xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision aa0dc6a7)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * xfrm_state.c
4  *
5  * Changes:
6  *	Mitsuru KANDA @USAGI
7  * 	Kazunori MIYAZAWA @USAGI
8  * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9  * 		IPv6 support
10  * 	YOSHIFUJI Hideaki @USAGI
11  * 		Split up af-specific functions
12  *	Derek Atkins <derek@ihtfp.com>
13  *		Add UDP Encapsulation
14  *
15  */
16 
17 #include <linux/workqueue.h>
18 #include <net/xfrm.h>
19 #include <linux/pfkeyv2.h>
20 #include <linux/ipsec.h>
21 #include <linux/module.h>
22 #include <linux/cache.h>
23 #include <linux/audit.h>
24 #include <linux/uaccess.h>
25 #include <linux/ktime.h>
26 #include <linux/slab.h>
27 #include <linux/interrupt.h>
28 #include <linux/kernel.h>
29 
30 #include <crypto/aead.h>
31 
32 #include "xfrm_hash.h"
33 
34 #define xfrm_state_deref_prot(table, net) \
35 	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
36 
37 static void xfrm_state_gc_task(struct work_struct *work);
38 
39 /* Each xfrm_state may be linked to two tables:
40 
41    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
42    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
43       destination/tunnel endpoint. (output)
44  */
45 
46 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
47 static struct kmem_cache *xfrm_state_cache __ro_after_init;
48 
49 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
50 static HLIST_HEAD(xfrm_state_gc_list);
51 
52 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
53 {
54 	return refcount_inc_not_zero(&x->refcnt);
55 }
56 
57 static inline unsigned int xfrm_dst_hash(struct net *net,
58 					 const xfrm_address_t *daddr,
59 					 const xfrm_address_t *saddr,
60 					 u32 reqid,
61 					 unsigned short family)
62 {
63 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
64 }
65 
66 static inline unsigned int xfrm_src_hash(struct net *net,
67 					 const xfrm_address_t *daddr,
68 					 const xfrm_address_t *saddr,
69 					 unsigned short family)
70 {
71 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
72 }
73 
74 static inline unsigned int
75 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
76 	      __be32 spi, u8 proto, unsigned short family)
77 {
78 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
79 }
80 
81 static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
82 {
83 	return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
84 }
85 
86 static void xfrm_hash_transfer(struct hlist_head *list,
87 			       struct hlist_head *ndsttable,
88 			       struct hlist_head *nsrctable,
89 			       struct hlist_head *nspitable,
90 			       struct hlist_head *nseqtable,
91 			       unsigned int nhashmask)
92 {
93 	struct hlist_node *tmp;
94 	struct xfrm_state *x;
95 
96 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
97 		unsigned int h;
98 
99 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
100 				    x->props.reqid, x->props.family,
101 				    nhashmask);
102 		hlist_add_head_rcu(&x->bydst, ndsttable + h);
103 
104 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
105 				    x->props.family,
106 				    nhashmask);
107 		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
108 
109 		if (x->id.spi) {
110 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
111 					    x->id.proto, x->props.family,
112 					    nhashmask);
113 			hlist_add_head_rcu(&x->byspi, nspitable + h);
114 		}
115 
116 		if (x->km.seq) {
117 			h = __xfrm_seq_hash(x->km.seq, nhashmask);
118 			hlist_add_head_rcu(&x->byseq, nseqtable + h);
119 		}
120 	}
121 }
122 
123 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
124 {
125 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
126 }
127 
128 static void xfrm_hash_resize(struct work_struct *work)
129 {
130 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
131 	struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
132 	unsigned long nsize, osize;
133 	unsigned int nhashmask, ohashmask;
134 	int i;
135 
136 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
137 	ndst = xfrm_hash_alloc(nsize);
138 	if (!ndst)
139 		return;
140 	nsrc = xfrm_hash_alloc(nsize);
141 	if (!nsrc) {
142 		xfrm_hash_free(ndst, nsize);
143 		return;
144 	}
145 	nspi = xfrm_hash_alloc(nsize);
146 	if (!nspi) {
147 		xfrm_hash_free(ndst, nsize);
148 		xfrm_hash_free(nsrc, nsize);
149 		return;
150 	}
151 	nseq = xfrm_hash_alloc(nsize);
152 	if (!nseq) {
153 		xfrm_hash_free(ndst, nsize);
154 		xfrm_hash_free(nsrc, nsize);
155 		xfrm_hash_free(nspi, nsize);
156 		return;
157 	}
158 
159 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
160 	write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
161 
162 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
163 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
164 	for (i = net->xfrm.state_hmask; i >= 0; i--)
165 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
166 
167 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
168 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
169 	oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
170 	ohashmask = net->xfrm.state_hmask;
171 
172 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
173 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
174 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
175 	rcu_assign_pointer(net->xfrm.state_byseq, nseq);
176 	net->xfrm.state_hmask = nhashmask;
177 
178 	write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
179 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
180 
181 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
182 
183 	synchronize_rcu();
184 
185 	xfrm_hash_free(odst, osize);
186 	xfrm_hash_free(osrc, osize);
187 	xfrm_hash_free(ospi, osize);
188 	xfrm_hash_free(oseq, osize);
189 }
190 
191 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
192 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
193 
194 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
195 
196 int __xfrm_state_delete(struct xfrm_state *x);
197 
198 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
199 static bool km_is_alive(const struct km_event *c);
200 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
201 
202 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
203 {
204 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
205 	int err = 0;
206 
207 	if (!afinfo)
208 		return -EAFNOSUPPORT;
209 
210 #define X(afi, T, name) do {			\
211 		WARN_ON((afi)->type_ ## name);	\
212 		(afi)->type_ ## name = (T);	\
213 	} while (0)
214 
215 	switch (type->proto) {
216 	case IPPROTO_COMP:
217 		X(afinfo, type, comp);
218 		break;
219 	case IPPROTO_AH:
220 		X(afinfo, type, ah);
221 		break;
222 	case IPPROTO_ESP:
223 		X(afinfo, type, esp);
224 		break;
225 	case IPPROTO_IPIP:
226 		X(afinfo, type, ipip);
227 		break;
228 	case IPPROTO_DSTOPTS:
229 		X(afinfo, type, dstopts);
230 		break;
231 	case IPPROTO_ROUTING:
232 		X(afinfo, type, routing);
233 		break;
234 	case IPPROTO_IPV6:
235 		X(afinfo, type, ipip6);
236 		break;
237 	default:
238 		WARN_ON(1);
239 		err = -EPROTONOSUPPORT;
240 		break;
241 	}
242 #undef X
243 	rcu_read_unlock();
244 	return err;
245 }
246 EXPORT_SYMBOL(xfrm_register_type);
247 
248 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
249 {
250 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
251 
252 	if (unlikely(afinfo == NULL))
253 		return;
254 
255 #define X(afi, T, name) do {				\
256 		WARN_ON((afi)->type_ ## name != (T));	\
257 		(afi)->type_ ## name = NULL;		\
258 	} while (0)
259 
260 	switch (type->proto) {
261 	case IPPROTO_COMP:
262 		X(afinfo, type, comp);
263 		break;
264 	case IPPROTO_AH:
265 		X(afinfo, type, ah);
266 		break;
267 	case IPPROTO_ESP:
268 		X(afinfo, type, esp);
269 		break;
270 	case IPPROTO_IPIP:
271 		X(afinfo, type, ipip);
272 		break;
273 	case IPPROTO_DSTOPTS:
274 		X(afinfo, type, dstopts);
275 		break;
276 	case IPPROTO_ROUTING:
277 		X(afinfo, type, routing);
278 		break;
279 	case IPPROTO_IPV6:
280 		X(afinfo, type, ipip6);
281 		break;
282 	default:
283 		WARN_ON(1);
284 		break;
285 	}
286 #undef X
287 	rcu_read_unlock();
288 }
289 EXPORT_SYMBOL(xfrm_unregister_type);
290 
291 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
292 {
293 	const struct xfrm_type *type = NULL;
294 	struct xfrm_state_afinfo *afinfo;
295 	int modload_attempted = 0;
296 
297 retry:
298 	afinfo = xfrm_state_get_afinfo(family);
299 	if (unlikely(afinfo == NULL))
300 		return NULL;
301 
302 	switch (proto) {
303 	case IPPROTO_COMP:
304 		type = afinfo->type_comp;
305 		break;
306 	case IPPROTO_AH:
307 		type = afinfo->type_ah;
308 		break;
309 	case IPPROTO_ESP:
310 		type = afinfo->type_esp;
311 		break;
312 	case IPPROTO_IPIP:
313 		type = afinfo->type_ipip;
314 		break;
315 	case IPPROTO_DSTOPTS:
316 		type = afinfo->type_dstopts;
317 		break;
318 	case IPPROTO_ROUTING:
319 		type = afinfo->type_routing;
320 		break;
321 	case IPPROTO_IPV6:
322 		type = afinfo->type_ipip6;
323 		break;
324 	default:
325 		break;
326 	}
327 
328 	if (unlikely(type && !try_module_get(type->owner)))
329 		type = NULL;
330 
331 	rcu_read_unlock();
332 
333 	if (!type && !modload_attempted) {
334 		request_module("xfrm-type-%d-%d", family, proto);
335 		modload_attempted = 1;
336 		goto retry;
337 	}
338 
339 	return type;
340 }
341 
342 static void xfrm_put_type(const struct xfrm_type *type)
343 {
