xref: /openbmc/linux/net/xfrm/xfrm_state.c (revision 2c684d89)
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 void xfrm_hash_resize(struct work_struct *work)
101 {
102 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
103 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
104 	unsigned long nsize, osize;
105 	unsigned int nhashmask, ohashmask;
106 	int i;
107 
108 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
109 	ndst = xfrm_hash_alloc(nsize);
110 	if (!ndst)
111 		return;
112 	nsrc = xfrm_hash_alloc(nsize);
113 	if (!nsrc) {
114 		xfrm_hash_free(ndst, nsize);
115 		return;
116 	}
117 	nspi = xfrm_hash_alloc(nsize);
118 	if (!nspi) {
119 		xfrm_hash_free(ndst, nsize);
120 		xfrm_hash_free(nsrc, nsize);
121 		return;
122 	}
123 
124 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
125 
126 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
127 	for (i = net->xfrm.state_hmask; i >= 0; i--)
128 		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
129 				   nhashmask);
130 
131 	odst = net->xfrm.state_bydst;
132 	osrc = net->xfrm.state_bysrc;
133 	ospi = net->xfrm.state_byspi;
134 	ohashmask = net->xfrm.state_hmask;
135 
136 	net->xfrm.state_bydst = ndst;
137 	net->xfrm.state_bysrc = nsrc;
138 	net->xfrm.state_byspi = nspi;
139 	net->xfrm.state_hmask = nhashmask;
140 
141 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
142 
143 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
144 	xfrm_hash_free(odst, osize);
145 	xfrm_hash_free(osrc, osize);
146 	xfrm_hash_free(ospi, osize);
147 }
148 
149 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
150 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
151 
152 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
153 
154 int __xfrm_state_delete(struct xfrm_state *x);
155 
156 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
157 bool km_is_alive(const struct km_event *c);
158 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
159 
160 static DEFINE_SPINLOCK(xfrm_type_lock);
161 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
162 {
163 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
164 	const struct xfrm_type **typemap;
165 	int err = 0;
166 
167 	if (unlikely(afinfo == NULL))
168 		return -EAFNOSUPPORT;
169 	typemap = afinfo->type_map;
170 	spin_lock_bh(&xfrm_type_lock);
171 
172 	if (likely(typemap[type->proto] == NULL))
173 		typemap[type->proto] = type;
174 	else
175 		err = -EEXIST;
176 	spin_unlock_bh(&xfrm_type_lock);
177 	xfrm_state_put_afinfo(afinfo);
178 	return err;
179 }
180 EXPORT_SYMBOL(xfrm_register_type);
181 
182 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
183 {
184 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
185 	const struct xfrm_type **typemap;
186 	int err = 0;
187 
188 	if (unlikely(afinfo == NULL))
189 		return -EAFNOSUPPORT;
190 	typemap = afinfo->type_map;
191 	spin_lock_bh(&xfrm_type_lock);
192 
193 	if (unlikely(typemap[type->proto] != type))
194 		err = -ENOENT;
195 	else
196 		typemap[type->proto] = NULL;
197 	spin_unlock_bh(&xfrm_type_lock);
198 	xfrm_state_put_afinfo(afinfo);
199 	return err;
200 }
201 EXPORT_SYMBOL(xfrm_unregister_type);
202 
203 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
204 {
205 	struct xfrm_state_afinfo *afinfo;
206 	const struct xfrm_type **typemap;
207 	const struct xfrm_type *type;
208 	int modload_attempted = 0;
209 
210 retry:
211 	afinfo = xfrm_state_get_afinfo(family);
212 	if (unlikely(afinfo == NULL))
213 		return NULL;
214 	typemap = afinfo->type_map;
215 
216 	type = typemap[proto];
217 	if (unlikely(type && !try_module_get(type->owner)))
218 		type = NULL;
219 	if (!type && !modload_attempted) {
220 		xfrm_state_put_afinfo(afinfo);
221 		request_module("xfrm-type-%d-%d", family, proto);
222 		modload_attempted = 1;
223 		goto retry;
224 	}
225 
226 	xfrm_state_put_afinfo(afinfo);
227 	return type;
228 }
229 
230 static void xfrm_put_type(const struct xfrm_type *type)
231 {
232 	module_put(type->owner);
233 }
234 
235 static DEFINE_SPINLOCK(xfrm_mode_lock);
236 int xfrm_register_mode(struct xfrm_mode *mode, int family)
237 {
238 	struct xfrm_state_afinfo *afinfo;
239 	struct xfrm_mode **modemap;
240 	int err;
241 
242 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
243 		return -EINVAL;
244 
245 	afinfo = xfrm_state_get_afinfo(family);
246 	if (unlikely(afinfo == NULL))
247 		return -EAFNOSUPPORT;
248 
249 	err = -EEXIST;
250 	modemap = afinfo->mode_map;
251 	spin_lock_bh(&xfrm_mode_lock);
252 	if (modemap[mode->encap])
253 		goto out;
254 
255 	err = -ENOENT;
256 	if (!try_module_get(afinfo->owner))
257 		goto out;
258 
259 	mode->afinfo = afinfo;
260 	modemap[mode->encap] = mode;
261 	err = 0;
262 
263 out:
264 	spin_unlock_bh(&xfrm_mode_lock);
265 	xfrm_state_put_afinfo(afinfo);
266 	return err;
267 }
268 EXPORT_SYMBOL(xfrm_register_mode);
269 
270 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
271 {
272 	struct xfrm_state_afinfo *afinfo;
273 	struct xfrm_mode **modemap;
274 	int err;
275 
276 	if (unlikely(mode->encap >= XFRM_MODE_MAX))
277 		return -EINVAL;
278 
279 	afinfo = xfrm_state_get_afinfo(family);
280 	if (unlikely(afinfo == NULL))
281 		return -EAFNOSUPPORT;
282 
283 	err = -ENOENT;
284 	modemap = afinfo->mode_map;
285 	spin_lock_bh(&xfrm_mode_lock);
286 	if (likely(modemap[mode->encap] == mode)) {
287 		modemap[mode->encap] = NULL;
288 		module_put(mode->afinfo->owner);
289 		err = 0;
290 	}
291 
292 	spin_unlock_bh(&xfrm_mode_lock);
293 	xfrm_state_put_afinfo(afinfo);
294 	return err;
295 }
296 EXPORT_SYMBOL(xfrm_unregister_mode);
297 
298 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
299 {
300 	struct xfrm_state_afinfo *afinfo;
301 	struct xfrm_mode *mode;
302 	int modload_attempted = 0;
303 
304 	if (unlikely(encap >= XFRM_MODE_MAX))
305 		return NULL;
306 
307 retry:
308 	afinfo = xfrm_state_get_afinfo(family);
309 	if (unlikely(afinfo == NULL))
310 		return NULL;
311 
312 	mode = afinfo->mode_map[encap];
313 	if (unlikely(mode && !try_module_get(mode->owner)))
314 		mode = NULL;
315 	if (!mode && !modload_attempted) {
316 		xfrm_state_put_afinfo(afinfo);
317 		request_module("xfrm-mode-%d-%d", family, encap);
318 		modload_attempted = 1;
319 		goto retry;
320 	}
321 
322 	xfrm_state_put_afinfo(afinfo);
323 	return mode;
324 }
325 
326 static void xfrm_put_mode(struct xfrm_mode *mode)
327 {
328 	module_put(mode->owner);
329 }
330 
331 static void xfrm_state_gc_destroy(struct xfrm_state *x)
332 {
333 	tasklet_hrtimer_cancel(&x->mtimer);
334 	del_timer_sync(&x->rtimer);
335 	kfree(x->aalg);
336 	kfree(x->ealg);
337 	kfree(x->calg);
338 	kfree(x->encap);
339 	kfree(x->coaddr);
340 	kfree(x->replay_esn);
341 	kfree(x->preplay_esn);
342 	if (x->inner_mode)
343 		xfrm_put_mode(x->inner_mode);
344 	if (x->inner_mode_iaf)
345 		xfrm_put_mode(x->inner_mode_iaf);
346 	if (x->outer_mode)
347 		xfrm_put_mode(x->outer_mode);
348 	if (x->type) {
349 		x->type->destructor(x);
350 		xfrm_put_type(x->type);
351 	}
352 	security_xfrm_state_free(x);
353 	kfree(x);
354 }
355 
356 static void xfrm_state_gc_task(struct work_struct *work)
357 {
358 	struct net *net = container_of(work, struct net, xfrm.state_gc_work);
359 	struct xfrm_state *x;
360 	struct hlist_node *tmp;
361 	struct hlist_head gc_list;
362 
363 	spin_lock_bh(&xfrm_state_gc_lock);
364 	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
365 	spin_unlock_bh(&xfrm_state_gc_lock);
366 
367 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
368 		xfrm_state_gc_destroy(x);
369 }
370 
371 static inline unsigned long make_jiffies(long secs)
372 {
373 	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
374 		return MAX_SCHEDULE_TIMEOUT-1;
375 	else
376 		return secs*HZ;
377 }
378 
379 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
380 {
381 	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
382 	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
383 	unsigned long now = get_seconds();
384 	long next = LONG_MAX;
385 	int warn = 0;
386 	int err = 0;
387 
388 	spin_lock(&x->lock);
389 	if (x->km.state == XFRM_STATE_DEAD)
390 		goto out;
391 	if (x->km.state == XFRM_STATE_EXPIRED)
392 		goto expired;
393 	if (x->lft.hard_add_expires_seconds) {
394 		long tmo = x->lft.hard_add_expires_seconds +
395 			x->curlft.add_time - now;
396 		if (tmo <= 0) {
397 			if (x->xflags & XFRM_SOFT_EXPIRE) {
398 				/* enter hard expire without soft expire first?!
