xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 8a10bc9d)
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 unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
39 
40 static inline unsigned int xfrm_dst_hash(struct net *net,
41 					 const xfrm_address_t *daddr,
42 					 const xfrm_address_t *saddr,
43 					 u32 reqid,
44 					 unsigned short family)
45 {
46 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
47 }
48 
49 static inline unsigned int xfrm_src_hash(struct net *net,
50 					 const xfrm_address_t *daddr,
51 					 const xfrm_address_t *saddr,
52 					 unsigned short family)
53 {
54 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
55 }
56 
57 static inline unsigned int
58 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
59 	      __be32 spi, u8 proto, unsigned short family)
60 {
61 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
62 }
63 
64 static void xfrm_hash_transfer(struct hlist_head *list,
65 			       struct hlist_head *ndsttable,
66 			       struct hlist_head *nsrctable,
67 			       struct hlist_head *nspitable,
68 			       unsigned int nhashmask)
69 {
70 	struct hlist_node *tmp;
71 	struct xfrm_state *x;
72 
73 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
74 		unsigned int h;
75 
76 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
77 				    x->props.reqid, x->props.family,
78 				    nhashmask);
79 		hlist_add_head(&x->bydst, ndsttable+h);
80 
81 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
82 				    x->props.family,
83 				    nhashmask);
84 		hlist_add_head(&x->bysrc, nsrctable+h);
85 
86 		if (x->id.spi) {
87 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
88 					    x->id.proto, x->props.family,
89 					    nhashmask);
90 			hlist_add_head(&x->byspi, nspitable+h);
91 		}
92 	}
93 }
94 
95 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
96 {
97 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
98 }
99 
100 static DEFINE_MUTEX(hash_resize_mutex);
101 
102 static void xfrm_hash_resize(struct work_struct *work)
103 {
104 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
105 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
106 	unsigned long nsize, osize;
107 	unsigned int nhashmask, ohashmask;
108 	int i;
109 
110 	mutex_lock(&hash_resize_mutex);
111 
112 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
113 	ndst = xfrm_hash_alloc(nsize);
114 	if (!ndst)
115 		goto out_unlock;
116 	nsrc = xfrm_hash_alloc(nsize);
117 	if (!nsrc) {
118 		xfrm_hash_free(ndst, nsize);
119 		goto out_unlock;
120 	}
121 	nspi = xfrm_hash_alloc(nsize);
122 	if (!nspi) {
123 		xfrm_hash_free(ndst, nsize);
124 		xfrm_hash_free(nsrc, nsize);
125 		goto out_unlock;
126 	}
127 
128 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
129 
130 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
131 	for (i = net->xfrm.state_hmask; i >= 0; i--)
132 		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
133 				   nhashmask);
134 
135 	odst = net->xfrm.state_bydst;
136 	osrc = net->xfrm.state_bysrc;
137 	ospi = net->xfrm.state_byspi;
138 	ohashmask = net->xfrm.state_hmask;
139 
140 	net->xfrm.state_bydst = ndst;
141 	net->xfrm.state_bysrc = nsrc;
142 	net->xfrm.state_byspi = nspi;
143 	net->xfrm.state_hmask = nhashmask;
144 
145 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
146 
147 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
148 	xfrm_hash_free(odst, osize);
149 	xfrm_hash_free(osrc, osize);
150 	xfrm_hash_free(ospi, osize);
151 
152 out_unlock:
153 	mutex_unlock(&hash_resize_mutex);
154 }
155 
156 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
157 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
158 
159 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
160 
161 int __xfrm_state_delete(struct xfrm_state *x);
162 
163 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
164 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
165 
166 static DEFINE_SPINLOCK(xfrm_type_lock);
167 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
168 {
169 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
170 	const struct xfrm_type **typemap;
171 	int err = 0;
172 
173 	if (unlikely(afinfo == NULL))
174 		return -EAFNOSUPPORT;
175 	typemap = afinfo->type_map;
176 	spin_lock_bh(&xfrm_type_lock);
177 
178 	if (likely(typemap[type->proto] == NULL))
179 		typemap[type->proto] = type;
180 	else
181 		err = -EEXIST;
182 	spin_unlock_bh(&xfrm_type_lock);
183 	xfrm_state_put_afinfo(afinfo);
184 	return err;
185 }
186 EXPORT_SYMBOL(xfrm_register_type);
187 
188 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
189 {
190 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
191 	const struct xfrm_type **typemap;
192 	int err = 0;
193 
194 	if (unlikely(afinfo == NULL))
195 		return -EAFNOSUPPORT;
196 	typemap = afinfo->type_map;
197 	spin_lock_bh(&xfrm_type_lock);
198 
199 	if (unlikely(typemap[type->proto] != type))
200 		err = -ENOENT;
201 	else
202 		typemap[type->proto] = NULL;
203 	spin_unlock_bh(&xfrm_type_lock);
204 	xfrm_state_put_afinfo(afinfo);
205 	return err;
206 }
207 EXPORT_SYMBOL(xfrm_unregister_type);
208 
209 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
210 {
211 	struct xfrm_state_afinfo *afinfo;
212 	const struct xfrm_type **typemap;
213 	const struct xfrm_type *type;
214 	int modload_attempted = 0;
215 
216 retry:
217 	afinfo = xfrm_state_get_afinfo(family);
218 	if (unlikely(afinfo == NULL))
219 		return NULL;
220 	typemap = afinfo->type_map;
221 
222 	type = typemap[proto];
223 	if (unlikely(type && !try_module_get(type->owner)))
224 		type = NULL;
225 	if (!type && !modload_attempted) {
226 		xfrm_state_put_afinfo(afinfo);
227 		request_module("xfrm-type-%d-%d", family, proto);
228 		modload_attempted = 1;
229 		goto retry;
230 	}
231 
232 	xfrm_state_put_afinfo(afinfo);
233 	return type;
234 }
235 
236 static void xfrm_put_type(const struct xfrm_type *type)
237 {
238 	module_put(type->owner);
239 }
240 
241 static DEFINE_SPINLOCK(xfrm_mode_lock);
242 int xfrm_register_mode(struct xfrm_mode *mode, int family)
243 {
244 	struct xfrm_state_afinfo *afinfo;
245 	struct xfrm_mode **modemap;
246 	int err;
247 
248 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
249 		return -EINVAL;
250 
251 	afinfo = xfrm_state_get_afinfo(family);
252 	if (unlikely(afinfo == NULL))
253 		return -EAFNOSUPPORT;
254 
255 	err = -EEXIST;
256 	modemap = afinfo->mode_map;
257 	spin_lock_bh(&xfrm_mode_lock);
258 	if (modemap[mode->encap])
259 		goto out;
260 
261 	err = -ENOENT;
262 	if (!try_module_get(afinfo->owner))
263 		goto out;
264 
265 	mode->afinfo = afinfo;
266 	modemap[mode->encap] = mode;
267 	err = 0;
268 
269 out:
270 	spin_unlock_bh(&xfrm_mode_lock);
271 	xfrm_state_put_afinfo(afinfo);
272 	return err;
273 }
274 EXPORT_SYMBOL(xfrm_register_mode);
275 
276 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
277 {
278 	struct xfrm_state_afinfo *afinfo;
279 	struct xfrm_mode **modemap;
280 	int err;
281 
282 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
283 		return -EINVAL;
284 
285 	afinfo = xfrm_state_get_afinfo(family);
