xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision cff4fa84)
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 /* Each xfrm_state may be linked to two tables:
32 
33    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
34    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
35       destination/tunnel endpoint. (output)
36  */
37 
38 static DEFINE_SPINLOCK(xfrm_state_lock);
39 
40 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
41 
42 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
44 
45 static inline unsigned int xfrm_dst_hash(struct net *net,
46 					 const xfrm_address_t *daddr,
47 					 const xfrm_address_t *saddr,
48 					 u32 reqid,
49 					 unsigned short family)
50 {
51 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
52 }
53 
54 static inline unsigned int xfrm_src_hash(struct net *net,
55 					 const xfrm_address_t *daddr,
56 					 const xfrm_address_t *saddr,
57 					 unsigned short family)
58 {
59 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
60 }
61 
62 static inline unsigned int
63 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
64 	      __be32 spi, u8 proto, unsigned short family)
65 {
66 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
67 }
68 
69 static void xfrm_hash_transfer(struct hlist_head *list,
70 			       struct hlist_head *ndsttable,
71 			       struct hlist_head *nsrctable,
72 			       struct hlist_head *nspitable,
73 			       unsigned int nhashmask)
74 {
75 	struct hlist_node *entry, *tmp;
76 	struct xfrm_state *x;
77 
78 	hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
79 		unsigned int h;
80 
81 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
82 				    x->props.reqid, x->props.family,
83 				    nhashmask);
84 		hlist_add_head(&x->bydst, ndsttable+h);
85 
86 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
87 				    x->props.family,
88 				    nhashmask);
89 		hlist_add_head(&x->bysrc, nsrctable+h);
90 
91 		if (x->id.spi) {
92 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
93 					    x->id.proto, x->props.family,
94 					    nhashmask);
95 			hlist_add_head(&x->byspi, nspitable+h);
96 		}
97 	}
98 }
99 
100 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
101 {
102 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
103 }
104 
105 static DEFINE_MUTEX(hash_resize_mutex);
106 
107 static void xfrm_hash_resize(struct work_struct *work)
108 {
109 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
110 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
111 	unsigned long nsize, osize;
112 	unsigned int nhashmask, ohashmask;
113 	int i;
114 
115 	mutex_lock(&hash_resize_mutex);
116 
117 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
118 	ndst = xfrm_hash_alloc(nsize);
119 	if (!ndst)
120 		goto out_unlock;
121 	nsrc = xfrm_hash_alloc(nsize);
122 	if (!nsrc) {
123 		xfrm_hash_free(ndst, nsize);
124 		goto out_unlock;
125 	}
126 	nspi = xfrm_hash_alloc(nsize);
127 	if (!nspi) {
128 		xfrm_hash_free(ndst, nsize);
129 		xfrm_hash_free(nsrc, nsize);
130 		goto out_unlock;
131 	}
132 
133 	spin_lock_bh(&xfrm_state_lock);
134 
135 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
136 	for (i = net->xfrm.state_hmask; i >= 0; i--)
137 		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
138 				   nhashmask);
139 
140 	odst = net->xfrm.state_bydst;
141 	osrc = net->xfrm.state_bysrc;
142 	ospi = net->xfrm.state_byspi;
143 	ohashmask = net->xfrm.state_hmask;
144 
145 	net->xfrm.state_bydst = ndst;
146 	net->xfrm.state_bysrc = nsrc;
147 	net->xfrm.state_byspi = nspi;
148 	net->xfrm.state_hmask = nhashmask;
149 
150 	spin_unlock_bh(&xfrm_state_lock);
151 
152 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
153 	xfrm_hash_free(odst, osize);
154 	xfrm_hash_free(osrc, osize);
155 	xfrm_hash_free(ospi, osize);
156 
157 out_unlock:
158 	mutex_unlock(&hash_resize_mutex);
159 }
160 
161 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
162 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
163 
164 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
165 
166 int __xfrm_state_delete(struct xfrm_state *x);
167 
168 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
169 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
170 
171 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
172 {
173 	struct xfrm_state_afinfo *afinfo;
174 	if (unlikely(family >= NPROTO))
175 		return NULL;
176 	write_lock_bh(&xfrm_state_afinfo_lock);
177 	afinfo = xfrm_state_afinfo[family];
178 	if (unlikely(!afinfo))
179 		write_unlock_bh(&xfrm_state_afinfo_lock);
180 	return afinfo;
181 }
182 
183 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
184 	__releases(xfrm_state_afinfo_lock)
185 {
186 	write_unlock_bh(&xfrm_state_afinfo_lock);
187 }
188 
189 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
190 {
191 	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
192 	const struct xfrm_type **typemap;
193 	int err = 0;
194 
195 	if (unlikely(afinfo == NULL))
196 		return -EAFNOSUPPORT;
197 	typemap = afinfo->type_map;
198 
199 	if (likely(typemap[type->proto] == NULL))
200 		typemap[type->proto] = type;
201 	else
202 		err = -EEXIST;
203 	xfrm_state_unlock_afinfo(afinfo);
204 	return err;
205 }
206 EXPORT_SYMBOL(xfrm_register_type);
207 
208 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
209 {
210 	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
211 	const struct xfrm_type **typemap;
212 	int err = 0;
213 
214 	if (unlikely(afinfo == NULL))
215 		return -EAFNOSUPPORT;
216 	typemap = afinfo->type_map;
217 
218 	if (unlikely(typemap[type->proto] != type))
219 		err = -ENOENT;
220 	else
221 		typemap[type->proto] = NULL;
222 	xfrm_state_unlock_afinfo(afinfo);
223 	return err;
224 }
225 EXPORT_SYMBOL(xfrm_unregister_type);
226 
227 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
228 {
229 	struct xfrm_state_afinfo *afinfo;
230 	const struct xfrm_type **typemap;
231 	const struct xfrm_type *type;
232 	int modload_attempted = 0;
233 
234 retry:
235 	afinfo = xfrm_state_get_afinfo(family);
236 	if (unlikely(afinfo == NULL))
237 		return NULL;
238 	typemap = afinfo->type_map;
239 
240 	type = typemap[proto];
241 	if (unlikely(type && !try_module_get(type->owner)))
242 		type = NULL;
243 	if (!type && !modload_attempted) {