399 				 * setting a new date could trigger this.
400 				 * workarbound: fix x->curflt.add_time by below:
401 				 */
402 				x->curlft.add_time = now - x->saved_tmo - 1;
403 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
404 			} else
405 				goto expired;
406 		}
407 		if (tmo < next)
408 			next = tmo;
409 	}
410 	if (x->lft.hard_use_expires_seconds) {
411 		long tmo = x->lft.hard_use_expires_seconds +
412 			(x->curlft.use_time ? : now) - now;
413 		if (tmo <= 0)
414 			goto expired;
415 		if (tmo < next)
416 			next = tmo;
417 	}
418 	if (x->km.dying)
419 		goto resched;
420 	if (x->lft.soft_add_expires_seconds) {
421 		long tmo = x->lft.soft_add_expires_seconds +
422 			x->curlft.add_time - now;
423 		if (tmo <= 0) {
424 			warn = 1;
425 			x->xflags &= ~XFRM_SOFT_EXPIRE;
426 		} else if (tmo < next) {
427 			next = tmo;
428 			x->xflags |= XFRM_SOFT_EXPIRE;
429 			x->saved_tmo = tmo;
430 		}
431 	}
432 	if (x->lft.soft_use_expires_seconds) {
433 		long tmo = x->lft.soft_use_expires_seconds +
434 			(x->curlft.use_time ? : now) - now;
435 		if (tmo <= 0)
436 			warn = 1;
437 		else if (tmo < next)
438 			next = tmo;
439 	}
440 
441 	x->km.dying = warn;
442 	if (warn)
443 		km_state_expired(x, 0, 0);
444 resched:
445 	if (next != LONG_MAX) {
446 		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
447 	}
448 
449 	goto out;
450 
451 expired:
452 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
453 		x->km.state = XFRM_STATE_EXPIRED;
454 
455 	err = __xfrm_state_delete(x);
456 	if (!err)
457 		km_state_expired(x, 1, 0);
458 
459 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
460 
461 out:
462 	spin_unlock(&x->lock);
463 	return HRTIMER_NORESTART;
464 }
465 
466 static void xfrm_replay_timer_handler(unsigned long data);
467 
468 struct xfrm_state *xfrm_state_alloc(struct net *net)
469 {
470 	struct xfrm_state *x;
471 
472 	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
473 
474 	if (x) {
475 		write_pnet(&x->xs_net, net);
476 		atomic_set(&x->refcnt, 1);
477 		atomic_set(&x->tunnel_users, 0);
478 		INIT_LIST_HEAD(&x->km.all);
479 		INIT_HLIST_NODE(&x->bydst);
480 		INIT_HLIST_NODE(&x->bysrc);
481 		INIT_HLIST_NODE(&x->byspi);
482 		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
483 					CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
484 		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
485 				(unsigned long)x);
486 		x->curlft.add_time = get_seconds();
487 		x->lft.soft_byte_limit = XFRM_INF;
488 		x->lft.soft_packet_limit = XFRM_INF;
489 		x->lft.hard_byte_limit = XFRM_INF;
490 		x->lft.hard_packet_limit = XFRM_INF;
491 		x->replay_maxage = 0;
492 		x->replay_maxdiff = 0;
493 		x->inner_mode = NULL;
494 		x->inner_mode_iaf = NULL;
495 		spin_lock_init(&x->lock);
496 	}
497 	return x;
498 }
499 EXPORT_SYMBOL(xfrm_state_alloc);
500 
501 void __xfrm_state_destroy(struct xfrm_state *x)
502 {
503 	struct net *net = xs_net(x);
504 
505 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
506 
507 	spin_lock_bh(&xfrm_state_gc_lock);
508 	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
509 	spin_unlock_bh(&xfrm_state_gc_lock);
510 	schedule_work(&net->xfrm.state_gc_work);
511 }
512 EXPORT_SYMBOL(__xfrm_state_destroy);
513 
514 int __xfrm_state_delete(struct xfrm_state *x)
515 {
516 	struct net *net = xs_net(x);
517 	int err = -ESRCH;
518 
519 	if (x->km.state != XFRM_STATE_DEAD) {
520 		x->km.state = XFRM_STATE_DEAD;
521 		spin_lock(&net->xfrm.xfrm_state_lock);
522 		list_del(&x->km.all);
523 		hlist_del(&x->bydst);
524 		hlist_del(&x->bysrc);
525 		if (x->id.spi)
526 			hlist_del(&x->byspi);
527 		net->xfrm.state_num--;
528 		spin_unlock(&net->xfrm.xfrm_state_lock);
529 
530 		/* All xfrm_state objects are created by xfrm_state_alloc.
531 		 * The xfrm_state_alloc call gives a reference, and that
532 		 * is what we are dropping here.