286 	if (unlikely(afinfo == NULL))
287 		return -EAFNOSUPPORT;
288 
289 	err = -ENOENT;
290 	modemap = afinfo->mode_map;
291 	spin_lock_bh(&xfrm_mode_lock);
292 	if (likely(modemap[mode->encap] == mode)) {
293 		modemap[mode->encap] = NULL;
294 		module_put(mode->afinfo->owner);
295 		err = 0;
296 	}
297 
298 	spin_unlock_bh(&xfrm_mode_lock);
299 	xfrm_state_put_afinfo(afinfo);
300 	return err;
301 }
302 EXPORT_SYMBOL(xfrm_unregister_mode);
303 
304 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
305 {
306 	struct xfrm_state_afinfo *afinfo;
307 	struct xfrm_mode *mode;
308 	int modload_attempted = 0;
309 
310 	if (unlikely(encap >= XFRM_MODE_MAX))
311 		return NULL;
312 
313 retry:
314 	afinfo = xfrm_state_get_afinfo(family);
315 	if (unlikely(afinfo == NULL))
316 		return NULL;
317 
318 	mode = afinfo->mode_map[encap];
319 	if (unlikely(mode && !try_module_get(mode->owner)))
320 		mode = NULL;
321 	if (!mode && !modload_attempted) {
322 		xfrm_state_put_afinfo(afinfo);
323 		request_module("xfrm-mode-%d-%d", family, encap);
324 		modload_attempted = 1;
325 		goto retry;
326 	}
327 
328 	xfrm_state_put_afinfo(afinfo);
329 	return mode;
330 }
331 
332 static void xfrm_put_mode(struct xfrm_mode *mode)
333 {
334 	module_put(mode->owner);
335 }
336 
337 static void xfrm_state_gc_destroy(struct xfrm_state *x)
338 {
339 	tasklet_hrtimer_cancel(&x->mtimer);
340 	del_timer_sync(&x->rtimer);
341 	kfree(x->aalg);
342 	kfree(x->ealg);
343 	kfree(x->calg);
344 	kfree(x->encap);
345 	kfree(x->coaddr);
346 	kfree(x->replay_esn);
347 	kfree(x->preplay_esn);
348 	if (x->inner_mode)
349 		xfrm_put_mode(x->inner_mode);
350 	if (x->inner_mode_iaf)
351 		xfrm_put_mode(x->inner_mode_iaf);
352 	if (x->outer_mode)
353 		xfrm_put_mode(x->outer_mode);
354 	if (x->type) {
355 		x->type->destructor(x);
356 		xfrm_put_type(x->type);
357 	}
358 	security_xfrm_state_free(x);
359 	kfree(x);
360 }
361 
362 static void xfrm_state_gc_task(struct work_struct *work)
363 {
364 	struct net *net = container_of(work, struct net, xfrm.state_gc_work);
365 	struct xfrm_state *x;
366 	struct hlist_node *tmp;
367 	struct hlist_head gc_list;
368 
369 	spin_lock_bh(&xfrm_state_gc_lock);
370 	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
371 	spin_unlock_bh(&xfrm_state_gc_lock);
372 
373 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
374 		xfrm_state_gc_destroy(x);
375 }
376 
377 static inline unsigned long make_jiffies(long secs)
378 {
379 	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
380 		return MAX_SCHEDULE_TIMEOUT-1;
381 	else
382 		return secs*HZ;
383 }
384 
385 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
386 {
387 	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
388 	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
389 	unsigned long now = get_seconds();
390 	long next = LONG_MAX;
391 	int warn = 0;
392 	int err = 0;
393 
394 	spin_lock(&x->lock);
395 	if (x->km.state == XFRM_STATE_DEAD)
396 		goto out;
397 	if (x->km.state == XFRM_STATE_EXPIRED)
398 		goto expired;
399 	if (x->lft.hard_add_expires_seconds) {
400 		long tmo = x->lft.hard_add_expires_seconds +
401 			x->curlft.add_time - now;
402 		if (tmo <= 0) {
403 			if (x->xflags & XFRM_SOFT_EXPIRE) {
404 				/* enter hard expire without soft expire first?!
405 				 * setting a new date could trigger this.
406 				 * workarbound: fix x->curflt.add_time by below:
407 				 */
408 				x->curlft.add_time = now - x->saved_tmo - 1;
409 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
410 			} else
411 				goto expired;
412 		}
413 		if (tmo < next)
414 			next = tmo;
415 	}
416 	if (x->lft.hard_use_expires_seconds) {
417 		long tmo = x->lft.hard_use_expires_seconds +
418 			(x->curlft.use_time ? : now) - now;
419 		if (tmo <= 0)
420 			goto expired;
421 		if (tmo < next)
422 			next = tmo;
423 	}
424 	if (x->km.dying)
425 		goto resched;
426 	if (x->lft.soft_add_expires_seconds) {
427 		long tmo = x->lft.soft_add_expires_seconds +
428 			x->curlft.add_time - now;
429 		if (tmo <= 0) {
430 			warn = 1;
431 			x->xflags &= ~XFRM_SOFT_EXPIRE;
432 		} else if (tmo < next) {
433 			next = tmo;
434 			x->xflags |= XFRM_SOFT_EXPIRE;
435 			x->saved_tmo = tmo;
436 		}
437 	}
438 	if (x->lft.soft_use_expires_seconds) {
439 		long tmo = x->lft.soft_use_expires_seconds +
440 			(x->curlft.use_time ? : now) - now;
441 		if (tmo <= 0)
442 			warn = 1;
443 		else if (tmo < next)
444 			next = tmo;
445 	}
446 
447 	x->km.dying = warn;
448 	if (warn)
449 		km_state_expired(x, 0, 0);
450 resched:
451 	if (next != LONG_MAX) {
452 		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
453 	}
454 
455 	goto out;
456 
457 expired:
458 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
459 		x->km.state = XFRM_STATE_EXPIRED;
460 
461 	err = __xfrm_state_delete(x);
462 	if (!err)
463 		km_state_expired(x, 1, 0);
464 
465 	xfrm_audit_state_delete(x, err ? 0 : 1,
466 				audit_get_loginuid(current),
467 				audit_get_sessionid(current), 0);
468 
469 out:
470 	spin_unlock(&x->lock);
471 	return HRTIMER_NORESTART;
472 }
473 
474 static void xfrm_replay_timer_handler(unsigned long data);
475 
476 struct xfrm_state *xfrm_state_alloc(struct net *net)
477 {
478 	struct xfrm_state *x;
479 
480 	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
481 
482 	if (x) {
483 		write_pnet(&x->xs_net, net);
484 		atomic_set(&x->refcnt, 1);
485 		atomic_set(&x->tunnel_users, 0);
486 		INIT_LIST_HEAD(&x->km.all);
487 		INIT_HLIST_NODE(&x->bydst);
488 		INIT_HLIST_NODE(&x->bysrc);
489 		INIT_HLIST_NODE(&x->byspi);
490 		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
491 					CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
492 		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
493 				(unsigned long)x);
494 		x->curlft.add_time = get_seconds();
495 		x->lft.soft_byte_limit = XFRM_INF;
496 		x->lft.soft_packet_limit = XFRM_INF;
497 		x->lft.hard_byte_limit = XFRM_INF;
498 		x->lft.hard_packet_limit = XFRM_INF;
499 		x->replay_maxage = 0;
500 		x->replay_maxdiff = 0;
501 		x->inner_mode = NULL;
502 		x->inner_mode_iaf = NULL;
503 		spin_lock_init(&x->lock);
504 	}
505 	return x;
506 }
507 EXPORT_SYMBOL(xfrm_state_alloc);
508 
509 void __xfrm_state_destroy(struct xfrm_state *x)
510 {
511 	struct net *net = xs_net(x);
512 
513 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
514 
515 	spin_lock_bh(&xfrm_state_gc_lock);
516 	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
517 	spin_unlock_bh(&xfrm_state_gc_lock);
518 	schedule_work(&net->xfrm.state_gc_work);
519 }
520 EXPORT_SYMBOL(__xfrm_state_destroy);
521 
522 int __xfrm_state_delete(struct xfrm_state *x)
523 {
524 	struct net *net = xs_net(x);
525 	int err = -ESRCH;
526 
527 	if (x->km.state != XFRM_STATE_DEAD) {
528 		x->km.state = XFRM_STATE_DEAD;
529 		spin_lock(&net->xfrm.xfrm_state_lock);
530 		list_del(&x->km.all);
531 		hlist_del(&x->bydst);
532 		hlist_del(&x->bysrc);
533 		if (x->id.spi)
534 			hlist_del(&x->byspi);
535 		net->xfrm.state_num--;
536 		spin_unlock(&net->xfrm.xfrm_state_lock);
537 
538 		/* All xfrm_state objects are created by xfrm_state_alloc.
