xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 5f5bac82)
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 			continue;
673 
674 		switch (family) {
675 		case AF_INET:
676 			if (x->id.daddr.a4 != daddr->a4)
677 				continue;
678 			break;
679 		case AF_INET6:
680 			if (!ipv6_addr_equal((struct in6_addr *)daddr,
681 					     (struct in6_addr *)
682 					     x->id.daddr.a6))
683 				continue;
684 			break;
685 		}
686 
687 		xfrm_state_hold(x);
688 		return x;
689 	}
690 
691 	return NULL;
692 }
693 
694 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
695 {
696 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
697 	struct xfrm_state *x;
698 	struct hlist_node *entry;
699 
700 	hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
701 		if (x->props.family != family ||
702 		    x->id.proto     != proto)
703 			continue;
704 
705 		switch (family) {
706 		case AF_INET:
707 			if (x->id.daddr.a4 != daddr->a4 ||
708 			    x->props.saddr.a4 != saddr->a4)
709 				continue;
710 			break;
711 		case AF_INET6:
712 			if (!ipv6_addr_equal((struct in6_addr *)daddr,
713 					     (struct in6_addr *)
714 					     x->id.daddr.a6) ||
715 			    !ipv6_addr_equal((struct in6_addr *)saddr,
716 					     (struct in6_addr *)
717 					     x->props.saddr.a6))
718 				continue;
719 			break;
720 		}
721 
722 		xfrm_state_hold(x);
723 		return x;
724 	}
725 
726 	return NULL;
727 }
728 
729 static inline struct xfrm_state *
730 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
731 {
732 	struct net *net = xs_net(x);
733 
734 	if (use_spi)
735 		return __xfrm_state_lookup(net, &x->id.daddr, x->id.spi,
736 					   x->id.proto, family);
737 	else
738 		return __xfrm_state_lookup_byaddr(net, &x->id.daddr,
739 						  &x->props.saddr,
740 						  x->id.proto, family);
741 }
742 
743 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
744 {
745 	if (have_hash_collision &&
746 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
747 	    net->xfrm.state_num > net->xfrm.state_hmask)
748 		schedule_work(&net->xfrm.state_hash_work);
749 }
750 
751 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
752 			       struct flowi *fl, unsigned short family,
753 			       xfrm_address_t *daddr, xfrm_address_t *saddr,
754 			       struct xfrm_state **best, int *acq_in_progress,
755 			       int *error)
756 {
757 	/* Resolution logic:
758 	 * 1. There is a valid state with matching selector. Done.
759 	 * 2. Valid state with inappropriate selector. Skip.
760 	 *
761 	 * Entering area of "sysdeps".
762 	 *
763 	 * 3. If state is not valid, selector is temporary, it selects
764 	 *    only session which triggered previous resolution. Key
765 	 *    manager will do something to install a state with proper
766 	 *    selector.
767 	 */
768 	if (x->km.state == XFRM_STATE_VALID) {
769 		if ((x->sel.family &&
770 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
771 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
772 			return;
773 
774 		if (!*best ||
775 		    (*best)->km.dying > x->km.dying ||
776 		    ((*best)->km.dying == x->km.dying &&
777 		     (*best)->curlft.add_time < x->curlft.add_time))
778 			*best = x;
779 	} else if (x->km.state == XFRM_STATE_ACQ) {
780 		*acq_in_progress = 1;
781 	} else if (x->km.state == XFRM_STATE_ERROR ||
782 		   x->km.state == XFRM_STATE_EXPIRED) {
783 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
784 		    security_xfrm_state_pol_flow_match(x, pol, fl))
785 			*error = -ESRCH;
786 	}
787 }
788 
789 struct xfrm_state *
790 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
791 		struct flowi *fl, struct xfrm_tmpl *tmpl,
792 		struct xfrm_policy *pol, int *err,
793 		unsigned short family)
794 {
795 	static xfrm_address_t saddr_wildcard = { };
796 	struct net *net = xp_net(pol);
797 	unsigned int h, h_wildcard;
798 	struct hlist_node *entry;
799 	struct xfrm_state *x, *x0, *to_put;
800 	int acquire_in_progress = 0;
801 	int error = 0;
802 	struct xfrm_state *best = NULL;
803 
804 	to_put = NULL;
805 
806 	spin_lock_bh(&xfrm_state_lock);
807 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
808 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
809 		if (x->props.family == family &&
810 		    x->props.reqid == tmpl->reqid &&
811 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
812 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
813 		    tmpl->mode == x->props.mode &&
814 		    tmpl->id.proto == x->id.proto &&
815 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
816 			xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
817 					   &best, &acquire_in_progress, &error);
818 	}
819 	if (best)
820 		goto found;
821 
822 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
823 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
824 		if (x->props.family == family &&
825 		    x->props.reqid == tmpl->reqid &&
826 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
827 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
828 		    tmpl->mode == x->props.mode &&
829 		    tmpl->id.proto == x->id.proto &&
830 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
831 			xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
832 					   &best, &acquire_in_progress, &error);
833 	}
834 
835 found:
836 	x = best;
837 	if (!x && !error && !acquire_in_progress) {
838 		if (tmpl->id.spi &&
839 		    (x0 = __xfrm_state_lookup(net, daddr, tmpl->id.spi,
840 					      tmpl->id.proto, family)) != NULL) {
841 			to_put = x0;
842 			error = -EEXIST;
843 			goto out;
844 		}
845 		x = xfrm_state_alloc(net);
846 		if (x == NULL) {
847 			error = -ENOMEM;
848 			goto out;
849 		}
850 		/* Initialize temporary selector matching only
851 		 * to current session. */
852 		xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
853 
