xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision e7bae9bb)
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 __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
48 static struct kmem_cache *xfrm_state_cache __ro_after_init;
49 
50 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
51 static HLIST_HEAD(xfrm_state_gc_list);
52 
53 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54 {
55 	return refcount_inc_not_zero(&x->refcnt);
56 }
57 
58 static inline unsigned int xfrm_dst_hash(struct net *net,
59 					 const xfrm_address_t *daddr,
60 					 const xfrm_address_t *saddr,
61 					 u32 reqid,
62 					 unsigned short family)
63 {
64 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
65 }
66 
67 static inline unsigned int xfrm_src_hash(struct net *net,
68 					 const xfrm_address_t *daddr,
69 					 const xfrm_address_t *saddr,
70 					 unsigned short family)
71 {
72 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
73 }
74 
75 static inline unsigned int
76 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
77 	      __be32 spi, u8 proto, unsigned short family)
78 {
79 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
80 }
81 
82 static void xfrm_hash_transfer(struct hlist_head *list,
83 			       struct hlist_head *ndsttable,
84 			       struct hlist_head *nsrctable,
85 			       struct hlist_head *nspitable,
86 			       unsigned int nhashmask)
87 {
88 	struct hlist_node *tmp;
89 	struct xfrm_state *x;
90 
91 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
92 		unsigned int h;
93 
94 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
95 				    x->props.reqid, x->props.family,
96 				    nhashmask);
97 		hlist_add_head_rcu(&x->bydst, ndsttable + h);
98 
99 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
100 				    x->props.family,
101 				    nhashmask);
102 		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
103 
104 		if (x->id.spi) {
105 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
106 					    x->id.proto, x->props.family,
107 					    nhashmask);
108 			hlist_add_head_rcu(&x->byspi, nspitable + h);
109 		}
110 	}
111 }
112 
113 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
114 {
115 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
116 }
117 
118 static void xfrm_hash_resize(struct work_struct *work)
119 {
120 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
121 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
122 	unsigned long nsize, osize;
123 	unsigned int nhashmask, ohashmask;
124 	int i;
125 
126 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
127 	ndst = xfrm_hash_alloc(nsize);
128 	if (!ndst)
129 		return;
130 	nsrc = xfrm_hash_alloc(nsize);
131 	if (!nsrc) {
132 		xfrm_hash_free(ndst, nsize);
133 		return;
134 	}
135 	nspi = xfrm_hash_alloc(nsize);
136 	if (!nspi) {
137 		xfrm_hash_free(ndst, nsize);
138 		xfrm_hash_free(nsrc, nsize);
139 		return;
140 	}
141 
142 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
143 	write_seqcount_begin(&xfrm_state_hash_generation);
144 
145 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
146 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
147 	for (i = net->xfrm.state_hmask; i >= 0; i--)
148 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
149 
150 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
151 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
152 	ohashmask = net->xfrm.state_hmask;
153 
154 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
155 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
156 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
157 	net->xfrm.state_hmask = nhashmask;
158 
159 	write_seqcount_end(&xfrm_state_hash_generation);
160 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
161 
162 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
163 
164 	synchronize_rcu();
165 
166 	xfrm_hash_free(odst, osize);
167 	xfrm_hash_free(osrc, osize);
168 	xfrm_hash_free(ospi, osize);
169 }
170 
171 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
172 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
173 
174 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
175 
176 int __xfrm_state_delete(struct xfrm_state *x);
177 
178 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
179 static bool km_is_alive(const struct km_event *c);
180 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
181 
182 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
183 {
184 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
185 	int err = 0;
186 
187 	if (!afinfo)
188 		return -EAFNOSUPPORT;
189 
190 #define X(afi, T, name) do {			\
191 		WARN_ON((afi)->type_ ## name);	\
192 		(afi)->type_ ## name = (T);	\
193 	} while (0)
194 
195 	switch (type->proto) {
196 	case IPPROTO_COMP:
197 		X(afinfo, type, comp);
198 		break;
199 	case IPPROTO_AH:
200 		X(afinfo, type, ah);
201 		break;
202 	case IPPROTO_ESP:
203 		X(afinfo, type, esp);
204 		break;
205 	case IPPROTO_IPIP:
206 		X(afinfo, type, ipip);
207 		break;
208 	case IPPROTO_DSTOPTS:
209 		X(afinfo, type, dstopts);
210 		break;
211 	case IPPROTO_ROUTING:
212 		X(afinfo, type, routing);
213 		break;
214 	case IPPROTO_IPV6:
215 		X(afinfo, type, ipip6);
216 		break;
217 	default:
218 		WARN_ON(1);
219 		err = -EPROTONOSUPPORT;
220 		break;
221 	}
222 #undef X
223 	rcu_read_unlock();
224 	return err;
225 }
226 EXPORT_SYMBOL(xfrm_register_type);
227 
228 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
229 {
230 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
231 
232 	if (unlikely(afinfo == NULL))
233 		return;
234 
235 #define X(afi, T, name) do {				\
236 		WARN_ON((afi)->type_ ## name != (T));	\
237 		(afi)->type_ ## name = NULL;		\
238 	} while (0)
239 
240 	switch (type->proto) {
241 	case IPPROTO_COMP:
242 		X(afinfo, type, comp);
243 		break;
244 	case IPPROTO_AH:
245 		X(afinfo, type, ah);
246 		break;
247 	case IPPROTO_ESP:
248 		X(afinfo, type, esp);
249 		break;
250 	case IPPROTO_IPIP:
251 		X(afinfo, type, ipip);
252 		break;
253 	case IPPROTO_DSTOPTS:
254 		X(afinfo, type, dstopts);
255 		break;
256 	case IPPROTO_ROUTING:
257 		X(afinfo, type, routing);
258 		break;
259 	case IPPROTO_IPV6:
260 		X(afinfo, type, ipip6);
261 		break;
262 	default:
263 		WARN_ON(1);
264 		break;
265 	}
266 #undef X
267 	rcu_read_unlock();
268 }
269 EXPORT_SYMBOL(xfrm_unregister_type);
270 
271 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
272 {
273 	const struct xfrm_type *type = NULL;
274 	struct xfrm_state_afinfo *afinfo;
275 	int modload_attempted = 0;
276 
277 retry:
278 	afinfo = xfrm_state_get_afinfo(family);
279 	if (unlikely(afinfo == NULL))
280 		return NULL;
281 
282 	switch (proto) {
283 	case IPPROTO_COMP:
284 		type = afinfo->type_comp;
285 		break;
286 	case IPPROTO_AH:
287 		type = afinfo->type_ah;
288 		break;
289 	case IPPROTO_ESP:
290 		type = afinfo->type_esp;
291 		break;
292 	case IPPROTO_IPIP:
293 		type = afinfo->type_ipip;
294 		break;
295 	case IPPROTO_DSTOPTS:
296 		type = afinfo->type_dstopts;
297 		break;
298 	case IPPROTO_ROUTING:
299 		type = afinfo->type_routing;
300 		break;
301 	case IPPROTO_IPV6:
302 		type = afinfo->type_ipip6;
303 		break;
304 	default:
305 		break;
306 	}
307 
308 	if (unlikely(type && !try_module_get(type->owner)))
309 		type = NULL;
310 
311 	rcu_read_unlock();
312 
313 	if (!type && !modload_attempted) {
314 		request_module("xfrm-type-%d-%d", family, proto);
315 		modload_attempted = 1;
316 		goto retry;
317 	}
318 
319 	return type;
320 }
321 
322 static void xfrm_put_type(const struct xfrm_type *type)
323 {
324 	module_put(type->owner);
325 }
326 
327 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
328 			       unsigned short family)
329 {
330 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
331 	int err = 0;
332 
333 	if (unlikely(afinfo == NULL))
334 		return -EAFNOSUPPORT;
335 
336 	switch (type->proto) {
337 	case IPPROTO_ESP:
338 		WARN_ON(afinfo->type_offload_esp);
339 		afinfo->type_offload_esp = type;
340 		break;
341 	default:
342 		WARN_ON(1);
343 		err = -EPROTONOSUPPORT;
344 		break;
345 	}
346 
347 	rcu_read_unlock();
348 	return err;
349 }
350 EXPORT_SYMBOL(xfrm_register_type_offload);
351 
352 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
353 				  unsigned short family)
354 {
355 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
356 
357 	if (unlikely(afinfo == NULL))
358 		return;
359 
360 	switch (type->proto) {
361 	case IPPROTO_ESP:
