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 <linux/uaccess.h>
  24#include <linux/ktime.h>
  25#include <linux/slab.h>
  26#include <linux/interrupt.h>
  27#include <linux/kernel.h>
  28
  29#include "xfrm_hash.h"
  30
  31#define xfrm_state_deref_prot(table, net) \
  32	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
  33
  34static void xfrm_state_gc_task(struct work_struct *work);
  35
  36/* Each xfrm_state may be linked to two tables:
  37
  38   1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  39   2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  40      destination/tunnel endpoint. (output)
  41 */
  42
  43static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  44static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
  45
  46static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
  47static HLIST_HEAD(xfrm_state_gc_list);
  48
  49static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
  50{
  51	return refcount_inc_not_zero(&x->refcnt);
  52}
  53
  54static inline unsigned int xfrm_dst_hash(struct net *net,
  55					 const xfrm_address_t *daddr,
  56					 const xfrm_address_t *saddr,
  57					 u32 reqid,
  58					 unsigned short family)
  59{
  60	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  61}
  62
  63static inline unsigned int xfrm_src_hash(struct net *net,
  64					 const xfrm_address_t *daddr,
  65					 const xfrm_address_t *saddr,
  66					 unsigned short family)
  67{
  68	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  69}
  70
  71static inline unsigned int
  72xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  73	      __be32 spi, u8 proto, unsigned short family)
  74{
  75	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  76}
  77
  78static void xfrm_hash_transfer(struct hlist_head *list,
  79			       struct hlist_head *ndsttable,
  80			       struct hlist_head *nsrctable,
  81			       struct hlist_head *nspitable,
  82			       unsigned int nhashmask)
  83{
  84	struct hlist_node *tmp;
  85	struct xfrm_state *x;
  86
  87	hlist_for_each_entry_safe(x, tmp, list, bydst) {
  88		unsigned int h;
  89
  90		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  91				    x->props.reqid, x->props.family,
  92				    nhashmask);
  93		hlist_add_head_rcu(&x->bydst, ndsttable + h);
  94
  95		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  96				    x->props.family,
  97				    nhashmask);
  98		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
  99
 100		if (x->id.spi) {
 101			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
 102					    x->id.proto, x->props.family,
 103					    nhashmask);
 104			hlist_add_head_rcu(&x->byspi, nspitable + h);
 105		}
 106	}
 107}
 108
 109static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 110{
 111	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 112}
 113
 114static void xfrm_hash_resize(struct work_struct *work)
 115{
 116	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 117	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 118	unsigned long nsize, osize;
 119	unsigned int nhashmask, ohashmask;
 120	int i;
 121
 122	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 123	ndst = xfrm_hash_alloc(nsize);
 124	if (!ndst)
 125		return;
 126	nsrc = xfrm_hash_alloc(nsize);
 127	if (!nsrc) {
 128		xfrm_hash_free(ndst, nsize);
 129		return;
 130	}
 131	nspi = xfrm_hash_alloc(nsize);
 132	if (!nspi) {
 133		xfrm_hash_free(ndst, nsize);
 134		xfrm_hash_free(nsrc, nsize);
 135		return;
 136	}
 137
 138	spin_lock_bh(&net->xfrm.xfrm_state_lock);
 139	write_seqcount_begin(&xfrm_state_hash_generation);
 140
 141	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 142	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
 143	for (i = net->xfrm.state_hmask; i >= 0; i--)
 144		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
 145
 146	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
 147	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
 148	ohashmask = net->xfrm.state_hmask;
 149
 150	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
 151	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
 152	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
 153	net->xfrm.state_hmask = nhashmask;
 154
 155	write_seqcount_end(&xfrm_state_hash_generation);
 156	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 157
 158	osize = (ohashmask + 1) * sizeof(struct hlist_head);
 159
 160	synchronize_rcu();
 161
 162	xfrm_hash_free(odst, osize);
 163	xfrm_hash_free(osrc, osize);
 164	xfrm_hash_free(ospi, osize);
 165}
 166
 167static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 168static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
 169
 170static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 171
 172int __xfrm_state_delete(struct xfrm_state *x);
 173
 174int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 175bool km_is_alive(const struct km_event *c);
 176void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 177
 178static DEFINE_SPINLOCK(xfrm_type_lock);
 179int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 180{
 181	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 182	const struct xfrm_type **typemap;
 183	int err = 0;
 184
 185	if (unlikely(afinfo == NULL))
 186		return -EAFNOSUPPORT;
 187	typemap = afinfo->type_map;
 188	spin_lock_bh(&xfrm_type_lock);
 189
 190	if (likely(typemap[type->proto] == NULL))
 191		typemap[type->proto] = type;
 192	else
 193		err = -EEXIST;
 194	spin_unlock_bh(&xfrm_type_lock);
 195	rcu_read_unlock();
 196	return err;
 197}
 198EXPORT_SYMBOL(xfrm_register_type);
 199
 200int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 201{
 202	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 203	const struct xfrm_type **typemap;
 204	int err = 0;
 205
 206	if (unlikely(afinfo == NULL))
 207		return -EAFNOSUPPORT;
 208	typemap = afinfo->type_map;
 209	spin_lock_bh(&xfrm_type_lock);
 210
 211	if (unlikely(typemap[type->proto] != type))
 212		err = -ENOENT;
 213	else
 214		typemap[type->proto] = NULL;
 215	spin_unlock_bh(&xfrm_type_lock);
 216	rcu_read_unlock();
 217	return err;
 218}
 219EXPORT_SYMBOL(xfrm_unregister_type);
 220
 221static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 222{
 223	struct xfrm_state_afinfo *afinfo;
 224	const struct xfrm_type **typemap;
 225	const struct xfrm_type *type;
 226	int modload_attempted = 0;
 227
 228retry:
 229	afinfo = xfrm_state_get_afinfo(family);
 230	if (unlikely(afinfo == NULL))
 231		return NULL;
 232	typemap = afinfo->type_map;
 233
 234	type = READ_ONCE(typemap[proto]);
 235	if (unlikely(type && !try_module_get(type->owner)))
 236		type = NULL;
 237
 238	rcu_read_unlock();
 239
 240	if (!type && !modload_attempted) {
 241		request_module("xfrm-type-%d-%d", family, proto);
 242		modload_attempted = 1;
 243		goto retry;
 244	}
 245
 246	return type;
 247}
 248
 249static void xfrm_put_type(const struct xfrm_type *type)
 250{
 251	module_put(type->owner);
 252}
 253
 254static DEFINE_SPINLOCK(xfrm_type_offload_lock);
 255int xfrm_register_type_offload(const struct xfrm_type_offload *type,
 256			       unsigned short family)
 257{
 258	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 259	const struct xfrm_type_offload **typemap;
 260	int err = 0;
 261
 262	if (unlikely(afinfo == NULL))
 263		return -EAFNOSUPPORT;
 264	typemap = afinfo->type_offload_map;
 265	spin_lock_bh(&xfrm_type_offload_lock);
 266
 267	if (likely(typemap[type->proto] == NULL))
 268		typemap[type->proto] = type;
 269	else
 270		err = -EEXIST;
 271	spin_unlock_bh(&xfrm_type_offload_lock);
 272	rcu_read_unlock();
 273	return err;
 274}
 275EXPORT_SYMBOL(xfrm_register_type_offload);
 276
 277int xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
 278				 unsigned short family)
 279{
 280	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 281	const struct xfrm_type_offload **typemap;
 282	int err = 0;
 283
 284	if (unlikely(afinfo == NULL))
 285		return -EAFNOSUPPORT;
 286	typemap = afinfo->type_offload_map;
 287	spin_lock_bh(&xfrm_type_offload_lock);
 288
 289	if (unlikely(typemap[type->proto] != type))
 290		err = -ENOENT;
 291	else
 292		typemap[type->proto] = NULL;
 293	spin_unlock_bh(&xfrm_type_offload_lock);
 294	rcu_read_unlock();
 295	return err;
 296}
 297EXPORT_SYMBOL(xfrm_unregister_type_offload);
 298
 299static const struct xfrm_type_offload *
 300xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
 301{
 302	struct xfrm_state_afinfo *afinfo;
 303	const struct xfrm_type_offload **typemap;