854 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
855 		if (error) {
856 			x->km.state = XFRM_STATE_DEAD;
857 			to_put = x;
858 			x = NULL;
859 			goto out;
860 		}
861 
862 		if (km_query(x, tmpl, pol) == 0) {
863 			x->km.state = XFRM_STATE_ACQ;
864 			list_add(&x->km.all, &net->xfrm.state_all);
865 			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
866 			h = xfrm_src_hash(net, daddr, saddr, family);
867 			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
868 			if (x->id.spi) {
869 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
870 				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
871 			}
872 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
873 			x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ;
874 			add_timer(&x->timer);
875 			net->xfrm.state_num++;
876 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
877 		} else {
878 			x->km.state = XFRM_STATE_DEAD;
879 			to_put = x;
880 			x = NULL;
881 			error = -ESRCH;
882 		}
883 	}
884 out:
885 	if (x)
886 		xfrm_state_hold(x);
887 	else
888 		*err = acquire_in_progress ? -EAGAIN : error;
889 	spin_unlock_bh(&xfrm_state_lock);
890 	if (to_put)
891 		xfrm_state_put(to_put);
892 	return x;
893 }
894 
895 struct xfrm_state *
896 xfrm_stateonly_find(struct net *net,
897 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
898 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
899 {
900 	unsigned int h;
901 	struct xfrm_state *rx = NULL, *x = NULL;
902 	struct hlist_node *entry;
903 
904 	spin_lock(&xfrm_state_lock);
905 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
906 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
907 		if (x->props.family == family &&
908 		    x->props.reqid == reqid &&
909 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
910 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
911 		    mode == x->props.mode &&
912 		    proto == x->id.proto &&
913 		    x->km.state == XFRM_STATE_VALID) {
914 			rx = x;
915 			break;
916 		}
917 	}
918 
919 	if (rx)
920 		xfrm_state_hold(rx);
921 	spin_unlock(&xfrm_state_lock);
922 
923 
924 	return rx;
925 }
926 EXPORT_SYMBOL(xfrm_stateonly_find);
927 
928 static void __xfrm_state_insert(struct xfrm_state *x)
929 {
930 	struct net *net = xs_net(x);
931 	unsigned int h;
932 
933 	x->genid = ++xfrm_state_genid;
934 
935 	list_add(&x->km.all, &net->xfrm.state_all);
936 
937 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
938 			  x->props.reqid, x->props.family);
939 	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
940 
941 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
942 	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
943 
944 	if (x->id.spi) {
945 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
946 				  x->props.family);
947 
948 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
949 	}
950 
951 	mod_timer(&x->timer, jiffies + HZ);
952 	if (x->replay_maxage)
953 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
954 
955 	wake_up(&net->xfrm.km_waitq);
956 
957 	net->xfrm.state_num++;
958 
959 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
960 }
961 
962 /* xfrm_state_lock is held */
963 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
964 {
965 	struct net *net = xs_net(xnew);
966 	unsigned short family = xnew->props.family;
967 	u32 reqid = xnew->props.reqid;
968 	struct xfrm_state *x;
969 	struct hlist_node *entry;
970 	unsigned int h;
971 
972 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
973 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
974 		if (x->props.family	== family &&
975 		    x->props.reqid	== reqid &&
976 		    !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
977 		    !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
978 			x->genid = xfrm_state_genid;
979 	}
980 }
981 
982 void xfrm_state_insert(struct xfrm_state *x)
983 {
984 	spin_lock_bh(&xfrm_state_lock);
985 	__xfrm_state_bump_genids(x);
986 	__xfrm_state_insert(x);
987 	spin_unlock_bh(&xfrm_state_lock);
988 }
989 EXPORT_SYMBOL(xfrm_state_insert);
990 
991 /* xfrm_state_lock is held */
992 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)
993 {
994 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
995 	struct hlist_node *entry;
996 	struct xfrm_state *x;
997 
998 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
999 		if (x->props.reqid  != reqid ||
1000 		    x->props.mode   != mode ||
1001 		    x->props.family != family ||
1002 		    x->km.state     != XFRM_STATE_ACQ ||
1003 		    x->id.spi       != 0 ||
1004 		    x->id.proto	    != proto)
1005 			continue;
1006 
1007 		switch (family) {
1008 		case AF_INET:
1009 			if (x->id.daddr.a4    != daddr->a4 ||
1010 			    x->props.saddr.a4 != saddr->a4)
1011 				continue;
1012 			break;
1013 		case AF_INET6:
1014 			if (!ipv6_addr_equal((struct in6_addr *)x->id.daddr.a6,
1015 					     (struct in6_addr *)daddr) ||
1016 			    !ipv6_addr_equal((struct in6_addr *)
1017 					     x->props.saddr.a6,
1018 					     (struct in6_addr *)saddr))
1019 				continue;
1020 			break;
1021 		}
1022 
1023 		xfrm_state_hold(x);
1024 		return x;
1025 	}
1026 
1027 	if (!create)
1028 		return NULL;
1029 
1030 	x = xfrm_state_alloc(net);
1031 	if (likely(x)) {
1032 		switch (family) {
1033 		case AF_INET:
1034 			x->sel.daddr.a4 = daddr->a4;
1035 			x->sel.saddr.a4 = saddr->a4;
1036 			x->sel.prefixlen_d = 32;
1037 			x->sel.prefixlen_s = 32;
1038 			x->props.saddr.a4 = saddr->a4;
1039 			x->id.daddr.a4 = daddr->a4;
1040 			break;
1041 
1042 		case AF_INET6:
1043 			ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1044 				       (struct in6_addr *)daddr);
1045 			ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1046 				       (struct in6_addr *)saddr);
1047 			x->sel.prefixlen_d = 128;
1048 			x->sel.prefixlen_s = 128;