539 		 * The xfrm_state_alloc call gives a reference, and that
540 		 * is what we are dropping here.
541 		 */
542 		xfrm_state_put(x);
543 		err = 0;
544 	}
545 
546 	return err;
547 }
548 EXPORT_SYMBOL(__xfrm_state_delete);
549 
550 int xfrm_state_delete(struct xfrm_state *x)
551 {
552 	int err;
553 
554 	spin_lock_bh(&x->lock);
555 	err = __xfrm_state_delete(x);
556 	spin_unlock_bh(&x->lock);
557 
558 	return err;
559 }
560 EXPORT_SYMBOL(xfrm_state_delete);
561 
562 #ifdef CONFIG_SECURITY_NETWORK_XFRM
563 static inline int
564 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
565 {
566 	int i, err = 0;
567 
568 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
569 		struct xfrm_state *x;
570 
571 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
572 			if (xfrm_id_proto_match(x->id.proto, proto) &&
573 			   (err = security_xfrm_state_delete(x)) != 0) {
574 				xfrm_audit_state_delete(x, 0,
575 							audit_info->loginuid,
576 							audit_info->sessionid,
577 							audit_info->secid);
578 				return err;
579 			}
580 		}
581 	}
582 
583 	return err;
584 }
585 #else
586 static inline int
587 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
588 {
589 	return 0;
590 }
591 #endif
592 
593 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
594 {
595 	int i, err = 0, cnt = 0;
596 
597 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
598 	err = xfrm_state_flush_secctx_check(net, proto, audit_info);
599 	if (err)
600 		goto out;
601 
602 	err = -ESRCH;
603 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
604 		struct xfrm_state *x;
605 restart:
606 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
607 			if (!xfrm_state_kern(x) &&
608 			    xfrm_id_proto_match(x->id.proto, proto)) {
609 				xfrm_state_hold(x);
610 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
611 
612 				err = xfrm_state_delete(x);
613 				xfrm_audit_state_delete(x, err ? 0 : 1,
614 							audit_info->loginuid,
615 							audit_info->sessionid,
616 							audit_info->secid);
617 				xfrm_state_put(x);
618 				if (!err)
619 					cnt++;
620 
621 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
622 				goto restart;
623 			}
624 		}
625 	}
626 	if (cnt)
627 		err = 0;
628 
629 out:
630 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
631 	return err;
632 }
633 EXPORT_SYMBOL(xfrm_state_flush);
634 
635 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
636 {
637 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
638 	si->sadcnt = net->xfrm.state_num;
639 	si->sadhcnt = net->xfrm.state_hmask;
640 	si->sadhmcnt = xfrm_state_hashmax;
641 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
642 }
643 EXPORT_SYMBOL(xfrm_sad_getinfo);
644 
645 static int
646 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
647 		    const struct xfrm_tmpl *tmpl,
648 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
649 		    unsigned short family)
650 {
651 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
652 	if (!afinfo)
653 		return -1;
654 	afinfo->init_tempsel(&x->sel, fl);
655 
656 	if (family != tmpl->encap_family) {
657 		xfrm_state_put_afinfo(afinfo);
658 		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
659 		if (!afinfo)
660 			return -1;
661 	}
662 	afinfo->init_temprop(x, tmpl, daddr, saddr);
663 	xfrm_state_put_afinfo(afinfo);
664 	return 0;
665 }
666 
667 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
668 					      const xfrm_address_t *daddr,
669 					      __be32 spi, u8 proto,
670 					      unsigned short family)
671 {
672 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
673 	struct xfrm_state *x;
674 
675 	hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
676 		if (x->props.family != family ||
677 		    x->id.spi       != spi ||
678 		    x->id.proto     != proto ||
679 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
680 			continue;
681 
682 		if ((mark & x->mark.m) != x->mark.v)
683 			continue;
684 		xfrm_state_hold(x);
685 		return x;
686 	}
687 
688 	return NULL;
689 }
690 
691 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
692 						     const xfrm_address_t *daddr,
693 						     const xfrm_address_t *saddr,
694 						     u8 proto, unsigned short family)
695 {
696 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
697 	struct xfrm_state *x;
698 
699 	hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
700 		if (x->props.family != family ||
701 		    x->id.proto     != proto ||
702 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
703 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
704 			continue;
705 
706 		if ((mark & x->mark.m) != x->mark.v)
707 			continue;
708 		xfrm_state_hold(x);
709 		return x;
710 	}
711 
712 	return NULL;
713 }
714 
715 static inline struct xfrm_state *
716 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
717 {
718 	struct net *net = xs_net(x);
719 	u32 mark = x->mark.v & x->mark.m;
720 
721 	if (use_spi)
722 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
723 					   x->id.spi, x->id.proto, family);
724 	else
725 		return __xfrm_state_lookup_byaddr(net, mark,
726 						  &x->id.daddr,
727 						  &x->props.saddr,
728 						  x->id.proto, family);
729 }
730 
731 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
732 {
733 	if (have_hash_collision &&
734 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
735 	    net->xfrm.state_num > net->xfrm.state_hmask)
736 		schedule_work(&net->xfrm.state_hash_work);
737 }
738 
739 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
740 			       const struct flowi *fl, unsigned short family,
741 			       struct xfrm_state **best, int *acq_in_progress,
742 			       int *error)
743 {
744 	/* Resolution logic:
745 	 * 1. There is a valid state with matching selector. Done.
746 	 * 2. Valid state with inappropriate selector. Skip.
747 	 *
748 	 * Entering area of "sysdeps".
749 	 *
750 	 * 3. If state is not valid, selector is temporary, it selects
751 	 *    only session which triggered previous resolution. Key
752 	 *    manager will do something to install a state with proper
753 	 *    selector.
