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
  38static DEFINE_SPINLOCK(xfrm_state_lock);
  39
  40static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  41
  42static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
  43static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
  44
  45static inline unsigned int xfrm_dst_hash(struct net *net,
  46					 const xfrm_address_t *daddr,
  47					 const xfrm_address_t *saddr,
  48					 u32 reqid,
  49					 unsigned short family)
  50{
  51	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  52}
  53
  54static inline unsigned int xfrm_src_hash(struct net *net,
  55					 const xfrm_address_t *daddr,
  56					 const xfrm_address_t *saddr,
  57					 unsigned short family)
  58{
  59	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  60}
  61
  62static inline unsigned int
  63xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  64	      __be32 spi, u8 proto, unsigned short family)
  65{
  66	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  67}
  68
  69static void xfrm_hash_transfer(struct hlist_head *list,
  70			       struct hlist_head *ndsttable,
  71			       struct hlist_head *nsrctable,
  72			       struct hlist_head *nspitable,
  73			       unsigned int nhashmask)
  74{
  75	struct hlist_node *entry, *tmp;
  76	struct xfrm_state *x;
  77
  78	hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
  79		unsigned int h;
  80
  81		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  82				    x->props.reqid, x->props.family,
  83				    nhashmask);
  84		hlist_add_head(&x->bydst, ndsttable+h);
  85
  86		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  87				    x->props.family,
  88				    nhashmask);
  89		hlist_add_head(&x->bysrc, nsrctable+h);
  90
  91		if (x->id.spi) {
  92			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  93					    x->id.proto, x->props.family,
  94					    nhashmask);
  95			hlist_add_head(&x->byspi, nspitable+h);
  96		}
  97	}
  98}
  99
 100static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 101{
 102	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 103}
 104
 105static DEFINE_MUTEX(hash_resize_mutex);
 106
 107static void xfrm_hash_resize(struct work_struct *work)
 108{
 109	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 110	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 111	unsigned long nsize, osize;
 112	unsigned int nhashmask, ohashmask;
 113	int i;
 114
 115	mutex_lock(&hash_resize_mutex);
 116
 117	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 118	ndst = xfrm_hash_alloc(nsize);
 119	if (!ndst)
 120		goto out_unlock;
 121	nsrc = xfrm_hash_alloc(nsize);
 122	if (!nsrc) {
 123		xfrm_hash_free(ndst, nsize);
 124		goto out_unlock;
 125	}
 126	nspi = xfrm_hash_alloc(nsize);
 127	if (!nspi) {
 128		xfrm_hash_free(ndst, nsize);
 129		xfrm_hash_free(nsrc, nsize);
 130		goto out_unlock;
 131	}
 132
 133	spin_lock_bh(&xfrm_state_lock);
 134
 135	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 136	for (i = net->xfrm.state_hmask; i >= 0; i--)
 137		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 138				   nhashmask);
 139
 140	odst = net->xfrm.state_bydst;
 141	osrc = net->xfrm.state_bysrc;
 142	ospi = net->xfrm.state_byspi;
 143	ohashmask = net->xfrm.state_hmask;
 144
 145	net->xfrm.state_bydst = ndst;
 146	net->xfrm.state_bysrc = nsrc;
 147	net->xfrm.state_byspi = nspi;
 148	net->xfrm.state_hmask = nhashmask;
 149
 150	spin_unlock_bh(&xfrm_state_lock);
 151
 152	osize = (ohashmask + 1) * sizeof(struct hlist_head);
 153	xfrm_hash_free(odst, osize);
 154	xfrm_hash_free(osrc, osize);
 155	xfrm_hash_free(ospi, osize);
 156
 157out_unlock:
 158	mutex_unlock(&hash_resize_mutex);
 159}
 160
 161static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
 162static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
 163
 164static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 165
 166int __xfrm_state_delete(struct xfrm_state *x);
 167
 168int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 169void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
 170
 171static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 172{
 173	struct xfrm_state_afinfo *afinfo;
 174	if (unlikely(family >= NPROTO))
 175		return NULL;
 176	write_lock_bh(&xfrm_state_afinfo_lock);
 177	afinfo = xfrm_state_afinfo[family];
 178	if (unlikely(!afinfo))
 179		write_unlock_bh(&xfrm_state_afinfo_lock);
 180	return afinfo;
 181}
 182
 183static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 184	__releases(xfrm_state_afinfo_lock)
 185{
 186	write_unlock_bh(&xfrm_state_afinfo_lock);
 187}
 188
 189int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 190{
 191	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 192	const struct xfrm_type **typemap;
 193	int err = 0;
 194
 195	if (unlikely(afinfo == NULL))
 196		return -EAFNOSUPPORT;
 197	typemap = afinfo->type_map;
 198
 199	if (likely(typemap[type->proto] == NULL))
 200		typemap[type->proto] = type;
 201	else
 202		err = -EEXIST;
 203	xfrm_state_unlock_afinfo(afinfo);
 204	return err;
 205}
 206EXPORT_SYMBOL(xfrm_register_type);
 207
 208int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 209{
 210	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 211	const struct xfrm_type **typemap;
 212	int err = 0;
 213
 214	if (unlikely(afinfo == NULL))
 215		return -EAFNOSUPPORT;
 216	typemap = afinfo->type_map;
 217
 218	if (unlikely(typemap[type->proto] != type))
 219		err = -ENOENT;
 220	else
 221		typemap[type->proto] = NULL;
 222	xfrm_state_unlock_afinfo(afinfo);
 223	return err;
 224}
 225EXPORT_SYMBOL(xfrm_unregister_type);
 226
 227static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 228{
 229	struct xfrm_state_afinfo *afinfo;
 230	const struct xfrm_type **typemap;
 231	const struct xfrm_type *type;
 232	int modload_attempted = 0;
 233
 234retry:
 235	afinfo = xfrm_state_get_afinfo(family);
 236	if (unlikely(afinfo == NULL))
 237		return NULL;
 238	typemap = afinfo->type_map;
 239
 240	type = typemap[proto];
 241	if (unlikely(type && !try_module_get(type->owner)))
 242		type = NULL;
 243	if (!type && !modload_attempted) {
 244		xfrm_state_put_afinfo(afinfo);
 245		request_module("xfrm-type-%d-%d", family, proto);
 246		modload_attempted = 1;
 247		goto retry;
 248	}
 249
 250	xfrm_state_put_afinfo(afinfo);
 251	return type;
 252}
 253
 254static void xfrm_put_type(const struct xfrm_type *type)
 255{
 256	module_put(type->owner);
 257}
 258
 259int xfrm_register_mode(struct xfrm_mode *mode, int family)
 260{
 261	struct xfrm_state_afinfo *afinfo;
 262	struct xfrm_mode **modemap;
 263	int err;
 264
 265	if (unlikely(mode->encap >= XFRM_MODE_MAX))
 266		return -EINVAL;
 267
 268	afinfo = xfrm_state_lock_afinfo(family);
 269	if (unlikely(afinfo == NULL))
 270		return -EAFNOSUPPORT;
 271
 272	err = -EEXIST;
 273	modemap = afinfo->mode_map;
 274	if (modemap[mode->encap])
 275		goto out;
 276
 277	err = -ENOENT;
 278	if (!try_module_get(afinfo->owner))
 279		goto out;
 280
 281	mode->afinfo = afinfo;