 304	const struct xfrm_type_offload *type;
 305
 306retry:
 307	afinfo = xfrm_state_get_afinfo(family);
 308	if (unlikely(afinfo == NULL))
 309		return NULL;
 310	typemap = afinfo->type_offload_map;
 311
 312	type = typemap[proto];
 313	if ((type && !try_module_get(type->owner)))
 314		type = NULL;
 315
 316	rcu_read_unlock();
 317
 318	if (!type && try_load) {
 319		request_module("xfrm-offload-%d-%d", family, proto);
 320		try_load = false;
 321		goto retry;
 322	}
 323
 324	return type;
 325}
 326
 327static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
 328{
 329	module_put(type->owner);
 330}
 331
 332static DEFINE_SPINLOCK(xfrm_mode_lock);
 333int xfrm_register_mode(struct xfrm_mode *mode, int family)
 334{
 335	struct xfrm_state_afinfo *afinfo;
 336	struct xfrm_mode **modemap;
 337	int err;
 338
 339	if (unlikely(mode->encap >= XFRM_MODE_MAX))
 340		return -EINVAL;
 341
 342	afinfo = xfrm_state_get_afinfo(family);
 343	if (unlikely(afinfo == NULL))
 344		return -EAFNOSUPPORT;
 345
 346	err = -EEXIST;
 347	modemap = afinfo->mode_map;
 348	spin_lock_bh(&xfrm_mode_lock);
 349	if (modemap[mode->encap])
 350		goto out;
 351
 352	err = -ENOENT;
 353	if (!try_module_get(afinfo->owner))
 354		goto out;
 355
 356	mode->afinfo = afinfo;
 357	modemap[mode->encap] = mode;
 358	err = 0;
 359
 360out:
 361	spin_unlock_bh(&xfrm_mode_lock);
 362	rcu_read_unlock();
 363	return err;
 364}
 365EXPORT_SYMBOL(xfrm_register_mode);
 366
 367int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 368{
 369	struct xfrm_state_afinfo *afinfo;
 370	struct xfrm_mode **modemap;
 371	int err;
 372
 373	if (unlikely(mode->encap >= XFRM_MODE_MAX))
 374		return -EINVAL;
 375
 376	afinfo = xfrm_state_get_afinfo(family);
 377	if (unlikely(afinfo == NULL))
 378		return -EAFNOSUPPORT;
 379
 380	err = -ENOENT;
 381	modemap = afinfo->mode_map;
 382	spin_lock_bh(&xfrm_mode_lock);
 383	if (likely(modemap[mode->encap] == mode)) {
 384		modemap[mode->encap] = NULL;
 385		module_put(mode->afinfo->owner);
 386		err = 0;
 387	}
 388
 389	spin_unlock_bh(&xfrm_mode_lock);
 390	rcu_read_unlock();
 391	return err;
 392}
 393EXPORT_SYMBOL(xfrm_unregister_mode);
 394
 395static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 396{
 397	struct xfrm_state_afinfo *afinfo;
 398	struct xfrm_mode *mode;
 399	int modload_attempted = 0;
 400
 401	if (unlikely(encap >= XFRM_MODE_MAX))
 402		return NULL;
 403
 404retry:
 405	afinfo = xfrm_state_get_afinfo(family);
 406	if (unlikely(afinfo == NULL))
 407		return NULL;
 408
 409	mode = READ_ONCE(afinfo->mode_map[encap]);
 410	if (unlikely(mode && !try_module_get(mode->owner)))
 411		mode = NULL;
 412
 413	rcu_read_unlock();
 414	if (!mode && !modload_attempted) {
 415		request_module("xfrm-mode-%d-%d", family, encap);
 416		modload_attempted = 1;
 417		goto retry;
 418	}
 419
 420	return mode;
 421}
 422
 423static void xfrm_put_mode(struct xfrm_mode *mode)
 424{
 425	module_put(mode->owner);
 426}
 427
 428static void xfrm_state_gc_destroy(struct xfrm_state *x)
 429{
 430	tasklet_hrtimer_cancel(&x->mtimer);
 431	del_timer_sync(&x->rtimer);
 432	kfree(x->aead);
 433	kfree(x->aalg);
 434	kfree(x->ealg);
 435	kfree(x->calg);
 436	kfree(x->encap);
 437	kfree(x->coaddr);
 438	kfree(x->replay_esn);
 439	kfree(x->preplay_esn);
 440	if (x->inner_mode)
 441		xfrm_put_mode(x->inner_mode);
 442	if (x->inner_mode_iaf)
 443		xfrm_put_mode(x->inner_mode_iaf);
 444	if (x->outer_mode)
 445		xfrm_put_mode(x->outer_mode);
 446	if (x->type_offload)
 447		xfrm_put_type_offload(x->type_offload);
 448	if (x->type) {
 449		x->type->destructor(x);
 450		xfrm_put_type(x->type);
 451	}
 452	xfrm_dev_state_free(x);
 453	security_xfrm_state_free(x);
 454	kfree(x);
 455}
 456
 457static void xfrm_state_gc_task(struct work_struct *work)
 458{
 459	struct xfrm_state *x;
 460	struct hlist_node *tmp;
 461	struct hlist_head gc_list;
 462
 463	spin_lock_bh(&xfrm_state_gc_lock);
 464	hlist_move_list(&xfrm_state_gc_list, &gc_list);
 465	spin_unlock_bh(&xfrm_state_gc_lock);
 466
 467	synchronize_rcu();
 468
 469	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 470		xfrm_state_gc_destroy(x);
 471}
 472
 473static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
 474{
 475	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 476	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 477	unsigned long now = get_seconds();
 478	long next = LONG_MAX;
 479	int warn = 0;
 480	int err = 0;
 481
 482	spin_lock(&x->lock);
 483	if (x->km.state == XFRM_STATE_DEAD)
 484		goto out;
 485	if (x->km.state == XFRM_STATE_EXPIRED)
 486		goto expired;
 487	if (x->lft.hard_add_expires_seconds) {
 488		long tmo = x->lft.hard_add_expires_seconds +
 489			x->curlft.add_time - now;
 490		if (tmo <= 0) {
 491			if (x->xflags & XFRM_SOFT_EXPIRE) {
 492				/* enter hard expire without soft expire first?!
 493				 * setting a new date could trigger this.
 494				 * workaround: fix x->curflt.add_time by below:
 495				 */
 496				x->curlft.add_time = now - x->saved_tmo - 1;
 497				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 498			} else
 499				goto expired;
 500		}
 501		if (tmo < next)
 502			next = tmo;
 503	}
 504	if (x->lft.hard_use_expires_seconds) {
 505		long tmo = x->lft.hard_use_expires_seconds +
 506			(x->curlft.use_time ? : now) - now;
 507		if (tmo <= 0)
 508			goto expired;
 509		if (tmo < next)
 510			next = tmo;
 511	}
 512	if (x->km.dying)
 513		goto resched;
 514	if (x->lft.soft_add_expires_seconds) {
 515		long tmo = x->lft.soft_add_expires_seconds +
 516			x->curlft.add_time - now;
 517		if (tmo <= 0) {
 518			warn = 1;
 519			x->xflags &= ~XFRM_SOFT_EXPIRE;
 520		} else if (tmo < next) {
 521			next = tmo;
 522			x->xflags |= XFRM_SOFT_EXPIRE;
 523			x->saved_tmo = tmo;
 524		}
 525	}
 526	if (x->lft.soft_use_expires_seconds) {
 527		long tmo = x->lft.soft_use_expires_seconds +
 528			(x->curlft.use_time ? : now) - now;
 529		if (tmo <= 0)
 530			warn = 1;
 531		else if (tmo < next)
 532			next = tmo;
 533	}
 534
 535	x->km.dying = warn;
 536	if (warn)
 537		km_state_expired(x, 0, 0);
 538resched:
 539	if (next != LONG_MAX) {
 540		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 541	}
 542
 543	goto out;
 544
 545expired:
 546	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
 547		x->km.state = XFRM_STATE_EXPIRED;
 548
 549	err = __xfrm_state_delete(x);
 550	if (!err)
 551		km_state_expired(x, 1, 0);
 552
 553	xfrm_audit_state_delete(x, err ? 0 : 1, true);
 554
 555out:
 556	spin_unlock(&x->lock);
 557	return HRTIMER_NORESTART;
 558}
 559
 560static void xfrm_replay_timer_handler(struct timer_list *t);
 561
 562struct xfrm_state *xfrm_state_alloc(struct net *net)
 563{
 564	struct xfrm_state *x;
 565
 566	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 567
 568	if (x) {
 569		write_pnet(&x->xs_net, net);
 570		refcount_set(&x->refcnt, 1);
 571		atomic_set(&x->tunnel_users, 0);
 572		INIT_LIST_HEAD(&x->km.all);
 573		INIT_HLIST_NODE(&x->bydst);
 574		INIT_HLIST_NODE(&x->bysrc);
 575		INIT_HLIST_NODE(&x->byspi);
 576		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
 577					CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
 578		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
 579		x->curlft.add_time = get_seconds();
 580		x->lft.soft_byte_limit = XFRM_INF;
 581		x->lft.soft_packet_limit = XFRM_INF;
 582		x->lft.hard_byte_limit = XFRM_INF;
 583		x->lft.hard_packet_limit = XFRM_INF;
 584		x->replay_maxage = 0;
 585		x->replay_maxdiff = 0;
 586		x->inner_mode = NULL;
 587		x->inner_mode_iaf = NULL;
 588		spin_lock_init(&x->lock);
 589	}
 590	return x;
 591}
 592EXPORT_SYMBOL(xfrm_state_alloc);
 593
 594void __xfrm_state_destroy(struct xfrm_state *x)
 595{
 596	WARN_ON(x->km.state != XFRM_STATE_DEAD);
 597
 598	spin_lock_bh(&xfrm_state_gc_lock);
 599	hlist_add_head(&x->gclist, &xfrm_state_gc_list);
 600	spin_unlock_bh(&xfrm_state_gc_lock);
 601	schedule_work(&xfrm_state_gc_work);
 602}
 603EXPORT_SYMBOL(__xfrm_state_destroy);
 604
 605int __xfrm_state_delete(struct xfrm_state *x)
 606{
 607	struct net *net = xs_net(x);
 608	int err = -ESRCH;
 609
 610	if (x->km.state != XFRM_STATE_DEAD) {
 611		x->km.state = XFRM_STATE_DEAD;
 612		spin_lock(&net->xfrm.xfrm_state_lock);
 613		list_del(&x->km.all);
 614		hlist_del_rcu(&x->bydst);
 615		hlist_del_rcu(&x->bysrc);
 616		if (x->id.spi)
 617			hlist_del_rcu(&x->byspi);
 618		net->xfrm.state_num--;
 619		spin_unlock(&net->xfrm.xfrm_state_lock);
 620
 621		xfrm_dev_state_delete(x);
 622
 623		/* All xfrm_state objects are created by xfrm_state_alloc.