754 	 */
755 	if (x->km.state == XFRM_STATE_VALID) {
756 		if ((x->sel.family &&
757 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
758 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
759 			return;
760 
761 		if (!*best ||
762 		    (*best)->km.dying > x->km.dying ||
763 		    ((*best)->km.dying == x->km.dying &&
764 		     (*best)->curlft.add_time < x->curlft.add_time))
765 			*best = x;
766 	} else if (x->km.state == XFRM_STATE_ACQ) {
767 		*acq_in_progress = 1;
768 	} else if (x->km.state == XFRM_STATE_ERROR ||
769 		   x->km.state == XFRM_STATE_EXPIRED) {
770 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
771 		    security_xfrm_state_pol_flow_match(x, pol, fl))
772 			*error = -ESRCH;
773 	}
774 }
775 
776 struct xfrm_state *
777 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
778 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
779 		struct xfrm_policy *pol, int *err,
780 		unsigned short family)
781 {
782 	static xfrm_address_t saddr_wildcard = { };
783 	struct net *net = xp_net(pol);
784 	unsigned int h, h_wildcard;
785 	struct xfrm_state *x, *x0, *to_put;
786 	int acquire_in_progress = 0;
787 	int error = 0;
788 	struct xfrm_state *best = NULL;
789 	u32 mark = pol->mark.v & pol->mark.m;
790 	unsigned short encap_family = tmpl->encap_family;
791 
792 	to_put = NULL;
793 
794 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
795 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
796 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
797 		if (x->props.family == encap_family &&
798 		    x->props.reqid == tmpl->reqid &&
799 		    (mark & x->mark.m) == x->mark.v &&
800 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
801 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
802 		    tmpl->mode == x->props.mode &&
803 		    tmpl->id.proto == x->id.proto &&
804 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
805 			xfrm_state_look_at(pol, x, fl, encap_family,
806 					   &best, &acquire_in_progress, &error);
807 	}
808 	if (best || acquire_in_progress)
809 		goto found;
810 
811 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
812 	hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
813 		if (x->props.family == encap_family &&
814 		    x->props.reqid == tmpl->reqid &&
815 		    (mark & x->mark.m) == x->mark.v &&
816 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
817 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
818 		    tmpl->mode == x->props.mode &&
819 		    tmpl->id.proto == x->id.proto &&
820 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
821 			xfrm_state_look_at(pol, x, fl, encap_family,
822 					   &best, &acquire_in_progress, &error);
823 	}
824 
825 found:
826 	x = best;
827 	if (!x && !error && !acquire_in_progress) {
828 		if (tmpl->id.spi &&
829 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
830 					      tmpl->id.proto, encap_family)) != NULL) {
831 			to_put = x0;
832 			error = -EEXIST;
833 			goto out;
834 		}
835 		x = xfrm_state_alloc(net);
836 		if (x == NULL) {
837 			error = -ENOMEM;
838 			goto out;
839 		}
840 		/* Initialize temporary state matching only
841 		 * to current session. */
842 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
843 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
844 
845 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
846 		if (error) {
847 			x->km.state = XFRM_STATE_DEAD;
848 			to_put = x;
849 			x = NULL;
850 			goto out;
851 		}
852 
853 		if (km_query(x, tmpl, pol) == 0) {
854 			x->km.state = XFRM_STATE_ACQ;
855 			list_add(&x->km.all, &net->xfrm.state_all);
856 			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
857 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
858 			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
859 			if (x->id.spi) {
860 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
861 				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
862 			}
863 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
864 			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
865 			net->xfrm.state_num++;
866 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
867 		} else {
868 			x->km.state = XFRM_STATE_DEAD;
869 			to_put = x;
870 			x = NULL;
871 			error = -ESRCH;
872 		}
873 	}
874 out:
875 	if (x)
876 		xfrm_state_hold(x);
877 	else
878 		*err = acquire_in_progress ? -EAGAIN : error;
879 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
880 	if (to_put)
881 		xfrm_state_put(to_put);
882 	return x;
883 }
884 
885 struct xfrm_state *
886 xfrm_stateonly_find(struct net *net, u32 mark,
887 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
888 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
889 {
890 	unsigned int h;
891 	struct xfrm_state *rx = NULL, *x = NULL;
892 
893 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
894 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
895 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
896 		if (x->props.family == family &&
897 		    x->props.reqid == reqid &&
898 		    (mark & x->mark.m) == x->mark.v &&
899 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
900 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
901 		    mode == x->props.mode &&
902 		    proto == x->id.proto &&
903 		    x->km.state == XFRM_STATE_VALID) {
904 			rx = x;
905 			break;
906 		}
907 	}
908 
909 	if (rx)
910 		xfrm_state_hold(rx);
911 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
912 
913 
914 	return rx;
915 }
916 EXPORT_SYMBOL(xfrm_stateonly_find);
917 
918 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
919 					      unsigned short family)
920 {
921 	struct xfrm_state *x;
922 	struct xfrm_state_walk *w;
923 
924 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
925 	list_for_each_entry(w, &net->xfrm.state_all, all) {
926 		x = container_of(w, struct xfrm_state, km);
927 		if (x->props.family != family ||
928 			x->id.spi != spi)
929 			continue;
930 
931 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
932 		xfrm_state_hold(x);
933 		return x;
934 	}
935 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
936 	return NULL;
937 }
938 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
939 
940 static void __xfrm_state_insert(struct xfrm_state *x)
941 {
942 	struct net *net = xs_net(x);
943 	unsigned int h;
944 
945 	list_add(&x->km.all, &net->xfrm.state_all);
946 
947 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
948 			  x->props.reqid, x->props.family);
949 	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
950 
951 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
952 	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
953 
954 	if (x->id.spi) {
955 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
956 				  x->props.family);
957 
958 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
959 	}
960 
961 	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
962 	if (x->replay_maxage)
963 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
964 
965 	net->xfrm.state_num++;
966 
967 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
968 }
969 
970 /* net->xfrm.xfrm_state_lock is held */
971 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
972 {
973 	struct net *net = xs_net(xnew);
974 	unsigned short family = xnew->props.family;
975 	u32 reqid = xnew->props.reqid;
976 	struct xfrm_state *x;
977 	unsigned int h;
978 	u32 mark = xnew->mark.v & xnew->mark.m;
979 
980 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
981 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
982 		if (x->props.family	== family &&
983 		    x->props.reqid	== reqid &&
984 		    (mark & x->mark.m) == x->mark.v &&
985 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
986 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
987 			x->genid++;
988 	}
989 }
990 
991 void xfrm_state_insert(struct xfrm_state *x)
992 {
993 	struct net *net = xs_net(x);
994 
995 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
996 	__xfrm_state_bump_genids(x);
997 	__xfrm_state_insert(x);
998 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
999 }
1000 EXPORT_SYMBOL(xfrm_state_insert);
1001 
1002 /* net->xfrm.xfrm_state_lock is held */
1003 static struct xfrm_state *__find_acq_core(struct net *net,
1004 					  const struct xfrm_mark *m,
1005 					  unsigned short family, u8 mode,
1006 					  u32 reqid, u8 proto,
1007 					  const xfrm_address_t *daddr,
1008 					  const xfrm_address_t *saddr,
1009 					  int create)
1010 {
1011 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1012 	struct xfrm_state *x;
1013 	u32 mark = m->v & m->m;
1014 