362 		WARN_ON(afinfo->type_offload_esp != type);
363 		afinfo->type_offload_esp = NULL;
364 		break;
365 	default:
366 		WARN_ON(1);
367 		break;
368 	}
369 	rcu_read_unlock();
370 }
371 EXPORT_SYMBOL(xfrm_unregister_type_offload);
372 
373 static const struct xfrm_type_offload *
374 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
375 {
376 	const struct xfrm_type_offload *type = NULL;
377 	struct xfrm_state_afinfo *afinfo;
378 
379 retry:
380 	afinfo = xfrm_state_get_afinfo(family);
381 	if (unlikely(afinfo == NULL))
382 		return NULL;
383 
384 	switch (proto) {
385 	case IPPROTO_ESP:
386 		type = afinfo->type_offload_esp;
387 		break;
388 	default:
389 		break;
390 	}
391 
392 	if ((type && !try_module_get(type->owner)))
393 		type = NULL;
394 
395 	rcu_read_unlock();
396 
397 	if (!type && try_load) {
398 		request_module("xfrm-offload-%d-%d", family, proto);
399 		try_load = false;
400 		goto retry;
401 	}
402 
403 	return type;
404 }
405 
406 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
407 {
408 	module_put(type->owner);
409 }
410 
411 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
412 	[XFRM_MODE_BEET] = {
413 		.encap = XFRM_MODE_BEET,
414 		.flags = XFRM_MODE_FLAG_TUNNEL,
415 		.family = AF_INET,
416 	},
417 	[XFRM_MODE_TRANSPORT] = {
418 		.encap = XFRM_MODE_TRANSPORT,
419 		.family = AF_INET,
420 	},
421 	[XFRM_MODE_TUNNEL] = {
422 		.encap = XFRM_MODE_TUNNEL,
423 		.flags = XFRM_MODE_FLAG_TUNNEL,
424 		.family = AF_INET,
425 	},
426 };
427 
428 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
429 	[XFRM_MODE_BEET] = {
430 		.encap = XFRM_MODE_BEET,
431 		.flags = XFRM_MODE_FLAG_TUNNEL,
432 		.family = AF_INET6,
433 	},
434 	[XFRM_MODE_ROUTEOPTIMIZATION] = {
435 		.encap = XFRM_MODE_ROUTEOPTIMIZATION,
436 		.family = AF_INET6,
437 	},
438 	[XFRM_MODE_TRANSPORT] = {
439 		.encap = XFRM_MODE_TRANSPORT,
440 		.family = AF_INET6,
441 	},
442 	[XFRM_MODE_TUNNEL] = {
443 		.encap = XFRM_MODE_TUNNEL,
444 		.flags = XFRM_MODE_FLAG_TUNNEL,
445 		.family = AF_INET6,
446 	},
447 };
448 
449 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
450 {
451 	const struct xfrm_mode *mode;
452 
453 	if (unlikely(encap >= XFRM_MODE_MAX))
454 		return NULL;
455 
456 	switch (family) {
457 	case AF_INET:
458 		mode = &xfrm4_mode_map[encap];
459 		if (mode->family == family)
460 			return mode;
461 		break;
462 	case AF_INET6:
463 		mode = &xfrm6_mode_map[encap];
464 		if (mode->family == family)
465 			return mode;
466 		break;
467 	default:
468 		break;
469 	}
470 
471 	return NULL;
472 }
473 
474 void xfrm_state_free(struct xfrm_state *x)
475 {
476 	kmem_cache_free(xfrm_state_cache, x);
477 }
478 EXPORT_SYMBOL(xfrm_state_free);
479 
480 static void ___xfrm_state_destroy(struct xfrm_state *x)
481 {
482 	hrtimer_cancel(&x->mtimer);
483 	del_timer_sync(&x->rtimer);
484 	kfree(x->aead);
485 	kfree(x->aalg);
486 	kfree(x->ealg);
487 	kfree(x->calg);
488 	kfree(x->encap);
489 	kfree(x->coaddr);
490 	kfree(x->replay_esn);
491 	kfree(x->preplay_esn);
492 	if (x->type_offload)
493 		xfrm_put_type_offload(x->type_offload);
494 	if (x->type) {
495 		x->type->destructor(x);
496 		xfrm_put_type(x->type);
497 	}
498 	if (x->xfrag.page)
499 		put_page(x->xfrag.page);
500 	xfrm_dev_state_free(x);
501 	security_xfrm_state_free(x);
502 	xfrm_state_free(x);
503 }
504 
505 static void xfrm_state_gc_task(struct work_struct *work)
506 {
507 	struct xfrm_state *x;
508 	struct hlist_node *tmp;
509 	struct hlist_head gc_list;
510 
511 	spin_lock_bh(&xfrm_state_gc_lock);
512 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
513 	spin_unlock_bh(&xfrm_state_gc_lock);
514 
515 	synchronize_rcu();
516 
517 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
518 		___xfrm_state_destroy(x);
519 }
520 
521 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
522 {
523 	struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
524 	enum hrtimer_restart ret = HRTIMER_NORESTART;
525 	time64_t now = ktime_get_real_seconds();
526 	time64_t next = TIME64_MAX;
527 	int warn = 0;
528 	int err = 0;
529 
530 	spin_lock(&x->lock);
531 	if (x->km.state == XFRM_STATE_DEAD)
532 		goto out;
533 	if (x->km.state == XFRM_STATE_EXPIRED)
534 		goto expired;
535 	if (x->lft.hard_add_expires_seconds) {
536 		long tmo = x->lft.hard_add_expires_seconds +
537 			x->curlft.add_time - now;
538 		if (tmo <= 0) {
539 			if (x->xflags & XFRM_SOFT_EXPIRE) {
540 				/* enter hard expire without soft expire first?!
541 				 * setting a new date could trigger this.
542 				 * workaround: fix x->curflt.add_time by below:
543 				 */
544 				x->curlft.add_time = now - x->saved_tmo - 1;
545 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
546 			} else
547 				goto expired;
548 		}
549 		if (tmo < next)
550 			next = tmo;
551 	}
552 	if (x->lft.hard_use_expires_seconds) {
553 		long tmo = x->lft.hard_use_expires_seconds +
554 			(x->curlft.use_time ? : now) - now;
555 		if (tmo <= 0)
556 			goto expired;
557 		if (tmo < next)
558 			next = tmo;
559 	}
560 	if (x->km.dying)
561 		goto resched;
562 	if (x->lft.soft_add_expires_seconds) {
563 		long tmo = x->lft.soft_add_expires_seconds +
564 			x->curlft.add_time - now;
565 		if (tmo <= 0) {
566 			warn = 1;
567 			x->xflags &= ~XFRM_SOFT_EXPIRE;
568 		} else if (tmo < next) {
569 			next = tmo;
570 			x->xflags |= XFRM_SOFT_EXPIRE;
571 			x->saved_tmo = tmo;
572 		}
573 	}
574 	if (x->lft.soft_use_expires_seconds) {
575 		long tmo = x->lft.soft_use_expires_seconds +
576 			(x->curlft.use_time ? : now) - now;
577 		if (tmo <= 0)
578 			warn = 1;
579 		else if (tmo < next)
580 			next = tmo;
581 	}
582 
583 	x->km.dying = warn;
584 	if (warn)
585 		km_state_expired(x, 0, 0);
586 resched:
587 	if (next != TIME64_MAX) {
588 		hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
589 		ret = HRTIMER_RESTART;
590 	}
591 
592 	goto out;
593 
594 expired:
595 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
596 		x->km.state = XFRM_STATE_EXPIRED;
597 
598 	err = __xfrm_state_delete(x);
599 	if (!err)
600 		km_state_expired(x, 1, 0);
601 
602 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
603 
604 out:
605 	spin_unlock(&x->lock);
606 	return ret;
607 }
608 
609 static void xfrm_replay_timer_handler(struct timer_list *t);
610 
611 struct xfrm_state *xfrm_state_alloc(struct net *net)
612 {
613 	struct xfrm_state *x;
614 
615 	x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
616 
617 	if (x) {
618 		write_pnet(&x->xs_net, net);
619 		refcount_set(&x->refcnt, 1);
620 		atomic_set(&x->tunnel_users, 0);
621 		INIT_LIST_HEAD(&x->km.all);
622 		INIT_HLIST_NODE(&x->bydst);
623 		INIT_HLIST_NODE(&x->bysrc);
624 		INIT_HLIST_NODE(&x->byspi);
625 		hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
626 		x->mtimer.function = xfrm_timer_handler;
627 		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
628 		x->curlft.add_time = ktime_get_real_seconds();
629 		x->lft.soft_byte_limit = XFRM_INF;
630 		x->lft.soft_packet_limit = XFRM_INF;
631 		x->lft.hard_byte_limit = XFRM_INF;
632 		x->lft.hard_packet_limit = XFRM_INF;
633 		x->replay_maxage = 0;
634 		x->replay_maxdiff = 0;
635 		spin_lock_init(&x->lock);
636 	}
637 	return x;
638 }
639 EXPORT_SYMBOL(xfrm_state_alloc);
640 
641 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
642 {
643 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
644 
645 	if (sync) {
646 		synchronize_rcu();
647 		___xfrm_state_destroy(x);
648 	} else {
649 		spin_lock_bh(&xfrm_state_gc_lock);
650 		hlist_add_head(&x->gclist, &xfrm_state_gc_list);
651 		spin_unlock_bh(&xfrm_state_gc_lock);
652 		schedule_work(&xfrm_state_gc_work);
653 	}
654 }
655 EXPORT_SYMBOL(__xfrm_state_destroy);
656 
657 int __xfrm_state_delete(struct xfrm_state *x)
658 {
659 	struct net *net = xs_net(x);
660 	int err = -ESRCH;
661 
662 	if (x->km.state != XFRM_STATE_DEAD) {
663 		x->km.state = XFRM_STATE_DEAD;
664 		spin_lock(&net->xfrm.xfrm_state_lock);
665 		list_del(&x->km.all);
666 		hlist_del_rcu(&x->bydst);
667 		hlist_del_rcu(&x->bysrc);
668 		if (x->id.spi)
669 			hlist_del_rcu(&x->byspi);
670 		net->xfrm.state_num--;
671 		spin_unlock(&net->xfrm.xfrm_state_lock);
672 
673 		if (x->encap_sk)
674 			sock_put(rcu_dereference_raw(x->encap_sk));
675 
676 		xfrm_dev_state_delete(x);
677 
678 		/* All xfrm_state objects are created by xfrm_state_alloc.
679 		 * The xfrm_state_alloc call gives a reference, and that
680 		 * is what we are dropping here.
