xfrm_state.c 53.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12
/*
 * xfrm_state.c
 *
 * Changes:
 *	Mitsuru KANDA @USAGI
 * 	Kazunori MIYAZAWA @USAGI
 * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
 * 		IPv6 support
 * 	YOSHIFUJI Hideaki @USAGI
 * 		Split up af-specific functions
 *	Derek Atkins <derek@ihtfp.com>
 *		Add UDP Encapsulation
13
 *
L
Linus Torvalds 已提交
14 15 16 17 18 19 20
 */

#include <linux/workqueue.h>
#include <net/xfrm.h>
#include <linux/pfkeyv2.h>
#include <linux/ipsec.h>
#include <linux/module.h>
21
#include <linux/cache.h>
P
Paul Moore 已提交
22
#include <linux/audit.h>
23
#include <asm/uaccess.h>
24
#include <linux/ktime.h>
25
#include <linux/slab.h>
26 27
#include <linux/interrupt.h>
#include <linux/kernel.h>
L
Linus Torvalds 已提交
28

29 30
#include "xfrm_hash.h"

L
Linus Torvalds 已提交
31 32 33
/* Each xfrm_state may be linked to two tables:

   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
L
Linus Torvalds 已提交
35 36 37 38 39
      destination/tunnel endpoint. (output)
 */

static DEFINE_SPINLOCK(xfrm_state_lock);

40 41
static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;

42 43 44
static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);

45
static inline unsigned int xfrm_dst_hash(struct net *net,
46 47
					 const xfrm_address_t *daddr,
					 const xfrm_address_t *saddr,
48
					 u32 reqid,
49
					 unsigned short family)
50
{
51
	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
52 53
}

54
static inline unsigned int xfrm_src_hash(struct net *net,
55 56
					 const xfrm_address_t *daddr,
					 const xfrm_address_t *saddr,
57
					 unsigned short family)
58
{
59
	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
60 61 62
}

static inline unsigned int
63 64
xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
	      __be32 spi, u8 proto, unsigned short family)
65
{
66
	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
67 68 69 70 71 72 73 74 75 76 77 78 79 80
}

static void xfrm_hash_transfer(struct hlist_head *list,
			       struct hlist_head *ndsttable,
			       struct hlist_head *nsrctable,
			       struct hlist_head *nspitable,
			       unsigned int nhashmask)
{
	struct hlist_node *entry, *tmp;
	struct xfrm_state *x;

	hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
		unsigned int h;

81 82 83
		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
				    x->props.reqid, x->props.family,
				    nhashmask);
84 85
		hlist_add_head(&x->bydst, ndsttable+h);

86 87
		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
				    x->props.family,
88 89 90
				    nhashmask);
		hlist_add_head(&x->bysrc, nsrctable+h);

91 92 93 94 95 96
		if (x->id.spi) {
			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
					    x->id.proto, x->props.family,
					    nhashmask);
			hlist_add_head(&x->byspi, nspitable+h);
		}
97 98 99
	}
}

100
static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
101
{
102
	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
103 104 105 106
}

static DEFINE_MUTEX(hash_resize_mutex);

107
static void xfrm_hash_resize(struct work_struct *work)
108
{
109
	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
110 111 112 113 114 115 116
	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
	unsigned long nsize, osize;
	unsigned int nhashmask, ohashmask;
	int i;

	mutex_lock(&hash_resize_mutex);

117
	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
118
	ndst = xfrm_hash_alloc(nsize);
119 120
	if (!ndst)
		goto out_unlock;
121
	nsrc = xfrm_hash_alloc(nsize);
122
	if (!nsrc) {
123
		xfrm_hash_free(ndst, nsize);
124 125
		goto out_unlock;
	}
126
	nspi = xfrm_hash_alloc(nsize);
127
	if (!nspi) {
128 129
		xfrm_hash_free(ndst, nsize);
		xfrm_hash_free(nsrc, nsize);
130 131 132 133 134 135
		goto out_unlock;
	}

	spin_lock_bh(&xfrm_state_lock);

	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
136 137
	for (i = net->xfrm.state_hmask; i >= 0; i--)
		xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
138 139
				   nhashmask);

140 141 142 143
	odst = net->xfrm.state_bydst;
	osrc = net->xfrm.state_bysrc;
	ospi = net->xfrm.state_byspi;
	ohashmask = net->xfrm.state_hmask;
144

145 146 147 148
	net->xfrm.state_bydst = ndst;
	net->xfrm.state_bysrc = nsrc;
	net->xfrm.state_byspi = nspi;
	net->xfrm.state_hmask = nhashmask;
149 150 151 152

	spin_unlock_bh(&xfrm_state_lock);

	osize = (ohashmask + 1) * sizeof(struct hlist_head);
153 154 155
	xfrm_hash_free(odst, osize);
	xfrm_hash_free(osrc, osize);
	xfrm_hash_free(ospi, osize);
156 157 158 159 160

out_unlock:
	mutex_unlock(&hash_resize_mutex);
}

161 162
static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
L
Linus Torvalds 已提交
163 164 165

static DEFINE_SPINLOCK(xfrm_state_gc_lock);

