xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 2208f39c)
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 		     (x->sel.family != family ||
1023 		      !xfrm_selector_match(&x->sel, fl, family))) ||
1024 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
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, fl))
1040 			*error = -ESRCH;
1041 	}
1042 }
1043 
1044 struct xfrm_state *
1045 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1046 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1047 		struct xfrm_policy *pol, int *err,
1048 		unsigned short family, u32 if_id)
1049 {
1050 	static xfrm_address_t saddr_wildcard = { };
1051 	struct net *net = xp_net(pol);
1052 	unsigned int h, h_wildcard;
1053 	struct xfrm_state *x, *x0, *to_put;
1054 	int acquire_in_progress = 0;
1055 	int error = 0;
1056 	struct xfrm_state *best = NULL;
1057 	u32 mark = pol->mark.v & pol->mark.m;
1058 	unsigned short encap_family = tmpl->encap_family;
1059 	unsigned int sequence;
1060 	struct km_event c;
1061 
1062 	to_put = NULL;
1063 
1064 	sequence = read_seqcount_begin(&xfrm_state_hash_generation);
1065 
1066 	rcu_read_lock();
1067 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1068 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1069 		if (x->props.family == encap_family &&
1070 		    x->props.reqid == tmpl->reqid &&
1071 		    (mark & x->mark.m) == x->mark.v &&
1072 		    x->if_id == if_id &&
1073 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1074 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1075 		    tmpl->mode == x->props.mode &&
1076 		    tmpl->id.proto == x->id.proto &&
1077 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1078 			xfrm_state_look_at(pol, x, fl, family,
1079 					   &best, &acquire_in_progress, &error);
1080 	}
1081 	if (best || acquire_in_progress)
1082 		goto found;
1083 
1084 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1085 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1086 		if (x->props.family == encap_family &&
1087 		    x->props.reqid == tmpl->reqid &&
1088 		    (mark & x->mark.m) == x->mark.v &&
1089 		    x->if_id == if_id &&
1090 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1091 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1092 		    tmpl->mode == x->props.mode &&
1093 		    tmpl->id.proto == x->id.proto &&
1094 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1095 			xfrm_state_look_at(pol, x, fl, family,
1096 					   &best, &acquire_in_progress, &error);
1097 	}
1098 
1099 found:
1100 	x = best;
1101 	if (!x && !error && !acquire_in_progress) {
1102 		if (tmpl->id.spi &&
1103 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1104 					      tmpl->id.proto, encap_family)) != NULL) {
1105 			to_put = x0;
1106 			error = -EEXIST;
1107 			goto out;
1108 		}
1109 
1110 		c.net = net;
1111 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1112 		 * for each and every packet - garbage collection might not
1113 		 * handle the flood.
1114 		 */
1115 		if (!km_is_alive(&c)) {
1116 			error = -ESRCH;
1117 			goto out;
1118 		}
1119 
1120 		x = xfrm_state_alloc(net);
1121 		if (x == NULL) {
1122 			error = -ENOMEM;
1123 			goto out;
1124 		}
1125 		/* Initialize temporary state matching only
1126 		 * to current session. */
1127 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1128 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1129 		x->if_id = if_id;
1130 
1131 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1132 		if (error) {
1133 			x->km.state = XFRM_STATE_DEAD;
1134 			to_put = x;
1135 			x = NULL;
1136 			goto out;
1137 		}
1138 
1139 		if (km_query(x, tmpl, pol) == 0) {
1140 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1141 			x->km.state = XFRM_STATE_ACQ;
1142 			list_add(&x->km.all, &net->xfrm.state_all);
1143 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1144 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1145 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1146 			if (x->id.spi) {
1147 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1148 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1149 			}
1150 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1151 			hrtimer_start(&x->mtimer,
1152 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1153 				      HRTIMER_MODE_REL_SOFT);
1154 			net->xfrm.state_num++;
1155 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1156 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1157 		} else {
1158 			x->km.state = XFRM_STATE_DEAD;
1159 			to_put = x;
1160 			x = NULL;
1161 			error = -ESRCH;
1162 		}
1163 	}
1164 out:
1165 	if (x) {
1166 		if (!xfrm_state_hold_rcu(x)) {
1167 			*err = -EAGAIN;
1168 			x = NULL;
1169 		}
1170 	} else {
1171 		*err = acquire_in_progress ? -EAGAIN : error;
1172 	}
1173 	rcu_read_unlock();
1174 	if (to_put)
1175 		xfrm_state_put(to_put);
1176 
1177 	if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1178 		*err = -EAGAIN;
1179 		if (x) {
1180 			xfrm_state_put(x);
1181 			x = NULL;
1182 		}
1183 	}
1184 
1185 	return x;
1186 }
1187 
1188 struct xfrm_state *
1189 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1190 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1191 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1192 {
1193 	unsigned int h;
1194 	struct xfrm_state *rx = NULL, *x = NULL;
1195 
1196 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1197 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1198 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1199 		if (x->props.family == family &&
1200 		    x->props.reqid == reqid &&
1201 		    (mark & x->mark.m) == x->mark.v &&
1202 		    x->if_id == if_id &&
1203 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1204 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1205 		    mode == x->props.mode &&
1206 		    proto == x->id.proto &&
1207 		    x->km.state == XFRM_STATE_VALID) {
1208 			rx = x;
1209 			break;
1210 		}
1211 	}
1212 
1213 	if (rx)
1214 		xfrm_state_hold(rx);
1215 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1216 
1217 
1218 	return rx;
1219 }
1220 EXPORT_SYMBOL(xfrm_stateonly_find);
1221 
1222 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1223 					      unsigned short family)
1224 {
1225 	struct xfrm_state *x;
1226 	struct xfrm_state_walk *w;
1227 
1228 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1229 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1230 		x = container_of(w, struct xfrm_state, km);