 282	modemap[mode->encap] = mode;
 283	err = 0;
 284
 285out:
 286	xfrm_state_unlock_afinfo(afinfo);
 287	return err;
 288}
 289EXPORT_SYMBOL(xfrm_register_mode);
 290
 291int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 292{
 293	struct xfrm_state_afinfo *afinfo;
 294	struct xfrm_mode **modemap;
 295	int err;
 296
 297	if (unlikely(mode->encap >= XFRM_MODE_MAX))
 298		return -EINVAL;
 299
 300	afinfo = xfrm_state_lock_afinfo(family);
 301	if (unlikely(afinfo == NULL))
 302		return -EAFNOSUPPORT;
 303
 304	err = -ENOENT;
 305	modemap = afinfo->mode_map;
 306	if (likely(modemap[mode->encap] == mode)) {
 307		modemap[mode->encap] = NULL;
 308		module_put(mode->afinfo->owner);
 309		err = 0;
 310	}
 311
 312	xfrm_state_unlock_afinfo(afinfo);
 313	return err;
 314}
 315EXPORT_SYMBOL(xfrm_unregister_mode);
 316
 317static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 318{
 319	struct xfrm_state_afinfo *afinfo;
 320	struct xfrm_mode *mode;
 321	int modload_attempted = 0;
 322
 323	if (unlikely(encap >= XFRM_MODE_MAX))
 324		return NULL;
 325
 326retry:
 327	afinfo = xfrm_state_get_afinfo(family);
 328	if (unlikely(afinfo == NULL))
 329		return NULL;
 330
 331	mode = afinfo->mode_map[encap];
 332	if (unlikely(mode && !try_module_get(mode->owner)))
 333		mode = NULL;
 334	if (!mode && !modload_attempted) {
 335		xfrm_state_put_afinfo(afinfo);
 336		request_module("xfrm-mode-%d-%d", family, encap);
 337		modload_attempted = 1;
 338		goto retry;
 339	}
 340
 341	xfrm_state_put_afinfo(afinfo);
 342	return mode;
 343}
 344
 345static void xfrm_put_mode(struct xfrm_mode *mode)
 346{
 347	module_put(mode->owner);
 348}
 349
 350static void xfrm_state_gc_destroy(struct xfrm_state *x)
 351{
 352	tasklet_hrtimer_cancel(&x->mtimer);
 353	del_timer_sync(&x->rtimer);
 354	kfree(x->aalg);
 355	kfree(x->ealg);
 356	kfree(x->calg);
 357	kfree(x->encap);
 358	kfree(x->coaddr);
 359	kfree(x->replay_esn);
 360	kfree(x->preplay_esn);
 361	if (x->inner_mode)
 362		xfrm_put_mode(x->inner_mode);
 363	if (x->inner_mode_iaf)
 364		xfrm_put_mode(x->inner_mode_iaf);
 365	if (x->outer_mode)
 366		xfrm_put_mode(x->outer_mode);
 367	if (x->type) {
 368		x->type->destructor(x);
 369		xfrm_put_type(x->type);
 370	}
 371	security_xfrm_state_free(x);
 372	kfree(x);
 373}
 374
 375static void xfrm_state_gc_task(struct work_struct *work)
 376{
 377	struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 378	struct xfrm_state *x;
 379	struct hlist_node *entry, *tmp;
 380	struct hlist_head gc_list;
 381
 382	spin_lock_bh(&xfrm_state_gc_lock);
 383	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 384	spin_unlock_bh(&xfrm_state_gc_lock);
 385
 386	hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 387		xfrm_state_gc_destroy(x);
 388
 389	wake_up(&net->xfrm.km_waitq);
 390}
 391
 392static inline unsigned long make_jiffies(long secs)
 393{
 394	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 395		return MAX_SCHEDULE_TIMEOUT-1;
 396	else
 397		return secs*HZ;
 398}
 399
 400static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 401{
 402	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 403	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 404	struct net *net = xs_net(x);
 405	unsigned long now = get_seconds();
 406	long next = LONG_MAX;
 407	int warn = 0;
 408	int err = 0;
 409
 410	spin_lock(&x->lock);
 411	if (x->km.state == XFRM_STATE_DEAD)
 412		goto out;
 413	if (x->km.state == XFRM_STATE_EXPIRED)
 414		goto expired;
 415	if (x->lft.hard_add_expires_seconds) {
 416		long tmo = x->lft.hard_add_expires_seconds +
 417			x->curlft.add_time - now;
 418		if (tmo <= 0)
 419			goto expired;
 420		if (tmo < next)
 421			next = tmo;
 422	}
 423	if (x->lft.hard_use_expires_seconds) {
 424		long tmo = x->lft.hard_use_expires_seconds +
 425			(x->curlft.use_time ? : now) - now;
 426		if (tmo <= 0)
 427			goto expired;
 428		if (tmo < next)
 429			next = tmo;
 430	}
 431	if (x->km.dying)
 432		goto resched;
 433	if (x->lft.soft_add_expires_seconds) {
 434		long tmo = x->lft.soft_add_expires_seconds +
 435			x->curlft.add_time - now;
 436		if (tmo <= 0)
 437			warn = 1;
 438		else if (tmo < next)
 439			next = tmo;
 440	}
 441	if (x->lft.soft_use_expires_seconds) {
 442		long tmo = x->lft.soft_use_expires_seconds +
 443			(x->curlft.use_time ? : now) - now;
 444		if (tmo <= 0)
 445			warn = 1;
 446		else if (tmo < next)
 447			next = tmo;
 448	}
 449
 450	x->km.dying = warn;
 451	if (warn)
 452		km_state_expired(x, 0, 0);
 453resched:
 454	if (next != LONG_MAX){
 455		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 456	}
 457
 458	goto out;
 459
 460expired:
 461	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 462		x->km.state = XFRM_STATE_EXPIRED;
 463		wake_up(&net->xfrm.km_waitq);
 464		next = 2;
 465		goto resched;
 466	}
 467
 468	err = __xfrm_state_delete(x);
 469	if (!err && x->id.spi)
 470		km_state_expired(x, 1, 0);
 471
 472	xfrm_audit_state_delete(x, err ? 0 : 1,
 473				audit_get_loginuid(current),
 474				audit_get_sessionid(current), 0);
 475
 476out:
 477	spin_unlock(&x->lock);
 478	return HRTIMER_NORESTART;
 479}
 480
 481static void xfrm_replay_timer_handler(unsigned long data);
 482
 483struct xfrm_state *xfrm_state_alloc(struct net *net)
 484{
 485	struct xfrm_state *x;
 486
 487	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 488
 489	if (x) {
 490		write_pnet(&x->xs_net, net);
 491		atomic_set(&x->refcnt, 1);
 492		atomic_set(&x->tunnel_users, 0);
 493		INIT_LIST_HEAD(&x->km.all);
 494		INIT_HLIST_NODE(&x->bydst);
 495		INIT_HLIST_NODE(&x->bysrc);
 496		INIT_HLIST_NODE(&x->byspi);
 497		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 498		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 499				(unsigned long)x);
 500		x->curlft.add_time = get_seconds();
 501		x->lft.soft_byte_limit = XFRM_INF;
 502		x->lft.soft_packet_limit = XFRM_INF;
 503		x->lft.hard_byte_limit = XFRM_INF;
 504		x->lft.hard_packet_limit = XFRM_INF;
 505		x->replay_maxage = 0;
 506		x->replay_maxdiff = 0;
 507		x->inner_mode = NULL;
 508		x->inner_mode_iaf = NULL;
 509		spin_lock_init(&x->lock);
 510	}
 511	return x;
 512}
 513EXPORT_SYMBOL(xfrm_state_alloc);
 514
 515void __xfrm_state_destroy(struct xfrm_state *x)
 516{
 517	struct net *net = xs_net(x);
 518
 519	WARN_ON(x->km.state != XFRM_STATE_DEAD);
 520
 521	spin_lock_bh(&xfrm_state_gc_lock);
 522	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 523	spin_unlock_bh(&xfrm_state_gc_lock);
 524	schedule_work(&net->xfrm.state_gc_work);
 525}
 526EXPORT_SYMBOL(__xfrm_state_destroy);
 527
 528int __xfrm_state_delete(struct xfrm_state *x)
 529{
 530	struct net *net = xs_net(x);
 531	int err = -ESRCH;
 532
 533	if (x->km.state != XFRM_STATE_DEAD) {
 534		x->km.state = XFRM_STATE_DEAD;
 535		spin_lock(&xfrm_state_lock);
 536		list_del(&x->km.all);
 537		hlist_del(&x->bydst);
 538		hlist_del(&x->bysrc);
 539		if (x->id.spi)
 540			hlist_del(&x->byspi);
 541		net->xfrm.state_num--;
 542		spin_unlock(&xfrm_state_lock);
 543
 544		/* All xfrm_state objects are created by xfrm_state_alloc.