344 	module_put(type->owner);
345 }
346 
347 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
348 			       unsigned short family)
349 {
350 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
351 	int err = 0;
352 
353 	if (unlikely(afinfo == NULL))
354 		return -EAFNOSUPPORT;
355 
356 	switch (type->proto) {
357 	case IPPROTO_ESP:
358 		WARN_ON(afinfo->type_offload_esp);
359 		afinfo->type_offload_esp = type;
360 		break;
361 	default:
362 		WARN_ON(1);
363 		err = -EPROTONOSUPPORT;
364 		break;
365 	}
366 
367 	rcu_read_unlock();
368 	return err;
369 }
370 EXPORT_SYMBOL(xfrm_register_type_offload);
371 
372 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
373 				  unsigned short family)
374 {
375 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
376 
377 	if (unlikely(afinfo == NULL))
378 		return;
379 
380 	switch (type->proto) {
381 	case IPPROTO_ESP:
382 		WARN_ON(afinfo->type_offload_esp != type);
383 		afinfo->type_offload_esp = NULL;
384 		break;
385 	default:
386 		WARN_ON(1);
387 		break;
388 	}
389 	rcu_read_unlock();
390 }
391 EXPORT_SYMBOL(xfrm_unregister_type_offload);
392 
393 static const struct xfrm_type_offload *
394 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
395 {
396 	const struct xfrm_type_offload *type = NULL;
397 	struct xfrm_state_afinfo *afinfo;
398 
399 retry:
400 	afinfo = xfrm_state_get_afinfo(family);
401 	if (unlikely(afinfo == NULL))
402 		return NULL;
403 
404 	switch (proto) {
405 	case IPPROTO_ESP:
406 		type = afinfo->type_offload_esp;
407 		break;
408 	default:
409 		break;
410 	}
411 
412 	if ((type && !try_module_get(type->owner)))
413 		type = NULL;
414 
415 	rcu_read_unlock();
416 
417 	if (!type && try_load) {
418 		request_module("xfrm-offload-%d-%d", family, proto);
419 		try_load = false;
420 		goto retry;
421 	}
422 
423 	return type;
424 }
425 
426 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
427 {
428 	module_put(type->owner);
429 }
430 
431 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
432 	[XFRM_MODE_BEET] = {
433 		.encap = XFRM_MODE_BEET,
434 		.flags = XFRM_MODE_FLAG_TUNNEL,
435 		.family = AF_INET,
436 	},
437 	[XFRM_MODE_TRANSPORT] = {
438 		.encap = XFRM_MODE_TRANSPORT,
439 		.family = AF_INET,
440 	},
441 	[XFRM_MODE_TUNNEL] = {
442 		.encap = XFRM_MODE_TUNNEL,
443 		.flags = XFRM_MODE_FLAG_TUNNEL,
444 		.family = AF_INET,
445 	},
446 };
447 
448 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
449 	[XFRM_MODE_BEET] = {
450 		.encap = XFRM_MODE_BEET,
451 		.flags = XFRM_MODE_FLAG_TUNNEL,
452 		.family = AF_INET6,
453 	},
454 	[XFRM_MODE_ROUTEOPTIMIZATION] = {
455 		.encap = XFRM_MODE_ROUTEOPTIMIZATION,
456 		.family = AF_INET6,
457 	},
458 	[XFRM_MODE_TRANSPORT] = {
459 		.encap = XFRM_MODE_TRANSPORT,
460 		.family = AF_INET6,
461 	},
462 	[XFRM_MODE_TUNNEL] = {
463 		.encap = XFRM_MODE_TUNNEL,
464 		.flags = XFRM_MODE_FLAG_TUNNEL,
465 		.family = AF_INET6,
466 	},
467 };
468 
469 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
470 {
471 	const struct xfrm_mode *mode;
472 
473 	if (unlikely(encap >= XFRM_MODE_MAX))
474 		return NULL;
475 
476 	switch (family) {
477 	case AF_INET:
478 		mode = &xfrm4_mode_map[encap];
479 		if (mode->family == family)
480 			return mode;
481 		break;
482 	case AF_INET6:
483 		mode = &xfrm6_mode_map[encap];
484 		if (mode->family == family)
485 			return mode;
486 		break;
487 	default:
488 		break;
489 	}
490 
491 	return NULL;
492 }
493 
494 void xfrm_state_free(struct xfrm_state *x)
495 {
496 	kmem_cache_free(xfrm_state_cache, x);
497 }
498 EXPORT_SYMBOL(xfrm_state_free);
499 
500 static void ___xfrm_state_destroy(struct xfrm_state *x)
501 {
502 	hrtimer_cancel(&x->mtimer);
503 	del_timer_sync(&x->rtimer);
504 	kfree(x->aead);
505 	kfree(x->aalg);
506 	kfree(x->ealg);
507 	kfree(x->calg);
508 	kfree(x->encap);
509 	kfree(x->coaddr);
510 	kfree(x->replay_esn);
511 	kfree(x->preplay_esn);
512 	if (x->type_offload)
513 		xfrm_put_type_offload(x->type_offload);
514 	if (x->type) {
515 		x->type->destructor(x);
516 		xfrm_put_type(x->type);
517 	}
518 	if (x->xfrag.page)
519 		put_page(x->xfrag.page);
520 	xfrm_dev_state_free(x);
521 	security_xfrm_state_free(x);
522 	xfrm_state_free(x);
523 }
524 
525 static void xfrm_state_gc_task(struct work_struct *work)
526 {
527 	struct xfrm_state *x;
528 	struct hlist_node *tmp;
529 	struct hlist_head gc_list;
530 
531 	spin_lock_bh(&xfrm_state_gc_lock);
532 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
533 	spin_unlock_bh(&xfrm_state_gc_lock);
534 
535 	synchronize_rcu();
536 
537 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
538 		___xfrm_state_destroy(x);
539 }
540 
541 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
542 {
543 	struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
544 	enum hrtimer_restart ret = HRTIMER_NORESTART;
545 	time64_t now = ktime_get_real_seconds();
546 	time64_t next = TIME64_MAX;
547 	int warn = 0;
548 	int err = 0;
549 
550 	spin_lock(&x->lock);
551 	if (x->km.state == XFRM_STATE_DEAD)
552 		goto out;
553 	if (x->km.state == XFRM_STATE_EXPIRED)
554 		goto expired;
555 	if (x->lft.hard_add_expires_seconds) {
556 		long tmo = x->lft.hard_add_expires_seconds +
557 			x->curlft.add_time - now;
558 		if (tmo <= 0) {
559 			if (x->xflags & XFRM_SOFT_EXPIRE) {
560 				/* enter hard expire without soft expire first?!
561 				 * setting a new date could trigger this.
562 				 * workaround: fix x->curflt.add_time by below:
563 				 */
564 				x->curlft.add_time = now - x->saved_tmo - 1;
565 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
566 			} else
567 				goto expired;
568 		}
569 		if (tmo < next)
570 			next = tmo;
571 	}
572 	if (x->lft.hard_use_expires_seconds) {
573 		long tmo = x->lft.hard_use_expires_seconds +
574 			(x->curlft.use_time ? : now) - now;
575 		if (tmo <= 0)
576 			goto expired;
577 		if (tmo < next)
578 			next = tmo;
579 	}
580 	if (x->km.dying)
581 		goto resched;
582 	if (x->lft.soft_add_expires_seconds) {
583 		long tmo = x->lft.soft_add_expires_seconds +
584 			x->curlft.add_time - now;
585 		if (tmo <= 0) {
586 			warn = 1;
587 			x->xflags &= ~XFRM_SOFT_EXPIRE;
588 		} else if (tmo < next) {
589 			next = tmo;
590 			x->xflags |= XFRM_SOFT_EXPIRE;
591 			x->saved_tmo = tmo;
592 		}
593 	}
594 	if (x->lft.soft_use_expires_seconds) {
595 		long tmo = x->lft.soft_use_expires_seconds +
596 			(x->curlft.use_time ? : now) - now;
597 		if (tmo <= 0)
598 			warn = 1;
599 		else if (tmo < next)
600 			next = tmo;
601 	}
602 
603 	x->km.dying = warn;
604 	if (warn)
605 		km_state_expired(x, 0, 0);
606 resched:
607 	if (next != TIME64_MAX) {
608 		hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
609 		ret = HRTIMER_RESTART;
610 	}
611 
612 	goto out;
613 
614 expired:
615 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
616 		x->km.state = XFRM_STATE_EXPIRED;
617 
618 	err = __xfrm_state_delete(x);
619 	if (!err)
620 		km_state_expired(x, 1, 0);
621 
622 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
623 
624 out:
625 	spin_unlock(&x->lock);
626 	return ret;
627 }
628 
629 static void xfrm_replay_timer_handler(struct timer_list *t);
630 
631 struct xfrm_state *xfrm_state_alloc(struct net *net)
632 {
633 	struct xfrm_state *x;
634 
635 	x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
636 
637 	if (x) {
638 		write_pnet(&x->xs_net, net);
639 		refcount_set(&x->refcnt, 1);
640 		atomic_set(&x->tunnel_users, 0);
641 		INIT_LIST_HEAD(&x->km.all);
642 		INIT_HLIST_NODE(&x->bydst);
643 		INIT_HLIST_NODE(&x->bysrc);
644 		INIT_HLIST_NODE(&x->byspi);
645 		INIT_HLIST_NODE(&x->byseq);
646 		hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
647 		x->mtimer.function = xfrm_timer_handler;
648 		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
649 		x->curlft.add_time = ktime_get_real_seconds();
650 		x->lft.soft_byte_limit = XFRM_INF;
651 		x->lft.soft_packet_limit = XFRM_INF;
652 		x->lft.hard_byte_limit = XFRM_INF;
653 		x->lft.hard_packet_limit = XFRM_INF;
654 		x->replay_maxage = 0;
655 		x->replay_maxdiff = 0;
656 		spin_lock_init(&x->lock);
657 	}
658 	return x;
659 }
660 EXPORT_SYMBOL(xfrm_state_alloc);
661 
662 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
663 {
664 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
665 
666 	if (sync) {
667 		synchronize_rcu();
668 		___xfrm_state_destroy(x);
669 	} else {
670 		spin_lock_bh(&xfrm_state_gc_lock);
671 		hlist_add_head(&x->gclist, &xfrm_state_gc_list);
672 		spin_unlock_bh(&xfrm_state_gc_lock);
673 		schedule_work(&xfrm_state_gc_work);
674 	}
675 }
676 EXPORT_SYMBOL(__xfrm_state_destroy);
677 
678 int __xfrm_state_delete(struct xfrm_state *x)
679 {
680 	struct net *net = xs_net(x);
681 	int err = -ESRCH;
682 
683 	if (x->km.state != XFRM_STATE_DEAD) {
684 		x->km.state = XFRM_STATE_DEAD;
685 		spin_lock(&net->xfrm.xfrm_state_lock);
686 		list_del(&x->km.all);
687 		hlist_del_rcu(&x->bydst);
688 		hlist_del_rcu(&x->bysrc);
689 		if (x->km.seq)
690 			hlist_del_rcu(&x->byseq);
691 		if (x->id.spi)
692 			hlist_del_rcu(&x->byspi);
693 		net->xfrm.state_num--;
694 		spin_unlock(&net->xfrm.xfrm_state_lock);
695 
696 		if (x->encap_sk)
697 			sock_put(rcu_dereference_raw(x->encap_sk));
698 
699 		xfrm_dev_state_delete(x);
700 
701 		/* All xfrm_state objects are created by xfrm_state_alloc.
702 		 * The xfrm_state_alloc call gives a reference, and that
703 		 * is what we are dropping here.
