xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 6dfcd296)
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 #include <linux/ktime.h>
25 #include <linux/slab.h>
26 #include <linux/interrupt.h>
27 #include <linux/kernel.h>
28 
29 #include "xfrm_hash.h"
30 
31 #define xfrm_state_deref_prot(table, net) \
32 	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
33 
34 static void xfrm_state_gc_task(struct work_struct *work);
35 
36 /* Each xfrm_state may be linked to two tables:
37 
38    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
39    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
40       destination/tunnel endpoint. (output)
41  */
42 
43 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
44 static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
45 
46 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
47 static HLIST_HEAD(xfrm_state_gc_list);
48 
49 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
50 {
51 	return atomic_inc_not_zero(&x->refcnt);
52 }
53 
54 static inline unsigned int xfrm_dst_hash(struct net *net,
55 					 const xfrm_address_t *daddr,
56 					 const xfrm_address_t *saddr,
57 					 u32 reqid,
58 					 unsigned short family)
59 {
60 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
61 }
62 
63 static inline unsigned int xfrm_src_hash(struct net *net,
64 					 const xfrm_address_t *daddr,
65 					 const xfrm_address_t *saddr,
66 					 unsigned short family)
67 {
68 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
69 }
70 
71 static inline unsigned int
72 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
73 	      __be32 spi, u8 proto, unsigned short family)
74 {
75 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
76 }
77 
78 static void xfrm_hash_transfer(struct hlist_head *list,
79 			       struct hlist_head *ndsttable,
80 			       struct hlist_head *nsrctable,
81 			       struct hlist_head *nspitable,
82 			       unsigned int nhashmask)
83 {
84 	struct hlist_node *tmp;
85 	struct xfrm_state *x;
86 
87 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
88 		unsigned int h;
89 
90 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
91 				    x->props.reqid, x->props.family,
92 				    nhashmask);
93 		hlist_add_head_rcu(&x->bydst, ndsttable + h);
94 
95 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
96 				    x->props.family,
97 				    nhashmask);
98 		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
99 
100 		if (x->id.spi) {
101 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
102 					    x->id.proto, x->props.family,
103 					    nhashmask);
104 			hlist_add_head_rcu(&x->byspi, nspitable + h);
105 		}
106 	}
107 }
108 
109 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
110 {
111 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
112 }
113 
114 static void xfrm_hash_resize(struct work_struct *work)
115 {
116 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
117 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
118 	unsigned long nsize, osize;
119 	unsigned int nhashmask, ohashmask;
120 	int i;
121 
122 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
123 	ndst = xfrm_hash_alloc(nsize);
124 	if (!ndst)
125 		return;
126 	nsrc = xfrm_hash_alloc(nsize);
127 	if (!nsrc) {
128 		xfrm_hash_free(ndst, nsize);
129 		return;
130 	}
131 	nspi = xfrm_hash_alloc(nsize);
132 	if (!nspi) {
133 		xfrm_hash_free(ndst, nsize);
134 		xfrm_hash_free(nsrc, nsize);
135 		return;
136 	}
137 
138 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
139 	write_seqcount_begin(&xfrm_state_hash_generation);
140 
141 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
142 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
143 	for (i = net->xfrm.state_hmask; i >= 0; i--)
144 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
145 
146 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
147 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
148 	ohashmask = net->xfrm.state_hmask;
149 
150 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
151 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
152 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
153 	net->xfrm.state_hmask = nhashmask;
154 
155 	write_seqcount_end(&xfrm_state_hash_generation);
156 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
157 
158 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
159 
160 	synchronize_rcu();
161 
162 	xfrm_hash_free(odst, osize);
163 	xfrm_hash_free(osrc, osize);
164 	xfrm_hash_free(ospi, osize);
165 }
166 
167 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
168 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
169 
170 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
171 
172 int __xfrm_state_delete(struct xfrm_state *x);
173 
174 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
175 bool km_is_alive(const struct km_event *c);
176 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
177 
178 static DEFINE_SPINLOCK(xfrm_type_lock);
179 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
180 {
181 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
182 	const struct xfrm_type **typemap;
183 	int err = 0;
184 
185 	if (unlikely(afinfo == NULL))
186 		return -EAFNOSUPPORT;
187 	typemap = afinfo->type_map;
188 	spin_lock_bh(&xfrm_type_lock);
189 
190 	if (likely(typemap[type->proto] == NULL))
191 		typemap[type->proto] = type;
192 	else
193 		err = -EEXIST;
194 	spin_unlock_bh(&xfrm_type_lock);
195 	xfrm_state_put_afinfo(afinfo);
196 	return err;
197 }
198 EXPORT_SYMBOL(xfrm_register_type);
199 
200 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
201 {
202 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
203 	const struct xfrm_type **typemap;
204 	int err = 0;
205 
206 	if (unlikely(afinfo == NULL))
207 		return -EAFNOSUPPORT;
208 	typemap = afinfo->type_map;
209 	spin_lock_bh(&xfrm_type_lock);
210 
211 	if (unlikely(typemap[type->proto] != type))
212 		err = -ENOENT;
213 	else
214 		typemap[type->proto] = NULL;
215 	spin_unlock_bh(&xfrm_type_lock);
216 	xfrm_state_put_afinfo(afinfo);
217 	return err;
218 }
219 EXPORT_SYMBOL(xfrm_unregister_type);
220 
221 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
222 {
223 	struct xfrm_state_afinfo *afinfo;
224 	const struct xfrm_type **typemap;
225 	const struct xfrm_type *type;
226 	int modload_attempted = 0;
227 
228 retry:
229 	afinfo = xfrm_state_get_afinfo(family);
230 	if (unlikely(afinfo == NULL))
231 		return NULL;
232 	typemap = afinfo->type_map;
233 
234 	type = typemap[proto];
235 	if (unlikely(type && !try_module_get(type->owner)))
236 		type = NULL;
237 	if (!type && !modload_attempted) {
238 		xfrm_state_put_afinfo(afinfo);
239 		request_module("xfrm-type-%d-%d", family, proto);
240 		modload_attempted = 1;
241 		goto retry;
242 	}
243 
244 	xfrm_state_put_afinfo(afinfo);
245 	return type;
246 }
247 
248 static void xfrm_put_type(const struct xfrm_type *type)
249 {
250 	module_put(type->owner);
251 }
252 
253 static DEFINE_SPINLOCK(xfrm_mode_lock);
254 int xfrm_register_mode(struct xfrm_mode *mode, int family)
255 {
256 	struct xfrm_state_afinfo *afinfo;
257 	struct xfrm_mode **modemap;
258 	int err;
259 
260 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
261 		return -EINVAL;
262 
263 	afinfo = xfrm_state_get_afinfo(family);
264 	if (unlikely(afinfo == NULL))
265 		return -EAFNOSUPPORT;
266 
267 	err = -EEXIST;
268 	modemap = afinfo->mode_map;
269 	spin_lock_bh(&xfrm_mode_lock);
270 	if (modemap[mode->encap])
271 		goto out;
272 
273 	err = -ENOENT;
274 	if (!try_module_get(afinfo->owner))
275 		goto out;
276 