166
int __xfrm_state_delete(struct xfrm_state *x);
L
Linus Torvalds 已提交
167

168
int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
169
void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
L
Linus Torvalds 已提交
170

171 172 173 174 175
static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
{
	struct xfrm_state_afinfo *afinfo;
	if (unlikely(family >= NPROTO))
		return NULL;
176
	spin_lock_bh(&xfrm_state_afinfo_lock);
177 178
	afinfo = xfrm_state_afinfo[family];
	if (unlikely(!afinfo))
179
		spin_unlock_bh(&xfrm_state_afinfo_lock);
180 181 182 183 184
	return afinfo;
}

static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
{
185
	spin_unlock_bh(&xfrm_state_afinfo_lock);
186 187
}

188
int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
189 190
{
	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
191
	const struct xfrm_type **typemap;
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206
	int err = 0;

	if (unlikely(afinfo == NULL))
		return -EAFNOSUPPORT;
	typemap = afinfo->type_map;

	if (likely(typemap[type->proto] == NULL))
		typemap[type->proto] = type;
	else
		err = -EEXIST;
	xfrm_state_unlock_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_register_type);

207
int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
208 209
{
	struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
210
	const struct xfrm_type **typemap;
211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
	int err = 0;

	if (unlikely(afinfo == NULL))
		return -EAFNOSUPPORT;
	typemap = afinfo->type_map;

	if (unlikely(typemap[type->proto] != type))
		err = -ENOENT;
	else
		typemap[type->proto] = NULL;
	xfrm_state_unlock_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_unregister_type);

226
static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
227 228
{
	struct xfrm_state_afinfo *afinfo;
229 230
	const struct xfrm_type **typemap;
	const struct xfrm_type *type;
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
	int modload_attempted = 0;

retry:
	afinfo = xfrm_state_get_afinfo(family);
	if (unlikely(afinfo == NULL))
		return NULL;
	typemap = afinfo->type_map;

	type = typemap[proto];
	if (unlikely(type && !try_module_get(type->owner)))
		type = NULL;
	if (!type && !modload_attempted) {
		xfrm_state_put_afinfo(afinfo);
		request_module("xfrm-type-%d-%d", family, proto);
		modload_attempted = 1;
		goto retry;
	}

	xfrm_state_put_afinfo(afinfo);
	return type;
}

253
static void xfrm_put_type(const struct xfrm_type *type)
254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
{
	module_put(type->owner);
}

int xfrm_register_mode(struct xfrm_mode *mode, int family)
{
	struct xfrm_state_afinfo *afinfo;
	struct xfrm_mode **modemap;
	int err;

	if (unlikely(mode->encap >= XFRM_MODE_MAX))
		return -EINVAL;

	afinfo = xfrm_state_lock_afinfo(family);
	if (unlikely(afinfo == NULL))
		return -EAFNOSUPPORT;

	err = -EEXIST;
	modemap = afinfo->mode_map;
273 274
	if (modemap[mode->encap])
		goto out;
275

276 277 278 279 280 281 282 283 284
	err = -ENOENT;
	if (!try_module_get(afinfo->owner))
		goto out;

	mode->afinfo = afinfo;
	modemap[mode->encap] = mode;
	err = 0;

out:
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
	xfrm_state_unlock_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_register_mode);

int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
{
	struct xfrm_state_afinfo *afinfo;
	struct xfrm_mode **modemap;
	int err;

	if (unlikely(mode->encap >= XFRM_MODE_MAX))
		return -EINVAL;

	afinfo = xfrm_state_lock_afinfo(family);
	if (unlikely(afinfo == NULL))
		return -EAFNOSUPPORT;

	err = -ENOENT;
	modemap = afinfo->mode_map;
	if (likely(modemap[mode->encap] == mode)) {
		modemap[mode->encap] = NULL;
307
		module_put(mode->afinfo->owner);
308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
		err = 0;
	}

	xfrm_state_unlock_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_unregister_mode);

static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
{
	struct xfrm_state_afinfo *afinfo;
	struct xfrm_mode *mode;
	int modload_attempted = 0;

	if (unlikely(encap >= XFRM_MODE_MAX))
		return NULL;

retry:
	afinfo = xfrm_state_get_afinfo(family);
	if (unlikely(afinfo == NULL))
		return NULL;

	mode = afinfo->mode_map[encap];
	if (unlikely(mode && !try_module_get(mode->owner)))
		mode = NULL;
	if (!mode && !modload_attempted) {
		xfrm_state_put_afinfo(afinfo);
		request_module("xfrm-mode-%d-%d", family, encap);
		modload_attempted = 1;
		goto retry;
	}