1015 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1016 		if (x->props.reqid  != reqid ||
1017 		    x->props.mode   != mode ||
1018 		    x->props.family != family ||
1019 		    x->km.state     != XFRM_STATE_ACQ ||
1020 		    x->id.spi       != 0 ||
1021 		    x->id.proto	    != proto ||
1022 		    (mark & x->mark.m) != x->mark.v ||
1023 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1024 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1025 			continue;
1026 
1027 		xfrm_state_hold(x);
1028 		return x;
1029 	}
1030 
1031 	if (!create)
1032 		return NULL;
1033 
1034 	x = xfrm_state_alloc(net);
1035 	if (likely(x)) {
1036 		switch (family) {
1037 		case AF_INET:
1038 			x->sel.daddr.a4 = daddr->a4;
1039 			x->sel.saddr.a4 = saddr->a4;
1040 			x->sel.prefixlen_d = 32;
1041 			x->sel.prefixlen_s = 32;
1042 			x->props.saddr.a4 = saddr->a4;
1043 			x->id.daddr.a4 = daddr->a4;
1044 			break;
1045 
1046 		case AF_INET6:
1047 			*(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1048 			*(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1049 			x->sel.prefixlen_d = 128;
1050 			x->sel.prefixlen_s = 128;
1051 			*(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1052 			*(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1053 			break;
1054 		}
1055 
1056 		x->km.state = XFRM_STATE_ACQ;
1057 		x->id.proto = proto;
1058 		x->props.family = family;
1059 		x->props.mode = mode;
1060 		x->props.reqid = reqid;
1061 		x->mark.v = m->v;
1062 		x->mark.m = m->m;
1063 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1064 		xfrm_state_hold(x);
1065 		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1066 		list_add(&x->km.all, &net->xfrm.state_all);
1067 		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1068 		h = xfrm_src_hash(net, daddr, saddr, family);
1069 		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1070 
1071 		net->xfrm.state_num++;
1072 
1073 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1074 	}
1075 
1076 	return x;
1077 }
1078 
1079 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1080 
1081 int xfrm_state_add(struct xfrm_state *x)
1082 {
1083 	struct net *net = xs_net(x);
1084 	struct xfrm_state *x1, *to_put;
1085 	int family;
1086 	int err;
1087 	u32 mark = x->mark.v & x->mark.m;
1088 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1089 
1090 	family = x->props.family;
1091 
1092 	to_put = NULL;
1093 
1094 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1095 
1096 	x1 = __xfrm_state_locate(x, use_spi, family);
1097 	if (x1) {
1098 		to_put = x1;
1099 		x1 = NULL;
1100 		err = -EEXIST;
1101 		goto out;
1102 	}
1103 
1104 	if (use_spi && x->km.seq) {
1105 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1106 		if (x1 && ((x1->id.proto != x->id.proto) ||
1107 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1108 			to_put = x1;
1109 			x1 = NULL;
1110 		}
1111 	}
1112 
1113 	if (use_spi && !x1)
1114 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1115 				     x->props.reqid, x->id.proto,
1116 				     &x->id.daddr, &x->props.saddr, 0);
1117 
1118 	__xfrm_state_bump_genids(x);
1119 	__xfrm_state_insert(x);
1120 	err = 0;
1121 
1122 out:
1123 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1124 
1125 	if (x1) {
1126 		xfrm_state_delete(x1);
1127 		xfrm_state_put(x1);
1128 	}
1129 
1130 	if (to_put)
1131 		xfrm_state_put(to_put);
1132 
1133 	return err;
1134 }
1135 EXPORT_SYMBOL(xfrm_state_add);
1136 
1137 #ifdef CONFIG_XFRM_MIGRATE
1138 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1139 {
1140 	struct net *net = xs_net(orig);
1141 	int err = -ENOMEM;
1142 	struct xfrm_state *x = xfrm_state_alloc(net);
1143 	if (!x)
1144 		goto out;
1145 
1146 	memcpy(&x->id, &orig->id, sizeof(x->id));
1147 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1148 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1149 	x->props.mode = orig->props.mode;
1150 	x->props.replay_window = orig->props.replay_window;
1151 	x->props.reqid = orig->props.reqid;
1152 	x->props.family = orig->props.family;
1153 	x->props.saddr = orig->props.saddr;
1154 
1155 	if (orig->aalg) {
1156 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1157 		if (!x->aalg)
1158 			goto error;
1159 	}
1160 	x->props.aalgo = orig->props.aalgo;
1161 
1162 	if (orig->ealg) {
1163 		x->ealg = xfrm_algo_clone(orig->ealg);
1164 		if (!x->ealg)
1165 			goto error;
1166 	}
1167 	x->props.ealgo = orig->props.ealgo;
1168 
1169 	if (orig->calg) {
1170 		x->calg = xfrm_algo_clone(orig->calg);
1171 		if (!x->calg)
1172 			goto error;
1173 	}
1174 	x->props.calgo = orig->props.calgo;
1175 
1176 	if (orig->encap) {
1177 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1178 		if (!x->encap)
1179 			goto error;
1180 	}
1181 
1182 	if (orig->coaddr) {
1183 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1184 				    GFP_KERNEL);
1185 		if (!x->coaddr)
1186 			goto error;
1187 	}
1188 
1189 	if (orig->replay_esn) {
1190 		err = xfrm_replay_clone(x, orig);
1191 		if (err)
1192 			goto error;
1193 	}
1194 
1195 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1196 
1197 	err = xfrm_init_state(x);
1198 	if (err)
1199 		goto error;
1200 
1201 	x->props.flags = orig->props.flags;
1202 	x->props.extra_flags = orig->props.extra_flags;
1203 
1204 	x->curlft.add_time = orig->curlft.add_time;
1205 	x->km.state = orig->km.state;
1206 	x->km.seq = orig->km.seq;
1207 
1208 	return x;
1209 
1210  error:
1211 	xfrm_state_put(x);
1212 out:
1213 	if (errp)
1214 		*errp = err;
1215 	return NULL;
1216 }
1217 
1218 /* net->xfrm.xfrm_state_lock is held */
1219 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1220 {
1221 	unsigned int h;
1222 	struct xfrm_state *x;
1223 
1224 	if (m->reqid) {
1225 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1226 				  m->reqid, m->old_family);
1227 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1228 			if (x->props.mode != m->mode ||
1229 			    x->id.proto != m->proto)
1230 				continue;
1231 			if (m->reqid && x->props.reqid != m->reqid)
1232 				continue;
1233 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1234 					     m->old_family) ||
1235 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1236 					     m->old_family))
1237 				continue;
1238 			xfrm_state_hold(x);
1239 			return x;
1240 		}
1241 	} else {
1242 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1243 				  m->old_family);
1244 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1245 			if (x->props.mode != m->mode ||
1246 			    x->id.proto != m->proto)
1247 				continue;
1248 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1249 					     m->old_family) ||
1250 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1251 					     m->old_family))
1252 				continue;
1253 			xfrm_state_hold(x);
1254 			return x;
1255 		}
1256 	}
1257 
1258 	return NULL;
1259 }
1260 EXPORT_SYMBOL(xfrm_migrate_state_find);
1261 
1262 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1263 				      struct xfrm_migrate *m)
1264 {
1265 	struct xfrm_state *xc;
1266 	int err;
1267 
1268 	xc = xfrm_state_clone(x, &err);
1269 	if (!xc)
1270 		return NULL;
1271 
1272 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1273 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1274 
1275 	/* add state */
1276 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1277 		/* a care is needed when the destination address of the
1278 		   state is to be updated as it is a part of triplet */
1279 		xfrm_state_insert(xc);
1280 	} else {
1281 		if ((err = xfrm_state_add(xc)) < 0)
1282 			goto error;
1283 	}
1284 
1285 	return xc;
1286 error:
1287 	xfrm_state_put(xc);
1288 	return NULL;
1289 }
1290 EXPORT_SYMBOL(xfrm_state_migrate);
1291 #endif
1292 
1293 int xfrm_state_update(struct xfrm_state *x)
1294 {
1295 	struct xfrm_state *x1, *to_put;
1296 	int err;
1297 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1298 	struct net *net = xs_net(x);
1299 
1300 	to_put = NULL;
1301 
1302 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1303 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1304 
1305 	err = -ESRCH;
1306 	if (!x1)
1307 		goto out;
1308 
1309 	if (xfrm_state_kern(x1)) {
1310 		to_put = x1;
1311 		err = -EEXIST;
1312 		goto out;
1313 	}
1314 
1315 	if (x1->km.state == XFRM_STATE_ACQ) {
1316 		__xfrm_state_insert(x);
1317 		x = NULL;
1318 	}
1319 	err = 0;
1320 
1321 out:
1322 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1323 
1324 	if (to_put)
1325 		xfrm_state_put(to_put);
1326 
1327 	if (err)
1328 		return err;
1329 
1330 	if (!x) {
1331 		xfrm_state_delete(x1);
1332 		xfrm_state_put(x1);
1333 		return 0;
1334 	}
