xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision e8e0929d)
1 /*
2  * xfrm_state.c
3  *
4  * Changes:
5  *	Mitsuru KANDA @USAGI
6  * 	Kazunori MIYAZAWA @USAGI
7  * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8  * 		IPv6 support
9  * 	YOSHIFUJI Hideaki @USAGI
10  * 		Split up af-specific functions
11  *	Derek Atkins <derek@ihtfp.com>
12  *		Add UDP Encapsulation
13  *
14  */
15 
16 #include <linux/workqueue.h>
17 #include <net/xfrm.h>
18 #include <linux/pfkeyv2.h>
19 #include <linux/ipsec.h>
20 #include <linux/module.h>
21 #include <linux/cache.h>
22 #include <linux/audit.h>
23 #include <asm/uaccess.h>
24 
25 #include "xfrm_hash.h"
26 
27 /* Each xfrm_state may be linked to two tables:
28 
29    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
30    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
31       destination/tunnel endpoint. (output)
32  */
33 
34 static DEFINE_SPINLOCK(xfrm_state_lock);
35 
36 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
37 static unsigned int xfrm_state_genid;
38 
39 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
40 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
41 
42 #ifdef CONFIG_AUDITSYSCALL
43 static void xfrm_audit_state_replay(struct xfrm_state *x,
44 				    struct sk_buff *skb, __be32 net_seq);
45 #else
46 #define xfrm_audit_state_replay(x, s, sq)	do { ; } while (0)
47 #endif /* CONFIG_AUDITSYSCALL */
48 
49 static inline unsigned int xfrm_dst_hash(struct net *net,
50 					 xfrm_address_t *daddr,
51 					 xfrm_address_t *saddr,
52 					 u32 reqid,
53 					 unsigned short family)
54 {
55 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
56 }
57 
58 static inline unsigned int xfrm_src_hash(struct net *net,
59 					 xfrm_address_t *daddr,
60 					 xfrm_address_t *saddr,
61 					 unsigned short family)
62 {
63 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
64 }
65 
66 static inline unsigned int
67 xfrm_spi_hash(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
68 {
69 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
70 }
71 
72 static void xfrm_hash_transfer(struct hlist_head *list,
73 			       struct hlist_head *ndsttable,
74 			       struct hlist_head *nsrctable,
75 			       struct hlist_head *nspitable,
76 			       unsigned int nhashmask)
77 {
78 	struct hlist_node *entry, *tmp;
79 	struct xfrm_state *x;
80 
81 	hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
82 		unsigned int h;
83 
84 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
85 				    x->props.reqid, x->props.family,
86 				    nhashmask);
87 		hlist_add_head(&x->bydst, ndsttable+h);
88 
89 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
90 				    x->props.family,
91 				    nhashmask);
92 		hlist_add_head(&x->bysrc, nsrctable+h);
93 
94 		if (x->id.spi) {
95 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
96 					    x->id.proto, x->props.family,
97 					    nhashmask);
98 			hlist_add_head(&x->byspi, nspitable+h);
99 		}
100 	}
101 }
102 
103 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
104 {
105 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
106 }
107 
108 static DEFINE_MUTEX(hash_resize_mutex);
109 
110 static void xfrm_hash_resize(struct work_struct *work)
111 {
112 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
113 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
114 	unsigned long nsize, osize;
115 	unsigned int nhashmask, ohashmask;
116 	int i;
117 
118 	mutex_lock(&hash_resize_mutex);
119 
120 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
121 	ndst = xfrm_hash_alloc(nsize);
122 	if (!ndst)
123 		goto out_unlock;
124 	nsrc = xfrm_hash_alloc(nsize);
125 	if (!nsrc) {
126 		xfrm_hash_free(ndst, nsize);
127 		goto out_unlock;
128 	}
129 	nspi = xfrm_hash_alloc(nsize);
130 	if (!nspi) {
131 		xfrm_hash_free(ndst, nsize);
132 		xfrm_hash_free(nsrc, nsize);
133 		goto out_unlock;
134 	}
135 
136 	spin_lock_bh(&xfrm_state_lock);
137 
138 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
139 	for (i = net->xfrm.state_hmask; i >= 0; i--)
140 		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
141 				   nhashmask);
142 
143 	odst = net->xfrm.state_bydst;
144 	osrc = net->xfrm.state_bysrc;
145 	ospi = net->xfrm.state_byspi;
146 	ohashmask = net->xfrm.state_hmask;
147 
148 	net->xfrm.state_bydst = ndst;
149 	net->xfrm.state_bysrc = nsrc;
150 	net->xfrm.state_byspi = nspi;
151 	net->xfrm.state_hmask = nhashmask;
152 
153 	spin_unlock_bh(&xfrm_state_lock);
154 
155 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
156 	xfrm_hash_free(odst, osize);
157 	xfrm_hash_free(osrc, osize);
158 	xfrm_hash_free(ospi, osize);
159 
160 out_unlock:
161 	mutex_unlock(&hash_resize_mutex);
162 }
163 
164 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
165 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
166 
167 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
168 
169 int __xfrm_state_delete(struct xfrm_state *x);
170 
171 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
172 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
173 
174 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
175 {
176 	struct xfrm_state_afinfo *afinfo;
177 	if (unlikely(family >= NPROTO))
178 		return NULL;
179 	write_lock_bh(&xfrm_state_afinfo_lock);
180 	afinfo = xfrm_state_afinfo[family];
181 	if (unlikely(!afinfo))
182 		write_unlock_bh(&xfrm_state_afinfo_lock);
183 	return afinfo;
184 }
185 
186 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
187 	__releases(xfrm_state_afinfo_lock)
188 {
189 	write_unlock_bh(&xfrm_state_afinfo_lock);
190 }
191 
192 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
193 {
194 	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
195 	const struct xfrm_type **typemap;
196 	int err = 0;
197 
198 	if (unlikely(afinfo == NULL))
199 		return -EAFNOSUPPORT;
200 	typemap = afinfo->type_map;
201 
202 	if (likely(typemap[type->proto] == NULL))
203 		typemap[type->proto] = type;
204 	else
205 		err = -EEXIST;
206 	xfrm_state_unlock_afinfo(afinfo);
207 	return err;
208 }
209 EXPORT_SYMBOL(xfrm_register_type);
210 
211 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
212 {
213 	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
214 	const struct xfrm_type **typemap;
215 	int err = 0;
216 
217 	if (unlikely(afinfo == NULL))
218 		return -EAFNOSUPPORT;
219 	typemap = afinfo->type_map;
220 
221 	if (unlikely(typemap[type->proto] != type))
222 		err = -ENOENT;
223 	else
224 		typemap[type->proto] = NULL;
225 	xfrm_state_unlock_afinfo(afinfo);
226 	return err;
227 }
228 EXPORT_SYMBOL(xfrm_unregister_type);
229 
230 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
231 {
232 	struct xfrm_state_afinfo *afinfo;
233 	const struct xfrm_type **typemap;
234 	const struct xfrm_type *type;
235 	int modload_attempted = 0;
236 
237 retry:
238 	afinfo = xfrm_state_get_afinfo(family);
239 	if (unlikely(afinfo == NULL))
240 		return NULL;
241 	typemap = afinfo->type_map;
242 
243 	type = typemap[proto];
244 	if (unlikely(type && !try_module_get(type->owner)))
245 		type = NULL;
246 	if (!type && !modload_attempted) {
247 		xfrm_state_put_afinfo(afinfo);
248 		request_module("xfrm-type-%d-%d", family, proto);
249 		modload_attempted = 1;
250 		goto retry;
251 	}
252 
253 	xfrm_state_put_afinfo(afinfo);
254 	return type;
255 }
256 
257 static void xfrm_put_type(const struct xfrm_type *type)
258 {
259 	module_put(type->owner);
260 }
261 
262 int xfrm_register_mode(struct xfrm_mode *mode, int family)
263 {
264 	struct xfrm_state_afinfo *afinfo;