244 		xfrm_state_put_afinfo(afinfo);
245 		request_module("xfrm-type-%d-%d", family, proto);
246 		modload_attempted = 1;
247 		goto retry;
248 	}
249 
250 	xfrm_state_put_afinfo(afinfo);
251 	return type;
252 }
253 
254 static void xfrm_put_type(const struct xfrm_type *type)
255 {
256 	module_put(type->owner);
257 }
258 
259 int xfrm_register_mode(struct xfrm_mode *mode, int family)
260 {
261 	struct xfrm_state_afinfo *afinfo;
262 	struct xfrm_mode **modemap;
263 	int err;
264 
265 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
266 		return -EINVAL;
267 
268 	afinfo = xfrm_state_lock_afinfo(family);
269 	if (unlikely(afinfo == NULL))
270 		return -EAFNOSUPPORT;
271 
272 	err = -EEXIST;
273 	modemap = afinfo->mode_map;
274 	if (modemap[mode->encap])
275 		goto out;
276 
277 	err = -ENOENT;
278 	if (!try_module_get(afinfo->owner))
279 		goto out;
280 
281 	mode->afinfo = afinfo;
282 	modemap[mode->encap] = mode;
283 	err = 0;
284 
285 out:
286 	xfrm_state_unlock_afinfo(afinfo);
287 	return err;
288 }
289 EXPORT_SYMBOL(xfrm_register_mode);
290 
291 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
292 {
293 	struct xfrm_state_afinfo *afinfo;
294 	struct xfrm_mode **modemap;
295 	int err;
296 
297 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
298 		return -EINVAL;
299 
300 	afinfo = xfrm_state_lock_afinfo(family);
301 	if (unlikely(afinfo == NULL))
302 		return -EAFNOSUPPORT;
303 
304 	err = -ENOENT;
305 	modemap = afinfo->mode_map;
306 	if (likely(modemap[mode->encap] == mode)) {
307 		modemap[mode->encap] = NULL;
308 		module_put(mode->afinfo->owner);
309 		err = 0;
310 	}
311 
312 	xfrm_state_unlock_afinfo(afinfo);
313 	return err;
314 }
315 EXPORT_SYMBOL(xfrm_unregister_mode);
316 
317 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
318 {
319 	struct xfrm_state_afinfo *afinfo;
320 	struct xfrm_mode *mode;
321 	int modload_attempted = 0;
322 
323 	if (unlikely(encap >= XFRM_MODE_MAX))
324 		return NULL;
325 
326 retry:
327 	afinfo = xfrm_state_get_afinfo(family);
328 	if (unlikely(afinfo == NULL))
329 		return NULL;
330 
331 	mode = afinfo->mode_map[encap];
332 	if (unlikely(mode && !try_module_get(mode->owner)))
333 		mode = NULL;
334 	if (!mode && !modload_attempted) {
335 		xfrm_state_put_afinfo(afinfo);
336 		request_module("xfrm-mode-%d-%d", family, encap);
337 		modload_attempted = 1;
338 		goto retry;
339 	}
340 
341 	xfrm_state_put_afinfo(afinfo);
342 	return mode;
343 }
344 
345 static void xfrm_put_mode(struct xfrm_mode *mode)
346 {
347 	module_put(mode->owner);
348 }
349 
350 static void xfrm_state_gc_destroy(struct xfrm_state *x)
351 {
352 	tasklet_hrtimer_cancel(&x->mtimer);
353 	del_timer_sync(&x->rtimer);
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 net *net = container_of(work, struct net, xfrm.state_gc_work);
378 	struct xfrm_state *x;
379 	struct hlist_node *entry, *tmp;
380 	struct hlist_head gc_list;
381 
382 	spin_lock_bh(&xfrm_state_gc_lock);
383 	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
384 	spin_unlock_bh(&xfrm_state_gc_lock);
385 
386 	hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
387 		xfrm_state_gc_destroy(x);
388 
389 	wake_up(&net->xfrm.km_waitq);
390 }
391 
392 static inline unsigned long make_jiffies(long secs)
393 {
394 	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
395 		return MAX_SCHEDULE_TIMEOUT-1;
396 	else
397 		return secs*HZ;
398 }
399 
400 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
401 {
402 	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
403 	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
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 		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
456 	}
457 
458 	goto out;
459 
460 expired:
461 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
462 		x->km.state = XFRM_STATE_EXPIRED;
463 		wake_up(&net->xfrm.km_waitq);
464 		next = 2;
465 		goto resched;
466 	}
467 
468 	err = __xfrm_state_delete(x);
469 	if (!err && x->id.spi)
470 		km_state_expired(x, 1, 0);
471 
472 	xfrm_audit_state_delete(x, err ? 0 : 1,
473 				audit_get_loginuid(current),
474 				audit_get_sessionid(current), 0);
475 
476 out:
477 	spin_unlock(&x->lock);
478 	return HRTIMER_NORESTART;
479 }
480 
481 static void xfrm_replay_timer_handler(unsigned long data);
482 
483 struct xfrm_state *xfrm_state_alloc(struct net *net)
484 {
485 	struct xfrm_state *x;
486 
487 	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
488 
489 	if (x) {
490 		write_pnet(&x->xs_net, net);
491 		atomic_set(&x->refcnt, 1);
492 		atomic_set(&x->tunnel_users, 0);
493 		INIT_LIST_HEAD(&x->km.all);
494 		INIT_HLIST_NODE(&x->bydst);
495 		INIT_HLIST_NODE(&x->bysrc);
496 		INIT_HLIST_NODE(&x->byspi);
497 		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
498 		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
499 				(unsigned long)x);
500 		x->curlft.add_time = get_seconds();
501 		x->lft.soft_byte_limit = XFRM_INF;
502 		x->lft.soft_packet_limit = XFRM_INF;
503 		x->lft.hard_byte_limit = XFRM_INF;
504 		x->lft.hard_packet_limit = XFRM_INF;
505 		x->replay_maxage = 0;
506 		x->replay_maxdiff = 0;
507 		x->inner_mode = NULL;
508 		x->inner_mode_iaf = NULL;
509 		spin_lock_init(&x->lock);
510 	}
511 	return x;
512 }
513 EXPORT_SYMBOL(xfrm_state_alloc);
514 
515 void __xfrm_state_destroy(struct xfrm_state *x)
516 {
517 	struct net *net = xs_net(x);
518 
519 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
520 
521 	spin_lock_bh(&xfrm_state_gc_lock);
522 	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
523 	spin_unlock_bh(&xfrm_state_gc_lock);
524 	schedule_work(&net->xfrm.state_gc_work);
525 }
526 EXPORT_SYMBOL(__xfrm_state_destroy);
527 
528 int __xfrm_state_delete(struct xfrm_state *x)
529 {
530 	struct net *net = xs_net(x);
531 	int err = -ESRCH;
532 
533 	if (x->km.state != XFRM_STATE_DEAD) {
534 		x->km.state = XFRM_STATE_DEAD;
535 		spin_lock(&xfrm_state_lock);
536 		list_del(&x->km.all);
537 		hlist_del(&x->bydst);
538 		hlist_del(&x->bysrc);
539 		if (x->id.spi)
540 			hlist_del(&x->byspi);
541 		net->xfrm.state_num--;
542 		spin_unlock(&xfrm_state_lock);
543 
544 		/* All xfrm_state objects are created by xfrm_state_alloc.
545 		 * The xfrm_state_alloc call gives a reference, and that
546 		 * is what we are dropping here.