704 		 */
705 		xfrm_state_put(x);
706 		err = 0;
707 	}
708 
709 	return err;
710 }
711 EXPORT_SYMBOL(__xfrm_state_delete);
712 
713 int xfrm_state_delete(struct xfrm_state *x)
714 {
715 	int err;
716 
717 	spin_lock_bh(&x->lock);
718 	err = __xfrm_state_delete(x);
719 	spin_unlock_bh(&x->lock);
720 
721 	return err;
722 }
723 EXPORT_SYMBOL(xfrm_state_delete);
724 
725 #ifdef CONFIG_SECURITY_NETWORK_XFRM
726 static inline int
727 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
728 {
729 	int i, err = 0;
730 
731 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
732 		struct xfrm_state *x;
733 
734 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
735 			if (xfrm_id_proto_match(x->id.proto, proto) &&
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 
746 static inline int
747 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
748 {
749 	int i, err = 0;
750 
751 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
752 		struct xfrm_state *x;
753 		struct xfrm_state_offload *xso;
754 
755 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
756 			xso = &x->xso;
757 
758 			if (xso->dev == dev &&
759 			   (err = security_xfrm_state_delete(x)) != 0) {
760 				xfrm_audit_state_delete(x, 0, task_valid);
761 				return err;
762 			}
763 		}
764 	}
765 
766 	return err;
767 }
768 #else
769 static inline int
770 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
771 {
772 	return 0;
773 }
774 
775 static inline int
776 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
777 {
778 	return 0;
779 }
780 #endif
781 
782 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
783 {
784 	int i, err = 0, cnt = 0;
785 
786 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
787 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
788 	if (err)
789 		goto out;
790 
791 	err = -ESRCH;
792 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
793 		struct xfrm_state *x;
794 restart:
795 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
796 			if (!xfrm_state_kern(x) &&
797 			    xfrm_id_proto_match(x->id.proto, proto)) {
798 				xfrm_state_hold(x);
799 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
800 
801 				err = xfrm_state_delete(x);
802 				xfrm_audit_state_delete(x, err ? 0 : 1,
803 							task_valid);
804 				if (sync)
805 					xfrm_state_put_sync(x);
806 				else
807 					xfrm_state_put(x);
808 				if (!err)
809 					cnt++;
810 
811 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
812 				goto restart;
813 			}
814 		}
815 	}
816 out:
817 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
818 	if (cnt)
819 		err = 0;
820 
821 	return err;
822 }
823 EXPORT_SYMBOL(xfrm_state_flush);
824 
825 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
826 {
827 	int i, err = 0, cnt = 0;
828 
829 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
830 	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
831 	if (err)
832 		goto out;
833 
834 	err = -ESRCH;
835 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
836 		struct xfrm_state *x;
837 		struct xfrm_state_offload *xso;
838 restart:
839 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
840 			xso = &x->xso;
841 
842 			if (!xfrm_state_kern(x) && xso->dev == dev) {
843 				xfrm_state_hold(x);
844 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
845 
846 				err = xfrm_state_delete(x);
847 				xfrm_audit_state_delete(x, err ? 0 : 1,
848 							task_valid);
849 				xfrm_state_put(x);
850 				if (!err)
851 					cnt++;
852 
853 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
854 				goto restart;
855 			}
856 		}
857 	}
858 	if (cnt)
859 		err = 0;
860 
861 out:
862 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
863 	return err;
864 }
865 EXPORT_SYMBOL(xfrm_dev_state_flush);
866 
867 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
868 {
869 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
870 	si->sadcnt = net->xfrm.state_num;
871 	si->sadhcnt = net->xfrm.state_hmask + 1;
872 	si->sadhmcnt = xfrm_state_hashmax;
873 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
874 }
875 EXPORT_SYMBOL(xfrm_sad_getinfo);
876 
877 static void
878 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
879 {
880 	const struct flowi4 *fl4 = &fl->u.ip4;
881 
882 	sel->daddr.a4 = fl4->daddr;
883 	sel->saddr.a4 = fl4->saddr;
884 	sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
885 	sel->dport_mask = htons(0xffff);
886 	sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
887 	sel->sport_mask = htons(0xffff);
888 	sel->family = AF_INET;
889 	sel->prefixlen_d = 32;
890 	sel->prefixlen_s = 32;
891 	sel->proto = fl4->flowi4_proto;
892 	sel->ifindex = fl4->flowi4_oif;
893 }
894 
895 static void
896 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
897 {
898 	const struct flowi6 *fl6 = &fl->u.ip6;
899 
900 	/* Initialize temporary selector matching only to current session. */
901 	*(struct in6_addr *)&sel->daddr = fl6->daddr;
902 	*(struct in6_addr *)&sel->saddr = fl6->saddr;
903 	sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
904 	sel->dport_mask = htons(0xffff);
905 	sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
906 	sel->sport_mask = htons(0xffff);
907 	sel->family = AF_INET6;
908 	sel->prefixlen_d = 128;
909 	sel->prefixlen_s = 128;
910 	sel->proto = fl6->flowi6_proto;
911 	sel->ifindex = fl6->flowi6_oif;
912 }
913 
914 static void
915 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
916 		    const struct xfrm_tmpl *tmpl,
917 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
918 		    unsigned short family)
919 {
920 	switch (family) {
921 	case AF_INET:
922 		__xfrm4_init_tempsel(&x->sel, fl);
923 		break;
924 	case AF_INET6:
925 		__xfrm6_init_tempsel(&x->sel, fl);
926 		break;
927 	}
928 
929 	x->id = tmpl->id;
930 
931 	switch (tmpl->encap_family) {
932 	case AF_INET:
933 		if (x->id.daddr.a4 == 0)
934 			x->id.daddr.a4 = daddr->a4;
935 		x->props.saddr = tmpl->saddr;
936 		if (x->props.saddr.a4 == 0)
937 			x->props.saddr.a4 = saddr->a4;
938 		break;
939 	case AF_INET6:
940 		if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
941 			memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
942 		memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
943 		if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
944 			memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
945 		break;
946 	}
947 
948 	x->props.mode = tmpl->mode;
949 	x->props.reqid = tmpl->reqid;
950 	x->props.family = tmpl->encap_family;
951 }
952 
953 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
954 					      const xfrm_address_t *daddr,
955 					      __be32 spi, u8 proto,
956 					      unsigned short family)
957 {
958 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
959 	struct xfrm_state *x;
960 
961 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
962 		if (x->props.family != family ||
963 		    x->id.spi       != spi ||
964 		    x->id.proto     != proto ||
965 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
966 			continue;
967 
968 		if ((mark & x->mark.m) != x->mark.v)
969 			continue;
970 		if (!xfrm_state_hold_rcu(x))
971 			continue;
972 		return x;
973 	}
974 
975 	return NULL;
976 }
977 
978 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
979 						     const xfrm_address_t *daddr,
980 						     const xfrm_address_t *saddr,
981 						     u8 proto, unsigned short family)
982 {
983 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
984 	struct xfrm_state *x;
985 
986 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
987 		if (x->props.family != family ||
988 		    x->id.proto     != proto ||
989 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
990 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
991 			continue;
992 
993 		if ((mark & x->mark.m) != x->mark.v)
994 			continue;
995 		if (!xfrm_state_hold_rcu(x))
996 			continue;
997 		return x;
998 	}
999 
1000 	return NULL;
1001 }
1002 
1003 static inline struct xfrm_state *
1004 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1005 {
1006 	struct net *net = xs_net(x);
1007 	u32 mark = x->mark.v & x->mark.m;
1008 
1009 	if (use_spi)
1010 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
1011 					   x->id.spi, x->id.proto, family);
1012 	else
1013 		return __xfrm_state_lookup_byaddr(net, mark,
1014 						  &x->id.daddr,
1015 						  &x->props.saddr,
1016 						  x->id.proto, family);
1017 }
1018 
1019 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1020 {
1021 	if (have_hash_collision &&
1022 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1023 	    net->xfrm.state_num > net->xfrm.state_hmask)
1024 		schedule_work(&net->xfrm.state_hash_work);
1025 }
1026 
1027 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1028 			       const struct flowi *fl, unsigned short family,
1029 			       struct xfrm_state **best, int *acq_in_progress,
1030 			       int *error)
1031 {
1032 	/* Resolution logic:
1033 	 * 1. There is a valid state with matching selector. Done.
1034 	 * 2. Valid state with inappropriate selector. Skip.
1035 	 *
1036 	 * Entering area of "sysdeps".
1037 	 *
1038 	 * 3. If state is not valid, selector is temporary, it selects
1039 	 *    only session which triggered previous resolution. Key
1040 	 *    manager will do something to install a state with proper
1041 	 *    selector.
1042 	 */
1043 	if (x->km.state == XFRM_STATE_VALID) {
1044 		if ((x->sel.family &&
1045 		     (x->sel.family != family ||
1046 		      !xfrm_selector_match(&x->sel, fl, family))) ||
1047 		    !security_xfrm_state_pol_flow_match(x, pol,
1048 							&fl->u.__fl_common))
1049 			return;
1050 
1051 		if (!*best ||
1052 		    (*best)->km.dying > x->km.dying ||
1053 		    ((*best)->km.dying == x->km.dying &&
1054 		     (*best)->curlft.add_time < x->curlft.add_time))
1055 			*best = x;
1056 	} else if (x->km.state == XFRM_STATE_ACQ) {
1057 		*acq_in_progress = 1;
1058 	} else if (x->km.state == XFRM_STATE_ERROR ||
1059 		   x->km.state == XFRM_STATE_EXPIRED) {
1060 		if ((!x->sel.family ||
1061 		     (x->sel.family == family &&
1062 		      xfrm_selector_match(&x->sel, fl, family))) &&
1063 		    security_xfrm_state_pol_flow_match(x, pol,
1064 						       &fl->u.__fl_common))
1065 			*error = -ESRCH;
1066 	}
1067 }
1068 
1069 struct xfrm_state *
1070 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1071 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1072 		struct xfrm_policy *pol, int *err,
1073 		unsigned short family, u32 if_id)
1074 {
1075 	static xfrm_address_t saddr_wildcard = { };
1076 	struct net *net = xp_net(pol);
1077 	unsigned int h, h_wildcard;
1078 	struct xfrm_state *x, *x0, *to_put;
1079 	int acquire_in_progress = 0;
1080 	int error = 0;
1081 	struct xfrm_state *best = NULL;
1082 	u32 mark = pol->mark.v & pol->mark.m;
1083 	unsigned short encap_family = tmpl->encap_family;
1084 	unsigned int sequence;
1085 	struct km_event c;
1086 
1087 	to_put = NULL;
1088 
1089 	sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1090 
1091 	rcu_read_lock();
1092 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1093 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1094 		if (x->props.family == encap_family &&
1095 		    x->props.reqid == tmpl->reqid &&
1096 		    (mark & x->mark.m) == x->mark.v &&
1097 		    x->if_id == if_id &&
1098 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1099 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1100 		    tmpl->mode == x->props.mode &&
1101 		    tmpl->id.proto == x->id.proto &&
1102 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1103 			xfrm_state_look_at(pol, x, fl, family,
1104 					   &best, &acquire_in_progress, &error);
1105 	}
1106 	if (best || acquire_in_progress)
1107 		goto found;
1108 
1109 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1110 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1111 		if (x->props.family == encap_family &&
1112 		    x->props.reqid == tmpl->reqid &&
1113 		    (mark & x->mark.m) == x->mark.v &&
1114 		    x->if_id == if_id &&
1115 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1116 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1117 		    tmpl->mode == x->props.mode &&
1118 		    tmpl->id.proto == x->id.proto &&
1119 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1120 			xfrm_state_look_at(pol, x, fl, family,
1121 					   &best, &acquire_in_progress, &error);
1122 	}
1123 
1124 found:
1125 	x = best;
1126 	if (!x && !error && !acquire_in_progress) {
1127 		if (tmpl->id.spi &&
1128 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1129 					      tmpl->id.proto, encap_family)) != NULL) {
1130 			to_put = x0;
1131 			error = -EEXIST;
1132 			goto out;
1133 		}
1134 
1135 		c.net = net;
1136 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1137 		 * for each and every packet - garbage collection might not
1138 		 * handle the flood.