1049 			ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1050 				       (struct in6_addr *)saddr);
1051 			ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1052 				       (struct in6_addr *)daddr);
1053 			break;
1054 		}
1055 
1056 		x->km.state = XFRM_STATE_ACQ;
1057 		x->id.proto = proto;
1058 		x->props.family = family;
1059 		x->props.mode = mode;
1060 		x->props.reqid = reqid;
1061 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1062 		xfrm_state_hold(x);
1063 		x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ;
1064 		add_timer(&x->timer);
1065 		list_add(&x->km.all, &net->xfrm.state_all);
1066 		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1067 		h = xfrm_src_hash(net, daddr, saddr, family);
1068 		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1069 
1070 		net->xfrm.state_num++;
1071 
1072 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1073 	}
1074 
1075 	return x;
1076 }
1077 
1078 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq);
1079 
1080 int xfrm_state_add(struct xfrm_state *x)
1081 {
1082 	struct net *net = xs_net(x);
1083 	struct xfrm_state *x1, *to_put;
1084 	int family;
1085 	int err;
1086 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1087 
1088 	family = x->props.family;
1089 
1090 	to_put = NULL;
1091 
1092 	spin_lock_bh(&xfrm_state_lock);
1093 
1094 	x1 = __xfrm_state_locate(x, use_spi, family);
1095 	if (x1) {
1096 		to_put = x1;
1097 		x1 = NULL;
1098 		err = -EEXIST;
1099 		goto out;
1100 	}
1101 
1102 	if (use_spi && x->km.seq) {
1103 		x1 = __xfrm_find_acq_byseq(net, x->km.seq);
1104 		if (x1 && ((x1->id.proto != x->id.proto) ||
1105 		    xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1106 			to_put = x1;
1107 			x1 = NULL;
1108 		}
1109 	}
1110 
1111 	if (use_spi && !x1)
1112 		x1 = __find_acq_core(net, family, x->props.mode, x->props.reqid,
1113 				     x->id.proto,
1114 				     &x->id.daddr, &x->props.saddr, 0);
1115 
1116 	__xfrm_state_bump_genids(x);
1117 	__xfrm_state_insert(x);
1118 	err = 0;
1119 
1120 out:
1121 	spin_unlock_bh(&xfrm_state_lock);
1122 
1123 	if (x1) {
1124 		xfrm_state_delete(x1);
1125 		xfrm_state_put(x1);
1126 	}
1127 
1128 	if (to_put)
1129 		xfrm_state_put(to_put);
1130 
1131 	return err;
1132 }
1133 EXPORT_SYMBOL(xfrm_state_add);
1134 
1135 #ifdef CONFIG_XFRM_MIGRATE
1136 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1137 {
1138 	struct net *net = xs_net(orig);
1139 	int err = -ENOMEM;
1140 	struct xfrm_state *x = xfrm_state_alloc(net);
1141 	if (!x)
1142 		goto error;
1143 
1144 	memcpy(&x->id, &orig->id, sizeof(x->id));
1145 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1146 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1147 	x->props.mode = orig->props.mode;
1148 	x->props.replay_window = orig->props.replay_window;
1149 	x->props.reqid = orig->props.reqid;
1150 	x->props.family = orig->props.family;
1151 	x->props.saddr = orig->props.saddr;
1152 
1153 	if (orig->aalg) {
1154 		x->aalg = xfrm_algo_clone(orig->aalg);
1155 		if (!x->aalg)
1156 			goto error;
1157 	}
1158 	x->props.aalgo = orig->props.aalgo;
1159 
1160 	if (orig->ealg) {
1161 		x->ealg = xfrm_algo_clone(orig->ealg);
1162 		if (!x->ealg)
1163 			goto error;
1164 	}
1165 	x->props.ealgo = orig->props.ealgo;
1166 
1167 	if (orig->calg) {
1168 		x->calg = xfrm_algo_clone(orig->calg);
1169 		if (!x->calg)
1170 			goto error;
1171 	}
1172 	x->props.calgo = orig->props.calgo;
1173 
1174 	if (orig->encap) {
1175 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1176 		if (!x->encap)
1177 			goto error;
1178 	}
1179 
1180 	if (orig->coaddr) {
1181 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1182 				    GFP_KERNEL);
1183 		if (!x->coaddr)
1184 			goto error;
1185 	}
1186 
1187 	err = xfrm_init_state(x);
1188 	if (err)
1189 		goto error;
1190 
1191 	x->props.flags = orig->props.flags;
1192 
1193 	x->curlft.add_time = orig->curlft.add_time;
1194 	x->km.state = orig->km.state;
1195 	x->km.seq = orig->km.seq;
1196 
1197 	return x;
1198 
1199  error:
1200 	if (errp)
1201 		*errp = err;
1202 	if (x) {
1203 		kfree(x->aalg);
1204 		kfree(x->ealg);
1205 		kfree(x->calg);
1206 		kfree(x->encap);
1207 		kfree(x->coaddr);
1208 	}
1209 	kfree(x);
1210 	return NULL;
1211 }
1212 
1213 /* xfrm_state_lock is held */
1214 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1215 {
1216 	unsigned int h;
1217 	struct xfrm_state *x;
1218 	struct hlist_node *entry;
1219 
1220 	if (m->reqid) {
1221 		h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1222 				  m->reqid, m->old_family);
1223 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1224 			if (x->props.mode != m->mode ||
1225 			    x->id.proto != m->proto)
1226 				continue;
1227 			if (m->reqid && x->props.reqid != m->reqid)
1228 				continue;
1229 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1230 					  m->old_family) ||
1231 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1232 					  m->old_family))
1233 				continue;
1234 			xfrm_state_hold(x);
1235 			return x;
1236 		}
1237 	} else {
1238 		h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1239 				  m->old_family);
1240 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1241 			if (x->props.mode != m->mode ||
1242 			    x->id.proto != m->proto)
1243 				continue;
1244 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1245 					  m->old_family) ||
1246 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1247 					  m->old_family))
1248 				continue;
1249 			xfrm_state_hold(x);
1250 			return x;
1251 		}
1252 	}
1253 
1254 	return NULL;
1255 }
1256 EXPORT_SYMBOL(xfrm_migrate_state_find);
1257 
1258 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1259 				       struct xfrm_migrate *m)
1260 {
1261 	struct xfrm_state *xc;
1262 	int err;
1263 
1264 	xc = xfrm_state_clone(x, &err);