547 		 */
548 		xfrm_state_put(x);
549 		err = 0;
550 	}
551 
552 	return err;
553 }
554 EXPORT_SYMBOL(__xfrm_state_delete);
555 
556 int xfrm_state_delete(struct xfrm_state *x)
557 {
558 	int err;
559 
560 	spin_lock_bh(&x->lock);
561 	err = __xfrm_state_delete(x);
562 	spin_unlock_bh(&x->lock);
563 
564 	return err;
565 }
566 EXPORT_SYMBOL(xfrm_state_delete);
567 
568 #ifdef CONFIG_SECURITY_NETWORK_XFRM
569 static inline int
570 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
571 {
572 	int i, err = 0;
573 
574 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
575 		struct hlist_node *entry;
576 		struct xfrm_state *x;
577 
578 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
579 			if (xfrm_id_proto_match(x->id.proto, proto) &&
580 			   (err = security_xfrm_state_delete(x)) != 0) {
581 				xfrm_audit_state_delete(x, 0,
582 							audit_info->loginuid,
583 							audit_info->sessionid,
584 							audit_info->secid);
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, struct xfrm_audit *audit_info)
595 {
596 	return 0;
597 }
598 #endif
599 
600 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
601 {
602 	int i, err = 0, cnt = 0;
603 
604 	spin_lock_bh(&xfrm_state_lock);
605 	err = xfrm_state_flush_secctx_check(net, proto, audit_info);
606 	if (err)
607 		goto out;
608 
609 	err = -ESRCH;
610 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
611 		struct hlist_node *entry;
612 		struct xfrm_state *x;
613 restart:
614 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
615 			if (!xfrm_state_kern(x) &&
616 			    xfrm_id_proto_match(x->id.proto, proto)) {
617 				xfrm_state_hold(x);
618 				spin_unlock_bh(&xfrm_state_lock);
619 
620 				err = xfrm_state_delete(x);
621 				xfrm_audit_state_delete(x, err ? 0 : 1,
622 							audit_info->loginuid,
623 							audit_info->sessionid,
624 							audit_info->secid);
625 				xfrm_state_put(x);
626 				if (!err)
627 					cnt++;
628 
629 				spin_lock_bh(&xfrm_state_lock);
630 				goto restart;
631 			}
632 		}
633 	}
634 	if (cnt)
635 		err = 0;
636 
637 out:
638 	spin_unlock_bh(&xfrm_state_lock);
639 	wake_up(&net->xfrm.km_waitq);
640 	return err;
641 }
642 EXPORT_SYMBOL(xfrm_state_flush);
643 
644 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
645 {
646 	spin_lock_bh(&xfrm_state_lock);
647 	si->sadcnt = net->xfrm.state_num;
648 	si->sadhcnt = net->xfrm.state_hmask;
649 	si->sadhmcnt = xfrm_state_hashmax;
650 	spin_unlock_bh(&xfrm_state_lock);
651 }
652 EXPORT_SYMBOL(xfrm_sad_getinfo);
653 
654 static int
655 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
656 		    const struct xfrm_tmpl *tmpl,
657 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
658 		    unsigned short family)
659 {
660 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
661 	if (!afinfo)
662 		return -1;
663 	afinfo->init_tempsel(&x->sel, fl);
664 
665 	if (family != tmpl->encap_family) {
666 		xfrm_state_put_afinfo(afinfo);
667 		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
668 		if (!afinfo)
669 			return -1;
670 	}
671 	afinfo->init_temprop(x, tmpl, daddr, saddr);
672 	xfrm_state_put_afinfo(afinfo);
673 	return 0;
674 }
675 
676 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
677 					      const xfrm_address_t *daddr,
678 					      __be32 spi, u8 proto,
679 					      unsigned short family)
680 {
681 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
682 	struct xfrm_state *x;
683 	struct hlist_node *entry;
684 
685 	hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
686 		if (x->props.family != family ||
687 		    x->id.spi       != spi ||
688 		    x->id.proto     != proto ||
689 		    xfrm_addr_cmp(&x->id.daddr, daddr, family))
690 			continue;
691 
692 		if ((mark & x->mark.m) != x->mark.v)
693 			continue;
694 		xfrm_state_hold(x);
695 		return x;
696 	}
697 
698 	return NULL;
699 }
700 
701 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
702 						     const xfrm_address_t *daddr,
703 						     const xfrm_address_t *saddr,
704 						     u8 proto, unsigned short family)
705 {
706 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
707 	struct xfrm_state *x;
708 	struct hlist_node *entry;
709 
710 	hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
711 		if (x->props.family != family ||
712 		    x->id.proto     != proto ||
713 		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
714 		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
715 			continue;
716 
717 		if ((mark & x->mark.m) != x->mark.v)
718 			continue;
719 		xfrm_state_hold(x);
720 		return x;
721 	}
722 
723 	return NULL;
724 }
725 
726 static inline struct xfrm_state *
727 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
728 {
729 	struct net *net = xs_net(x);
730 	u32 mark = x->mark.v & x->mark.m;
731 
732 	if (use_spi)
733 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
734 					   x->id.spi, x->id.proto, family);
735 	else
736 		return __xfrm_state_lookup_byaddr(net, mark,
737 						  &x->id.daddr,
738 						  &x->props.saddr,
739 						  x->id.proto, family);
740 }
741 
742 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
743 {
744 	if (have_hash_collision &&
745 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
746 	    net->xfrm.state_num > net->xfrm.state_hmask)
747 		schedule_work(&net->xfrm.state_hash_work);
748 }
749 
750 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
751 			       const struct flowi *fl, unsigned short family,
752 			       struct xfrm_state **best, int *acq_in_progress,
753 			       int *error)
754 {
755 	/* Resolution logic:
756 	 * 1. There is a valid state with matching selector. Done.
757 	 * 2. Valid state with inappropriate selector. Skip.
758 	 *
759 	 * Entering area of "sysdeps".
760 	 *
761 	 * 3. If state is not valid, selector is temporary, it selects
762 	 *    only session which triggered previous resolution. Key
763 	 *    manager will do something to install a state with proper
764 	 *    selector.
765 	 */
766 	if (x->km.state == XFRM_STATE_VALID) {
767 		if ((x->sel.family &&
768 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
769 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
770 			return;
771 
772 		if (!*best ||
773 		    (*best)->km.dying > x->km.dying ||
774 		    ((*best)->km.dying == x->km.dying &&
775 		     (*best)->curlft.add_time < x->curlft.add_time))
776 			*best = x;
777 	} else if (x->km.state == XFRM_STATE_ACQ) {
778 		*acq_in_progress = 1;
779 	} else if (x->km.state == XFRM_STATE_ERROR ||
780 		   x->km.state == XFRM_STATE_EXPIRED) {
781 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
782 		    security_xfrm_state_pol_flow_match(x, pol, fl))
783 			*error = -ESRCH;
784 	}
785 }
786 
787 struct xfrm_state *
788 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
789 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
790 		struct xfrm_policy *pol, int *err,
791 		unsigned short family)
792 {
793 	static xfrm_address_t saddr_wildcard = { };
794 	struct net *net = xp_net(pol);
795 	unsigned int h, h_wildcard;
796 	struct hlist_node *entry;
797 	struct xfrm_state *x, *x0, *to_put;
798 	int acquire_in_progress = 0;
799 	int error = 0;
800 	struct xfrm_state *best = NULL;
801 	u32 mark = pol->mark.v & pol->mark.m;
802 	unsigned short encap_family = tmpl->encap_family;
803 
804 	to_put = NULL;
805 
806 	spin_lock_bh(&xfrm_state_lock);
807 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
808 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
809 		if (x->props.family == encap_family &&
810 		    x->props.reqid == tmpl->reqid &&
811 		    (mark & x->mark.m) == x->mark.v &&
812 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
813 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
814 		    tmpl->mode == x->props.mode &&
815 		    tmpl->id.proto == x->id.proto &&
816 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
817 			xfrm_state_look_at(pol, x, fl, encap_family,