1231 		if (x->props.family != family ||
1232 			x->id.spi != spi)
1233 			continue;
1234 
1235 		xfrm_state_hold(x);
1236 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1237 		return x;
1238 	}
1239 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1240 	return NULL;
1241 }
1242 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1243 
1244 static void __xfrm_state_insert(struct xfrm_state *x)
1245 {
1246 	struct net *net = xs_net(x);
1247 	unsigned int h;
1248 
1249 	list_add(&x->km.all, &net->xfrm.state_all);
1250 
1251 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1252 			  x->props.reqid, x->props.family);
1253 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1254 
1255 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1256 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1257 
1258 	if (x->id.spi) {
1259 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1260 				  x->props.family);
1261 
1262 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1263 	}
1264 
1265 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1266 	if (x->replay_maxage)
1267 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1268 
1269 	net->xfrm.state_num++;
1270 
1271 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1272 }
1273 
1274 /* net->xfrm.xfrm_state_lock is held */
1275 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1276 {
1277 	struct net *net = xs_net(xnew);
1278 	unsigned short family = xnew->props.family;
1279 	u32 reqid = xnew->props.reqid;
1280 	struct xfrm_state *x;
1281 	unsigned int h;
1282 	u32 mark = xnew->mark.v & xnew->mark.m;
1283 	u32 if_id = xnew->if_id;
1284 
1285 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1286 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1287 		if (x->props.family	== family &&
1288 		    x->props.reqid	== reqid &&
1289 		    x->if_id		== if_id &&
1290 		    (mark & x->mark.m) == x->mark.v &&
1291 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1292 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1293 			x->genid++;
1294 	}
1295 }
1296 
1297 void xfrm_state_insert(struct xfrm_state *x)
1298 {
1299 	struct net *net = xs_net(x);
1300 
1301 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1302 	__xfrm_state_bump_genids(x);
1303 	__xfrm_state_insert(x);
1304 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1305 }
1306 EXPORT_SYMBOL(xfrm_state_insert);
1307 
1308 /* net->xfrm.xfrm_state_lock is held */
1309 static struct xfrm_state *__find_acq_core(struct net *net,
1310 					  const struct xfrm_mark *m,
1311 					  unsigned short family, u8 mode,
1312 					  u32 reqid, u32 if_id, u8 proto,
1313 					  const xfrm_address_t *daddr,
1314 					  const xfrm_address_t *saddr,
1315 					  int create)
1316 {
1317 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1318 	struct xfrm_state *x;
1319 	u32 mark = m->v & m->m;
1320 
1321 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1322 		if (x->props.reqid  != reqid ||
1323 		    x->props.mode   != mode ||
1324 		    x->props.family != family ||
1325 		    x->km.state     != XFRM_STATE_ACQ ||
1326 		    x->id.spi       != 0 ||
1327 		    x->id.proto	    != proto ||
1328 		    (mark & x->mark.m) != x->mark.v ||
1329 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1330 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1331 			continue;
1332 
1333 		xfrm_state_hold(x);
1334 		return x;
1335 	}
1336 
1337 	if (!create)
1338 		return NULL;
1339 
1340 	x = xfrm_state_alloc(net);
1341 	if (likely(x)) {
1342 		switch (family) {
1343 		case AF_INET:
1344 			x->sel.daddr.a4 = daddr->a4;
1345 			x->sel.saddr.a4 = saddr->a4;
1346 			x->sel.prefixlen_d = 32;
1347 			x->sel.prefixlen_s = 32;
1348 			x->props.saddr.a4 = saddr->a4;
1349 			x->id.daddr.a4 = daddr->a4;
1350 			break;
1351 
1352 		case AF_INET6:
1353 			x->sel.daddr.in6 = daddr->in6;
1354 			x->sel.saddr.in6 = saddr->in6;
1355 			x->sel.prefixlen_d = 128;
1356 			x->sel.prefixlen_s = 128;
1357 			x->props.saddr.in6 = saddr->in6;
1358 			x->id.daddr.in6 = daddr->in6;
1359 			break;
1360 		}
1361 
1362 		x->km.state = XFRM_STATE_ACQ;
1363 		x->id.proto = proto;
1364 		x->props.family = family;
1365 		x->props.mode = mode;
1366 		x->props.reqid = reqid;
1367 		x->if_id = if_id;
1368 		x->mark.v = m->v;
1369 		x->mark.m = m->m;
1370 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1371 		xfrm_state_hold(x);
1372 		hrtimer_start(&x->mtimer,
1373 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1374 			      HRTIMER_MODE_REL_SOFT);
1375 		list_add(&x->km.all, &net->xfrm.state_all);
1376 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1377 		h = xfrm_src_hash(net, daddr, saddr, family);
1378 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1379 
1380 		net->xfrm.state_num++;
1381 
1382 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1383 	}
1384 
1385 	return x;
1386 }
1387 
1388 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1389 
1390 int xfrm_state_add(struct xfrm_state *x)
1391 {
1392 	struct net *net = xs_net(x);
1393 	struct xfrm_state *x1, *to_put;
1394 	int family;
1395 	int err;
1396 	u32 mark = x->mark.v & x->mark.m;
1397 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1398 
1399 	family = x->props.family;
1400 
1401 	to_put = NULL;
1402 
1403 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1404 
1405 	x1 = __xfrm_state_locate(x, use_spi, family);
1406 	if (x1) {
1407 		to_put = x1;
1408 		x1 = NULL;
1409 		err = -EEXIST;
1410 		goto out;
1411 	}
1412 
1413 	if (use_spi && x->km.seq) {
1414 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1415 		if (x1 && ((x1->id.proto != x->id.proto) ||
1416 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1417 			to_put = x1;
1418 			x1 = NULL;
1419 		}
1420 	}
1421 
1422 	if (use_spi && !x1)
1423 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1424 				     x->props.reqid, x->if_id, x->id.proto,
1425 				     &x->id.daddr, &x->props.saddr, 0);
1426 
1427 	__xfrm_state_bump_genids(x);
1428 	__xfrm_state_insert(x);
1429 	err = 0;
1430 
1431 out:
1432 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1433 
1434 	if (x1) {
1435 		xfrm_state_delete(x1);
1436 		xfrm_state_put(x1);
1437 	}
1438 
1439 	if (to_put)
1440 		xfrm_state_put(to_put);
1441 
1442 	return err;
1443 }
1444 EXPORT_SYMBOL(xfrm_state_add);
1445 
1446 #ifdef CONFIG_XFRM_MIGRATE