 545		 * The xfrm_state_alloc call gives a reference, and that
 546		 * is what we are dropping here.
 547		 */
 548		xfrm_state_put(x);
 549		err = 0;
 550	}
 551
 552	return err;
 553}
 554EXPORT_SYMBOL(__xfrm_state_delete);
 555
 556int xfrm_state_delete(struct xfrm_state *x)
 557{
 558	int err;
 559
 560	spin_lock_bh(&x->lock);
 561	err = __xfrm_state_delete(x);
 562	spin_unlock_bh(&x->lock);
 563
 564	return err;
 565}
 566EXPORT_SYMBOL(xfrm_state_delete);
 567
 568#ifdef CONFIG_SECURITY_NETWORK_XFRM
 569static inline int
 570xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 571{
 572	int i, err = 0;
 573
 574	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 575		struct hlist_node *entry;
 576		struct xfrm_state *x;
 577
 578		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 579			if (xfrm_id_proto_match(x->id.proto, proto) &&
 580			   (err = security_xfrm_state_delete(x)) != 0) {
 581				xfrm_audit_state_delete(x, 0,
 582							audit_info->loginuid,
 583							audit_info->sessionid,
 584							audit_info->secid);
 585				return err;
 586			}
 587		}
 588	}
 589
 590	return err;
 591}
 592#else
 593static inline int
 594xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 595{
 596	return 0;
 597}
 598#endif
 599
 600int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 601{
 602	int i, err = 0, cnt = 0;
 603
 604	spin_lock_bh(&xfrm_state_lock);
 605	err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 606	if (err)
 607		goto out;
 608
 609	err = -ESRCH;
 610	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 611		struct hlist_node *entry;
 612		struct xfrm_state *x;
 613restart:
 614		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 615			if (!xfrm_state_kern(x) &&
 616			    xfrm_id_proto_match(x->id.proto, proto)) {
 617				xfrm_state_hold(x);
 618				spin_unlock_bh(&xfrm_state_lock);
 619
 620				err = xfrm_state_delete(x);
 621				xfrm_audit_state_delete(x, err ? 0 : 1,
 622							audit_info->loginuid,
 623							audit_info->sessionid,
 624							audit_info->secid);
 625				xfrm_state_put(x);
 626				if (!err)
 627					cnt++;
 628
 629				spin_lock_bh(&xfrm_state_lock);
 630				goto restart;
 631			}
 632		}
 633	}
 634	if (cnt)
 635		err = 0;
 636
 637out:
 638	spin_unlock_bh(&xfrm_state_lock);
 639	wake_up(&net->xfrm.km_waitq);
 640	return err;
 641}
 642EXPORT_SYMBOL(xfrm_state_flush);
 643
 644void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 645{
 646	spin_lock_bh(&xfrm_state_lock);
 647	si->sadcnt = net->xfrm.state_num;
 648	si->sadhcnt = net->xfrm.state_hmask;
 649	si->sadhmcnt = xfrm_state_hashmax;
 650	spin_unlock_bh(&xfrm_state_lock);
 651}
 652EXPORT_SYMBOL(xfrm_sad_getinfo);
 653
 654static int
 655xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 656		    const struct xfrm_tmpl *tmpl,
 657		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 658		    unsigned short family)
 659{
 660	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 661	if (!afinfo)
 662		return -1;
 663	afinfo->init_tempsel(&x->sel, fl);
 664
 665	if (family != tmpl->encap_family) {
 666		xfrm_state_put_afinfo(afinfo);
 667		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 668		if (!afinfo)
 669			return -1;
 670	}
 671	afinfo->init_temprop(x, tmpl, daddr, saddr);
 672	xfrm_state_put_afinfo(afinfo);
 673	return 0;
 674}
 675
 676static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 677					      const xfrm_address_t *daddr,
 678					      __be32 spi, u8 proto,
 679					      unsigned short family)
 680{
 681	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 682	struct xfrm_state *x;
 683	struct hlist_node *entry;
 684
 685	hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 686		if (x->props.family != family ||
 687		    x->id.spi       != spi ||
 688		    x->id.proto     != proto ||
 689		    xfrm_addr_cmp(&x->id.daddr, daddr, family))
 690			continue;
 691
 692		if ((mark & x->mark.m) != x->mark.v)
 693			continue;
 694		xfrm_state_hold(x);
 695		return x;
 696	}
 697
 698	return NULL;
 699}
 700
 701static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 702						     const xfrm_address_t *daddr,
 703						     const xfrm_address_t *saddr,
 704						     u8 proto, unsigned short family)
 705{
 706	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 707	struct xfrm_state *x;
 708	struct hlist_node *entry;
 709
 710	hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 711		if (x->props.family != family ||
 712		    x->id.proto     != proto ||
 713		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 714		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
 715			continue;
 716
 717		if ((mark & x->mark.m) != x->mark.v)
 718			continue;
 719		xfrm_state_hold(x);
 720		return x;
 721	}
 722
 723	return NULL;
 724}
 725
 726static inline struct xfrm_state *
 727__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 728{
 729	struct net *net = xs_net(x);
 730	u32 mark = x->mark.v & x->mark.m;
 731
 732	if (use_spi)
 733		return __xfrm_state_lookup(net, mark, &x->id.daddr,
 734					   x->id.spi, x->id.proto, family);
 735	else
 736		return __xfrm_state_lookup_byaddr(net, mark,
 737						  &x->id.daddr,
 738						  &x->props.saddr,
 739						  x->id.proto, family);
 740}
 741
 742static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 743{
 744	if (have_hash_collision &&
 745	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 746	    net->xfrm.state_num > net->xfrm.state_hmask)
 747		schedule_work(&net->xfrm.state_hash_work);
 748}
 749
 750static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 751			       const struct flowi *fl, unsigned short family,
 752			       struct xfrm_state **best, int *acq_in_progress,
 753			       int *error)
 754{
 755	/* Resolution logic:
 756	 * 1. There is a valid state with matching selector. Done.
 757	 * 2. Valid state with inappropriate selector. Skip.
 758	 *
 759	 * Entering area of "sysdeps".
 760	 *
 761	 * 3. If state is not valid, selector is temporary, it selects
 762	 *    only session which triggered previous resolution. Key
 763	 *    manager will do something to install a state with proper
 764	 *    selector.