1265 	if (!xc)
1266 		return NULL;
1267 
1268 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1269 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1270 
1271 	/* add state */
1272 	if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1273 		/* a care is needed when the destination address of the
1274 		   state is to be updated as it is a part of triplet */
1275 		xfrm_state_insert(xc);
1276 	} else {
1277 		if ((err = xfrm_state_add(xc)) < 0)
1278 			goto error;
1279 	}
1280 
1281 	return xc;
1282 error:
1283 	kfree(xc);
1284 	return NULL;
1285 }
1286 EXPORT_SYMBOL(xfrm_state_migrate);
1287 #endif
1288 
1289 int xfrm_state_update(struct xfrm_state *x)
1290 {
1291 	struct xfrm_state *x1, *to_put;
1292 	int err;
1293 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1294 
1295 	to_put = NULL;
1296 
1297 	spin_lock_bh(&xfrm_state_lock);
1298 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1299 
1300 	err = -ESRCH;
1301 	if (!x1)
1302 		goto out;
1303 
1304 	if (xfrm_state_kern(x1)) {
1305 		to_put = x1;
1306 		err = -EEXIST;
1307 		goto out;
1308 	}
1309 
1310 	if (x1->km.state == XFRM_STATE_ACQ) {
1311 		__xfrm_state_insert(x);
1312 		x = NULL;
1313 	}
1314 	err = 0;
1315 
1316 out:
1317 	spin_unlock_bh(&xfrm_state_lock);
1318 
1319 	if (to_put)
1320 		xfrm_state_put(to_put);
1321 
1322 	if (err)
1323 		return err;
1324 
1325 	if (!x) {
1326 		xfrm_state_delete(x1);
1327 		xfrm_state_put(x1);
1328 		return 0;
1329 	}
1330 
1331 	err = -EINVAL;
1332 	spin_lock_bh(&x1->lock);
1333 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1334 		if (x->encap && x1->encap)
1335 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1336 		if (x->coaddr && x1->coaddr) {
1337 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1338 		}
1339 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1340 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1341 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1342 		x1->km.dying = 0;
1343 
1344 		mod_timer(&x1->timer, jiffies + HZ);
1345 		if (x1->curlft.use_time)
1346 			xfrm_state_check_expire(x1);
1347 
1348 		err = 0;
1349 	}
1350 	spin_unlock_bh(&x1->lock);
1351 
1352 	xfrm_state_put(x1);
1353 
1354 	return err;
1355 }
1356 EXPORT_SYMBOL(xfrm_state_update);
1357 
1358 int xfrm_state_check_expire(struct xfrm_state *x)
1359 {
1360 	if (!x->curlft.use_time)
1361 		x->curlft.use_time = get_seconds();
1362 
1363 	if (x->km.state != XFRM_STATE_VALID)
1364 		return -EINVAL;
1365 
1366 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1367 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1368 		x->km.state = XFRM_STATE_EXPIRED;
1369 		mod_timer(&x->timer, jiffies);
1370 		return -EINVAL;
1371 	}
1372 
1373 	if (!x->km.dying &&
1374 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1375 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1376 		x->km.dying = 1;
1377 		km_state_expired(x, 0, 0);
1378 	}
1379 	return 0;
1380 }
1381 EXPORT_SYMBOL(xfrm_state_check_expire);
1382 
1383 struct xfrm_state *
1384 xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto,
1385 		  unsigned short family)
1386 {
1387 	struct xfrm_state *x;
1388 
1389 	spin_lock_bh(&xfrm_state_lock);
1390 	x = __xfrm_state_lookup(net, daddr, spi, proto, family);
1391 	spin_unlock_bh(&xfrm_state_lock);
1392 	return x;
1393 }
1394 EXPORT_SYMBOL(xfrm_state_lookup);
1395 
1396 struct xfrm_state *
1397 xfrm_state_lookup_byaddr(struct net *net,
1398 			 xfrm_address_t *daddr, xfrm_address_t *saddr,
1399 			 u8 proto, unsigned short family)
1400 {
1401 	struct xfrm_state *x;
1402 
1403 	spin_lock_bh(&xfrm_state_lock);
1404 	x = __xfrm_state_lookup_byaddr(net, daddr, saddr, proto, family);
1405 	spin_unlock_bh(&xfrm_state_lock);
1406 	return x;
1407 }
1408 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1409 
1410 struct xfrm_state *
1411 xfrm_find_acq(struct net *net, u8 mode, u32 reqid, u8 proto,
1412 	      xfrm_address_t *daddr, xfrm_address_t *saddr,
1413 	      int create, unsigned short family)
1414 {
1415 	struct xfrm_state *x;
1416 
1417 	spin_lock_bh(&xfrm_state_lock);
1418 	x = __find_acq_core(net, family, mode, reqid, proto, daddr, saddr, create);
1419 	spin_unlock_bh(&xfrm_state_lock);
1420 
1421 	return x;
1422 }
1423 EXPORT_SYMBOL(xfrm_find_acq);
1424 
1425 #ifdef CONFIG_XFRM_SUB_POLICY
1426 int
1427 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1428 	       unsigned short family)
1429 {
1430 	int err = 0;
1431 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1432 	if (!afinfo)
1433 		return -EAFNOSUPPORT;
1434 
1435 	spin_lock_bh(&xfrm_state_lock);
1436 	if (afinfo->tmpl_sort)
1437 		err = afinfo->tmpl_sort(dst, src, n);
1438 	spin_unlock_bh(&xfrm_state_lock);
1439 	xfrm_state_put_afinfo(afinfo);
1440 	return err;
1441 }
1442 EXPORT_SYMBOL(xfrm_tmpl_sort);
1443 
1444 int
1445 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1446 		unsigned short family)
1447 {
1448 	int err = 0;
1449 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1450 	if (!afinfo)
1451 		return -EAFNOSUPPORT;
1452 
1453 	spin_lock_bh(&xfrm_state_lock);
1454 	if (afinfo->state_sort)
1455 		err = afinfo->state_sort(dst, src, n);
1456 	spin_unlock_bh(&xfrm_state_lock);
1457 	xfrm_state_put_afinfo(afinfo);
1458 	return err;
1459 }
1460 EXPORT_SYMBOL(xfrm_state_sort);
1461 #endif
1462 
1463 /* Silly enough, but I'm lazy to build resolution list */
1464 
1465 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq)
1466 {
1467 	int i;
1468 
1469 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1470 		struct hlist_node *entry;
1471 		struct xfrm_state *x;
1472 
1473 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1474 			if (x->km.seq == seq &&
1475 			    x->km.state == XFRM_STATE_ACQ) {
1476 				xfrm_state_hold(x);
1477 				return x;
1478 			}
1479 		}