277 	mode->afinfo = afinfo;
278 	modemap[mode->encap] = mode;
279 	err = 0;
280 
281 out:
282 	spin_unlock_bh(&xfrm_mode_lock);
283 	xfrm_state_put_afinfo(afinfo);
284 	return err;
285 }
286 EXPORT_SYMBOL(xfrm_register_mode);
287 
288 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
289 {
290 	struct xfrm_state_afinfo *afinfo;
291 	struct xfrm_mode **modemap;
292 	int err;
293 
294 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
295 		return -EINVAL;
296 
297 	afinfo = xfrm_state_get_afinfo(family);
298 	if (unlikely(afinfo == NULL))
299 		return -EAFNOSUPPORT;
300 
301 	err = -ENOENT;
302 	modemap = afinfo->mode_map;
303 	spin_lock_bh(&xfrm_mode_lock);
304 	if (likely(modemap[mode->encap] == mode)) {
305 		modemap[mode->encap] = NULL;
306 		module_put(mode->afinfo->owner);
307 		err = 0;
308 	}
309 
310 	spin_unlock_bh(&xfrm_mode_lock);
311 	xfrm_state_put_afinfo(afinfo);
312 	return err;
313 }
314 EXPORT_SYMBOL(xfrm_unregister_mode);
315 
316 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
317 {
318 	struct xfrm_state_afinfo *afinfo;
319 	struct xfrm_mode *mode;
320 	int modload_attempted = 0;
321 
322 	if (unlikely(encap >= XFRM_MODE_MAX))
323 		return NULL;
324 
325 retry:
326 	afinfo = xfrm_state_get_afinfo(family);
327 	if (unlikely(afinfo == NULL))
328 		return NULL;
329 
330 	mode = afinfo->mode_map[encap];
331 	if (unlikely(mode && !try_module_get(mode->owner)))
332 		mode = NULL;
333 	if (!mode && !modload_attempted) {
334 		xfrm_state_put_afinfo(afinfo);
335 		request_module("xfrm-mode-%d-%d", family, encap);
336 		modload_attempted = 1;
337 		goto retry;
338 	}
339 
340 	xfrm_state_put_afinfo(afinfo);
341 	return mode;
342 }
343 
344 static void xfrm_put_mode(struct xfrm_mode *mode)
345 {
346 	module_put(mode->owner);
347 }
348 
349 static void xfrm_state_gc_destroy(struct xfrm_state *x)
350 {
351 	tasklet_hrtimer_cancel(&x->mtimer);
352 	del_timer_sync(&x->rtimer);
353 	kfree(x->aead);
354 	kfree(x->aalg);
355 	kfree(x->ealg);
356 	kfree(x->calg);
357 	kfree(x->encap);
358 	kfree(x->coaddr);
359 	kfree(x->replay_esn);
360 	kfree(x->preplay_esn);
361 	if (x->inner_mode)
362 		xfrm_put_mode(x->inner_mode);
363 	if (x->inner_mode_iaf)
364 		xfrm_put_mode(x->inner_mode_iaf);
365 	if (x->outer_mode)
366 		xfrm_put_mode(x->outer_mode);
367 	if (x->type) {
368 		x->type->destructor(x);
369 		xfrm_put_type(x->type);
370 	}
371 	security_xfrm_state_free(x);
372 	kfree(x);
373 }
374 
375 static void xfrm_state_gc_task(struct work_struct *work)
376 {
377 	struct xfrm_state *x;
378 	struct hlist_node *tmp;
379 	struct hlist_head gc_list;
380 
381 	spin_lock_bh(&xfrm_state_gc_lock);
382 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
383 	spin_unlock_bh(&xfrm_state_gc_lock);
384 
385 	synchronize_rcu();
386 
387 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
388 		xfrm_state_gc_destroy(x);
389 }
390 
391 static inline unsigned long make_jiffies(long secs)
392 {
393 	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
394 		return MAX_SCHEDULE_TIMEOUT-1;
395 	else
396 		return secs*HZ;
397 }
398 
399 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
400 {
401 	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
402 	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
403 	unsigned long now = get_seconds();
404 	long next = LONG_MAX;
405 	int warn = 0;
406 	int err = 0;
407 
408 	spin_lock(&x->lock);
409 	if (x->km.state == XFRM_STATE_DEAD)
410 		goto out;
411 	if (x->km.state == XFRM_STATE_EXPIRED)
412 		goto expired;
413 	if (x->lft.hard_add_expires_seconds) {
414 		long tmo = x->lft.hard_add_expires_seconds +
415 			x->curlft.add_time - now;
416 		if (tmo <= 0) {
417 			if (x->xflags & XFRM_SOFT_EXPIRE) {
418 				/* enter hard expire without soft expire first?!
419 				 * setting a new date could trigger this.
420 				 * workarbound: fix x->curflt.add_time by below:
421 				 */
422 				x->curlft.add_time = now - x->saved_tmo - 1;
423 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
424 			} else
425 				goto expired;
426 		}
427 		if (tmo < next)
428 			next = tmo;
429 	}
430 	if (x->lft.hard_use_expires_seconds) {
431 		long tmo = x->lft.hard_use_expires_seconds +
432 			(x->curlft.use_time ? : now) - now;
433 		if (tmo <= 0)
434 			goto expired;
435 		if (tmo < next)
436 			next = tmo;
437 	}
438 	if (x->km.dying)
439 		goto resched;
440 	if (x->lft.soft_add_expires_seconds) {
441 		long tmo = x->lft.soft_add_expires_seconds +
442 			x->curlft.add_time - now;
443 		if (tmo <= 0) {
444 			warn = 1;
445 			x->xflags &= ~XFRM_SOFT_EXPIRE;
446 		} else if (tmo < next) {
447 			next = tmo;
448 			x->xflags |= XFRM_SOFT_EXPIRE;
449 			x->saved_tmo = tmo;
450 		}
451 	}
452 	if (x->lft.soft_use_expires_seconds) {
453 		long tmo = x->lft.soft_use_expires_seconds +
454 			(x->curlft.use_time ? : now) - now;
455 		if (tmo <= 0)
456 			warn = 1;
457 		else if (tmo < next)
458 			next = tmo;
459 	}
460 
461 	x->km.dying = warn;
462 	if (warn)
463 		km_state_expired(x, 0, 0);
464 resched:
465 	if (next != LONG_MAX) {
466 		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
467 	}
468 
469 	goto out;
470 
471 expired:
472 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
473 		x->km.state = XFRM_STATE_EXPIRED;
474 
475 	err = __xfrm_state_delete(x);
476 	if (!err)
477 		km_state_expired(x, 1, 0);
478 
479 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
480 
481 out:
482 	spin_unlock(&x->lock);
483 	return HRTIMER_NORESTART;
484 }
485 
486 static void xfrm_replay_timer_handler(unsigned long data);
487 
488 struct xfrm_state *xfrm_state_alloc(struct net *net)
489 {
490 	struct xfrm_state *x;
491 
492 	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
493 
494 	if (x) {
495 		write_pnet(&x->xs_net, net);
496 		atomic_set(&x->refcnt, 1);
497 		atomic_set(&x->tunnel_users, 0);
498 		INIT_LIST_HEAD(&x->km.all);
499 		INIT_HLIST_NODE(&x->bydst);
500 		INIT_HLIST_NODE(&x->bysrc);
501 		INIT_HLIST_NODE(&x->byspi);
502 		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
503 					CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
504 		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
505 				(unsigned long)x);
506 		x->curlft.add_time = get_seconds();
507 		x->lft.soft_byte_limit = XFRM_INF;
508 		x->lft.soft_packet_limit = XFRM_INF;
509 		x->lft.hard_byte_limit = XFRM_INF;
510 		x->lft.hard_packet_limit = XFRM_INF;
511 		x->replay_maxage = 0;
512 		x->replay_maxdiff = 0;
513 		x->inner_mode = NULL;
514 		x->inner_mode_iaf = NULL;
515 		spin_lock_init(&x->lock);
516 	}
517 	return x;
518 }
519 EXPORT_SYMBOL(xfrm_state_alloc);
520 
521 void __xfrm_state_destroy(struct xfrm_state *x)
522 {
523 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
524 
525 	spin_lock_bh(&xfrm_state_gc_lock);
526 	hlist_add_head(&x->gclist, &xfrm_state_gc_list);
527 	spin_unlock_bh(&xfrm_state_gc_lock);
528 	schedule_work(&xfrm_state_gc_work);
529 }
530 EXPORT_SYMBOL(__xfrm_state_destroy);
531 
532 int __xfrm_state_delete(struct xfrm_state *x)
533 {
534 	struct net *net = xs_net(x);
535 	int err = -ESRCH;
536 
537 	if (x->km.state != XFRM_STATE_DEAD) {
538 		x->km.state = XFRM_STATE_DEAD;
539 		spin_lock(&net->xfrm.xfrm_state_lock);
540 		list_del(&x->km.all);
541 		hlist_del_rcu(&x->bydst);
542 		hlist_del_rcu(&x->bysrc);
543 		if (x->id.spi)
544 			hlist_del_rcu(&x->byspi);
545 		net->xfrm.state_num--;
546 		spin_unlock(&net->xfrm.xfrm_state_lock);
547 
548 		/* All xfrm_state objects are created by xfrm_state_alloc.
549 		 * The xfrm_state_alloc call gives a reference, and that
550 		 * is what we are dropping here.