 765	 */
 766	if (x->km.state == XFRM_STATE_VALID) {
 767		if ((x->sel.family &&
 768		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 769		    !security_xfrm_state_pol_flow_match(x, pol, fl))
 770			return;
 771
 772		if (!*best ||
 773		    (*best)->km.dying > x->km.dying ||
 774		    ((*best)->km.dying == x->km.dying &&
 775		     (*best)->curlft.add_time < x->curlft.add_time))
 776			*best = x;
 777	} else if (x->km.state == XFRM_STATE_ACQ) {
 778		*acq_in_progress = 1;
 779	} else if (x->km.state == XFRM_STATE_ERROR ||
 780		   x->km.state == XFRM_STATE_EXPIRED) {
 781		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 782		    security_xfrm_state_pol_flow_match(x, pol, fl))
 783			*error = -ESRCH;
 784	}
 785}
 786
 787struct xfrm_state *
 788xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 789		const struct flowi *fl, struct xfrm_tmpl *tmpl,
 790		struct xfrm_policy *pol, int *err,
 791		unsigned short family)
 792{
 793	static xfrm_address_t saddr_wildcard = { };
 794	struct net *net = xp_net(pol);
 795	unsigned int h, h_wildcard;
 796	struct hlist_node *entry;
 797	struct xfrm_state *x, *x0, *to_put;
 798	int acquire_in_progress = 0;
 799	int error = 0;
 800	struct xfrm_state *best = NULL;
 801	u32 mark = pol->mark.v & pol->mark.m;
 802	unsigned short encap_family = tmpl->encap_family;
 803
 804	to_put = NULL;
 805
 806	spin_lock_bh(&xfrm_state_lock);
 807	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 808	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 809		if (x->props.family == encap_family &&
 810		    x->props.reqid == tmpl->reqid &&
 811		    (mark & x->mark.m) == x->mark.v &&
 812		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
 813		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 814		    tmpl->mode == x->props.mode &&
 815		    tmpl->id.proto == x->id.proto &&
 816		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 817			xfrm_state_look_at(pol, x, fl, encap_family,
 818					   &best, &acquire_in_progress, &error);
 819	}
 820	if (best)
 821		goto found;
 822
 823	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 824	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 825		if (x->props.family == encap_family &&
 826		    x->props.reqid == tmpl->reqid &&
 827		    (mark & x->mark.m) == x->mark.v &&
 828		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
 829		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 830		    tmpl->mode == x->props.mode &&
 831		    tmpl->id.proto == x->id.proto &&
 832		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 833			xfrm_state_look_at(pol, x, fl, encap_family,
 834					   &best, &acquire_in_progress, &error);
 835	}
 836
 837found:
 838	x = best;
 839	if (!x && !error && !acquire_in_progress) {
 840		if (tmpl->id.spi &&
 841		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 842					      tmpl->id.proto, encap_family)) != NULL) {
 843			to_put = x0;
 844			error = -EEXIST;
 845			goto out;
 846		}
 847		x = xfrm_state_alloc(net);
 848		if (x == NULL) {
 849			error = -ENOMEM;
 850			goto out;
 851		}
 852		/* Initialize temporary state matching only
 853		 * to current session. */
 854		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 855		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 856
 857		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 858		if (error) {
 859			x->km.state = XFRM_STATE_DEAD;
 860			to_put = x;
 861			x = NULL;
 862			goto out;
 863		}
 864
 865		if (km_query(x, tmpl, pol) == 0) {
 866			x->km.state = XFRM_STATE_ACQ;
 867			list_add(&x->km.all, &net->xfrm.state_all);
 868			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 869			h = xfrm_src_hash(net, daddr, saddr, encap_family);
 870			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 871			if (x->id.spi) {
 872				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 873				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 874			}
 875			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 876			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 877			net->xfrm.state_num++;
 878			xfrm_hash_grow_check(net, x->bydst.next != NULL);
 879		} else {
 880			x->km.state = XFRM_STATE_DEAD;
 881			to_put = x;
 882			x = NULL;
 883			error = -ESRCH;
 884		}
 885	}
 886out:
 887	if (x)
 888		xfrm_state_hold(x);
 889	else
 890		*err = acquire_in_progress ? -EAGAIN : error;
 891	spin_unlock_bh(&xfrm_state_lock);
 892	if (to_put)
 893		xfrm_state_put(to_put);
 894	return x;
 895}
 896
 897struct xfrm_state *
 898xfrm_stateonly_find(struct net *net, u32 mark,
 899		    xfrm_address_t *daddr, xfrm_address_t *saddr,
 900		    unsigned short family, u8 mode, u8 proto, u32 reqid)
 901{
 902	unsigned int h;
 903	struct xfrm_state *rx = NULL, *x = NULL;
 904	struct hlist_node *entry;
 905
 906	spin_lock(&xfrm_state_lock);
 907	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 908	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 909		if (x->props.family == family &&
 910		    x->props.reqid == reqid &&
 911		    (mark & x->mark.m) == x->mark.v &&
 912		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
 913		    xfrm_state_addr_check(x, daddr, saddr, family) &&
 914		    mode == x->props.mode &&
 915		    proto == x->id.proto &&
 916		    x->km.state == XFRM_STATE_VALID) {
 917			rx = x;
 918			break;
 919		}
 920	}
 921
 922	if (rx)
 923		xfrm_state_hold(rx);
 924	spin_unlock(&xfrm_state_lock);
 925
 926
 927	return rx;
 928}
 929EXPORT_SYMBOL(xfrm_stateonly_find);
 930
 931static void __xfrm_state_insert(struct xfrm_state *x)
 932{
 933	struct net *net = xs_net(x);
 934	unsigned int h;
 935
 936	list_add(&x->km.all, &net->xfrm.state_all);
 937
 938	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 939			  x->props.reqid, x->props.family);
 940	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 941
 942	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 943	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 944
 945	if (x->id.spi) {
 946		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 947				  x->props.family);
 948
 949		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 950	}
 951
 952	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 953	if (x->replay_maxage)
 954		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 955
 956	wake_up(&net->xfrm.km_waitq);
 957
 958	net->xfrm.state_num++;
 959
 960	xfrm_hash_grow_check(net, x->bydst.next != NULL);
 961}
 962
 963/* xfrm_state_lock is held */
 964static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 965{
 966	struct net *net = xs_net(xnew);
 967	unsigned short family = xnew->props.family;
 968	u32 reqid = xnew->props.reqid;
 969	struct xfrm_state *x;
 970	struct hlist_node *entry;
 971	unsigned int h;
 972	u32 mark = xnew->mark.v & xnew->mark.m;
 973
 974	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 975	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 976		if (x->props.family	== family &&
 977		    x->props.reqid	== reqid &&
 978		    (mark & x->mark.m) == x->mark.v &&
 979		    !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 980		    !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 981			x->genid++;
 982	}
 983}
 984
 985void xfrm_state_insert(struct xfrm_state *x)
 986{
 987	spin_lock_bh(&xfrm_state_lock);
 988	__xfrm_state_bump_genids(x);
 989	__xfrm_state_insert(x);
 990	spin_unlock_bh(&xfrm_state_lock);
 991}
 992EXPORT_SYMBOL(xfrm_state_insert);
 993
 994/* xfrm_state_lock is held */
 995static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
 996					  unsigned short family, u8 mode,
 997					  u32 reqid, u8 proto,
 998					  const xfrm_address_t *daddr,
 999					  const xfrm_address_t *saddr, int create)
1000{
1001	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1002	struct hlist_node *entry;
1003	struct xfrm_state *x;
1004	u32 mark = m->v & m->m;
1005
1006	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1007		if (x->props.reqid  != reqid ||
1008		    x->props.mode   != mode ||
1009		    x->props.family != family ||
1010		    x->km.state     != XFRM_STATE_ACQ ||
1011		    x->id.spi       != 0 ||