1480 	}
1481 	return NULL;
1482 }
1483 
1484 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 seq)
1485 {
1486 	struct xfrm_state *x;
1487 
1488 	spin_lock_bh(&xfrm_state_lock);
1489 	x = __xfrm_find_acq_byseq(net, seq);
1490 	spin_unlock_bh(&xfrm_state_lock);
1491 	return x;
1492 }
1493 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1494 
1495 u32 xfrm_get_acqseq(void)
1496 {
1497 	u32 res;
1498 	static u32 acqseq;
1499 	static DEFINE_SPINLOCK(acqseq_lock);
1500 
1501 	spin_lock_bh(&acqseq_lock);
1502 	res = (++acqseq ? : ++acqseq);
1503 	spin_unlock_bh(&acqseq_lock);
1504 	return res;
1505 }
1506 EXPORT_SYMBOL(xfrm_get_acqseq);
1507 
1508 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1509 {
1510 	struct net *net = xs_net(x);
1511 	unsigned int h;
1512 	struct xfrm_state *x0;
1513 	int err = -ENOENT;
1514 	__be32 minspi = htonl(low);
1515 	__be32 maxspi = htonl(high);
1516 
1517 	spin_lock_bh(&x->lock);
1518 	if (x->km.state == XFRM_STATE_DEAD)
1519 		goto unlock;
1520 
1521 	err = 0;
1522 	if (x->id.spi)
1523 		goto unlock;
1524 
1525 	err = -ENOENT;
1526 
1527 	if (minspi == maxspi) {
1528 		x0 = xfrm_state_lookup(net, &x->id.daddr, minspi, x->id.proto, x->props.family);
1529 		if (x0) {
1530 			xfrm_state_put(x0);
1531 			goto unlock;
1532 		}
1533 		x->id.spi = minspi;
1534 	} else {
1535 		u32 spi = 0;
1536 		for (h=0; h<high-low+1; h++) {
1537 			spi = low + net_random()%(high-low+1);
1538 			x0 = xfrm_state_lookup(net, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1539 			if (x0 == NULL) {
1540 				x->id.spi = htonl(spi);
1541 				break;
1542 			}
1543 			xfrm_state_put(x0);
1544 		}
1545 	}
1546 	if (x->id.spi) {
1547 		spin_lock_bh(&xfrm_state_lock);
1548 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1549 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1550 		spin_unlock_bh(&xfrm_state_lock);
1551 
1552 		err = 0;
1553 	}
1554 
1555 unlock:
1556 	spin_unlock_bh(&x->lock);
1557 
1558 	return err;
1559 }
1560 EXPORT_SYMBOL(xfrm_alloc_spi);
1561 
1562 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1563 		    int (*func)(struct xfrm_state *, int, void*),
1564 		    void *data)
1565 {
1566 	struct xfrm_state *state;
1567 	struct xfrm_state_walk *x;
1568 	int err = 0;
1569 
1570 	if (walk->seq != 0 && list_empty(&walk->all))
1571 		return 0;
1572 
1573 	spin_lock_bh(&xfrm_state_lock);
1574 	if (list_empty(&walk->all))
1575 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1576 	else
1577 		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1578 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1579 		if (x->state == XFRM_STATE_DEAD)
1580 			continue;
1581 		state = container_of(x, struct xfrm_state, km);
1582 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1583 			continue;
1584 		err = func(state, walk->seq, data);
1585 		if (err) {
1586 			list_move_tail(&walk->all, &x->all);
1587 			goto out;
1588 		}
1589 		walk->seq++;
1590 	}
1591 	if (walk->seq == 0) {
1592 		err = -ENOENT;
1593 		goto out;
1594 	}
1595 	list_del_init(&walk->all);
1596 out:
1597 	spin_unlock_bh(&xfrm_state_lock);
1598 	return err;
1599 }
1600 EXPORT_SYMBOL(xfrm_state_walk);
1601 
1602 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1603 {
1604 	INIT_LIST_HEAD(&walk->all);
1605 	walk->proto = proto;
1606 	walk->state = XFRM_STATE_DEAD;
1607 	walk->seq = 0;
1608 }
1609 EXPORT_SYMBOL(xfrm_state_walk_init);
1610 
1611 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1612 {
1613 	if (list_empty(&walk->all))
1614 		return;
1615 
1616 	spin_lock_bh(&xfrm_state_lock);
1617 	list_del(&walk->all);
1618 	spin_unlock_bh(&xfrm_state_lock);
1619 }
1620 EXPORT_SYMBOL(xfrm_state_walk_done);
1621 
1622 
1623 void xfrm_replay_notify(struct xfrm_state *x, int event)
1624 {
1625 	struct km_event c;
1626 	/* we send notify messages in case
1627 	 *  1. we updated on of the sequence numbers, and the seqno difference
1628 	 *     is at least x->replay_maxdiff, in this case we also update the
1629 	 *     timeout of our timer function
1630 	 *  2. if x->replay_maxage has elapsed since last update,
1631 	 *     and there were changes
1632 	 *
1633 	 *  The state structure must be locked!
1634 	 */
1635 
1636 	switch (event) {
1637 	case XFRM_REPLAY_UPDATE:
1638 		if (x->replay_maxdiff &&
1639 		    (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1640 		    (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1641 			if (x->xflags & XFRM_TIME_DEFER)
1642 				event = XFRM_REPLAY_TIMEOUT;
1643 			else
1644 				return;
1645 		}
1646 
1647 		break;
1648 
1649 	case XFRM_REPLAY_TIMEOUT:
1650 		if ((x->replay.seq == x->preplay.seq) &&
1651 		    (x->replay.bitmap == x->preplay.bitmap) &&
1652 		    (x->replay.oseq == x->preplay.oseq)) {
1653 			x->xflags |= XFRM_TIME_DEFER;
1654 			return;
1655 		}
1656 
1657 		break;
1658 	}
1659 
1660 	memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1661 	c.event = XFRM_MSG_NEWAE;
1662 	c.data.aevent = event;
1663 	km_state_notify(x, &c);
1664 
1665 	if (x->replay_maxage &&
1666 	    !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1667 		x->xflags &= ~XFRM_TIME_DEFER;
1668 }
1669 
1670 static void xfrm_replay_timer_handler(unsigned long data)
1671 {
1672 	struct xfrm_state *x = (struct xfrm_state*)data;
1673 
1674 	spin_lock(&x->lock);
1675 
1676 	if (x->km.state == XFRM_STATE_VALID) {
1677 		if (xfrm_aevent_is_on(xs_net(x)))
1678 			xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1679 		else
1680 			x->xflags |= XFRM_TIME_DEFER;
1681 	}
1682 
1683 	spin_unlock(&x->lock);
1684 }
1685 
1686 int xfrm_replay_check(struct xfrm_state *x,
1687 		      struct sk_buff *skb, __be32 net_seq)
1688 {
1689 	u32 diff;
1690 	u32 seq = ntohl(net_seq);
1691 
1692 	if (unlikely(seq == 0))
1693 		goto err;
1694 
1695 	if (likely(seq > x->replay.seq))