 624		 * The xfrm_state_alloc call gives a reference, and that
 625		 * is what we are dropping here.
 626		 */
 627		xfrm_state_put(x);
 628		err = 0;
 629	}
 630
 631	return err;
 632}
 633EXPORT_SYMBOL(__xfrm_state_delete);
 634
 635int xfrm_state_delete(struct xfrm_state *x)
 636{
 637	int err;
 638
 639	spin_lock_bh(&x->lock);
 640	err = __xfrm_state_delete(x);
 641	spin_unlock_bh(&x->lock);
 642
 643	return err;
 644}
 645EXPORT_SYMBOL(xfrm_state_delete);
 646
 647#ifdef CONFIG_SECURITY_NETWORK_XFRM
 648static inline int
 649xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 650{
 651	int i, err = 0;
 652
 653	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 654		struct xfrm_state *x;
 655
 656		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 657			if (xfrm_id_proto_match(x->id.proto, proto) &&
 658			   (err = security_xfrm_state_delete(x)) != 0) {
 659				xfrm_audit_state_delete(x, 0, task_valid);
 660				return err;
 661			}
 662		}
 663	}
 664
 665	return err;
 666}
 667
 668static inline int
 669xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
 670{
 671	int i, err = 0;
 672
 673	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 674		struct xfrm_state *x;
 675		struct xfrm_state_offload *xso;
 676
 677		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 678			xso = &x->xso;
 679
 680			if (xso->dev == dev &&
 681			   (err = security_xfrm_state_delete(x)) != 0) {
 682				xfrm_audit_state_delete(x, 0, task_valid);
 683				return err;
 684			}
 685		}
 686	}
 687
 688	return err;
 689}
 690#else
 691static inline int
 692xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 693{
 694	return 0;
 695}
 696
 697static inline int
 698xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
 699{
 700	return 0;
 701}
 702#endif
 703
 704int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
 705{
 706	int i, err = 0, cnt = 0;
 707
 708	spin_lock_bh(&net->xfrm.xfrm_state_lock);
 709	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
 710	if (err)
 711		goto out;
 712
 713	err = -ESRCH;
 714	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 715		struct xfrm_state *x;
 716restart:
 717		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 718			if (!xfrm_state_kern(x) &&
 719			    xfrm_id_proto_match(x->id.proto, proto)) {
 720				xfrm_state_hold(x);
 721				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 722
 723				err = xfrm_state_delete(x);
 724				xfrm_audit_state_delete(x, err ? 0 : 1,
 725							task_valid);
 726				xfrm_state_put(x);
 727				if (!err)
 728					cnt++;
 729
 730				spin_lock_bh(&net->xfrm.xfrm_state_lock);
 731				goto restart;
 732			}
 733		}
 734	}
 735out:
 736	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 737	if (cnt) {
 738		err = 0;
 739		xfrm_policy_cache_flush();
 740	}
 741	return err;
 742}
 743EXPORT_SYMBOL(xfrm_state_flush);
 744
 745int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
 746{
 747	int i, err = 0, cnt = 0;
 748
 749	spin_lock_bh(&net->xfrm.xfrm_state_lock);
 750	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
 751	if (err)
 752		goto out;
 753
 754	err = -ESRCH;
 755	for (i = 0; i <= net->xfrm.state_hmask; i++) {
 756		struct xfrm_state *x;
 757		struct xfrm_state_offload *xso;
 758restart:
 759		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 760			xso = &x->xso;
 761
 762			if (!xfrm_state_kern(x) && xso->dev == dev) {
 763				xfrm_state_hold(x);
 764				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 765
 766				err = xfrm_state_delete(x);
 767				xfrm_audit_state_delete(x, err ? 0 : 1,
 768							task_valid);
 769				xfrm_state_put(x);
 770				if (!err)
 771					cnt++;
 772
 773				spin_lock_bh(&net->xfrm.xfrm_state_lock);
 774				goto restart;
 775			}
 776		}
 777	}
 778	if (cnt)
 779		err = 0;
 780
 781out:
 782	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 783	return err;
 784}
 785EXPORT_SYMBOL(xfrm_dev_state_flush);
 786
 787void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 788{
 789	spin_lock_bh(&net->xfrm.xfrm_state_lock);
 790	si->sadcnt = net->xfrm.state_num;
 791	si->sadhcnt = net->xfrm.state_hmask;
 792	si->sadhmcnt = xfrm_state_hashmax;
 793	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 794}
 795EXPORT_SYMBOL(xfrm_sad_getinfo);
 796
 797static void
 798xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 799		    const struct xfrm_tmpl *tmpl,
 800		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 801		    unsigned short family)
 802{
 803	struct xfrm_state_afinfo *afinfo = xfrm_state_afinfo_get_rcu(family);
 804
 805	if (!afinfo)
 806		return;
 807
 808	afinfo->init_tempsel(&x->sel, fl);
 809
 810	if (family != tmpl->encap_family) {
 811		afinfo = xfrm_state_afinfo_get_rcu(tmpl->encap_family);
 812		if (!afinfo)
 813			return;
 814	}
 815	afinfo->init_temprop(x, tmpl, daddr, saddr);
 816}
 817
 818static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 819					      const xfrm_address_t *daddr,
 820					      __be32 spi, u8 proto,
 821					      unsigned short family)
 822{
 823	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 824	struct xfrm_state *x;
 825
 826	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
 827		if (x->props.family != family ||
 828		    x->id.spi       != spi ||
 829		    x->id.proto     != proto ||
 830		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
 831			continue;
 832
 833		if ((mark & x->mark.m) != x->mark.v)
 834			continue;
 835		if (!xfrm_state_hold_rcu(x))
 836			continue;
 837		return x;
 838	}
 839
 840	return NULL;
 841}
 842
 843static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 844						     const xfrm_address_t *daddr,
 845						     const xfrm_address_t *saddr,
 846						     u8 proto, unsigned short family)
 847{
 848	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 849	struct xfrm_state *x;
 850
 851	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
 852		if (x->props.family != family ||
 853		    x->id.proto     != proto ||
 854		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 855		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
 856			continue;
 857
 858		if ((mark & x->mark.m) != x->mark.v)
 859			continue;
 860		if (!xfrm_state_hold_rcu(x))
 861			continue;
 862		return x;
 863	}
 864
 865	return NULL;
 866}
 867
 868static inline struct xfrm_state *
 869__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 870{
 871	struct net *net = xs_net(x);
 872	u32 mark = x->mark.v & x->mark.m;
 873
 874	if (use_spi)
 875		return __xfrm_state_lookup(net, mark, &x->id.daddr,
 876					   x->id.spi, x->id.proto, family);
 877	else
 878		return __xfrm_state_lookup_byaddr(net, mark,
 879						  &x->id.daddr,
 880						  &x->props.saddr,
 881						  x->id.proto, family);
 882}
 883
 884static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 885{
 886	if (have_hash_collision &&
 887	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 888	    net->xfrm.state_num > net->xfrm.state_hmask)
 889		schedule_work(&net->xfrm.state_hash_work);
 890}
 891
 892static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 893			       const struct flowi *fl, unsigned short family,
 894			       struct xfrm_state **best, int *acq_in_progress,
 895			       int *error)
 896{
 897	/* Resolution logic:
 898	 * 1. There is a valid state with matching selector. Done.
 899	 * 2. Valid state with inappropriate selector. Skip.