265 	struct xfrm_mode **modemap;
266 	int err;
267 
268 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
269 		return -EINVAL;
270 
271 	afinfo = xfrm_state_lock_afinfo(family);
272 	if (unlikely(afinfo == NULL))
273 		return -EAFNOSUPPORT;
274 
275 	err = -EEXIST;
276 	modemap = afinfo->mode_map;
277 	if (modemap[mode->encap])
278 		goto out;
279 
280 	err = -ENOENT;
281 	if (!try_module_get(afinfo->owner))
282 		goto out;
283 
284 	mode->afinfo = afinfo;
285 	modemap[mode->encap] = mode;
286 	err = 0;
287 
288 out:
289 	xfrm_state_unlock_afinfo(afinfo);
290 	return err;
291 }
292 EXPORT_SYMBOL(xfrm_register_mode);
293 
294 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
295 {
296 	struct xfrm_state_afinfo *afinfo;
297 	struct xfrm_mode **modemap;
298 	int err;
299 
300 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
301 		return -EINVAL;
302 
303 	afinfo = xfrm_state_lock_afinfo(family);
304 	if (unlikely(afinfo == NULL))
305 		return -EAFNOSUPPORT;
306 
307 	err = -ENOENT;
308 	modemap = afinfo->mode_map;
309 	if (likely(modemap[mode->encap] == mode)) {
310 		modemap[mode->encap] = NULL;
311 		module_put(mode->afinfo->owner);
312 		err = 0;
313 	}
314 
315 	xfrm_state_unlock_afinfo(afinfo);
316 	return err;
317 }
318 EXPORT_SYMBOL(xfrm_unregister_mode);
319 
320 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
321 {
322 	struct xfrm_state_afinfo *afinfo;
323 	struct xfrm_mode *mode;
324 	int modload_attempted = 0;
325 
326 	if (unlikely(encap >= XFRM_MODE_MAX))
327 		return NULL;
328 
329 retry:
330 	afinfo = xfrm_state_get_afinfo(family);
331 	if (unlikely(afinfo == NULL))
332 		return NULL;
333 
334 	mode = afinfo->mode_map[encap];
335 	if (unlikely(mode && !try_module_get(mode->owner)))
336 		mode = NULL;
337 	if (!mode && !modload_attempted) {
338 		xfrm_state_put_afinfo(afinfo);
339 		request_module("xfrm-mode-%d-%d", family, encap);
340 		modload_attempted = 1;
341 		goto retry;
342 	}
343 
344 	xfrm_state_put_afinfo(afinfo);
345 	return mode;
346 }
347 
348 static void xfrm_put_mode(struct xfrm_mode *mode)
349 {
350 	module_put(mode->owner);
351 }
352 
353 static void xfrm_state_gc_destroy(struct xfrm_state *x)
354 {
355 	del_timer_sync(&x->timer);
356 	del_timer_sync(&x->rtimer);
357 	kfree(x->aalg);
358 	kfree(x->ealg);
359 	kfree(x->calg);
360 	kfree(x->encap);
361 	kfree(x->coaddr);
362 	if (x->inner_mode)
363 		xfrm_put_mode(x->inner_mode);
364 	if (x->inner_mode_iaf)
365 		xfrm_put_mode(x->inner_mode_iaf);
366 	if (x->outer_mode)
367 		xfrm_put_mode(x->outer_mode);
368 	if (x->type) {
369 		x->type->destructor(x);
370 		xfrm_put_type(x->type);
371 	}
372 	security_xfrm_state_free(x);
373 	kfree(x);
374 }
375 
376 static void xfrm_state_gc_task(struct work_struct *work)
377 {
378 	struct net *net = container_of(work, struct net, xfrm.state_gc_work);
379 	struct xfrm_state *x;
380 	struct hlist_node *entry, *tmp;
381 	struct hlist_head gc_list;
382 
383 	spin_lock_bh(&xfrm_state_gc_lock);
384 	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
385 	spin_unlock_bh(&xfrm_state_gc_lock);
386 
387 	hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
388 		xfrm_state_gc_destroy(x);
389 
390 	wake_up(&net->xfrm.km_waitq);
391 }
392 
393 static inline unsigned long make_jiffies(long secs)
394 {
395 	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
396 		return MAX_SCHEDULE_TIMEOUT-1;
397 	else
398 		return secs*HZ;
399 }
400 
401 static void xfrm_timer_handler(unsigned long data)
402 {
403 	struct xfrm_state *x = (struct xfrm_state*)data;
404 	struct net *net = xs_net(x);
405 	unsigned long now = get_seconds();
406 	long next = LONG_MAX;
407 	int warn = 0;
408 	int err = 0;
409 
410 	spin_lock(&x->lock);
411 	if (x->km.state == XFRM_STATE_DEAD)
412 		goto out;
413 	if (x->km.state == XFRM_STATE_EXPIRED)
414 		goto expired;
415 	if (x->lft.hard_add_expires_seconds) {
416 		long tmo = x->lft.hard_add_expires_seconds +
417 			x->curlft.add_time - now;
418 		if (tmo <= 0)
419 			goto expired;
420 		if (tmo < next)
421 			next = tmo;
422 	}
423 	if (x->lft.hard_use_expires_seconds) {
424 		long tmo = x->lft.hard_use_expires_seconds +
425 			(x->curlft.use_time ? : now) - now;
426 		if (tmo <= 0)
427 			goto expired;
428 		if (tmo < next)
429 			next = tmo;
430 	}
431 	if (x->km.dying)
432 		goto resched;
433 	if (x->lft.soft_add_expires_seconds) {
434 		long tmo = x->lft.soft_add_expires_seconds +
435 			x->curlft.add_time - now;
436 		if (tmo <= 0)
437 			warn = 1;
438 		else if (tmo < next)
439 			next = tmo;
440 	}
441 	if (x->lft.soft_use_expires_seconds) {
442 		long tmo = x->lft.soft_use_expires_seconds +
443 			(x->curlft.use_time ? : now) - now;
444 		if (tmo <= 0)
445 			warn = 1;
446 		else if (tmo < next)
447 			next = tmo;
448 	}
449 
450 	x->km.dying = warn;
451 	if (warn)
452 		km_state_expired(x, 0, 0);
453 resched:
454 	if (next != LONG_MAX)
455 		mod_timer(&x->timer, jiffies + make_jiffies(next));
456 
457 	goto out;
458 
459 expired:
460 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
461 		x->km.state = XFRM_STATE_EXPIRED;
462 		wake_up(&net->xfrm.km_waitq);
463 		next = 2;
464 		goto resched;
465 	}
466 
467 	err = __xfrm_state_delete(x);
468 	if (!err && x->id.spi)
469 		km_state_expired(x, 1, 0);
470 
471 	xfrm_audit_state_delete(x, err ? 0 : 1,
472 				audit_get_loginuid(current),
473 				audit_get_sessionid(current), 0);
474 
475 out:
476 	spin_unlock(&x->lock);
477 }
478 
479 static void xfrm_replay_timer_handler(unsigned long data);
480 
481 struct xfrm_state *xfrm_state_alloc(struct net *net)
482 {
483 	struct xfrm_state *x;
484 
485 	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
486 
487 	if (x) {
488 		write_pnet(&x->xs_net, net);
489 		atomic_set(&x->refcnt, 1);
490 		atomic_set(&x->tunnel_users, 0);
491 		INIT_LIST_HEAD(&x->km.all);
492 		INIT_HLIST_NODE(&x->bydst);
493 		INIT_HLIST_NODE(&x->bysrc);
494 		INIT_HLIST_NODE(&x->byspi);
495 		setup_timer(&x->timer, xfrm_timer_handler, (unsigned long)x);
496 		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
497 				(unsigned long)x);
498 		x->curlft.add_time = get_seconds();
499 		x->lft.soft_byte_limit = XFRM_INF;
500 		x->lft.soft_packet_limit = XFRM_INF;
501 		x->lft.hard_byte_limit = XFRM_INF;
502 		x->lft.hard_packet_limit = XFRM_INF;
503 		x->replay_maxage = 0;
504 		x->replay_maxdiff = 0;
505 		x->inner_mode = NULL;
506 		x->inner_mode_iaf = NULL;
507 		spin_lock_init(&x->lock);
508 	}
509 	return x;
510 }
511 EXPORT_SYMBOL(xfrm_state_alloc);
512 
513 void __xfrm_state_destroy(struct xfrm_state *x)
514 {
515 	struct net *net = xs_net(x);
516 
517 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
518 
519 	spin_lock_bh(&xfrm_state_gc_lock);
520 	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
521 	spin_unlock_bh(&xfrm_state_gc_lock);
522 	schedule_work(&net->xfrm.state_gc_work);
523 }
524 EXPORT_SYMBOL(__xfrm_state_destroy);
525 
526 int __xfrm_state_delete(struct xfrm_state *x)
527 {
528 	struct net *net = xs_net(x);
529 	int err = -ESRCH;
530 
531 	if (x->km.state != XFRM_STATE_DEAD) {
532 		x->km.state = XFRM_STATE_DEAD;
533 		spin_lock(&xfrm_state_lock);
534 		list_del(&x->km.all);
535 		hlist_del(&x->bydst);
536 		hlist_del(&x->bysrc);
537 		if (x->id.spi)
538 			hlist_del(&x->byspi);
539 		net->xfrm.state_num--;
540 		spin_unlock(&xfrm_state_lock);
541 
542 		/* All xfrm_state objects are created by xfrm_state_alloc.
543 		 * The xfrm_state_alloc call gives a reference, and that
544 		 * is what we are dropping here.