818 					   &best, &acquire_in_progress, &error);
819 	}
820 	if (best)
821 		goto found;
822 
823 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
824 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
825 		if (x->props.family == encap_family &&
826 		    x->props.reqid == tmpl->reqid &&
827 		    (mark & x->mark.m) == x->mark.v &&
828 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
829 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
830 		    tmpl->mode == x->props.mode &&
831 		    tmpl->id.proto == x->id.proto &&
832 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
833 			xfrm_state_look_at(pol, x, fl, encap_family,
834 					   &best, &acquire_in_progress, &error);
835 	}
836 
837 found:
838 	x = best;
839 	if (!x && !error && !acquire_in_progress) {
840 		if (tmpl->id.spi &&
841 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
842 					      tmpl->id.proto, encap_family)) != NULL) {
843 			to_put = x0;
844 			error = -EEXIST;
845 			goto out;
846 		}
847 		x = xfrm_state_alloc(net);
848 		if (x == NULL) {
849 			error = -ENOMEM;
850 			goto out;
851 		}
852 		/* Initialize temporary state matching only
853 		 * to current session. */
854 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
855 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
856 
857 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
858 		if (error) {
859 			x->km.state = XFRM_STATE_DEAD;
860 			to_put = x;
861 			x = NULL;
862 			goto out;
863 		}
864 
865 		if (km_query(x, tmpl, pol) == 0) {
866 			x->km.state = XFRM_STATE_ACQ;
867 			list_add(&x->km.all, &net->xfrm.state_all);
868 			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
869 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
870 			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
871 			if (x->id.spi) {
872 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
873 				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
874 			}
875 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
876 			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
877 			net->xfrm.state_num++;
878 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
879 		} else {
880 			x->km.state = XFRM_STATE_DEAD;
881 			to_put = x;
882 			x = NULL;
883 			error = -ESRCH;
884 		}
885 	}
886 out:
887 	if (x)
888 		xfrm_state_hold(x);
889 	else
890 		*err = acquire_in_progress ? -EAGAIN : error;
891 	spin_unlock_bh(&xfrm_state_lock);
892 	if (to_put)
893 		xfrm_state_put(to_put);
894 	return x;
895 }
896 
897 struct xfrm_state *
898 xfrm_stateonly_find(struct net *net, u32 mark,
899 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
900 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
901 {
902 	unsigned int h;
903 	struct xfrm_state *rx = NULL, *x = NULL;
904 	struct hlist_node *entry;
905 
906 	spin_lock(&xfrm_state_lock);
907 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
908 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
909 		if (x->props.family == family &&
910 		    x->props.reqid == reqid &&
911 		    (mark & x->mark.m) == x->mark.v &&
912 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
913 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
914 		    mode == x->props.mode &&
915 		    proto == x->id.proto &&
916 		    x->km.state == XFRM_STATE_VALID) {
917 			rx = x;
918 			break;
919 		}
920 	}
921 
922 	if (rx)
923 		xfrm_state_hold(rx);
924 	spin_unlock(&xfrm_state_lock);
925 
926 
927 	return rx;
928 }
929 EXPORT_SYMBOL(xfrm_stateonly_find);
930 
931 static void __xfrm_state_insert(struct xfrm_state *x)
932 {
933 	struct net *net = xs_net(x);
934 	unsigned int h;
935 
936 	list_add(&x->km.all, &net->xfrm.state_all);
937 
938 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
939 			  x->props.reqid, x->props.family);
940 	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
941 
942 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
943 	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
944 
945 	if (x->id.spi) {
946 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
947 				  x->props.family);
948 
949 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
950 	}
951 
952 	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
953 	if (x->replay_maxage)
954 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
955 
956 	wake_up(&net->xfrm.km_waitq);
957 
958 	net->xfrm.state_num++;
959 
960 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
961 }
962 
963 /* xfrm_state_lock is held */
964 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
965 {
966 	struct net *net = xs_net(xnew);
967 	unsigned short family = xnew->props.family;
968 	u32 reqid = xnew->props.reqid;
969 	struct xfrm_state *x;
970 	struct hlist_node *entry;
971 	unsigned int h;
972 	u32 mark = xnew->mark.v & xnew->mark.m;
973 
974 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
975 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
976 		if (x->props.family	== family &&
977 		    x->props.reqid	== reqid &&
978 		    (mark & x->mark.m) == x->mark.v &&
979 		    !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
980 		    !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
981 			x->genid++;
982 	}
983 }
984 
985 void xfrm_state_insert(struct xfrm_state *x)
986 {
987 	spin_lock_bh(&xfrm_state_lock);
988 	__xfrm_state_bump_genids(x);
989 	__xfrm_state_insert(x);
990 	spin_unlock_bh(&xfrm_state_lock);
991 }
992 EXPORT_SYMBOL(xfrm_state_insert);
993 
994 /* xfrm_state_lock is held */
995 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
996 					  unsigned short family, u8 mode,
997 					  u32 reqid, u8 proto,
998 					  const xfrm_address_t *daddr,
999 					  const xfrm_address_t *saddr, int create)
1000 {
1001 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1002 	struct hlist_node *entry;
1003 	struct xfrm_state *x;
1004 	u32 mark = m->v & m->m;
1005 
1006 	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1007 		if (x->props.reqid  != reqid ||
1008 		    x->props.mode   != mode ||
1009 		    x->props.family != family ||
1010 		    x->km.state     != XFRM_STATE_ACQ ||
1011 		    x->id.spi       != 0 ||
1012 		    x->id.proto	    != proto ||
1013 		    (mark & x->mark.m) != x->mark.v ||
1014 		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1015 		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
1016 			continue;
1017 
1018 		xfrm_state_hold(x);
1019 		return x;
1020 	}
1021 
1022 	if (!create)
1023 		return NULL;
1024 
1025 	x = xfrm_state_alloc(net);
1026 	if (likely(x)) {
1027 		switch (family) {
1028 		case AF_INET:
1029 			x->sel.daddr.a4 = daddr->a4;
1030 			x->sel.saddr.a4 = saddr->a4;
1031 			x->sel.prefixlen_d = 32;
1032 			x->sel.prefixlen_s = 32;
1033 			x->props.saddr.a4 = saddr->a4;
1034 			x->id.daddr.a4 = daddr->a4;
1035 			break;
1036 
1037 		case AF_INET6:
1038 			ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1039 				       (const struct in6_addr *)daddr);
1040 			ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1041 				       (const struct in6_addr *)saddr);
1042 			x->sel.prefixlen_d = 128;
1043 			x->sel.prefixlen_s = 128;
1044 			ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1045 				       (const struct in6_addr *)saddr);
1046 			ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1047 				       (const struct in6_addr *)daddr);
1048 			break;
1049 		}
1050 
1051 		x->km.state = XFRM_STATE_ACQ;
1052 		x->id.proto = proto;
1053 		x->props.family = family;
1054 		x->props.mode = mode;
1055 		x->props.reqid = reqid;
1056 		x->mark.v = m->v;
1057 		x->mark.m = m->m;
1058 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1059 		xfrm_state_hold(x);
1060 		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1061 		list_add(&x->km.all, &net->xfrm.state_all);
1062 		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1063 		h = xfrm_src_hash(net, daddr, saddr, family);
1064 		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1065 
1066 		net->xfrm.state_num++;