 900	 *
 901	 * Entering area of "sysdeps".
 902	 *
 903	 * 3. If state is not valid, selector is temporary, it selects
 904	 *    only session which triggered previous resolution. Key
 905	 *    manager will do something to install a state with proper
 906	 *    selector.
 907	 */
 908	if (x->km.state == XFRM_STATE_VALID) {
 909		if ((x->sel.family &&
 910		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 911		    !security_xfrm_state_pol_flow_match(x, pol, fl))
 912			return;
 913
 914		if (!*best ||
 915		    (*best)->km.dying > x->km.dying ||
 916		    ((*best)->km.dying == x->km.dying &&
 917		     (*best)->curlft.add_time < x->curlft.add_time))
 918			*best = x;
 919	} else if (x->km.state == XFRM_STATE_ACQ) {
 920		*acq_in_progress = 1;
 921	} else if (x->km.state == XFRM_STATE_ERROR ||
 922		   x->km.state == XFRM_STATE_EXPIRED) {
 923		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 924		    security_xfrm_state_pol_flow_match(x, pol, fl))
 925			*error = -ESRCH;
 926	}
 927}
 928
 929struct xfrm_state *
 930xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 931		const struct flowi *fl, struct xfrm_tmpl *tmpl,
 932		struct xfrm_policy *pol, int *err,
 933		unsigned short family)
 934{
 935	static xfrm_address_t saddr_wildcard = { };
 936	struct net *net = xp_net(pol);
 937	unsigned int h, h_wildcard;
 938	struct xfrm_state *x, *x0, *to_put;
 939	int acquire_in_progress = 0;
 940	int error = 0;
 941	struct xfrm_state *best = NULL;
 942	u32 mark = pol->mark.v & pol->mark.m;
 943	unsigned short encap_family = tmpl->encap_family;
 944	unsigned int sequence;
 945	struct km_event c;
 946
 947	to_put = NULL;
 948
 949	sequence = read_seqcount_begin(&xfrm_state_hash_generation);
 950
 951	rcu_read_lock();
 952	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 953	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
 954		if (x->props.family == encap_family &&
 955		    x->props.reqid == tmpl->reqid &&
 956		    (mark & x->mark.m) == x->mark.v &&
 957		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
 958		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 959		    tmpl->mode == x->props.mode &&
 960		    tmpl->id.proto == x->id.proto &&
 961		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 962			xfrm_state_look_at(pol, x, fl, encap_family,
 963					   &best, &acquire_in_progress, &error);
 964	}
 965	if (best || acquire_in_progress)
 966		goto found;
 967
 968	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 969	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
 970		if (x->props.family == encap_family &&
 971		    x->props.reqid == tmpl->reqid &&
 972		    (mark & x->mark.m) == x->mark.v &&
 973		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
 974		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
 975		    tmpl->mode == x->props.mode &&
 976		    tmpl->id.proto == x->id.proto &&
 977		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 978			xfrm_state_look_at(pol, x, fl, encap_family,
 979					   &best, &acquire_in_progress, &error);
 980	}
 981
 982found:
 983	x = best;
 984	if (!x && !error && !acquire_in_progress) {
 985		if (tmpl->id.spi &&
 986		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 987					      tmpl->id.proto, encap_family)) != NULL) {
 988			to_put = x0;
 989			error = -EEXIST;
 990			goto out;
 991		}
 992
 993		c.net = net;
 994		/* If the KMs have no listeners (yet...), avoid allocating an SA
 995		 * for each and every packet - garbage collection might not
 996		 * handle the flood.
 997		 */
 998		if (!km_is_alive(&c)) {
 999			error = -ESRCH;
1000			goto out;
1001		}
1002
1003		x = xfrm_state_alloc(net);
1004		if (x == NULL) {
1005			error = -ENOMEM;
1006			goto out;
1007		}
1008		/* Initialize temporary state matching only
1009		 * to current session. */
1010		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1011		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1012
1013		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1014		if (error) {
1015			x->km.state = XFRM_STATE_DEAD;
1016			to_put = x;
1017			x = NULL;
1018			goto out;
1019		}
1020
1021		if (km_query(x, tmpl, pol) == 0) {
1022			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1023			x->km.state = XFRM_STATE_ACQ;
1024			list_add(&x->km.all, &net->xfrm.state_all);
1025			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1026			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1027			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1028			if (x->id.spi) {
1029				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1030				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1031			}
1032			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1033			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1034			net->xfrm.state_num++;
1035			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1036			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1037		} else {
1038			x->km.state = XFRM_STATE_DEAD;
1039			to_put = x;
1040			x = NULL;
1041			error = -ESRCH;
1042		}
1043	}
1044out:
1045	if (x) {
1046		if (!xfrm_state_hold_rcu(x)) {
1047			*err = -EAGAIN;
1048			x = NULL;
1049		}
1050	} else {
1051		*err = acquire_in_progress ? -EAGAIN : error;
1052	}
1053	rcu_read_unlock();
1054	if (to_put)
1055		xfrm_state_put(to_put);
1056
1057	if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1058		*err = -EAGAIN;
1059		if (x) {
1060			xfrm_state_put(x);
1061			x = NULL;
1062		}
1063	}
1064
1065	return x;
1066}
1067
1068struct xfrm_state *
1069xfrm_stateonly_find(struct net *net, u32 mark,
1070		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1071		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1072{
1073	unsigned int h;
1074	struct xfrm_state *rx = NULL, *x = NULL;
1075
1076	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1077	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1078	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1079		if (x->props.family == family &&
1080		    x->props.reqid == reqid &&
1081		    (mark & x->mark.m) == x->mark.v &&
1082		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1083		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1084		    mode == x->props.mode &&
1085		    proto == x->id.proto &&
1086		    x->km.state == XFRM_STATE_VALID) {
1087			rx = x;
1088			break;
1089		}
1090	}
1091
1092	if (rx)
1093		xfrm_state_hold(rx);
1094	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1095
1096
1097	return rx;
1098}
1099EXPORT_SYMBOL(xfrm_stateonly_find);
1100
1101struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1102					      unsigned short family)
1103{
1104	struct xfrm_state *x;
1105	struct xfrm_state_walk *w;
1106
1107	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1108	list_for_each_entry(w, &net->xfrm.state_all, all) {
1109		x = container_of(w, struct xfrm_state, km);
1110		if (x->props.family != family ||
1111			x->id.spi != spi)
1112			continue;
1113
1114		xfrm_state_hold(x);
1115		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1116		return x;
1117	}
1118	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1119	return NULL;
1120}
1121EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1122
1123static void __xfrm_state_insert(struct xfrm_state *x)
1124{
1125	struct net *net = xs_net(x);
1126	unsigned int h;
1127
1128	list_add(&x->km.all, &net->xfrm.state_all);
1129
1130	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1131			  x->props.reqid, x->props.family);
1132	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1133
1134	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1135	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1136
1137	if (x->id.spi) {
1138		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1139				  x->props.family);
1140
1141		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1142	}
1143
1144	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1145	if (x->replay_maxage)
1146		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1147
1148	net->xfrm.state_num++;
1149
1150	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1151}
1152
1153/* net->xfrm.xfrm_state_lock is held */
1154static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1155{
1156	struct net *net = xs_net(xnew);
1157	unsigned short family = xnew->props.family;
1158	u32 reqid = xnew->props.reqid;
1159	struct xfrm_state *x;
1160	unsigned int h;
1161	u32 mark = xnew->mark.v & xnew->mark.m;
1162
1163	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1164	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1165		if (x->props.family	== family &&
1166		    x->props.reqid	== reqid &&
1167		    (mark & x->mark.m) == x->mark.v &&
1168		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1169		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1170			x->genid++;
1171	}
1172}
1173
1174void xfrm_state_insert(struct xfrm_state *x)
1175{
1176	struct net *net = xs_net(x);
1177
1178	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1179	__xfrm_state_bump_genids(x);