1139 		 */
1140 		if (!km_is_alive(&c)) {
1141 			error = -ESRCH;
1142 			goto out;
1143 		}
1144 
1145 		x = xfrm_state_alloc(net);
1146 		if (x == NULL) {
1147 			error = -ENOMEM;
1148 			goto out;
1149 		}
1150 		/* Initialize temporary state matching only
1151 		 * to current session. */
1152 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1153 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1154 		x->if_id = if_id;
1155 
1156 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1157 		if (error) {
1158 			x->km.state = XFRM_STATE_DEAD;
1159 			to_put = x;
1160 			x = NULL;
1161 			goto out;
1162 		}
1163 
1164 		if (km_query(x, tmpl, pol) == 0) {
1165 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1166 			x->km.state = XFRM_STATE_ACQ;
1167 			list_add(&x->km.all, &net->xfrm.state_all);
1168 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1169 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1170 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1171 			if (x->id.spi) {
1172 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1173 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1174 			}
1175 			if (x->km.seq) {
1176 				h = xfrm_seq_hash(net, x->km.seq);
1177 				hlist_add_head_rcu(&x->byseq, net->xfrm.state_byseq + h);
1178 			}
1179 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1180 			hrtimer_start(&x->mtimer,
1181 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1182 				      HRTIMER_MODE_REL_SOFT);
1183 			net->xfrm.state_num++;
1184 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1185 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1186 		} else {
1187 			x->km.state = XFRM_STATE_DEAD;
1188 			to_put = x;
1189 			x = NULL;
1190 			error = -ESRCH;
1191 		}
1192 	}
1193 out:
1194 	if (x) {
1195 		if (!xfrm_state_hold_rcu(x)) {
1196 			*err = -EAGAIN;
1197 			x = NULL;
1198 		}
1199 	} else {
1200 		*err = acquire_in_progress ? -EAGAIN : error;
1201 	}
1202 	rcu_read_unlock();
1203 	if (to_put)
1204 		xfrm_state_put(to_put);
1205 
1206 	if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1207 		*err = -EAGAIN;
1208 		if (x) {
1209 			xfrm_state_put(x);
1210 			x = NULL;
1211 		}
1212 	}
1213 
1214 	return x;
1215 }
1216 
1217 struct xfrm_state *
1218 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1219 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1220 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1221 {
1222 	unsigned int h;
1223 	struct xfrm_state *rx = NULL, *x = NULL;
1224 
1225 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1226 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1227 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1228 		if (x->props.family == family &&
1229 		    x->props.reqid == reqid &&
1230 		    (mark & x->mark.m) == x->mark.v &&
1231 		    x->if_id == if_id &&
1232 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1233 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1234 		    mode == x->props.mode &&
1235 		    proto == x->id.proto &&
1236 		    x->km.state == XFRM_STATE_VALID) {
1237 			rx = x;
1238 			break;
1239 		}
1240 	}
1241 
1242 	if (rx)
1243 		xfrm_state_hold(rx);
1244 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1245 
1246 
1247 	return rx;
1248 }
1249 EXPORT_SYMBOL(xfrm_stateonly_find);
1250 
1251 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1252 					      unsigned short family)
1253 {
1254 	struct xfrm_state *x;
1255 	struct xfrm_state_walk *w;
1256 
1257 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1258 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1259 		x = container_of(w, struct xfrm_state, km);
1260 		if (x->props.family != family ||
1261 			x->id.spi != spi)
1262 			continue;
1263 
1264 		xfrm_state_hold(x);
1265 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1266 		return x;
1267 	}
1268 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1269 	return NULL;
1270 }
1271 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1272 
1273 static void __xfrm_state_insert(struct xfrm_state *x)
1274 {
1275 	struct net *net = xs_net(x);
1276 	unsigned int h;
1277 
1278 	list_add(&x->km.all, &net->xfrm.state_all);
1279 
1280 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1281 			  x->props.reqid, x->props.family);
1282 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1283 
1284 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1285 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1286 
1287 	if (x->id.spi) {
1288 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1289 				  x->props.family);
1290 
1291 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1292 	}
1293 
1294 	if (x->km.seq) {
1295 		h = xfrm_seq_hash(net, x->km.seq);
1296 
1297 		hlist_add_head_rcu(&x->byseq, net->xfrm.state_byseq + h);
1298 	}
1299 
1300 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1301 	if (x->replay_maxage)
1302 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1303 
1304 	net->xfrm.state_num++;
1305 
1306 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1307 }
1308 
1309 /* net->xfrm.xfrm_state_lock is held */
1310 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1311 {
1312 	struct net *net = xs_net(xnew);
1313 	unsigned short family = xnew->props.family;
1314 	u32 reqid = xnew->props.reqid;
1315 	struct xfrm_state *x;
1316 	unsigned int h;
1317 	u32 mark = xnew->mark.v & xnew->mark.m;
1318 	u32 if_id = xnew->if_id;
1319 
1320 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1321 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1322 		if (x->props.family	== family &&
1323 		    x->props.reqid	== reqid &&
1324 		    x->if_id		== if_id &&
1325 		    (mark & x->mark.m) == x->mark.v &&
1326 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1327 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1328 			x->genid++;
1329 	}
1330 }
1331 
1332 void xfrm_state_insert(struct xfrm_state *x)
1333 {
1334 	struct net *net = xs_net(x);
1335 
1336 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1337 	__xfrm_state_bump_genids(x);
1338 	__xfrm_state_insert(x);
1339 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1340 }
1341 EXPORT_SYMBOL(xfrm_state_insert);
1342 
1343 /* net->xfrm.xfrm_state_lock is held */
1344 static struct xfrm_state *__find_acq_core(struct net *net,
1345 					  const struct xfrm_mark *m,
1346 					  unsigned short family, u8 mode,
1347 					  u32 reqid, u32 if_id, u8 proto,
1348 					  const xfrm_address_t *daddr,
1349 					  const xfrm_address_t *saddr,
1350 					  int create)
1351 {
1352 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1353 	struct xfrm_state *x;
1354 	u32 mark = m->v & m->m;
1355 
1356 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1357 		if (x->props.reqid  != reqid ||
1358 		    x->props.mode   != mode ||
1359 		    x->props.family != family ||
1360 		    x->km.state     != XFRM_STATE_ACQ ||
1361 		    x->id.spi       != 0 ||
1362 		    x->id.proto	    != proto ||
1363 		    (mark & x->mark.m) != x->mark.v ||
1364 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1365 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1366 			continue;
1367 
1368 		xfrm_state_hold(x);
1369 		return x;
1370 	}
1371 
1372 	if (!create)
1373 		return NULL;
1374 
1375 	x = xfrm_state_alloc(net);
1376 	if (likely(x)) {
1377 		switch (family) {
1378 		case AF_INET:
1379 			x->sel.daddr.a4 = daddr->a4;
1380 			x->sel.saddr.a4 = saddr->a4;
1381 			x->sel.prefixlen_d = 32;
1382 			x->sel.prefixlen_s = 32;
1383 			x->props.saddr.a4 = saddr->a4;
1384 			x->id.daddr.a4 = daddr->a4;
1385 			break;
1386 
1387 		case AF_INET6:
1388 			x->sel.daddr.in6 = daddr->in6;
1389 			x->sel.saddr.in6 = saddr->in6;
1390 			x->sel.prefixlen_d = 128;
1391 			x->sel.prefixlen_s = 128;
1392 			x->props.saddr.in6 = saddr->in6;
1393 			x->id.daddr.in6 = daddr->in6;
1394 			break;
1395 		}
1396 
1397 		x->km.state = XFRM_STATE_ACQ;
1398 		x->id.proto = proto;
1399 		x->props.family = family;
1400 		x->props.mode = mode;
1401 		x->props.reqid = reqid;
1402 		x->if_id = if_id;
1403 		x->mark.v = m->v;
1404 		x->mark.m = m->m;
1405 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1406 		xfrm_state_hold(x);
1407 		hrtimer_start(&x->mtimer,
1408 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1409 			      HRTIMER_MODE_REL_SOFT);
1410 		list_add(&x->km.all, &net->xfrm.state_all);
1411 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1412 		h = xfrm_src_hash(net, daddr, saddr, family);
1413 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1414 
1415 		net->xfrm.state_num++;
1416 
1417 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1418 	}
1419 
1420 	return x;
1421 }
1422 
1423 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1424 
1425 int xfrm_state_add(struct xfrm_state *x)
1426 {
1427 	struct net *net = xs_net(x);
1428 	struct xfrm_state *x1, *to_put;
1429 	int family;
1430 	int err;
1431 	u32 mark = x->mark.v & x->mark.m;
1432 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1433 
1434 	family = x->props.family;
1435 
1436 	to_put = NULL;
1437 
1438 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1439 
1440 	x1 = __xfrm_state_locate(x, use_spi, family);
1441 	if (x1) {
1442 		to_put = x1;
1443 		x1 = NULL;
1444 		err = -EEXIST;
1445 		goto out;
1446 	}
1447 
1448 	if (use_spi && x->km.seq) {
1449 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1450 		if (x1 && ((x1->id.proto != x->id.proto) ||
1451 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1452 			to_put = x1;
1453 			x1 = NULL;
1454 		}
1455 	}
1456 
1457 	if (use_spi && !x1)
1458 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1459 				     x->props.reqid, x->if_id, x->id.proto,
1460 				     &x->id.daddr, &x->props.saddr, 0);
1461 
1462 	__xfrm_state_bump_genids(x);
1463 	__xfrm_state_insert(x);
1464 	err = 0;
1465 
1466 out:
1467 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1468 
1469 	if (x1) {
1470 		xfrm_state_delete(x1);
1471 		xfrm_state_put(x1);
1472 	}
1473 
1474 	if (to_put)
1475 		xfrm_state_put(to_put);
1476 
1477 	return err;
1478 }
1479 EXPORT_SYMBOL(xfrm_state_add);
1480 
1481 #ifdef CONFIG_XFRM_MIGRATE
1482 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1483 {
1484 	struct xfrm_user_sec_ctx *uctx;
1485 	int size = sizeof(*uctx) + security->ctx_len;
1486 	int err;
1487 
1488 	uctx = kmalloc(size, GFP_KERNEL);
1489 	if (!uctx)
1490 		return -ENOMEM;
1491 
1492 	uctx->exttype = XFRMA_SEC_CTX;
1493 	uctx->len = size;
1494 	uctx->ctx_doi = security->ctx_doi;
1495 	uctx->ctx_alg = security->ctx_alg;
1496 	uctx->ctx_len = security->ctx_len;
1497 	memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1498 	err = security_xfrm_state_alloc(x, uctx);
1499 	kfree(uctx);
1500 	if (err)
1501 		return err;
1502 
1503 	return 0;
1504 }
1505 
1506 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1507 					   struct xfrm_encap_tmpl *encap)
1508 {
1509 	struct net *net = xs_net(orig);
1510 	struct xfrm_state *x = xfrm_state_alloc(net);
1511 	if (!x)
1512 		goto out;
1513 
1514 	memcpy(&x->id, &orig->id, sizeof(x->id));
1515 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1516 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1517 	x->props.mode = orig->props.mode;
1518 	x->props.replay_window = orig->props.replay_window;
1519 	x->props.reqid = orig->props.reqid;
1520 	x->props.family = orig->props.family;
1521 	x->props.saddr = orig->props.saddr;
1522 
1523 	if (orig->aalg) {
1524 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1525 		if (!x->aalg)
1526 			goto error;
1527 	}
1528 	x->props.aalgo = orig->props.aalgo;
1529 
1530 	if (orig->aead) {
1531 		x->aead = xfrm_algo_aead_clone(orig->aead);
1532 		x->geniv = orig->geniv;
1533 		if (!x->aead)
1534 			goto error;
1535 	}
1536 	if (orig->ealg) {
1537 		x->ealg = xfrm_algo_clone(orig->ealg);
1538 		if (!x->ealg)
1539 			goto error;
1540 	}
1541 	x->props.ealgo = orig->props.ealgo;
1542 
1543 	if (orig->calg) {
1544 		x->calg = xfrm_algo_clone(orig->calg);
1545 		if (!x->calg)
1546 			goto error;
1547 	}
1548 	x->props.calgo = orig->props.calgo;
1549 
1550 	if (encap || orig->encap) {
1551 		if (encap)
1552 			x->encap = kmemdup(encap, sizeof(*x->encap),
1553 					GFP_KERNEL);
1554 		else
1555 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1556 					GFP_KERNEL);
1557 
1558 		if (!x->encap)
1559 			goto error;
1560 	}
1561 
1562 	if (orig->security)
1563 		if (clone_security(x, orig->security))
1564 			goto error;
1565 
1566 	if (orig->coaddr) {
1567 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1568 				    GFP_KERNEL);
1569 		if (!x->coaddr)
1570 			goto error;
1571 	}
1572 
1573 	if (orig->replay_esn) {
1574 		if (xfrm_replay_clone(x, orig))
1575 			goto error;
1576 	}
1577 
1578 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1579 	memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1580 
1581 	if (xfrm_init_state(x) < 0)
1582 		goto error;
1583 
1584 	x->props.flags = orig->props.flags;
1585 	x->props.extra_flags = orig->props.extra_flags;
1586 
1587 	x->if_id = orig->if_id;
1588 	x->tfcpad = orig->tfcpad;
1589 	x->replay_maxdiff = orig->replay_maxdiff;
1590 	x->replay_maxage = orig->replay_maxage;
1591 	memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1592 	x->km.state = orig->km.state;
1593 	x->km.seq = orig->km.seq;
1594 	x->replay = orig->replay;
1595 	x->preplay = orig->preplay;
1596 
1597 	return x;
1598 
1599  error:
1600 	xfrm_state_put(x);
1601 out:
1602 	return NULL;
1603 }
1604 
1605 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1606 {
1607 	unsigned int h;
1608 	struct xfrm_state *x = NULL;
1609 
1610 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1611 
1612 	if (m->reqid) {
1613 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1614 				  m->reqid, m->old_family);
1615 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1616 			if (x->props.mode != m->mode ||
1617 			    x->id.proto != m->proto)
1618 				continue;
1619 			if (m->reqid && x->props.reqid != m->reqid)
1620 				continue;
1621 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1622 					     m->old_family) ||
1623 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1624 					     m->old_family))
1625 				continue;
1626 			xfrm_state_hold(x);
1627 			break;
1628 		}
1629 	} else {
1630 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1631 				  m->old_family);
1632 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1633 			if (x->props.mode != m->mode ||
1634 			    x->id.proto != m->proto)
1635 				continue;
1636 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1637 					     m->old_family) ||
1638 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1639 					     m->old_family))
1640 				continue;
1641 			xfrm_state_hold(x);
1642 			break;
1643 		}
1644 	}
1645 
1646 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1647 
1648 	return x;
1649 }
1650 EXPORT_SYMBOL(xfrm_migrate_state_find);
1651 
1652 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1653 				      struct xfrm_migrate *m,
1654 				      struct xfrm_encap_tmpl *encap)
1655 {
1656 	struct xfrm_state *xc;
1657 
1658 	xc = xfrm_state_clone(x, encap);
1659 	if (!xc)
1660 		return NULL;
1661 
1662 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1663 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1664 
1665 	/* add state */
1666 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1667 		/* a care is needed when the destination address of the
1668 		   state is to be updated as it is a part of triplet */
1669 		xfrm_state_insert(xc);
1670 	} else {
1671 		if (xfrm_state_add(xc) < 0)
1672 			goto error;
1673 	}
1674 
1675 	return xc;
1676 error:
1677 	xfrm_state_put(xc);
1678 	return NULL;
1679 }
1680 EXPORT_SYMBOL(xfrm_state_migrate);
1681 #endif
1682 
1683 int xfrm_state_update(struct xfrm_state *x)
1684 {
1685 	struct xfrm_state *x1, *to_put;
1686 	int err;
1687 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1688 	struct net *net = xs_net(x);
1689 
1690 	to_put = NULL;
1691 
1692 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1693 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1694 
1695 	err = -ESRCH;
1696 	if (!x1)
1697 		goto out;
1698 
1699 	if (xfrm_state_kern(x1)) {
1700 		to_put = x1;
1701 		err = -EEXIST;
1702 		goto out;
1703 	}
1704 
1705 	if (x1->km.state == XFRM_STATE_ACQ) {
1706 		__xfrm_state_insert(x);
1707 		x = NULL;
1708 	}
1709 	err = 0;
1710 
1711 out:
1712 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1713 
1714 	if (to_put)
1715 		xfrm_state_put(to_put);
1716 
1717 	if (err)
1718 		return err;
1719 
1720 	if (!x) {
1721 		xfrm_state_delete(x1);
1722 		xfrm_state_put(x1);
1723 		return 0;
1724 	}
1725 
1726 	err = -EINVAL;
1727 	spin_lock_bh(&x1->lock);
1728 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1729 		if (x->encap && x1->encap &&
1730 		    x->encap->encap_type == x1->encap->encap_type)
1731 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1732 		else if (x->encap || x1->encap)
1733 			goto fail;
1734 
1735 		if (x->coaddr && x1->coaddr) {
1736 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1737 		}
1738 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1739 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1740 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1741 		x1->km.dying = 0;
1742 
1743 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1744 			      HRTIMER_MODE_REL_SOFT);
1745 		if (x1->curlft.use_time)
1746 			xfrm_state_check_expire(x1);
1747 
1748 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1749 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1750 
1751 			if (x->props.smark.m || x->props.smark.v)
1752 				x1->props.smark = x->props.smark;
1753 
1754 			if (x->if_id)
1755 				x1->if_id = x->if_id;
1756 
1757 			__xfrm_state_bump_genids(x1);
1758 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1759 		}
1760 
1761 		err = 0;
1762 		x->km.state = XFRM_STATE_DEAD;
1763 		__xfrm_state_put(x);
1764 	}
1765 
1766 fail:
1767 	spin_unlock_bh(&x1->lock);
1768 
1769 	xfrm_state_put(x1);
1770 
1771 	return err;
1772 }
1773 EXPORT_SYMBOL(xfrm_state_update);
1774 
1775 int xfrm_state_check_expire(struct xfrm_state *x)
1776 {
1777 	if (!x->curlft.use_time)
1778 		x->curlft.use_time = ktime_get_real_seconds();
1779 
1780 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1781 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1782 		x->km.state = XFRM_STATE_EXPIRED;
1783 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1784 		return -EINVAL;
1785 	}
1786 
1787 	if (!x->km.dying &&
1788 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1789 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1790 		x->km.dying = 1;
1791 		km_state_expired(x, 0, 0);
1792 	}
1793 	return 0;
1794 }
1795 EXPORT_SYMBOL(xfrm_state_check_expire);
1796 
1797 struct xfrm_state *
1798 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1799 		  u8 proto, unsigned short family)
1800 {
1801 	struct xfrm_state *x;
1802 
1803 	rcu_read_lock();
1804 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1805 	rcu_read_unlock();
1806 	return x;
1807 }
1808 EXPORT_SYMBOL(xfrm_state_lookup);
1809 
1810 struct xfrm_state *
1811 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1812 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1813 			 u8 proto, unsigned short family)
1814 {
1815 	struct xfrm_state *x;
1816 
1817 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1818 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1819 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1820 	return x;
1821 }
1822 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1823 
1824 struct xfrm_state *