533 		 */
534 		xfrm_state_put(x);
535 		err = 0;
536 	}
537 
538 	return err;
539 }
540 EXPORT_SYMBOL(__xfrm_state_delete);
541 
542 int xfrm_state_delete(struct xfrm_state *x)
543 {
544 	int err;
545 
546 	spin_lock_bh(&x->lock);
547 	err = __xfrm_state_delete(x);
548 	spin_unlock_bh(&x->lock);
549 
550 	return err;
551 }
552 EXPORT_SYMBOL(xfrm_state_delete);
553 
554 #ifdef CONFIG_SECURITY_NETWORK_XFRM
555 static inline int
556 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
557 {
558 	int i, err = 0;
559 
560 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
561 		struct xfrm_state *x;
562 
563 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
564 			if (xfrm_id_proto_match(x->id.proto, proto) &&
565 			   (err = security_xfrm_state_delete(x)) != 0) {
566 				xfrm_audit_state_delete(x, 0, task_valid);
567 				return err;
568 			}
569 		}
570 	}
571 
572 	return err;
573 }
574 #else
575 static inline int
576 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
577 {
578 	return 0;
579 }
580 #endif
581 
582 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
583 {
584 	int i, err = 0, cnt = 0;
585 
586 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
587 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
588 	if (err)
589 		goto out;
590 
591 	err = -ESRCH;
592 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
593 		struct xfrm_state *x;
594 restart:
595 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
596 			if (!xfrm_state_kern(x) &&
597 			    xfrm_id_proto_match(x->id.proto, proto)) {
598 				xfrm_state_hold(x);
599 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
600 
601 				err = xfrm_state_delete(x);
602 				xfrm_audit_state_delete(x, err ? 0 : 1,
603 							task_valid);
604 				xfrm_state_put(x);
605 				if (!err)
606 					cnt++;
607 
608 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
609 				goto restart;
610 			}
611 		}
612 	}
613 	if (cnt)
614 		err = 0;
615 
616 out:
617 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
618 	return err;
619 }
620 EXPORT_SYMBOL(xfrm_state_flush);
621 
622 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
623 {
624 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
625 	si->sadcnt = net->xfrm.state_num;
626 	si->sadhcnt = net->xfrm.state_hmask;
627 	si->sadhmcnt = xfrm_state_hashmax;
628 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
629 }
630 EXPORT_SYMBOL(xfrm_sad_getinfo);
631 
632 static int
633 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
634 		    const struct xfrm_tmpl *tmpl,
635 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
636 		    unsigned short family)
637 {
638 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
639 	if (!afinfo)
640 		return -1;
641 	afinfo->init_tempsel(&x->sel, fl);
642 
643 	if (family != tmpl->encap_family) {
644 		xfrm_state_put_afinfo(afinfo);
645 		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
646 		if (!afinfo)
647 			return -1;
648 	}
649 	afinfo->init_temprop(x, tmpl, daddr, saddr);
650 	xfrm_state_put_afinfo(afinfo);
651 	return 0;
652 }
653 
654 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
655 					      const xfrm_address_t *daddr,
656 					      __be32 spi, u8 proto,
657 					      unsigned short family)
658 {
659 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
660 	struct xfrm_state *x;
661 
662 	hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
663 		if (x->props.family != family ||
664 		    x->id.spi       != spi ||
665 		    x->id.proto     != proto ||
666 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
667 			continue;
668 
669 		if ((mark & x->mark.m) != x->mark.v)
670 			continue;
671 		xfrm_state_hold(x);
672 		return x;
673 	}
674 
675 	return NULL;
676 }
677 
678 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
679 						     const xfrm_address_t *daddr,
680 						     const xfrm_address_t *saddr,
681 						     u8 proto, unsigned short family)
682 {
683 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
684 	struct xfrm_state *x;
685 
686 	hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
687 		if (x->props.family != family ||
688 		    x->id.proto     != proto ||
689 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
690 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
691 			continue;
692 
693 		if ((mark & x->mark.m) != x->mark.v)
694 			continue;
695 		xfrm_state_hold(x);
696 		return x;
697 	}
698 
699 	return NULL;
700 }
701 
702 static inline struct xfrm_state *
703 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
704 {
705 	struct net *net = xs_net(x);
706 	u32 mark = x->mark.v & x->mark.m;
707 
708 	if (use_spi)
709 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
710 					   x->id.spi, x->id.proto, family);
711 	else
712 		return __xfrm_state_lookup_byaddr(net, mark,
713 						  &x->id.daddr,
714 						  &x->props.saddr,
715 						  x->id.proto, family);
716 }
717 
718 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
719 {
720 	if (have_hash_collision &&
721 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
722 	    net->xfrm.state_num > net->xfrm.state_hmask)
723 		schedule_work(&net->xfrm.state_hash_work);
724 }
725 
726 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
727 			       const struct flowi *fl, unsigned short family,
728 			       struct xfrm_state **best, int *acq_in_progress,
729 			       int *error)
730 {
731 	/* Resolution logic:
732 	 * 1. There is a valid state with matching selector. Done.
733 	 * 2. Valid state with inappropriate selector. Skip.
734 	 *
735 	 * Entering area of "sysdeps".
736 	 *
737 	 * 3. If state is not valid, selector is temporary, it selects
738 	 *    only session which triggered previous resolution. Key
739 	 *    manager will do something to install a state with proper
740 	 *    selector.
741 	 */
742 	if (x->km.state == XFRM_STATE_VALID) {
743 		if ((x->sel.family &&
744 		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
745 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
746 			return;
747 
748 		if (!*best ||
749 		    (*best)->km.dying > x->km.dying ||
750 		    ((*best)->km.dying == x->km.dying &&
751 		     (*best)->curlft.add_time < x->curlft.add_time))
752 			*best = x;
753 	} else if (x->km.state == XFRM_STATE_ACQ) {
754 		*acq_in_progress = 1;
755 	} else if (x->km.state == XFRM_STATE_ERROR ||
756 		   x->km.state == XFRM_STATE_EXPIRED) {
757 		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
758 		    security_xfrm_state_pol_flow_match(x, pol, fl))
759 			*error = -ESRCH;
760 	}
761 }
762 
763 struct xfrm_state *
764 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
765 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
766 		struct xfrm_policy *pol, int *err,
767 		unsigned short family)
768 {
769 	static xfrm_address_t saddr_wildcard = { };
770 	struct net *net = xp_net(pol);
771 	unsigned int h, h_wildcard;
772 	struct xfrm_state *x, *x0, *to_put;
773 	int acquire_in_progress = 0;
774 	int error = 0;
775 	struct xfrm_state *best = NULL;
776 	u32 mark = pol->mark.v & pol->mark.m;
777 	unsigned short encap_family = tmpl->encap_family;
778 	struct km_event c;
779 
780 	to_put = NULL;
781 
782 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
783 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
784 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
785 		if (x->props.family == encap_family &&
786 		    x->props.reqid == tmpl->reqid &&
787 		    (mark & x->mark.m) == x->mark.v &&
788 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
789 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
790 		    tmpl->mode == x->props.mode &&
791 		    tmpl->id.proto == x->id.proto &&
792 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
793 			xfrm_state_look_at(pol, x, fl, encap_family,
794 					   &best, &acquire_in_progress, &error);
795 	}
796 	if (best || acquire_in_progress)
797 		goto found;
798 
799 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
800 	hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
801 		if (x->props.family == encap_family &&
802 		    x->props.reqid == tmpl->reqid &&
803 		    (mark & x->mark.m) == x->mark.v &&
804 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
805 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
806 		    tmpl->mode == x->props.mode &&
807 		    tmpl->id.proto == x->id.proto &&
808 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
809 			xfrm_state_look_at(pol, x, fl, encap_family,
810 					   &best, &acquire_in_progress, &error);
811 	}
812 
813 found:
814 	x = best;
815 	if (!x && !error && !acquire_in_progress) {
816 		if (tmpl->id.spi &&
817 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
818 					      tmpl->id.proto, encap_family)) != NULL) {
819 			to_put = x0;
820 			error = -EEXIST;
821 			goto out;
822 		}
823 
824 		c.net = net;
825 		/* If the KMs have no listeners (yet...), avoid allocating an SA
826 		 * for each and every packet - garbage collection might not
827 		 * handle the flood.