545 		 */
546 		xfrm_state_put(x);
547 		err = 0;
548 	}
549 
550 	return err;
551 }
552 EXPORT_SYMBOL(__xfrm_state_delete);
553 
554 int xfrm_state_delete(struct xfrm_state *x)
555 {
556 	int err;
557 
558 	spin_lock_bh(&x->lock);
559 	err = __xfrm_state_delete(x);
560 	spin_unlock_bh(&x->lock);
561 
562 	return err;
563 }
564 EXPORT_SYMBOL(xfrm_state_delete);
565 
566 #ifdef CONFIG_SECURITY_NETWORK_XFRM
567 static inline int
568 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
569 {
570 	int i, err = 0;
571 
572 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
573 		struct hlist_node *entry;
574 		struct xfrm_state *x;
575 
576 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
577 			if (xfrm_id_proto_match(x->id.proto, proto) &&
578 			   (err = security_xfrm_state_delete(x)) != 0) {
579 				xfrm_audit_state_delete(x, 0,
580 							audit_info->loginuid,
581 							audit_info->sessionid,
582 							audit_info->secid);
583 				return err;
584 			}
585 		}
586 	}
587 
588 	return err;
589 }
590 #else
591 static inline int
592 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
593 {
594 	return 0;
595 }
596 #endif
597 
598 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
599 {
600 	int i, err = 0;
601 
602 	spin_lock_bh(&xfrm_state_lock);
603 	err = xfrm_state_flush_secctx_check(net, proto, audit_info);
604 	if (err)
605 		goto out;
606 
607 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
608 		struct hlist_node *entry;
609 		struct xfrm_state *x;
610 restart:
611 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
612 			if (!xfrm_state_kern(x) &&
613 			    xfrm_id_proto_match(x->id.proto, proto)) {
614 				xfrm_state_hold(x);
615 				spin_unlock_bh(&xfrm_state_lock);
616 
617 				err = xfrm_state_delete(x);
618 				xfrm_audit_state_delete(x, err ? 0 : 1,
619 							audit_info->loginuid,
620 							audit_info->sessionid,
621 							audit_info->secid);
622 				xfrm_state_put(x);
623 
624 				spin_lock_bh(&xfrm_state_lock);
625 				goto restart;
626 			}
627 		}
628 	}
629 	err = 0;
630 
631 out:
632 	spin_unlock_bh(&xfrm_state_lock);
633 	wake_up(&net->xfrm.km_waitq);
634 	return err;
635 }
636 EXPORT_SYMBOL(xfrm_state_flush);
637 
638 void xfrm_sad_getinfo(struct xfrmk_sadinfo *si)
639 {
640 	spin_lock_bh(&xfrm_state_lock);
641 	si->sadcnt = init_net.xfrm.state_num;
642 	si->sadhcnt = init_net.xfrm.state_hmask;
643 	si->sadhmcnt = xfrm_state_hashmax;
644 	spin_unlock_bh(&xfrm_state_lock);
645 }
646 EXPORT_SYMBOL(xfrm_sad_getinfo);
647 
648 static int
649 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
650 		  struct xfrm_tmpl *tmpl,
651 		  xfrm_address_t *daddr, xfrm_address_t *saddr,
652 		  unsigned short family)
653 {
654 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
655 	if (!afinfo)
656 		return -1;
657 	afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
658 	xfrm_state_put_afinfo(afinfo);
659 	return 0;
660 }
661 
662 static struct xfrm_state *__xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
663 {
664 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
665 	struct xfrm_state *x;
666 	struct hlist_node *entry;
667 
668 	hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
669 		if (x->props.family != family ||
670 		    x->id.spi       != spi ||
671 		    x->id.proto     != proto ||
672 		    xfrm_addr_cmp(&x->id.daddr, daddr, family))
673 			continue;
674 
675 		xfrm_state_hold(x);
676 		return x;
677 	}
678 
679 	return NULL;
680 }
681 
682 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
683 {
684 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
685 	struct xfrm_state *x;
686 	struct hlist_node *entry;
687 
688 	hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
689 		if (x->props.family != family ||
690 		    x->id.proto     != proto ||
691 		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
692 		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
693 			continue;
694 
695 		xfrm_state_hold(x);
696 		return x;
697 	}
698 
699 	return NULL;
700 }
701 
702 static inline struct xfrm_state *
703 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
704 {
705 	struct net *net = xs_net(x);
706 
707 	if (use_spi)
708 		return __xfrm_state_lookup(net, &x->id.daddr, x->id.spi,
709 					   x->id.proto, family);
710 	else
711 		return __xfrm_state_lookup_byaddr(net, &x->id.daddr,
712 						  &x->props.saddr,
713 						  x->id.proto, family);
714 }
715 
716 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
717 {
718 	if (have_hash_collision &&
719 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
720 	    net->xfrm.state_num > net->xfrm.state_hmask)
721 		schedule_work(&net->xfrm.state_hash_work);
722 }
723 
724 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
725 			       struct flowi *fl, unsigned short family,
726 			       xfrm_address_t *daddr, xfrm_address_t *saddr,
727 			       struct xfrm_state **best, int *acq_in_progress,
728 			       int *error)
729 {
730 	/* Resolution logic:
731 	 * 1. There is a valid state with matching selector. Done.
732 	 * 2. Valid state with inappropriate selector. Skip.
733 	 *
734 	 * Entering area of "sysdeps".
735 	 *
736 	 * 3. If state is not valid, selector is temporary, it selects
737 	 *    only session which triggered previous resolution. Key
738 	 *    manager will do something to install a state with proper
739 	 *    selector.
740 	 */
741 	if (x->km.state == XFRM_STATE_VALID) {
742 		if ((x->sel.family &&
743 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
744 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
745 			return;
746 
747 		if (!*best ||
748 		    (*best)->km.dying > x->km.dying ||
749 		    ((*best)->km.dying == x->km.dying &&
750 		     (*best)->curlft.add_time < x->curlft.add_time))
751 			*best = x;
752 	} else if (x->km.state == XFRM_STATE_ACQ) {
753 		*acq_in_progress = 1;
754 	} else if (x->km.state == XFRM_STATE_ERROR ||
755 		   x->km.state == XFRM_STATE_EXPIRED) {
756 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
757 		    security_xfrm_state_pol_flow_match(x, pol, fl))
758 			*error = -ESRCH;
759 	}
760 }
761 
762 struct xfrm_state *
763 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
764 		struct flowi *fl, struct xfrm_tmpl *tmpl,
765 		struct xfrm_policy *pol, int *err,
766 		unsigned short family)
767 {
768 	static xfrm_address_t saddr_wildcard = { };
769 	struct net *net = xp_net(pol);
770 	unsigned int h, h_wildcard;
771 	struct hlist_node *entry;
772 	struct xfrm_state *x, *x0, *to_put;
773 	int acquire_in_progress = 0;
774 	int error = 0;
775 	struct xfrm_state *best = NULL;
776 
777 	to_put = NULL;
778 
779 	spin_lock_bh(&xfrm_state_lock);
780 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
781 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
782 		if (x->props.family == family &&
783 		    x->props.reqid == tmpl->reqid &&
784 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
785 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
786 		    tmpl->mode == x->props.mode &&
787 		    tmpl->id.proto == x->id.proto &&
788 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
789 			xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
790 					   &best, &acquire_in_progress, &error);
791 	}
792 	if (best)
793 		goto found;
794 
795 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
796 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
797 		if (x->props.family == family &&
798 		    x->props.reqid == tmpl->reqid &&
799 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
800 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
801 		    tmpl->mode == x->props.mode &&
802 		    tmpl->id.proto == x->id.proto &&
803 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
804 			xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
805 					   &best, &acquire_in_progress, &error);
806 	}
807 
808 found:
809 	x = best;
810 	if (!x && !error && !acquire_in_progress) {
811 		if (tmpl->id.spi &&
812 		    (x0 = __xfrm_state_lookup(net, daddr, tmpl->id.spi,
813 					      tmpl->id.proto, family)) != NULL) {
814 			to_put = x0;
815 			error = -EEXIST;