1335 
1336 	err = -EINVAL;
1337 	spin_lock_bh(&x1->lock);
1338 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1339 		if (x->encap && x1->encap)
1340 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1341 		if (x->coaddr && x1->coaddr) {
1342 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1343 		}
1344 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1345 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1346 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1347 		x1->km.dying = 0;
1348 
1349 		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1350 		if (x1->curlft.use_time)
1351 			xfrm_state_check_expire(x1);
1352 
1353 		err = 0;
1354 		x->km.state = XFRM_STATE_DEAD;
1355 		__xfrm_state_put(x);
1356 	}
1357 	spin_unlock_bh(&x1->lock);
1358 
1359 	xfrm_state_put(x1);
1360 
1361 	return err;
1362 }
1363 EXPORT_SYMBOL(xfrm_state_update);
1364 
1365 int xfrm_state_check_expire(struct xfrm_state *x)
1366 {
1367 	if (!x->curlft.use_time)
1368 		x->curlft.use_time = get_seconds();
1369 
1370 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1371 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1372 		x->km.state = XFRM_STATE_EXPIRED;
1373 		tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1374 		return -EINVAL;
1375 	}
1376 
1377 	if (!x->km.dying &&
1378 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1379 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1380 		x->km.dying = 1;
1381 		km_state_expired(x, 0, 0);
1382 	}
1383 	return 0;
1384 }
1385 EXPORT_SYMBOL(xfrm_state_check_expire);
1386 
1387 struct xfrm_state *
1388 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1389 		  u8 proto, unsigned short family)
1390 {
1391 	struct xfrm_state *x;
1392 
1393 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1394 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1395 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1396 	return x;
1397 }
1398 EXPORT_SYMBOL(xfrm_state_lookup);
1399 
1400 struct xfrm_state *
1401 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1402 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1403 			 u8 proto, unsigned short family)
1404 {
1405 	struct xfrm_state *x;
1406 
1407 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1408 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1409 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1410 	return x;
1411 }
1412 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1413 
1414 struct xfrm_state *
1415 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1416 	      u8 proto, const xfrm_address_t *daddr,
1417 	      const xfrm_address_t *saddr, int create, unsigned short family)
1418 {
1419 	struct xfrm_state *x;
1420 
1421 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1422 	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1423 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1424 
1425 	return x;
1426 }
1427 EXPORT_SYMBOL(xfrm_find_acq);
1428 
1429 #ifdef CONFIG_XFRM_SUB_POLICY
1430 int
1431 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1432 	       unsigned short family, struct net *net)
1433 {
1434 	int err = 0;
1435 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1436 	if (!afinfo)
1437 		return -EAFNOSUPPORT;
1438 
1439 	spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1440 	if (afinfo->tmpl_sort)
1441 		err = afinfo->tmpl_sort(dst, src, n);
1442 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1443 	xfrm_state_put_afinfo(afinfo);
1444 	return err;
1445 }
1446 EXPORT_SYMBOL(xfrm_tmpl_sort);
1447 
1448 int
1449 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1450 		unsigned short family)
1451 {
1452 	int err = 0;
1453 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1454 	struct net *net = xs_net(*dst);
1455 
1456 	if (!afinfo)
1457 		return -EAFNOSUPPORT;
1458 
1459 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1460 	if (afinfo->state_sort)
1461 		err = afinfo->state_sort(dst, src, n);
1462 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1463 	xfrm_state_put_afinfo(afinfo);
1464 	return err;
1465 }
1466 EXPORT_SYMBOL(xfrm_state_sort);
1467 #endif
1468 
1469 /* Silly enough, but I'm lazy to build resolution list */
1470 
1471 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1472 {
1473 	int i;
1474 
1475 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1476 		struct xfrm_state *x;
1477 
1478 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1479 			if (x->km.seq == seq &&
1480 			    (mark & x->mark.m) == x->mark.v &&
1481 			    x->km.state == XFRM_STATE_ACQ) {
1482 				xfrm_state_hold(x);
1483 				return x;
1484 			}
1485 		}
1486 	}
1487 	return NULL;
1488 }
1489 
1490 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1491 {
1492 	struct xfrm_state *x;
1493 
1494 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1495 	x = __xfrm_find_acq_byseq(net, mark, seq);
1496 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1497 	return x;
1498 }
1499 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1500 
1501 u32 xfrm_get_acqseq(void)
1502 {
1503 	u32 res;
1504 	static atomic_t acqseq;
1505 
1506 	do {
1507 		res = atomic_inc_return(&acqseq);
1508 	} while (!res);
1509 
1510 	return res;
1511 }
1512 EXPORT_SYMBOL(xfrm_get_acqseq);
1513 
1514 int verify_spi_info(u8 proto, u32 min, u32 max)
1515 {
1516 	switch (proto) {
1517 	case IPPROTO_AH:
1518 	case IPPROTO_ESP:
1519 		break;
1520 
1521 	case IPPROTO_COMP:
1522 		/* IPCOMP spi is 16-bits. */
1523 		if (max >= 0x10000)
1524 			return -EINVAL;
1525 		break;
1526 
1527 	default:
1528 		return -EINVAL;
1529 	}
1530 
1531 	if (min > max)
1532 		return -EINVAL;
1533 
1534 	return 0;
1535 }
1536 EXPORT_SYMBOL(verify_spi_info);
1537 
1538 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1539 {
1540 	struct net *net = xs_net(x);
1541 	unsigned int h;
1542 	struct xfrm_state *x0;
1543 	int err = -ENOENT;
1544 	__be32 minspi = htonl(low);
1545 	__be32 maxspi = htonl(high);
1546 	u32 mark = x->mark.v & x->mark.m;
1547 
1548 	spin_lock_bh(&x->lock);
1549 	if (x->km.state == XFRM_STATE_DEAD)
1550 		goto unlock;
1551 
1552 	err = 0;
1553 	if (x->id.spi)
1554 		goto unlock;
1555 
1556 	err = -ENOENT;
1557 
1558 	if (minspi == maxspi) {
1559 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1560 		if (x0) {
1561 			xfrm_state_put(x0);
1562 			goto unlock;
1563 		}
1564 		x->id.spi = minspi;
1565 	} else {
1566 		u32 spi = 0;
1567 		for (h = 0; h < high-low+1; h++) {
1568 			spi = low + prandom_u32()%(high-low+1);
1569 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1570 			if (x0 == NULL) {
1571 				x->id.spi = htonl(spi);
1572 				break;
1573 			}
1574 			xfrm_state_put(x0);
1575 		}
1576 	}
1577 	if (x->id.spi) {
1578 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
1579 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1580 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1581 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1582 
1583 		err = 0;
1584 	}
1585 
1586 unlock:
1587 	spin_unlock_bh(&x->lock);
1588 
1589 	return err;
1590 }
1591 EXPORT_SYMBOL(xfrm_alloc_spi);
1592 
1593 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1594 		    int (*func)(struct xfrm_state *, int, void*),
1595 		    void *data)
1596 {
1597 	struct xfrm_state *state;
1598 	struct xfrm_state_walk *x;
1599 	int err = 0;
1600 
1601 	if (walk->seq != 0 && list_empty(&walk->all))
1602 		return 0;
1603 
1604 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1605 	if (list_empty(&walk->all))
1606 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1607 	else
1608 		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1609 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1610 		if (x->state == XFRM_STATE_DEAD)
1611 			continue;
1612 		state = container_of(x, struct xfrm_state, km);
1613 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1614 			continue;
1615 		err = func(state, walk->seq, data);
1616 		if (err) {
1617 			list_move_tail(&walk->all, &x->all);
1618 			goto out;
1619 		}
1620 		walk->seq++;
1621 	}
1622 	if (walk->seq == 0) {
1623 		err = -ENOENT;
1624 		goto out;
1625 	}
1626 	list_del_init(&walk->all);
1627 out:
1628 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1629 	return err;
1630 }
1631 EXPORT_SYMBOL(xfrm_state_walk);
1632 
1633 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1634 {
1635 	INIT_LIST_HEAD(&walk->all);
1636 	walk->proto = proto;
1637 	walk->state = XFRM_STATE_DEAD;
1638 	walk->seq = 0;
1639 }
1640 EXPORT_SYMBOL(xfrm_state_walk_init);
1641 
1642 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1643 {
1644 	if (list_empty(&walk->all))