1696 		return 0;
1697 
1698 	diff = x->replay.seq - seq;
1699 	if (diff >= min_t(unsigned int, x->props.replay_window,
1700 			  sizeof(x->replay.bitmap) * 8)) {
1701 		x->stats.replay_window++;
1702 		goto err;
1703 	}
1704 
1705 	if (x->replay.bitmap & (1U << diff)) {
1706 		x->stats.replay++;
1707 		goto err;
1708 	}
1709 	return 0;
1710 
1711 err:
1712 	xfrm_audit_state_replay(x, skb, net_seq);
1713 	return -EINVAL;
1714 }
1715 
1716 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1717 {
1718 	u32 diff;
1719 	u32 seq = ntohl(net_seq);
1720 
1721 	if (seq > x->replay.seq) {
1722 		diff = seq - x->replay.seq;
1723 		if (diff < x->props.replay_window)
1724 			x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1725 		else
1726 			x->replay.bitmap = 1;
1727 		x->replay.seq = seq;
1728 	} else {
1729 		diff = x->replay.seq - seq;
1730 		x->replay.bitmap |= (1U << diff);
1731 	}
1732 
1733 	if (xfrm_aevent_is_on(xs_net(x)))
1734 		xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1735 }
1736 
1737 static LIST_HEAD(xfrm_km_list);
1738 static DEFINE_RWLOCK(xfrm_km_lock);
1739 
1740 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1741 {
1742 	struct xfrm_mgr *km;
1743 
1744 	read_lock(&xfrm_km_lock);
1745 	list_for_each_entry(km, &xfrm_km_list, list)
1746 		if (km->notify_policy)
1747 			km->notify_policy(xp, dir, c);
1748 	read_unlock(&xfrm_km_lock);
1749 }
1750 
1751 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1752 {
1753 	struct xfrm_mgr *km;
1754 	read_lock(&xfrm_km_lock);
1755 	list_for_each_entry(km, &xfrm_km_list, list)
1756 		if (km->notify)
1757 			km->notify(x, c);
1758 	read_unlock(&xfrm_km_lock);
1759 }
1760 
1761 EXPORT_SYMBOL(km_policy_notify);
1762 EXPORT_SYMBOL(km_state_notify);
1763 
1764 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1765 {
1766 	struct net *net = xs_net(x);
1767 	struct km_event c;
1768 
1769 	c.data.hard = hard;
1770 	c.pid = pid;
1771 	c.event = XFRM_MSG_EXPIRE;
1772 	km_state_notify(x, &c);
1773 
1774 	if (hard)
1775 		wake_up(&net->xfrm.km_waitq);
1776 }
1777 
1778 EXPORT_SYMBOL(km_state_expired);
1779 /*
1780  * We send to all registered managers regardless of failure
1781  * We are happy with one success
1782 */
1783 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1784 {
1785 	int err = -EINVAL, acqret;
1786 	struct xfrm_mgr *km;
1787 
1788 	read_lock(&xfrm_km_lock);
1789 	list_for_each_entry(km, &xfrm_km_list, list) {
1790 		acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1791 		if (!acqret)
1792 			err = acqret;
1793 	}
1794 	read_unlock(&xfrm_km_lock);
1795 	return err;
1796 }
1797 EXPORT_SYMBOL(km_query);
1798 
1799 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1800 {
1801 	int err = -EINVAL;
1802 	struct xfrm_mgr *km;
1803 
1804 	read_lock(&xfrm_km_lock);
1805 	list_for_each_entry(km, &xfrm_km_list, list) {
1806 		if (km->new_mapping)
1807 			err = km->new_mapping(x, ipaddr, sport);
1808 		if (!err)
1809 			break;
1810 	}
1811 	read_unlock(&xfrm_km_lock);
1812 	return err;
1813 }
1814 EXPORT_SYMBOL(km_new_mapping);
1815 
1816 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1817 {
1818 	struct net *net = xp_net(pol);
1819 	struct km_event c;
1820 
1821 	c.data.hard = hard;
1822 	c.pid = pid;
1823 	c.event = XFRM_MSG_POLEXPIRE;
1824 	km_policy_notify(pol, dir, &c);
1825 
1826 	if (hard)
1827 		wake_up(&net->xfrm.km_waitq);
1828 }
1829 EXPORT_SYMBOL(km_policy_expired);
1830 
1831 #ifdef CONFIG_XFRM_MIGRATE
1832 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1833 	       struct xfrm_migrate *m, int num_migrate,
1834 	       struct xfrm_kmaddress *k)
1835 {
1836 	int err = -EINVAL;
1837 	int ret;
1838 	struct xfrm_mgr *km;
1839 
1840 	read_lock(&xfrm_km_lock);
1841 	list_for_each_entry(km, &xfrm_km_list, list) {
1842 		if (km->migrate) {
1843 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1844 			if (!ret)
1845 				err = ret;
1846 		}
1847 	}
1848 	read_unlock(&xfrm_km_lock);
1849 	return err;
1850 }
1851 EXPORT_SYMBOL(km_migrate);
1852 #endif
1853 
1854 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1855 {
1856 	int err = -EINVAL;
1857 	int ret;
1858 	struct xfrm_mgr *km;
1859 
1860 	read_lock(&xfrm_km_lock);
1861 	list_for_each_entry(km, &xfrm_km_list, list) {
1862 		if (km->report) {
1863 			ret = km->report(net, proto, sel, addr);
1864 			if (!ret)
1865 				err = ret;
1866 		}
1867 	}
1868 	read_unlock(&xfrm_km_lock);
1869 	return err;
1870 }
1871 EXPORT_SYMBOL(km_report);
1872 
1873 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1874 {
1875 	int err;
1876 	u8 *data;
1877 	struct xfrm_mgr *km;
1878 	struct xfrm_policy *pol = NULL;
1879 
1880 	if (optlen <= 0 || optlen > PAGE_SIZE)
1881 		return -EMSGSIZE;
1882 
1883 	data = kmalloc(optlen, GFP_KERNEL);
1884 	if (!data)
1885 		return -ENOMEM;
1886 
1887 	err = -EFAULT;
1888 	if (copy_from_user(data, optval, optlen))
1889 		goto out;
1890 
1891 	err = -EINVAL;
1892 	read_lock(&xfrm_km_lock);
1893 	list_for_each_entry(km, &xfrm_km_list, list) {
1894 		pol = km->compile_policy(sk, optname, data,
1895 					 optlen, &err);
1896 		if (err >= 0)
1897 			break;
1898 	}
1899 	read_unlock(&xfrm_km_lock);
1900 
1901 	if (err >= 0) {
1902 		xfrm_sk_policy_insert(sk, err, pol);
1903 		xfrm_pol_put(pol);
1904 		err = 0;
1905 	}
1906 
1907 out:
1908 	kfree(data);
1909 	return err;
1910 }
1911 EXPORT_SYMBOL(xfrm_user_policy);
1912 
1913 int xfrm_register_km(struct xfrm_mgr *km)
1914 {
1915 	write_lock_bh(&xfrm_km_lock);
1916 	list_add_tail(&km->list, &xfrm_km_list);
1917 	write_unlock_bh(&xfrm_km_lock);
1918 	return 0;
1919 }
1920 EXPORT_SYMBOL(xfrm_register_km);
1921 
1922 int xfrm_unregister_km(struct xfrm_mgr *km)
1923 {
1924 	write_lock_bh(&xfrm_km_lock);
1925 	list_del(&km->list);
1926 	write_unlock_bh(&xfrm_km_lock);
1927 	return 0;