816 			goto out;
817 		}
818 		x = xfrm_state_alloc(net);
819 		if (x == NULL) {
820 			error = -ENOMEM;
821 			goto out;
822 		}
823 		/* Initialize temporary selector matching only
824 		 * to current session. */
825 		xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
826 
827 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
828 		if (error) {
829 			x->km.state = XFRM_STATE_DEAD;
830 			to_put = x;
831 			x = NULL;
832 			goto out;
833 		}
834 
835 		if (km_query(x, tmpl, pol) == 0) {
836 			x->km.state = XFRM_STATE_ACQ;
837 			list_add(&x->km.all, &net->xfrm.state_all);
838 			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
839 			h = xfrm_src_hash(net, daddr, saddr, family);
840 			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
841 			if (x->id.spi) {
842 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
843 				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
844 			}
845 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
846 			x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ;
847 			add_timer(&x->timer);
848 			net->xfrm.state_num++;
849 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
850 		} else {
851 			x->km.state = XFRM_STATE_DEAD;
852 			to_put = x;
853 			x = NULL;
854 			error = -ESRCH;
855 		}
856 	}
857 out:
858 	if (x)
859 		xfrm_state_hold(x);
860 	else
861 		*err = acquire_in_progress ? -EAGAIN : error;
862 	spin_unlock_bh(&xfrm_state_lock);
863 	if (to_put)
864 		xfrm_state_put(to_put);
865 	return x;
866 }
867 
868 struct xfrm_state *
869 xfrm_stateonly_find(struct net *net,
870 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
871 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
872 {
873 	unsigned int h;
874 	struct xfrm_state *rx = NULL, *x = NULL;
875 	struct hlist_node *entry;
876 
877 	spin_lock(&xfrm_state_lock);
878 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
879 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
880 		if (x->props.family == family &&
881 		    x->props.reqid == reqid &&
882 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
883 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
884 		    mode == x->props.mode &&
885 		    proto == x->id.proto &&
886 		    x->km.state == XFRM_STATE_VALID) {
887 			rx = x;
888 			break;
889 		}
890 	}
891 
892 	if (rx)
893 		xfrm_state_hold(rx);
894 	spin_unlock(&xfrm_state_lock);
895 
896 
897 	return rx;
898 }
899 EXPORT_SYMBOL(xfrm_stateonly_find);
900 
901 static void __xfrm_state_insert(struct xfrm_state *x)
902 {
903 	struct net *net = xs_net(x);
904 	unsigned int h;
905 
906 	x->genid = ++xfrm_state_genid;
907 
908 	list_add(&x->km.all, &net->xfrm.state_all);
909 
910 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
911 			  x->props.reqid, x->props.family);
912 	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
913 
914 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
915 	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
916 
917 	if (x->id.spi) {
918 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
919 				  x->props.family);
920 
921 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
922 	}
923 
924 	mod_timer(&x->timer, jiffies + HZ);
925 	if (x->replay_maxage)
926 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
927 
928 	wake_up(&net->xfrm.km_waitq);
929 
930 	net->xfrm.state_num++;
931 
932 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
933 }
934 
935 /* xfrm_state_lock is held */
936 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
937 {
938 	struct net *net = xs_net(xnew);
939 	unsigned short family = xnew->props.family;
940 	u32 reqid = xnew->props.reqid;
941 	struct xfrm_state *x;
942 	struct hlist_node *entry;
943 	unsigned int h;
944 
945 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
946 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
947 		if (x->props.family	== family &&
948 		    x->props.reqid	== reqid &&
949 		    !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
950 		    !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
951 			x->genid = xfrm_state_genid;
952 	}
953 }
954 
955 void xfrm_state_insert(struct xfrm_state *x)
956 {
957 	spin_lock_bh(&xfrm_state_lock);
958 	__xfrm_state_bump_genids(x);
959 	__xfrm_state_insert(x);
960 	spin_unlock_bh(&xfrm_state_lock);
961 }
962 EXPORT_SYMBOL(xfrm_state_insert);
963 
964 /* xfrm_state_lock is held */
965 static struct xfrm_state *__find_acq_core(struct net *net, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
966 {
967 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
968 	struct hlist_node *entry;
969 	struct xfrm_state *x;
970 
971 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
972 		if (x->props.reqid  != reqid ||
973 		    x->props.mode   != mode ||
974 		    x->props.family != family ||
975 		    x->km.state     != XFRM_STATE_ACQ ||
976 		    x->id.spi       != 0 ||
977 		    x->id.proto	    != proto ||
978 		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
979 		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
980 			continue;
981 
982 		xfrm_state_hold(x);
983 		return x;
984 	}
985 
986 	if (!create)
987 		return NULL;
988 
989 	x = xfrm_state_alloc(net);
990 	if (likely(x)) {
991 		switch (family) {
992 		case AF_INET:
993 			x->sel.daddr.a4 = daddr->a4;
994 			x->sel.saddr.a4 = saddr->a4;
995 			x->sel.prefixlen_d = 32;
996 			x->sel.prefixlen_s = 32;
997 			x->props.saddr.a4 = saddr->a4;
998 			x->id.daddr.a4 = daddr->a4;
999 			break;
1000 
1001 		case AF_INET6:
1002 			ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1003 				       (struct in6_addr *)daddr);
1004 			ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1005 				       (struct in6_addr *)saddr);
1006 			x->sel.prefixlen_d = 128;
1007 			x->sel.prefixlen_s = 128;
1008 			ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1009 				       (struct in6_addr *)saddr);
1010 			ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1011 				       (struct in6_addr *)daddr);
1012 			break;
1013 		}
1014 
1015 		x->km.state = XFRM_STATE_ACQ;
1016 		x->id.proto = proto;
1017 		x->props.family = family;
1018 		x->props.mode = mode;
1019 		x->props.reqid = reqid;
1020 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1021 		xfrm_state_hold(x);
1022 		x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ;
1023 		add_timer(&x->timer);
1024 		list_add(&x->km.all, &net->xfrm.state_all);
1025 		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1026 		h = xfrm_src_hash(net, daddr, saddr, family);
1027 		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1028 
1029 		net->xfrm.state_num++;
1030 
1031 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1032 	}
1033 
1034 	return x;
1035 }
1036 
1037 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq);
1038 
1039 int xfrm_state_add(struct xfrm_state *x)
1040 {
1041 	struct net *net = xs_net(x);
1042 	struct xfrm_state *x1, *to_put;
1043 	int family;
1044 	int err;
1045 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1046 
1047 	family = x->props.family;
1048 
1049 	to_put = NULL;
1050 
1051 	spin_lock_bh(&xfrm_state_lock);
1052 
1053 	x1 = __xfrm_state_locate(x, use_spi, family);
1054 	if (x1) {
1055 		to_put = x1;
1056 		x1 = NULL;
1057 		err = -EEXIST;
1058 		goto out;
1059 	}
1060 
1061 	if (use_spi && x->km.seq) {
1062 		x1 = __xfrm_find_acq_byseq(net, x->km.seq);
1063 		if (x1 && ((x1->id.proto != x->id.proto) ||
1064 		    xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1065 			to_put = x1;
1066 			x1 = NULL;
1067 		}
1068 	}
1069 
1070 	if (use_spi && !x1)
1071 		x1 = __find_acq_core(net, family, x->props.mode, x->props.reqid,
1072 				     x->id.proto,
1073 				     &x->id.daddr, &x->props.saddr, 0);
1074 
1075 	__xfrm_state_bump_genids(x);
1076 	__xfrm_state_insert(x);
1077 	err = 0;
1078 
1079 out:
1080 	spin_unlock_bh(&xfrm_state_lock);
1081 
1082 	if (x1) {
1083 		xfrm_state_delete(x1);
1084 		xfrm_state_put(x1);
1085 	}
1086 
1087 	if (to_put)
1088 		xfrm_state_put(to_put);
1089 
1090 	return err;
1091 }
1092 EXPORT_SYMBOL(xfrm_state_add);
1093 
1094 #ifdef CONFIG_XFRM_MIGRATE