551 		 */
552 		xfrm_state_put(x);
553 		err = 0;
554 	}
555 
556 	return err;
557 }
558 EXPORT_SYMBOL(__xfrm_state_delete);
559 
560 int xfrm_state_delete(struct xfrm_state *x)
561 {
562 	int err;
563 
564 	spin_lock_bh(&x->lock);
565 	err = __xfrm_state_delete(x);
566 	spin_unlock_bh(&x->lock);
567 
568 	return err;
569 }
570 EXPORT_SYMBOL(xfrm_state_delete);
571 
572 #ifdef CONFIG_SECURITY_NETWORK_XFRM
573 static inline int
574 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
575 {
576 	int i, err = 0;
577 
578 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
579 		struct xfrm_state *x;
580 
581 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
582 			if (xfrm_id_proto_match(x->id.proto, proto) &&
583 			   (err = security_xfrm_state_delete(x)) != 0) {
584 				xfrm_audit_state_delete(x, 0, task_valid);
585 				return err;
586 			}
587 		}
588 	}
589 
590 	return err;
591 }
592 #else
593 static inline int
594 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
595 {
596 	return 0;
597 }
598 #endif
599 
600 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
601 {
602 	int i, err = 0, cnt = 0;
603 
604 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
605 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
606 	if (err)
607 		goto out;
608 
609 	err = -ESRCH;
610 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
611 		struct xfrm_state *x;
612 restart:
613 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
614 			if (!xfrm_state_kern(x) &&
615 			    xfrm_id_proto_match(x->id.proto, proto)) {
616 				xfrm_state_hold(x);
617 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
618 
619 				err = xfrm_state_delete(x);
620 				xfrm_audit_state_delete(x, err ? 0 : 1,
621 							task_valid);
622 				xfrm_state_put(x);
623 				if (!err)
624 					cnt++;
625 
626 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
627 				goto restart;
628 			}
629 		}
630 	}
631 	if (cnt)
632 		err = 0;
633 
634 out:
635 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
636 	return err;
637 }
638 EXPORT_SYMBOL(xfrm_state_flush);
639 
640 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
641 {
642 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
643 	si->sadcnt = net->xfrm.state_num;
644 	si->sadhcnt = net->xfrm.state_hmask;
645 	si->sadhmcnt = xfrm_state_hashmax;
646 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
647 }
648 EXPORT_SYMBOL(xfrm_sad_getinfo);
649 
650 static int
651 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
652 		    const struct xfrm_tmpl *tmpl,
653 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
654 		    unsigned short family)
655 {
656 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
657 	if (!afinfo)
658 		return -1;
659 	afinfo->init_tempsel(&x->sel, fl);
660 
661 	if (family != tmpl->encap_family) {
662 		xfrm_state_put_afinfo(afinfo);
663 		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
664 		if (!afinfo)
665 			return -1;
666 	}
667 	afinfo->init_temprop(x, tmpl, daddr, saddr);
668 	xfrm_state_put_afinfo(afinfo);
669 	return 0;
670 }
671 
672 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
673 					      const xfrm_address_t *daddr,
674 					      __be32 spi, u8 proto,
675 					      unsigned short family)
676 {
677 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
678 	struct xfrm_state *x;
679 
680 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
681 		if (x->props.family != family ||
682 		    x->id.spi       != spi ||
683 		    x->id.proto     != proto ||
684 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
685 			continue;
686 
687 		if ((mark & x->mark.m) != x->mark.v)
688 			continue;
689 		if (!xfrm_state_hold_rcu(x))
690 			continue;
691 		return x;
692 	}
693 
694 	return NULL;
695 }
696 
697 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
698 						     const xfrm_address_t *daddr,
699 						     const xfrm_address_t *saddr,
700 						     u8 proto, unsigned short family)
701 {
702 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
703 	struct xfrm_state *x;
704 
705 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
706 		if (x->props.family != family ||
707 		    x->id.proto     != proto ||
708 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
709 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
710 			continue;
711 
712 		if ((mark & x->mark.m) != x->mark.v)
713 			continue;
714 		if (!xfrm_state_hold_rcu(x))
715 			continue;
716 		return x;
717 	}
718 
719 	return NULL;
720 }
721 
722 static inline struct xfrm_state *
723 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
724 {
725 	struct net *net = xs_net(x);
726 	u32 mark = x->mark.v & x->mark.m;
727 
728 	if (use_spi)
729 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
730 					   x->id.spi, x->id.proto, family);
731 	else
732 		return __xfrm_state_lookup_byaddr(net, mark,
733 						  &x->id.daddr,
734 						  &x->props.saddr,
735 						  x->id.proto, family);
736 }
737 
738 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
739 {
740 	if (have_hash_collision &&
741 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
742 	    net->xfrm.state_num > net->xfrm.state_hmask)
743 		schedule_work(&net->xfrm.state_hash_work);
744 }
745 
746 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
747 			       const struct flowi *fl, unsigned short family,
748 			       struct xfrm_state **best, int *acq_in_progress,
749 			       int *error)
750 {
751 	/* Resolution logic:
752 	 * 1. There is a valid state with matching selector. Done.
753 	 * 2. Valid state with inappropriate selector. Skip.
754 	 *
755 	 * Entering area of "sysdeps".
756 	 *
757 	 * 3. If state is not valid, selector is temporary, it selects
758 	 *    only session which triggered previous resolution. Key
759 	 *    manager will do something to install a state with proper
760 	 *    selector.
761 	 */
762 	if (x->km.state == XFRM_STATE_VALID) {
763 		if ((x->sel.family &&
764 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
765 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
766 			return;
767 
768 		if (!*best ||
769 		    (*best)->km.dying > x->km.dying ||
770 		    ((*best)->km.dying == x->km.dying &&
771 		     (*best)->curlft.add_time < x->curlft.add_time))
772 			*best = x;
773 	} else if (x->km.state == XFRM_STATE_ACQ) {
774 		*acq_in_progress = 1;
775 	} else if (x->km.state == XFRM_STATE_ERROR ||
776 		   x->km.state == XFRM_STATE_EXPIRED) {
777 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
778 		    security_xfrm_state_pol_flow_match(x, pol, fl))
779 			*error = -ESRCH;
780 	}
781 }
782 
783 struct xfrm_state *
784 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
785 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
786 		struct xfrm_policy *pol, int *err,
787 		unsigned short family)
788 {
789 	static xfrm_address_t saddr_wildcard = { };
790 	struct net *net = xp_net(pol);
791 	unsigned int h, h_wildcard;
792 	struct xfrm_state *x, *x0, *to_put;
793 	int acquire_in_progress = 0;
794 	int error = 0;
795 	struct xfrm_state *best = NULL;
796 	u32 mark = pol->mark.v & pol->mark.m;
797 	unsigned short encap_family = tmpl->encap_family;
798 	unsigned int sequence;
799 	struct km_event c;
800 
801 	to_put = NULL;
802 
803 	sequence = read_seqcount_begin(&xfrm_state_hash_generation);
804 
805 	rcu_read_lock();
806 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
807 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
808 		if (x->props.family == encap_family &&
809 		    x->props.reqid == tmpl->reqid &&
810 		    (mark & x->mark.m) == x->mark.v &&
811 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
812 		    xfrm_state_addr_check(x, daddr, saddr, encap_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, encap_family,
817 					   &best, &acquire_in_progress, &error);
818 	}
819 	if (best || acquire_in_progress)
820 		goto found;
821 
822 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
823 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
824 		if (x->props.family == encap_family &&
825 		    x->props.reqid == tmpl->reqid &&
826 		    (mark & x->mark.m) == x->mark.v &&
827 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
828 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
829 		    tmpl->mode == x->props.mode &&
830 		    tmpl->id.proto == x->id.proto &&
831 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
832 			xfrm_state_look_at(pol, x, fl, encap_family,
833 					   &best, &acquire_in_progress, &error);
834 	}
835 
836 found:
837 	x = best;
838 	if (!x && !error && !acquire_in_progress) {
839 		if (tmpl->id.spi &&
840 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
841 					      tmpl->id.proto, encap_family)) != NULL) {
842 			to_put = x0;
843 			error = -EEXIST;
844 			goto out;
845 		}
846 
847 		c.net = net;
848 		/* If the KMs have no listeners (yet...), avoid allocating an SA
849 		 * for each and every packet - garbage collection might not
850 		 * handle the flood.