1825 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1826 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1827 	      const xfrm_address_t *saddr, int create, unsigned short family)
1828 {
1829 	struct xfrm_state *x;
1830 
1831 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1832 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1833 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1834 
1835 	return x;
1836 }
1837 EXPORT_SYMBOL(xfrm_find_acq);
1838 
1839 #ifdef CONFIG_XFRM_SUB_POLICY
1840 #if IS_ENABLED(CONFIG_IPV6)
1841 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1842 static void
1843 __xfrm6_sort(void **dst, void **src, int n,
1844 	     int (*cmp)(const void *p), int maxclass)
1845 {
1846 	int count[XFRM_MAX_DEPTH] = { };
1847 	int class[XFRM_MAX_DEPTH];
1848 	int i;
1849 
1850 	for (i = 0; i < n; i++) {
1851 		int c = cmp(src[i]);
1852 
1853 		class[i] = c;
1854 		count[c]++;
1855 	}
1856 
1857 	for (i = 2; i < maxclass; i++)
1858 		count[i] += count[i - 1];
1859 
1860 	for (i = 0; i < n; i++) {
1861 		dst[count[class[i] - 1]++] = src[i];
1862 		src[i] = NULL;
1863 	}
1864 }
1865 
1866 /* Rule for xfrm_state:
1867  *
1868  * rule 1: select IPsec transport except AH
1869  * rule 2: select MIPv6 RO or inbound trigger
1870  * rule 3: select IPsec transport AH
1871  * rule 4: select IPsec tunnel
1872  * rule 5: others
1873  */
1874 static int __xfrm6_state_sort_cmp(const void *p)
1875 {
1876 	const struct xfrm_state *v = p;
1877 
1878 	switch (v->props.mode) {
1879 	case XFRM_MODE_TRANSPORT:
1880 		if (v->id.proto != IPPROTO_AH)
1881 			return 1;
1882 		else
1883 			return 3;
1884 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1885 	case XFRM_MODE_ROUTEOPTIMIZATION:
1886 	case XFRM_MODE_IN_TRIGGER:
1887 		return 2;
1888 #endif
1889 	case XFRM_MODE_TUNNEL:
1890 	case XFRM_MODE_BEET:
1891 		return 4;
1892 	}
1893 	return 5;
1894 }
1895 
1896 /* Rule for xfrm_tmpl:
1897  *
1898  * rule 1: select IPsec transport
1899  * rule 2: select MIPv6 RO or inbound trigger
1900  * rule 3: select IPsec tunnel
1901  * rule 4: others
1902  */
1903 static int __xfrm6_tmpl_sort_cmp(const void *p)
1904 {
1905 	const struct xfrm_tmpl *v = p;
1906 
1907 	switch (v->mode) {
1908 	case XFRM_MODE_TRANSPORT:
1909 		return 1;
1910 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1911 	case XFRM_MODE_ROUTEOPTIMIZATION:
1912 	case XFRM_MODE_IN_TRIGGER:
1913 		return 2;
1914 #endif
1915 	case XFRM_MODE_TUNNEL:
1916 	case XFRM_MODE_BEET:
1917 		return 3;
1918 	}
1919 	return 4;
1920 }
1921 #else
1922 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1923 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1924 
1925 static inline void
1926 __xfrm6_sort(void **dst, void **src, int n,
1927 	     int (*cmp)(const void *p), int maxclass)
1928 {
1929 	int i;
1930 
1931 	for (i = 0; i < n; i++)
1932 		dst[i] = src[i];
1933 }
1934 #endif /* CONFIG_IPV6 */
1935 
1936 void
1937 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1938 	       unsigned short family)
1939 {
1940 	int i;
1941 
1942 	if (family == AF_INET6)
1943 		__xfrm6_sort((void **)dst, (void **)src, n,
1944 			     __xfrm6_tmpl_sort_cmp, 5);
1945 	else
1946 		for (i = 0; i < n; i++)
1947 			dst[i] = src[i];
1948 }
1949 
1950 void
1951 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1952 		unsigned short family)
1953 {
1954 	int i;
1955 
1956 	if (family == AF_INET6)
1957 		__xfrm6_sort((void **)dst, (void **)src, n,
1958 			     __xfrm6_state_sort_cmp, 6);
1959 	else
1960 		for (i = 0; i < n; i++)
1961 			dst[i] = src[i];
1962 }
1963 #endif
1964 
1965 /* Silly enough, but I'm lazy to build resolution list */
1966 
1967 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1968 {
1969 	unsigned int h = xfrm_seq_hash(net, seq);
1970 	struct xfrm_state *x;
1971 
1972 	hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
1973 		if (x->km.seq == seq &&
1974 		    (mark & x->mark.m) == x->mark.v &&
1975 		    x->km.state == XFRM_STATE_ACQ) {
1976 			xfrm_state_hold(x);
1977 			return x;
1978 		}
1979 	}
1980 
1981 	return NULL;
1982 }
1983 
1984 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1985 {
1986 	struct xfrm_state *x;
1987 
1988 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1989 	x = __xfrm_find_acq_byseq(net, mark, seq);
1990 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1991 	return x;
1992 }
1993 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1994 
1995 u32 xfrm_get_acqseq(void)
1996 {
1997 	u32 res;
1998 	static atomic_t acqseq;
1999 
2000 	do {
2001 		res = atomic_inc_return(&acqseq);
2002 	} while (!res);
2003 
2004 	return res;
2005 }
2006 EXPORT_SYMBOL(xfrm_get_acqseq);
2007 
2008 int verify_spi_info(u8 proto, u32 min, u32 max)
2009 {
2010 	switch (proto) {
2011 	case IPPROTO_AH:
2012 	case IPPROTO_ESP:
2013 		break;
2014 
2015 	case IPPROTO_COMP:
2016 		/* IPCOMP spi is 16-bits. */
2017 		if (max >= 0x10000)
2018 			return -EINVAL;
2019 		break;
2020 
2021 	default:
2022 		return -EINVAL;
2023 	}
2024 
2025 	if (min > max)
2026 		return -EINVAL;
2027 
2028 	return 0;
2029 }
2030 EXPORT_SYMBOL(verify_spi_info);
2031 
2032 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2033 {
2034 	struct net *net = xs_net(x);
2035 	unsigned int h;
2036 	struct xfrm_state *x0;
2037 	int err = -ENOENT;
2038 	__be32 minspi = htonl(low);
2039 	__be32 maxspi = htonl(high);
2040 	__be32 newspi = 0;
2041 	u32 mark = x->mark.v & x->mark.m;
2042 
2043 	spin_lock_bh(&x->lock);
2044 	if (x->km.state == XFRM_STATE_DEAD)
2045 		goto unlock;
2046 
2047 	err = 0;
2048 	if (x->id.spi)
2049 		goto unlock;
2050 
2051 	err = -ENOENT;
2052 
2053 	if (minspi == maxspi) {
2054 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2055 		if (x0) {
2056 			xfrm_state_put(x0);
2057 			goto unlock;
2058 		}
2059 		newspi = minspi;
2060 	} else {
2061 		u32 spi = 0;
2062 		for (h = 0; h < high-low+1; h++) {
2063 			spi = low + prandom_u32()%(high-low+1);
2064 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2065 			if (x0 == NULL) {
2066 				newspi = htonl(spi);
2067 				break;
2068 			}
2069 			xfrm_state_put(x0);
2070 		}
2071 	}
2072 	if (newspi) {
2073 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2074 		x->id.spi = newspi;
2075 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2076 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2077 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2078 
2079 		err = 0;
2080 	}
2081 
2082 unlock:
2083 	spin_unlock_bh(&x->lock);
2084 
2085 	return err;
2086 }
2087 EXPORT_SYMBOL(xfrm_alloc_spi);
2088 
2089 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2090 				      struct xfrm_address_filter *filter)
2091 {
2092 	if (filter) {
2093 		if ((filter->family == AF_INET ||
2094 		     filter->family == AF_INET6) &&
2095 		    x->props.family != filter->family)
2096 			return false;
2097 
2098 		return addr_match(&x->props.saddr, &filter->saddr,
2099 				  filter->splen) &&
2100 		       addr_match(&x->id.daddr, &filter->daddr,
2101 				  filter->dplen);
2102 	}
2103 	return true;
2104 }
2105 
2106 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2107 		    int (*func)(struct xfrm_state *, int, void*),
2108 		    void *data)
2109 {
2110 	struct xfrm_state *state;
2111 	struct xfrm_state_walk *x;
2112 	int err = 0;
2113 
2114 	if (walk->seq != 0 && list_empty(&walk->all))
2115 		return 0;
2116 
2117 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2118 	if (list_empty(&walk->all))
2119 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2120 	else
2121 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2122 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2123 		if (x->state == XFRM_STATE_DEAD)
2124 			continue;
2125 		state = container_of(x, struct xfrm_state, km);
2126 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2127 			continue;
2128 		if (!__xfrm_state_filter_match(state, walk->filter))
2129 			continue;
2130 		err = func(state, walk->seq, data);
2131 		if (err) {
2132 			list_move_tail(&walk->all, &x->all);
2133 			goto out;
2134 		}
2135 		walk->seq++;
2136 	}
2137 	if (walk->seq == 0) {
2138 		err = -ENOENT;
2139 		goto out;
2140 	}
2141 	list_del_init(&walk->all);
2142 out:
2143 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2144 	return err;
2145 }
2146 EXPORT_SYMBOL(xfrm_state_walk);
2147 
2148 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2149 			  struct xfrm_address_filter *filter)
2150 {
2151 	INIT_LIST_HEAD(&walk->all);
2152 	walk->proto = proto;
2153 	walk->state = XFRM_STATE_DEAD;
2154 	walk->seq = 0;
2155 	walk->filter = filter;
2156 }
2157 EXPORT_SYMBOL(xfrm_state_walk_init);
2158 
2159 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2160 {
2161 	kfree(walk->filter);
2162 
2163 	if (list_empty(&walk->all))
2164 		return;
2165 
2166 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2167 	list_del(&walk->all);
2168 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2169 }
2170 EXPORT_SYMBOL(xfrm_state_walk_done);
2171 
2172 static void xfrm_replay_timer_handler(struct timer_list *t)
2173 {
2174 	struct xfrm_state *x = from_timer(x, t, rtimer);
2175 
2176 	spin_lock(&x->lock);
2177 
2178 	if (x->km.state == XFRM_STATE_VALID) {
2179 		if (xfrm_aevent_is_on(xs_net(x)))
2180 			xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2181 		else
2182 			x->xflags |= XFRM_TIME_DEFER;
2183 	}
2184 
2185 	spin_unlock(&x->lock);
2186 }
2187 
2188 static LIST_HEAD(xfrm_km_list);
2189 
2190 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2191 {
2192 	struct xfrm_mgr *km;
2193 
2194 	rcu_read_lock();
2195 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2196 		if (km->notify_policy)
2197 			km->notify_policy(xp, dir, c);
2198 	rcu_read_unlock();
2199 }
2200 
2201 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2202 {
2203 	struct xfrm_mgr *km;
2204 	rcu_read_lock();
2205 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2206 		if (km->notify)