1928 }
1929 EXPORT_SYMBOL(xfrm_unregister_km);
1930 
1931 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1932 {
1933 	int err = 0;
1934 	if (unlikely(afinfo == NULL))
1935 		return -EINVAL;
1936 	if (unlikely(afinfo->family >= NPROTO))
1937 		return -EAFNOSUPPORT;
1938 	write_lock_bh(&xfrm_state_afinfo_lock);
1939 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1940 		err = -ENOBUFS;
1941 	else
1942 		xfrm_state_afinfo[afinfo->family] = afinfo;
1943 	write_unlock_bh(&xfrm_state_afinfo_lock);
1944 	return err;
1945 }
1946 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1947 
1948 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1949 {
1950 	int err = 0;
1951 	if (unlikely(afinfo == NULL))
1952 		return -EINVAL;
1953 	if (unlikely(afinfo->family >= NPROTO))
1954 		return -EAFNOSUPPORT;
1955 	write_lock_bh(&xfrm_state_afinfo_lock);
1956 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1957 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1958 			err = -EINVAL;
1959 		else
1960 			xfrm_state_afinfo[afinfo->family] = NULL;
1961 	}
1962 	write_unlock_bh(&xfrm_state_afinfo_lock);
1963 	return err;
1964 }
1965 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1966 
1967 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1968 {
1969 	struct xfrm_state_afinfo *afinfo;
1970 	if (unlikely(family >= NPROTO))
1971 		return NULL;
1972 	read_lock(&xfrm_state_afinfo_lock);
1973 	afinfo = xfrm_state_afinfo[family];
1974 	if (unlikely(!afinfo))
1975 		read_unlock(&xfrm_state_afinfo_lock);
1976 	return afinfo;
1977 }
1978 
1979 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1980 	__releases(xfrm_state_afinfo_lock)
1981 {
1982 	read_unlock(&xfrm_state_afinfo_lock);
1983 }
1984 
1985 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1986 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1987 {
1988 	if (x->tunnel) {
1989 		struct xfrm_state *t = x->tunnel;
1990 
1991 		if (atomic_read(&t->tunnel_users) == 2)
1992 			xfrm_state_delete(t);
1993 		atomic_dec(&t->tunnel_users);
1994 		xfrm_state_put(t);
1995 		x->tunnel = NULL;
1996 	}
1997 }
1998 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1999 
2000 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2001 {
2002 	int res;
2003 
2004 	spin_lock_bh(&x->lock);
2005 	if (x->km.state == XFRM_STATE_VALID &&
2006 	    x->type && x->type->get_mtu)
2007 		res = x->type->get_mtu(x, mtu);
2008 	else
2009 		res = mtu - x->props.header_len;
2010 	spin_unlock_bh(&x->lock);
2011 	return res;
2012 }
2013 
2014 int xfrm_init_state(struct xfrm_state *x)
2015 {
2016 	struct xfrm_state_afinfo *afinfo;
2017 	struct xfrm_mode *inner_mode;
2018 	int family = x->props.family;
2019 	int err;
2020 
2021 	err = -EAFNOSUPPORT;
2022 	afinfo = xfrm_state_get_afinfo(family);
2023 	if (!afinfo)
2024 		goto error;
2025 
2026 	err = 0;
2027 	if (afinfo->init_flags)
2028 		err = afinfo->init_flags(x);
2029 
2030 	xfrm_state_put_afinfo(afinfo);
2031 
2032 	if (err)
2033 		goto error;
2034 
2035 	err = -EPROTONOSUPPORT;
2036 
2037 	if (x->sel.family != AF_UNSPEC) {
2038 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2039 		if (inner_mode == NULL)
2040 			goto error;
2041 
2042 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2043 		    family != x->sel.family) {
2044 			xfrm_put_mode(inner_mode);
2045 			goto error;
2046 		}
2047 
2048 		x->inner_mode = inner_mode;
2049 	} else {
2050 		struct xfrm_mode *inner_mode_iaf;
2051 		int iafamily = AF_INET;
2052 
2053 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2054 		if (inner_mode == NULL)
2055 			goto error;
2056 
2057 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2058 			xfrm_put_mode(inner_mode);
2059 			goto error;
2060 		}
2061 		x->inner_mode = inner_mode;
2062 
2063 		if (x->props.family == AF_INET)
2064 			iafamily = AF_INET6;
2065 
2066 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2067 		if (inner_mode_iaf) {
2068 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2069 				x->inner_mode_iaf = inner_mode_iaf;
2070 			else
2071 				xfrm_put_mode(inner_mode_iaf);
2072 		}
2073 	}
2074 
2075 	x->type = xfrm_get_type(x->id.proto, family);
2076 	if (x->type == NULL)
2077 		goto error;
2078 
2079 	err = x->type->init_state(x);
2080 	if (err)
2081 		goto error;
2082 
2083 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2084 	if (x->outer_mode == NULL)
2085 		goto error;
2086 
2087 	x->km.state = XFRM_STATE_VALID;
2088 
2089 error:
2090 	return err;
2091 }
2092 
2093 EXPORT_SYMBOL(xfrm_init_state);
2094 
2095 int __net_init xfrm_state_init(struct net *net)
2096 {
2097 	unsigned int sz;
2098 
2099 	INIT_LIST_HEAD(&net->xfrm.state_all);
2100 
2101 	sz = sizeof(struct hlist_head) * 8;
2102 
2103 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2104 	if (!net->xfrm.state_bydst)
2105 		goto out_bydst;
2106 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2107 	if (!net->xfrm.state_bysrc)
2108 		goto out_bysrc;
2109 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2110 	if (!net->xfrm.state_byspi)
2111 		goto out_byspi;
2112 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2113 
2114 	net->xfrm.state_num = 0;
2115 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2116 	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2117 	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2118 	init_waitqueue_head(&net->xfrm.km_waitq);
2119 	return 0;
2120 
2121 out_byspi:
2122 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2123 out_bysrc:
2124 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2125 out_bydst:
2126 	return -ENOMEM;
2127 }
2128 
2129 void xfrm_state_fini(struct net *net)
2130 {
2131 	struct xfrm_audit audit_info;
2132 	unsigned int sz;
2133 
2134 	flush_work(&net->xfrm.state_hash_work);
2135 	audit_info.loginuid = -1;