1067 
1068 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1069 	}
1070 
1071 	return x;
1072 }
1073 
1074 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1075 
1076 int xfrm_state_add(struct xfrm_state *x)
1077 {
1078 	struct net *net = xs_net(x);
1079 	struct xfrm_state *x1, *to_put;
1080 	int family;
1081 	int err;
1082 	u32 mark = x->mark.v & x->mark.m;
1083 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1084 
1085 	family = x->props.family;
1086 
1087 	to_put = NULL;
1088 
1089 	spin_lock_bh(&xfrm_state_lock);
1090 
1091 	x1 = __xfrm_state_locate(x, use_spi, family);
1092 	if (x1) {
1093 		to_put = x1;
1094 		x1 = NULL;
1095 		err = -EEXIST;
1096 		goto out;
1097 	}
1098 
1099 	if (use_spi && x->km.seq) {
1100 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1101 		if (x1 && ((x1->id.proto != x->id.proto) ||
1102 		    xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1103 			to_put = x1;
1104 			x1 = NULL;
1105 		}
1106 	}
1107 
1108 	if (use_spi && !x1)
1109 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1110 				     x->props.reqid, x->id.proto,
1111 				     &x->id.daddr, &x->props.saddr, 0);
1112 
1113 	__xfrm_state_bump_genids(x);
1114 	__xfrm_state_insert(x);
1115 	err = 0;
1116 
1117 out:
1118 	spin_unlock_bh(&xfrm_state_lock);
1119 
1120 	if (x1) {
1121 		xfrm_state_delete(x1);
1122 		xfrm_state_put(x1);
1123 	}
1124 
1125 	if (to_put)
1126 		xfrm_state_put(to_put);
1127 
1128 	return err;
1129 }
1130 EXPORT_SYMBOL(xfrm_state_add);
1131 
1132 #ifdef CONFIG_XFRM_MIGRATE
1133 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1134 {
1135 	struct net *net = xs_net(orig);
1136 	int err = -ENOMEM;
1137 	struct xfrm_state *x = xfrm_state_alloc(net);
1138 	if (!x)
1139 		goto out;
1140 
1141 	memcpy(&x->id, &orig->id, sizeof(x->id));
1142 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1143 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1144 	x->props.mode = orig->props.mode;
1145 	x->props.replay_window = orig->props.replay_window;
1146 	x->props.reqid = orig->props.reqid;
1147 	x->props.family = orig->props.family;
1148 	x->props.saddr = orig->props.saddr;
1149 
1150 	if (orig->aalg) {
1151 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1152 		if (!x->aalg)
1153 			goto error;
1154 	}
1155 	x->props.aalgo = orig->props.aalgo;
1156 
1157 	if (orig->ealg) {
1158 		x->ealg = xfrm_algo_clone(orig->ealg);
1159 		if (!x->ealg)
1160 			goto error;
1161 	}
1162 	x->props.ealgo = orig->props.ealgo;
1163 
1164 	if (orig->calg) {
1165 		x->calg = xfrm_algo_clone(orig->calg);
1166 		if (!x->calg)
1167 			goto error;
1168 	}
1169 	x->props.calgo = orig->props.calgo;
1170 
1171 	if (orig->encap) {
1172 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1173 		if (!x->encap)
1174 			goto error;
1175 	}
1176 
1177 	if (orig->coaddr) {
1178 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1179 				    GFP_KERNEL);
1180 		if (!x->coaddr)
1181 			goto error;
1182 	}
1183 
1184 	if (orig->replay_esn) {
1185 		err = xfrm_replay_clone(x, orig);
1186 		if (err)
1187 			goto error;
1188 	}
1189 
1190 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1191 
1192 	err = xfrm_init_state(x);
1193 	if (err)
1194 		goto error;
1195 
1196 	x->props.flags = orig->props.flags;
1197 
1198 	x->curlft.add_time = orig->curlft.add_time;
1199 	x->km.state = orig->km.state;
1200 	x->km.seq = orig->km.seq;
1201 
1202 	return x;
1203 
1204  error:
1205 	xfrm_state_put(x);
1206 out:
1207 	if (errp)
1208 		*errp = err;
1209 	return NULL;
1210 }
1211 
1212 /* xfrm_state_lock is held */
1213 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1214 {
1215 	unsigned int h;
1216 	struct xfrm_state *x;
1217 	struct hlist_node *entry;
1218 
1219 	if (m->reqid) {
1220 		h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1221 				  m->reqid, m->old_family);
1222 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1223 			if (x->props.mode != m->mode ||
1224 			    x->id.proto != m->proto)
1225 				continue;
1226 			if (m->reqid && x->props.reqid != m->reqid)
1227 				continue;
1228 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1229 					  m->old_family) ||
1230 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1231 					  m->old_family))
1232 				continue;
1233 			xfrm_state_hold(x);
1234 			return x;
1235 		}
1236 	} else {
1237 		h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1238 				  m->old_family);
1239 		hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1240 			if (x->props.mode != m->mode ||
1241 			    x->id.proto != m->proto)
1242 				continue;
1243 			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1244 					  m->old_family) ||
1245 			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1246 					  m->old_family))
1247 				continue;
1248 			xfrm_state_hold(x);
1249 			return x;
1250 		}
1251 	}
1252 
1253 	return NULL;
1254 }
1255 EXPORT_SYMBOL(xfrm_migrate_state_find);
1256 
1257 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1258 				       struct xfrm_migrate *m)
1259 {
1260 	struct xfrm_state *xc;
1261 	int err;
1262 
1263 	xc = xfrm_state_clone(x, &err);
1264 	if (!xc)
1265 		return NULL;
1266 
1267 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1268 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1269 
1270 	/* add state */
1271 	if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1272 		/* a care is needed when the destination address of the
1273 		   state is to be updated as it is a part of triplet */
1274 		xfrm_state_insert(xc);
1275 	} else {
1276 		if ((err = xfrm_state_add(xc)) < 0)
1277 			goto error;
1278 	}
1279 
1280 	return xc;
1281 error:
1282 	xfrm_state_put(xc);
1283 	return NULL;
1284 }
1285 EXPORT_SYMBOL(xfrm_state_migrate);
1286 #endif
1287 
1288 int xfrm_state_update(struct xfrm_state *x)
1289 {
1290 	struct xfrm_state *x1, *to_put;
1291 	int err;
1292 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1293 
1294 	to_put = NULL;
1295 
1296 	spin_lock_bh(&xfrm_state_lock);
1297 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1298 
1299 	err = -ESRCH;
1300 	if (!x1)
1301 		goto out;
1302 
1303 	if (xfrm_state_kern(x1)) {
1304 		to_put = x1;
1305 		err = -EEXIST;
1306 		goto out;
1307 	}
1308 
1309 	if (x1->km.state == XFRM_STATE_ACQ) {
1310 		__xfrm_state_insert(x);
1311 		x = NULL;
1312 	}
1313 	err = 0;
1314 
1315 out:
1316 	spin_unlock_bh(&xfrm_state_lock);
1317 
1318 	if (to_put)
1319 		xfrm_state_put(to_put);
1320 
1321 	if (err)
1322 		return err;
1323 
1324 	if (!x) {
1325 		xfrm_state_delete(x1);
1326 		xfrm_state_put(x1);
1327 		return 0;
1328 	}
1329 
1330 	err = -EINVAL;
1331 	spin_lock_bh(&x1->lock);
1332 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1333 		if (x->encap && x1->encap)
1334 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1335 		if (x->coaddr && x1->coaddr) {
1336 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1337 		}
1338 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1339 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1340 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1341 		x1->km.dying = 0;
1342 
1343 		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1344 		if (x1->curlft.use_time)
1345 			xfrm_state_check_expire(x1);
1346 
1347 		err = 0;
1348 		x->km.state = XFRM_STATE_DEAD;
1349 		__xfrm_state_put(x);
1350 	}
1351 	spin_unlock_bh(&x1->lock);
1352 
1353 	xfrm_state_put(x1);
1354 
1355 	return err;
1356 }
1357 EXPORT_SYMBOL(xfrm_state_update);
1358 
1359 int xfrm_state_check_expire(struct xfrm_state *x)
1360 {
1361 	if (!x->curlft.use_time)
1362 		x->curlft.use_time = get_seconds();
1363 
1364 	if (x->km.state != XFRM_STATE_VALID)
1365 		return -EINVAL;
1366 
1367 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1368 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1369 		x->km.state = XFRM_STATE_EXPIRED;
1370 		tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1371 		return -EINVAL;