681 		 */
682 		xfrm_state_put(x);
683 		err = 0;
684 	}
685 
686 	return err;
687 }
688 EXPORT_SYMBOL(__xfrm_state_delete);
689 
690 int xfrm_state_delete(struct xfrm_state *x)
691 {
692 	int err;
693 
694 	spin_lock_bh(&x->lock);
695 	err = __xfrm_state_delete(x);
696 	spin_unlock_bh(&x->lock);
697 
698 	return err;
699 }
700 EXPORT_SYMBOL(xfrm_state_delete);
701 
702 #ifdef CONFIG_SECURITY_NETWORK_XFRM
703 static inline int
704 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
705 {
706 	int i, err = 0;
707 
708 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
709 		struct xfrm_state *x;
710 
711 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
712 			if (xfrm_id_proto_match(x->id.proto, proto) &&
713 			   (err = security_xfrm_state_delete(x)) != 0) {
714 				xfrm_audit_state_delete(x, 0, task_valid);
715 				return err;
716 			}
717 		}
718 	}
719 
720 	return err;
721 }
722 
723 static inline int
724 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
725 {
726 	int i, err = 0;
727 
728 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
729 		struct xfrm_state *x;
730 		struct xfrm_state_offload *xso;
731 
732 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
733 			xso = &x->xso;
734 
735 			if (xso->dev == dev &&
736 			   (err = security_xfrm_state_delete(x)) != 0) {
737 				xfrm_audit_state_delete(x, 0, task_valid);
738 				return err;
739 			}
740 		}
741 	}
742 
743 	return err;
744 }
745 #else
746 static inline int
747 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
748 {
749 	return 0;
750 }
751 
752 static inline int
753 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
754 {
755 	return 0;
756 }
757 #endif
758 
759 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
760 {
761 	int i, err = 0, cnt = 0;
762 
763 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
764 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
765 	if (err)
766 		goto out;
767 
768 	err = -ESRCH;
769 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
770 		struct xfrm_state *x;
771 restart:
772 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
773 			if (!xfrm_state_kern(x) &&
774 			    xfrm_id_proto_match(x->id.proto, proto)) {
775 				xfrm_state_hold(x);
776 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
777 
778 				err = xfrm_state_delete(x);
779 				xfrm_audit_state_delete(x, err ? 0 : 1,
780 							task_valid);
781 				if (sync)
782 					xfrm_state_put_sync(x);
783 				else
784 					xfrm_state_put(x);
785 				if (!err)
786 					cnt++;
787 
788 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
789 				goto restart;
790 			}
791 		}
792 	}
793 out:
794 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
795 	if (cnt)
796 		err = 0;
797 
798 	return err;
799 }
800 EXPORT_SYMBOL(xfrm_state_flush);
801 
802 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
803 {
804 	int i, err = 0, cnt = 0;
805 
806 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
807 	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
808 	if (err)
809 		goto out;
810 
811 	err = -ESRCH;
812 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
813 		struct xfrm_state *x;
814 		struct xfrm_state_offload *xso;
815 restart:
816 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
817 			xso = &x->xso;
818 
819 			if (!xfrm_state_kern(x) && xso->dev == dev) {
820 				xfrm_state_hold(x);
821 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
822 
823 				err = xfrm_state_delete(x);
824 				xfrm_audit_state_delete(x, err ? 0 : 1,
825 							task_valid);
826 				xfrm_state_put(x);
827 				if (!err)
828 					cnt++;
829 
830 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
831 				goto restart;
832 			}
833 		}
834 	}
835 	if (cnt)
836 		err = 0;
837 
838 out:
839 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
840 	return err;
841 }
842 EXPORT_SYMBOL(xfrm_dev_state_flush);
843 
844 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
845 {
846 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
847 	si->sadcnt = net->xfrm.state_num;
848 	si->sadhcnt = net->xfrm.state_hmask + 1;
849 	si->sadhmcnt = xfrm_state_hashmax;
850 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
851 }
852 EXPORT_SYMBOL(xfrm_sad_getinfo);
853 
854 static void
855 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
856 {
857 	const struct flowi4 *fl4 = &fl->u.ip4;
858 
859 	sel->daddr.a4 = fl4->daddr;
860 	sel->saddr.a4 = fl4->saddr;
861 	sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
862 	sel->dport_mask = htons(0xffff);
863 	sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
864 	sel->sport_mask = htons(0xffff);
865 	sel->family = AF_INET;
866 	sel->prefixlen_d = 32;
867 	sel->prefixlen_s = 32;
868 	sel->proto = fl4->flowi4_proto;
869 	sel->ifindex = fl4->flowi4_oif;
870 }
871 
872 static void
873 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
874 {
875 	const struct flowi6 *fl6 = &fl->u.ip6;
876 
877 	/* Initialize temporary selector matching only to current session. */
878 	*(struct in6_addr *)&sel->daddr = fl6->daddr;
879 	*(struct in6_addr *)&sel->saddr = fl6->saddr;
880 	sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
881 	sel->dport_mask = htons(0xffff);
882 	sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
883 	sel->sport_mask = htons(0xffff);
884 	sel->family = AF_INET6;
885 	sel->prefixlen_d = 128;
886 	sel->prefixlen_s = 128;
887 	sel->proto = fl6->flowi6_proto;
888 	sel->ifindex = fl6->flowi6_oif;
889 }
890 
891 static void
892 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
893 		    const struct xfrm_tmpl *tmpl,
894 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
895 		    unsigned short family)
896 {
897 	switch (family) {
898 	case AF_INET:
899 		__xfrm4_init_tempsel(&x->sel, fl);
900 		break;
901 	case AF_INET6:
902 		__xfrm6_init_tempsel(&x->sel, fl);
903 		break;
904 	}
905 
906 	x->id = tmpl->id;
907 
908 	switch (tmpl->encap_family) {
909 	case AF_INET:
910 		if (x->id.daddr.a4 == 0)
911 			x->id.daddr.a4 = daddr->a4;
912 		x->props.saddr = tmpl->saddr;
913 		if (x->props.saddr.a4 == 0)
914 			x->props.saddr.a4 = saddr->a4;
915 		break;
916 	case AF_INET6:
917 		if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
918 			memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
919 		memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
920 		if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
921 			memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
922 		break;
923 	}
924 
925 	x->props.mode = tmpl->mode;
926 	x->props.reqid = tmpl->reqid;
927 	x->props.family = tmpl->encap_family;
928 }
929 
930 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
931 					      const xfrm_address_t *daddr,
932 					      __be32 spi, u8 proto,
933 					      unsigned short family)
934 {
935 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
936 	struct xfrm_state *x;
937 
938 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
939 		if (x->props.family != family ||
940 		    x->id.spi       != spi ||
941 		    x->id.proto     != proto ||
942 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
943 			continue;
944 
945 		if ((mark & x->mark.m) != x->mark.v)
946 			continue;
947 		if (!xfrm_state_hold_rcu(x))
948 			continue;
949 		return x;
950 	}
951 
952 	return NULL;
953 }
954 
955 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
956 						     const xfrm_address_t *daddr,
957 						     const xfrm_address_t *saddr,
958 						     u8 proto, unsigned short family)
959 {
960 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
961 	struct xfrm_state *x;
962 
963 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
964 		if (x->props.family != family ||
965 		    x->id.proto     != proto ||
966 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
967 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
968 			continue;
969 
970 		if ((mark & x->mark.m) != x->mark.v)
971 			continue;
972 		if (!xfrm_state_hold_rcu(x))
973 			continue;
974 		return x;
975 	}
976 
977 	return NULL;
978 }
979 
980 static inline struct xfrm_state *
981 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
982 {
983 	struct net *net = xs_net(x);
984 	u32 mark = x->mark.v & x->mark.m;
985 
986 	if (use_spi)
987 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
988 					   x->id.spi, x->id.proto, family);
989 	else
990 		return __xfrm_state_lookup_byaddr(net, mark,
991 						  &x->id.daddr,
992 						  &x->props.saddr,
993 						  x->id.proto, family);
994 }
995 
996 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
997 {
998 	if (have_hash_collision &&
999 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1000 	    net->xfrm.state_num > net->xfrm.state_hmask)
1001 		schedule_work(&net->xfrm.state_hash_work);
1002 }
1003 
1004 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1005 			       const struct flowi *fl, unsigned short family,
1006 			       struct xfrm_state **best, int *acq_in_progress,
1007 			       int *error)
1008 {
1009 	/* Resolution logic:
1010 	 * 1. There is a valid state with matching selector. Done.
1011 	 * 2. Valid state with inappropriate selector. Skip.
1012 	 *
1013 	 * Entering area of "sysdeps".
1014 	 *
1015 	 * 3. If state is not valid, selector is temporary, it selects
1016 	 *    only session which triggered previous resolution. Key
1017 	 *    manager will do something to install a state with proper
1018 	 *    selector.