2136 	audit_info.sessionid = -1;
2137 	audit_info.secid = 0;
2138 	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2139 	flush_work(&net->xfrm.state_gc_work);
2140 
2141 	WARN_ON(!list_empty(&net->xfrm.state_all));
2142 
2143 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2144 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2145 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2146 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2147 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2148 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2149 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2150 }
2151 
2152 #ifdef CONFIG_AUDITSYSCALL
2153 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2154 				     struct audit_buffer *audit_buf)
2155 {
2156 	struct xfrm_sec_ctx *ctx = x->security;
2157 	u32 spi = ntohl(x->id.spi);
2158 
2159 	if (ctx)
2160 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2161 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2162 
2163 	switch(x->props.family) {
2164 	case AF_INET:
2165 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2166 				 &x->props.saddr.a4, &x->id.daddr.a4);
2167 		break;
2168 	case AF_INET6:
2169 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2170 				 x->props.saddr.a6, x->id.daddr.a6);
2171 		break;
2172 	}
2173 
2174 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2175 }
2176 
2177 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2178 				      struct audit_buffer *audit_buf)
2179 {
2180 	struct iphdr *iph4;
2181 	struct ipv6hdr *iph6;
2182 
2183 	switch (family) {
2184 	case AF_INET:
2185 		iph4 = ip_hdr(skb);
2186 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2187 				 &iph4->saddr, &iph4->daddr);
2188 		break;
2189 	case AF_INET6:
2190 		iph6 = ipv6_hdr(skb);
2191 		audit_log_format(audit_buf,
2192 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2193 				 &iph6->saddr,&iph6->daddr,
2194 				 iph6->flow_lbl[0] & 0x0f,
2195 				 iph6->flow_lbl[1],
2196 				 iph6->flow_lbl[2]);
2197 		break;
2198 	}
2199 }
2200 
2201 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2202 			  uid_t auid, u32 sessionid, u32 secid)
2203 {
2204 	struct audit_buffer *audit_buf;
2205 
2206 	audit_buf = xfrm_audit_start("SAD-add");
2207 	if (audit_buf == NULL)
2208 		return;
2209 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2210 	xfrm_audit_helper_sainfo(x, audit_buf);
2211 	audit_log_format(audit_buf, " res=%u", result);
2212 	audit_log_end(audit_buf);
2213 }
2214 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2215 
2216 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2217 			     uid_t auid, u32 sessionid, u32 secid)
2218 {
2219 	struct audit_buffer *audit_buf;
2220 
2221 	audit_buf = xfrm_audit_start("SAD-delete");
2222 	if (audit_buf == NULL)
2223 		return;
2224 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2225 	xfrm_audit_helper_sainfo(x, audit_buf);
2226 	audit_log_format(audit_buf, " res=%u", result);
2227 	audit_log_end(audit_buf);
2228 }
2229 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2230 
2231 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2232 				      struct sk_buff *skb)
2233 {
2234 	struct audit_buffer *audit_buf;
2235 	u32 spi;
2236 
2237 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2238 	if (audit_buf == NULL)
2239 		return;
2240 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2241 	/* don't record the sequence number because it's inherent in this kind
2242 	 * of audit message */
2243 	spi = ntohl(x->id.spi);
2244 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2245 	audit_log_end(audit_buf);
2246 }
2247 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2248 
2249 static void xfrm_audit_state_replay(struct xfrm_state *x,
2250 			     struct sk_buff *skb, __be32 net_seq)
2251 {
2252 	struct audit_buffer *audit_buf;
2253 	u32 spi;
2254 
2255 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2256 	if (audit_buf == NULL)
2257 		return;
2258 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2259 	spi = ntohl(x->id.spi);
2260 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2261 			 spi, spi, ntohl(net_seq));
2262 	audit_log_end(audit_buf);
2263 }
2264 
2265 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2266 {
2267 	struct audit_buffer *audit_buf;
2268 
2269 	audit_buf = xfrm_audit_start("SA-notfound");
2270 	if (audit_buf == NULL)
2271 		return;
2272 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2273 	audit_log_end(audit_buf);
2274 }
2275 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2276 
2277 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2278 			       __be32 net_spi, __be32 net_seq)
2279 {
2280 	struct audit_buffer *audit_buf;
2281 	u32 spi;
2282 
2283 	audit_buf = xfrm_audit_start("SA-notfound");
2284 	if (audit_buf == NULL)
2285 		return;
2286 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2287 	spi = ntohl(net_spi);
2288 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2289 			 spi, spi, ntohl(net_seq));
2290 	audit_log_end(audit_buf);
2291 }
2292 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2293 
2294 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2295 			      struct sk_buff *skb, u8 proto)
2296 {
2297 	struct audit_buffer *audit_buf;
2298 	__be32 net_spi;
2299 	__be32 net_seq;
2300 
2301 	audit_buf = xfrm_audit_start("SA-icv-failure");
2302 	if (audit_buf == NULL)
2303 		return;
2304 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2305 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2306 		u32 spi = ntohl(net_spi);
2307 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2308 				 spi, spi, ntohl(net_seq));
2309 	}
2310 	audit_log_end(audit_buf);
2311 }
2312 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2313 #endif /* CONFIG_AUDITSYSCALL */
2314