1372 	}
1373 
1374 	if (!x->km.dying &&
1375 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1376 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1377 		x->km.dying = 1;
1378 		km_state_expired(x, 0, 0);
1379 	}
1380 	return 0;
1381 }
1382 EXPORT_SYMBOL(xfrm_state_check_expire);
1383 
1384 struct xfrm_state *
1385 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1386 		  u8 proto, unsigned short family)
1387 {
1388 	struct xfrm_state *x;
1389 
1390 	spin_lock_bh(&xfrm_state_lock);
1391 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1392 	spin_unlock_bh(&xfrm_state_lock);
1393 	return x;
1394 }
1395 EXPORT_SYMBOL(xfrm_state_lookup);
1396 
1397 struct xfrm_state *
1398 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1399 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1400 			 u8 proto, unsigned short family)
1401 {
1402 	struct xfrm_state *x;
1403 
1404 	spin_lock_bh(&xfrm_state_lock);
1405 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1406 	spin_unlock_bh(&xfrm_state_lock);
1407 	return x;
1408 }
1409 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1410 
1411 struct xfrm_state *
1412 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1413 	      const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1414 	      int create, unsigned short family)
1415 {
1416 	struct xfrm_state *x;
1417 
1418 	spin_lock_bh(&xfrm_state_lock);
1419 	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1420 	spin_unlock_bh(&xfrm_state_lock);
1421 
1422 	return x;
1423 }
1424 EXPORT_SYMBOL(xfrm_find_acq);
1425 
1426 #ifdef CONFIG_XFRM_SUB_POLICY
1427 int
1428 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1429 	       unsigned short family)
1430 {
1431 	int err = 0;
1432 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1433 	if (!afinfo)
1434 		return -EAFNOSUPPORT;
1435 
1436 	spin_lock_bh(&xfrm_state_lock);
1437 	if (afinfo->tmpl_sort)
1438 		err = afinfo->tmpl_sort(dst, src, n);
1439 	spin_unlock_bh(&xfrm_state_lock);
1440 	xfrm_state_put_afinfo(afinfo);
1441 	return err;
1442 }
1443 EXPORT_SYMBOL(xfrm_tmpl_sort);
1444 
1445 int
1446 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1447 		unsigned short family)
1448 {
1449 	int err = 0;
1450 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1451 	if (!afinfo)
1452 		return -EAFNOSUPPORT;
1453 
1454 	spin_lock_bh(&xfrm_state_lock);
1455 	if (afinfo->state_sort)
1456 		err = afinfo->state_sort(dst, src, n);
1457 	spin_unlock_bh(&xfrm_state_lock);
1458 	xfrm_state_put_afinfo(afinfo);
1459 	return err;
1460 }
1461 EXPORT_SYMBOL(xfrm_state_sort);
1462 #endif
1463 
1464 /* Silly enough, but I'm lazy to build resolution list */
1465 
1466 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1467 {
1468 	int i;
1469 
1470 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1471 		struct hlist_node *entry;
1472 		struct xfrm_state *x;
1473 
1474 		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1475 			if (x->km.seq == seq &&
1476 			    (mark & x->mark.m) == x->mark.v &&
1477 			    x->km.state == XFRM_STATE_ACQ) {
1478 				xfrm_state_hold(x);
1479 				return x;
1480 			}
1481 		}
1482 	}
1483 	return NULL;
1484 }
1485 
1486 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1487 {
1488 	struct xfrm_state *x;
1489 
1490 	spin_lock_bh(&xfrm_state_lock);
1491 	x = __xfrm_find_acq_byseq(net, mark, seq);
1492 	spin_unlock_bh(&xfrm_state_lock);
1493 	return x;
1494 }
1495 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1496 
1497 u32 xfrm_get_acqseq(void)
1498 {
1499 	u32 res;
1500 	static atomic_t acqseq;
1501 
1502 	do {
1503 		res = atomic_inc_return(&acqseq);
1504 	} while (!res);
1505 
1506 	return res;
1507 }
1508 EXPORT_SYMBOL(xfrm_get_acqseq);
1509 
1510 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1511 {
1512 	struct net *net = xs_net(x);
1513 	unsigned int h;
1514 	struct xfrm_state *x0;
1515 	int err = -ENOENT;
1516 	__be32 minspi = htonl(low);
1517 	__be32 maxspi = htonl(high);
1518 	u32 mark = x->mark.v & x->mark.m;
1519 
1520 	spin_lock_bh(&x->lock);
1521 	if (x->km.state == XFRM_STATE_DEAD)
1522 		goto unlock;
1523 
1524 	err = 0;
1525 	if (x->id.spi)
1526 		goto unlock;
1527 
1528 	err = -ENOENT;
1529 
1530 	if (minspi == maxspi) {
1531 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1532 		if (x0) {
1533 			xfrm_state_put(x0);
1534 			goto unlock;
1535 		}
1536 		x->id.spi = minspi;
1537 	} else {
1538 		u32 spi = 0;
1539 		for (h=0; h<high-low+1; h++) {
1540 			spi = low + net_random()%(high-low+1);
1541 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1542 			if (x0 == NULL) {
1543 				x->id.spi = htonl(spi);
1544 				break;
1545 			}
1546 			xfrm_state_put(x0);
1547 		}
1548 	}
1549 	if (x->id.spi) {
1550 		spin_lock_bh(&xfrm_state_lock);
1551 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1552 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1553 		spin_unlock_bh(&xfrm_state_lock);
1554 
1555 		err = 0;
1556 	}
1557 
1558 unlock:
1559 	spin_unlock_bh(&x->lock);
1560 
1561 	return err;
1562 }
1563 EXPORT_SYMBOL(xfrm_alloc_spi);
1564 
1565 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1566 		    int (*func)(struct xfrm_state *, int, void*),
1567 		    void *data)
1568 {
1569 	struct xfrm_state *state;
1570 	struct xfrm_state_walk *x;
1571 	int err = 0;
1572 
1573 	if (walk->seq != 0 && list_empty(&walk->all))
1574 		return 0;
1575 
1576 	spin_lock_bh(&xfrm_state_lock);
1577 	if (list_empty(&walk->all))
1578 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1579 	else
1580 		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1581 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1582 		if (x->state == XFRM_STATE_DEAD)
1583 			continue;
1584 		state = container_of(x, struct xfrm_state, km);
1585 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1586 			continue;
1587 		err = func(state, walk->seq, data);
1588 		if (err) {
1589 			list_move_tail(&walk->all, &x->all);
1590 			goto out;
1591 		}
1592 		walk->seq++;
1593 	}
1594 	if (walk->seq == 0) {
1595 		err = -ENOENT;
1596 		goto out;
1597 	}
1598 	list_del_init(&walk->all);
1599 out:
1600 	spin_unlock_bh(&xfrm_state_lock);
1601 	return err;
1602 }
1603 EXPORT_SYMBOL(xfrm_state_walk);
1604 
1605 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1606 {
1607 	INIT_LIST_HEAD(&walk->all);
1608 	walk->proto = proto;
1609 	walk->state = XFRM_STATE_DEAD;
1610 	walk->seq = 0;
1611 }
1612 EXPORT_SYMBOL(xfrm_state_walk_init);
1613 
1614 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1615 {
1616 	if (list_empty(&walk->all))
1617 		return;
1618 
1619 	spin_lock_bh(&xfrm_state_lock);
1620 	list_del(&walk->all);
1621 	spin_unlock_bh(&xfrm_state_lock);
1622 }
1623 EXPORT_SYMBOL(xfrm_state_walk_done);
1624 
1625 static void xfrm_replay_timer_handler(unsigned long data)
1626 {
1627 	struct xfrm_state *x = (struct xfrm_state*)data;
1628 
1629 	spin_lock(&x->lock);
1630 
1631 	if (x->km.state == XFRM_STATE_VALID) {
1632 		if (xfrm_aevent_is_on(xs_net(x)))
1633 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1634 		else
1635 			x->xflags |= XFRM_TIME_DEFER;
1636 	}
1637 
1638 	spin_unlock(&x->lock);
1639 }
1640 
1641 static LIST_HEAD(xfrm_km_list);
1642 static DEFINE_RWLOCK(xfrm_km_lock);
1643 
1644 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1645 {
1646 	struct xfrm_mgr *km;
1647 
1648 	read_lock(&xfrm_km_lock);
1649 	list_for_each_entry(km, &xfrm_km_list, list)
1650 		if (km->notify_policy)
1651 			km->notify_policy(xp, dir, c);
1652 	read_unlock(&xfrm_km_lock);
1653 }
1654 
1655 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1656 {
1657 	struct xfrm_mgr *km;
1658 	read_lock(&xfrm_km_lock);
1659 	list_for_each_entry(km, &xfrm_km_list, list)
1660 		if (km->notify)
1661 			km->notify(x, c);
1662 	read_unlock(&xfrm_km_lock);
1663 }
1664 
1665 EXPORT_SYMBOL(km_policy_notify);
1666 EXPORT_SYMBOL(km_state_notify);
1667 
1668 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1669 {