1012		    x->id.proto	    != proto ||
1013		    (mark & x->mark.m) != x->mark.v ||
1014		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1015		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
1016			continue;
1017
1018		xfrm_state_hold(x);
1019		return x;
1020	}
1021
1022	if (!create)
1023		return NULL;
1024
1025	x = xfrm_state_alloc(net);
1026	if (likely(x)) {
1027		switch (family) {
1028		case AF_INET:
1029			x->sel.daddr.a4 = daddr->a4;
1030			x->sel.saddr.a4 = saddr->a4;
1031			x->sel.prefixlen_d = 32;
1032			x->sel.prefixlen_s = 32;
1033			x->props.saddr.a4 = saddr->a4;
1034			x->id.daddr.a4 = daddr->a4;
1035			break;
1036
1037		case AF_INET6:
1038			*(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1039			*(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
 
 
1040			x->sel.prefixlen_d = 128;
1041			x->sel.prefixlen_s = 128;
1042			*(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1043			*(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
 
 
1044			break;
1045		}
1046
1047		x->km.state = XFRM_STATE_ACQ;
1048		x->id.proto = proto;
1049		x->props.family = family;
1050		x->props.mode = mode;
1051		x->props.reqid = reqid;
1052		x->mark.v = m->v;
1053		x->mark.m = m->m;
1054		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1055		xfrm_state_hold(x);
1056		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1057		list_add(&x->km.all, &net->xfrm.state_all);
1058		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1059		h = xfrm_src_hash(net, daddr, saddr, family);
1060		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1061
1062		net->xfrm.state_num++;
1063
1064		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1065	}
1066
1067	return x;
1068}
1069
1070static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1071
1072int xfrm_state_add(struct xfrm_state *x)
1073{
1074	struct net *net = xs_net(x);
1075	struct xfrm_state *x1, *to_put;
1076	int family;
1077	int err;
1078	u32 mark = x->mark.v & x->mark.m;
1079	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1080
1081	family = x->props.family;
1082
1083	to_put = NULL;
1084
1085	spin_lock_bh(&xfrm_state_lock);
1086
1087	x1 = __xfrm_state_locate(x, use_spi, family);
1088	if (x1) {
1089		to_put = x1;
1090		x1 = NULL;
1091		err = -EEXIST;
1092		goto out;
1093	}
1094
1095	if (use_spi && x->km.seq) {
1096		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1097		if (x1 && ((x1->id.proto != x->id.proto) ||
1098		    xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1099			to_put = x1;
1100			x1 = NULL;
1101		}
1102	}
1103
1104	if (use_spi && !x1)
1105		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1106				     x->props.reqid, x->id.proto,
1107				     &x->id.daddr, &x->props.saddr, 0);
1108
1109	__xfrm_state_bump_genids(x);
1110	__xfrm_state_insert(x);
1111	err = 0;
1112
1113out:
1114	spin_unlock_bh(&xfrm_state_lock);
1115
1116	if (x1) {
1117		xfrm_state_delete(x1);
1118		xfrm_state_put(x1);
1119	}
1120
1121	if (to_put)
1122		xfrm_state_put(to_put);
1123
1124	return err;
1125}
1126EXPORT_SYMBOL(xfrm_state_add);
1127
1128#ifdef CONFIG_XFRM_MIGRATE
1129static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1130{
1131	struct net *net = xs_net(orig);
1132	int err = -ENOMEM;
1133	struct xfrm_state *x = xfrm_state_alloc(net);
1134	if (!x)
1135		goto out;
1136
1137	memcpy(&x->id, &orig->id, sizeof(x->id));
1138	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1139	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1140	x->props.mode = orig->props.mode;
1141	x->props.replay_window = orig->props.replay_window;
1142	x->props.reqid = orig->props.reqid;
1143	x->props.family = orig->props.family;
1144	x->props.saddr = orig->props.saddr;
1145
1146	if (orig->aalg) {
1147		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1148		if (!x->aalg)
1149			goto error;
1150	}
1151	x->props.aalgo = orig->props.aalgo;
1152
1153	if (orig->ealg) {
1154		x->ealg = xfrm_algo_clone(orig->ealg);
1155		if (!x->ealg)
1156			goto error;
1157	}
1158	x->props.ealgo = orig->props.ealgo;
1159
1160	if (orig->calg) {
1161		x->calg = xfrm_algo_clone(orig->calg);
1162		if (!x->calg)
1163			goto error;
1164	}
1165	x->props.calgo = orig->props.calgo;
1166
1167	if (orig->encap) {
1168		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1169		if (!x->encap)
1170			goto error;
1171	}
1172
1173	if (orig->coaddr) {
1174		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1175				    GFP_KERNEL);
1176		if (!x->coaddr)
1177			goto error;
1178	}
1179
1180	if (orig->replay_esn) {
1181		err = xfrm_replay_clone(x, orig);
1182		if (err)
1183			goto error;
1184	}
1185
1186	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1187
1188	err = xfrm_init_state(x);
1189	if (err)
1190		goto error;
1191
1192	x->props.flags = orig->props.flags;
1193
1194	x->curlft.add_time = orig->curlft.add_time;
1195	x->km.state = orig->km.state;
1196	x->km.seq = orig->km.seq;
1197
1198	return x;
1199
1200 error:
1201	xfrm_state_put(x);
1202out:
1203	if (errp)
1204		*errp = err;
1205	return NULL;
1206}
1207
1208/* xfrm_state_lock is held */
1209struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1210{
1211	unsigned int h;
1212	struct xfrm_state *x;
1213	struct hlist_node *entry;
1214
1215	if (m->reqid) {
1216		h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1217				  m->reqid, m->old_family);
1218		hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1219			if (x->props.mode != m->mode ||
1220			    x->id.proto != m->proto)
1221				continue;
1222			if (m->reqid && x->props.reqid != m->reqid)
1223				continue;
1224			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1225					  m->old_family) ||
1226			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1227					  m->old_family))
1228				continue;
1229			xfrm_state_hold(x);
1230			return x;
1231		}
1232	} else {
1233		h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1234				  m->old_family);
1235		hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1236			if (x->props.mode != m->mode ||
1237			    x->id.proto != m->proto)
1238				continue;
1239			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1240					  m->old_family) ||
1241			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1242					  m->old_family))
1243				continue;
1244			xfrm_state_hold(x);
1245			return x;
1246		}
1247	}
1248
1249	return NULL;
1250}
1251EXPORT_SYMBOL(xfrm_migrate_state_find);
1252
1253struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1254				       struct xfrm_migrate *m)
1255{
1256	struct xfrm_state *xc;
1257	int err;
1258
1259	xc = xfrm_state_clone(x, &err);
1260	if (!xc)
1261		return NULL;
1262
1263	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1264	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1265
1266	/* add state */
1267	if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1268		/* a care is needed when the destination address of the
1269		   state is to be updated as it is a part of triplet */
1270		xfrm_state_insert(xc);
1271	} else {
1272		if ((err = xfrm_state_add(xc)) < 0)
1273			goto error;
1274	}
1275
1276	return xc;
1277error:
1278	xfrm_state_put(xc);
1279	return NULL;
1280}
1281EXPORT_SYMBOL(xfrm_state_migrate);
1282#endif
1283
1284int xfrm_state_update(struct xfrm_state *x)
1285{
1286	struct xfrm_state *x1, *to_put;
1287	int err;
1288	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1289
1290	to_put = NULL;
1291
1292	spin_lock_bh(&xfrm_state_lock);
1293	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1294
1295	err = -ESRCH;
1296	if (!x1)
1297		goto out;
1298
1299	if (xfrm_state_kern(x1)) {
1300		to_put = x1;
1301		err = -EEXIST;
1302		goto out;
1303	}
1304
1305	if (x1->km.state == XFRM_STATE_ACQ) {
1306		__xfrm_state_insert(x);
1307		x = NULL;
1308	}
1309	err = 0;
1310
1311out:
1312	spin_unlock_bh(&xfrm_state_lock);
1313
1314	if (to_put)
1315		xfrm_state_put(to_put);
1316
1317	if (err)
1318		return err;
1319
1320	if (!x) {
1321		xfrm_state_delete(x1);
1322		xfrm_state_put(x1);
1323		return 0;
1324	}
1325
1326	err = -EINVAL;
1327	spin_lock_bh(&x1->lock);
1328	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1329		if (x->encap && x1->encap)