1180	__xfrm_state_insert(x);
1181	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1182}
1183EXPORT_SYMBOL(xfrm_state_insert);
1184
1185/* net->xfrm.xfrm_state_lock is held */
1186static struct xfrm_state *__find_acq_core(struct net *net,
1187					  const struct xfrm_mark *m,
1188					  unsigned short family, u8 mode,
1189					  u32 reqid, u8 proto,
1190					  const xfrm_address_t *daddr,
1191					  const xfrm_address_t *saddr,
1192					  int create)
1193{
1194	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1195	struct xfrm_state *x;
1196	u32 mark = m->v & m->m;
1197
1198	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1199		if (x->props.reqid  != reqid ||
1200		    x->props.mode   != mode ||
1201		    x->props.family != family ||
1202		    x->km.state     != XFRM_STATE_ACQ ||
1203		    x->id.spi       != 0 ||
1204		    x->id.proto	    != proto ||
1205		    (mark & x->mark.m) != x->mark.v ||
1206		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1207		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1208			continue;
1209
1210		xfrm_state_hold(x);
1211		return x;
1212	}
1213
1214	if (!create)
1215		return NULL;
1216
1217	x = xfrm_state_alloc(net);
1218	if (likely(x)) {
1219		switch (family) {
1220		case AF_INET:
1221			x->sel.daddr.a4 = daddr->a4;
1222			x->sel.saddr.a4 = saddr->a4;
1223			x->sel.prefixlen_d = 32;
1224			x->sel.prefixlen_s = 32;
1225			x->props.saddr.a4 = saddr->a4;
1226			x->id.daddr.a4 = daddr->a4;
1227			break;
1228
1229		case AF_INET6:
1230			x->sel.daddr.in6 = daddr->in6;
1231			x->sel.saddr.in6 = saddr->in6;
1232			x->sel.prefixlen_d = 128;
1233			x->sel.prefixlen_s = 128;
1234			x->props.saddr.in6 = saddr->in6;
1235			x->id.daddr.in6 = daddr->in6;
1236			break;
1237		}
1238
1239		x->km.state = XFRM_STATE_ACQ;
1240		x->id.proto = proto;
1241		x->props.family = family;
1242		x->props.mode = mode;
1243		x->props.reqid = reqid;
1244		x->mark.v = m->v;
1245		x->mark.m = m->m;
1246		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1247		xfrm_state_hold(x);
1248		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1249		list_add(&x->km.all, &net->xfrm.state_all);
1250		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1251		h = xfrm_src_hash(net, daddr, saddr, family);
1252		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1253
1254		net->xfrm.state_num++;
1255
1256		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1257	}
1258
1259	return x;
1260}
1261
1262static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1263
1264int xfrm_state_add(struct xfrm_state *x)
1265{
1266	struct net *net = xs_net(x);
1267	struct xfrm_state *x1, *to_put;
1268	int family;
1269	int err;
1270	u32 mark = x->mark.v & x->mark.m;
1271	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1272
1273	family = x->props.family;
1274
1275	to_put = NULL;
1276
1277	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1278
1279	x1 = __xfrm_state_locate(x, use_spi, family);
1280	if (x1) {
1281		to_put = x1;
1282		x1 = NULL;
1283		err = -EEXIST;
1284		goto out;
1285	}
1286
1287	if (use_spi && x->km.seq) {
1288		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1289		if (x1 && ((x1->id.proto != x->id.proto) ||
1290		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1291			to_put = x1;
1292			x1 = NULL;
1293		}
1294	}
1295
1296	if (use_spi && !x1)
1297		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1298				     x->props.reqid, x->id.proto,
1299				     &x->id.daddr, &x->props.saddr, 0);
1300
1301	__xfrm_state_bump_genids(x);
1302	__xfrm_state_insert(x);
1303	err = 0;
1304
1305out:
1306	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1307
1308	if (x1) {
1309		xfrm_state_delete(x1);
1310		xfrm_state_put(x1);
1311	}
1312
1313	if (to_put)
1314		xfrm_state_put(to_put);
1315
1316	return err;
1317}
1318EXPORT_SYMBOL(xfrm_state_add);
1319
1320#ifdef CONFIG_XFRM_MIGRATE
1321static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1322					   struct xfrm_encap_tmpl *encap)
1323{
1324	struct net *net = xs_net(orig);
1325	struct xfrm_state *x = xfrm_state_alloc(net);
1326	if (!x)
1327		goto out;
1328
1329	memcpy(&x->id, &orig->id, sizeof(x->id));
1330	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1331	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1332	x->props.mode = orig->props.mode;
1333	x->props.replay_window = orig->props.replay_window;
1334	x->props.reqid = orig->props.reqid;
1335	x->props.family = orig->props.family;
1336	x->props.saddr = orig->props.saddr;
1337
1338	if (orig->aalg) {
1339		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1340		if (!x->aalg)
1341			goto error;
1342	}
1343	x->props.aalgo = orig->props.aalgo;
1344
1345	if (orig->aead) {
1346		x->aead = xfrm_algo_aead_clone(orig->aead);
1347		x->geniv = orig->geniv;
1348		if (!x->aead)
1349			goto error;
1350	}
1351	if (orig->ealg) {
1352		x->ealg = xfrm_algo_clone(orig->ealg);
1353		if (!x->ealg)
1354			goto error;
1355	}
1356	x->props.ealgo = orig->props.ealgo;
1357
1358	if (orig->calg) {
1359		x->calg = xfrm_algo_clone(orig->calg);
1360		if (!x->calg)
1361			goto error;
1362	}
1363	x->props.calgo = orig->props.calgo;
1364
1365	if (encap || orig->encap) {
1366		if (encap)
1367			x->encap = kmemdup(encap, sizeof(*x->encap),
1368					GFP_KERNEL);
1369		else
1370			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1371					GFP_KERNEL);
1372
1373		if (!x->encap)
1374			goto error;
1375	}
1376
1377	if (orig->coaddr) {
1378		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1379				    GFP_KERNEL);
1380		if (!x->coaddr)
1381			goto error;
1382	}
1383
1384	if (orig->replay_esn) {
1385		if (xfrm_replay_clone(x, orig))
1386			goto error;
1387	}
1388
1389	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1390
1391	if (xfrm_init_state(x) < 0)
1392		goto error;
1393
1394	x->props.flags = orig->props.flags;
1395	x->props.extra_flags = orig->props.extra_flags;
1396
1397	x->tfcpad = orig->tfcpad;
1398	x->replay_maxdiff = orig->replay_maxdiff;
1399	x->replay_maxage = orig->replay_maxage;
1400	x->curlft.add_time = orig->curlft.add_time;
1401	x->km.state = orig->km.state;
1402	x->km.seq = orig->km.seq;
1403	x->replay = orig->replay;
1404	x->preplay = orig->preplay;
1405
1406	return x;
1407
1408 error:
1409	xfrm_state_put(x);
1410out:
1411	return NULL;
1412}
1413
1414struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1415{
1416	unsigned int h;
1417	struct xfrm_state *x = NULL;
1418
1419	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1420
1421	if (m->reqid) {
1422		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1423				  m->reqid, m->old_family);
1424		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1425			if (x->props.mode != m->mode ||
1426			    x->id.proto != m->proto)
1427				continue;
1428			if (m->reqid && x->props.reqid != m->reqid)
1429				continue;
1430			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1431					     m->old_family) ||
1432			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1433					     m->old_family))
1434				continue;
1435			xfrm_state_hold(x);
1436			break;
1437		}
1438	} else {
1439		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1440				  m->old_family);
1441		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1442			if (x->props.mode != m->mode ||
1443			    x->id.proto != m->proto)
1444				continue;
1445			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1446					     m->old_family) ||
1447			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1448					     m->old_family))
1449				continue;
1450			xfrm_state_hold(x);
1451			break;
1452		}
1453	}
1454
1455	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1456
1457	return x;
1458}
1459EXPORT_SYMBOL(xfrm_migrate_state_find);
1460
1461struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1462				      struct xfrm_migrate *m,
1463				      struct xfrm_encap_tmpl *encap)
1464{
1465	struct xfrm_state *xc;
1466
1467	xc = xfrm_state_clone(x, encap);
1468	if (!xc)
1469		return NULL;
1470
1471	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1472	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1473
1474	/* add state */
1475	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1476		/* a care is needed when the destination address of the
1477		   state is to be updated as it is a part of triplet */
1478		xfrm_state_insert(xc);
1479	} else {
1480		if (xfrm_state_add(xc) < 0)
1481			goto error;
1482	}
1483
1484	return xc;
1485error:
1486	xfrm_state_put(xc);
1487	return NULL;
1488}
1489EXPORT_SYMBOL(xfrm_state_migrate);
1490#endif
1491
1492int xfrm_state_update(struct xfrm_state *x)
1493{
1494	struct xfrm_state *x1, *to_put;
1495	int err;
1496	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1497	struct net *net = xs_net(x);
1498
1499	to_put = NULL;
1500
1501	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1502	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1503
1504	err = -ESRCH;
1505	if (!x1)
1506		goto out;
1507