1019 	 */
1020 	if (x->km.state == XFRM_STATE_VALID) {
1021 		if ((x->sel.family &&
1022 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
1023 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
1024 			return;
1025 
1026 		if (!*best ||
1027 		    (*best)->km.dying > x->km.dying ||
1028 		    ((*best)->km.dying == x->km.dying &&
1029 		     (*best)->curlft.add_time < x->curlft.add_time))
1030 			*best = x;
1031 	} else if (x->km.state == XFRM_STATE_ACQ) {
1032 		*acq_in_progress = 1;
1033 	} else if (x->km.state == XFRM_STATE_ERROR ||
1034 		   x->km.state == XFRM_STATE_EXPIRED) {
1035 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
1036 		    security_xfrm_state_pol_flow_match(x, pol, fl))
1037 			*error = -ESRCH;
1038 	}
1039 }
1040 
1041 struct xfrm_state *
1042 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1043 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1044 		struct xfrm_policy *pol, int *err,
1045 		unsigned short family, u32 if_id)
1046 {
1047 	static xfrm_address_t saddr_wildcard = { };
1048 	struct net *net = xp_net(pol);
1049 	unsigned int h, h_wildcard;
1050 	struct xfrm_state *x, *x0, *to_put;
1051 	int acquire_in_progress = 0;
1052 	int error = 0;
1053 	struct xfrm_state *best = NULL;
1054 	u32 mark = pol->mark.v & pol->mark.m;
1055 	unsigned short encap_family = tmpl->encap_family;
1056 	unsigned int sequence;
1057 	struct km_event c;
1058 
1059 	to_put = NULL;
1060 
1061 	sequence = read_seqcount_begin(&xfrm_state_hash_generation);
1062 
1063 	rcu_read_lock();
1064 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1065 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1066 		if (x->props.family == encap_family &&
1067 		    x->props.reqid == tmpl->reqid &&
1068 		    (mark & x->mark.m) == x->mark.v &&
1069 		    x->if_id == if_id &&
1070 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1071 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1072 		    tmpl->mode == x->props.mode &&
1073 		    tmpl->id.proto == x->id.proto &&
1074 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1075 			xfrm_state_look_at(pol, x, fl, encap_family,
1076 					   &best, &acquire_in_progress, &error);
1077 	}
1078 	if (best || acquire_in_progress)
1079 		goto found;
1080 
1081 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1082 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1083 		if (x->props.family == encap_family &&
1084 		    x->props.reqid == tmpl->reqid &&
1085 		    (mark & x->mark.m) == x->mark.v &&
1086 		    x->if_id == if_id &&
1087 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1088 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1089 		    tmpl->mode == x->props.mode &&
1090 		    tmpl->id.proto == x->id.proto &&
1091 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1092 			xfrm_state_look_at(pol, x, fl, encap_family,
1093 					   &best, &acquire_in_progress, &error);
1094 	}
1095 
1096 found:
1097 	x = best;
1098 	if (!x && !error && !acquire_in_progress) {
1099 		if (tmpl->id.spi &&
1100 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1101 					      tmpl->id.proto, encap_family)) != NULL) {
1102 			to_put = x0;
1103 			error = -EEXIST;
1104 			goto out;
1105 		}
1106 
1107 		c.net = net;
1108 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1109 		 * for each and every packet - garbage collection might not
1110 		 * handle the flood.
1111 		 */
1112 		if (!km_is_alive(&c)) {
1113 			error = -ESRCH;
1114 			goto out;
1115 		}
1116 
1117 		x = xfrm_state_alloc(net);
1118 		if (x == NULL) {
1119 			error = -ENOMEM;
1120 			goto out;
1121 		}
1122 		/* Initialize temporary state matching only
1123 		 * to current session. */
1124 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1125 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1126 		x->if_id = if_id;
1127 
1128 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1129 		if (error) {
1130 			x->km.state = XFRM_STATE_DEAD;
1131 			to_put = x;
1132 			x = NULL;
1133 			goto out;
1134 		}
1135 
1136 		if (km_query(x, tmpl, pol) == 0) {
1137 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1138 			x->km.state = XFRM_STATE_ACQ;
1139 			list_add(&x->km.all, &net->xfrm.state_all);
1140 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1141 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1142 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1143 			if (x->id.spi) {
1144 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1145 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1146 			}
1147 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1148 			hrtimer_start(&x->mtimer,
1149 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1150 				      HRTIMER_MODE_REL_SOFT);
1151 			net->xfrm.state_num++;
1152 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1153 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1154 		} else {
1155 			x->km.state = XFRM_STATE_DEAD;
1156 			to_put = x;
1157 			x = NULL;
1158 			error = -ESRCH;
1159 		}
1160 	}
1161 out:
1162 	if (x) {
1163 		if (!xfrm_state_hold_rcu(x)) {
1164 			*err = -EAGAIN;
1165 			x = NULL;
1166 		}
1167 	} else {
1168 		*err = acquire_in_progress ? -EAGAIN : error;
1169 	}
1170 	rcu_read_unlock();
1171 	if (to_put)
1172 		xfrm_state_put(to_put);
1173 
1174 	if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1175 		*err = -EAGAIN;
1176 		if (x) {
1177 			xfrm_state_put(x);
1178 			x = NULL;
1179 		}
1180 	}
1181 
1182 	return x;
1183 }
1184 
1185 struct xfrm_state *
1186 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1187 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1188 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1189 {
1190 	unsigned int h;
1191 	struct xfrm_state *rx = NULL, *x = NULL;
1192 
1193 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1194 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1195 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1196 		if (x->props.family == family &&
1197 		    x->props.reqid == reqid &&
1198 		    (mark & x->mark.m) == x->mark.v &&
1199 		    x->if_id == if_id &&
1200 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1201 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1202 		    mode == x->props.mode &&
1203 		    proto == x->id.proto &&
1204 		    x->km.state == XFRM_STATE_VALID) {
1205 			rx = x;
1206 			break;
1207 		}
1208 	}
1209 
1210 	if (rx)
1211 		xfrm_state_hold(rx);
1212 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1213 
1214 
1215 	return rx;
1216 }
1217 EXPORT_SYMBOL(xfrm_stateonly_find);
1218 
1219 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1220 					      unsigned short family)
1221 {
1222 	struct xfrm_state *x;
1223 	struct xfrm_state_walk *w;
1224 
1225 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1226 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1227 		x = container_of(w, struct xfrm_state, km);
1228 		if (x->props.family != family ||
1229 			x->id.spi != spi)
1230 			continue;
1231 
1232 		xfrm_state_hold(x);
1233 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1234 		return x;
1235 	}
1236 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1237 	return NULL;
1238 }
1239 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1240 
1241 static void __xfrm_state_insert(struct xfrm_state *x)
1242 {
1243 	struct net *net = xs_net(x);
1244 	unsigned int h;
1245 
1246 	list_add(&x->km.all, &net->xfrm.state_all);
1247 
1248 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1249 			  x->props.reqid, x->props.family);
1250 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1251 
1252 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1253 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1254 
1255 	if (x->id.spi) {
1256 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1257 				  x->props.family);
1258 
1259 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1260 	}
1261 
1262 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1263 	if (x->replay_maxage)
1264 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1265 
1266 	net->xfrm.state_num++;
1267 
1268 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1269 }
1270 
1271 /* net->xfrm.xfrm_state_lock is held */
1272 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1273 {
1274 	struct net *net = xs_net(xnew);
1275 	unsigned short family = xnew->props.family;
1276 	u32 reqid = xnew->props.reqid;
1277 	struct xfrm_state *x;
1278 	unsigned int h;
1279 	u32 mark = xnew->mark.v & xnew->mark.m;
1280 	u32 if_id = xnew->if_id;
1281 
1282 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1283 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1284 		if (x->props.family	== family &&
1285 		    x->props.reqid	== reqid &&
1286 		    x->if_id		== if_id &&
1287 		    (mark & x->mark.m) == x->mark.v &&
1288 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1289 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1290 			x->genid++;
1291 	}
1292 }
1293 
1294 void xfrm_state_insert(struct xfrm_state *x)
1295 {
1296 	struct net *net = xs_net(x);
1297 
1298 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1299 	__xfrm_state_bump_genids(x);
1300 	__xfrm_state_insert(x);
1301 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1302 }
1303 EXPORT_SYMBOL(xfrm_state_insert);
1304 
1305 /* net->xfrm.xfrm_state_lock is held */
1306 static struct xfrm_state *__find_acq_core(struct net *net,
1307 					  const struct xfrm_mark *m,
1308 					  unsigned short family, u8 mode,
1309 					  u32 reqid, u32 if_id, u8 proto,
1310 					  const xfrm_address_t *daddr,
1311 					  const xfrm_address_t *saddr,
1312 					  int create)
1313 {
1314 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1315 	struct xfrm_state *x;
1316 	u32 mark = m->v & m->m;
1317 
1318 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1319 		if (x->props.reqid  != reqid ||