1447 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1448 {
1449 	struct xfrm_user_sec_ctx *uctx;
1450 	int size = sizeof(*uctx) + security->ctx_len;
1451 	int err;
1452 
1453 	uctx = kmalloc(size, GFP_KERNEL);
1454 	if (!uctx)
1455 		return -ENOMEM;
1456 
1457 	uctx->exttype = XFRMA_SEC_CTX;
1458 	uctx->len = size;
1459 	uctx->ctx_doi = security->ctx_doi;
1460 	uctx->ctx_alg = security->ctx_alg;
1461 	uctx->ctx_len = security->ctx_len;
1462 	memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1463 	err = security_xfrm_state_alloc(x, uctx);
1464 	kfree(uctx);
1465 	if (err)
1466 		return err;
1467 
1468 	return 0;
1469 }
1470 
1471 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1472 					   struct xfrm_encap_tmpl *encap)
1473 {
1474 	struct net *net = xs_net(orig);
1475 	struct xfrm_state *x = xfrm_state_alloc(net);
1476 	if (!x)
1477 		goto out;
1478 
1479 	memcpy(&x->id, &orig->id, sizeof(x->id));
1480 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1481 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1482 	x->props.mode = orig->props.mode;
1483 	x->props.replay_window = orig->props.replay_window;
1484 	x->props.reqid = orig->props.reqid;
1485 	x->props.family = orig->props.family;
1486 	x->props.saddr = orig->props.saddr;
1487 
1488 	if (orig->aalg) {
1489 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1490 		if (!x->aalg)
1491 			goto error;
1492 	}
1493 	x->props.aalgo = orig->props.aalgo;
1494 
1495 	if (orig->aead) {
1496 		x->aead = xfrm_algo_aead_clone(orig->aead);
1497 		x->geniv = orig->geniv;
1498 		if (!x->aead)
1499 			goto error;
1500 	}
1501 	if (orig->ealg) {
1502 		x->ealg = xfrm_algo_clone(orig->ealg);
1503 		if (!x->ealg)
1504 			goto error;
1505 	}
1506 	x->props.ealgo = orig->props.ealgo;
1507 
1508 	if (orig->calg) {
1509 		x->calg = xfrm_algo_clone(orig->calg);
1510 		if (!x->calg)
1511 			goto error;
1512 	}
1513 	x->props.calgo = orig->props.calgo;
1514 
1515 	if (encap || orig->encap) {
1516 		if (encap)
1517 			x->encap = kmemdup(encap, sizeof(*x->encap),
1518 					GFP_KERNEL);
1519 		else
1520 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1521 					GFP_KERNEL);
1522 
1523 		if (!x->encap)
1524 			goto error;
1525 	}
1526 
1527 	if (orig->security)
1528 		if (clone_security(x, orig->security))
1529 			goto error;
1530 
1531 	if (orig->coaddr) {
1532 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1533 				    GFP_KERNEL);
1534 		if (!x->coaddr)
1535 			goto error;
1536 	}
1537 
1538 	if (orig->replay_esn) {
1539 		if (xfrm_replay_clone(x, orig))
1540 			goto error;
1541 	}
1542 
1543 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1544 	memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1545 
1546 	if (xfrm_init_state(x) < 0)
1547 		goto error;
1548 
1549 	x->props.flags = orig->props.flags;
1550 	x->props.extra_flags = orig->props.extra_flags;
1551 
1552 	x->if_id = orig->if_id;
1553 	x->tfcpad = orig->tfcpad;
1554 	x->replay_maxdiff = orig->replay_maxdiff;
1555 	x->replay_maxage = orig->replay_maxage;
1556 	memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1557 	x->km.state = orig->km.state;
1558 	x->km.seq = orig->km.seq;
1559 	x->replay = orig->replay;
1560 	x->preplay = orig->preplay;
1561 
1562 	return x;
1563 
1564  error:
1565 	xfrm_state_put(x);
1566 out:
1567 	return NULL;
1568 }
1569 
1570 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1571 {
1572 	unsigned int h;
1573 	struct xfrm_state *x = NULL;
1574 
1575 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1576 
1577 	if (m->reqid) {
1578 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1579 				  m->reqid, m->old_family);
1580 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1581 			if (x->props.mode != m->mode ||
1582 			    x->id.proto != m->proto)
1583 				continue;
1584 			if (m->reqid && x->props.reqid != m->reqid)
1585 				continue;
1586 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1587 					     m->old_family) ||
1588 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1589 					     m->old_family))
1590 				continue;
1591 			xfrm_state_hold(x);
1592 			break;
1593 		}
1594 	} else {
1595 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1596 				  m->old_family);
1597 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1598 			if (x->props.mode != m->mode ||
1599 			    x->id.proto != m->proto)
1600 				continue;
1601 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1602 					     m->old_family) ||
1603 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1604 					     m->old_family))
1605 				continue;
1606 			xfrm_state_hold(x);
1607 			break;
1608 		}
1609 	}
1610 
1611 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1612 
1613 	return x;
1614 }
1615 EXPORT_SYMBOL(xfrm_migrate_state_find);
1616 
1617 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1618 				      struct xfrm_migrate *m,
1619 				      struct xfrm_encap_tmpl *encap)
1620 {
1621 	struct xfrm_state *xc;
1622 
1623 	xc = xfrm_state_clone(x, encap);
1624 	if (!xc)
1625 		return NULL;
1626 
1627 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1628 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1629 
1630 	/* add state */
1631 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1632 		/* a care is needed when the destination address of the
1633 		   state is to be updated as it is a part of triplet */
1634 		xfrm_state_insert(xc);
1635 	} else {
1636 		if (xfrm_state_add(xc) < 0)
1637 			goto error;
1638 	}
1639 
1640 	return xc;
1641 error:
1642 	xfrm_state_put(xc);
1643 	return NULL;
1644 }
1645 EXPORT_SYMBOL(xfrm_state_migrate);
1646 #endif
1647 
1648 int xfrm_state_update(struct xfrm_state *x)
1649 {
1650 	struct xfrm_state *x1, *to_put;
1651 	int err;
1652 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1653 	struct net *net = xs_net(x);
1654 
1655 	to_put = NULL;
1656 
1657 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1658 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1659 
1660 	err = -ESRCH;
1661 	if (!x1)
1662 		goto out;
1663 
1664 	if (xfrm_state_kern(x1)) {
1665 		to_put = x1;
1666 		err = -EEXIST;
1667 		goto out;
1668 	}
1669 
1670 	if (x1->km.state == XFRM_STATE_ACQ) {
1671 		__xfrm_state_insert(x);
1672 		x = NULL;
1673 	}
1674 	err = 0;
1675 
1676 out:
1677 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1678 
1679 	if (to_put)
1680 		xfrm_state_put(to_put);
1681 
1682 	if (err)
1683 		return err;
1684 