1508	if (xfrm_state_kern(x1)) {
1509		to_put = x1;
1510		err = -EEXIST;
1511		goto out;
1512	}
1513
1514	if (x1->km.state == XFRM_STATE_ACQ) {
1515		__xfrm_state_insert(x);
1516		x = NULL;
1517	}
1518	err = 0;
1519
1520out:
1521	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1522
1523	if (to_put)
1524		xfrm_state_put(to_put);
1525
1526	if (err)
1527		return err;
1528
1529	if (!x) {
1530		xfrm_state_delete(x1);
1531		xfrm_state_put(x1);
1532		return 0;
1533	}
1534
1535	err = -EINVAL;
1536	spin_lock_bh(&x1->lock);
1537	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1538		if (x->encap && x1->encap &&
1539		    x->encap->encap_type == x1->encap->encap_type)
1540			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1541		else if (x->encap || x1->encap)
1542			goto fail;
1543
1544		if (x->coaddr && x1->coaddr) {
1545			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1546		}
1547		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1548			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1549		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1550		x1->km.dying = 0;
1551
1552		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1553		if (x1->curlft.use_time)
1554			xfrm_state_check_expire(x1);
1555
1556		err = 0;
1557		x->km.state = XFRM_STATE_DEAD;
1558		__xfrm_state_put(x);
1559	}
1560
1561fail:
1562	spin_unlock_bh(&x1->lock);
1563
1564	xfrm_state_put(x1);
1565
1566	return err;
1567}
1568EXPORT_SYMBOL(xfrm_state_update);
1569
1570int xfrm_state_check_expire(struct xfrm_state *x)
1571{
1572	if (!x->curlft.use_time)
1573		x->curlft.use_time = get_seconds();
1574
1575	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1576	    x->curlft.packets >= x->lft.hard_packet_limit) {
1577		x->km.state = XFRM_STATE_EXPIRED;
1578		tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1579		return -EINVAL;
1580	}
1581
1582	if (!x->km.dying &&
1583	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1584	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1585		x->km.dying = 1;
1586		km_state_expired(x, 0, 0);
1587	}
1588	return 0;
1589}
1590EXPORT_SYMBOL(xfrm_state_check_expire);
1591
1592struct xfrm_state *
1593xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1594		  u8 proto, unsigned short family)
1595{
1596	struct xfrm_state *x;
1597
1598	rcu_read_lock();
1599	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1600	rcu_read_unlock();
1601	return x;
1602}
1603EXPORT_SYMBOL(xfrm_state_lookup);
1604
1605struct xfrm_state *
1606xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1607			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1608			 u8 proto, unsigned short family)
1609{
1610	struct xfrm_state *x;
1611
1612	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1613	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1614	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1615	return x;
1616}
1617EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1618
1619struct xfrm_state *
1620xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1621	      u8 proto, const xfrm_address_t *daddr,
1622	      const xfrm_address_t *saddr, int create, unsigned short family)
1623{
1624	struct xfrm_state *x;
1625
1626	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1627	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1628	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1629
1630	return x;
1631}
1632EXPORT_SYMBOL(xfrm_find_acq);
1633
1634#ifdef CONFIG_XFRM_SUB_POLICY
1635int
1636xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1637	       unsigned short family, struct net *net)
1638{
1639	int i;
1640	int err = 0;
1641	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1642	if (!afinfo)
1643		return -EAFNOSUPPORT;
1644
1645	spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1646	if (afinfo->tmpl_sort)
1647		err = afinfo->tmpl_sort(dst, src, n);
1648	else
1649		for (i = 0; i < n; i++)
1650			dst[i] = src[i];
1651	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1652	rcu_read_unlock();
1653	return err;
1654}
1655EXPORT_SYMBOL(xfrm_tmpl_sort);
1656
1657int
1658xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1659		unsigned short family)
1660{
1661	int i;
1662	int err = 0;
1663	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1664	struct net *net = xs_net(*src);
1665
1666	if (!afinfo)
1667		return -EAFNOSUPPORT;
1668
1669	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1670	if (afinfo->state_sort)
1671		err = afinfo->state_sort(dst, src, n);
1672	else
1673		for (i = 0; i < n; i++)
1674			dst[i] = src[i];
1675	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1676	rcu_read_unlock();
1677	return err;
1678}
1679EXPORT_SYMBOL(xfrm_state_sort);
1680#endif
1681
1682/* Silly enough, but I'm lazy to build resolution list */
1683
1684static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1685{
1686	int i;
1687
1688	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1689		struct xfrm_state *x;
1690
1691		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1692			if (x->km.seq == seq &&
1693			    (mark & x->mark.m) == x->mark.v &&
1694			    x->km.state == XFRM_STATE_ACQ) {
1695				xfrm_state_hold(x);
1696				return x;
1697			}
1698		}
1699	}
1700	return NULL;
1701}
1702
1703struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1704{
1705	struct xfrm_state *x;
1706
1707	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1708	x = __xfrm_find_acq_byseq(net, mark, seq);
1709	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1710	return x;
1711}
1712EXPORT_SYMBOL(xfrm_find_acq_byseq);
1713
1714u32 xfrm_get_acqseq(void)
1715{
1716	u32 res;
1717	static atomic_t acqseq;
1718
1719	do {
1720		res = atomic_inc_return(&acqseq);
1721	} while (!res);
1722
1723	return res;
1724}
1725EXPORT_SYMBOL(xfrm_get_acqseq);
1726
1727int verify_spi_info(u8 proto, u32 min, u32 max)
1728{
1729	switch (proto) {
1730	case IPPROTO_AH:
1731	case IPPROTO_ESP:
1732		break;
1733
1734	case IPPROTO_COMP:
1735		/* IPCOMP spi is 16-bits. */
1736		if (max >= 0x10000)
1737			return -EINVAL;
1738		break;
1739
1740	default:
1741		return -EINVAL;
1742	}
1743
1744	if (min > max)
1745		return -EINVAL;
1746
1747	return 0;
1748}
1749EXPORT_SYMBOL(verify_spi_info);
1750
1751int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1752{
1753	struct net *net = xs_net(x);
1754	unsigned int h;
1755	struct xfrm_state *x0;
1756	int err = -ENOENT;
1757	__be32 minspi = htonl(low);
1758	__be32 maxspi = htonl(high);
1759	u32 mark = x->mark.v & x->mark.m;
1760
1761	spin_lock_bh(&x->lock);
1762	if (x->km.state == XFRM_STATE_DEAD)
1763		goto unlock;
1764
1765	err = 0;
1766	if (x->id.spi)
1767		goto unlock;
1768
1769	err = -ENOENT;
1770
1771	if (minspi == maxspi) {
1772		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1773		if (x0) {
1774			xfrm_state_put(x0);
1775			goto unlock;
1776		}
1777		x->id.spi = minspi;
1778	} else {
1779		u32 spi = 0;
1780		for (h = 0; h < high-low+1; h++) {
1781			spi = low + prandom_u32()%(high-low+1);
1782			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1783			if (x0 == NULL) {
1784				x->id.spi = htonl(spi);
1785				break;
1786			}
1787			xfrm_state_put(x0);
1788		}
1789	}
1790	if (x->id.spi) {
1791		spin_lock_bh(&net->xfrm.xfrm_state_lock);
1792		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1793		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1794		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1795
1796		err = 0;
1797	}
1798
1799unlock:
1800	spin_unlock_bh(&x->lock);
1801
1802	return err;
1803}
1804EXPORT_SYMBOL(xfrm_alloc_spi);
1805
1806static bool __xfrm_state_filter_match(struct xfrm_state *x,
1807				      struct xfrm_address_filter *filter)
1808{
1809	if (filter) {
1810		if ((filter->family == AF_INET ||
1811		     filter->family == AF_INET6) &&
1812		    x->props.family != filter->family)
1813			return false;
1814
1815		return addr_match(&x->props.saddr, &filter->saddr,
1816				  filter->splen) &&
1817		       addr_match(&x->id.daddr, &filter->daddr,
1818				  filter->dplen);
1819	}
1820	return true;
1821}
1822
1823int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1824		    int (*func)(struct xfrm_state *, int, void*),
1825		    void *data)
1826{
1827	struct xfrm_state *state;
1828	struct xfrm_state_walk *x;
1829	int err = 0;
1830
1831	if (walk->seq != 0 && list_empty(&walk->all))
1832		return 0;
1833
1834	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1835	if (list_empty(&walk->all))
1836		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1837	else
1838		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1839	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1840		if (x->state == XFRM_STATE_DEAD)
1841			continue;
1842		state = container_of(x, struct xfrm_state, km);
1843		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1844			continue;
1845		if (!__xfrm_state_filter_match(state, walk->filter))
1846			continue;
1847		err = func(state, walk->seq, data);
1848		if (err) {
1849			list_move_tail(&walk->all, &x->all);