1095 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1096 {
1097 	struct net *net = xs_net(orig);
1098 	int err = -ENOMEM;
1099 	struct xfrm_state *x = xfrm_state_alloc(net);
1100 	if (!x)
1101 		goto error;
1102 
1103 	memcpy(&x->id, &orig->id, sizeof(x->id));
1104 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1105 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1106 	x->props.mode = orig->props.mode;
1107 	x->props.replay_window = orig->props.replay_window;
1108 	x->props.reqid = orig->props.reqid;
1109 	x->props.family = orig->props.family;
1110 	x->props.saddr = orig->props.saddr;
1111 
1112 	if (orig->aalg) {
1113 		x->aalg = xfrm_algo_clone(orig->aalg);
1114 		if (!x->aalg)
1115 			goto error;
1116 	}
1117 	x->props.aalgo = orig->props.aalgo;
1118 
1119 	if (orig->ealg) {
1120 		x->ealg = xfrm_algo_clone(orig->ealg);
1121 		if (!x->ealg)
1122 			goto error;
1123 	}
1124 	x->props.ealgo = orig->props.ealgo;
1125 
1126 	if (orig->calg) {
1127 		x->calg = xfrm_algo_clone(orig->calg);
1128 		if (!x->calg)
1129 			goto error;
1130 	}
1131 	x->props.calgo = orig->props.calgo;
1132 
1133 	if (orig->encap) {
1134 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1135 		if (!x->encap)
1136 			goto error;
1137 	}
1138 
1139 	if (orig->coaddr) {
1140 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1141 				    GFP_KERNEL);
1142 		if (!x->coaddr)
1143 			goto error;
1144 	}
1145 
1146 	err = xfrm_init_state(x);
1147 	if (err)
1148 		goto error;
1149 
1150 	x->props.flags = orig->props.flags;
1151 
1152 	x->curlft.add_time = orig->curlft.add_time;
1153 	x->km.state = orig->km.state;
1154 	x->km.seq = orig->km.seq;
1155 
1156 	return x;
1157 
1158  error:
1159 	if (errp)
1160 		*errp = err;
1161 	if (x) {
1162 		kfree(x->aalg);
1163 		kfree(x->ealg);
1164 		kfree(x->calg);
1165 		kfree(x->encap);
1166 		kfree(x->coaddr);
1167 	}
1168 	kfree(x);
1169 	return NULL;
1170 }
1171 
1172 /* xfrm_state_lock is held */
1173 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1174 {
1175 	unsigned int h;
1176 	struct xfrm_state *x;
1177 	struct hlist_node *entry;
1178 
1179 	if (m->reqid) {
1180 		h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1181 				  m->reqid, m->old_family);
1182 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1183 			if (x->props.mode != m->mode ||
1184 			    x->id.proto != m->proto)
1185 				continue;
1186 			if (m->reqid && x->props.reqid != m->reqid)
1187 				continue;
1188 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1189 					  m->old_family) ||
1190 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1191 					  m->old_family))
1192 				continue;
1193 			xfrm_state_hold(x);
1194 			return x;
1195 		}
1196 	} else {
1197 		h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1198 				  m->old_family);
1199 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1200 			if (x->props.mode != m->mode ||
1201 			    x->id.proto != m->proto)
1202 				continue;
1203 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1204 					  m->old_family) ||
1205 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1206 					  m->old_family))
1207 				continue;
1208 			xfrm_state_hold(x);
1209 			return x;
1210 		}
1211 	}
1212 
1213 	return NULL;
1214 }
1215 EXPORT_SYMBOL(xfrm_migrate_state_find);
1216 
1217 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1218 				       struct xfrm_migrate *m)
1219 {
1220 	struct xfrm_state *xc;
1221 	int err;
1222 
1223 	xc = xfrm_state_clone(x, &err);
1224 	if (!xc)
1225 		return NULL;
1226 
1227 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1228 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1229 
1230 	/* add state */
1231 	if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1232 		/* a care is needed when the destination address of the
1233 		   state is to be updated as it is a part of triplet */
1234 		xfrm_state_insert(xc);
1235 	} else {
1236 		if ((err = xfrm_state_add(xc)) < 0)
1237 			goto error;
1238 	}
1239 
1240 	return xc;
1241 error:
1242 	kfree(xc);
1243 	return NULL;
1244 }
1245 EXPORT_SYMBOL(xfrm_state_migrate);
1246 #endif
1247 
1248 int xfrm_state_update(struct xfrm_state *x)
1249 {
1250 	struct xfrm_state *x1, *to_put;
1251 	int err;
1252 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1253 
1254 	to_put = NULL;
1255 
1256 	spin_lock_bh(&xfrm_state_lock);
1257 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1258 
1259 	err = -ESRCH;
1260 	if (!x1)
1261 		goto out;
1262 
1263 	if (xfrm_state_kern(x1)) {
1264 		to_put = x1;
1265 		err = -EEXIST;
1266 		goto out;
1267 	}
1268 
1269 	if (x1->km.state == XFRM_STATE_ACQ) {
1270 		__xfrm_state_insert(x);
1271 		x = NULL;
1272 	}
1273 	err = 0;
1274 
1275 out:
1276 	spin_unlock_bh(&xfrm_state_lock);
1277 
1278 	if (to_put)
1279 		xfrm_state_put(to_put);
1280 
1281 	if (err)
1282 		return err;
1283 
1284 	if (!x) {
1285 		xfrm_state_delete(x1);
1286 		xfrm_state_put(x1);
1287 		return 0;
1288 	}
1289 
1290 	err = -EINVAL;
1291 	spin_lock_bh(&x1->lock);
1292 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1293 		if (x->encap && x1->encap)
1294 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1295 		if (x->coaddr && x1->coaddr) {
1296 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1297 		}
1298 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1299 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1300 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1301 		x1->km.dying = 0;
1302 
1303 		mod_timer(&x1->timer, jiffies + HZ);
1304 		if (x1->curlft.use_time)
1305 			xfrm_state_check_expire(x1);
1306 
1307 		err = 0;
1308 	}
1309 	spin_unlock_bh(&x1->lock);
1310 
1311 	xfrm_state_put(x1);
1312 
1313 	return err;
1314 }
1315 EXPORT_SYMBOL(xfrm_state_update);
1316 
1317 int xfrm_state_check_expire(struct xfrm_state *x)
1318 {
1319 	if (!x->curlft.use_time)
1320 		x->curlft.use_time = get_seconds();
1321 
1322 	if (x->km.state != XFRM_STATE_VALID)
1323 		return -EINVAL;
1324 
1325 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1326 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1327 		x->km.state = XFRM_STATE_EXPIRED;
1328 		mod_timer(&x->timer, jiffies);
1329 		return -EINVAL;
1330 	}
1331 
1332 	if (!x->km.dying &&
1333 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1334 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1335 		x->km.dying = 1;
1336 		km_state_expired(x, 0, 0);
1337 	}
1338 	return 0;
1339 }
1340 EXPORT_SYMBOL(xfrm_state_check_expire);
1341 
1342 struct xfrm_state *
1343 xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto,
1344 		  unsigned short family)
1345 {
1346 	struct xfrm_state *x;
1347 
1348 	spin_lock_bh(&xfrm_state_lock);
1349 	x = __xfrm_state_lookup(net, daddr, spi, proto, family);
1350 	spin_unlock_bh(&xfrm_state_lock);
1351 	return x;
1352 }
1353 EXPORT_SYMBOL(xfrm_state_lookup);
1354 
1355 struct xfrm_state *
1356 xfrm_state_lookup_byaddr(struct net *net,
1357 			 xfrm_address_t *daddr, xfrm_address_t *saddr,
1358 			 u8 proto, unsigned short family)
1359 {
1360 	struct xfrm_state *x;
1361 
1362 	spin_lock_bh(&xfrm_state_lock);
1363 	x = __xfrm_state_lookup_byaddr(net, daddr, saddr, proto, family);
1364 	spin_unlock_bh(&xfrm_state_lock);
1365 	return x;
1366 }
1367 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1368 
1369 struct xfrm_state *
1370 xfrm_find_acq(struct net *net, u8 mode, u32 reqid, u8 proto,
1371 	      xfrm_address_t *daddr, xfrm_address_t *saddr,
1372 	      int create, unsigned short family)
1373 {
1374 	struct xfrm_state *x;
1375 
1376 	spin_lock_bh(&xfrm_state_lock);
1377 	x = __find_acq_core(net, family, mode, reqid, proto, daddr, saddr, create);
1378 	spin_unlock_bh(&xfrm_state_lock);
1379 
1380 	return x;
1381 }
1382 EXPORT_SYMBOL(xfrm_find_acq);
1383 
1384 #ifdef CONFIG_XFRM_SUB_POLICY
1385 int
1386 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1387 	       unsigned short family)
1388 {
1389 	int err = 0;
1390 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1391 	if (!afinfo)
1392 		return -EAFNOSUPPORT;
1393 
1394 	spin_lock_bh(&xfrm_state_lock);
1395 	if (afinfo->tmpl_sort)
1396 		err = afinfo->tmpl_sort(dst, src, n);
