo[family]);
2485 }
2486 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2487 
2488 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2489 {
2490 	struct xfrm_state_afinfo *afinfo;
2491 	if (unlikely(family >= NPROTO))
2492 		return NULL;
2493 	rcu_read_lock();
2494 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2495 	if (unlikely(!afinfo))
2496 		rcu_read_unlock();
2497 	return afinfo;
2498 }
2499 
2500 void xfrm_flush_gc(void)
2501 {
2502 	flush_work(&xfrm_state_gc_work);
2503 }
2504 EXPORT_SYMBOL(xfrm_flush_gc);
2505 
2506 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2507 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2508 {
2509 	if (x->tunnel) {
2510 		struct xfrm_state *t = x->tunnel;
2511 
2512 		if (atomic_read(&t->tunnel_users) == 2)
2513 			xfrm_state_delete(t);
2514 		atomic_dec(&t->tunnel_users);
2515 		xfrm_state_put_sync(t);
2516 		x->tunnel = NULL;
2517 	}
2518 }
2519 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2520 
2521 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2522 {
2523 	const struct xfrm_type *type = READ_ONCE(x->type);
2524 	struct crypto_aead *aead;
2525 	u32 blksize, net_adj = 0;
2526 
2527 	if (x->km.state != XFRM_STATE_VALID ||
2528 	    !type || type->proto != IPPROTO_ESP)
2529 		return mtu - x->props.header_len;
2530 
2531 	aead = x->data;
2532 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2533 
2534 	switch (x->props.mode) {
2535 	case XFRM_MODE_TRANSPORT:
2536 	case XFRM_MODE_BEET:
2537 		if (x->props.family == AF_INET)
2538 			net_adj = sizeof(struct iphdr);
2539 		else if (x->props.family == AF_INET6)
2540 			net_adj = sizeof(struct ipv6hdr);
2541 		break;
2542 	case XFRM_MODE_TUNNEL:
2543 		break;
2544 	default:
2545 		WARN_ON_ONCE(1);
2546 		break;
2547 	}
2548 
2549 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2550 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2551 }
2552 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2553 
2554 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2555 {
2556 	const struct xfrm_mode *inner_mode;
2557 	const struct xfrm_mode *outer_mode;
2558 	int family = x->props.family;
2559 	int err;
2560 
2561 	if (family == AF_INET &&
2562 	    xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2563 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2564 
2565 	err = -EPROTONOSUPPORT;
2566 
2567 	if (x->sel.family != AF_UNSPEC) {
2568 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2569 		if (inner_mode == NULL)
2570 			goto error;
2571 
2572 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2573 		    family != x->sel.family)
2574 			goto error;
2575 
2576 		x->inner_mode = *inner_mode;
2577 	} else {
2578 		const struct xfrm_mode *inner_mode_iaf;
2579 		int iafamily = AF_INET;
2580 
2581 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2582 		if (inner_mode == NULL)
2583 			goto error;
2584 
2585 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2586 			goto error;
2587 
2588 		x->inner_mode = *inner_mode;
2589 
2590 		if (x->props.family == AF_INET)
2591 			iafamily = AF_INET6;
2592 
2593 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2594 		if (inner_mode_iaf) {
2595 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2596 				x->inner_mode_iaf = *inner_mode_iaf;
2597 		}
2598 	}
2599 
2600 	x->type = xfrm_get_type(x->id.proto, family);
2601 	if (x->type == NULL)
2602 		goto error;
2603 
2604 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2605 
2606 	err = x->type->init_state(x);
2607 	if (err)
2608 		goto error;
2609 
2610 	outer_mode = xfrm_get_mode(x->props.mode, family);
2611 	if (!outer_mode) {
2612 		err = -EPROTONOSUPPORT;
2613 		goto error;
2614 	}
2615 
2616 	x->outer_mode = *outer_mode;
2617 	if (init_replay) {
2618 		err = xfrm_init_replay(x);
2619 		if (err)
2620 			goto error;
2621 	}
2622 
2623 error:
2624 	return err;
2625 }
2626 
2627 EXPORT_SYMBOL(__xfrm_init_state);
2628 
2629 int xfrm_init_state(struct xfrm_state *x)
2630 {
2631 	int err;
2632 
2633 	err = __xfrm_init_state(x, true, false);
2634 	if (!err)
2635 		x->km.state = XFRM_STATE_VALID;
2636 
2637 	return err;
2638 }
2639 
2640 EXPORT_SYMBOL(xfrm_init_state);
2641 
2642 int __net_init xfrm_state_init(struct net *net)
2643 {
2644 	unsigned int sz;
2645 
2646 	if (net_eq(net, &init_net))
2647 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2648 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2649 
2650 	INIT_LIST_HEAD(&net->xfrm.state_all);
2651 
2652 	sz = sizeof(struct hlist_head) * 8;
2653 
2654 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2655 	if (!net->xfrm.state_bydst)
2656 		goto out_bydst;
2657 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2658 	if (!net->xfrm.state_bysrc)
2659 		goto out_bysrc;
2660 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2661 	if (!net->xfrm.state_byspi)
2662 		goto out_byspi;
2663 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2664 
2665 	net->xfrm.state_num = 0;
2666 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2667 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2668 	seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2669 			       &net->xfrm.xfrm_state_lock);
2670 	return 0;
2671 
2672 out_byspi:
2673 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2674 out_bysrc:
2675 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2676 out_bydst:
2677 	return -ENOMEM;
2678 }
2679 
2680 void xfrm_state_fini(struct net *net)
2681 {
2682 	unsigned int sz;
2683 
2684 	flush_work(&net->xfrm.state_hash_work);
2685 	flush_work(&xfrm_state_gc_work);
2686 	xfrm_state_flush(net, 0, false, true);