1850			goto out;
1851		}
1852		walk->seq++;
1853	}
1854	if (walk->seq == 0) {
1855		err = -ENOENT;
1856		goto out;
1857	}
1858	list_del_init(&walk->all);
1859out:
1860	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1861	return err;
1862}
1863EXPORT_SYMBOL(xfrm_state_walk);
1864
1865void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1866			  struct xfrm_address_filter *filter)
1867{
1868	INIT_LIST_HEAD(&walk->all);
1869	walk->proto = proto;
1870	walk->state = XFRM_STATE_DEAD;
1871	walk->seq = 0;
1872	walk->filter = filter;
1873}
1874EXPORT_SYMBOL(xfrm_state_walk_init);
1875
1876void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1877{
1878	kfree(walk->filter);
1879
1880	if (list_empty(&walk->all))
1881		return;
1882
1883	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1884	list_del(&walk->all);
1885	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1886}
1887EXPORT_SYMBOL(xfrm_state_walk_done);
1888
1889static void xfrm_replay_timer_handler(struct timer_list *t)
1890{
1891	struct xfrm_state *x = from_timer(x, t, rtimer);
1892
1893	spin_lock(&x->lock);
1894
1895	if (x->km.state == XFRM_STATE_VALID) {
1896		if (xfrm_aevent_is_on(xs_net(x)))
1897			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1898		else
1899			x->xflags |= XFRM_TIME_DEFER;
1900	}
1901
1902	spin_unlock(&x->lock);
1903}
1904
1905static LIST_HEAD(xfrm_km_list);
1906
1907void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1908{
1909	struct xfrm_mgr *km;
1910
1911	rcu_read_lock();
1912	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1913		if (km->notify_policy)
1914			km->notify_policy(xp, dir, c);
1915	rcu_read_unlock();
1916}
1917
1918void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1919{
1920	struct xfrm_mgr *km;
1921	rcu_read_lock();
1922	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1923		if (km->notify)
1924			km->notify(x, c);
1925	rcu_read_unlock();
1926}
1927
1928EXPORT_SYMBOL(km_policy_notify);
1929EXPORT_SYMBOL(km_state_notify);
1930
1931void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1932{
1933	struct km_event c;
1934
1935	c.data.hard = hard;
1936	c.portid = portid;
1937	c.event = XFRM_MSG_EXPIRE;
1938	km_state_notify(x, &c);
1939}
1940
1941EXPORT_SYMBOL(km_state_expired);
1942/*
1943 * We send to all registered managers regardless of failure
1944 * We are happy with one success
1945*/
1946int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1947{
1948	int err = -EINVAL, acqret;
1949	struct xfrm_mgr *km;
1950
1951	rcu_read_lock();
1952	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1953		acqret = km->acquire(x, t, pol);
1954		if (!acqret)
1955			err = acqret;
1956	}
1957	rcu_read_unlock();
1958	return err;
1959}
1960EXPORT_SYMBOL(km_query);
1961
1962int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1963{
1964	int err = -EINVAL;
1965	struct xfrm_mgr *km;
1966
1967	rcu_read_lock();
1968	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1969		if (km->new_mapping)
1970			err = km->new_mapping(x, ipaddr, sport);
1971		if (!err)
1972			break;
1973	}
1974	rcu_read_unlock();
1975	return err;
1976}
1977EXPORT_SYMBOL(km_new_mapping);
1978
1979void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1980{
1981	struct km_event c;
1982
1983	c.data.hard = hard;
1984	c.portid = portid;
1985	c.event = XFRM_MSG_POLEXPIRE;
1986	km_policy_notify(pol, dir, &c);
1987}
1988EXPORT_SYMBOL(km_policy_expired);
1989
1990#ifdef CONFIG_XFRM_MIGRATE
1991int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1992	       const struct xfrm_migrate *m, int num_migrate,
1993	       const struct xfrm_kmaddress *k,
1994	       const struct xfrm_encap_tmpl *encap)
1995{
1996	int err = -EINVAL;
1997	int ret;
1998	struct xfrm_mgr *km;
1999
2000	rcu_read_lock();
2001	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2002		if (km->migrate) {
2003			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2004					  encap);
2005			if (!ret)
2006				err = ret;
2007		}
2008	}
2009	rcu_read_unlock();
2010	return err;
2011}
2012EXPORT_SYMBOL(km_migrate);
2013#endif
2014
2015int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2016{
2017	int err = -EINVAL;
2018	int ret;
2019	struct xfrm_mgr *km;
2020
2021	rcu_read_lock();
2022	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2023		if (km->report) {
2024			ret = km->report(net, proto, sel, addr);
2025			if (!ret)
2026				err = ret;
2027		}
2028	}
2029	rcu_read_unlock();
2030	return err;
2031}
2032EXPORT_SYMBOL(km_report);
2033
2034bool km_is_alive(const struct km_event *c)
2035{
2036	struct xfrm_mgr *km;
2037	bool is_alive = false;
2038
2039	rcu_read_lock();
2040	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2041		if (km->is_alive && km->is_alive(c)) {
2042			is_alive = true;
2043			break;
2044		}
2045	}
2046	rcu_read_unlock();
2047
2048	return is_alive;
2049}
2050EXPORT_SYMBOL(km_is_alive);
2051
2052int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2053{
2054	int err;
2055	u8 *data;
2056	struct xfrm_mgr *km;
2057	struct xfrm_policy *pol = NULL;
2058
2059#ifdef CONFIG_COMPAT
2060	if (in_compat_syscall())
2061		return -EOPNOTSUPP;
2062#endif
2063
2064	if (!optval && !optlen) {
2065		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2066		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2067		__sk_dst_reset(sk);
2068		return 0;
2069	}
2070
2071	if (optlen <= 0 || optlen > PAGE_SIZE)
2072		return -EMSGSIZE;
2073
2074	data = memdup_user(optval, optlen);
2075	if (IS_ERR(data))
2076		return PTR_ERR(data);
2077
2078	err = -EINVAL;
2079	rcu_read_lock();
2080	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2081		pol = km->compile_policy(sk, optname, data,
2082					 optlen, &err);
2083		if (err >= 0)
2084			break;
2085	}
2086	rcu_read_unlock();
2087
2088	if (err >= 0) {
2089		xfrm_sk_policy_insert(sk, err, pol);
2090		xfrm_pol_put(pol);
2091		__sk_dst_reset(sk);
2092		err = 0;
2093	}
2094
2095	kfree(data);
2096	return err;
2097}
2098EXPORT_SYMBOL(xfrm_user_policy);
2099
2100static DEFINE_SPINLOCK(xfrm_km_lock);
2101
2102int xfrm_register_km(struct xfrm_mgr *km)
2103{
2104	spin_lock_bh(&xfrm_km_lock);
2105	list_add_tail_rcu(&km->list, &xfrm_km_list);
2106	spin_unlock_bh(&xfrm_km_lock);
2107	return 0;
2108}
2109EXPORT_SYMBOL(xfrm_register_km);
2110
2111int xfrm_unregister_km(struct xfrm_mgr *km)
2112{
2113	spin_lock_bh(&xfrm_km_lock);
2114	list_del_rcu(&km->list);
2115	spin_unlock_bh(&xfrm_km_lock);
2116	synchronize_rcu();
2117	return 0;
2118}
2119EXPORT_SYMBOL(xfrm_unregister_km);
2120
2121int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2122{
2123	int err = 0;
2124
2125	if (WARN_ON(afinfo->family >= NPROTO))
2126		return -EAFNOSUPPORT;
2127
2128	spin_lock_bh(&xfrm_state_afinfo_lock);
2129	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2130		err = -EEXIST;
2131	else
2132		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2133	spin_unlock_bh(&xfrm_state_afinfo_lock);
2134	return err;
2135}
2136EXPORT_SYMBOL(xfrm_state_register_afinfo);
2137
2138int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2139{
2140	int err = 0, family = afinfo->family;
2141
2142	if (WARN_ON(family >= NPROTO))
2143		return -EAFNOSUPPORT;
2144
2145	spin_lock_bh(&xfrm_state_afinfo_lock);
2146	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2147		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2148			err = -EINVAL;
2149		else
2150			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2151	}
2152	spin_unlock_bh(&xfrm_state_afinfo_lock);
2153	synchronize_rcu();
2154	return err;
2155}
2156EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2157
2158struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2159{
2160	if (unlikely(family >= NPROTO))
2161		return NULL;
2162
2163	return rcu_dereference(xfrm_state_afinfo[family]);
2164}
2165
2166struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2167{
2168	struct xfrm_state_afinfo *afinfo;
2169	if (unlikely(family >= NPROTO))
2170		return NULL;
2171	rcu_read_lock();
2172	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2173	if (unlikely(!afinfo))
2174		rcu_read_unlock();
2175	return afinfo;
2176}
2177
2178void xfrm_flush_gc(void)
2179{
2180	flush_work(&xfrm_state_gc_work);
2181}
2182EXPORT_SYMBOL(xfrm_flush_gc);
2183
2184/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2185void xfrm_state_delete_tunnel(struct xfrm_state *x)
2186{
2187	if (x->tunnel) {
2188		struct xfrm_state *t = x->tunnel;
2189
2190		if (atomic_read(&t->tunnel_users) == 2)
2191			xfrm_state_delete(t);
2192		atomic_dec(&t->tunnel_users);
2193		xfrm_state_put(t);
2194		x->tunnel = NULL;
2195	}
2196}
2197EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2198
2199int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2200{
2201	const struct xfrm_type *type = READ_ONCE(x->type);
2202
2203	if (x->km.state == XFRM_STATE_VALID &&
2204	    type && type->get_mtu)
2205		return type->get_mtu(x, mtu);
2206
2207	return mtu - x->props.header_len;
2208}
2209