851 		 */
852 		if (!km_is_alive(&c)) {
853 			error = -ESRCH;
854 			goto out;
855 		}
856 
857 		x = xfrm_state_alloc(net);
858 		if (x == NULL) {
859 			error = -ENOMEM;
860 			goto out;
861 		}
862 		/* Initialize temporary state matching only
863 		 * to current session. */
864 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
865 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
866 
867 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
868 		if (error) {
869 			x->km.state = XFRM_STATE_DEAD;
870 			to_put = x;
871 			x = NULL;
872 			goto out;
873 		}
874 
875 		if (km_query(x, tmpl, pol) == 0) {
876 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
877 			x->km.state = XFRM_STATE_ACQ;
878 			list_add(&x->km.all, &net->xfrm.state_all);
879 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
880 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
881 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
882 			if (x->id.spi) {
883 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
884 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
885 			}
886 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
887 			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
888 			net->xfrm.state_num++;
889 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
890 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
891 		} else {
892 			x->km.state = XFRM_STATE_DEAD;
893 			to_put = x;
894 			x = NULL;
895 			error = -ESRCH;
896 		}
897 	}
898 out:
899 	if (x) {
900 		if (!xfrm_state_hold_rcu(x)) {
901 			*err = -EAGAIN;
902 			x = NULL;
903 		}
904 	} else {
905 		*err = acquire_in_progress ? -EAGAIN : error;
906 	}
907 	rcu_read_unlock();
908 	if (to_put)
909 		xfrm_state_put(to_put);
910 
911 	if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
912 		*err = -EAGAIN;
913 		if (x) {
914 			xfrm_state_put(x);
915 			x = NULL;
916 		}
917 	}
918 
919 	return x;
920 }
921 
922 struct xfrm_state *
923 xfrm_stateonly_find(struct net *net, u32 mark,
924 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
925 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
926 {
927 	unsigned int h;
928 	struct xfrm_state *rx = NULL, *x = NULL;
929 
930 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
931 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
932 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
933 		if (x->props.family == family &&
934 		    x->props.reqid == reqid &&
935 		    (mark & x->mark.m) == x->mark.v &&
936 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
937 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
938 		    mode == x->props.mode &&
939 		    proto == x->id.proto &&
940 		    x->km.state == XFRM_STATE_VALID) {
941 			rx = x;
942 			break;
943 		}
944 	}
945 
946 	if (rx)
947 		xfrm_state_hold(rx);
948 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
949 
950 
951 	return rx;
952 }
953 EXPORT_SYMBOL(xfrm_stateonly_find);
954 
955 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
956 					      unsigned short family)
957 {
958 	struct xfrm_state *x;
959 	struct xfrm_state_walk *w;
960 
961 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
962 	list_for_each_entry(w, &net->xfrm.state_all, all) {
963 		x = container_of(w, struct xfrm_state, km);
964 		if (x->props.family != family ||
965 			x->id.spi != spi)
966 			continue;
967 
968 		xfrm_state_hold(x);
969 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
970 		return x;
971 	}
972 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
973 	return NULL;
974 }
975 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
976 
977 static void __xfrm_state_insert(struct xfrm_state *x)
978 {
979 	struct net *net = xs_net(x);
980 	unsigned int h;
981 
982 	list_add(&x->km.all, &net->xfrm.state_all);
983 
984 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
985 			  x->props.reqid, x->props.family);
986 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
987 
988 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
989 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
990 
991 	if (x->id.spi) {
992 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
993 				  x->props.family);
994 
995 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
996 	}
997 
998 	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
999 	if (x->replay_maxage)
1000 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1001 
1002 	net->xfrm.state_num++;
1003 
1004 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1005 }
1006 
1007 /* net->xfrm.xfrm_state_lock is held */
1008 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1009 {
1010 	struct net *net = xs_net(xnew);
1011 	unsigned short family = xnew->props.family;
1012 	u32 reqid = xnew->props.reqid;
1013 	struct xfrm_state *x;
1014 	unsigned int h;
1015 	u32 mark = xnew->mark.v & xnew->mark.m;
1016 
1017 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1018 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1019 		if (x->props.family	== family &&
1020 		    x->props.reqid	== reqid &&
1021 		    (mark & x->mark.m) == x->mark.v &&
1022 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1023 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1024 			x->genid++;
1025 	}
1026 }
1027 
1028 void xfrm_state_insert(struct xfrm_state *x)
1029 {
1030 	struct net *net = xs_net(x);
1031 
1032 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1033 	__xfrm_state_bump_genids(x);
1034 	__xfrm_state_insert(x);
1035 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1036 }
1037 EXPORT_SYMBOL(xfrm_state_insert);
1038 
1039 /* net->xfrm.xfrm_state_lock is held */
1040 static struct xfrm_state *__find_acq_core(struct net *net,
1041 					  const struct xfrm_mark *m,
1042 					  unsigned short family, u8 mode,
1043 					  u32 reqid, u8 proto,
1044 					  const xfrm_address_t *daddr,
1045 					  const xfrm_address_t *saddr,
1046 					  int create)
1047 {
1048 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1049 	struct xfrm_state *x;
1050 	u32 mark = m->v & m->m;
1051 
1052 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1053 		if (x->props.reqid  != reqid ||
1054 		    x->props.mode   != mode ||
1055 		    x->props.family != family ||
1056 		    x->km.state     != XFRM_STATE_ACQ ||
1057 		    x->id.spi       != 0 ||
1058 		    x->id.proto	    != proto ||
1059 		    (mark & x->mark.m) != x->mark.v ||
1060 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1061 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1062 			continue;
1063 
1064 		xfrm_state_hold(x);
1065 		return x;
1066 	}
1067 
1068 	if (!create)
1069 		return NULL;
1070 
1071 	x = xfrm_state_alloc(net);
1072 	if (likely(x)) {
1073 		switch (family) {
1074 		case AF_INET:
1075 			x->sel.daddr.a4 = daddr->a4;
1076 			x->sel.saddr.a4 = saddr->a4;
1077 			x->sel.prefixlen_d = 32;
1078 			x->sel.prefixlen_s = 32;
1079 			x->props.saddr.a4 = saddr->a4;
1080 			x->id.daddr.a4 = daddr->a4;
1081 			break;
1082 
1083 		case AF_INET6:
1084 			x->sel.daddr.in6 = daddr->in6;
1085 			x->sel.saddr.in6 = saddr->in6;
1086 			x->sel.prefixlen_d = 128;
1087 			x->sel.prefixlen_s = 128;
1088 			x->props.saddr.in6 = saddr->in6;
1089 			x->id.daddr.in6 = daddr->in6;
1090 			break;
1091 		}
1092 
1093 		x->km.state = XFRM_STATE_ACQ;
1094 		x->id.proto = proto;
1095 		x->props.family = family;
1096 		x->props.mode = mode;
1097 		x->props.reqid = reqid;
1098 		x->mark.v = m->v;
1099 		x->mark.m = m->m;
1100 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1101 		xfrm_state_hold(x);
1102 		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1103 		list_add(&x->km.all, &net->xfrm.state_all);
1104 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1105 		h = xfrm_src_hash(net, daddr, saddr, family);
1106 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1107 
1108 		net->xfrm.state_num++;
1109 
1110 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1111 	}
1112 
1113 	return x;
1114 }
1115 
1116 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1117 
1118 int xfrm_state_add(struct xfrm_state *x)
1119 {
1120 	struct net *net = xs_net(x);
1121 	struct xfrm_state *x1, *to_put;
1122 	int family;
1123 	int err;
1124 	u32 mark = x->mark.v & x->mark.m;