1645 		return;
1646 
1647 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1648 	list_del(&walk->all);
1649 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1650 }
1651 EXPORT_SYMBOL(xfrm_state_walk_done);
1652 
1653 static void xfrm_replay_timer_handler(unsigned long data)
1654 {
1655 	struct xfrm_state *x = (struct xfrm_state *)data;
1656 
1657 	spin_lock(&x->lock);
1658 
1659 	if (x->km.state == XFRM_STATE_VALID) {
1660 		if (xfrm_aevent_is_on(xs_net(x)))
1661 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1662 		else
1663 			x->xflags |= XFRM_TIME_DEFER;
1664 	}
1665 
1666 	spin_unlock(&x->lock);
1667 }
1668 
1669 static LIST_HEAD(xfrm_km_list);
1670 
1671 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1672 {
1673 	struct xfrm_mgr *km;
1674 
1675 	rcu_read_lock();
1676 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1677 		if (km->notify_policy)
1678 			km->notify_policy(xp, dir, c);
1679 	rcu_read_unlock();
1680 }
1681 
1682 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1683 {
1684 	struct xfrm_mgr *km;
1685 	rcu_read_lock();
1686 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1687 		if (km->notify)
1688 			km->notify(x, c);
1689 	rcu_read_unlock();
1690 }
1691 
1692 EXPORT_SYMBOL(km_policy_notify);
1693 EXPORT_SYMBOL(km_state_notify);
1694 
1695 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1696 {
1697 	struct km_event c;
1698 
1699 	c.data.hard = hard;
1700 	c.portid = portid;
1701 	c.event = XFRM_MSG_EXPIRE;
1702 	km_state_notify(x, &c);
1703 }
1704 
1705 EXPORT_SYMBOL(km_state_expired);
1706 /*
1707  * We send to all registered managers regardless of failure
1708  * We are happy with one success
1709 */
1710 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1711 {
1712 	int err = -EINVAL, acqret;
1713 	struct xfrm_mgr *km;
1714 
1715 	rcu_read_lock();
1716 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1717 		acqret = km->acquire(x, t, pol);
1718 		if (!acqret)
1719 			err = acqret;
1720 	}
1721 	rcu_read_unlock();
1722 	return err;
1723 }
1724 EXPORT_SYMBOL(km_query);
1725 
1726 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1727 {
1728 	int err = -EINVAL;
1729 	struct xfrm_mgr *km;
1730 
1731 	rcu_read_lock();
1732 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1733 		if (km->new_mapping)
1734 			err = km->new_mapping(x, ipaddr, sport);
1735 		if (!err)
1736 			break;
1737 	}
1738 	rcu_read_unlock();
1739 	return err;
1740 }
1741 EXPORT_SYMBOL(km_new_mapping);
1742 
1743 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1744 {
1745 	struct km_event c;
1746 
1747 	c.data.hard = hard;
1748 	c.portid = portid;
1749 	c.event = XFRM_MSG_POLEXPIRE;
1750 	km_policy_notify(pol, dir, &c);
1751 }
1752 EXPORT_SYMBOL(km_policy_expired);
1753 
1754 #ifdef CONFIG_XFRM_MIGRATE
1755 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1756 	       const struct xfrm_migrate *m, int num_migrate,
1757 	       const struct xfrm_kmaddress *k)
1758 {
1759 	int err = -EINVAL;
1760 	int ret;
1761 	struct xfrm_mgr *km;
1762 
1763 	rcu_read_lock();
1764 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1765 		if (km->migrate) {
1766 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1767 			if (!ret)
1768 				err = ret;
1769 		}
1770 	}
1771 	rcu_read_unlock();
1772 	return err;
1773 }
1774 EXPORT_SYMBOL(km_migrate);
1775 #endif
1776 
1777 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1778 {
1779 	int err = -EINVAL;
1780 	int ret;
1781 	struct xfrm_mgr *km;
1782 
1783 	rcu_read_lock();
1784 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1785 		if (km->report) {
1786 			ret = km->report(net, proto, sel, addr);
1787 			if (!ret)
1788 				err = ret;
1789 		}
1790 	}
1791 	rcu_read_unlock();
1792 	return err;
1793 }
1794 EXPORT_SYMBOL(km_report);
1795 
1796 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1797 {
1798 	int err;
1799 	u8 *data;
1800 	struct xfrm_mgr *km;
1801 	struct xfrm_policy *pol = NULL;
1802 
1803 	if (optlen <= 0 || optlen > PAGE_SIZE)
1804 		return -EMSGSIZE;
1805 
1806 	data = kmalloc(optlen, GFP_KERNEL);
1807 	if (!data)
1808 		return -ENOMEM;
1809 
1810 	err = -EFAULT;
1811 	if (copy_from_user(data, optval, optlen))
1812 		goto out;
1813 
1814 	err = -EINVAL;
1815 	rcu_read_lock();
1816 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1817 		pol = km->compile_policy(sk, optname, data,
1818 					 optlen, &err);
1819 		if (err >= 0)
1820 			break;
1821 	}
1822 	rcu_read_unlock();
1823 
1824 	if (err >= 0) {
1825 		xfrm_sk_policy_insert(sk, err, pol);
1826 		xfrm_pol_put(pol);
1827 		err = 0;
1828 	}
1829 
1830 out:
1831 	kfree(data);
1832 	return err;
1833 }
1834 EXPORT_SYMBOL(xfrm_user_policy);
1835 
1836 static DEFINE_SPINLOCK(xfrm_km_lock);
1837 
1838 int xfrm_register_km(struct xfrm_mgr *km)
1839 {
1840 	spin_lock_bh(&xfrm_km_lock);
1841 	list_add_tail_rcu(&km->list, &xfrm_km_list);
1842 	spin_unlock_bh(&xfrm_km_lock);
1843 	return 0;
1844 }
1845 EXPORT_SYMBOL(xfrm_register_km);
1846 
1847 int xfrm_unregister_km(struct xfrm_mgr *km)
1848 {
1849 	spin_lock_bh(&xfrm_km_lock);
1850 	list_del_rcu(&km->list);
1851 	spin_unlock_bh(&xfrm_km_lock);
1852 	synchronize_rcu();
1853 	return 0;
1854 }
1855 EXPORT_SYMBOL(xfrm_unregister_km);
1856 
1857 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1858 {
1859 	int err = 0;
1860 	if (unlikely(afinfo == NULL))
1861 		return -EINVAL;
1862 	if (unlikely(afinfo->family >= NPROTO))
1863 		return -EAFNOSUPPORT;
1864 	spin_lock_bh(&xfrm_state_afinfo_lock);
1865 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1866 		err = -ENOBUFS;
1867 	else
1868 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1869 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1870 	return err;
1871 }
1872 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1873 
1874 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1875 {
1876 	int err = 0;
1877 	if (unlikely(afinfo == NULL))
1878 		return -EINVAL;
1879 	if (unlikely(afinfo->family >= NPROTO))
1880 		return -EAFNOSUPPORT;
1881 	spin_lock_bh(&xfrm_state_afinfo_lock);
1882 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1883 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1884 			err = -EINVAL;
1885 		else
1886 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1887 	}
1888 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1889 	synchronize_rcu();
1890 	return err;
1891 }
1892 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1893 
1894 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1895 {
1896 	struct xfrm_state_afinfo *afinfo;
1897 	if (unlikely(family >= NPROTO))
1898 		return NULL;
1899 	rcu_read_lock();
1900 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1901 	if (unlikely(!afinfo))
1902 		rcu_read_unlock();
1903 	return afinfo;
1904 }
1905 
1906 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1907 {
1908 	rcu_read_unlock();
1909 }
1910 
1911 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1912 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1913 {
1914 	if (x->tunnel) {
1915 		struct xfrm_state *t = x->tunnel;
1916 
1917 		if (atomic_read(&t->tunnel_users) == 2)
1918 			xfrm_state_delete(t);
1919 		atomic_dec(&t->tunnel_users);
1920 		xfrm_state_put(t);
1921 		x->tunnel = NULL;
1922 	}
1923 }
1924 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1925 
1926 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1927 {
1928 	int res;
1929 
1930 	spin_lock_bh(&x->lock);
1931 	if (x->km.state == XFRM_STATE_VALID &&
1932 	    x->type && x->type->get_mtu)
1933 		res = x->type->get_mtu(x, mtu);
1934 	else
1935 		res = mtu - x->props.header_len;
1936 	spin_unlock_bh(&x->lock);
1937 	return res;
1938 }
1939 
1940 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1941 {
1942 	struct xfrm_state_afinfo *afinfo;
1943 	struct xfrm_mode *inner_mode;
1944 	int family = x->props.family;
1945 	int err;
1946 
1947 	err = -EAFNOSUPPORT;
1948 	afinfo = xfrm_state_get_afinfo(family);
1949 	if (!afinfo)
1950 		goto error;
1951 
1952 	err = 0;
1953 	if (afinfo->init_flags)
1954 		err = afinfo->init_flags(x);
1955 
1956 	xfrm_state_put_afinfo(afinfo);
1957 
1958 	if (err)
1959 		goto error;
1960 
1961 	err = -EPROTONOSUPPORT;
1962 
1963 	if (x->sel.family != AF_UNSPEC) {
1964 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1965 		if (inner_mode == NULL)
1966 			goto error;
1967 
1968 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1969 		    family != x->sel.family) {
1970 			xfrm_put_mode(inner_mode);
1971 			goto error;
1972 		}
1973 
1974 		x->inner_mode = inner_mode;
1975 	} else {
1976 		struct xfrm_mode *inner_mode_iaf;