2210int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2211{
2212	struct xfrm_state_afinfo *afinfo;
2213	struct xfrm_mode *inner_mode;
2214	int family = x->props.family;
2215	int err;
2216
2217	err = -EAFNOSUPPORT;
2218	afinfo = xfrm_state_get_afinfo(family);
2219	if (!afinfo)
2220		goto error;
2221
2222	err = 0;
2223	if (afinfo->init_flags)
2224		err = afinfo->init_flags(x);
2225
2226	rcu_read_unlock();
2227
2228	if (err)
2229		goto error;
2230
2231	err = -EPROTONOSUPPORT;
2232
2233	if (x->sel.family != AF_UNSPEC) {
2234		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2235		if (inner_mode == NULL)
2236			goto error;
2237
2238		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2239		    family != x->sel.family) {
2240			xfrm_put_mode(inner_mode);
2241			goto error;
2242		}
2243
2244		x->inner_mode = inner_mode;
2245	} else {
2246		struct xfrm_mode *inner_mode_iaf;
2247		int iafamily = AF_INET;
2248
2249		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2250		if (inner_mode == NULL)
2251			goto error;
2252
2253		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2254			xfrm_put_mode(inner_mode);
2255			goto error;
2256		}
2257		x->inner_mode = inner_mode;
2258
2259		if (x->props.family == AF_INET)
2260			iafamily = AF_INET6;
2261
2262		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2263		if (inner_mode_iaf) {
2264			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2265				x->inner_mode_iaf = inner_mode_iaf;
2266			else
2267				xfrm_put_mode(inner_mode_iaf);
2268		}
2269	}
2270
2271	x->type = xfrm_get_type(x->id.proto, family);
2272	if (x->type == NULL)
2273		goto error;
2274
2275	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2276
2277	err = x->type->init_state(x);
2278	if (err)
2279		goto error;
2280
2281	x->outer_mode = xfrm_get_mode(x->props.mode, family);
2282	if (x->outer_mode == NULL) {
2283		err = -EPROTONOSUPPORT;
2284		goto error;
2285	}
2286
2287	if (init_replay) {
2288		err = xfrm_init_replay(x);
2289		if (err)
2290			goto error;
2291	}
2292
2293error:
2294	return err;
2295}
2296
2297EXPORT_SYMBOL(__xfrm_init_state);
2298
2299int xfrm_init_state(struct xfrm_state *x)
2300{
2301	int err;
2302
2303	err = __xfrm_init_state(x, true, false);
2304	if (!err)
2305		x->km.state = XFRM_STATE_VALID;
2306
2307	return err;
2308}
2309
2310EXPORT_SYMBOL(xfrm_init_state);
2311
2312int __net_init xfrm_state_init(struct net *net)
2313{
2314	unsigned int sz;
2315
2316	INIT_LIST_HEAD(&net->xfrm.state_all);
2317
2318	sz = sizeof(struct hlist_head) * 8;
2319
2320	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2321	if (!net->xfrm.state_bydst)
2322		goto out_bydst;
2323	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2324	if (!net->xfrm.state_bysrc)
2325		goto out_bysrc;
2326	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2327	if (!net->xfrm.state_byspi)
2328		goto out_byspi;
2329	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2330
2331	net->xfrm.state_num = 0;
2332	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2333	spin_lock_init(&net->xfrm.xfrm_state_lock);
2334	return 0;
2335
2336out_byspi:
2337	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2338out_bysrc:
2339	xfrm_hash_free(net->xfrm.state_bydst, sz);
2340out_bydst:
2341	return -ENOMEM;
2342}
2343
2344void xfrm_state_fini(struct net *net)
2345{
2346	unsigned int sz;
2347
2348	flush_work(&net->xfrm.state_hash_work);
2349	xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2350	flush_work(&xfrm_state_gc_work);
2351
2352	WARN_ON(!list_empty(&net->xfrm.state_all));
2353
2354	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2355	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2356	xfrm_hash_free(net->xfrm.state_byspi, sz);
2357	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2358	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2359	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2360	xfrm_hash_free(net->xfrm.state_bydst, sz);
2361}
2362
2363#ifdef CONFIG_AUDITSYSCALL
2364static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2365				     struct audit_buffer *audit_buf)
2366{
2367	struct xfrm_sec_ctx *ctx = x->security;
2368	u32 spi = ntohl(x->id.spi);
2369
2370	if (ctx)
2371		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2372				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2373
2374	switch (x->props.family) {
2375	case AF_INET:
2376		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2377				 &x->props.saddr.a4, &x->id.daddr.a4);
2378		break;
2379	case AF_INET6:
2380		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2381				 x->props.saddr.a6, x->id.daddr.a6);
2382		break;
2383	}
2384
2385	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2386}
2387
2388static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2389				      struct audit_buffer *audit_buf)
2390{
2391	const struct iphdr *iph4;
2392	const struct ipv6hdr *iph6;
2393
2394	switch (family) {
2395	case AF_INET:
2396		iph4 = ip_hdr(skb);
2397		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2398				 &iph4->saddr, &iph4->daddr);
2399		break;
2400	case AF_INET6:
2401		iph6 = ipv6_hdr(skb);
2402		audit_log_format(audit_buf,
2403				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2404				 &iph6->saddr, &iph6->daddr,
2405				 iph6->flow_lbl[0] & 0x0f,
2406				 iph6->flow_lbl[1],
2407				 iph6->flow_lbl[2]);
2408		break;
2409	}
2410}
2411
2412void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2413{
2414	struct audit_buffer *audit_buf;
2415
2416	audit_buf = xfrm_audit_start("SAD-add");
2417	if (audit_buf == NULL)
2418		return;
2419	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2420	xfrm_audit_helper_sainfo(x, audit_buf);
2421	audit_log_format(audit_buf, " res=%u", result);
2422	audit_log_end(audit_buf);
2423}
2424EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2425
2426void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2427{
2428	struct audit_buffer *audit_buf;
2429
2430	audit_buf = xfrm_audit_start("SAD-delete");
2431	if (audit_buf == NULL)
2432		return;
2433	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2434	xfrm_audit_helper_sainfo(x, audit_buf);
2435	audit_log_format(audit_buf, " res=%u", result);
2436	audit_log_end(audit_buf);
2437}
2438EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2439
2440void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2441				      struct sk_buff *skb)
2442{
2443	struct audit_buffer *audit_buf;
2444	u32 spi;
2445
2446	audit_buf = xfrm_audit_start("SA-replay-overflow");
2447	if (audit_buf == NULL)
2448		return;
2449	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2450	/* don't record the sequence number because it's inherent in this kind
2451	 * of audit message */
2452	spi = ntohl(x->id.spi);
2453	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2454	audit_log_end(audit_buf);
2455}
2456EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2457
2458void xfrm_audit_state_replay(struct xfrm_state *x,
2459			     struct sk_buff *skb, __be32 net_seq)
2460{
2461	struct audit_buffer *audit_buf;
2462	u32 spi;
2463
2464	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2465	if (audit_buf == NULL)
2466		return;
2467	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2468	spi = ntohl(x->id.spi);
2469	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2470			 spi, spi, ntohl(net_seq));
2471	audit_log_end(audit_buf);
2472}
2473EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2474
2475void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2476{
2477	struct audit_buffer *audit_buf;
2478
2479	audit_buf = xfrm_audit_start("SA-notfound");
2480	if (audit_buf == NULL)
2481		return;
2482	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2483	audit_log_end(audit_buf);
2484}
2485EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2486
2487void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2488			       __be32 net_spi, __be32 net_seq)
2489{
2490	struct audit_buffer *audit_buf;
2491	u32 spi;
2492
2493	audit_buf = xfrm_audit_start("SA-notfound");
2494	if (audit_buf == NULL)
2495		return;
2496	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2497	spi = ntohl(net_spi);
2498	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2499			 spi, spi, ntohl(net_seq));
2500	audit_log_end(audit_buf);
2501}
2502EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2503
2504void xfrm_audit_state_icvfail(struct xfrm_state *x,
2505			      struct sk_buff *skb, u8 proto)
2506{
2507	struct audit_buffer *audit_buf;
2508	__be32 net_spi;
2509	__be32 net_seq;
2510
2511	audit_buf = xfrm_audit_start("SA-icv-failure");
2512	if (audit_buf == NULL)
2513		return;
2514	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2515	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2516		u32 spi = ntohl(net_spi);
2517		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2518				 spi, spi, ntohl(net_seq));
2519	}
2520	audit_log_end(audit_buf);
2521}
2522EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2523#endif /* CONFIG_AUDITSYSCALL */