2207 			km->notify(x, c);
2208 	rcu_read_unlock();
2209 }
2210 
2211 EXPORT_SYMBOL(km_policy_notify);
2212 EXPORT_SYMBOL(km_state_notify);
2213 
2214 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2215 {
2216 	struct km_event c;
2217 
2218 	c.data.hard = hard;
2219 	c.portid = portid;
2220 	c.event = XFRM_MSG_EXPIRE;
2221 	km_state_notify(x, &c);
2222 }
2223 
2224 EXPORT_SYMBOL(km_state_expired);
2225 /*
2226  * We send to all registered managers regardless of failure
2227  * We are happy with one success
2228 */
2229 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2230 {
2231 	int err = -EINVAL, acqret;
2232 	struct xfrm_mgr *km;
2233 
2234 	rcu_read_lock();
2235 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2236 		acqret = km->acquire(x, t, pol);
2237 		if (!acqret)
2238 			err = acqret;
2239 	}
2240 	rcu_read_unlock();
2241 	return err;
2242 }
2243 EXPORT_SYMBOL(km_query);
2244 
2245 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2246 {
2247 	int err = -EINVAL;
2248 	struct xfrm_mgr *km;
2249 
2250 	rcu_read_lock();
2251 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2252 		if (km->new_mapping)
2253 			err = km->new_mapping(x, ipaddr, sport);
2254 		if (!err)
2255 			break;
2256 	}
2257 	rcu_read_unlock();
2258 	return err;
2259 }
2260 EXPORT_SYMBOL(km_new_mapping);
2261 
2262 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2263 {
2264 	struct km_event c;
2265 
2266 	c.data.hard = hard;
2267 	c.portid = portid;
2268 	c.event = XFRM_MSG_POLEXPIRE;
2269 	km_policy_notify(pol, dir, &c);
2270 }
2271 EXPORT_SYMBOL(km_policy_expired);
2272 
2273 #ifdef CONFIG_XFRM_MIGRATE
2274 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2275 	       const struct xfrm_migrate *m, int num_migrate,
2276 	       const struct xfrm_kmaddress *k,
2277 	       const struct xfrm_encap_tmpl *encap)
2278 {
2279 	int err = -EINVAL;
2280 	int ret;
2281 	struct xfrm_mgr *km;
2282 
2283 	rcu_read_lock();
2284 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2285 		if (km->migrate) {
2286 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2287 					  encap);
2288 			if (!ret)
2289 				err = ret;
2290 		}
2291 	}
2292 	rcu_read_unlock();
2293 	return err;
2294 }
2295 EXPORT_SYMBOL(km_migrate);
2296 #endif
2297 
2298 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2299 {
2300 	int err = -EINVAL;
2301 	int ret;
2302 	struct xfrm_mgr *km;
2303 
2304 	rcu_read_lock();
2305 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2306 		if (km->report) {
2307 			ret = km->report(net, proto, sel, addr);
2308 			if (!ret)
2309 				err = ret;
2310 		}
2311 	}
2312 	rcu_read_unlock();
2313 	return err;
2314 }
2315 EXPORT_SYMBOL(km_report);
2316 
2317 static bool km_is_alive(const struct km_event *c)
2318 {
2319 	struct xfrm_mgr *km;
2320 	bool is_alive = false;
2321 
2322 	rcu_read_lock();
2323 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2324 		if (km->is_alive && km->is_alive(c)) {
2325 			is_alive = true;
2326 			break;
2327 		}
2328 	}
2329 	rcu_read_unlock();
2330 
2331 	return is_alive;
2332 }
2333 
2334 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2335 static DEFINE_SPINLOCK(xfrm_translator_lock);
2336 static struct xfrm_translator __rcu *xfrm_translator;
2337 
2338 struct xfrm_translator *xfrm_get_translator(void)
2339 {
2340 	struct xfrm_translator *xtr;
2341 
2342 	rcu_read_lock();
2343 	xtr = rcu_dereference(xfrm_translator);
2344 	if (unlikely(!xtr))
2345 		goto out;
2346 	if (!try_module_get(xtr->owner))
2347 		xtr = NULL;
2348 out:
2349 	rcu_read_unlock();
2350 	return xtr;
2351 }
2352 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2353 
2354 void xfrm_put_translator(struct xfrm_translator *xtr)
2355 {
2356 	module_put(xtr->owner);
2357 }
2358 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2359 
2360 int xfrm_register_translator(struct xfrm_translator *xtr)
2361 {
2362 	int err = 0;
2363 
2364 	spin_lock_bh(&xfrm_translator_lock);
2365 	if (unlikely(xfrm_translator != NULL))
2366 		err = -EEXIST;
2367 	else
2368 		rcu_assign_pointer(xfrm_translator, xtr);
2369 	spin_unlock_bh(&xfrm_translator_lock);
2370 
2371 	return err;
2372 }
2373 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2374 
2375 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2376 {
2377 	int err = 0;
2378 
2379 	spin_lock_bh(&xfrm_translator_lock);
2380 	if (likely(xfrm_translator != NULL)) {
2381 		if (rcu_access_pointer(xfrm_translator) != xtr)
2382 			err = -EINVAL;
2383 		else
2384 			RCU_INIT_POINTER(xfrm_translator, NULL);
2385 	}
2386 	spin_unlock_bh(&xfrm_translator_lock);
2387 	synchronize_rcu();
2388 
2389 	return err;
2390 }
2391 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2392 #endif
2393 
2394 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2395 {
2396 	int err;
2397 	u8 *data;
2398 	struct xfrm_mgr *km;
2399 	struct xfrm_policy *pol = NULL;
2400 
2401 	if (sockptr_is_null(optval) && !optlen) {
2402 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2403 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2404 		__sk_dst_reset(sk);
2405 		return 0;
2406 	}
2407 
2408 	if (optlen <= 0 || optlen > PAGE_SIZE)
2409 		return -EMSGSIZE;
2410 
2411 	data = memdup_sockptr(optval, optlen);
2412 	if (IS_ERR(data))
2413 		return PTR_ERR(data);
2414 
2415 	if (in_compat_syscall()) {
2416 		struct xfrm_translator *xtr = xfrm_get_translator();
2417 
2418 		if (!xtr) {
2419 			kfree(data);
2420 			return -EOPNOTSUPP;
2421 		}
2422 
2423 		err = xtr->xlate_user_policy_sockptr(&data, optlen);
2424 		xfrm_put_translator(xtr);
2425 		if (err) {
2426 			kfree(data);
2427 			return err;
2428 		}
2429 	}
2430 
2431 	err = -EINVAL;
2432 	rcu_read_lock();
2433 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2434 		pol = km->compile_policy(sk, optname, data,
2435 					 optlen, &err);
2436 		if (err >= 0)
2437 			break;
2438 	}
2439 	rcu_read_unlock();
2440 
2441 	if (err >= 0) {
2442 		xfrm_sk_policy_insert(sk, err, pol);
2443 		xfrm_pol_put(pol);
2444 		__sk_dst_reset(sk);
2445 		err = 0;
2446 	}
2447 
2448 	kfree(data);
2449 	return err;
2450 }
2451 EXPORT_SYMBOL(xfrm_user_policy);
2452 
2453 static DEFINE_SPINLOCK(xfrm_km_lock);
2454 
2455 int xfrm_register_km(struct xfrm_mgr *km)
2456 {
2457 	spin_lock_bh(&xfrm_km_lock);
2458 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2459 	spin_unlock_bh(&xfrm_km_lock);
2460 	return 0;
2461 }
2462 EXPORT_SYMBOL(xfrm_register_km);
2463 
2464 int xfrm_unregister_km(struct xfrm_mgr *km)
2465 {
2466 	spin_lock_bh(&xfrm_km_lock);
2467 	list_del_rcu(&km->list);
2468 	spin_unlock_bh(&xfrm_km_lock);
2469 	synchronize_rcu();
2470 	return 0;
2471 }
2472 EXPORT_SYMBOL(xfrm_unregister_km);
2473 
2474 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2475 {
2476 	int err = 0;
2477 
2478 	if (WARN_ON(afinfo->family >= NPROTO))
2479 		return -EAFNOSUPPORT;
2480 
2481 	spin_lock_bh(&xfrm_state_afinfo_lock);
2482 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2483 		err = -EEXIST;
2484 	else
2485 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2486 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2487 	return err;
2488 }
2489 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2490 
2491 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2492 {
2493 	int err = 0, family = afinfo->family;
2494 
2495 	if (WARN_ON(family >= NPROTO))
2496 		return -EAFNOSUPPORT;
2497 
2498 	spin_lock_bh(&xfrm_state_afinfo_lock);
2499 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2500 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2501 			err = -EINVAL;
2502 		else
2503 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2504 	}
2505 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2506 	synchronize_rcu();
2507 	return err;
2508 }
2509 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2510 
2511 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2512 {
2513 	if (unlikely(family >= NPROTO))
2514 		return NULL;
2515 
2516 	return rcu_dereference(xfrm_state_afinfo[family]);
2517 }
2518 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2519 
2520 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2521 {
2522 	struct xfrm_state_afinfo *afinfo;
2523 	if (unlikely(family >= NPROTO))
2524 		return NULL;
2525 	rcu_read_lock();
2526 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2527 	if (unlikely(!afinfo))
2528 		rcu_read_unlock();
2529 	return afinfo;
2530 }
2531 
2532 void xfrm_flush_gc(void)
2533 {
2534 	flush_work(&xfrm_state_gc_work);
2535 }
2536 EXPORT_SYMBOL(xfrm_flush_gc);
2537 
2538 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2539 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2540 {
2541 	if (x->tunnel) {
2542 		struct xfrm_state *t = x->tunnel;
2543 
2544 		if (atomic_read(&t->tunnel_users) == 2)
2545 			xfrm_state_delete(t);
2546 		atomic_dec(&t->tunnel_users);
2547 		xfrm_state_put_sync(t);
2548 		x->tunnel = NULL;
2549 	}
2550 }
2551 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2552 
2553 u32 __xfrm_state_mtu(struct xfrm_state *x, int mtu)
2554 {
2555 	const struct xfrm_type *type = READ_ONCE(x->type);
2556 	struct crypto_aead *aead;
2557 	u32 blksize, net_adj = 0;
2558 
2559 	if (x->km.state != XFRM_STATE_VALID ||
2560 	    !type || type->proto != IPPROTO_ESP)
2561 		return mtu - x->props.header_len;
2562 
2563 	aead = x->data;
2564 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2565 
2566 	switch (x->props.mode) {
2567 	case XFRM_MODE_TRANSPORT:
2568 	case XFRM_MODE_BEET:
2569 		if (x->props.family == AF_INET)
2570 			net_adj = sizeof(struct iphdr);
2571 		else if (x->props.family == AF_INET6)
2572 			net_adj = sizeof(struct ipv6hdr);
2573 		break;
2574 	case XFRM_MODE_TUNNEL:
2575 		break;
2576 	default:
2577 		WARN_ON_ONCE(1);
2578 		break;
2579 	}
2580 
2581 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2582 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2583 }