1320 		    x->props.mode   != mode ||
1321 		    x->props.family != family ||
1322 		    x->km.state     != XFRM_STATE_ACQ ||
1323 		    x->id.spi       != 0 ||
1324 		    x->id.proto	    != proto ||
1325 		    (mark & x->mark.m) != x->mark.v ||
1326 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1327 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1328 			continue;
1329 
1330 		xfrm_state_hold(x);
1331 		return x;
1332 	}
1333 
1334 	if (!create)
1335 		return NULL;
1336 
1337 	x = xfrm_state_alloc(net);
1338 	if (likely(x)) {
1339 		switch (family) {
1340 		case AF_INET:
1341 			x->sel.daddr.a4 = daddr->a4;
1342 			x->sel.saddr.a4 = saddr->a4;
1343 			x->sel.prefixlen_d = 32;
1344 			x->sel.prefixlen_s = 32;
1345 			x->props.saddr.a4 = saddr->a4;
1346 			x->id.daddr.a4 = daddr->a4;
1347 			break;
1348 
1349 		case AF_INET6:
1350 			x->sel.daddr.in6 = daddr->in6;
1351 			x->sel.saddr.in6 = saddr->in6;
1352 			x->sel.prefixlen_d = 128;
1353 			x->sel.prefixlen_s = 128;
1354 			x->props.saddr.in6 = saddr->in6;
1355 			x->id.daddr.in6 = daddr->in6;
1356 			break;
1357 		}
1358 
1359 		x->km.state = XFRM_STATE_ACQ;
1360 		x->id.proto = proto;
1361 		x->props.family = family;
1362 		x->props.mode = mode;
1363 		x->props.reqid = reqid;
1364 		x->if_id = if_id;
1365 		x->mark.v = m->v;
1366 		x->mark.m = m->m;
1367 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1368 		xfrm_state_hold(x);
1369 		hrtimer_start(&x->mtimer,
1370 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1371 			      HRTIMER_MODE_REL_SOFT);
1372 		list_add(&x->km.all, &net->xfrm.state_all);
1373 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1374 		h = xfrm_src_hash(net, daddr, saddr, family);
1375 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1376 
1377 		net->xfrm.state_num++;
1378 
1379 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1380 	}
1381 
1382 	return x;
1383 }
1384 
1385 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1386 
1387 int xfrm_state_add(struct xfrm_state *x)
1388 {
1389 	struct net *net = xs_net(x);
1390 	struct xfrm_state *x1, *to_put;
1391 	int family;
1392 	int err;
1393 	u32 mark = x->mark.v & x->mark.m;
1394 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1395 
1396 	family = x->props.family;
1397 
1398 	to_put = NULL;
1399 
1400 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1401 
1402 	x1 = __xfrm_state_locate(x, use_spi, family);
1403 	if (x1) {
1404 		to_put = x1;
1405 		x1 = NULL;
1406 		err = -EEXIST;
1407 		goto out;
1408 	}
1409 
1410 	if (use_spi && x->km.seq) {
1411 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1412 		if (x1 && ((x1->id.proto != x->id.proto) ||
1413 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1414 			to_put = x1;
1415 			x1 = NULL;
1416 		}
1417 	}
1418 
1419 	if (use_spi && !x1)
1420 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1421 				     x->props.reqid, x->if_id, x->id.proto,
1422 				     &x->id.daddr, &x->props.saddr, 0);
1423 
1424 	__xfrm_state_bump_genids(x);
1425 	__xfrm_state_insert(x);
1426 	err = 0;
1427 
1428 out:
1429 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1430 
1431 	if (x1) {
1432 		xfrm_state_delete(x1);
1433 		xfrm_state_put(x1);
1434 	}
1435 
1436 	if (to_put)
1437 		xfrm_state_put(to_put);
1438 
1439 	return err;
1440 }
1441 EXPORT_SYMBOL(xfrm_state_add);
1442 
1443 #ifdef CONFIG_XFRM_MIGRATE
1444 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1445 					   struct xfrm_encap_tmpl *encap)
1446 {
1447 	struct net *net = xs_net(orig);
1448 	struct xfrm_state *x = xfrm_state_alloc(net);
1449 	if (!x)
1450 		goto out;
1451 
1452 	memcpy(&x->id, &orig->id, sizeof(x->id));
1453 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1454 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1455 	x->props.mode = orig->props.mode;
1456 	x->props.replay_window = orig->props.replay_window;
1457 	x->props.reqid = orig->props.reqid;
1458 	x->props.family = orig->props.family;
1459 	x->props.saddr = orig->props.saddr;
1460 
1461 	if (orig->aalg) {
1462 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1463 		if (!x->aalg)
1464 			goto error;
1465 	}
1466 	x->props.aalgo = orig->props.aalgo;
1467 
1468 	if (orig->aead) {
1469 		x->aead = xfrm_algo_aead_clone(orig->aead);
1470 		x->geniv = orig->geniv;
1471 		if (!x->aead)
1472 			goto error;
1473 	}
1474 	if (orig->ealg) {
1475 		x->ealg = xfrm_algo_clone(orig->ealg);
1476 		if (!x->ealg)
1477 			goto error;
1478 	}
1479 	x->props.ealgo = orig->props.ealgo;
1480 
1481 	if (orig->calg) {
1482 		x->calg = xfrm_algo_clone(orig->calg);
1483 		if (!x->calg)
1484 			goto error;
1485 	}
1486 	x->props.calgo = orig->props.calgo;
1487 
1488 	if (encap || orig->encap) {
1489 		if (encap)
1490 			x->encap = kmemdup(encap, sizeof(*x->encap),
1491 					GFP_KERNEL);
1492 		else
1493 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1494 					GFP_KERNEL);
1495 
1496 		if (!x->encap)
1497 			goto error;
1498 	}
1499 
1500 	if (orig->coaddr) {
1501 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1502 				    GFP_KERNEL);
1503 		if (!x->coaddr)
1504 			goto error;
1505 	}
1506 
1507 	if (orig->replay_esn) {
1508 		if (xfrm_replay_clone(x, orig))
1509 			goto error;
1510 	}
1511 
1512 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1513 
1514 	if (xfrm_init_state(x) < 0)
1515 		goto error;
1516 
1517 	x->props.flags = orig->props.flags;
1518 	x->props.extra_flags = orig->props.extra_flags;
1519 
1520 	x->if_id = orig->if_id;
1521 	x->tfcpad = orig->tfcpad;
1522 	x->replay_maxdiff = orig->replay_maxdiff;
1523 	x->replay_maxage = orig->replay_maxage;
1524 	x->curlft.add_time = orig->curlft.add_time;
1525 	x->km.state = orig->km.state;
1526 	x->km.seq = orig->km.seq;
1527 	x->replay = orig->replay;
1528 	x->preplay = orig->preplay;
1529 
1530 	return x;
1531 
1532  error:
1533 	xfrm_state_put(x);
1534 out:
1535 	return NULL;
1536 }
1537 
1538 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1539 {
1540 	unsigned int h;
1541 	struct xfrm_state *x = NULL;
1542 
1543 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1544 
1545 	if (m->reqid) {
1546 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1547 				  m->reqid, m->old_family);
1548 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1549 			if (x->props.mode != m->mode ||
1550 			    x->id.proto != m->proto)
1551 				continue;
1552 			if (m->reqid && x->props.reqid != m->reqid)
1553 				continue;
1554 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1555 					     m->old_family) ||
1556 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1557 					     m->old_family))
1558 				continue;
1559 			xfrm_state_hold(x);
1560 			break;
1561 		}
1562 	} else {
1563 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1564 				  m->old_family);
1565 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1566 			if (x->props.mode != m->mode ||
1567 			    x->id.proto != m->proto)
1568 				continue;
1569 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1570 					     m->old_family) ||
1571 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1572 					     m->old_family))
1573 				continue;
1574 			xfrm_state_hold(x);
1575 			break;
1576 		}
1577 	}
1578 
1579 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1580 
1581 	return x;
1582 }
1583 EXPORT_SYMBOL(xfrm_migrate_state_find);
1584 
1585 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1586 				      struct xfrm_migrate *m,
1587 				      struct xfrm_encap_tmpl *encap)
1588 {
1589 	struct xfrm_state *xc;
1590 
1591 	xc = xfrm_state_clone(x, encap);
1592 	if (!xc)
1593 		return NULL;
1594 
1595 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1596 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1597 
1598 	/* add state */
1599 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1600 		/* a care is needed when the destination address of the
1601 		   state is to be updated as it is a part of triplet */
1602 		xfrm_state_insert(xc);
1603 	} else {
1604 		if (xfrm_state_add(xc) < 0)
1605 			goto error;
1606 	}
1607 
1608 	return xc;
1609 error:
1610 	xfrm_state_put(xc);
1611 	return NULL;
1612 }
1613 EXPORT_SYMBOL(xfrm_state_migrate);
1614 #endif
1615 
1616 int xfrm_state_update(struct xfrm_state *x)
1617 {
1618 	struct xfrm_state *x1, *to_put;
1619 	int err;
1620 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1621 	struct net *net = xs_net(x);
1622 
1623 	to_put = NULL;
1624 
1625 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1626 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1627 
1628 	err = -ESRCH;
1629 	if (!x1)
1630 		goto out;
1631 
1632 	if (xfrm_state_kern(x1)) {
1633 		to_put = x1;
1634 		err = -EEXIST;
1635 		goto out;
1636 	}
1637 
1638 	if (x1->km.state == XFRM_STATE_ACQ) {
1639 		__xfrm_state_insert(x);
1640 		x = NULL;
1641 	}
1642 	err = 0;
1643 
1644 out:
1645 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1646 
1647 	if (to_put)
1648 		xfrm_state_put(to_put);
1649 
1650 	if (err)
1651 		return err;
1652 
1653 	if (!x) {
1654 		xfrm_state_delete(x1);
1655 		xfrm_state_put(x1);
1656 		return 0;
1657 	}
1658 
1659 	err = -EINVAL;
1660 	spin_lock_bh(&x1->lock);
1661 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1662 		if (x->encap && x1->encap &&
1663 		    x->encap->encap_type == x1->encap->encap_type)
1664 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1665 		else if (x->encap || x1->encap)
1666 			goto fail;
1667 
1668 		if (x->coaddr && x1->coaddr) {
1669 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1670 		}
1671 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1672 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1673 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1674 		x1->km.dying = 0;
1675 