828 		 */
829 		if (!km_is_alive(&c)) {
830 			error = -ESRCH;
831 			goto out;
832 		}
833 
834 		x = xfrm_state_alloc(net);
835 		if (x == NULL) {
836 			error = -ENOMEM;
837 			goto out;
838 		}
839 		/* Initialize temporary state matching only
840 		 * to current session. */
841 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
842 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
843 
844 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
845 		if (error) {
846 			x->km.state = XFRM_STATE_DEAD;
847 			to_put = x;
848 			x = NULL;
849 			goto out;
850 		}
851 
852 		if (km_query(x, tmpl, pol) == 0) {
853 			x->km.state = XFRM_STATE_ACQ;
854 			list_add(&x->km.all, &net->xfrm.state_all);
855 			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
856 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
857 			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
858 			if (x->id.spi) {
859 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
860 				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
861 			}
862 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
863 			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
864 			net->xfrm.state_num++;
865 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
866 		} else {
867 			x->km.state = XFRM_STATE_DEAD;
868 			to_put = x;
869 			x = NULL;
870 			error = -ESRCH;
871 		}
872 	}
873 out:
874 	if (x)
875 		xfrm_state_hold(x);
876 	else
877 		*err = acquire_in_progress ? -EAGAIN : error;
878 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
879 	if (to_put)
880 		xfrm_state_put(to_put);
881 	return x;
882 }
883 
884 struct xfrm_state *
885 xfrm_stateonly_find(struct net *net, u32 mark,
886 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
887 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
888 {
889 	unsigned int h;
890 	struct xfrm_state *rx = NULL, *x = NULL;
891 
892 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
893 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
894 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
895 		if (x->props.family == family &&
896 		    x->props.reqid == reqid &&
897 		    (mark & x->mark.m) == x->mark.v &&
898 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
899 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
900 		    mode == x->props.mode &&
901 		    proto == x->id.proto &&
902 		    x->km.state == XFRM_STATE_VALID) {
903 			rx = x;
904 			break;
905 		}
906 	}
907 
908 	if (rx)
909 		xfrm_state_hold(rx);
910 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
911 
912 
913 	return rx;
914 }
915 EXPORT_SYMBOL(xfrm_stateonly_find);
916 
917 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
918 					      unsigned short family)
919 {
920 	struct xfrm_state *x;
921 	struct xfrm_state_walk *w;
922 
923 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
924 	list_for_each_entry(w, &net->xfrm.state_all, all) {
925 		x = container_of(w, struct xfrm_state, km);
926 		if (x->props.family != family ||
927 			x->id.spi != spi)
928 			continue;
929 
930 		xfrm_state_hold(x);
931 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
932 		return x;
933 	}
934 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
935 	return NULL;
936 }
937 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
938 
939 static void __xfrm_state_insert(struct xfrm_state *x)
940 {
941 	struct net *net = xs_net(x);
942 	unsigned int h;
943 
944 	list_add(&x->km.all, &net->xfrm.state_all);
945 
946 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
947 			  x->props.reqid, x->props.family);
948 	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
949 
950 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
951 	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
952 
953 	if (x->id.spi) {
954 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
955 				  x->props.family);
956 
957 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
958 	}
959 
960 	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
961 	if (x->replay_maxage)
962 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
963 
964 	net->xfrm.state_num++;
965 
966 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
967 }
968 
969 /* net->xfrm.xfrm_state_lock is held */
970 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
971 {
972 	struct net *net = xs_net(xnew);
973 	unsigned short family = xnew->props.family;
974 	u32 reqid = xnew->props.reqid;
975 	struct xfrm_state *x;
976 	unsigned int h;
977 	u32 mark = xnew->mark.v & xnew->mark.m;
978 
979 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
980 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
981 		if (x->props.family	== family &&
982 		    x->props.reqid	== reqid &&
983 		    (mark & x->mark.m) == x->mark.v &&
984 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
985 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
986 			x->genid++;
987 	}
988 }
989 
990 void xfrm_state_insert(struct xfrm_state *x)
991 {
992 	struct net *net = xs_net(x);
993 
994 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
995 	__xfrm_state_bump_genids(x);
996 	__xfrm_state_insert(x);
997 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
998 }
999 EXPORT_SYMBOL(xfrm_state_insert);
1000 
1001 /* net->xfrm.xfrm_state_lock is held */
1002 static struct xfrm_state *__find_acq_core(struct net *net,
1003 					  const struct xfrm_mark *m,
1004 					  unsigned short family, u8 mode,
1005 					  u32 reqid, u8 proto,
1006 					  const xfrm_address_t *daddr,
1007 					  const xfrm_address_t *saddr,
1008 					  int create)
1009 {
1010 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1011 	struct xfrm_state *x;
1012 	u32 mark = m->v & m->m;
1013 
1014 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1015 		if (x->props.reqid  != reqid ||
1016 		    x->props.mode   != mode ||
1017 		    x->props.family != family ||
1018 		    x->km.state     != XFRM_STATE_ACQ ||
1019 		    x->id.spi       != 0 ||
1020 		    x->id.proto	    != proto ||
1021 		    (mark & x->mark.m) != x->mark.v ||
1022 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1023 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1024 			continue;
1025 
1026 		xfrm_state_hold(x);
1027 		return x;
1028 	}
1029 
1030 	if (!create)
1031 		return NULL;
1032 
1033 	x = xfrm_state_alloc(net);
1034 	if (likely(x)) {
1035 		switch (family) {
1036 		case AF_INET:
1037 			x->sel.daddr.a4 = daddr->a4;
1038 			x->sel.saddr.a4 = saddr->a4;
1039 			x->sel.prefixlen_d = 32;
1040 			x->sel.prefixlen_s = 32;
1041 			x->props.saddr.a4 = saddr->a4;
1042 			x->id.daddr.a4 = daddr->a4;
1043 			break;
1044 
1045 		case AF_INET6:
1046 			x->sel.daddr.in6 = daddr->in6;
1047 			x->sel.saddr.in6 = saddr->in6;
1048 			x->sel.prefixlen_d = 128;
1049 			x->sel.prefixlen_s = 128;
1050 			x->props.saddr.in6 = saddr->in6;
1051 			x->id.daddr.in6 = daddr->in6;
1052 			break;
1053 		}
1054 
1055 		x->km.state = XFRM_STATE_ACQ;
1056 		x->id.proto = proto;
1057 		x->props.family = family;
1058 		x->props.mode = mode;
1059 		x->props.reqid = reqid;
1060 		x->mark.v = m->v;
1061 		x->mark.m = m->m;
1062 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1063 		xfrm_state_hold(x);
1064 		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1065 		list_add(&x->km.all, &net->xfrm.state_all);
1066 		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1067 		h = xfrm_src_hash(net, daddr, saddr, family);
1068 		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1069 
1070 		net->xfrm.state_num++;
1071 
1072 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1073 	}
1074 
1075 	return x;
1076 }
1077 
1078 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1079 
1080 int xfrm_state_add(struct xfrm_state *x)
1081 {
1082 	struct net *net = xs_net(x);
1083 	struct xfrm_state *x1, *to_put;
1084 	int family;
1085 	int err;
1086 	u32 mark = x->mark.v & x->mark.m;
1087 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1088 
1089 	family = x->props.family;
1090 
1091 	to_put = NULL;
1092 
1093 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1094 
1095 	x1 = __xfrm_state_locate(x, use_spi, family);
1096 	if (x1) {
1097 		to_put = x1;
1098 		x1 = NULL;
1099 		err = -EEXIST;