1397 	spin_unlock_bh(&xfrm_state_lock);
1398 	xfrm_state_put_afinfo(afinfo);
1399 	return err;
1400 }
1401 EXPORT_SYMBOL(xfrm_tmpl_sort);
1402 
1403 int
1404 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1405 		unsigned short family)
1406 {
1407 	int err = 0;
1408 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1409 	if (!afinfo)
1410 		return -EAFNOSUPPORT;
1411 
1412 	spin_lock_bh(&xfrm_state_lock);
1413 	if (afinfo->state_sort)
1414 		err = afinfo->state_sort(dst, src, n);
1415 	spin_unlock_bh(&xfrm_state_lock);
1416 	xfrm_state_put_afinfo(afinfo);
1417 	return err;
1418 }
1419 EXPORT_SYMBOL(xfrm_state_sort);
1420 #endif
1421 
1422 /* Silly enough, but I'm lazy to build resolution list */
1423 
1424 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq)
1425 {
1426 	int i;
1427 
1428 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1429 		struct hlist_node *entry;
1430 		struct xfrm_state *x;
1431 
1432 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1433 			if (x->km.seq == seq &&
1434 			    x->km.state == XFRM_STATE_ACQ) {
1435 				xfrm_state_hold(x);
1436 				return x;
1437 			}
1438 		}
1439 	}
1440 	return NULL;
1441 }
1442 
1443 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 seq)
1444 {
1445 	struct xfrm_state *x;
1446 
1447 	spin_lock_bh(&xfrm_state_lock);
1448 	x = __xfrm_find_acq_byseq(net, seq);
1449 	spin_unlock_bh(&xfrm_state_lock);
1450 	return x;
1451 }
1452 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1453 
1454 u32 xfrm_get_acqseq(void)
1455 {
1456 	u32 res;
1457 	static u32 acqseq;
1458 	static DEFINE_SPINLOCK(acqseq_lock);
1459 
1460 	spin_lock_bh(&acqseq_lock);
1461 	res = (++acqseq ? : ++acqseq);
1462 	spin_unlock_bh(&acqseq_lock);
1463 	return res;
1464 }
1465 EXPORT_SYMBOL(xfrm_get_acqseq);
1466 
1467 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1468 {
1469 	struct net *net = xs_net(x);
1470 	unsigned int h;
1471 	struct xfrm_state *x0;
1472 	int err = -ENOENT;
1473 	__be32 minspi = htonl(low);
1474 	__be32 maxspi = htonl(high);
1475 
1476 	spin_lock_bh(&x->lock);
1477 	if (x->km.state == XFRM_STATE_DEAD)
1478 		goto unlock;
1479 
1480 	err = 0;
1481 	if (x->id.spi)
1482 		goto unlock;
1483 
1484 	err = -ENOENT;
1485 
1486 	if (minspi == maxspi) {
1487 		x0 = xfrm_state_lookup(net, &x->id.daddr, minspi, x->id.proto, x->props.family);
1488 		if (x0) {
1489 			xfrm_state_put(x0);
1490 			goto unlock;
1491 		}
1492 		x->id.spi = minspi;
1493 	} else {
1494 		u32 spi = 0;
1495 		for (h=0; h<high-low+1; h++) {
1496 			spi = low + net_random()%(high-low+1);
1497 			x0 = xfrm_state_lookup(net, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1498 			if (x0 == NULL) {
1499 				x->id.spi = htonl(spi);
1500 				break;
1501 			}
1502 			xfrm_state_put(x0);
1503 		}
1504 	}
1505 	if (x->id.spi) {
1506 		spin_lock_bh(&xfrm_state_lock);
1507 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1508 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1509 		spin_unlock_bh(&xfrm_state_lock);
1510 
1511 		err = 0;
1512 	}
1513 
1514 unlock:
1515 	spin_unlock_bh(&x->lock);
1516 
1517 	return err;
1518 }
1519 EXPORT_SYMBOL(xfrm_alloc_spi);
1520 
1521 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1522 		    int (*func)(struct xfrm_state *, int, void*),
1523 		    void *data)
1524 {
1525 	struct xfrm_state *state;
1526 	struct xfrm_state_walk *x;
1527 	int err = 0;
1528 
1529 	if (walk->seq != 0 && list_empty(&walk->all))
1530 		return 0;
1531 
1532 	spin_lock_bh(&xfrm_state_lock);
1533 	if (list_empty(&walk->all))
1534 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1535 	else
1536 		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1537 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1538 		if (x->state == XFRM_STATE_DEAD)
1539 			continue;
1540 		state = container_of(x, struct xfrm_state, km);
1541 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1542 			continue;
1543 		err = func(state, walk->seq, data);
1544 		if (err) {
1545 			list_move_tail(&walk->all, &x->all);
1546 			goto out;
1547 		}
1548 		walk->seq++;
1549 	}
1550 	if (walk->seq == 0) {
1551 		err = -ENOENT;
1552 		goto out;
1553 	}
1554 	list_del_init(&walk->all);
1555 out:
1556 	spin_unlock_bh(&xfrm_state_lock);
1557 	return err;
1558 }
1559 EXPORT_SYMBOL(xfrm_state_walk);
1560 
1561 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1562 {
1563 	INIT_LIST_HEAD(&walk->all);
1564 	walk->proto = proto;
1565 	walk->state = XFRM_STATE_DEAD;
1566 	walk->seq = 0;
1567 }
1568 EXPORT_SYMBOL(xfrm_state_walk_init);
1569 
1570 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1571 {
1572 	if (list_empty(&walk->all))
1573 		return;
1574 
1575 	spin_lock_bh(&xfrm_state_lock);
1576 	list_del(&walk->all);
1577 	spin_unlock_bh(&xfrm_state_lock);
1578 }
1579 EXPORT_SYMBOL(xfrm_state_walk_done);
1580 
1581 
1582 void xfrm_replay_notify(struct xfrm_state *x, int event)
1583 {
1584 	struct km_event c;
1585 	/* we send notify messages in case
1586 	 *  1. we updated on of the sequence numbers, and the seqno difference
1587 	 *     is at least x->replay_maxdiff, in this case we also update the
1588 	 *     timeout of our timer function
1589 	 *  2. if x->replay_maxage has elapsed since last update,
1590 	 *     and there were changes
1591 	 *
1592 	 *  The state structure must be locked!
1593 	 */
1594 
1595 	switch (event) {
1596 	case XFRM_REPLAY_UPDATE:
1597 		if (x->replay_maxdiff &&
1598 		    (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1599 		    (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1600 			if (x->xflags & XFRM_TIME_DEFER)
1601 				event = XFRM_REPLAY_TIMEOUT;
1602 			else
1603 				return;
1604 		}
1605 
1606 		break;
1607 
1608 	case XFRM_REPLAY_TIMEOUT:
1609 		if ((x->replay.seq == x->preplay.seq) &&
1610 		    (x->replay.bitmap == x->preplay.bitmap) &&
1611 		    (x->replay.oseq == x->preplay.oseq)) {
1612 			x->xflags |= XFRM_TIME_DEFER;
1613 			return;
1614 		}
1615 
1616 		break;
1617 	}
1618 
1619 	memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1620 	c.event = XFRM_MSG_NEWAE;
1621 	c.data.aevent = event;
1622 	km_state_notify(x, &c);
1623 
1624 	if (x->replay_maxage &&
1625 	    !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1626 		x->xflags &= ~XFRM_TIME_DEFER;
1627 }
1628 
1629 static void xfrm_replay_timer_handler(unsigned long data)
1630 {
1631 	struct xfrm_state *x = (struct xfrm_state*)data;
1632 
1633 	spin_lock(&x->lock);
1634 
1635 	if (x->km.state == XFRM_STATE_VALID) {
1636 		if (xfrm_aevent_is_on(xs_net(x)))
1637 			xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1638 		else
1639 			x->xflags |= XFRM_TIME_DEFER;
1640 	}
1641 
1642 	spin_unlock(&x->lock);
1643 }
1644 
1645 int xfrm_replay_check(struct xfrm_state *x,
1646 		      struct sk_buff *skb, __be32 net_seq)
1647 {
1648 	u32 diff;
1649 	u32 seq = ntohl(net_seq);
1650 
1651 	if (unlikely(seq == 0))
1652 		goto err;
1653 
1654 	if (likely(seq > x->replay.seq))
1655 		return 0;
1656 
1657 	diff = x->replay.seq - seq;
1658 	if (diff >= min_t(unsigned int, x->props.replay_window,
1659 			  sizeof(x->replay.bitmap) * 8)) {
1660 		x->stats.replay_window++;
1661 		goto err;
1662 	}
1663 
1664 	if (x->replay.bitmap & (1U << diff)) {
1665 		x->stats.replay++;
1666 		goto err;
1667 	}
1668 	return 0;
1669 
1670 err:
1671 	xfrm_audit_state_replay(x, skb, net_seq);
1672 	return -EINVAL;
1673 }
1674 
1675 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1676 {
1677 	u32 diff;
1678 	u32 seq = ntohl(net_seq);
1679 
1680 	if (seq > x->replay.seq) {
1681 		diff = seq - x->replay.seq;
1682 		if (diff < x->props.replay_window)
1683 			x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1684 		else
1685 			x->replay.bitmap = 1;
1686 		x->replay.seq = seq;
1687 	} else {
1688 		diff = x->replay.seq - seq;
1689 		x->replay.bitmap |= (1U << diff);
1690 	}
1691 
1692 	if (xfrm_aevent_is_on(xs_net(x)))
1693 		xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1694 }
1695 
1696 static LIST_HEAD(xfrm_km_list);
1697 static DEFINE_RWLOCK(xfrm_km_lock);
1698 
1699 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1700 {