1676 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1677 			      HRTIMER_MODE_REL_SOFT);
1678 		if (x1->curlft.use_time)
1679 			xfrm_state_check_expire(x1);
1680 
1681 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1682 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1683 
1684 			if (x->props.smark.m || x->props.smark.v)
1685 				x1->props.smark = x->props.smark;
1686 
1687 			if (x->if_id)
1688 				x1->if_id = x->if_id;
1689 
1690 			__xfrm_state_bump_genids(x1);
1691 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1692 		}
1693 
1694 		err = 0;
1695 		x->km.state = XFRM_STATE_DEAD;
1696 		__xfrm_state_put(x);
1697 	}
1698 
1699 fail:
1700 	spin_unlock_bh(&x1->lock);
1701 
1702 	xfrm_state_put(x1);
1703 
1704 	return err;
1705 }
1706 EXPORT_SYMBOL(xfrm_state_update);
1707 
1708 int xfrm_state_check_expire(struct xfrm_state *x)
1709 {
1710 	if (!x->curlft.use_time)
1711 		x->curlft.use_time = ktime_get_real_seconds();
1712 
1713 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1714 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1715 		x->km.state = XFRM_STATE_EXPIRED;
1716 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1717 		return -EINVAL;
1718 	}
1719 
1720 	if (!x->km.dying &&
1721 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1722 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1723 		x->km.dying = 1;
1724 		km_state_expired(x, 0, 0);
1725 	}
1726 	return 0;
1727 }
1728 EXPORT_SYMBOL(xfrm_state_check_expire);
1729 
1730 struct xfrm_state *
1731 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1732 		  u8 proto, unsigned short family)
1733 {
1734 	struct xfrm_state *x;
1735 
1736 	rcu_read_lock();
1737 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1738 	rcu_read_unlock();
1739 	return x;
1740 }
1741 EXPORT_SYMBOL(xfrm_state_lookup);
1742 
1743 struct xfrm_state *
1744 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1745 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1746 			 u8 proto, unsigned short family)
1747 {
1748 	struct xfrm_state *x;
1749 
1750 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1751 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1752 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1753 	return x;
1754 }
1755 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1756 
1757 struct xfrm_state *
1758 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1759 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1760 	      const xfrm_address_t *saddr, int create, unsigned short family)
1761 {
1762 	struct xfrm_state *x;
1763 
1764 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1765 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1766 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1767 
1768 	return x;
1769 }
1770 EXPORT_SYMBOL(xfrm_find_acq);
1771 
1772 #ifdef CONFIG_XFRM_SUB_POLICY
1773 #if IS_ENABLED(CONFIG_IPV6)
1774 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1775 static void
1776 __xfrm6_sort(void **dst, void **src, int n,
1777 	     int (*cmp)(const void *p), int maxclass)
1778 {
1779 	int count[XFRM_MAX_DEPTH] = { };
1780 	int class[XFRM_MAX_DEPTH];
1781 	int i;
1782 
1783 	for (i = 0; i < n; i++) {
1784 		int c = cmp(src[i]);
1785 
1786 		class[i] = c;
1787 		count[c]++;
1788 	}
1789 
1790 	for (i = 2; i < maxclass; i++)
1791 		count[i] += count[i - 1];
1792 
1793 	for (i = 0; i < n; i++) {
1794 		dst[count[class[i] - 1]++] = src[i];
1795 		src[i] = NULL;
1796 	}
1797 }
1798 
1799 /* Rule for xfrm_state:
1800  *
1801  * rule 1: select IPsec transport except AH
1802  * rule 2: select MIPv6 RO or inbound trigger
1803  * rule 3: select IPsec transport AH
1804  * rule 4: select IPsec tunnel
1805  * rule 5: others
1806  */
1807 static int __xfrm6_state_sort_cmp(const void *p)
1808 {
1809 	const struct xfrm_state *v = p;
1810 
1811 	switch (v->props.mode) {
1812 	case XFRM_MODE_TRANSPORT:
1813 		if (v->id.proto != IPPROTO_AH)
1814 			return 1;
1815 		else
1816 			return 3;
1817 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1818 	case XFRM_MODE_ROUTEOPTIMIZATION:
1819 	case XFRM_MODE_IN_TRIGGER:
1820 		return 2;
1821 #endif
1822 	case XFRM_MODE_TUNNEL:
1823 	case XFRM_MODE_BEET:
1824 		return 4;
1825 	}
1826 	return 5;
1827 }
1828 
1829 /* Rule for xfrm_tmpl:
1830  *
1831  * rule 1: select IPsec transport
1832  * rule 2: select MIPv6 RO or inbound trigger
1833  * rule 3: select IPsec tunnel
1834  * rule 4: others
1835  */
1836 static int __xfrm6_tmpl_sort_cmp(const void *p)
1837 {
1838 	const struct xfrm_tmpl *v = p;
1839 
1840 	switch (v->mode) {
1841 	case XFRM_MODE_TRANSPORT:
1842 		return 1;
1843 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1844 	case XFRM_MODE_ROUTEOPTIMIZATION:
1845 	case XFRM_MODE_IN_TRIGGER:
1846 		return 2;
1847 #endif
1848 	case XFRM_MODE_TUNNEL:
1849 	case XFRM_MODE_BEET:
1850 		return 3;
1851 	}
1852 	return 4;
1853 }
1854 #else
1855 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1856 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1857 
1858 static inline void
1859 __xfrm6_sort(void **dst, void **src, int n,
1860 	     int (*cmp)(const void *p), int maxclass)
1861 {
1862 	int i;
1863 
1864 	for (i = 0; i < n; i++)
1865 		dst[i] = src[i];
1866 }
1867 #endif /* CONFIG_IPV6 */
1868 
1869 void
1870 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1871 	       unsigned short family)
1872 {
1873 	int i;
1874 
1875 	if (family == AF_INET6)
1876 		__xfrm6_sort((void **)dst, (void **)src, n,
1877 			     __xfrm6_tmpl_sort_cmp, 5);
1878 	else
1879 		for (i = 0; i < n; i++)
1880 			dst[i] = src[i];
1881 }
1882 
1883 void
1884 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1885 		unsigned short family)
1886 {
1887 	int i;
1888 
1889 	if (family == AF_INET6)
1890 		__xfrm6_sort((void **)dst, (void **)src, n,
1891 			     __xfrm6_state_sort_cmp, 6);
1892 	else
1893 		for (i = 0; i < n; i++)
1894 			dst[i] = src[i];
1895 }
1896 #endif
1897 
1898 /* Silly enough, but I'm lazy to build resolution list */
1899 
1900 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1901 {
1902 	int i;
1903 
1904 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1905 		struct xfrm_state *x;
1906 
1907 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1908 			if (x->km.seq == seq &&
1909 			    (mark & x->mark.m) == x->mark.v &&
1910 			    x->km.state == XFRM_STATE_ACQ) {
1911 				xfrm_state_hold(x);
1912 				return x;
1913 			}
1914 		}
1915 	}
1916 	return NULL;
1917 }
1918 
1919 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1920 {
1921 	struct xfrm_state *x;
1922 
1923 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1924 	x = __xfrm_find_acq_byseq(net, mark, seq);
1925 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1926 	return x;
1927 }
1928 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1929 
1930 u32 xfrm_get_acqseq(void)
1931 {
1932 	u32 res;
1933 	static atomic_t acqseq;
1934 
1935 	do {
1936 		res = atomic_inc_return(&acqseq);
1937 	} while (!res);
1938 
1939 	return res;
1940 }
1941 EXPORT_SYMBOL(xfrm_get_acqseq);
1942 
1943 int verify_spi_info(u8 proto, u32 min, u32 max)
1944 {
1945 	switch (proto) {
1946 	case IPPROTO_AH:
1947 	case IPPROTO_ESP:
1948 		break;
1949 
1950 	case IPPROTO_COMP:
1951 		/* IPCOMP spi is 16-bits. */
1952 		if (max >= 0x10000)
1953 			return -EINVAL;
1954 		break;
1955 
1956 	default:
1957 		return -EINVAL;
1958 	}
1959 
1960 	if (min > max)
1961 		return -EINVAL;
1962 
1963 	return 0;
1964 }
1965 EXPORT_SYMBOL(verify_spi_info);
1966 
1967 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1968 {
1969 	struct net *net = xs_net(x);
1970 	unsigned int h;
1971 	struct xfrm_state *x0;
1972 	int err = -ENOENT;
1973 	__be32 minspi = htonl(low);
1974 	__be32 maxspi = htonl(high);
1975 	u32 mark = x->mark.v & x->mark.m;
1976 
1977 	spin_lock_bh(&x->lock);
1978 	if (x->km.state == XFRM_STATE_DEAD)
1979 		goto unlock;
1980 
1981 	err = 0;
1982 	if (x->id.spi)
1983 		goto unlock;
1984 
1985 	err = -ENOENT;
1986 
1987 	if (minspi == maxspi) {
1988 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1989 		if (x0) {
1990 			xfrm_state_put(x0);
1991 			goto unlock;
1992 		}
1993 		x->id.spi = minspi;
1994 	} else {
1995 		u32 spi = 0;
1996 		for (h = 0; h < high-low+1; h++) {
1997 			spi = low + prandom_u32()%(high-low+1);
1998 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1999 			if (x0 == NULL) {
2000 				x->id.spi = htonl(spi);
2001 				break;
2002 			}
2003 			xfrm_state_put(x0);
2004 		}
2005 	}
2006 	if (x->id.spi) {
2007 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2008 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2009 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2010 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2011 
2012 		err = 0;
2013 	}
2014 
2015 unlock:
2016 	spin_unlock_bh(&x->lock);
2017 
2018 	return err;
2019 }
2020 EXPORT_SYMBOL(xfrm_alloc_spi);
2021 
2022 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2023 				      struct xfrm_address_filter *filter)
2024 {
2025 	if (filter) {
2026 		if ((filter->family == AF_INET ||
2027 		     filter->family == AF_INET6) &&
2028 		    x->props.family != filter->family)
2029 			return false;
2030 
2031 		return addr_match(&x->props.saddr, &filter->saddr,
2032 				  filter->splen) &&
2033 		       addr_match(&x->id.daddr, &filter->daddr,
2034 				  filter->dplen);
2035 	}
2036 	return true;
2037 }
2038 
2039 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2040 		    int (*func)(struct xfrm_state *, int, void*),
2041 		    void *data)
2042 {
2043 	struct xfrm_state *state;
2044 	struct xfrm_state_walk *x;
2045 	int err = 0;
2046 
2047 	if (walk->seq != 0 && list_empty(&walk->all))
2048 		return 0;
2049 
2050 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2051 	if (list_empty(&walk->all))