2584 EXPORT_SYMBOL_GPL(__xfrm_state_mtu);
2585 
2586 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2587 {
2588 	mtu = __xfrm_state_mtu(x, mtu);
2589 
2590 	if (x->props.family == AF_INET6 && mtu < IPV6_MIN_MTU)
2591 		return IPV6_MIN_MTU;
2592 
2593 	return mtu;
2594 }
2595 
2596 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2597 {
2598 	const struct xfrm_mode *inner_mode;
2599 	const struct xfrm_mode *outer_mode;
2600 	int family = x->props.family;
2601 	int err;
2602 
2603 	if (family == AF_INET &&
2604 	    xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2605 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2606 
2607 	err = -EPROTONOSUPPORT;
2608 
2609 	if (x->sel.family != AF_UNSPEC) {
2610 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2611 		if (inner_mode == NULL)
2612 			goto error;
2613 
2614 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2615 		    family != x->sel.family)
2616 			goto error;
2617 
2618 		x->inner_mode = *inner_mode;
2619 	} else {
2620 		const struct xfrm_mode *inner_mode_iaf;
2621 		int iafamily = AF_INET;
2622 
2623 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2624 		if (inner_mode == NULL)
2625 			goto error;
2626 
2627 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2628 			goto error;
2629 
2630 		x->inner_mode = *inner_mode;
2631 
2632 		if (x->props.family == AF_INET)
2633 			iafamily = AF_INET6;
2634 
2635 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2636 		if (inner_mode_iaf) {
2637 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2638 				x->inner_mode_iaf = *inner_mode_iaf;
2639 		}
2640 	}
2641 
2642 	x->type = xfrm_get_type(x->id.proto, family);
2643 	if (x->type == NULL)
2644 		goto error;
2645 
2646 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2647 
2648 	err = x->type->init_state(x);
2649 	if (err)
2650 		goto error;
2651 
2652 	outer_mode = xfrm_get_mode(x->props.mode, family);
2653 	if (!outer_mode) {
2654 		err = -EPROTONOSUPPORT;
2655 		goto error;
2656 	}
2657 
2658 	x->outer_mode = *outer_mode;
2659 	if (init_replay) {
2660 		err = xfrm_init_replay(x);
2661 		if (err)
2662 			goto error;
2663 	}
2664 
2665 error:
2666 	return err;
2667 }
2668 
2669 EXPORT_SYMBOL(__xfrm_init_state);
2670 
2671 int xfrm_init_state(struct xfrm_state *x)
2672 {
2673 	int err;
2674 
2675 	err = __xfrm_init_state(x, true, false);
2676 	if (!err)
2677 		x->km.state = XFRM_STATE_VALID;
2678 
2679 	return err;
2680 }
2681 
2682 EXPORT_SYMBOL(xfrm_init_state);
2683 
2684 int __net_init xfrm_state_init(struct net *net)
2685 {
2686 	unsigned int sz;
2687 
2688 	if (net_eq(net, &init_net))
2689 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2690 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2691 
2692 	INIT_LIST_HEAD(&net->xfrm.state_all);
2693 
2694 	sz = sizeof(struct hlist_head) * 8;
2695 
2696 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2697 	if (!net->xfrm.state_bydst)
2698 		goto out_bydst;
2699 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2700 	if (!net->xfrm.state_bysrc)
2701 		goto out_bysrc;
2702 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2703 	if (!net->xfrm.state_byspi)
2704 		goto out_byspi;
2705 	net->xfrm.state_byseq = xfrm_hash_alloc(sz);
2706 	if (!net->xfrm.state_byseq)
2707 		goto out_byseq;
2708 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2709 
2710 	net->xfrm.state_num = 0;
2711 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2712 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2713 	seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2714 			       &net->xfrm.xfrm_state_lock);
2715 	return 0;
2716 
2717 out_byseq:
2718 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2719 out_byspi:
2720 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2721 out_bysrc:
2722 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2723 out_bydst:
2724 	return -ENOMEM;
2725 }
2726 
2727 void xfrm_state_fini(struct net *net)
2728 {
2729 	unsigned int sz;
2730 
2731 	flush_work(&net->xfrm.state_hash_work);
2732 	flush_work(&xfrm_state_gc_work);
2733 	xfrm_state_flush(net, 0, false, true);
2734 
2735 	WARN_ON(!list_empty(&net->xfrm.state_all));
2736 
2737 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2738 	WARN_ON(!hlist_empty(net->xfrm.state_byseq));
2739 	xfrm_hash_free(net->xfrm.state_byseq, sz);
2740 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2741 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2742 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2743 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2744 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2745 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2746 }
2747 
2748 #ifdef CONFIG_AUDITSYSCALL
2749 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2750 				     struct audit_buffer *audit_buf)
2751 {
2752 	struct xfrm_sec_ctx *ctx = x->security;
2753 	u32 spi = ntohl(x->id.spi);
2754 
2755 	if (ctx)
2756 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2757 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2758 
2759 	switch (x->props.family) {
2760 	case AF_INET:
2761 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2762 				 &x->props.saddr.a4, &x->id.daddr.a4);
2763 		break;
2764 	case AF_INET6:
2765 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2766 				 x->props.saddr.a6, x->id.daddr.a6);
2767 		break;
2768 	}
2769 
2770 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2771 }
2772 
2773 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2774 				      struct audit_buffer *audit_buf)
2775 {
2776 	const struct iphdr *iph4;
2777 	const struct ipv6hdr *iph6;
2778 
2779 	switch (family) {
2780 	case AF_INET:
2781 		iph4 = ip_hdr(skb);
2782 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2783 				 &iph4->saddr, &iph4->daddr);
2784 		break;
2785 	case AF_INET6:
2786 		iph6 = ipv6_hdr(skb);
2787 		audit_log_format(audit_buf,
2788 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2789 				 &iph6->saddr, &iph6->daddr,
2790 				 iph6->flow_lbl[0] & 0x0f,
2791 				 iph6->flow_lbl[1],
2792 				 iph6->flow_lbl[2]);
2793 		break;
2794 	}
2795 }
2796 
2797 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2798 {
2799 	struct audit_buffer *audit_buf;
2800 
2801 	audit_buf = xfrm_audit_start("SAD-add");
2802 	if (audit_buf == NULL)
2803 		return;
2804 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2805 	xfrm_audit_helper_sainfo(x, audit_buf);
2806 	audit_log_format(audit_buf, " res=%u", result);
2807 	audit_log_end(audit_buf);
2808 }
2809 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2810 
2811 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2812 {
2813 	struct audit_buffer *audit_buf;
2814 
2815 	audit_buf = xfrm_audit_start("SAD-delete");
2816 	if (audit_buf == NULL)
2817 		return;
2818 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2819 	xfrm_audit_helper_sainfo(x, audit_buf);
2820 	audit_log_format(audit_buf, " res=%u", result);
2821 	audit_log_end(audit_buf);
2822 }
2823 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2824 
2825 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2826 				      struct sk_buff *skb)
2827 {
2828 	struct audit_buffer *audit_buf;
2829 	u32 spi;
2830 
2831 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2832 	if (audit_buf == NULL)
2833 		return;
2834 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2835 	/* don't record the sequence number because it's inherent in this kind
2836 	 * of audit message */
2837 	spi = ntohl(x->id.spi);
2838 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2839 	audit_log_end(audit_buf);
2840 }
2841 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2842 
2843 void xfrm_audit_state_replay(struct xfrm_state *x,
2844 			     struct sk_buff *skb, __be32 net_seq)
2845 {
2846 	struct audit_buffer *audit_buf;
2847 	u32 spi;
2848 
2849 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2850 	if (audit_buf == NULL)
2851 		return;
2852 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2853 	spi = ntohl(x->id.spi);
2854 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2855 			 spi, spi, ntohl(net_seq));
2856 	audit_log_end(audit_buf);
2857 }
2858 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2859 
2860 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2861 {
2862 	struct audit_buffer *audit_buf;
2863 
2864 	audit_buf = xfrm_audit_start("SA-notfound");
2865 	if (audit_buf == NULL)
2866 		return;
2867 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2868 	audit_log_end(audit_buf);
2869 }
2870 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2871 
2872 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2873 			       __be32 net_spi, __be32 net_seq)
2874 {
2875 	struct audit_buffer *audit_buf;
2876 	u32 spi;
2877 
2878 	audit_buf = xfrm_audit_start("SA-notfound");
2879 	if (audit_buf == NULL)
2880 		return;
2881 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2882 	spi = ntohl(net_spi);
2883 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2884 			 spi, spi, ntohl(net_seq));
2885 	audit_log_end(audit_buf);
2886 }
2887 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2888 
2889 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2890 			      struct sk_buff *skb, u8 proto)
2891 {
2892 	struct audit_buffer *audit_buf;
2893 	__be32 net_spi;
2894 	__be32 net_seq;
2895 
2896 	audit_buf = xfrm_audit_start("SA-icv-failure");
2897 	if (audit_buf == NULL)
2898 		return;
2899 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2900 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2901 		u32 spi = ntohl(net_spi);
2902 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2903 				 spi, spi, ntohl(net_seq));
2904 	}
2905 	audit_log_end(audit_buf);
2906 }
2907 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2908 #endif /* CONFIG_AUDITSYSCALL */
2909