1670 	struct net *net = xs_net(x);
1671 	struct km_event c;
1672 
1673 	c.data.hard = hard;
1674 	c.pid = pid;
1675 	c.event = XFRM_MSG_EXPIRE;
1676 	km_state_notify(x, &c);
1677 
1678 	if (hard)
1679 		wake_up(&net->xfrm.km_waitq);
1680 }
1681 
1682 EXPORT_SYMBOL(km_state_expired);
1683 /*
1684  * We send to all registered managers regardless of failure
1685  * We are happy with one success
1686 */
1687 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1688 {
1689 	int err = -EINVAL, acqret;
1690 	struct xfrm_mgr *km;
1691 
1692 	read_lock(&xfrm_km_lock);
1693 	list_for_each_entry(km, &xfrm_km_list, list) {
1694 		acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1695 		if (!acqret)
1696 			err = acqret;
1697 	}
1698 	read_unlock(&xfrm_km_lock);
1699 	return err;
1700 }
1701 EXPORT_SYMBOL(km_query);
1702 
1703 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1704 {
1705 	int err = -EINVAL;
1706 	struct xfrm_mgr *km;
1707 
1708 	read_lock(&xfrm_km_lock);
1709 	list_for_each_entry(km, &xfrm_km_list, list) {
1710 		if (km->new_mapping)
1711 			err = km->new_mapping(x, ipaddr, sport);
1712 		if (!err)
1713 			break;
1714 	}
1715 	read_unlock(&xfrm_km_lock);
1716 	return err;
1717 }
1718 EXPORT_SYMBOL(km_new_mapping);
1719 
1720 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1721 {
1722 	struct net *net = xp_net(pol);
1723 	struct km_event c;
1724 
1725 	c.data.hard = hard;
1726 	c.pid = pid;
1727 	c.event = XFRM_MSG_POLEXPIRE;
1728 	km_policy_notify(pol, dir, &c);
1729 
1730 	if (hard)
1731 		wake_up(&net->xfrm.km_waitq);
1732 }
1733 EXPORT_SYMBOL(km_policy_expired);
1734 
1735 #ifdef CONFIG_XFRM_MIGRATE
1736 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1737 	       const struct xfrm_migrate *m, int num_migrate,
1738 	       const struct xfrm_kmaddress *k)
1739 {
1740 	int err = -EINVAL;
1741 	int ret;
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->migrate) {
1747 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1748 			if (!ret)
1749 				err = ret;
1750 		}
1751 	}
1752 	read_unlock(&xfrm_km_lock);
1753 	return err;
1754 }
1755 EXPORT_SYMBOL(km_migrate);
1756 #endif
1757 
1758 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1759 {
1760 	int err = -EINVAL;
1761 	int ret;
1762 	struct xfrm_mgr *km;
1763 
1764 	read_lock(&xfrm_km_lock);
1765 	list_for_each_entry(km, &xfrm_km_list, list) {
1766 		if (km->report) {
1767 			ret = km->report(net, proto, sel, addr);
1768 			if (!ret)
1769 				err = ret;
1770 		}
1771 	}
1772 	read_unlock(&xfrm_km_lock);
1773 	return err;
1774 }
1775 EXPORT_SYMBOL(km_report);
1776 
1777 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1778 {
1779 	int err;
1780 	u8 *data;
1781 	struct xfrm_mgr *km;
1782 	struct xfrm_policy *pol = NULL;
1783 
1784 	if (optlen <= 0 || optlen > PAGE_SIZE)
1785 		return -EMSGSIZE;
1786 
1787 	data = kmalloc(optlen, GFP_KERNEL);
1788 	if (!data)
1789 		return -ENOMEM;
1790 
1791 	err = -EFAULT;
1792 	if (copy_from_user(data, optval, optlen))
1793 		goto out;
1794 
1795 	err = -EINVAL;
1796 	read_lock(&xfrm_km_lock);
1797 	list_for_each_entry(km, &xfrm_km_list, list) {
1798 		pol = km->compile_policy(sk, optname, data,
1799 					 optlen, &err);
1800 		if (err >= 0)
1801 			break;
1802 	}
1803 	read_unlock(&xfrm_km_lock);
1804 
1805 	if (err >= 0) {
1806 		xfrm_sk_policy_insert(sk, err, pol);
1807 		xfrm_pol_put(pol);
1808 		err = 0;
1809 	}
1810 
1811 out:
1812 	kfree(data);
1813 	return err;
1814 }
1815 EXPORT_SYMBOL(xfrm_user_policy);
1816 
1817 int xfrm_register_km(struct xfrm_mgr *km)
1818 {
1819 	write_lock_bh(&xfrm_km_lock);
1820 	list_add_tail(&km->list, &xfrm_km_list);
1821 	write_unlock_bh(&xfrm_km_lock);
1822 	return 0;
1823 }
1824 EXPORT_SYMBOL(xfrm_register_km);
1825 
1826 int xfrm_unregister_km(struct xfrm_mgr *km)
1827 {
1828 	write_lock_bh(&xfrm_km_lock);
1829 	list_del(&km->list);
1830 	write_unlock_bh(&xfrm_km_lock);
1831 	return 0;
1832 }
1833 EXPORT_SYMBOL(xfrm_unregister_km);
1834 
1835 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1836 {
1837 	int err = 0;
1838 	if (unlikely(afinfo == NULL))
1839 		return -EINVAL;
1840 	if (unlikely(afinfo->family >= NPROTO))
1841 		return -EAFNOSUPPORT;
1842 	write_lock_bh(&xfrm_state_afinfo_lock);
1843 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1844 		err = -ENOBUFS;
1845 	else
1846 		xfrm_state_afinfo[afinfo->family] = afinfo;
1847 	write_unlock_bh(&xfrm_state_afinfo_lock);
1848 	return err;
1849 }
1850 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1851 
1852 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1853 {
1854 	int err = 0;
1855 	if (unlikely(afinfo == NULL))
1856 		return -EINVAL;
1857 	if (unlikely(afinfo->family >= NPROTO))
1858 		return -EAFNOSUPPORT;
1859 	write_lock_bh(&xfrm_state_afinfo_lock);
1860 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1861 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1862 			err = -EINVAL;
1863 		else
1864 			xfrm_state_afinfo[afinfo->family] = NULL;
1865 	}
1866 	write_unlock_bh(&xfrm_state_afinfo_lock);
1867 	return err;
1868 }
1869 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1870 
1871 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1872 {
1873 	struct xfrm_state_afinfo *afinfo;
1874 	if (unlikely(family >= NPROTO))
1875 		return NULL;
1876 	read_lock(&xfrm_state_afinfo_lock);
1877 	afinfo = xfrm_state_afinfo[family];
1878 	if (unlikely(!afinfo))
1879 		read_unlock(&xfrm_state_afinfo_lock);
1880 	return afinfo;
1881 }
1882 
1883 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1884 	__releases(xfrm_state_afinfo_lock)
1885 {
1886 	read_unlock(&xfrm_state_afinfo_lock);
1887 }
1888 
1889 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1890 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1891 {
1892 	if (x->tunnel) {
1893 		struct xfrm_state *t = x->tunnel;
1894 
1895 		if (atomic_read(&t->tunnel_users) == 2)
1896 			xfrm_state_delete(t);
1897 		atomic_dec(&t->tunnel_users);
1898 		xfrm_state_put(t);
1899 		x->tunnel = NULL;
1900 	}
1901 }
1902 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1903 
1904 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1905 {
1906 	int res;
1907 
1908 	spin_lock_bh(&x->lock);
1909 	if (x->km.state == XFRM_STATE_VALID &&
1910 	    x->type && x->type->get_mtu)
1911 		res = x->type->get_mtu(x, mtu);
1912 	else
1913 		res = mtu - x->props.header_len;
1914 	spin_unlock_bh(&x->lock);
1915 	return res;
1916 }
1917 
1918 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1919 {
1920 	struct xfrm_state_afinfo *afinfo;
1921 	struct xfrm_mode *inner_mode;
1922 	int family = x->props.family;
1923 	int err;
1924 
1925 	err = -EAFNOSUPPORT;
1926 	afinfo = xfrm_state_get_afinfo(family);
1927 	if (!afinfo)
1928 		goto error;
1929 
1930 	err = 0;
1931 	if (afinfo->init_flags)
1932 		err = afinfo->init_flags(x);
1933 
1934 	xfrm_state_put_afinfo(afinfo);
1935 
1936 	if (err)
1937 		goto error;
1938 
1939 	err = -EPROTONOSUPPORT;
1940 
1941 	if (x->sel.family != AF_UNSPEC) {
1942 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1943 		if (inner_mode == NULL)
1944 			goto error;
1945 
1946 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1947 		    family != x->sel.family) {
1948 			xfrm_put_mode(inner_mode);
1949 			goto error;
1950 		}
1951 
1952 		x->inner_mode = inner_mode;
1953 	} else {
1954 		struct xfrm_mode *inner_mode_iaf;
1955 		int iafamily = AF_INET;
1956 
1957 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1958 		if (inner_mode == NULL)
1959 			goto error;
1960 
1961 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1962 			xfrm_put_mode(inner_mode);
1963 			goto error;
1964 		}
1965 		x->inner_mode = inner_mode;
1966 
1967 		if (x->props.family == AF_INET)
1968 			iafamily = AF_INET6;
1969 
1970 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1971 		if (inner_mode_iaf) {
1972 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1973 				x->inner_mode_iaf = inner_mode_iaf;