1685 	if (!x) {
1686 		xfrm_state_delete(x1);
1687 		xfrm_state_put(x1);
1688 		return 0;
1689 	}
1690 
1691 	err = -EINVAL;
1692 	spin_lock_bh(&x1->lock);
1693 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1694 		if (x->encap && x1->encap &&
1695 		    x->encap->encap_type == x1->encap->encap_type)
1696 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1697 		else if (x->encap || x1->encap)
1698 			goto fail;
1699 
1700 		if (x->coaddr && x1->coaddr) {
1701 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1702 		}
1703 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1704 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1705 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1706 		x1->km.dying = 0;
1707 
1708 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1709 			      HRTIMER_MODE_REL_SOFT);
1710 		if (x1->curlft.use_time)
1711 			xfrm_state_check_expire(x1);
1712 
1713 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1714 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1715 
1716 			if (x->props.smark.m || x->props.smark.v)
1717 				x1->props.smark = x->props.smark;
1718 
1719 			if (x->if_id)
1720 				x1->if_id = x->if_id;
1721 
1722 			__xfrm_state_bump_genids(x1);
1723 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1724 		}
1725 
1726 		err = 0;
1727 		x->km.state = XFRM_STATE_DEAD;
1728 		__xfrm_state_put(x);
1729 	}
1730 
1731 fail:
1732 	spin_unlock_bh(&x1->lock);
1733 
1734 	xfrm_state_put(x1);
1735 
1736 	return err;
1737 }
1738 EXPORT_SYMBOL(xfrm_state_update);
1739 
1740 int xfrm_state_check_expire(struct xfrm_state *x)
1741 {
1742 	if (!x->curlft.use_time)
1743 		x->curlft.use_time = ktime_get_real_seconds();
1744 
1745 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1746 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1747 		x->km.state = XFRM_STATE_EXPIRED;
1748 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1749 		return -EINVAL;
1750 	}
1751 
1752 	if (!x->km.dying &&
1753 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1754 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1755 		x->km.dying = 1;
1756 		km_state_expired(x, 0, 0);
1757 	}
1758 	return 0;
1759 }
1760 EXPORT_SYMBOL(xfrm_state_check_expire);
1761 
1762 struct xfrm_state *
1763 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1764 		  u8 proto, unsigned short family)
1765 {
1766 	struct xfrm_state *x;
1767 
1768 	rcu_read_lock();
1769 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1770 	rcu_read_unlock();
1771 	return x;
1772 }
1773 EXPORT_SYMBOL(xfrm_state_lookup);
1774 
1775 struct xfrm_state *
1776 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1777 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1778 			 u8 proto, unsigned short family)
1779 {
1780 	struct xfrm_state *x;
1781 
1782 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1783 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1784 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1785 	return x;
1786 }
1787 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1788 
1789 struct xfrm_state *
1790 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1791 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1792 	      const xfrm_address_t *saddr, int create, unsigned short family)
1793 {
1794 	struct xfrm_state *x;
1795 
1796 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1797 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1798 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1799 
1800 	return x;
1801 }
1802 EXPORT_SYMBOL(xfrm_find_acq);
1803 
1804 #ifdef CONFIG_XFRM_SUB_POLICY
1805 #if IS_ENABLED(CONFIG_IPV6)
1806 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1807 static void
1808 __xfrm6_sort(void **dst, void **src, int n,
1809 	     int (*cmp)(const void *p), int maxclass)
1810 {
1811 	int count[XFRM_MAX_DEPTH] = { };
1812 	int class[XFRM_MAX_DEPTH];
1813 	int i;
1814 
1815 	for (i = 0; i < n; i++) {
1816 		int c = cmp(src[i]);
1817 
1818 		class[i] = c;
1819 		count[c]++;
1820 	}
1821 
1822 	for (i = 2; i < maxclass; i++)
1823 		count[i] += count[i - 1];
1824 
1825 	for (i = 0; i < n; i++) {
1826 		dst[count[class[i] - 1]++] = src[i];
1827 		src[i] = NULL;
1828 	}
1829 }
1830 
1831 /* Rule for xfrm_state:
1832  *
1833  * rule 1: select IPsec transport except AH
1834  * rule 2: select MIPv6 RO or inbound trigger
1835  * rule 3: select IPsec transport AH
1836  * rule 4: select IPsec tunnel
1837  * rule 5: others
1838  */
1839 static int __xfrm6_state_sort_cmp(const void *p)
1840 {
1841 	const struct xfrm_state *v = p;
1842 
1843 	switch (v->props.mode) {
1844 	case XFRM_MODE_TRANSPORT:
1845 		if (v->id.proto != IPPROTO_AH)
1846 			return 1;
1847 		else
1848 			return 3;
1849 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1850 	case XFRM_MODE_ROUTEOPTIMIZATION:
1851 	case XFRM_MODE_IN_TRIGGER:
1852 		return 2;
1853 #endif
1854 	case XFRM_MODE_TUNNEL:
1855 	case XFRM_MODE_BEET:
1856 		return 4;
1857 	}
1858 	return 5;
1859 }
1860 
1861 /* Rule for xfrm_tmpl:
1862  *
1863  * rule 1: select IPsec transport
1864  * rule 2: select MIPv6 RO or inbound trigger
1865  * rule 3: select IPsec tunnel
1866  * rule 4: others
1867  */
1868 static int __xfrm6_tmpl_sort_cmp(const void *p)
1869 {
1870 	const struct xfrm_tmpl *v = p;
1871 
1872 	switch (v->mode) {
1873 	case XFRM_MODE_TRANSPORT:
1874 		return 1;
1875 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1876 	case XFRM_MODE_ROUTEOPTIMIZATION:
1877 	case XFRM_MODE_IN_TRIGGER:
1878 		return 2;
1879 #endif
1880 	case XFRM_MODE_TUNNEL:
1881 	case XFRM_MODE_BEET:
1882 		return 3;
1883 	}
1884 	return 4;
1885 }
1886 #else
1887 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1888 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1889 
1890 static inline void
1891 __xfrm6_sort(void **dst, void **src, int n,
1892 	     int (*cmp)(const void *p), int maxclass)
1893 {
1894 	int i;
1895 
1896 	for (i = 0; i < n; i++)
1897 		dst[i] = src[i];
1898 }
1899 #endif /* CONFIG_IPV6 */
1900 
1901 void
1902 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1903 	       unsigned short family)
1904 {
1905 	int i;
1906 
1907 	if (family == AF_INET6)
1908 		__xfrm6_sort((void **)dst, (void **)src, n,
1909 			     __xfrm6_tmpl_sort_cmp, 5);
1910 	else
1911 		for (i = 0; i < n; i++)
1912 			dst[i] = src[i];
1913 }
1914 
1915 void
1916 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1917 		unsigned short family)