2052 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2053 	else
2054 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2055 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2056 		if (x->state == XFRM_STATE_DEAD)
2057 			continue;
2058 		state = container_of(x, struct xfrm_state, km);
2059 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2060 			continue;
2061 		if (!__xfrm_state_filter_match(state, walk->filter))
2062 			continue;
2063 		err = func(state, walk->seq, data);
2064 		if (err) {
2065 			list_move_tail(&walk->all, &x->all);
2066 			goto out;
2067 		}
2068 		walk->seq++;
2069 	}
2070 	if (walk->seq == 0) {
2071 		err = -ENOENT;
2072 		goto out;
2073 	}
2074 	list_del_init(&walk->all);
2075 out:
2076 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2077 	return err;
2078 }
2079 EXPORT_SYMBOL(xfrm_state_walk);
2080 
2081 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2082 			  struct xfrm_address_filter *filter)
2083 {
2084 	INIT_LIST_HEAD(&walk->all);
2085 	walk->proto = proto;
2086 	walk->state = XFRM_STATE_DEAD;
2087 	walk->seq = 0;
2088 	walk->filter = filter;
2089 }
2090 EXPORT_SYMBOL(xfrm_state_walk_init);
2091 
2092 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2093 {
2094 	kfree(walk->filter);
2095 
2096 	if (list_empty(&walk->all))
2097 		return;
2098 
2099 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2100 	list_del(&walk->all);
2101 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2102 }
2103 EXPORT_SYMBOL(xfrm_state_walk_done);
2104 
2105 static void xfrm_replay_timer_handler(struct timer_list *t)
2106 {
2107 	struct xfrm_state *x = from_timer(x, t, rtimer);
2108 
2109 	spin_lock(&x->lock);
2110 
2111 	if (x->km.state == XFRM_STATE_VALID) {
2112 		if (xfrm_aevent_is_on(xs_net(x)))
2113 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2114 		else
2115 			x->xflags |= XFRM_TIME_DEFER;
2116 	}
2117 
2118 	spin_unlock(&x->lock);
2119 }
2120 
2121 static LIST_HEAD(xfrm_km_list);
2122 
2123 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2124 {
2125 	struct xfrm_mgr *km;
2126 
2127 	rcu_read_lock();
2128 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2129 		if (km->notify_policy)
2130 			km->notify_policy(xp, dir, c);
2131 	rcu_read_unlock();
2132 }
2133 
2134 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2135 {
2136 	struct xfrm_mgr *km;
2137 	rcu_read_lock();
2138 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2139 		if (km->notify)
2140 			km->notify(x, c);
2141 	rcu_read_unlock();
2142 }
2143 
2144 EXPORT_SYMBOL(km_policy_notify);
2145 EXPORT_SYMBOL(km_state_notify);
2146 
2147 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2148 {
2149 	struct km_event c;
2150 
2151 	c.data.hard = hard;
2152 	c.portid = portid;
2153 	c.event = XFRM_MSG_EXPIRE;
2154 	km_state_notify(x, &c);
2155 }
2156 
2157 EXPORT_SYMBOL(km_state_expired);
2158 /*
2159  * We send to all registered managers regardless of failure
2160  * We are happy with one success
2161 */
2162 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2163 {
2164 	int err = -EINVAL, acqret;
2165 	struct xfrm_mgr *km;
2166 
2167 	rcu_read_lock();
2168 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2169 		acqret = km->acquire(x, t, pol);
2170 		if (!acqret)
2171 			err = acqret;
2172 	}
2173 	rcu_read_unlock();
2174 	return err;
2175 }
2176 EXPORT_SYMBOL(km_query);
2177 
2178 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2179 {
2180 	int err = -EINVAL;
2181 	struct xfrm_mgr *km;
2182 
2183 	rcu_read_lock();
2184 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2185 		if (km->new_mapping)
2186 			err = km->new_mapping(x, ipaddr, sport);
2187 		if (!err)
2188 			break;
2189 	}
2190 	rcu_read_unlock();
2191 	return err;
2192 }
2193 EXPORT_SYMBOL(km_new_mapping);
2194 
2195 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2196 {
2197 	struct km_event c;
2198 
2199 	c.data.hard = hard;
2200 	c.portid = portid;
2201 	c.event = XFRM_MSG_POLEXPIRE;
2202 	km_policy_notify(pol, dir, &c);
2203 }
2204 EXPORT_SYMBOL(km_policy_expired);
2205 
2206 #ifdef CONFIG_XFRM_MIGRATE
2207 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2208 	       const struct xfrm_migrate *m, int num_migrate,
2209 	       const struct xfrm_kmaddress *k,
2210 	       const struct xfrm_encap_tmpl *encap)
2211 {
2212 	int err = -EINVAL;
2213 	int ret;
2214 	struct xfrm_mgr *km;
2215 
2216 	rcu_read_lock();
2217 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2218 		if (km->migrate) {
2219 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2220 					  encap);
2221 			if (!ret)
2222 				err = ret;
2223 		}
2224 	}
2225 	rcu_read_unlock();
2226 	return err;
2227 }
2228 EXPORT_SYMBOL(km_migrate);
2229 #endif
2230 
2231 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2232 {
2233 	int err = -EINVAL;
2234 	int ret;
2235 	struct xfrm_mgr *km;
2236 
2237 	rcu_read_lock();
2238 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2239 		if (km->report) {
2240 			ret = km->report(net, proto, sel, addr);
2241 			if (!ret)
2242 				err = ret;
2243 		}
2244 	}
2245 	rcu_read_unlock();
2246 	return err;
2247 }
2248 EXPORT_SYMBOL(km_report);
2249 
2250 static bool km_is_alive(const struct km_event *c)
2251 {
2252 	struct xfrm_mgr *km;
2253 	bool is_alive = false;
2254 
2255 	rcu_read_lock();
2256 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2257 		if (km->is_alive && km->is_alive(c)) {
2258 			is_alive = true;
2259 			break;
2260 		}
2261 	}
2262 	rcu_read_unlock();
2263 
2264 	return is_alive;
2265 }
2266 
2267 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2268 {
2269 	int err;
2270 	u8 *data;
2271 	struct xfrm_mgr *km;
2272 	struct xfrm_policy *pol = NULL;
2273 
2274 	if (in_compat_syscall())
2275 		return -EOPNOTSUPP;
2276 
2277 	if (sockptr_is_null(optval) && !optlen) {
2278 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2279 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2280 		__sk_dst_reset(sk);
2281 		return 0;
2282 	}
2283 
2284 	if (optlen <= 0 || optlen > PAGE_SIZE)
2285 		return -EMSGSIZE;
2286 
2287 	data = memdup_sockptr(optval, optlen);
2288 	if (IS_ERR(data))
2289 		return PTR_ERR(data);
2290 
2291 	err = -EINVAL;
2292 	rcu_read_lock();
2293 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2294 		pol = km->compile_policy(sk, optname, data,
2295 					 optlen, &err);
2296 		if (err >= 0)
2297 			break;
2298 	}
2299 	rcu_read_unlock();
2300 
2301 	if (err >= 0) {
2302 		xfrm_sk_policy_insert(sk, err, pol);
2303 		xfrm_pol_put(pol);
2304 		__sk_dst_reset(sk);
2305 		err = 0;
2306 	}
2307 
2308 	kfree(data);
2309 	return err;
2310 }
2311 EXPORT_SYMBOL(xfrm_user_policy);
2312 
2313 static DEFINE_SPINLOCK(xfrm_km_lock);
2314 
2315 int xfrm_register_km(struct xfrm_mgr *km)
2316 {
2317 	spin_lock_bh(&xfrm_km_lock);
2318 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2319 	spin_unlock_bh(&xfrm_km_lock);
2320 	return 0;
2321 }
2322 EXPORT_SYMBOL(xfrm_register_km);
2323 
2324 int xfrm_unregister_km(struct xfrm_mgr *km)
2325 {
2326 	spin_lock_bh(&xfrm_km_lock);
2327 	list_del_rcu(&km->list);
2328 	spin_unlock_bh(&xfrm_km_lock);
2329 	synchronize_rcu();
2330 	return 0;
2331 }
2332 EXPORT_SYMBOL(xfrm_unregister_km);
2333 
2334 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2335 {
2336 	int err = 0;
2337 
2338 	if (WARN_ON(afinfo->family >= NPROTO))
2339 		return -EAFNOSUPPORT;
2340 
2341 	spin_lock_bh(&xfrm_state_afinfo_lock);
2342 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2343 		err = -EEXIST;
2344 	else
2345 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2346 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2347 	return err;
2348 }
2349 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2350 
2351 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2352 {
2353 	int err = 0, family = afinfo->family;
2354 
2355 	if (WARN_ON(family >= NPROTO))
2356 		return -EAFNOSUPPORT;
2357 
2358 	spin_lock_bh(&xfrm_state_afinfo_lock);
2359 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2360 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2361 			err = -EINVAL;
2362 		else
2363 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2364 	}
2365 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2366 	synchronize_rcu();
2367 	return err;
2368 }
2369 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2370 
2371 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2372 {
2373 	if (unlikely(family >= NPROTO))
2374 		return NULL;
2375 
2376 	return rcu_dereference(xfrm_state_afinfo[family]);
2377 }
2378 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2379 
2380 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2381 {
2382 	struct xfrm_state_afinfo *afinfo;
2383 	if (unlikely(family >= NPROTO))
2384 		return NULL;
2385 	rcu_read_lock();
2386 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2387 	if (unlikely(!afinfo))
2388 		rcu_read_unlock();
2389 	return afinfo;
2390 }
2391 
2392 void xfrm_flush_gc(void)
2393 {
2394 	flush_work(&xfrm_state_gc_work);
2395 }
2396 EXPORT_SYMBOL(xfrm_flush_gc);
2397 
2398 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2399 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2400 {
2401 	if (x->tunnel) {
2402 		struct xfrm_state *t = x->tunnel;
2403 
2404 		if (atomic_read(&t->tunnel_users) == 2)
2405 			xfrm_state_delete(t);
2406 		atomic_dec(&t->tunnel_users);
2407 		xfrm_state_put_sync(t);
2408 		x->tunnel = NULL;
2409 	}
2410 }
2411 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2412 
2413 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2414 {
2415 	const struct xfrm_type *type = READ_ONCE(x->type);
2416 	struct crypto_aead *aead;
2417 	u32 blksize, net_adj = 0;
2418 
2419 	if (x->km.state != XFRM_STATE_VALID ||
2420 	    !type || type->proto != IPPROTO_ESP)
2421 		return mtu - x->props.header_len;
2422 
2423 	aead = x->data;