1977 		int iafamily = AF_INET;
1978 
1979 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1980 		if (inner_mode == NULL)
1981 			goto error;
1982 
1983 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1984 			xfrm_put_mode(inner_mode);
1985 			goto error;
1986 		}
1987 		x->inner_mode = inner_mode;
1988 
1989 		if (x->props.family == AF_INET)
1990 			iafamily = AF_INET6;
1991 
1992 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1993 		if (inner_mode_iaf) {
1994 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1995 				x->inner_mode_iaf = inner_mode_iaf;
1996 			else
1997 				xfrm_put_mode(inner_mode_iaf);
1998 		}
1999 	}
2000 
2001 	x->type = xfrm_get_type(x->id.proto, family);
2002 	if (x->type == NULL)
2003 		goto error;
2004 
2005 	err = x->type->init_state(x);
2006 	if (err)
2007 		goto error;
2008 
2009 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2010 	if (x->outer_mode == NULL) {
2011 		err = -EPROTONOSUPPORT;
2012 		goto error;
2013 	}
2014 
2015 	if (init_replay) {
2016 		err = xfrm_init_replay(x);
2017 		if (err)
2018 			goto error;
2019 	}
2020 
2021 	x->km.state = XFRM_STATE_VALID;
2022 
2023 error:
2024 	return err;
2025 }
2026 
2027 EXPORT_SYMBOL(__xfrm_init_state);
2028 
2029 int xfrm_init_state(struct xfrm_state *x)
2030 {
2031 	return __xfrm_init_state(x, true);
2032 }
2033 
2034 EXPORT_SYMBOL(xfrm_init_state);
2035 
2036 int __net_init xfrm_state_init(struct net *net)
2037 {
2038 	unsigned int sz;
2039 
2040 	INIT_LIST_HEAD(&net->xfrm.state_all);
2041 
2042 	sz = sizeof(struct hlist_head) * 8;
2043 
2044 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2045 	if (!net->xfrm.state_bydst)
2046 		goto out_bydst;
2047 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2048 	if (!net->xfrm.state_bysrc)
2049 		goto out_bysrc;
2050 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2051 	if (!net->xfrm.state_byspi)
2052 		goto out_byspi;
2053 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2054 
2055 	net->xfrm.state_num = 0;
2056 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2057 	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2058 	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2059 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2060 	return 0;
2061 
2062 out_byspi:
2063 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2064 out_bysrc:
2065 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2066 out_bydst:
2067 	return -ENOMEM;
2068 }
2069 
2070 void xfrm_state_fini(struct net *net)
2071 {
2072 	struct xfrm_audit audit_info;
2073 	unsigned int sz;
2074 
2075 	flush_work(&net->xfrm.state_hash_work);
2076 	audit_info.loginuid = INVALID_UID;
2077 	audit_info.sessionid = (unsigned int)-1;
2078 	audit_info.secid = 0;
2079 	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2080 	flush_work(&net->xfrm.state_gc_work);
2081 
2082 	WARN_ON(!list_empty(&net->xfrm.state_all));
2083 
2084 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2085 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2086 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2087 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2088 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2089 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2090 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2091 }
2092 
2093 #ifdef CONFIG_AUDITSYSCALL
2094 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2095 				     struct audit_buffer *audit_buf)
2096 {
2097 	struct xfrm_sec_ctx *ctx = x->security;
2098 	u32 spi = ntohl(x->id.spi);
2099 
2100 	if (ctx)
2101 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2102 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2103 
2104 	switch (x->props.family) {
2105 	case AF_INET:
2106 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2107 				 &x->props.saddr.a4, &x->id.daddr.a4);
2108 		break;
2109 	case AF_INET6:
2110 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2111 				 x->props.saddr.a6, x->id.daddr.a6);
2112 		break;
2113 	}
2114 
2115 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2116 }
2117 
2118 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2119 				      struct audit_buffer *audit_buf)
2120 {
2121 	const struct iphdr *iph4;
2122 	const struct ipv6hdr *iph6;
2123 
2124 	switch (family) {
2125 	case AF_INET:
2126 		iph4 = ip_hdr(skb);
2127 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2128 				 &iph4->saddr, &iph4->daddr);
2129 		break;
2130 	case AF_INET6:
2131 		iph6 = ipv6_hdr(skb);
2132 		audit_log_format(audit_buf,
2133 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2134 				 &iph6->saddr, &iph6->daddr,
2135 				 iph6->flow_lbl[0] & 0x0f,
2136 				 iph6->flow_lbl[1],
2137 				 iph6->flow_lbl[2]);
2138 		break;
2139 	}
2140 }
2141 
2142 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2143 			  kuid_t auid, unsigned int sessionid, u32 secid)
2144 {
2145 	struct audit_buffer *audit_buf;
2146 
2147 	audit_buf = xfrm_audit_start("SAD-add");
2148 	if (audit_buf == NULL)
2149 		return;
2150 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2151 	xfrm_audit_helper_sainfo(x, audit_buf);
2152 	audit_log_format(audit_buf, " res=%u", result);
2153 	audit_log_end(audit_buf);
2154 }
2155 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2156 
2157 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2158 			     kuid_t auid, unsigned int sessionid, u32 secid)
2159 {
2160 	struct audit_buffer *audit_buf;
2161 
2162 	audit_buf = xfrm_audit_start("SAD-delete");
2163 	if (audit_buf == NULL)
2164 		return;
2165 	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2166 	xfrm_audit_helper_sainfo(x, audit_buf);
2167 	audit_log_format(audit_buf, " res=%u", result);
2168 	audit_log_end(audit_buf);
2169 }
2170 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2171 
2172 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2173 				      struct sk_buff *skb)
2174 {
2175 	struct audit_buffer *audit_buf;
2176 	u32 spi;
2177 
2178 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2179 	if (audit_buf == NULL)
2180 		return;
2181 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2182 	/* don't record the sequence number because it's inherent in this kind
2183 	 * of audit message */
2184 	spi = ntohl(x->id.spi);
2185 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2186 	audit_log_end(audit_buf);
2187 }
2188 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2189 
2190 void xfrm_audit_state_replay(struct xfrm_state *x,
2191 			     struct sk_buff *skb, __be32 net_seq)
2192 {
2193 	struct audit_buffer *audit_buf;
2194 	u32 spi;
2195 
2196 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2197 	if (audit_buf == NULL)
2198 		return;
2199 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2200 	spi = ntohl(x->id.spi);
2201 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2202 			 spi, spi, ntohl(net_seq));
2203 	audit_log_end(audit_buf);
2204 }
2205 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2206 
2207 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2208 {
2209 	struct audit_buffer *audit_buf;
2210 
2211 	audit_buf = xfrm_audit_start("SA-notfound");
2212 	if (audit_buf == NULL)
2213 		return;
2214 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2215 	audit_log_end(audit_buf);
2216 }
2217 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2218 
2219 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2220 			       __be32 net_spi, __be32 net_seq)
2221 {
2222 	struct audit_buffer *audit_buf;
2223 	u32 spi;
2224 
2225 	audit_buf = xfrm_audit_start("SA-notfound");
2226 	if (audit_buf == NULL)
2227 		return;
2228 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2229 	spi = ntohl(net_spi);
2230 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2231 			 spi, spi, ntohl(net_seq));
2232 	audit_log_end(audit_buf);
2233 }
2234 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2235 
2236 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2237 			      struct sk_buff *skb, u8 proto)
2238 {
2239 	struct audit_buffer *audit_buf;
2240 	__be32 net_spi;
2241 	__be32 net_seq;
2242 
2243 	audit_buf = xfrm_audit_start("SA-icv-failure");
2244 	if (audit_buf == NULL)
2245 		return;
2246 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2247 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2248 		u32 spi = ntohl(net_spi);
2249 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2250 				 spi, spi, ntohl(net_seq));
2251 	}
2252 	audit_log_end(audit_buf);
2253 }
2254 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2255 #endif /* CONFIG_AUDITSYSCALL */
2256