1918 {
1919 	int i;
1920 
1921 	if (family == AF_INET6)
1922 		__xfrm6_sort((void **)dst, (void **)src, n,
1923 			     __xfrm6_state_sort_cmp, 6);
1924 	else
1925 		for (i = 0; i < n; i++)
1926 			dst[i] = src[i];
1927 }
1928 #endif
1929 
1930 /* Silly enough, but I'm lazy to build resolution list */
1931 
1932 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1933 {
1934 	int i;
1935 
1936 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1937 		struct xfrm_state *x;
1938 
1939 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1940 			if (x->km.seq == seq &&
1941 			    (mark & x->mark.m) == x->mark.v &&
1942 			    x->km.state == XFRM_STATE_ACQ) {
1943 				xfrm_state_hold(x);
1944 				return x;
1945 			}
1946 		}
1947 	}
1948 	return NULL;
1949 }
1950 
1951 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1952 {
1953 	struct xfrm_state *x;
1954 
1955 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1956 	x = __xfrm_find_acq_byseq(net, mark, seq);
1957 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1958 	return x;
1959 }
1960 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1961 
1962 u32 xfrm_get_acqseq(void)
1963 {
1964 	u32 res;
1965 	static atomic_t acqseq;
1966 
1967 	do {
1968 		res = atomic_inc_return(&acqseq);
1969 	} while (!res);
1970 
1971 	return res;
1972 }
1973 EXPORT_SYMBOL(xfrm_get_acqseq);
1974 
1975 int verify_spi_info(u8 proto, u32 min, u32 max)
1976 {
1977 	switch (proto) {
1978 	case IPPROTO_AH:
1979 	case IPPROTO_ESP:
1980 		break;
1981 
1982 	case IPPROTO_COMP:
1983 		/* IPCOMP spi is 16-bits. */
1984 		if (max >= 0x10000)
1985 			return -EINVAL;
1986 		break;
1987 
1988 	default:
1989 		return -EINVAL;
1990 	}
1991 
1992 	if (min > max)
1993 		return -EINVAL;
1994 
1995 	return 0;
1996 }
1997 EXPORT_SYMBOL(verify_spi_info);
1998 
1999 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2000 {
2001 	struct net *net = xs_net(x);
2002 	unsigned int h;
2003 	struct xfrm_state *x0;
2004 	int err = -ENOENT;
2005 	__be32 minspi = htonl(low);
2006 	__be32 maxspi = htonl(high);
2007 	__be32 newspi = 0;
2008 	u32 mark = x->mark.v & x->mark.m;
2009 
2010 	spin_lock_bh(&x->lock);
2011 	if (x->km.state == XFRM_STATE_DEAD)
2012 		goto unlock;
2013 
2014 	err = 0;
2015 	if (x->id.spi)
2016 		goto unlock;
2017 
2018 	err = -ENOENT;
2019 
2020 	if (minspi == maxspi) {
2021 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2022 		if (x0) {
2023 			xfrm_state_put(x0);
2024 			goto unlock;
2025 		}
2026 		newspi = minspi;
2027 	} else {
2028 		u32 spi = 0;
2029 		for (h = 0; h < high-low+1; h++) {
2030 			spi = low + prandom_u32()%(high-low+1);
2031 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2032 			if (x0 == NULL) {
2033 				newspi = htonl(spi);
2034 				break;
2035 			}
2036 			xfrm_state_put(x0);
2037 		}
2038 	}
2039 	if (newspi) {
2040 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2041 		x->id.spi = newspi;
2042 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2043 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2044 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2045 
2046 		err = 0;
2047 	}
2048 
2049 unlock:
2050 	spin_unlock_bh(&x->lock);
2051 
2052 	return err;
2053 }
2054 EXPORT_SYMBOL(xfrm_alloc_spi);
2055 
2056 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2057 				      struct xfrm_address_filter *filter)
2058 {
2059 	if (filter) {
2060 		if ((filter->family == AF_INET ||
2061 		     filter->family == AF_INET6) &&
2062 		    x->props.family != filter->family)
2063 			return false;
2064 
2065 		return addr_match(&x->props.saddr, &filter->saddr,
2066 				  filter->splen) &&
2067 		       addr_match(&x->id.daddr, &filter->daddr,
2068 				  filter->dplen);
2069 	}
2070 	return true;
2071 }
2072 
2073 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2074 		    int (*func)(struct xfrm_state *, int, void*),
2075 		    void *data)
2076 {
2077 	struct xfrm_state *state;
2078 	struct xfrm_state_walk *x;
2079 	int err = 0;
2080 
2081 	if (walk->seq != 0 && list_empty(&walk->all))
2082 		return 0;
2083 
2084 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2085 	if (list_empty(&walk->all))
2086 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2087 	else
2088 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2089 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2090 		if (x->state == XFRM_STATE_DEAD)
2091 			continue;
2092 		state = container_of(x, struct xfrm_state, km);
2093 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2094 			continue;
2095 		if (!__xfrm_state_filter_match(state, walk->filter))
2096 			continue;
2097 		err = func(state, walk->seq, data);
2098 		if (err) {
2099 			list_move_tail(&walk->all, &x->all);
2100 			goto out;
2101 		}
2102 		walk->seq++;
2103 	}
2104 	if (walk->seq == 0) {
2105 		err = -ENOENT;
2106 		goto out;
2107 	}
2108 	list_del_init(&walk->all);
2109 out:
2110 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2111 	return err;
2112 }
2113 EXPORT_SYMBOL(xfrm_state_walk);
2114 
2115 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2116 			  struct xfrm_address_filter *filter)
2117 {
2118 	INIT_LIST_HEAD(&walk->all);
2119 	walk->proto = proto;
2120 	walk->state = XFRM_STATE_DEAD;
2121 	walk->seq = 0;
2122 	walk->filter = filter;
2123 }
2124 EXPORT_SYMBOL(xfrm_state_walk_init);
2125 
2126 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2127 {
2128 	kfree(walk->filter);
2129 
2130 	if (list_empty(&walk->all))
2131 		return;
2132 
2133 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2134 	list_del(&walk->all);
2135 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2136 }
2137 EXPORT_SYMBOL(xfrm_state_walk_done);
2138 
2139 static void xfrm_replay_timer_handler(struct timer_list *t)
2140 {
2141 	struct xfrm_state *x = from_timer(x, t, rtimer);
2142 
2143 	spin_lock(&x->lock);
2144 
2145 	if (x->km.state == XFRM_STATE_VALID) {
2146 		if (xfrm_aevent_is_on(xs_net(x)))
2147 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2148 		else
2149 			x->xflags |= XFRM_TIME_DEFER;
2150 	}
2151 
2152 	spin_unlock(&x->lock);
2153 }
2154 
2155 static LIST_HEAD(xfrm_km_list);
2156 
2157 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2158 {
2159 	struct xfrm_mgr *km;
2160 
2161 	rcu_read_lock();
2162 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2163 		if (km->notify_policy)
2164 			km->notify_policy(xp, dir, c);
2165 	rcu_read_unlock();
2166 }
2167 