2424 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2425 
2426 	switch (x->props.mode) {
2427 	case XFRM_MODE_TRANSPORT:
2428 	case XFRM_MODE_BEET:
2429 		if (x->props.family == AF_INET)
2430 			net_adj = sizeof(struct iphdr);
2431 		else if (x->props.family == AF_INET6)
2432 			net_adj = sizeof(struct ipv6hdr);
2433 		break;
2434 	case XFRM_MODE_TUNNEL:
2435 		break;
2436 	default:
2437 		WARN_ON_ONCE(1);
2438 		break;
2439 	}
2440 
2441 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2442 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2443 }
2444 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2445 
2446 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2447 {
2448 	const struct xfrm_mode *inner_mode;
2449 	const struct xfrm_mode *outer_mode;
2450 	int family = x->props.family;
2451 	int err;
2452 
2453 	if (family == AF_INET &&
2454 	    xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2455 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2456 
2457 	err = -EPROTONOSUPPORT;
2458 
2459 	if (x->sel.family != AF_UNSPEC) {
2460 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2461 		if (inner_mode == NULL)
2462 			goto error;
2463 
2464 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2465 		    family != x->sel.family)
2466 			goto error;
2467 
2468 		x->inner_mode = *inner_mode;
2469 	} else {
2470 		const struct xfrm_mode *inner_mode_iaf;
2471 		int iafamily = AF_INET;
2472 
2473 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2474 		if (inner_mode == NULL)
2475 			goto error;
2476 
2477 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2478 			goto error;
2479 
2480 		x->inner_mode = *inner_mode;
2481 
2482 		if (x->props.family == AF_INET)
2483 			iafamily = AF_INET6;
2484 
2485 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2486 		if (inner_mode_iaf) {
2487 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2488 				x->inner_mode_iaf = *inner_mode_iaf;
2489 		}
2490 	}
2491 
2492 	x->type = xfrm_get_type(x->id.proto, family);
2493 	if (x->type == NULL)
2494 		goto error;
2495 
2496 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2497 
2498 	err = x->type->init_state(x);
2499 	if (err)
2500 		goto error;
2501 
2502 	outer_mode = xfrm_get_mode(x->props.mode, family);
2503 	if (!outer_mode) {
2504 		err = -EPROTONOSUPPORT;
2505 		goto error;
2506 	}
2507 
2508 	x->outer_mode = *outer_mode;
2509 	if (init_replay) {
2510 		err = xfrm_init_replay(x);
2511 		if (err)
2512 			goto error;
2513 	}
2514 
2515 error:
2516 	return err;
2517 }
2518 
2519 EXPORT_SYMBOL(__xfrm_init_state);
2520 
2521 int xfrm_init_state(struct xfrm_state *x)
2522 {
2523 	int err;
2524 
2525 	err = __xfrm_init_state(x, true, false);
2526 	if (!err)
2527 		x->km.state = XFRM_STATE_VALID;
2528 
2529 	return err;
2530 }
2531 
2532 EXPORT_SYMBOL(xfrm_init_state);
2533 
2534 int __net_init xfrm_state_init(struct net *net)
2535 {
2536 	unsigned int sz;
2537 
2538 	if (net_eq(net, &init_net))
2539 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2540 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2541 
2542 	INIT_LIST_HEAD(&net->xfrm.state_all);
2543 
2544 	sz = sizeof(struct hlist_head) * 8;
2545 
2546 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2547 	if (!net->xfrm.state_bydst)
2548 		goto out_bydst;
2549 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2550 	if (!net->xfrm.state_bysrc)
2551 		goto out_bysrc;
2552 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2553 	if (!net->xfrm.state_byspi)
2554 		goto out_byspi;
2555 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2556 
2557 	net->xfrm.state_num = 0;
2558 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2559 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2560 	return 0;
2561 
2562 out_byspi:
2563 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2564 out_bysrc:
2565 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2566 out_bydst:
2567 	return -ENOMEM;
2568 }
2569 
2570 void xfrm_state_fini(struct net *net)
2571 {
2572 	unsigned int sz;
2573 
2574 	flush_work(&net->xfrm.state_hash_work);
2575 	flush_work(&xfrm_state_gc_work);
2576 	xfrm_state_flush(net, 0, false, true);
2577 
2578 	WARN_ON(!list_empty(&net->xfrm.state_all));
2579 
2580 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2581 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2582 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2583 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2584 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2585 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2586 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2587 }
2588 
2589 #ifdef CONFIG_AUDITSYSCALL
2590 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2591 				     struct audit_buffer *audit_buf)
2592 {
2593 	struct xfrm_sec_ctx *ctx = x->security;
2594 	u32 spi = ntohl(x->id.spi);
2595 
2596 	if (ctx)
2597 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2598 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2599 
2600 	switch (x->props.family) {
2601 	case AF_INET:
2602 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2603 				 &x->props.saddr.a4, &x->id.daddr.a4);
2604 		break;
2605 	case AF_INET6:
2606 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2607 				 x->props.saddr.a6, x->id.daddr.a6);
2608 		break;
2609 	}
2610 
2611 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2612 }
2613 
2614 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2615 				      struct audit_buffer *audit_buf)
2616 {
2617 	const struct iphdr *iph4;
2618 	const struct ipv6hdr *iph6;
2619 
2620 	switch (family) {
2621 	case AF_INET:
2622 		iph4 = ip_hdr(skb);
2623 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2624 				 &iph4->saddr, &iph4->daddr);
2625 		break;
2626 	case AF_INET6:
2627 		iph6 = ipv6_hdr(skb);
2628 		audit_log_format(audit_buf,
2629 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2630 				 &iph6->saddr, &iph6->daddr,
2631 				 iph6->flow_lbl[0] & 0x0f,
2632 				 iph6->flow_lbl[1],
2633 				 iph6->flow_lbl[2]);
2634 		break;
2635 	}
2636 }
2637 
2638 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2639 {
2640 	struct audit_buffer *audit_buf;
2641 
2642 	audit_buf = xfrm_audit_start("SAD-add");
2643 	if (audit_buf == NULL)
2644 		return;
2645 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2646 	xfrm_audit_helper_sainfo(x, audit_buf);
2647 	audit_log_format(audit_buf, " res=%u", result);
2648 	audit_log_end(audit_buf);
2649 }
2650 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2651 
2652 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2653 {
2654 	struct audit_buffer *audit_buf;
2655 
2656 	audit_buf = xfrm_audit_start("SAD-delete");
2657 	if (audit_buf == NULL)
2658 		return;
2659 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2660 	xfrm_audit_helper_sainfo(x, audit_buf);
2661 	audit_log_format(audit_buf, " res=%u", result);
2662 	audit_log_end(audit_buf);
2663 }
2664 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2665 
2666 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2667 				      struct sk_buff *skb)
2668 {
2669 	struct audit_buffer *audit_buf;
2670 	u32 spi;
2671 
2672 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2673 	if (audit_buf == NULL)
2674 		return;
2675 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2676 	/* don't record the sequence number because it's inherent in this kind
2677 	 * of audit message */
2678 	spi = ntohl(x->id.spi);
2679 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2680 	audit_log_end(audit_buf);
2681 }
2682 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2683 
2684 void xfrm_audit_state_replay(struct xfrm_state *x,
2685 			     struct sk_buff *skb, __be32 net_seq)
2686 {
2687 	struct audit_buffer *audit_buf;
2688 	u32 spi;
2689 
2690 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2691 	if (audit_buf == NULL)
2692 		return;
2693 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2694 	spi = ntohl(x->id.spi);
2695 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2696 			 spi, spi, ntohl(net_seq));
2697 	audit_log_end(audit_buf);
2698 }
2699 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2700 
2701 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2702 {
2703 	struct audit_buffer *audit_buf;
2704 
2705 	audit_buf = xfrm_audit_start("SA-notfound");
2706 	if (audit_buf == NULL)
2707 		return;
2708 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2709 	audit_log_end(audit_buf);
2710 }
2711 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2712 
2713 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2714 			       __be32 net_spi, __be32 net_seq)
2715 {
2716 	struct audit_buffer *audit_buf;
2717 	u32 spi;
2718 
2719 	audit_buf = xfrm_audit_start("SA-notfound");
2720 	if (audit_buf == NULL)
2721 		return;
2722 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2723 	spi = ntohl(net_spi);
2724 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2725 			 spi, spi, ntohl(net_seq));
2726 	audit_log_end(audit_buf);
2727 }
2728 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2729 
2730 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2731 			      struct sk_buff *skb, u8 proto)
2732 {
2733 	struct audit_buffer *audit_buf;
2734 	__be32 net_spi;
2735 	__be32 net_seq;
2736 
2737 	audit_buf = xfrm_audit_start("SA-icv-failure");
2738 	if (audit_buf == NULL)
2739 		return;
2740 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2741 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2742 		u32 spi = ntohl(net_spi);
2743 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2744 				 spi, spi, ntohl(net_seq));
2745 	}
2746 	audit_log_end(audit_buf);
2747 }
2748 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2749 #endif /* CONFIG_AUDITSYSCALL */
2750