1330			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1331		if (x->coaddr && x1->coaddr) {
1332			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1333		}
1334		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1335			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1336		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1337		x1->km.dying = 0;
1338
1339		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1340		if (x1->curlft.use_time)
1341			xfrm_state_check_expire(x1);
1342
1343		err = 0;
1344		x->km.state = XFRM_STATE_DEAD;
1345		__xfrm_state_put(x);
1346	}
1347	spin_unlock_bh(&x1->lock);
1348
1349	xfrm_state_put(x1);
1350
1351	return err;
1352}
1353EXPORT_SYMBOL(xfrm_state_update);
1354
1355int xfrm_state_check_expire(struct xfrm_state *x)
1356{
1357	if (!x->curlft.use_time)
1358		x->curlft.use_time = get_seconds();
1359
1360	if (x->km.state != XFRM_STATE_VALID)
1361		return -EINVAL;
1362
1363	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1364	    x->curlft.packets >= x->lft.hard_packet_limit) {
1365		x->km.state = XFRM_STATE_EXPIRED;
1366		tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1367		return -EINVAL;
1368	}
1369
1370	if (!x->km.dying &&
1371	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1372	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1373		x->km.dying = 1;
1374		km_state_expired(x, 0, 0);
1375	}
1376	return 0;
1377}
1378EXPORT_SYMBOL(xfrm_state_check_expire);
1379
1380struct xfrm_state *
1381xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1382		  u8 proto, unsigned short family)
1383{
1384	struct xfrm_state *x;
1385
1386	spin_lock_bh(&xfrm_state_lock);
1387	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1388	spin_unlock_bh(&xfrm_state_lock);
1389	return x;
1390}
1391EXPORT_SYMBOL(xfrm_state_lookup);
1392
1393struct xfrm_state *
1394xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1395			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1396			 u8 proto, unsigned short family)
1397{
1398	struct xfrm_state *x;
1399
1400	spin_lock_bh(&xfrm_state_lock);
1401	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1402	spin_unlock_bh(&xfrm_state_lock);
1403	return x;
1404}
1405EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1406
1407struct xfrm_state *
1408xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1409	      const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1410	      int create, unsigned short family)
1411{
1412	struct xfrm_state *x;
1413
1414	spin_lock_bh(&xfrm_state_lock);
1415	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1416	spin_unlock_bh(&xfrm_state_lock);
1417
1418	return x;
1419}
1420EXPORT_SYMBOL(xfrm_find_acq);
1421
1422#ifdef CONFIG_XFRM_SUB_POLICY
1423int
1424xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1425	       unsigned short family)
1426{
1427	int err = 0;
1428	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1429	if (!afinfo)
1430		return -EAFNOSUPPORT;
1431
1432	spin_lock_bh(&xfrm_state_lock);
1433	if (afinfo->tmpl_sort)
1434		err = afinfo->tmpl_sort(dst, src, n);
1435	spin_unlock_bh(&xfrm_state_lock);
1436	xfrm_state_put_afinfo(afinfo);
1437	return err;
1438}
1439EXPORT_SYMBOL(xfrm_tmpl_sort);
1440
1441int
1442xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1443		unsigned short family)
1444{
1445	int err = 0;
1446	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1447	if (!afinfo)
1448		return -EAFNOSUPPORT;
1449
1450	spin_lock_bh(&xfrm_state_lock);
1451	if (afinfo->state_sort)
1452		err = afinfo->state_sort(dst, src, n);
1453	spin_unlock_bh(&xfrm_state_lock);
1454	xfrm_state_put_afinfo(afinfo);
1455	return err;
1456}
1457EXPORT_SYMBOL(xfrm_state_sort);
1458#endif
1459
1460/* Silly enough, but I'm lazy to build resolution list */
1461
1462static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1463{
1464	int i;
1465
1466	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1467		struct hlist_node *entry;
1468		struct xfrm_state *x;
1469
1470		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1471			if (x->km.seq == seq &&
1472			    (mark & x->mark.m) == x->mark.v &&
1473			    x->km.state == XFRM_STATE_ACQ) {
1474				xfrm_state_hold(x);
1475				return x;
1476			}
1477		}
1478	}
1479	return NULL;
1480}
1481
1482struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1483{
1484	struct xfrm_state *x;
1485
1486	spin_lock_bh(&xfrm_state_lock);
1487	x = __xfrm_find_acq_byseq(net, mark, seq);
1488	spin_unlock_bh(&xfrm_state_lock);
1489	return x;
1490}
1491EXPORT_SYMBOL(xfrm_find_acq_byseq);
1492
1493u32 xfrm_get_acqseq(void)
1494{
1495	u32 res;
1496	static atomic_t acqseq;
1497
1498	do {
1499		res = atomic_inc_return(&acqseq);
1500	} while (!res);
1501
1502	return res;
1503}
1504EXPORT_SYMBOL(xfrm_get_acqseq);
1505
1506int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1507{
1508	struct net *net = xs_net(x);
1509	unsigned int h;
1510	struct xfrm_state *x0;
1511	int err = -ENOENT;
1512	__be32 minspi = htonl(low);
1513	__be32 maxspi = htonl(high);
1514	u32 mark = x->mark.v & x->mark.m;
1515
1516	spin_lock_bh(&x->lock);
1517	if (x->km.state == XFRM_STATE_DEAD)
1518		goto unlock;
1519
1520	err = 0;
1521	if (x->id.spi)
1522		goto unlock;
1523
1524	err = -ENOENT;
1525
1526	if (minspi == maxspi) {
1527		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1528		if (x0) {
1529			xfrm_state_put(x0);
1530			goto unlock;
1531		}
1532		x->id.spi = minspi;
1533	} else {
1534		u32 spi = 0;
1535		for (h=0; h<high-low+1; h++) {
1536			spi = low + net_random()%(high-low+1);
1537			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1538			if (x0 == NULL) {
1539				x->id.spi = htonl(spi);
1540				break;
1541			}
1542			xfrm_state_put(x0);
1543		}
1544	}
1545	if (x->id.spi) {
1546		spin_lock_bh(&xfrm_state_lock);
1547		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1548		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1549		spin_unlock_bh(&xfrm_state_lock);
1550
1551		err = 0;
1552	}
1553
1554unlock:
1555	spin_unlock_bh(&x->lock);
1556
1557	return err;
1558}
1559EXPORT_SYMBOL(xfrm_alloc_spi);
1560
1561int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1562		    int (*func)(struct xfrm_state *, int, void*),
1563		    void *data)
1564{
1565	struct xfrm_state *state;
1566	struct xfrm_state_walk *x;
1567	int err = 0;
1568
1569	if (walk->seq != 0 && list_empty(&walk->all))
1570		return 0;
1571
1572	spin_lock_bh(&xfrm_state_lock);
1573	if (list_empty(&walk->all))
1574		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1575	else
1576		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1577	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1578		if (x->state == XFRM_STATE_DEAD)
1579			continue;
1580		state = container_of(x, struct xfrm_state, km);
1581		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1582			continue;
1583		err = func(state, walk->seq, data);
1584		if (err) {
1585			list_move_tail(&walk->all, &x->all);
1586			goto out;
1587		}
1588		walk->seq++;
1589	}
1590	if (walk->seq == 0) {
1591		err = -ENOENT;
1592		goto out;
1593	}
1594	list_del_init(&walk->all);
1595out:
1596	spin_unlock_bh(&xfrm_state_lock);
1597	return err;
1598}
1599EXPORT_SYMBOL(xfrm_state_walk);
1600
1601void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1602{
1603	INIT_LIST_HEAD(&walk->all);
1604	walk->proto = proto;
1605	walk->state = XFRM_STATE_DEAD;
1606	walk->seq = 0;
1607}
1608EXPORT_SYMBOL(xfrm_state_walk_init);
1609
1610void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1611{
1612	if (list_empty(&walk->all))
1613		return;
1614
1615	spin_lock_bh(&xfrm_state_lock);
1616	list_del(&walk->all);
1617	spin_unlock_bh(&xfrm_state_lock);
1618}
1619EXPORT_SYMBOL(xfrm_state_walk_done);
1620
1621static void xfrm_replay_timer_handler(unsigned long data)
1622{
1623	struct xfrm_state *x = (struct xfrm_state*)data;
1624
1625	spin_lock(&x->lock);
1626
1627	if (x->km.state == XFRM_STATE_VALID) {
1628		if (xfrm_aevent_is_on(xs_net(x)))
1629			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1630		else
1631			x->xflags |= XFRM_TIME_DEFER;
1632	}
1633
1634	spin_unlock(&x->lock);
1635}
1636
1637static LIST_HEAD(xfrm_km_list);