1125 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1126 
1127 	family = x->props.family;
1128 
1129 	to_put = NULL;
1130 
1131 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1132 
1133 	x1 = __xfrm_state_locate(x, use_spi, family);
1134 	if (x1) {
1135 		to_put = x1;
1136 		x1 = NULL;
1137 		err = -EEXIST;
1138 		goto out;
1139 	}
1140 
1141 	if (use_spi && x->km.seq) {
1142 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1143 		if (x1 && ((x1->id.proto != x->id.proto) ||
1144 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1145 			to_put = x1;
1146 			x1 = NULL;
1147 		}
1148 	}
1149 
1150 	if (use_spi && !x1)
1151 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1152 				     x->props.reqid, x->id.proto,
1153 				     &x->id.daddr, &x->props.saddr, 0);
1154 
1155 	__xfrm_state_bump_genids(x);
1156 	__xfrm_state_insert(x);
1157 	err = 0;
1158 
1159 out:
1160 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1161 
1162 	if (x1) {
1163 		xfrm_state_delete(x1);
1164 		xfrm_state_put(x1);
1165 	}
1166 
1167 	if (to_put)
1168 		xfrm_state_put(to_put);
1169 
1170 	return err;
1171 }
1172 EXPORT_SYMBOL(xfrm_state_add);
1173 
1174 #ifdef CONFIG_XFRM_MIGRATE
1175 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1176 {
1177 	struct net *net = xs_net(orig);
1178 	struct xfrm_state *x = xfrm_state_alloc(net);
1179 	if (!x)
1180 		goto out;
1181 
1182 	memcpy(&x->id, &orig->id, sizeof(x->id));
1183 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1184 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1185 	x->props.mode = orig->props.mode;
1186 	x->props.replay_window = orig->props.replay_window;
1187 	x->props.reqid = orig->props.reqid;
1188 	x->props.family = orig->props.family;
1189 	x->props.saddr = orig->props.saddr;
1190 
1191 	if (orig->aalg) {
1192 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1193 		if (!x->aalg)
1194 			goto error;
1195 	}
1196 	x->props.aalgo = orig->props.aalgo;
1197 
1198 	if (orig->aead) {
1199 		x->aead = xfrm_algo_aead_clone(orig->aead);
1200 		if (!x->aead)
1201 			goto error;
1202 	}
1203 	if (orig->ealg) {
1204 		x->ealg = xfrm_algo_clone(orig->ealg);
1205 		if (!x->ealg)
1206 			goto error;
1207 	}
1208 	x->props.ealgo = orig->props.ealgo;
1209 
1210 	if (orig->calg) {
1211 		x->calg = xfrm_algo_clone(orig->calg);
1212 		if (!x->calg)
1213 			goto error;
1214 	}
1215 	x->props.calgo = orig->props.calgo;
1216 
1217 	if (orig->encap) {
1218 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1219 		if (!x->encap)
1220 			goto error;
1221 	}
1222 
1223 	if (orig->coaddr) {
1224 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1225 				    GFP_KERNEL);
1226 		if (!x->coaddr)
1227 			goto error;
1228 	}
1229 
1230 	if (orig->replay_esn) {
1231 		if (xfrm_replay_clone(x, orig))
1232 			goto error;
1233 	}
1234 
1235 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1236 
1237 	if (xfrm_init_state(x) < 0)
1238 		goto error;
1239 
1240 	x->props.flags = orig->props.flags;
1241 	x->props.extra_flags = orig->props.extra_flags;
1242 
1243 	x->tfcpad = orig->tfcpad;
1244 	x->replay_maxdiff = orig->replay_maxdiff;
1245 	x->replay_maxage = orig->replay_maxage;
1246 	x->curlft.add_time = orig->curlft.add_time;
1247 	x->km.state = orig->km.state;
1248 	x->km.seq = orig->km.seq;
1249 
1250 	return x;
1251 
1252  error:
1253 	xfrm_state_put(x);
1254 out:
1255 	return NULL;
1256 }
1257 
1258 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1259 {
1260 	unsigned int h;
1261 	struct xfrm_state *x = NULL;
1262 
1263 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1264 
1265 	if (m->reqid) {
1266 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1267 				  m->reqid, m->old_family);
1268 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1269 			if (x->props.mode != m->mode ||
1270 			    x->id.proto != m->proto)
1271 				continue;
1272 			if (m->reqid && x->props.reqid != m->reqid)
1273 				continue;
1274 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1275 					     m->old_family) ||
1276 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1277 					     m->old_family))
1278 				continue;
1279 			xfrm_state_hold(x);
1280 			break;
1281 		}
1282 	} else {
1283 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1284 				  m->old_family);
1285 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1286 			if (x->props.mode != m->mode ||
1287 			    x->id.proto != m->proto)
1288 				continue;
1289 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1290 					     m->old_family) ||
1291 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1292 					     m->old_family))
1293 				continue;
1294 			xfrm_state_hold(x);
1295 			break;
1296 		}
1297 	}
1298 
1299 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1300 
1301 	return x;
1302 }
1303 EXPORT_SYMBOL(xfrm_migrate_state_find);
1304 
1305 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1306 				      struct xfrm_migrate *m)
1307 {
1308 	struct xfrm_state *xc;
1309 
1310 	xc = xfrm_state_clone(x);
1311 	if (!xc)
1312 		return NULL;
1313 
1314 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1315 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1316 
1317 	/* add state */
1318 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1319 		/* a care is needed when the destination address of the
1320 		   state is to be updated as it is a part of triplet */
1321 		xfrm_state_insert(xc);
1322 	} else {
1323 		if (xfrm_state_add(xc) < 0)
1324 			goto error;
1325 	}
1326 
1327 	return xc;
1328 error:
1329 	xfrm_state_put(xc);
1330 	return NULL;
1331 }
1332 EXPORT_SYMBOL(xfrm_state_migrate);
1333 #endif
1334 
1335 int xfrm_state_update(struct xfrm_state *x)
1336 {
1337 	struct xfrm_state *x1, *to_put;
1338 	int err;
1339 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1340 	struct net *net = xs_net(x);
1341 
1342 	to_put = NULL;
1343 
1344 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1345 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1346 
1347 	err = -ESRCH;
1348 	if (!x1)
1349 		goto out;
1350 
1351 	if (xfrm_state_kern(x1)) {
1352 		to_put = x1;
1353 		err = -EEXIST;
1354 		goto out;
1355 	}
1356 
1357 	if (x1->km.state == XFRM_STATE_ACQ) {
1358 		__xfrm_state_insert(x);
1359 		x = NULL;
1360 	}
1361 	err = 0;
1362 
1363 out:
1364 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1365 
1366 	if (to_put)
1367 		xfrm_state_put(to_put);
1368 
1369 	if (err)
1370 		return err;
1371 
1372 	if (!x) {
1373 		xfrm_state_delete(x1);
1374 		xfrm_state_put(x1);
1375 		return 0;
1376 	}
1377 
1378 	err = -EINVAL;
1379 	spin_lock_bh(&x1->lock);
1380 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1381 		if (x->encap && x1->encap)
1382 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1383 		if (x->coaddr && x1->coaddr) {
1384 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1385 		}
1386 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1387 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1388 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1389 		x1->km.dying = 0;
1390 
1391 		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1392 		if (x1->curlft.use_time)
1393 			xfrm_state_check_expire(x1);
1394 
1395 		err = 0;
1396 		x->km.state = XFRM_STATE_DEAD;
1397 		__xfrm_state_put(x);
1398 	}
1399 	spin_unlock_bh(&x1->lock);
1400 
1401 	xfrm_state_put(x1);
1402 
1403 	return err;
1404 }
1405 EXPORT_SYMBOL(xfrm_state_update);
1406 
1407 int xfrm_state_check_expire(struct xfrm_state *x)
1408 {
1409 	if (!x->curlft.use_time)
1410 		x->curlft.use_time = get_seconds();
1411 
1412 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1413 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1414 		x->km.state = XFRM_STATE_EXPIRED;
1415 		tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1416 		return -EINVAL;
1417 	}
1418 
1419 	if (!x->km.dying &&
1420 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1421 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1422 		x->km.dying = 1;
1423 		km_state_expired(x, 0, 0);
1424 	}
1425 	return 0;
1426 }
1427 EXPORT_SYMBOL(xfrm_state_check_expire);
1428 
1429 struct xfrm_state *
1430 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1431 		  u8 proto, unsigned short family)
1432 {