2168 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2169 {
2170 	struct xfrm_mgr *km;
2171 	rcu_read_lock();
2172 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2173 		if (km->notify)
2174 			km->notify(x, c);
2175 	rcu_read_unlock();
2176 }
2177 
2178 EXPORT_SYMBOL(km_policy_notify);
2179 EXPORT_SYMBOL(km_state_notify);
2180 
2181 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2182 {
2183 	struct km_event c;
2184 
2185 	c.data.hard = hard;
2186 	c.portid = portid;
2187 	c.event = XFRM_MSG_EXPIRE;
2188 	km_state_notify(x, &c);
2189 }
2190 
2191 EXPORT_SYMBOL(km_state_expired);
2192 /*
2193  * We send to all registered managers regardless of failure
2194  * We are happy with one success
2195 */
2196 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2197 {
2198 	int err = -EINVAL, acqret;
2199 	struct xfrm_mgr *km;
2200 
2201 	rcu_read_lock();
2202 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2203 		acqret = km->acquire(x, t, pol);
2204 		if (!acqret)
2205 			err = acqret;
2206 	}
2207 	rcu_read_unlock();
2208 	return err;
2209 }
2210 EXPORT_SYMBOL(km_query);
2211 
2212 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2213 {
2214 	int err = -EINVAL;
2215 	struct xfrm_mgr *km;
2216 
2217 	rcu_read_lock();
2218 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2219 		if (km->new_mapping)
2220 			err = km->new_mapping(x, ipaddr, sport);
2221 		if (!err)
2222 			break;
2223 	}
2224 	rcu_read_unlock();
2225 	return err;
2226 }
2227 EXPORT_SYMBOL(km_new_mapping);
2228 
2229 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2230 {
2231 	struct km_event c;
2232 
2233 	c.data.hard = hard;
2234 	c.portid = portid;
2235 	c.event = XFRM_MSG_POLEXPIRE;
2236 	km_policy_notify(pol, dir, &c);
2237 }
2238 EXPORT_SYMBOL(km_policy_expired);
2239 
2240 #ifdef CONFIG_XFRM_MIGRATE
2241 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2242 	       const struct xfrm_migrate *m, int num_migrate,
2243 	       const struct xfrm_kmaddress *k,
2244 	       const struct xfrm_encap_tmpl *encap)
2245 {
2246 	int err = -EINVAL;
2247 	int ret;
2248 	struct xfrm_mgr *km;
2249 
2250 	rcu_read_lock();
2251 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2252 		if (km->migrate) {
2253 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2254 					  encap);
2255 			if (!ret)
2256 				err = ret;
2257 		}
2258 	}
2259 	rcu_read_unlock();
2260 	return err;
2261 }
2262 EXPORT_SYMBOL(km_migrate);
2263 #endif
2264 
2265 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2266 {
2267 	int err = -EINVAL;
2268 	int ret;
2269 	struct xfrm_mgr *km;
2270 
2271 	rcu_read_lock();
2272 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2273 		if (km->report) {
2274 			ret = km->report(net, proto, sel, addr);
2275 			if (!ret)
2276 				err = ret;
2277 		}
2278 	}
2279 	rcu_read_unlock();
2280 	return err;
2281 }
2282 EXPORT_SYMBOL(km_report);
2283 
2284 static bool km_is_alive(const struct km_event *c)
2285 {
2286 	struct xfrm_mgr *km;
2287 	bool is_alive = false;
2288 
2289 	rcu_read_lock();
2290 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2291 		if (km->is_alive && km->is_alive(c)) {
2292 			is_alive = true;
2293 			break;
2294 		}
2295 	}
2296 	rcu_read_unlock();
2297 
2298 	return is_alive;
2299 }
2300 
2301 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2302 static DEFINE_SPINLOCK(xfrm_translator_lock);
2303 static struct xfrm_translator __rcu *xfrm_translator;
2304 
2305 struct xfrm_translator *xfrm_get_translator(void)
2306 {
2307 	struct xfrm_translator *xtr;
2308 
2309 	rcu_read_lock();
2310 	xtr = rcu_dereference(xfrm_translator);
2311 	if (unlikely(!xtr))
2312 		goto out;
2313 	if (!try_module_get(xtr->owner))
2314 		xtr = NULL;
2315 out:
2316 	rcu_read_unlock();
2317 	return xtr;
2318 }
2319 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2320 
2321 void xfrm_put_translator(struct xfrm_translator *xtr)
2322 {
2323 	module_put(xtr->owner);
2324 }
2325 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2326 
2327 int xfrm_register_translator(struct xfrm_translator *xtr)
2328 {
2329 	int err = 0;
2330 
2331 	spin_lock_bh(&xfrm_translator_lock);
2332 	if (unlikely(xfrm_translator != NULL))
2333 		err = -EEXIST;
2334 	else
2335 		rcu_assign_pointer(xfrm_translator, xtr);
2336 	spin_unlock_bh(&xfrm_translator_lock);
2337 
2338 	return err;
2339 }
2340 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2341 
2342 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2343 {
2344 	int err = 0;
2345 
2346 	spin_lock_bh(&xfrm_translator_lock);
2347 	if (likely(xfrm_translator != NULL)) {
2348 		if (rcu_access_pointer(xfrm_translator) != xtr)
2349 			err = -EINVAL;
2350 		else
2351 			RCU_INIT_POINTER(xfrm_translator, NULL);
2352 	}
2353 	spin_unlock_bh(&xfrm_translator_lock);
2354 	synchronize_rcu();
2355 
2356 	return err;
2357 }
2358 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2359 #endif
2360 
2361 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2362 {
2363 	int err;
2364 	u8 *data;
2365 	struct xfrm_mgr *km;
2366 	struct xfrm_policy *pol = NULL;
2367 
2368 	if (sockptr_is_null(optval) && !optlen) {
2369 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2370 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2371 		__sk_dst_reset(sk);
2372 		return 0;
2373 	}
2374 
2375 	if (optlen <= 0 || optlen > PAGE_SIZE)
2376 		return -EMSGSIZE;
2377 
2378 	data = memdup_sockptr(optval, optlen);
2379 	if (IS_ERR(data))
2380 		return PTR_ERR(data);
2381 
2382 	if (in_compat_syscall()) {
2383 		struct xfrm_translator *xtr = xfrm_get_translator();
2384 
2385 		if (!xtr)
2386 			return -EOPNOTSUPP;
2387 
2388 		err = xtr->xlate_user_policy_sockptr(&data, optlen);
2389 		xfrm_put_translator(xtr);
2390 		if (err) {
2391 			kfree(data);
2392 			return err;
2393 		}
2394 	}
2395 
2396 	err = -EINVAL;
2397 	rcu_read_lock();
2398 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2399 		pol = km->compile_policy(sk, optname, data,
2400 					 optlen, &err);
2401 		if (err >= 0)
2402 			break;
2403 	}
2404 	rcu_read_unlock();
2405 
2406 	if (err >= 0) {
2407 		xfrm_sk_policy_insert(sk, err, pol);
2408 		xfrm_pol_put(pol);
2409 		__sk_dst_reset(sk);
2410 		err = 0;
2411 	}
2412 
2413 	kfree(data);
2414 	return err;
2415 }
2416 EXPORT_SYMBOL(xfrm_user_policy);
2417 
2418 static DEFINE_SPINLOCK(xfrm_km_lock);
2419 
2420 int xfrm_register_km(struct xfrm_mgr *km)
2421 {
2422 	spin_lock_bh(&xfrm_km_lock);