1638static DEFINE_RWLOCK(xfrm_km_lock);
1639
1640void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1641{
1642	struct xfrm_mgr *km;
1643
1644	read_lock(&xfrm_km_lock);
1645	list_for_each_entry(km, &xfrm_km_list, list)
1646		if (km->notify_policy)
1647			km->notify_policy(xp, dir, c);
1648	read_unlock(&xfrm_km_lock);
1649}
1650
1651void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1652{
1653	struct xfrm_mgr *km;
1654	read_lock(&xfrm_km_lock);
1655	list_for_each_entry(km, &xfrm_km_list, list)
1656		if (km->notify)
1657			km->notify(x, c);
1658	read_unlock(&xfrm_km_lock);
1659}
1660
1661EXPORT_SYMBOL(km_policy_notify);
1662EXPORT_SYMBOL(km_state_notify);
1663
1664void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1665{
1666	struct net *net = xs_net(x);
1667	struct km_event c;
1668
1669	c.data.hard = hard;
1670	c.pid = pid;
1671	c.event = XFRM_MSG_EXPIRE;
1672	km_state_notify(x, &c);
1673
1674	if (hard)
1675		wake_up(&net->xfrm.km_waitq);
1676}
1677
1678EXPORT_SYMBOL(km_state_expired);
1679/*
1680 * We send to all registered managers regardless of failure
1681 * We are happy with one success
1682*/
1683int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1684{
1685	int err = -EINVAL, acqret;
1686	struct xfrm_mgr *km;
1687
1688	read_lock(&xfrm_km_lock);
1689	list_for_each_entry(km, &xfrm_km_list, list) {
1690		acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1691		if (!acqret)
1692			err = acqret;
1693	}
1694	read_unlock(&xfrm_km_lock);
1695	return err;
1696}
1697EXPORT_SYMBOL(km_query);
1698
1699int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1700{
1701	int err = -EINVAL;
1702	struct xfrm_mgr *km;
1703
1704	read_lock(&xfrm_km_lock);
1705	list_for_each_entry(km, &xfrm_km_list, list) {
1706		if (km->new_mapping)
1707			err = km->new_mapping(x, ipaddr, sport);
1708		if (!err)
1709			break;
1710	}
1711	read_unlock(&xfrm_km_lock);
1712	return err;
1713}
1714EXPORT_SYMBOL(km_new_mapping);
1715
1716void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1717{
1718	struct net *net = xp_net(pol);
1719	struct km_event c;
1720
1721	c.data.hard = hard;
1722	c.pid = pid;
1723	c.event = XFRM_MSG_POLEXPIRE;
1724	km_policy_notify(pol, dir, &c);
1725
1726	if (hard)
1727		wake_up(&net->xfrm.km_waitq);
1728}
1729EXPORT_SYMBOL(km_policy_expired);
1730
1731#ifdef CONFIG_XFRM_MIGRATE
1732int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1733	       const struct xfrm_migrate *m, int num_migrate,
1734	       const struct xfrm_kmaddress *k)
1735{
1736	int err = -EINVAL;
1737	int ret;
1738	struct xfrm_mgr *km;
1739
1740	read_lock(&xfrm_km_lock);
1741	list_for_each_entry(km, &xfrm_km_list, list) {
1742		if (km->migrate) {
1743			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1744			if (!ret)
1745				err = ret;
1746		}
1747	}
1748	read_unlock(&xfrm_km_lock);
1749	return err;
1750}
1751EXPORT_SYMBOL(km_migrate);
1752#endif
1753
1754int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1755{
1756	int err = -EINVAL;
1757	int ret;
1758	struct xfrm_mgr *km;
1759
1760	read_lock(&xfrm_km_lock);
1761	list_for_each_entry(km, &xfrm_km_list, list) {
1762		if (km->report) {
1763			ret = km->report(net, proto, sel, addr);
1764			if (!ret)
1765				err = ret;
1766		}
1767	}
1768	read_unlock(&xfrm_km_lock);
1769	return err;
1770}
1771EXPORT_SYMBOL(km_report);
1772
1773int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1774{
1775	int err;
1776	u8 *data;
1777	struct xfrm_mgr *km;
1778	struct xfrm_policy *pol = NULL;
1779
1780	if (optlen <= 0 || optlen > PAGE_SIZE)
1781		return -EMSGSIZE;
1782
1783	data = kmalloc(optlen, GFP_KERNEL);
1784	if (!data)
1785		return -ENOMEM;
1786
1787	err = -EFAULT;
1788	if (copy_from_user(data, optval, optlen))
1789		goto out;
1790
1791	err = -EINVAL;
1792	read_lock(&xfrm_km_lock);
1793	list_for_each_entry(km, &xfrm_km_list, list) {
1794		pol = km->compile_policy(sk, optname, data,
1795					 optlen, &err);
1796		if (err >= 0)
1797			break;
1798	}
1799	read_unlock(&xfrm_km_lock);
1800
1801	if (err >= 0) {
1802		xfrm_sk_policy_insert(sk, err, pol);
1803		xfrm_pol_put(pol);
1804		err = 0;
1805	}
1806
1807out:
1808	kfree(data);
1809	return err;
1810}
1811EXPORT_SYMBOL(xfrm_user_policy);
1812
1813int xfrm_register_km(struct xfrm_mgr *km)
1814{
1815	write_lock_bh(&xfrm_km_lock);
1816	list_add_tail(&km->list, &xfrm_km_list);
1817	write_unlock_bh(&xfrm_km_lock);
1818	return 0;
1819}
1820EXPORT_SYMBOL(xfrm_register_km);
1821
1822int xfrm_unregister_km(struct xfrm_mgr *km)
1823{
1824	write_lock_bh(&xfrm_km_lock);
1825	list_del(&km->list);
1826	write_unlock_bh(&xfrm_km_lock);
1827	return 0;
1828}
1829EXPORT_SYMBOL(xfrm_unregister_km);
1830
1831int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1832{
1833	int err = 0;
1834	if (unlikely(afinfo == NULL))
1835		return -EINVAL;
1836	if (unlikely(afinfo->family >= NPROTO))
1837		return -EAFNOSUPPORT;
1838	write_lock_bh(&xfrm_state_afinfo_lock);
1839	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1840		err = -ENOBUFS;
1841	else
1842		xfrm_state_afinfo[afinfo->family] = afinfo;
1843	write_unlock_bh(&xfrm_state_afinfo_lock);
1844	return err;
1845}
1846EXPORT_SYMBOL(xfrm_state_register_afinfo);
1847
1848int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1849{
1850	int err = 0;
1851	if (unlikely(afinfo == NULL))
1852		return -EINVAL;
1853	if (unlikely(afinfo->family >= NPROTO))
1854		return -EAFNOSUPPORT;
1855	write_lock_bh(&xfrm_state_afinfo_lock);
1856	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1857		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1858			err = -EINVAL;
1859		else
1860			xfrm_state_afinfo[afinfo->family] = NULL;
1861	}
1862	write_unlock_bh(&xfrm_state_afinfo_lock);
1863	return err;
1864}
1865EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1866
1867static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1868{
1869	struct xfrm_state_afinfo *afinfo;
1870	if (unlikely(family >= NPROTO))
1871		return NULL;
1872	read_lock(&xfrm_state_afinfo_lock);
1873	afinfo = xfrm_state_afinfo[family];
1874	if (unlikely(!afinfo))
1875		read_unlock(&xfrm_state_afinfo_lock);
1876	return afinfo;
1877}
1878
1879static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1880	__releases(xfrm_state_afinfo_lock)
1881{
1882	read_unlock(&xfrm_state_afinfo_lock);
1883}
1884
1885/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1886void xfrm_state_delete_tunnel(struct xfrm_state *x)
1887{
1888	if (x->tunnel) {
1889		struct xfrm_state *t = x->tunnel;
1890
1891		if (atomic_read(&t->tunnel_users) == 2)
1892			xfrm_state_delete(t);
1893		atomic_dec(&t->tunnel_users);
1894		xfrm_state_put(t);
1895		x->tunnel = NULL;
1896	}
1897}
1898EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1899
1900int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1901{
1902	int res;
1903
1904	spin_lock_bh(&x->lock);
1905	if (x->km.state == XFRM_STATE_VALID &&
1906	    x->type && x->type->get_mtu)
1907		res = x->type->get_mtu(x, mtu);
1908	else
1909		res = mtu - x->props.header_len;
1910	spin_unlock_bh(&x->lock);
1911	return res;
1912}
1913
1914int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1915{
1916	struct xfrm_state_afinfo *afinfo;
1917	struct xfrm_mode *inner_mode;
1918	int family = x->props.family;
1919	int err;
1920
1921	err = -EAFNOSUPPORT;
1922	afinfo = xfrm_state_get_afinfo(family);
1923	if (!afinfo)
1924		goto error;
1925
1926	err = 0;
1927	if (afinfo->init_flags)
1928		err = afinfo->init_flags(x);
1929
1930	xfrm_state_put_afinfo(afinfo);
1931
1932	if (err)
1933		goto error;
1934
1935	err = -EPROTONOSUPPORT;
1936
1937	if (x->sel.family != AF_UNSPEC) {
1938		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1939		if (inner_mode == NULL)
1940			goto error;
1941
1942		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1943		    family != x->sel.family) {
1944			xfrm_put_mode(inner_mode);
1945			goto error;
1946		}
1947
1948		x->inner_mode = inner_mode;
1949	} else {
1950		struct xfrm_mode *inner_mode_iaf;
1951		int iafamily = AF_INET;
1952
1953		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1954		if (inner_mode == NULL)
1955			goto error;
1956