1433 	struct xfrm_state *x;
1434 
1435 	rcu_read_lock();
1436 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1437 	rcu_read_unlock();
1438 	return x;
1439 }
1440 EXPORT_SYMBOL(xfrm_state_lookup);
1441 
1442 struct xfrm_state *
1443 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1444 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1445 			 u8 proto, unsigned short family)
1446 {
1447 	struct xfrm_state *x;
1448 
1449 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1450 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1451 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1452 	return x;
1453 }
1454 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1455 
1456 struct xfrm_state *
1457 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1458 	      u8 proto, const xfrm_address_t *daddr,
1459 	      const xfrm_address_t *saddr, int create, unsigned short family)
1460 {
1461 	struct xfrm_state *x;
1462 
1463 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1464 	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1465 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1466 
1467 	return x;
1468 }
1469 EXPORT_SYMBOL(xfrm_find_acq);
1470 
1471 #ifdef CONFIG_XFRM_SUB_POLICY
1472 int
1473 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1474 	       unsigned short family, struct net *net)
1475 {
1476 	int err = 0;
1477 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1478 	if (!afinfo)
1479 		return -EAFNOSUPPORT;
1480 
1481 	spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1482 	if (afinfo->tmpl_sort)
1483 		err = afinfo->tmpl_sort(dst, src, n);
1484 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1485 	xfrm_state_put_afinfo(afinfo);
1486 	return err;
1487 }
1488 EXPORT_SYMBOL(xfrm_tmpl_sort);
1489 
1490 int
1491 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1492 		unsigned short family)
1493 {
1494 	int err = 0;
1495 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1496 	struct net *net = xs_net(*src);
1497 
1498 	if (!afinfo)
1499 		return -EAFNOSUPPORT;
1500 
1501 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1502 	if (afinfo->state_sort)
1503 		err = afinfo->state_sort(dst, src, n);
1504 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1505 	xfrm_state_put_afinfo(afinfo);
1506 	return err;
1507 }
1508 EXPORT_SYMBOL(xfrm_state_sort);
1509 #endif
1510 
1511 /* Silly enough, but I'm lazy to build resolution list */
1512 
1513 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1514 {
1515 	int i;
1516 
1517 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1518 		struct xfrm_state *x;
1519 
1520 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1521 			if (x->km.seq == seq &&
1522 			    (mark & x->mark.m) == x->mark.v &&
1523 			    x->km.state == XFRM_STATE_ACQ) {
1524 				xfrm_state_hold(x);
1525 				return x;
1526 			}
1527 		}
1528 	}
1529 	return NULL;
1530 }
1531 
1532 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1533 {
1534 	struct xfrm_state *x;
1535 
1536 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1537 	x = __xfrm_find_acq_byseq(net, mark, seq);
1538 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1539 	return x;
1540 }
1541 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1542 
1543 u32 xfrm_get_acqseq(void)
1544 {
1545 	u32 res;
1546 	static atomic_t acqseq;
1547 
1548 	do {
1549 		res = atomic_inc_return(&acqseq);
1550 	} while (!res);
1551 
1552 	return res;
1553 }
1554 EXPORT_SYMBOL(xfrm_get_acqseq);
1555 
1556 int verify_spi_info(u8 proto, u32 min, u32 max)
1557 {
1558 	switch (proto) {
1559 	case IPPROTO_AH:
1560 	case IPPROTO_ESP:
1561 		break;
1562 
1563 	case IPPROTO_COMP:
1564 		/* IPCOMP spi is 16-bits. */
1565 		if (max >= 0x10000)
1566 			return -EINVAL;
1567 		break;
1568 
1569 	default:
1570 		return -EINVAL;
1571 	}
1572 
1573 	if (min > max)
1574 		return -EINVAL;
1575 
1576 	return 0;
1577 }
1578 EXPORT_SYMBOL(verify_spi_info);
1579 
1580 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1581 {
1582 	struct net *net = xs_net(x);
1583 	unsigned int h;
1584 	struct xfrm_state *x0;
1585 	int err = -ENOENT;
1586 	__be32 minspi = htonl(low);
1587 	__be32 maxspi = htonl(high);
1588 	u32 mark = x->mark.v & x->mark.m;
1589 
1590 	spin_lock_bh(&x->lock);
1591 	if (x->km.state == XFRM_STATE_DEAD)
1592 		goto unlock;
1593 
1594 	err = 0;
1595 	if (x->id.spi)
1596 		goto unlock;
1597 
1598 	err = -ENOENT;
1599 
1600 	if (minspi == maxspi) {
1601 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1602 		if (x0) {
1603 			xfrm_state_put(x0);
1604 			goto unlock;
1605 		}
1606 		x->id.spi = minspi;
1607 	} else {
1608 		u32 spi = 0;
1609 		for (h = 0; h < high-low+1; h++) {
1610 			spi = low + prandom_u32()%(high-low+1);
1611 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1612 			if (x0 == NULL) {
1613 				x->id.spi = htonl(spi);
1614 				break;
1615 			}
1616 			xfrm_state_put(x0);
1617 		}
1618 	}
1619 	if (x->id.spi) {
1620 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
1621 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1622 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1623 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1624 
1625 		err = 0;
1626 	}
1627 
1628 unlock:
1629 	spin_unlock_bh(&x->lock);
1630 
1631 	return err;
1632 }
1633 EXPORT_SYMBOL(xfrm_alloc_spi);
1634 
1635 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1636 				      struct xfrm_address_filter *filter)
1637 {
1638 	if (filter) {
1639 		if ((filter->family == AF_INET ||
1640 		     filter->family == AF_INET6) &&
1641 		    x->props.family != filter->family)
1642 			return false;
1643 
1644 		return addr_match(&x->props.saddr, &filter->saddr,
1645 				  filter->splen) &&
1646 		       addr_match(&x->id.daddr, &filter->daddr,
1647 				  filter->dplen);
1648 	}
1649 	return true;
1650 }
1651 
1652 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1653 		    int (*func)(struct xfrm_state *, int, void*),
1654 		    void *data)
1655 {
1656 	struct xfrm_state *state;
1657 	struct xfrm_state_walk *x;
1658 	int err = 0;
1659 
1660 	if (walk->seq != 0 && list_empty(&walk->all))
1661 		return 0;
1662 
1663 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1664 	if (list_empty(&walk->all))
1665 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1666 	else
1667 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1668 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1669 		if (x->state == XFRM_STATE_DEAD)
1670 			continue;
1671 		state = container_of(x, struct xfrm_state, km);
1672 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1673 			continue;
1674 		if (!__xfrm_state_filter_match(state, walk->filter))
1675 			continue;
1676 		err = func(state, walk->seq, data);
1677 		if (err) {
1678 			list_move_tail(&walk->all, &x->all);
1679 			goto out;
1680 		}
1681 		walk->seq++;
1682 	}
1683 	if (walk->seq == 0) {
1684 		err = -ENOENT;
1685 		goto out;
1686 	}
1687 	list_del_init(&walk->all);
1688 out:
1689 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1690 	return err;
1691 }
1692 EXPORT_SYMBOL(xfrm_state_walk);
1693 
1694 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1695 			  struct xfrm_address_filter *filter)
1696 {
1697 	INIT_LIST_HEAD(&walk->all);
1698 	walk->proto = proto;
1699 	walk->state = XFRM_STATE_DEAD;
1700 	walk->seq = 0;
1701 	walk->filter = filter;
1702 }
1703 EXPORT_SYMBOL(xfrm_state_walk_init);
1704 
1705 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1706 {
1707 	kfree(walk->filter);
1708 
1709 	if (list_empty(&walk->all))
1710 		return;
1711 
1712 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1713 	list_del(&walk->all);
1714 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1715 }
1716 EXPORT_SYMBOL(xfrm_state_walk_done);
1717 
1718 static void xfrm_replay_timer_handler(unsigned long data)
1719 {
1720 	struct xfrm_state *x = (struct xfrm_state *)data;
1721 
1722 	spin_lock(&x->lock);
1723 
1724 	if (x->km.state == XFRM_STATE_VALID) {
1725 		if (xfrm_aevent_is_on(xs_net(x)))
1726 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1727 		else
1728 			x->xflags |= XFRM_TIME_DEFER;
1729 	}
1730 
1731 	spin_unlock(&x->lock);
1732 }
1733 
1734 static LIST_HEAD(xfrm_km_list);
1735 
1736 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1737 {