2423 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2424 	spin_unlock_bh(&xfrm_km_lock);
2425 	return 0;
2426 }
2427 EXPORT_SYMBOL(xfrm_register_km);
2428 
2429 int xfrm_unregister_km(struct xfrm_mgr *km)
2430 {
2431 	spin_lock_bh(&xfrm_km_lock);
2432 	list_del_rcu(&km->list);
2433 	spin_unlock_bh(&xfrm_km_lock);
2434 	synchronize_rcu();
2435 	return 0;
2436 }
2437 EXPORT_SYMBOL(xfrm_unregister_km);
2438 
2439 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2440 {
2441 	int err = 0;
2442 
2443 	if (WARN_ON(afinfo->family >= NPROTO))
2444 		return -EAFNOSUPPORT;
2445 
2446 	spin_lock_bh(&xfrm_state_afinfo_lock);
2447 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2448 		err = -EEXIST;
2449 	else
2450 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2451 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2452 	return err;
2453 }
2454 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2455 
2456 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2457 {
2458 	int err = 0, family = afinfo->family;
2459 
2460 	if (WARN_ON(family >= NPROTO))
2461 		return -EAFNOSUPPORT;
2462 
2463 	spin_lock_bh(&xfrm_state_afinfo_lock);
2464 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2465 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2466 			err = -EINVAL;
2467 		else
2468 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2469 	}
2470 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2471 	synchronize_rcu();
2472 	return err;
2473 }
2474 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2475 
2476 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2477 {
2478 	if (unlikely(family >= NPROTO))
2479 		return NULL;
2480 
2481 	return rcu_dereference(xfrm_state_afinfo[family]);
2482 }
2483 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2484 
2485 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2486 {
2487 	struct xfrm_state_afinfo *afinfo;
2488 	if (unlikely(family >= NPROTO))
2489 		return NULL;
2490 	rcu_read_lock();
2491 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2492 	if (unlikely(!afinfo))
2493 		rcu_read_unlock();
2494 	return afinfo;
2495 }
2496 
2497 void xfrm_flush_gc(void)
2498 {
2499 	flush_work(&xfrm_state_gc_work);
2500 }
2501 EXPORT_SYMBOL(xfrm_flush_gc);
2502 
2503 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2504 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2505 {
2506 	if (x->tunnel) {
2507 		struct xfrm_state *t = x->tunnel;
2508 
2509 		if (atomic_read(&t->tunnel_users) == 2)
2510 			xfrm_state_delete(t);
2511 		atomic_dec(&t->tunnel_users);
2512 		xfrm_state_put_sync(t);
2513 		x->tunnel = NULL;
2514 	}
2515 }
2516 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2517 
2518 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2519 {
2520 	const struct xfrm_type *type = READ_ONCE(x->type);
2521 	struct crypto_aead *aead;
2522 	u32 blksize, net_adj = 0;
2523 
2524 	if (x->km.state != XFRM_STATE_VALID ||
2525 	    !type || type->proto != IPPROTO_ESP)
2526 		return mtu - x->props.header_len;
2527 
2528 	aead = x->data;
2529 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2530 
2531 	switch (x->props.mode) {
2532 	case XFRM_MODE_TRANSPORT:
2533 	case XFRM_MODE_BEET:
2534 		if (x->props.family == AF_INET)
2535 			net_adj = sizeof(struct iphdr);
2536 		else if (x->props.family == AF_INET6)
2537 			net_adj = sizeof(struct ipv6hdr);
2538 		break;
2539 	case XFRM_MODE_TUNNEL:
2540 		break;
2541 	default:
2542 		WARN_ON_ONCE(1);
2543 		break;
2544 	}
2545 
2546 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2547 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2548 }
2549 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2550 
2551 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2552 {
2553 	const struct xfrm_mode *inner_mode;
2554 	const struct xfrm_mode *outer_mode;
2555 	int family = x->props.family;
2556 	int err;
2557 
2558 	if (family == AF_INET &&
2559 	    xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2560 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2561 
2562 	err = -EPROTONOSUPPORT;
2563 
2564 	if (x->sel.family != AF_UNSPEC) {
2565 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2566 		if (inner_mode == NULL)
2567 			goto error;
2568 
2569 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2570 		    family != x->sel.family)
2571 			goto error;
2572 
2573 		x->inner_mode = *inner_mode;
2574 	} else {
2575 		const struct xfrm_mode *inner_mode_iaf;
2576 		int iafamily = AF_INET;
2577 
2578 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2579 		if (inner_mode == NULL)
2580 			goto error;
2581 
2582 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2583 			goto error;
2584 
2585 		x->inner_mode = *inner_mode;
2586 
2587 		if (x->props.family == AF_INET)
2588 			iafamily = AF_INET6;
2589 
2590 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2591 		if (inner_mode_iaf) {
2592 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2593 				x->inner_mode_iaf = *inner_mode_iaf;
2594 		}
2595 	}
2596 
2597 	x->type = xfrm_get_type(x->id.proto, family);
2598 	if (x->type == NULL)
2599 		goto error;
2600 
2601 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2602 
2603 	err = x->type->init_state(x);
2604 	if (err)
2605 		goto error;
2606 
2607 	outer_mode = xfrm_get_mode(x->props.mode, family);
2608 	if (!outer_mode) {
2609 		err = -EPROTONOSUPPORT;
2610 		goto error;
2611 	}
2612 
2613 	x->outer_mode = *outer_mode;
2614 	if (init_replay) {
2615 		err = xfrm_init_replay(x);
2616 		if (err)
2617 			goto error;
2618 	}
2619 
2620 error:
2621 	return err;
2622 }
2623 
2624 EXPORT_SYMBOL(__xfrm_init_state);
2625 
2626 int xfrm_init_state(struct xfrm_state *x)
2627 {
2628 	int err;
2629 
2630 	err = __xfrm_init_state(x, true, false);
2631 	if (!err)
2632 		x->km.state = XFRM_STATE_VALID;
2633 
2634 	return err;
2635 }
2636 
2637 EXPORT_SYMBOL(xfrm_init_state);
2638 
2639 int __net_init xfrm_state_init(struct net *net)
2640 {
2641 	unsigned int sz;
2642 
2643 	if (net_eq(net, &init_net))
2644 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2645 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2646 
2647 	INIT_LIST_HEAD(&net->xfrm.state_all);
2648 
2649 	sz = sizeof(struct hlist_head) * 8;
2650 
2651 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2652 	if (!net->xfrm.state_bydst)
2653 		goto out_bydst;
2654 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2655 	if (!net->xfrm.state_bysrc)