1100 		goto out;
1101 	}
1102 
1103 	if (use_spi && x->km.seq) {
1104 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1105 		if (x1 && ((x1->id.proto != x->id.proto) ||
1106 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1107 			to_put = x1;
1108 			x1 = NULL;
1109 		}
1110 	}
1111 
1112 	if (use_spi && !x1)
1113 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1114 				     x->props.reqid, x->id.proto,
1115 				     &x->id.daddr, &x->props.saddr, 0);
1116 
1117 	__xfrm_state_bump_genids(x);
1118 	__xfrm_state_insert(x);
1119 	err = 0;
1120 
1121 out:
1122 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1123 
1124 	if (x1) {
1125 		xfrm_state_delete(x1);
1126 		xfrm_state_put(x1);
1127 	}
1128 
1129 	if (to_put)
1130 		xfrm_state_put(to_put);
1131 
1132 	return err;
1133 }
1134 EXPORT_SYMBOL(xfrm_state_add);
1135 
1136 #ifdef CONFIG_XFRM_MIGRATE
1137 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1138 {
1139 	struct net *net = xs_net(orig);
1140 	struct xfrm_state *x = xfrm_state_alloc(net);
1141 	if (!x)
1142 		goto out;
1143 
1144 	memcpy(&x->id, &orig->id, sizeof(x->id));
1145 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1146 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1147 	x->props.mode = orig->props.mode;
1148 	x->props.replay_window = orig->props.replay_window;
1149 	x->props.reqid = orig->props.reqid;
1150 	x->props.family = orig->props.family;
1151 	x->props.saddr = orig->props.saddr;
1152 
1153 	if (orig->aalg) {
1154 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1155 		if (!x->aalg)
1156 			goto error;
1157 	}
1158 	x->props.aalgo = orig->props.aalgo;
1159 
1160 	if (orig->aead) {
1161 		x->aead = xfrm_algo_aead_clone(orig->aead);
1162 		if (!x->aead)
1163 			goto error;
1164 	}
1165 	if (orig->ealg) {
1166 		x->ealg = xfrm_algo_clone(orig->ealg);
1167 		if (!x->ealg)
1168 			goto error;
1169 	}
1170 	x->props.ealgo = orig->props.ealgo;
1171 
1172 	if (orig->calg) {
1173 		x->calg = xfrm_algo_clone(orig->calg);
1174 		if (!x->calg)
1175 			goto error;
1176 	}
1177 	x->props.calgo = orig->props.calgo;
1178 
1179 	if (orig->encap) {
1180 		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1181 		if (!x->encap)
1182 			goto error;
1183 	}
1184 
1185 	if (orig->coaddr) {
1186 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1187 				    GFP_KERNEL);
1188 		if (!x->coaddr)
1189 			goto error;
1190 	}
1191 
1192 	if (orig->replay_esn) {
1193 		if (xfrm_replay_clone(x, orig))
1194 			goto error;
1195 	}
1196 
1197 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1198 
1199 	if (xfrm_init_state(x) < 0)
1200 		goto error;
1201 
1202 	x->props.flags = orig->props.flags;
1203 	x->props.extra_flags = orig->props.extra_flags;
1204 
1205 	x->tfcpad = orig->tfcpad;
1206 	x->replay_maxdiff = orig->replay_maxdiff;
1207 	x->replay_maxage = orig->replay_maxage;
1208 	x->curlft.add_time = orig->curlft.add_time;
1209 	x->km.state = orig->km.state;
1210 	x->km.seq = orig->km.seq;
1211 
1212 	return x;
1213 
1214  error:
1215 	xfrm_state_put(x);
1216 out:
1217 	return NULL;
1218 }
1219 
1220 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1221 {
1222 	unsigned int h;
1223 	struct xfrm_state *x = NULL;
1224 
1225 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1226 
1227 	if (m->reqid) {
1228 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1229 				  m->reqid, m->old_family);
1230 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1231 			if (x->props.mode != m->mode ||
1232 			    x->id.proto != m->proto)
1233 				continue;
1234 			if (m->reqid && x->props.reqid != m->reqid)
1235 				continue;
1236 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1237 					     m->old_family) ||
1238 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1239 					     m->old_family))
1240 				continue;
1241 			xfrm_state_hold(x);
1242 			break;
1243 		}
1244 	} else {
1245 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1246 				  m->old_family);
1247 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1248 			if (x->props.mode != m->mode ||
1249 			    x->id.proto != m->proto)
1250 				continue;
1251 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1252 					     m->old_family) ||
1253 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1254 					     m->old_family))
1255 				continue;
1256 			xfrm_state_hold(x);
1257 			break;
1258 		}
1259 	}
1260 
1261 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1262 
1263 	return x;
1264 }
1265 EXPORT_SYMBOL(xfrm_migrate_state_find);
1266 
1267 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1268 				      struct xfrm_migrate *m)
1269 {
1270 	struct xfrm_state *xc;
1271 
1272 	xc = xfrm_state_clone(x);
1273 	if (!xc)
1274 		return NULL;
1275 
1276 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1277 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1278 
1279 	/* add state */
1280 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1281 		/* a care is needed when the destination address of the
1282 		   state is to be updated as it is a part of triplet */
1283 		xfrm_state_insert(xc);
1284 	} else {
1285 		if (xfrm_state_add(xc) < 0)
1286 			goto error;
1287 	}
1288 
1289 	return xc;
1290 error:
1291 	xfrm_state_put(xc);
1292 	return NULL;
1293 }
1294 EXPORT_SYMBOL(xfrm_state_migrate);
1295 #endif
1296 
1297 int xfrm_state_update(struct xfrm_state *x)
1298 {
1299 	struct xfrm_state *x1, *to_put;
1300 	int err;
1301 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1302 	struct net *net = xs_net(x);
1303 
1304 	to_put = NULL;
1305 
1306 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1307 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1308 
1309 	err = -ESRCH;
1310 	if (!x1)
1311 		goto out;
1312 
1313 	if (xfrm_state_kern(x1)) {
1314 		to_put = x1;
1315 		err = -EEXIST;
1316 		goto out;
1317 	}
1318 
1319 	if (x1->km.state == XFRM_STATE_ACQ) {
1320 		__xfrm_state_insert(x);
1321 		x = NULL;
1322 	}
1323 	err = 0;
1324 
1325 out:
1326 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1327 
1328 	if (to_put)
1329 		xfrm_state_put(to_put);
1330 
1331 	if (err)
1332 		return err;
1333 
1334 	if (!x) {
1335 		xfrm_state_delete(x1);
1336 		xfrm_state_put(x1);
1337 		return 0;
1338 	}
1339 
1340 	err = -EINVAL;
1341 	spin_lock_bh(&x1->lock);
1342 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1343 		if (x->encap && x1->encap)
1344 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1345 		if (x->coaddr && x1->coaddr) {
1346 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1347 		}
1348 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1349 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1350 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1351 		x1->km.dying = 0;
1352 
1353 		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1354 		if (x1->curlft.use_time)
1355 			xfrm_state_check_expire(x1);
1356 
1357 		err = 0;
1358 		x->km.state = XFRM_STATE_DEAD;
1359 		__xfrm_state_put(x);
1360 	}
1361 	spin_unlock_bh(&x1->lock);
1362 
1363 	xfrm_state_put(x1);
1364 
1365 	return err;
1366 }
1367 EXPORT_SYMBOL(xfrm_state_update);
1368 
1369 int xfrm_state_check_expire(struct xfrm_state *x)
1370 {
1371 	if (!x->curlft.use_time)
1372 		x->curlft.use_time = get_seconds();
1373 
1374 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1375 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1376 		x->km.state = XFRM_STATE_EXPIRED;
1377 		tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1378 		return -EINVAL;
1379 	}
1380 
1381 	if (!x->km.dying &&
1382 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1383 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1384 		x->km.dying = 1;
1385 		km_state_expired(x, 0, 0);
1386 	}
1387 	return 0;
1388 }
1389 EXPORT_SYMBOL(xfrm_state_check_expire);
1390 
1391 struct xfrm_state *
1392 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1393 		  u8 proto, unsigned short family)
1394 {
1395 	struct xfrm_state *x;
1396 
1397 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1398 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1399 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1400 	return x;
1401 }
1402 EXPORT_SYMBOL(xfrm_state_lookup);
1403 