1974 			else
1975 				xfrm_put_mode(inner_mode_iaf);
1976 		}
1977 	}
1978 
1979 	x->type = xfrm_get_type(x->id.proto, family);
1980 	if (x->type == NULL)
1981 		goto error;
1982 
1983 	err = x->type->init_state(x);
1984 	if (err)
1985 		goto error;
1986 
1987 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
1988 	if (x->outer_mode == NULL)
1989 		goto error;
1990 
1991 	if (init_replay) {
1992 		err = xfrm_init_replay(x);
1993 		if (err)
1994 			goto error;
1995 	}
1996 
1997 	x->km.state = XFRM_STATE_VALID;
1998 
1999 error:
2000 	return err;
2001 }
2002 
2003 EXPORT_SYMBOL(__xfrm_init_state);
2004 
2005 int xfrm_init_state(struct xfrm_state *x)
2006 {
2007 	return __xfrm_init_state(x, true);
2008 }
2009 
2010 EXPORT_SYMBOL(xfrm_init_state);
2011 
2012 int __net_init xfrm_state_init(struct net *net)
2013 {
2014 	unsigned int sz;
2015 
2016 	INIT_LIST_HEAD(&net->xfrm.state_all);
2017 
2018 	sz = sizeof(struct hlist_head) * 8;
2019 
2020 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2021 	if (!net->xfrm.state_bydst)
2022 		goto out_bydst;
2023 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2024 	if (!net->xfrm.state_bysrc)
2025 		goto out_bysrc;
2026 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2027 	if (!net->xfrm.state_byspi)
2028 		goto out_byspi;
2029 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2030 
2031 	net->xfrm.state_num = 0;
2032 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2033 	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2034 	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2035 	init_waitqueue_head(&net->xfrm.km_waitq);
2036 	return 0;
2037 
2038 out_byspi:
2039 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2040 out_bysrc:
2041 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2042 out_bydst:
2043 	return -ENOMEM;
2044 }
2045 
2046 void xfrm_state_fini(struct net *net)
2047 {
2048 	struct xfrm_audit audit_info;
2049 	unsigned int sz;
2050 
2051 	flush_work(&net->xfrm.state_hash_work);
2052 	audit_info.loginuid = -1;
2053 	audit_info.sessionid = -1;
2054 	audit_info.secid = 0;
2055 	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2056 	flush_work(&net->xfrm.state_gc_work);
2057 
2058 	WARN_ON(!list_empty(&net->xfrm.state_all));
2059 
2060 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2061 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2062 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2063 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2064 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2065 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2066 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2067 }
2068 
2069 #ifdef CONFIG_AUDITSYSCALL
2070 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2071 				     struct audit_buffer *audit_buf)
2072 {
2073 	struct xfrm_sec_ctx *ctx = x->security;
2074 	u32 spi = ntohl(x->id.spi);
2075 
2076 	if (ctx)
2077 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2078 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2079 
2080 	switch(x->props.family) {
2081 	case AF_INET:
2082 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2083 				 &x->props.saddr.a4, &x->id.daddr.a4);
2084 		break;
2085 	case AF_INET6:
2086 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2087 				 x->props.saddr.a6, x->id.daddr.a6);
2088 		break;
2089 	}
2090 
2091 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2092 }
2093 
2094 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2095 				      struct audit_buffer *audit_buf)
2096 {
2097 	const struct iphdr *iph4;
2098 	const struct ipv6hdr *iph6;
2099 
2100 	switch (family) {
2101 	case AF_INET:
2102 		iph4 = ip_hdr(skb);
2103 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2104 				 &iph4->saddr, &iph4->daddr);
2105 		break;
2106 	case AF_INET6:
2107 		iph6 = ipv6_hdr(skb);
2108 		audit_log_format(audit_buf,
2109 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2110 				 &iph6->saddr,&iph6->daddr,
2111 				 iph6->flow_lbl[0] & 0x0f,
2112 				 iph6->flow_lbl[1],
2113 				 iph6->flow_lbl[2]);
2114 		break;
2115 	}
2116 }
2117 
2118 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2119 			  uid_t auid, u32 sessionid, u32 secid)
2120 {
2121 	struct audit_buffer *audit_buf;
2122 
2123 	audit_buf = xfrm_audit_start("SAD-add");
2124 	if (audit_buf == NULL)
2125 		return;
2126 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2127 	xfrm_audit_helper_sainfo(x, audit_buf);
2128 	audit_log_format(audit_buf, " res=%u", result);
2129 	audit_log_end(audit_buf);
2130 }
2131 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2132 
2133 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2134 			     uid_t auid, u32 sessionid, u32 secid)
2135 {
2136 	struct audit_buffer *audit_buf;
2137 
2138 	audit_buf = xfrm_audit_start("SAD-delete");
2139 	if (audit_buf == NULL)
2140 		return;
2141 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2142 	xfrm_audit_helper_sainfo(x, audit_buf);
2143 	audit_log_format(audit_buf, " res=%u", result);
2144 	audit_log_end(audit_buf);
2145 }
2146 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2147 
2148 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2149 				      struct sk_buff *skb)
2150 {
2151 	struct audit_buffer *audit_buf;
2152 	u32 spi;
2153 
2154 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2155 	if (audit_buf == NULL)
2156 		return;
2157 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2158 	/* don't record the sequence number because it's inherent in this kind
2159 	 * of audit message */
2160 	spi = ntohl(x->id.spi);
2161 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2162 	audit_log_end(audit_buf);
2163 }
2164 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2165 
2166 void xfrm_audit_state_replay(struct xfrm_state *x,
2167 			     struct sk_buff *skb, __be32 net_seq)
2168 {
2169 	struct audit_buffer *audit_buf;
2170 	u32 spi;
2171 
2172 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2173 	if (audit_buf == NULL)
2174 		return;
2175 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2176 	spi = ntohl(x->id.spi);
2177 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2178 			 spi, spi, ntohl(net_seq));
2179 	audit_log_end(audit_buf);
2180 }
2181 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2182 
2183 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2184 {
2185 	struct audit_buffer *audit_buf;
2186 
2187 	audit_buf = xfrm_audit_start("SA-notfound");
2188 	if (audit_buf == NULL)
2189 		return;
2190 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2191 	audit_log_end(audit_buf);
2192 }
2193 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2194 
2195 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2196 			       __be32 net_spi, __be32 net_seq)
2197 {
2198 	struct audit_buffer *audit_buf;
2199 	u32 spi;
2200 
2201 	audit_buf = xfrm_audit_start("SA-notfound");
2202 	if (audit_buf == NULL)
2203 		return;
2204 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2205 	spi = ntohl(net_spi);
2206 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2207 			 spi, spi, ntohl(net_seq));
2208 	audit_log_end(audit_buf);
2209 }
2210 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2211 
2212 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2213 			      struct sk_buff *skb, u8 proto)
2214 {
2215 	struct audit_buffer *audit_buf;
2216 	__be32 net_spi;
2217 	__be32 net_seq;
2218 
2219 	audit_buf = xfrm_audit_start("SA-icv-failure");
2220 	if (audit_buf == NULL)
2221 		return;
2222 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2223 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2224 		u32 spi = ntohl(net_spi);
2225 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2226 				 spi, spi, ntohl(net_seq));
2227 	}
2228 	audit_log_end(audit_buf);
2229 }
2230 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2231 #endif /* CONFIG_AUDITSYSCALL */
2232