1738 	struct xfrm_mgr *km;
1739 
1740 	rcu_read_lock();
1741 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1742 		if (km->notify_policy)
1743 			km->notify_policy(xp, dir, c);
1744 	rcu_read_unlock();
1745 }
1746 
1747 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1748 {
1749 	struct xfrm_mgr *km;
1750 	rcu_read_lock();
1751 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1752 		if (km->notify)
1753 			km->notify(x, c);
1754 	rcu_read_unlock();
1755 }
1756 
1757 EXPORT_SYMBOL(km_policy_notify);
1758 EXPORT_SYMBOL(km_state_notify);
1759 
1760 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1761 {
1762 	struct km_event c;
1763 
1764 	c.data.hard = hard;
1765 	c.portid = portid;
1766 	c.event = XFRM_MSG_EXPIRE;
1767 	km_state_notify(x, &c);
1768 }
1769 
1770 EXPORT_SYMBOL(km_state_expired);
1771 /*
1772  * We send to all registered managers regardless of failure
1773  * We are happy with one success
1774 */
1775 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1776 {
1777 	int err = -EINVAL, acqret;
1778 	struct xfrm_mgr *km;
1779 
1780 	rcu_read_lock();
1781 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1782 		acqret = km->acquire(x, t, pol);
1783 		if (!acqret)
1784 			err = acqret;
1785 	}
1786 	rcu_read_unlock();
1787 	return err;
1788 }
1789 EXPORT_SYMBOL(km_query);
1790 
1791 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1792 {
1793 	int err = -EINVAL;
1794 	struct xfrm_mgr *km;
1795 
1796 	rcu_read_lock();
1797 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1798 		if (km->new_mapping)
1799 			err = km->new_mapping(x, ipaddr, sport);
1800 		if (!err)
1801 			break;
1802 	}
1803 	rcu_read_unlock();
1804 	return err;
1805 }
1806 EXPORT_SYMBOL(km_new_mapping);
1807 
1808 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1809 {
1810 	struct km_event c;
1811 
1812 	c.data.hard = hard;
1813 	c.portid = portid;
1814 	c.event = XFRM_MSG_POLEXPIRE;
1815 	km_policy_notify(pol, dir, &c);
1816 }
1817 EXPORT_SYMBOL(km_policy_expired);
1818 
1819 #ifdef CONFIG_XFRM_MIGRATE
1820 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1821 	       const struct xfrm_migrate *m, int num_migrate,
1822 	       const struct xfrm_kmaddress *k)
1823 {
1824 	int err = -EINVAL;
1825 	int ret;
1826 	struct xfrm_mgr *km;
1827 
1828 	rcu_read_lock();
1829 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1830 		if (km->migrate) {
1831 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1832 			if (!ret)
1833 				err = ret;
1834 		}
1835 	}
1836 	rcu_read_unlock();
1837 	return err;
1838 }
1839 EXPORT_SYMBOL(km_migrate);
1840 #endif
1841 
1842 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1843 {
1844 	int err = -EINVAL;
1845 	int ret;
1846 	struct xfrm_mgr *km;
1847 
1848 	rcu_read_lock();
1849 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1850 		if (km->report) {
1851 			ret = km->report(net, proto, sel, addr);
1852 			if (!ret)
1853 				err = ret;
1854 		}
1855 	}
1856 	rcu_read_unlock();
1857 	return err;
1858 }
1859 EXPORT_SYMBOL(km_report);
1860 
1861 bool km_is_alive(const struct km_event *c)
1862 {
1863 	struct xfrm_mgr *km;
1864 	bool is_alive = false;
1865 
1866 	rcu_read_lock();
1867 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1868 		if (km->is_alive && km->is_alive(c)) {
1869 			is_alive = true;
1870 			break;
1871 		}
1872 	}
1873 	rcu_read_unlock();
1874 
1875 	return is_alive;
1876 }
1877 EXPORT_SYMBOL(km_is_alive);
1878 
1879 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1880 {
1881 	int err;
1882 	u8 *data;
1883 	struct xfrm_mgr *km;
1884 	struct xfrm_policy *pol = NULL;
1885 
1886 	if (optlen <= 0 || optlen > PAGE_SIZE)
1887 		return -EMSGSIZE;
1888 
1889 	data = kmalloc(optlen, GFP_KERNEL);
1890 	if (!data)
1891 		return -ENOMEM;
1892 
1893 	err = -EFAULT;
1894 	if (copy_from_user(data, optval, optlen))
1895 		goto out;
1896 
1897 	err = -EINVAL;
1898 	rcu_read_lock();
1899 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1900 		pol = km->compile_policy(sk, optname, data,
1901 					 optlen, &err);
1902 		if (err >= 0)
1903 			break;
1904 	}
1905 	rcu_read_unlock();
1906 
1907 	if (err >= 0) {
1908 		xfrm_sk_policy_insert(sk, err, pol);
1909 		xfrm_pol_put(pol);
1910 		err = 0;
1911 	}
1912 
1913 out:
1914 	kfree(data);
1915 	return err;
1916 }
1917 EXPORT_SYMBOL(xfrm_user_policy);
1918 
1919 static DEFINE_SPINLOCK(xfrm_km_lock);
1920 
1921 int xfrm_register_km(struct xfrm_mgr *km)
1922 {
1923 	spin_lock_bh(&xfrm_km_lock);
1924 	list_add_tail_rcu(&km->list, &xfrm_km_list);
1925 	spin_unlock_bh(&xfrm_km_lock);
1926 	return 0;
1927 }
1928 EXPORT_SYMBOL(xfrm_register_km);
1929 
1930 int xfrm_unregister_km(struct xfrm_mgr *km)
1931 {
1932 	spin_lock_bh(&xfrm_km_lock);
1933 	list_del_rcu(&km->list);
1934 	spin_unlock_bh(&xfrm_km_lock);
1935 	synchronize_rcu();
1936 	return 0;
1937 }
1938 EXPORT_SYMBOL(xfrm_unregister_km);
1939 
1940 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1941 {
1942 	int err = 0;
1943 	if (unlikely(afinfo == NULL))
1944 		return -EINVAL;
1945 	if (unlikely(afinfo->family >= NPROTO))
1946 		return -EAFNOSUPPORT;
1947 	spin_lock_bh(&xfrm_state_afinfo_lock);
1948 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1949 		err = -EEXIST;
1950 	else
1951 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1952 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1953 	return err;
1954 }
1955 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1956 
1957 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1958 {
1959 	int err = 0;
1960 	if (unlikely(afinfo == NULL))
1961 		return -EINVAL;
1962 	if (unlikely(afinfo->family >= NPROTO))
1963 		return -EAFNOSUPPORT;
1964 	spin_lock_bh(&xfrm_state_afinfo_lock);
1965 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1966 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1967 			err = -EINVAL;
1968 		else
1969 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1970 	}
1971 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1972 	synchronize_rcu();
1973 	return err;
1974 }
1975 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1976 
1977 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1978 {
1979 	struct xfrm_state_afinfo *afinfo;
1980 	if (unlikely(family >= NPROTO))
1981 		return NULL;
1982 	rcu_read_lock();
1983 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1984 	if (unlikely(!afinfo))
1985 		rcu_read_unlock();
1986 	return afinfo;
1987 }
1988 
1989 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1990 {
1991 	rcu_read_unlock();
1992 }
1993 
1994 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1995 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1996 {
1997 	if (x->tunnel) {
1998 		struct xfrm_state *t = x->tunnel;
1999 
2000 		if (atomic_read(&t->tunnel_users) == 2)
2001 			xfrm_state_delete(t);
2002 		atomic_dec(&t->tunnel_users);
2003 		xfrm_state_put(t);
2004 		x->tunnel = NULL;
2005 	}
2006 }
2007 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2008 
2009 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2010 {
2011 	int res;
2012 
2013 	spin_lock_bh(&x->lock);
2014 	if (x->km.state == XFRM_STATE_VALID &&
2015 	    x->type && x->type->get_mtu)
2016 		res = x->type->get_mtu(x, mtu);
2017 	else
2018 		res = mtu - x->props.header_len;
2019 	spin_unlock_bh(&x->lock);
2020 	return res;
2021 }
2022 
2023 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
2024 {
2025 	struct xfrm_state_afinfo *afinfo;
2026 	struct xfrm_mode *inner_mode;
2027 	int family = x->props.family;
2028 	int err;
2029 
2030 	err = -EAFNOSUPPORT;
2031 	afinfo = xfrm_state_get_afinfo(family);
2032 	if (!afinfo)
2033 		goto error;
2034 
2035 	err = 0;
2036 	if (afinfo->init_flags)
2037 		err = afinfo->init_flags(x);
2038 
2039 	xfrm_state_put_afinfo(afinfo);
2040 
2041 	if (err)
2042 		goto error;
2043 
2044 	err = -EPROTONOSUPPORT;
2045 
2046 	if (x->sel.family != AF_UNSPEC) {
2047 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2048 		if (inner_mode == NULL)
2049 			goto error;
2050 
2051 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2052 		    family != x->sel.family) {
2053 			xfrm_put_mode(inner_mode);
2054 			goto error;
2055 		}
2056 
2057 		x->inner_mode = inner_mode;
2058 	} else {