1404 struct xfrm_state *
1405 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1406 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1407 			 u8 proto, unsigned short family)
1408 {
1409 	struct xfrm_state *x;
1410 
1411 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1412 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1413 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1414 	return x;
1415 }
1416 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1417 
1418 struct xfrm_state *
1419 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1420 	      u8 proto, const xfrm_address_t *daddr,
1421 	      const xfrm_address_t *saddr, int create, unsigned short family)
1422 {
1423 	struct xfrm_state *x;
1424 
1425 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1426 	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1427 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1428 
1429 	return x;
1430 }
1431 EXPORT_SYMBOL(xfrm_find_acq);
1432 
1433 #ifdef CONFIG_XFRM_SUB_POLICY
1434 int
1435 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1436 	       unsigned short family, struct net *net)
1437 {
1438 	int err = 0;
1439 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1440 	if (!afinfo)
1441 		return -EAFNOSUPPORT;
1442 
1443 	spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1444 	if (afinfo->tmpl_sort)
1445 		err = afinfo->tmpl_sort(dst, src, n);
1446 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1447 	xfrm_state_put_afinfo(afinfo);
1448 	return err;
1449 }
1450 EXPORT_SYMBOL(xfrm_tmpl_sort);
1451 
1452 int
1453 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1454 		unsigned short family)
1455 {
1456 	int err = 0;
1457 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1458 	struct net *net = xs_net(*src);
1459 
1460 	if (!afinfo)
1461 		return -EAFNOSUPPORT;
1462 
1463 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1464 	if (afinfo->state_sort)
1465 		err = afinfo->state_sort(dst, src, n);
1466 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1467 	xfrm_state_put_afinfo(afinfo);
1468 	return err;
1469 }
1470 EXPORT_SYMBOL(xfrm_state_sort);
1471 #endif
1472 
1473 /* Silly enough, but I'm lazy to build resolution list */
1474 
1475 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1476 {
1477 	int i;
1478 
1479 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1480 		struct xfrm_state *x;
1481 
1482 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1483 			if (x->km.seq == seq &&
1484 			    (mark & x->mark.m) == x->mark.v &&
1485 			    x->km.state == XFRM_STATE_ACQ) {
1486 				xfrm_state_hold(x);
1487 				return x;
1488 			}
1489 		}
1490 	}
1491 	return NULL;
1492 }
1493 
1494 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1495 {
1496 	struct xfrm_state *x;
1497 
1498 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1499 	x = __xfrm_find_acq_byseq(net, mark, seq);
1500 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1501 	return x;
1502 }
1503 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1504 
1505 u32 xfrm_get_acqseq(void)
1506 {
1507 	u32 res;
1508 	static atomic_t acqseq;
1509 
1510 	do {
1511 		res = atomic_inc_return(&acqseq);
1512 	} while (!res);
1513 
1514 	return res;
1515 }
1516 EXPORT_SYMBOL(xfrm_get_acqseq);
1517 
1518 int verify_spi_info(u8 proto, u32 min, u32 max)
1519 {
1520 	switch (proto) {
1521 	case IPPROTO_AH:
1522 	case IPPROTO_ESP:
1523 		break;
1524 
1525 	case IPPROTO_COMP:
1526 		/* IPCOMP spi is 16-bits. */
1527 		if (max >= 0x10000)
1528 			return -EINVAL;
1529 		break;
1530 
1531 	default:
1532 		return -EINVAL;
1533 	}
1534 
1535 	if (min > max)
1536 		return -EINVAL;
1537 
1538 	return 0;
1539 }
1540 EXPORT_SYMBOL(verify_spi_info);
1541 
1542 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1543 {
1544 	struct net *net = xs_net(x);
1545 	unsigned int h;
1546 	struct xfrm_state *x0;
1547 	int err = -ENOENT;
1548 	__be32 minspi = htonl(low);
1549 	__be32 maxspi = htonl(high);
1550 	u32 mark = x->mark.v & x->mark.m;
1551 
1552 	spin_lock_bh(&x->lock);
1553 	if (x->km.state == XFRM_STATE_DEAD)
1554 		goto unlock;
1555 
1556 	err = 0;
1557 	if (x->id.spi)
1558 		goto unlock;
1559 
1560 	err = -ENOENT;
1561 
1562 	if (minspi == maxspi) {
1563 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1564 		if (x0) {
1565 			xfrm_state_put(x0);
1566 			goto unlock;
1567 		}
1568 		x->id.spi = minspi;
1569 	} else {
1570 		u32 spi = 0;
1571 		for (h = 0; h < high-low+1; h++) {
1572 			spi = low + prandom_u32()%(high-low+1);
1573 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1574 			if (x0 == NULL) {
1575 				x->id.spi = htonl(spi);
1576 				break;
1577 			}
1578 			xfrm_state_put(x0);
1579 		}
1580 	}
1581 	if (x->id.spi) {
1582 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
1583 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1584 		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1585 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1586 
1587 		err = 0;
1588 	}
1589 
1590 unlock:
1591 	spin_unlock_bh(&x->lock);
1592 
1593 	return err;
1594 }
1595 EXPORT_SYMBOL(xfrm_alloc_spi);
1596 
1597 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1598 				      struct xfrm_address_filter *filter)
1599 {
1600 	if (filter) {
1601 		if ((filter->family == AF_INET ||
1602 		     filter->family == AF_INET6) &&
1603 		    x->props.family != filter->family)
1604 			return false;
1605 
1606 		return addr_match(&x->props.saddr, &filter->saddr,
1607 				  filter->splen) &&
1608 		       addr_match(&x->id.daddr, &filter->daddr,
1609 				  filter->dplen);
1610 	}
1611 	return true;
1612 }
1613 
1614 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1615 		    int (*func)(struct xfrm_state *, int, void*),
1616 		    void *data)
1617 {
1618 	struct xfrm_state *state;
1619 	struct xfrm_state_walk *x;
1620 	int err = 0;
1621 
1622 	if (walk->seq != 0 && list_empty(&walk->all))
1623 		return 0;
1624 
1625 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1626 	if (list_empty(&walk->all))
1627 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1628 	else
1629 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1630 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1631 		if (x->state == XFRM_STATE_DEAD)
1632 			continue;
1633 		state = container_of(x, struct xfrm_state, km);
1634 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1635 			continue;
1636 		if (!__xfrm_state_filter_match(state, walk->filter))
1637 			continue;
1638 		err = func(state, walk->seq, data);
1639 		if (err) {
1640 			list_move_tail(&walk->all, &x->all);
1641 			goto out;
1642 		}
1643 		walk->seq++;
1644 	}
1645 	if (walk->seq == 0) {
1646 		err = -ENOENT;
1647 		goto out;
1648 	}
1649 	list_del_init(&walk->all);
1650 out:
1651 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1652 	return err;
1653 }
1654 EXPORT_SYMBOL(xfrm_state_walk);
1655 
1656 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1657 			  struct xfrm_address_filter *filter)
1658 {
1659 	INIT_LIST_HEAD(&walk->all);
1660 	walk->proto = proto;
1661 	walk->state = XFRM_STATE_DEAD;
1662 	walk->seq = 0;
1663 	walk->filter = filter;
1664 }
1665 EXPORT_SYMBOL(xfrm_state_walk_init);
1666 
1667 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1668 {
1669 	kfree(walk->filter);
1670 
1671 	if (list_empty(&walk->all))
1672 		return;
1673 
1674 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1675 	list_del(&walk->all);
1676 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1677 }
1678 EXPORT_SYMBOL(xfrm_state_walk_done);
1679 
1680 static void xfrm_replay_timer_handler(unsigned long data)
1681 {
1682 	struct xfrm_state *x = (struct xfrm_state *)data;
1683 
1684 	spin_lock(&x->lock);
1685 
1686 	if (x->km.state == XFRM_STATE_VALID) {
1687 		if (xfrm_aevent_is_on(xs_net(x)))
1688 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1689 		else
1690 			x->xflags |= XFRM_TIME_DEFER;
1691 	}
1692 
1693 	spin_unlock(&x->lock);
1694 }
1695 
1696 static LIST_HEAD(xfrm_km_list);
1697 
1698 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1699 {
1700 	struct xfrm_mgr *km;
1701 
1702 	rcu_read_lock();
1703 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1704 		if (km->notify_policy)