2656 		goto out_bysrc;
2657 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2658 	if (!net->xfrm.state_byspi)
2659 		goto out_byspi;
2660 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2661 
2662 	net->xfrm.state_num = 0;
2663 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2664 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2665 	return 0;
2666 
2667 out_byspi:
2668 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2669 out_bysrc:
2670 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2671 out_bydst:
2672 	return -ENOMEM;
2673 }
2674 
2675 void xfrm_state_fini(struct net *net)
2676 {
2677 	unsigned int sz;
2678 
2679 	flush_work(&net->xfrm.state_hash_work);
2680 	flush_work(&xfrm_state_gc_work);
2681 	xfrm_state_flush(net, 0, false, true);
2682 
2683 	WARN_ON(!list_empty(&net->xfrm.state_all));
2684 
2685 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2686 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2687 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2688 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2689 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2690 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2691 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2692 }
2693 
2694 #ifdef CONFIG_AUDITSYSCALL
2695 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2696 				     struct audit_buffer *audit_buf)
2697 {
2698 	struct xfrm_sec_ctx *ctx = x->security;
2699 	u32 spi = ntohl(x->id.spi);
2700 
2701 	if (ctx)
2702 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2703 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2704 
2705 	switch (x->props.family) {
2706 	case AF_INET:
2707 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2708 				 &x->props.saddr.a4, &x->id.daddr.a4);
2709 		break;
2710 	case AF_INET6:
2711 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2712 				 x->props.saddr.a6, x->id.daddr.a6);
2713 		break;
2714 	}
2715 
2716 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2717 }
2718 
2719 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2720 				      struct audit_buffer *audit_buf)
2721 {
2722 	const struct iphdr *iph4;
2723 	const struct ipv6hdr *iph6;
2724 
2725 	switch (family) {
2726 	case AF_INET:
2727 		iph4 = ip_hdr(skb);
2728 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2729 				 &iph4->saddr, &iph4->daddr);
2730 		break;
2731 	case AF_INET6:
2732 		iph6 = ipv6_hdr(skb);
2733 		audit_log_format(audit_buf,
2734 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2735 				 &iph6->saddr, &iph6->daddr,
2736 				 iph6->flow_lbl[0] & 0x0f,
2737 				 iph6->flow_lbl[1],
2738 				 iph6->flow_lbl[2]);
2739 		break;
2740 	}
2741 }
2742 
2743 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2744 {
2745 	struct audit_buffer *audit_buf;
2746 
2747 	audit_buf = xfrm_audit_start("SAD-add");
2748 	if (audit_buf == NULL)
2749 		return;
2750 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2751 	xfrm_audit_helper_sainfo(x, audit_buf);
2752 	audit_log_format(audit_buf, " res=%u", result);
2753 	audit_log_end(audit_buf);
2754 }
2755 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2756 
2757 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2758 {
2759 	struct audit_buffer *audit_buf;
2760 
2761 	audit_buf = xfrm_audit_start("SAD-delete");
2762 	if (audit_buf == NULL)
2763 		return;
2764 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2765 	xfrm_audit_helper_sainfo(x, audit_buf);
2766 	audit_log_format(audit_buf, " res=%u", result);
2767 	audit_log_end(audit_buf);
2768 }
2769 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2770 
2771 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2772 				      struct sk_buff *skb)
2773 {
2774 	struct audit_buffer *audit_buf;
2775 	u32 spi;
2776 
2777 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2778 	if (audit_buf == NULL)
2779 		return;
2780 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2781 	/* don't record the sequence number because it's inherent in this kind
2782 	 * of audit message */
2783 	spi = ntohl(x->id.spi);
2784 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2785 	audit_log_end(audit_buf);
2786 }
2787 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2788 
2789 void xfrm_audit_state_replay(struct xfrm_state *x,
2790 			     struct sk_buff *skb, __be32 net_seq)
2791 {
2792 	struct audit_buffer *audit_buf;
2793 	u32 spi;
2794 
2795 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2796 	if (audit_buf == NULL)
2797 		return;
2798 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2799 	spi = ntohl(x->id.spi);
2800 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2801 			 spi, spi, ntohl(net_seq));
2802 	audit_log_end(audit_buf);
2803 }
2804 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2805 
2806 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2807 {
2808 	struct audit_buffer *audit_buf;
2809 
2810 	audit_buf = xfrm_audit_start("SA-notfound");
2811 	if (audit_buf == NULL)
2812 		return;
2813 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2814 	audit_log_end(audit_buf);
2815 }
2816 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2817 
2818 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2819 			       __be32 net_spi, __be32 net_seq)
2820 {
2821 	struct audit_buffer *audit_buf;
2822 	u32 spi;
2823 
2824 	audit_buf = xfrm_audit_start("SA-notfound");
2825 	if (audit_buf == NULL)
2826 		return;
2827 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2828 	spi = ntohl(net_spi);
2829 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2830 			 spi, spi, ntohl(net_seq));
2831 	audit_log_end(audit_buf);
2832 }
2833 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2834 
2835 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2836 			      struct sk_buff *skb, u8 proto)
2837 {
2838 	struct audit_buffer *audit_buf;
2839 	__be32 net_spi;
2840 	__be32 net_seq;
2841 
2842 	audit_buf = xfrm_audit_start("SA-icv-failure");
2843 	if (audit_buf == NULL)
2844 		return;
2845 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2846 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2847 		u32 spi = ntohl(net_spi);
2848 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2849 				 spi, spi, ntohl(net_seq));
2850 	}
2851 	audit_log_end(audit_buf);
2852 }
2853 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2854 #endif /* CONFIG_AUDITSYSCALL */
2855