2059 		struct xfrm_mode *inner_mode_iaf;
2060 		int iafamily = AF_INET;
2061 
2062 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2063 		if (inner_mode == NULL)
2064 			goto error;
2065 
2066 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2067 			xfrm_put_mode(inner_mode);
2068 			goto error;
2069 		}
2070 		x->inner_mode = inner_mode;
2071 
2072 		if (x->props.family == AF_INET)
2073 			iafamily = AF_INET6;
2074 
2075 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2076 		if (inner_mode_iaf) {
2077 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2078 				x->inner_mode_iaf = inner_mode_iaf;
2079 			else
2080 				xfrm_put_mode(inner_mode_iaf);
2081 		}
2082 	}
2083 
2084 	x->type = xfrm_get_type(x->id.proto, family);
2085 	if (x->type == NULL)
2086 		goto error;
2087 
2088 	err = x->type->init_state(x);
2089 	if (err)
2090 		goto error;
2091 
2092 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2093 	if (x->outer_mode == NULL) {
2094 		err = -EPROTONOSUPPORT;
2095 		goto error;
2096 	}
2097 
2098 	if (init_replay) {
2099 		err = xfrm_init_replay(x);
2100 		if (err)
2101 			goto error;
2102 	}
2103 
2104 	x->km.state = XFRM_STATE_VALID;
2105 
2106 error:
2107 	return err;
2108 }
2109 
2110 EXPORT_SYMBOL(__xfrm_init_state);
2111 
2112 int xfrm_init_state(struct xfrm_state *x)
2113 {
2114 	return __xfrm_init_state(x, true);
2115 }
2116 
2117 EXPORT_SYMBOL(xfrm_init_state);
2118 
2119 int __net_init xfrm_state_init(struct net *net)
2120 {
2121 	unsigned int sz;
2122 
2123 	INIT_LIST_HEAD(&net->xfrm.state_all);
2124 
2125 	sz = sizeof(struct hlist_head) * 8;
2126 
2127 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2128 	if (!net->xfrm.state_bydst)
2129 		goto out_bydst;
2130 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2131 	if (!net->xfrm.state_bysrc)
2132 		goto out_bysrc;
2133 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2134 	if (!net->xfrm.state_byspi)
2135 		goto out_byspi;
2136 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2137 
2138 	net->xfrm.state_num = 0;
2139 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2140 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2141 	return 0;
2142 
2143 out_byspi:
2144 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2145 out_bysrc:
2146 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2147 out_bydst:
2148 	return -ENOMEM;
2149 }
2150 
2151 void xfrm_state_fini(struct net *net)
2152 {
2153 	unsigned int sz;
2154 
2155 	flush_work(&net->xfrm.state_hash_work);
2156 	xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2157 	flush_work(&xfrm_state_gc_work);
2158 
2159 	WARN_ON(!list_empty(&net->xfrm.state_all));
2160 
2161 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2162 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2163 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2164 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2165 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2166 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2167 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2168 }
2169 
2170 #ifdef CONFIG_AUDITSYSCALL
2171 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2172 				     struct audit_buffer *audit_buf)
2173 {
2174 	struct xfrm_sec_ctx *ctx = x->security;
2175 	u32 spi = ntohl(x->id.spi);
2176 
2177 	if (ctx)
2178 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2179 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2180 
2181 	switch (x->props.family) {
2182 	case AF_INET:
2183 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2184 				 &x->props.saddr.a4, &x->id.daddr.a4);
2185 		break;
2186 	case AF_INET6:
2187 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2188 				 x->props.saddr.a6, x->id.daddr.a6);
2189 		break;
2190 	}
2191 
2192 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2193 }
2194 
2195 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2196 				      struct audit_buffer *audit_buf)
2197 {
2198 	const struct iphdr *iph4;
2199 	const struct ipv6hdr *iph6;
2200 
2201 	switch (family) {
2202 	case AF_INET:
2203 		iph4 = ip_hdr(skb);
2204 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2205 				 &iph4->saddr, &iph4->daddr);
2206 		break;
2207 	case AF_INET6:
2208 		iph6 = ipv6_hdr(skb);
2209 		audit_log_format(audit_buf,
2210 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2211 				 &iph6->saddr, &iph6->daddr,
2212 				 iph6->flow_lbl[0] & 0x0f,
2213 				 iph6->flow_lbl[1],
2214 				 iph6->flow_lbl[2]);
2215 		break;
2216 	}
2217 }
2218 
2219 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2220 {
2221 	struct audit_buffer *audit_buf;
2222 
2223 	audit_buf = xfrm_audit_start("SAD-add");
2224 	if (audit_buf == NULL)
2225 		return;
2226 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2227 	xfrm_audit_helper_sainfo(x, audit_buf);
2228 	audit_log_format(audit_buf, " res=%u", result);
2229 	audit_log_end(audit_buf);
2230 }
2231 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2232 
2233 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2234 {
2235 	struct audit_buffer *audit_buf;
2236 
2237 	audit_buf = xfrm_audit_start("SAD-delete");
2238 	if (audit_buf == NULL)
2239 		return;
2240 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2241 	xfrm_audit_helper_sainfo(x, audit_buf);
2242 	audit_log_format(audit_buf, " res=%u", result);
2243 	audit_log_end(audit_buf);
2244 }
2245 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2246 
2247 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2248 				      struct sk_buff *skb)
2249 {
2250 	struct audit_buffer *audit_buf;
2251 	u32 spi;
2252 
2253 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2254 	if (audit_buf == NULL)
2255 		return;
2256 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2257 	/* don't record the sequence number because it's inherent in this kind
2258 	 * of audit message */
2259 	spi = ntohl(x->id.spi);
2260 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2261 	audit_log_end(audit_buf);
2262 }
2263 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2264 
2265 void xfrm_audit_state_replay(struct xfrm_state *x,
2266 			     struct sk_buff *skb, __be32 net_seq)
2267 {
2268 	struct audit_buffer *audit_buf;
2269 	u32 spi;
2270 
2271 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2272 	if (audit_buf == NULL)
2273 		return;
2274 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2275 	spi = ntohl(x->id.spi);
2276 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2277 			 spi, spi, ntohl(net_seq));
2278 	audit_log_end(audit_buf);
2279 }
2280 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2281 
2282 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2283 {
2284 	struct audit_buffer *audit_buf;
2285 
2286 	audit_buf = xfrm_audit_start("SA-notfound");
2287 	if (audit_buf == NULL)
2288 		return;
2289 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2290 	audit_log_end(audit_buf);
2291 }
2292 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2293 
2294 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2295 			       __be32 net_spi, __be32 net_seq)
2296 {
2297 	struct audit_buffer *audit_buf;
2298 	u32 spi;
2299 
2300 	audit_buf = xfrm_audit_start("SA-notfound");
2301 	if (audit_buf == NULL)
2302 		return;
2303 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2304 	spi = ntohl(net_spi);
2305 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2306 			 spi, spi, ntohl(net_seq));
2307 	audit_log_end(audit_buf);
2308 }
2309 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2310 
2311 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2312 			      struct sk_buff *skb, u8 proto)
2313 {
2314 	struct audit_buffer *audit_buf;
2315 	__be32 net_spi;
2316 	__be32 net_seq;
2317 
2318 	audit_buf = xfrm_audit_start("SA-icv-failure");
2319 	if (audit_buf == NULL)
2320 		return;
2321 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2322 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2323 		u32 spi = ntohl(net_spi);
2324 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2325 				 spi, spi, ntohl(net_seq));
2326 	}
2327 	audit_log_end(audit_buf);
2328 }
2329 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2330 #endif /* CONFIG_AUDITSYSCALL */
2331