1705 			km->notify_policy(xp, dir, c);
1706 	rcu_read_unlock();
1707 }
1708 
1709 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1710 {
1711 	struct xfrm_mgr *km;
1712 	rcu_read_lock();
1713 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1714 		if (km->notify)
1715 			km->notify(x, c);
1716 	rcu_read_unlock();
1717 }
1718 
1719 EXPORT_SYMBOL(km_policy_notify);
1720 EXPORT_SYMBOL(km_state_notify);
1721 
1722 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1723 {
1724 	struct km_event c;
1725 
1726 	c.data.hard = hard;
1727 	c.portid = portid;
1728 	c.event = XFRM_MSG_EXPIRE;
1729 	km_state_notify(x, &c);
1730 }
1731 
1732 EXPORT_SYMBOL(km_state_expired);
1733 /*
1734  * We send to all registered managers regardless of failure
1735  * We are happy with one success
1736 */
1737 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1738 {
1739 	int err = -EINVAL, acqret;
1740 	struct xfrm_mgr *km;
1741 
1742 	rcu_read_lock();
1743 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1744 		acqret = km->acquire(x, t, pol);
1745 		if (!acqret)
1746 			err = acqret;
1747 	}
1748 	rcu_read_unlock();
1749 	return err;
1750 }
1751 EXPORT_SYMBOL(km_query);
1752 
1753 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1754 {
1755 	int err = -EINVAL;
1756 	struct xfrm_mgr *km;
1757 
1758 	rcu_read_lock();
1759 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1760 		if (km->new_mapping)
1761 			err = km->new_mapping(x, ipaddr, sport);
1762 		if (!err)
1763 			break;
1764 	}
1765 	rcu_read_unlock();
1766 	return err;
1767 }
1768 EXPORT_SYMBOL(km_new_mapping);
1769 
1770 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1771 {
1772 	struct km_event c;
1773 
1774 	c.data.hard = hard;
1775 	c.portid = portid;
1776 	c.event = XFRM_MSG_POLEXPIRE;
1777 	km_policy_notify(pol, dir, &c);
1778 }
1779 EXPORT_SYMBOL(km_policy_expired);
1780 
1781 #ifdef CONFIG_XFRM_MIGRATE
1782 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1783 	       const struct xfrm_migrate *m, int num_migrate,
1784 	       const struct xfrm_kmaddress *k)
1785 {
1786 	int err = -EINVAL;
1787 	int ret;
1788 	struct xfrm_mgr *km;
1789 
1790 	rcu_read_lock();
1791 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1792 		if (km->migrate) {
1793 			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1794 			if (!ret)
1795 				err = ret;
1796 		}
1797 	}
1798 	rcu_read_unlock();
1799 	return err;
1800 }
1801 EXPORT_SYMBOL(km_migrate);
1802 #endif
1803 
1804 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1805 {
1806 	int err = -EINVAL;
1807 	int ret;
1808 	struct xfrm_mgr *km;
1809 
1810 	rcu_read_lock();
1811 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1812 		if (km->report) {
1813 			ret = km->report(net, proto, sel, addr);
1814 			if (!ret)
1815 				err = ret;
1816 		}
1817 	}
1818 	rcu_read_unlock();
1819 	return err;
1820 }
1821 EXPORT_SYMBOL(km_report);
1822 
1823 bool km_is_alive(const struct km_event *c)
1824 {
1825 	struct xfrm_mgr *km;
1826 	bool is_alive = false;
1827 
1828 	rcu_read_lock();
1829 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1830 		if (km->is_alive && km->is_alive(c)) {
1831 			is_alive = true;
1832 			break;
1833 		}
1834 	}
1835 	rcu_read_unlock();
1836 
1837 	return is_alive;
1838 }
1839 EXPORT_SYMBOL(km_is_alive);
1840 
1841 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1842 {
1843 	int err;
1844 	u8 *data;
1845 	struct xfrm_mgr *km;
1846 	struct xfrm_policy *pol = NULL;
1847 
1848 	if (optlen <= 0 || optlen > PAGE_SIZE)
1849 		return -EMSGSIZE;
1850 
1851 	data = kmalloc(optlen, GFP_KERNEL);
1852 	if (!data)
1853 		return -ENOMEM;
1854 
1855 	err = -EFAULT;
1856 	if (copy_from_user(data, optval, optlen))
1857 		goto out;
1858 
1859 	err = -EINVAL;
1860 	rcu_read_lock();
1861 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1862 		pol = km->compile_policy(sk, optname, data,
1863 					 optlen, &err);
1864 		if (err >= 0)
1865 			break;
1866 	}
1867 	rcu_read_unlock();
1868 
1869 	if (err >= 0) {
1870 		xfrm_sk_policy_insert(sk, err, pol);
1871 		xfrm_pol_put(pol);
1872 		err = 0;
1873 	}
1874 
1875 out:
1876 	kfree(data);
1877 	return err;
1878 }
1879 EXPORT_SYMBOL(xfrm_user_policy);
1880 
1881 static DEFINE_SPINLOCK(xfrm_km_lock);
1882 
1883 int xfrm_register_km(struct xfrm_mgr *km)
1884 {
1885 	spin_lock_bh(&xfrm_km_lock);
1886 	list_add_tail_rcu(&km->list, &xfrm_km_list);
1887 	spin_unlock_bh(&xfrm_km_lock);
1888 	return 0;
1889 }
1890 EXPORT_SYMBOL(xfrm_register_km);
1891 
1892 int xfrm_unregister_km(struct xfrm_mgr *km)
1893 {
1894 	spin_lock_bh(&xfrm_km_lock);
1895 	list_del_rcu(&km->list);
1896 	spin_unlock_bh(&xfrm_km_lock);
1897 	synchronize_rcu();
1898 	return 0;
1899 }
1900 EXPORT_SYMBOL(xfrm_unregister_km);
1901 
1902 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1903 {
1904 	int err = 0;
1905 	if (unlikely(afinfo == NULL))
1906 		return -EINVAL;
1907 	if (unlikely(afinfo->family >= NPROTO))
1908 		return -EAFNOSUPPORT;
1909 	spin_lock_bh(&xfrm_state_afinfo_lock);
1910 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1911 		err = -EEXIST;
1912 	else
1913 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1914 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1915 	return err;
1916 }
1917 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1918 
1919 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1920 {
1921 	int err = 0;
1922 	if (unlikely(afinfo == NULL))
1923 		return -EINVAL;
1924 	if (unlikely(afinfo->family >= NPROTO))
1925 		return -EAFNOSUPPORT;
1926 	spin_lock_bh(&xfrm_state_afinfo_lock);
1927 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1928 		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1929 			err = -EINVAL;
1930 		else
1931 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1932 	}
1933 	spin_unlock_bh(&xfrm_state_afinfo_lock);
1934 	synchronize_rcu();
1935 	return err;
1936 }
1937 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1938 
1939 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1940 {
1941 	struct xfrm_state_afinfo *afinfo;
1942 	if (unlikely(family >= NPROTO))
1943 		return NULL;
1944 	rcu_read_lock();
1945 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1946 	if (unlikely(!afinfo))
1947 		rcu_read_unlock();
1948 	return afinfo;
1949 }
1950 
1951 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1952 {
1953 	rcu_read_unlock();
1954 }
1955 
1956 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1957 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1958 {
1959 	if (x->tunnel) {
1960 		struct xfrm_state *t = x->tunnel;
1961 
1962 		if (atomic_read(&t->tunnel_users) == 2)
1963 			xfrm_state_delete(t);
1964 		atomic_dec(&t->tunnel_users);
1965 		xfrm_state_put(t);
1966 		x->tunnel = NULL;
1967 	}
1968 }
1969 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1970 
1971 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1972 {
1973 	int res;
1974 
1975 	spin_lock_bh(&x->lock);
1976 	if (x->km.state == XFRM_STATE_VALID &&
1977 	    x->type && x->type->get_mtu)
1978 		res = x->type->get_mtu(x, mtu);
1979 	else
1980 		res = mtu - x->props.header_len;
1981 	spin_unlock_bh(&x->lock);
1982 	return res;
1983 }
1984 
1985 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1986 {
1987 	struct xfrm_state_afinfo *afinfo;
1988 	struct xfrm_mode *inner_mode;
1989 	int family = x->props.family;
1990 	int err;
1991 
1992 	err = -EAFNOSUPPORT;
1993 	afinfo = xfrm_state_get_afinfo(family);
1994 	if (!afinfo)
1995 		goto error;
1996 
1997 	err = 0;
1998 	if (afinfo->init_flags)
1999 		err = afinfo->init_flags(x);
2000 
2001 	xfrm_state_put_afinfo(afinfo);
2002 
2003 	if (err)
2004 		goto error;
2005 
2006 	err = -EPROTONOSUPPORT;
2007 
2008 	if (x->sel.family != AF_UNSPEC) {
2009 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2010 		if (inner_mode == NULL)
2011 			goto error;
2012 
2013 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2014 		    family != x->sel.family) {
2015 			xfrm_put_mode(inner_mode);
2016 			goto error;
2017 		}
2018 
2019 		x->inner_mode = inner_mode;
2020 	} else {
2021 		struct xfrm_mode *inner_mode_iaf;
2022 		int iafamily = AF_INET;
2023 
2024 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2025 		if (inner_mode == NULL)