1701 	struct xfrm_mgr *km;
1702 
1703 	read_lock(&xfrm_km_lock);
1704 	list_for_each_entry(km, &xfrm_km_list, list)
1705 		if (km->notify_policy)
1706 			km->notify_policy(xp, dir, c);
1707 	read_unlock(&xfrm_km_lock);
1708 }
1709 
1710 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1711 {
1712 	struct xfrm_mgr *km;
1713 	read_lock(&xfrm_km_lock);
1714 	list_for_each_entry(km, &xfrm_km_list, list)
1715 		if (km->notify)
1716 			km->notify(x, c);
1717 	read_unlock(&xfrm_km_lock);
1718 }
1719 
1720 EXPORT_SYMBOL(km_policy_notify);
1721 EXPORT_SYMBOL(km_state_notify);
1722 
1723 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1724 {
1725 	struct net *net = xs_net(x);
1726 	struct km_event c;
1727 
1728 	c.data.hard = hard;
1729 	c.pid = pid;
1730 	c.event = XFRM_MSG_EXPIRE;
1731 	km_state_notify(x, &c);
1732 
1733 	if (hard)
1734 		wake_up(&net->xfrm.km_waitq);
1735 }
1736 
1737 EXPORT_SYMBOL(km_state_expired);
1738 /*
1739  * We send to all registered managers regardless of failure
1740  * We are happy with one success
1741 */
1742 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1743 {
1744 	int err = -EINVAL, acqret;
1745 	struct xfrm_mgr *km;
1746 
1747 	read_lock(&xfrm_km_lock);
1748 	list_for_each_entry(km, &xfrm_km_list, list) {
1749 		acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1750 		if (!acqret)
1751 			err = acqret;
1752 	}
1753 	read_unlock(&xfrm_km_lock);
1754 	return err;
1755 }
1756 EXPORT_SYMBOL(km_query);
1757 
1758 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1759 {
1760 	int err = -EINVAL;
1761 	struct xfrm_mgr *km;
1762 
1763 	read_lock(&xfrm_km_lock);
1764 	list_for_each_entry(km, &xfrm_km_list, list) {
1765 		if (km->new_mapping)
1766 			err = km->new_mapping(x, ipaddr, sport);
1767 		if (!err)
1768 			break;
1769 	}
1770 	read_unlock(&xfrm_km_lock);
1771 	return err;
1772 }
1773 EXPORT_SYMBOL(km_new_mapping);
1774 
1775 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1776 {
1777 	struct net *net = xp_net(pol);
1778 	struct km_event c;
1779 
1780 	c.data.hard = hard;
1781 	c.pid = pid;
1782 	c.event = XFRM_MSG_POLEXPIRE;
1783 	km_policy_notify(pol, dir, &c);
1784 
1785 	if (hard)
1786 		wake_up(&net->xfrm.km_waitq);
1787 }
1788 EXPORT_SYMBOL(km_policy_expired);
1789 
1790 #ifdef CONFIG_XFRM_MIGRATE
1791 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1792 	       struct xfrm_migrate *m, int num_migrate,
1793 	       struct xfrm_kmaddress *k)
1794 {
1795 	int err = -EINVAL;
1796 	int ret;
1797 	struct xfrm_mgr *km;
1798 
1799 	read_lock(&xfrm_km_lock);
1800 	list_for_each_entry(km, &xfrm_km_list, list) {
1801 		if (km->migrate) {
1802 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1803 			if (!ret)
1804 				err = ret;
1805 		}
1806 	}
1807 	read_unlock(&xfrm_km_lock);
1808 	return err;
1809 }
1810 EXPORT_SYMBOL(km_migrate);
1811 #endif
1812 
1813 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1814 {
1815 	int err = -EINVAL;
1816 	int ret;
1817 	struct xfrm_mgr *km;
1818 
1819 	read_lock(&xfrm_km_lock);
1820 	list_for_each_entry(km, &xfrm_km_list, list) {
1821 		if (km->report) {
1822 			ret = km->report(net, proto, sel, addr);
1823 			if (!ret)
1824 				err = ret;
1825 		}
1826 	}
1827 	read_unlock(&xfrm_km_lock);
1828 	return err;
1829 }
1830 EXPORT_SYMBOL(km_report);
1831 
1832 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1833 {
1834 	int err;
1835 	u8 *data;
1836 	struct xfrm_mgr *km;
1837 	struct xfrm_policy *pol = NULL;
1838 
1839 	if (optlen <= 0 || optlen > PAGE_SIZE)
1840 		return -EMSGSIZE;
1841 
1842 	data = kmalloc(optlen, GFP_KERNEL);
1843 	if (!data)
1844 		return -ENOMEM;
1845 
1846 	err = -EFAULT;
1847 	if (copy_from_user(data, optval, optlen))
1848 		goto out;
1849 
1850 	err = -EINVAL;
1851 	read_lock(&xfrm_km_lock);
1852 	list_for_each_entry(km, &xfrm_km_list, list) {
1853 		pol = km->compile_policy(sk, optname, data,
1854 					 optlen, &err);
1855 		if (err >= 0)
1856 			break;
1857 	}
1858 	read_unlock(&xfrm_km_lock);
1859 
1860 	if (err >= 0) {
1861 		xfrm_sk_policy_insert(sk, err, pol);
1862 		xfrm_pol_put(pol);
1863 		err = 0;
1864 	}
1865 
1866 out:
1867 	kfree(data);
1868 	return err;
1869 }
1870 EXPORT_SYMBOL(xfrm_user_policy);
1871 
1872 int xfrm_register_km(struct xfrm_mgr *km)
1873 {
1874 	write_lock_bh(&xfrm_km_lock);
1875 	list_add_tail(&km->list, &xfrm_km_list);
1876 	write_unlock_bh(&xfrm_km_lock);
1877 	return 0;
1878 }
1879 EXPORT_SYMBOL(xfrm_register_km);
1880 
1881 int xfrm_unregister_km(struct xfrm_mgr *km)
1882 {
1883 	write_lock_bh(&xfrm_km_lock);
1884 	list_del(&km->list);
1885 	write_unlock_bh(&xfrm_km_lock);
1886 	return 0;
1887 }
1888 EXPORT_SYMBOL(xfrm_unregister_km);
1889 
1890 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1891 {
1892 	int err = 0;
1893 	if (unlikely(afinfo == NULL))
1894 		return -EINVAL;
1895 	if (unlikely(afinfo->family >= NPROTO))
1896 		return -EAFNOSUPPORT;
1897 	write_lock_bh(&xfrm_state_afinfo_lock);
1898 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1899 		err = -ENOBUFS;
1900 	else
1901 		xfrm_state_afinfo[afinfo->family] = afinfo;
1902 	write_unlock_bh(&xfrm_state_afinfo_lock);
1903 	return err;
1904 }
1905 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1906 
1907 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1908 {
1909 	int err = 0;
1910 	if (unlikely(afinfo == NULL))
1911 		return -EINVAL;
1912 	if (unlikely(afinfo->family >= NPROTO))
1913 		return -EAFNOSUPPORT;
1914 	write_lock_bh(&xfrm_state_afinfo_lock);
1915 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1916 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1917 			err = -EINVAL;
1918 		else
1919 			xfrm_state_afinfo[afinfo->family] = NULL;
1920 	}
1921 	write_unlock_bh(&xfrm_state_afinfo_lock);
1922 	return err;
1923 }
1924 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1925 
1926 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1927 {
1928 	struct xfrm_state_afinfo *afinfo;
1929 	if (unlikely(family >= NPROTO))
1930 		return NULL;
1931 	read_lock(&xfrm_state_afinfo_lock);
1932 	afinfo = xfrm_state_afinfo[family];
1933 	if (unlikely(!afinfo))
1934 		read_unlock(&xfrm_state_afinfo_lock);
1935 	return afinfo;
1936 }
1937 
1938 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1939 	__releases(xfrm_state_afinfo_lock)
1940 {
1941 	read_unlock(&xfrm_state_afinfo_lock);
1942 }
1943 
1944 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1945 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1946 {
1947 	if (x->tunnel) {
1948 		struct xfrm_state *t = x->tunnel;
1949 
1950 		if (atomic_read(&t->tunnel_users) == 2)
1951 			xfrm_state_delete(t);
1952 		atomic_dec(&t->tunnel_users);
1953 		xfrm_state_put(t);
1954 		x->tunnel = NULL;
1955 	}
1956 }
1957 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1958 
1959 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1960 {
1961 	int res;
1962 
1963 	spin_lock_bh(&x->lock);
1964 	if (x->km.state == XFRM_STATE_VALID &&
1965 	    x->type && x->type->get_mtu)
1966 		res = x->type->get_mtu(x, mtu);
1967 	else
1968 		res = mtu - x->props.header_len;
1969 	spin_unlock_bh(&x->lock);
1970 	return res;
1971 }
1972 
1973 int xfrm_init_state(struct xfrm_state *x)
1974 {
1975 	struct xfrm_state_afinfo *afinfo;
1976 	struct xfrm_mode *inner_mode;
1977 	int family = x->props.family;
1978 	int err;
1979 
1980 	err = -EAFNOSUPPORT;
1981 	afinfo = xfrm_state_get_afinfo(family);
1982 	if (!afinfo)
1983 		goto error;
1984 
1985 	err = 0;
1986 	if (afinfo->init_flags)
1987 		err = afinfo->init_flags(x);
1988 
1989 	xfrm_state_put_afinfo(afinfo);
1990 
1991 	if (err)
1992 		goto error;
1993 
1994 	err = -EPROTONOSUPPORT;
1995 
1996 	if (x->sel.family != AF_UNSPEC) {
1997 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1998 		if (inner_mode == NULL)
1999 			goto error;
2000 
2001 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2002 		    family != x->sel.family) {
2003 			xfrm_put_mode(inner_mode);
2004 			goto error;
2005 		}
2006 
2007 		x->inner_mode = inner_mode;
2008 	} else {
2009 		struct xfrm_mode *inner_mode_iaf;
2010 		int iafamily = AF_INET;
2011 