2687 
2688 	WARN_ON(!list_empty(&net->xfrm.state_all));
2689 
2690 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2691 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2692 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2693 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2694 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2695 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2696 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2697 }
2698 
2699 #ifdef CONFIG_AUDITSYSCALL
2700 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2701 				     struct audit_buffer *audit_buf)
2702 {
2703 	struct xfrm_sec_ctx *ctx = x->security;
2704 	u32 spi = ntohl(x->id.spi);
2705 
2706 	if (ctx)
2707 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2708 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2709 
2710 	switch (x->props.family) {
2711 	case AF_INET:
2712 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2713 				 &x->props.saddr.a4, &x->id.daddr.a4);
2714 		break;
2715 	case AF_INET6:
2716 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2717 				 x->props.saddr.a6, x->id.daddr.a6);
2718 		break;
2719 	}
2720 
2721 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2722 }
2723 
2724 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2725 				      struct audit_buffer *audit_buf)
2726 {
2727 	const struct iphdr *iph4;
2728 	const struct ipv6hdr *iph6;
2729 
2730 	switch (family) {
2731 	case AF_INET:
2732 		iph4 = ip_hdr(skb);
2733 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2734 				 &iph4->saddr, &iph4->daddr);
2735 		break;
2736 	case AF_INET6:
2737 		iph6 = ipv6_hdr(skb);
2738 		audit_log_format(audit_buf,
2739 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2740 				 &iph6->saddr, &iph6->daddr,
2741 				 iph6->flow_lbl[0] & 0x0f,
2742 				 iph6->flow_lbl[1],
2743 				 iph6->flow_lbl[2]);
2744 		break;
2745 	}
2746 }
2747 
2748 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2749 {
2750 	struct audit_buffer *audit_buf;
2751 
2752 	audit_buf = xfrm_audit_start("SAD-add");
2753 	if (audit_buf == NULL)
2754 		return;
2755 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2756 	xfrm_audit_helper_sainfo(x, audit_buf);
2757 	audit_log_format(audit_buf, " res=%u", result);
2758 	audit_log_end(audit_buf);
2759 }
2760 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2761 
2762 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2763 {
2764 	struct audit_buffer *audit_buf;
2765 
2766 	audit_buf = xfrm_audit_start("SAD-delete");
2767 	if (audit_buf == NULL)
2768 		return;
2769 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2770 	xfrm_audit_helper_sainfo(x, audit_buf);
2771 	audit_log_format(audit_buf, " res=%u", result);
2772 	audit_log_end(audit_buf);
2773 }
2774 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2775 
2776 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2777 				      struct sk_buff *skb)
2778 {
2779 	struct audit_buffer *audit_buf;
2780 	u32 spi;
2781 
2782 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2783 	if (audit_buf == NULL)
2784 		return;
2785 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2786 	/* don't record the sequence number because it's inherent in this kind
2787 	 * of audit message */
2788 	spi = ntohl(x->id.spi);
2789 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2790 	audit_log_end(audit_buf);
2791 }
2792 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2793 
2794 void xfrm_audit_state_replay(struct xfrm_state *x,
2795 			     struct sk_buff *skb, __be32 net_seq)
2796 {
2797 	struct audit_buffer *audit_buf;
2798 	u32 spi;
2799 
2800 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2801 	if (audit_buf == NULL)
2802 		return;
2803 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2804 	spi = ntohl(x->id.spi);
2805 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2806 			 spi, spi, ntohl(net_seq));
2807 	audit_log_end(audit_buf);
2808 }
2809 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2810 
2811 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2812 {
2813 	struct audit_buffer *audit_buf;
2814 
2815 	audit_buf = xfrm_audit_start("SA-notfound");
2816 	if (audit_buf == NULL)
2817 		return;
2818 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2819 	audit_log_end(audit_buf);
2820 }
2821 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2822 
2823 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2824 			       __be32 net_spi, __be32 net_seq)
2825 {
2826 	struct audit_buffer *audit_buf;
2827 	u32 spi;
2828 
2829 	audit_buf = xfrm_audit_start("SA-notfound");
2830 	if (audit_buf == NULL)
2831 		return;
2832 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2833 	spi = ntohl(net_spi);
2834 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2835 			 spi, spi, ntohl(net_seq));
2836 	audit_log_end(audit_buf);
2837 }
2838 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2839 
2840 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2841 			      struct sk_buff *skb, u8 proto)
2842 {
2843 	struct audit_buffer *audit_buf;
2844 	__be32 net_spi;
2845 	__be32 net_seq;
2846 
2847 	audit_buf = xfrm_audit_start("SA-icv-failure");
2848 	if (audit_buf == NULL)
2849 		return;
2850 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2851 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2852 		u32 spi = ntohl(net_spi);
2853 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2854 				 spi, spi, ntohl(net_seq));
2855 	}
2856 	audit_log_end(audit_buf);
2857 }
2858 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2859 #endif /* CONFIG_AUDITSYSCALL */
2860