2026 			goto error;
2027 
2028 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2029 			xfrm_put_mode(inner_mode);
2030 			goto error;
2031 		}
2032 		x->inner_mode = inner_mode;
2033 
2034 		if (x->props.family == AF_INET)
2035 			iafamily = AF_INET6;
2036 
2037 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2038 		if (inner_mode_iaf) {
2039 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2040 				x->inner_mode_iaf = inner_mode_iaf;
2041 			else
2042 				xfrm_put_mode(inner_mode_iaf);
2043 		}
2044 	}
2045 
2046 	x->type = xfrm_get_type(x->id.proto, family);
2047 	if (x->type == NULL)
2048 		goto error;
2049 
2050 	err = x->type->init_state(x);
2051 	if (err)
2052 		goto error;
2053 
2054 	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2055 	if (x->outer_mode == NULL) {
2056 		err = -EPROTONOSUPPORT;
2057 		goto error;
2058 	}
2059 
2060 	if (init_replay) {
2061 		err = xfrm_init_replay(x);
2062 		if (err)
2063 			goto error;
2064 	}
2065 
2066 	x->km.state = XFRM_STATE_VALID;
2067 
2068 error:
2069 	return err;
2070 }
2071 
2072 EXPORT_SYMBOL(__xfrm_init_state);
2073 
2074 int xfrm_init_state(struct xfrm_state *x)
2075 {
2076 	return __xfrm_init_state(x, true);
2077 }
2078 
2079 EXPORT_SYMBOL(xfrm_init_state);
2080 
2081 int __net_init xfrm_state_init(struct net *net)
2082 {
2083 	unsigned int sz;
2084 
2085 	INIT_LIST_HEAD(&net->xfrm.state_all);
2086 
2087 	sz = sizeof(struct hlist_head) * 8;
2088 
2089 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2090 	if (!net->xfrm.state_bydst)
2091 		goto out_bydst;
2092 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2093 	if (!net->xfrm.state_bysrc)
2094 		goto out_bysrc;
2095 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2096 	if (!net->xfrm.state_byspi)
2097 		goto out_byspi;
2098 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2099 
2100 	net->xfrm.state_num = 0;
2101 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2102 	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2103 	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2104 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2105 	return 0;
2106 
2107 out_byspi:
2108 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2109 out_bysrc:
2110 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2111 out_bydst:
2112 	return -ENOMEM;
2113 }
2114 
2115 void xfrm_state_fini(struct net *net)
2116 {
2117 	unsigned int sz;
2118 
2119 	flush_work(&net->xfrm.state_hash_work);
2120 	xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2121 	flush_work(&net->xfrm.state_gc_work);
2122 
2123 	WARN_ON(!list_empty(&net->xfrm.state_all));
2124 
2125 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2126 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2127 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2128 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2129 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2130 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2131 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2132 }
2133 
2134 #ifdef CONFIG_AUDITSYSCALL
2135 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2136 				     struct audit_buffer *audit_buf)
2137 {
2138 	struct xfrm_sec_ctx *ctx = x->security;
2139 	u32 spi = ntohl(x->id.spi);
2140 
2141 	if (ctx)
2142 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2143 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2144 
2145 	switch (x->props.family) {
2146 	case AF_INET:
2147 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2148 				 &x->props.saddr.a4, &x->id.daddr.a4);
2149 		break;
2150 	case AF_INET6:
2151 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2152 				 x->props.saddr.a6, x->id.daddr.a6);
2153 		break;
2154 	}
2155 
2156 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2157 }
2158 
2159 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2160 				      struct audit_buffer *audit_buf)
2161 {
2162 	const struct iphdr *iph4;
2163 	const struct ipv6hdr *iph6;
2164 
2165 	switch (family) {
2166 	case AF_INET:
2167 		iph4 = ip_hdr(skb);
2168 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2169 				 &iph4->saddr, &iph4->daddr);
2170 		break;
2171 	case AF_INET6:
2172 		iph6 = ipv6_hdr(skb);
2173 		audit_log_format(audit_buf,
2174 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2175 				 &iph6->saddr, &iph6->daddr,
2176 				 iph6->flow_lbl[0] & 0x0f,
2177 				 iph6->flow_lbl[1],
2178 				 iph6->flow_lbl[2]);
2179 		break;
2180 	}
2181 }
2182 
2183 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2184 {
2185 	struct audit_buffer *audit_buf;
2186 
2187 	audit_buf = xfrm_audit_start("SAD-add");
2188 	if (audit_buf == NULL)
2189 		return;
2190 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2191 	xfrm_audit_helper_sainfo(x, audit_buf);
2192 	audit_log_format(audit_buf, " res=%u", result);
2193 	audit_log_end(audit_buf);
2194 }
2195 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2196 
2197 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2198 {
2199 	struct audit_buffer *audit_buf;
2200 
2201 	audit_buf = xfrm_audit_start("SAD-delete");
2202 	if (audit_buf == NULL)
2203 		return;
2204 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2205 	xfrm_audit_helper_sainfo(x, audit_buf);
2206 	audit_log_format(audit_buf, " res=%u", result);
2207 	audit_log_end(audit_buf);
2208 }
2209 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2210 
2211 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2212 				      struct sk_buff *skb)
2213 {
2214 	struct audit_buffer *audit_buf;
2215 	u32 spi;
2216 
2217 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2218 	if (audit_buf == NULL)
2219 		return;
2220 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2221 	/* don't record the sequence number because it's inherent in this kind
2222 	 * of audit message */
2223 	spi = ntohl(x->id.spi);
2224 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2225 	audit_log_end(audit_buf);
2226 }
2227 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2228 
2229 void xfrm_audit_state_replay(struct xfrm_state *x,
2230 			     struct sk_buff *skb, __be32 net_seq)
2231 {
2232 	struct audit_buffer *audit_buf;
2233 	u32 spi;
2234 
2235 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2236 	if (audit_buf == NULL)
2237 		return;
2238 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2239 	spi = ntohl(x->id.spi);
2240 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2241 			 spi, spi, ntohl(net_seq));
2242 	audit_log_end(audit_buf);
2243 }
2244 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2245 
2246 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2247 {
2248 	struct audit_buffer *audit_buf;
2249 
2250 	audit_buf = xfrm_audit_start("SA-notfound");
2251 	if (audit_buf == NULL)
2252 		return;
2253 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2254 	audit_log_end(audit_buf);
2255 }
2256 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2257 
2258 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2259 			       __be32 net_spi, __be32 net_seq)
2260 {
2261 	struct audit_buffer *audit_buf;
2262 	u32 spi;
2263 
2264 	audit_buf = xfrm_audit_start("SA-notfound");
2265 	if (audit_buf == NULL)
2266 		return;
2267 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2268 	spi = ntohl(net_spi);
2269 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2270 			 spi, spi, ntohl(net_seq));
2271 	audit_log_end(audit_buf);
2272 }
2273 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2274 
2275 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2276 			      struct sk_buff *skb, u8 proto)
2277 {
2278 	struct audit_buffer *audit_buf;
2279 	__be32 net_spi;
2280 	__be32 net_seq;
2281 
2282 	audit_buf = xfrm_audit_start("SA-icv-failure");
2283 	if (audit_buf == NULL)
2284 		return;
2285 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2286 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2287 		u32 spi = ntohl(net_spi);
2288 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2289 				 spi, spi, ntohl(net_seq));
2290 	}
2291 	audit_log_end(audit_buf);
2292 }
2293 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2294 #endif /* CONFIG_AUDITSYSCALL */
2295