1957		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1958			xfrm_put_mode(inner_mode);
1959			goto error;
1960		}
1961		x->inner_mode = inner_mode;
1962
1963		if (x->props.family == AF_INET)
1964			iafamily = AF_INET6;
1965
1966		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1967		if (inner_mode_iaf) {
1968			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1969				x->inner_mode_iaf = inner_mode_iaf;
1970			else
1971				xfrm_put_mode(inner_mode_iaf);
1972		}
1973	}
1974
1975	x->type = xfrm_get_type(x->id.proto, family);
1976	if (x->type == NULL)
1977		goto error;
1978
1979	err = x->type->init_state(x);
1980	if (err)
1981		goto error;
1982
1983	x->outer_mode = xfrm_get_mode(x->props.mode, family);
1984	if (x->outer_mode == NULL)
1985		goto error;
1986
1987	if (init_replay) {
1988		err = xfrm_init_replay(x);
1989		if (err)
1990			goto error;
1991	}
1992
1993	x->km.state = XFRM_STATE_VALID;
1994
1995error:
1996	return err;
1997}
1998
1999EXPORT_SYMBOL(__xfrm_init_state);
2000
2001int xfrm_init_state(struct xfrm_state *x)
2002{
2003	return __xfrm_init_state(x, true);
2004}
2005
2006EXPORT_SYMBOL(xfrm_init_state);
2007
2008int __net_init xfrm_state_init(struct net *net)
2009{
2010	unsigned int sz;
2011
2012	INIT_LIST_HEAD(&net->xfrm.state_all);
2013
2014	sz = sizeof(struct hlist_head) * 8;
2015
2016	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2017	if (!net->xfrm.state_bydst)
2018		goto out_bydst;
2019	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2020	if (!net->xfrm.state_bysrc)
2021		goto out_bysrc;
2022	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2023	if (!net->xfrm.state_byspi)
2024		goto out_byspi;
2025	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2026
2027	net->xfrm.state_num = 0;
2028	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2029	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2030	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2031	init_waitqueue_head(&net->xfrm.km_waitq);
2032	return 0;
2033
2034out_byspi:
2035	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2036out_bysrc:
2037	xfrm_hash_free(net->xfrm.state_bydst, sz);
2038out_bydst:
2039	return -ENOMEM;
2040}
2041
2042void xfrm_state_fini(struct net *net)
2043{
2044	struct xfrm_audit audit_info;
2045	unsigned int sz;
2046
2047	flush_work(&net->xfrm.state_hash_work);
2048	audit_info.loginuid = -1;
2049	audit_info.sessionid = -1;
2050	audit_info.secid = 0;
2051	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2052	flush_work(&net->xfrm.state_gc_work);
2053
2054	WARN_ON(!list_empty(&net->xfrm.state_all));
2055
2056	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2057	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2058	xfrm_hash_free(net->xfrm.state_byspi, sz);
2059	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2060	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2061	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2062	xfrm_hash_free(net->xfrm.state_bydst, sz);
2063}
2064
2065#ifdef CONFIG_AUDITSYSCALL
2066static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2067				     struct audit_buffer *audit_buf)
2068{
2069	struct xfrm_sec_ctx *ctx = x->security;
2070	u32 spi = ntohl(x->id.spi);
2071
2072	if (ctx)
2073		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2074				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2075
2076	switch(x->props.family) {
2077	case AF_INET:
2078		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2079				 &x->props.saddr.a4, &x->id.daddr.a4);
2080		break;
2081	case AF_INET6:
2082		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2083				 x->props.saddr.a6, x->id.daddr.a6);
2084		break;
2085	}
2086
2087	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2088}
2089
2090static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2091				      struct audit_buffer *audit_buf)
2092{
2093	const struct iphdr *iph4;
2094	const struct ipv6hdr *iph6;
2095
2096	switch (family) {
2097	case AF_INET:
2098		iph4 = ip_hdr(skb);
2099		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2100				 &iph4->saddr, &iph4->daddr);
2101		break;
2102	case AF_INET6:
2103		iph6 = ipv6_hdr(skb);
2104		audit_log_format(audit_buf,
2105				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2106				 &iph6->saddr,&iph6->daddr,
2107				 iph6->flow_lbl[0] & 0x0f,
2108				 iph6->flow_lbl[1],
2109				 iph6->flow_lbl[2]);
2110		break;
2111	}
2112}
2113
2114void xfrm_audit_state_add(struct xfrm_state *x, int result,
2115			  uid_t auid, u32 sessionid, u32 secid)
2116{
2117	struct audit_buffer *audit_buf;
2118
2119	audit_buf = xfrm_audit_start("SAD-add");
2120	if (audit_buf == NULL)
2121		return;
2122	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2123	xfrm_audit_helper_sainfo(x, audit_buf);
2124	audit_log_format(audit_buf, " res=%u", result);
2125	audit_log_end(audit_buf);
2126}
2127EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2128
2129void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2130			     uid_t auid, u32 sessionid, u32 secid)
2131{
2132	struct audit_buffer *audit_buf;
2133
2134	audit_buf = xfrm_audit_start("SAD-delete");
2135	if (audit_buf == NULL)
2136		return;
2137	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2138	xfrm_audit_helper_sainfo(x, audit_buf);
2139	audit_log_format(audit_buf, " res=%u", result);
2140	audit_log_end(audit_buf);
2141}
2142EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2143
2144void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2145				      struct sk_buff *skb)
2146{
2147	struct audit_buffer *audit_buf;
2148	u32 spi;
2149
2150	audit_buf = xfrm_audit_start("SA-replay-overflow");
2151	if (audit_buf == NULL)
2152		return;
2153	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2154	/* don't record the sequence number because it's inherent in this kind
2155	 * of audit message */
2156	spi = ntohl(x->id.spi);
2157	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158	audit_log_end(audit_buf);
2159}
2160EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2161
2162void xfrm_audit_state_replay(struct xfrm_state *x,
2163			     struct sk_buff *skb, __be32 net_seq)
2164{
2165	struct audit_buffer *audit_buf;
2166	u32 spi;
2167
2168	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2169	if (audit_buf == NULL)
2170		return;
2171	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2172	spi = ntohl(x->id.spi);
2173	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2174			 spi, spi, ntohl(net_seq));
2175	audit_log_end(audit_buf);
2176}
2177EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2178
2179void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2180{
2181	struct audit_buffer *audit_buf;
2182
2183	audit_buf = xfrm_audit_start("SA-notfound");
2184	if (audit_buf == NULL)
2185		return;
2186	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2187	audit_log_end(audit_buf);
2188}
2189EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2190
2191void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2192			       __be32 net_spi, __be32 net_seq)
2193{
2194	struct audit_buffer *audit_buf;
2195	u32 spi;
2196
2197	audit_buf = xfrm_audit_start("SA-notfound");
2198	if (audit_buf == NULL)
2199		return;
2200	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2201	spi = ntohl(net_spi);
2202	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2203			 spi, spi, ntohl(net_seq));
2204	audit_log_end(audit_buf);
2205}
2206EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2207
2208void xfrm_audit_state_icvfail(struct xfrm_state *x,
2209			      struct sk_buff *skb, u8 proto)
2210{
2211	struct audit_buffer *audit_buf;
2212	__be32 net_spi;
2213	__be32 net_seq;
2214
2215	audit_buf = xfrm_audit_start("SA-icv-failure");
2216	if (audit_buf == NULL)
2217		return;
2218	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2219	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2220		u32 spi = ntohl(net_spi);
2221		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2222				 spi, spi, ntohl(net_seq));
2223	}
2224	audit_log_end(audit_buf);
2225}
2226EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2227#endif /* CONFIG_AUDITSYSCALL */