2012 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2013 		if (inner_mode == NULL)
2014 			goto error;
2015 
2016 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2017 			xfrm_put_mode(inner_mode);
2018 			goto error;
2019 		}
2020 		x->inner_mode = inner_mode;
2021 
2022 		if (x->props.family == AF_INET)
2023 			iafamily = AF_INET6;
2024 
2025 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2026 		if (inner_mode_iaf) {
2027 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2028 				x->inner_mode_iaf = inner_mode_iaf;
2029 			else
2030 				xfrm_put_mode(inner_mode_iaf);
2031 		}
2032 	}
2033 
2034 	x->type = xfrm_get_type(x->id.proto, family);
2035 	if (x->type == NULL)
2036 		goto error;
2037 
2038 	err = x->type->init_state(x);
2039 	if (err)
2040 		goto error;
2041 
2042 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2043 	if (x->outer_mode == NULL)
2044 		goto error;
2045 
2046 	x->km.state = XFRM_STATE_VALID;
2047 
2048 error:
2049 	return err;
2050 }
2051 
2052 EXPORT_SYMBOL(xfrm_init_state);
2053 
2054 int __net_init xfrm_state_init(struct net *net)
2055 {
2056 	unsigned int sz;
2057 
2058 	INIT_LIST_HEAD(&net->xfrm.state_all);
2059 
2060 	sz = sizeof(struct hlist_head) * 8;
2061 
2062 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2063 	if (!net->xfrm.state_bydst)
2064 		goto out_bydst;
2065 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2066 	if (!net->xfrm.state_bysrc)
2067 		goto out_bysrc;
2068 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2069 	if (!net->xfrm.state_byspi)
2070 		goto out_byspi;
2071 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2072 
2073 	net->xfrm.state_num = 0;
2074 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2075 	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2076 	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2077 	init_waitqueue_head(&net->xfrm.km_waitq);
2078 	return 0;
2079 
2080 out_byspi:
2081 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2082 out_bysrc:
2083 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2084 out_bydst:
2085 	return -ENOMEM;
2086 }
2087 
2088 void xfrm_state_fini(struct net *net)
2089 {
2090 	struct xfrm_audit audit_info;
2091 	unsigned int sz;
2092 
2093 	flush_work(&net->xfrm.state_hash_work);
2094 	audit_info.loginuid = -1;
2095 	audit_info.sessionid = -1;
2096 	audit_info.secid = 0;
2097 	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2098 	flush_work(&net->xfrm.state_gc_work);
2099 
2100 	WARN_ON(!list_empty(&net->xfrm.state_all));
2101 
2102 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2103 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2104 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2105 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2106 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2107 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2108 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2109 }
2110 
2111 #ifdef CONFIG_AUDITSYSCALL
2112 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2113 				     struct audit_buffer *audit_buf)
2114 {
2115 	struct xfrm_sec_ctx *ctx = x->security;
2116 	u32 spi = ntohl(x->id.spi);
2117 
2118 	if (ctx)
2119 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2120 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2121 
2122 	switch(x->props.family) {
2123 	case AF_INET:
2124 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2125 				 &x->props.saddr.a4, &x->id.daddr.a4);
2126 		break;
2127 	case AF_INET6:
2128 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2129 				 x->props.saddr.a6, x->id.daddr.a6);
2130 		break;
2131 	}
2132 
2133 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2134 }
2135 
2136 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2137 				      struct audit_buffer *audit_buf)
2138 {
2139 	struct iphdr *iph4;
2140 	struct ipv6hdr *iph6;
2141 
2142 	switch (family) {
2143 	case AF_INET:
2144 		iph4 = ip_hdr(skb);
2145 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2146 				 &iph4->saddr, &iph4->daddr);
2147 		break;
2148 	case AF_INET6:
2149 		iph6 = ipv6_hdr(skb);
2150 		audit_log_format(audit_buf,
2151 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2152 				 &iph6->saddr,&iph6->daddr,
2153 				 iph6->flow_lbl[0] & 0x0f,
2154 				 iph6->flow_lbl[1],
2155 				 iph6->flow_lbl[2]);
2156 		break;
2157 	}
2158 }
2159 
2160 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2161 			  uid_t auid, u32 sessionid, u32 secid)
2162 {
2163 	struct audit_buffer *audit_buf;
2164 
2165 	audit_buf = xfrm_audit_start("SAD-add");
2166 	if (audit_buf == NULL)
2167 		return;
2168 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2169 	xfrm_audit_helper_sainfo(x, audit_buf);
2170 	audit_log_format(audit_buf, " res=%u", result);
2171 	audit_log_end(audit_buf);
2172 }
2173 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2174 
2175 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2176 			     uid_t auid, u32 sessionid, u32 secid)
2177 {
2178 	struct audit_buffer *audit_buf;
2179 
2180 	audit_buf = xfrm_audit_start("SAD-delete");
2181 	if (audit_buf == NULL)
2182 		return;
2183 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2184 	xfrm_audit_helper_sainfo(x, audit_buf);
2185 	audit_log_format(audit_buf, " res=%u", result);
2186 	audit_log_end(audit_buf);
2187 }
2188 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2189 
2190 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2191 				      struct sk_buff *skb)
2192 {
2193 	struct audit_buffer *audit_buf;
2194 	u32 spi;
2195 
2196 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2197 	if (audit_buf == NULL)
2198 		return;
2199 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2200 	/* don't record the sequence number because it's inherent in this kind
2201 	 * of audit message */
2202 	spi = ntohl(x->id.spi);
2203 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2204 	audit_log_end(audit_buf);
2205 }
2206 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2207 
2208 static void xfrm_audit_state_replay(struct xfrm_state *x,
2209 			     struct sk_buff *skb, __be32 net_seq)
2210 {
2211 	struct audit_buffer *audit_buf;
2212 	u32 spi;
2213 
2214 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2215 	if (audit_buf == NULL)
2216 		return;
2217 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2218 	spi = ntohl(x->id.spi);
2219 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2220 			 spi, spi, ntohl(net_seq));
2221 	audit_log_end(audit_buf);
2222 }
2223 
2224 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2225 {
2226 	struct audit_buffer *audit_buf;
2227 
2228 	audit_buf = xfrm_audit_start("SA-notfound");
2229 	if (audit_buf == NULL)
2230 		return;
2231 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2232 	audit_log_end(audit_buf);
2233 }
2234 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2235 
2236 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2237 			       __be32 net_spi, __be32 net_seq)
2238 {
2239 	struct audit_buffer *audit_buf;
2240 	u32 spi;
2241 
2242 	audit_buf = xfrm_audit_start("SA-notfound");
2243 	if (audit_buf == NULL)
2244 		return;
2245 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2246 	spi = ntohl(net_spi);
2247 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2248 			 spi, spi, ntohl(net_seq));
2249 	audit_log_end(audit_buf);
2250 }
2251 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2252 
2253 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2254 			      struct sk_buff *skb, u8 proto)
2255 {
2256 	struct audit_buffer *audit_buf;
2257 	__be32 net_spi;
2258 	__be32 net_seq;
2259 
2260 	audit_buf = xfrm_audit_start("SA-icv-failure");
2261 	if (audit_buf == NULL)
2262 		return;
2263 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2264 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2265 		u32 spi = ntohl(net_spi);
2266 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2267 				 spi, spi, ntohl(net_seq));
2268 	}
2269 	audit_log_end(audit_buf);
2270 }
2271 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2272 #endif /* CONFIG_AUDITSYSCALL */
2273