	xfrm_state_put_afinfo(afinfo);
	return mode;
}

static void xfrm_put_mode(struct xfrm_mode *mode)
{
	module_put(mode->owner);
}

L
Linus Torvalds 已提交
349 350
static void xfrm_state_gc_destroy(struct xfrm_state *x)
{
351
	tasklet_hrtimer_cancel(&x->mtimer);
352
	del_timer_sync(&x->rtimer);
J
Jesper Juhl 已提交
353 354 355 356
	kfree(x->aalg);
	kfree(x->ealg);
	kfree(x->calg);
	kfree(x->encap);
357
	kfree(x->coaddr);
358 359
	kfree(x->replay_esn);
	kfree(x->preplay_esn);
360 361
	if (x->inner_mode)
		xfrm_put_mode(x->inner_mode);
362 363
	if (x->inner_mode_iaf)
		xfrm_put_mode(x->inner_mode_iaf);
364 365
	if (x->outer_mode)
		xfrm_put_mode(x->outer_mode);
L
Linus Torvalds 已提交
366 367 368 369
	if (x->type) {
		x->type->destructor(x);
		xfrm_put_type(x->type);
	}
370
	security_xfrm_state_free(x);
L
Linus Torvalds 已提交
371 372 373
	kfree(x);
}

374
static void xfrm_state_gc_task(struct work_struct *work)
L
Linus Torvalds 已提交
375
{
376
	struct net *net = container_of(work, struct net, xfrm.state_gc_work);
H
Herbert Xu 已提交
377 378 379
	struct xfrm_state *x;
	struct hlist_node *entry, *tmp;
	struct hlist_head gc_list;
L
Linus Torvalds 已提交
380 381

	spin_lock_bh(&xfrm_state_gc_lock);
382
	hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
L
Linus Torvalds 已提交
383 384
	spin_unlock_bh(&xfrm_state_gc_lock);

H
Herbert Xu 已提交
385
	hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
L
Linus Torvalds 已提交
386
		xfrm_state_gc_destroy(x);
387

A
Alexey Dobriyan 已提交
388
	wake_up(&net->xfrm.km_waitq);
L
Linus Torvalds 已提交
389 390 391 392 393 394 395
}

static inline unsigned long make_jiffies(long secs)
{
	if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
		return MAX_SCHEDULE_TIMEOUT-1;
	else
396
		return secs*HZ;
L
Linus Torvalds 已提交
397 398
}

399
static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
L
Linus Torvalds 已提交
400
{
401 402
	struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
	struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
403
	struct net *net = xs_net(x);
404
	unsigned long now = get_seconds();
L
Linus Torvalds 已提交
405 406
	long next = LONG_MAX;
	int warn = 0;
J
Joy Latten 已提交
407
	int err = 0;
L
Linus Torvalds 已提交
408 409 410 411 412 413 414 415 416

	spin_lock(&x->lock);
	if (x->km.state == XFRM_STATE_DEAD)
		goto out;
	if (x->km.state == XFRM_STATE_EXPIRED)
		goto expired;
	if (x->lft.hard_add_expires_seconds) {
		long tmo = x->lft.hard_add_expires_seconds +
			x->curlft.add_time - now;
417 418 419 420 421 422 423 424 425 426 427
		if (tmo <= 0) {
			if (x->xflags & XFRM_SOFT_EXPIRE) {
				/* enter hard expire without soft expire first?!
				 * setting a new date could trigger this.
				 * workarbound: fix x->curflt.add_time by below:
				 */
				x->curlft.add_time = now - x->saved_tmo - 1;
				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
			} else
				goto expired;
		}
L
Linus Torvalds 已提交
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
		if (tmo < next)
			next = tmo;
	}
	if (x->lft.hard_use_expires_seconds) {
		long tmo = x->lft.hard_use_expires_seconds +
			(x->curlft.use_time ? : now) - now;
		if (tmo <= 0)
			goto expired;
		if (tmo < next)
			next = tmo;
	}
	if (x->km.dying)
		goto resched;
	if (x->lft.soft_add_expires_seconds) {
		long tmo = x->lft.soft_add_expires_seconds +
			x->curlft.add_time - now;
444
		if (tmo <= 0) {
L
Linus Torvalds 已提交
445
			warn = 1;
446 447
			x->xflags &= ~XFRM_SOFT_EXPIRE;
		} else if (tmo < next) {
L
Linus Torvalds 已提交
448
			next = tmo;
449 450 451
			x->xflags |= XFRM_SOFT_EXPIRE;
			x->saved_tmo = tmo;
		}
L
Linus Torvalds 已提交
452 453 454 455 456 457 458 459 460 461
	}
	if (x->lft.soft_use_expires_seconds) {
		long tmo = x->lft.soft_use_expires_seconds +
			(x->curlft.use_time ? : now) - now;
		if (tmo <= 0)
			warn = 1;
		else if (tmo < next)
			next = tmo;
	}

462
	x->km.dying = warn;
L
Linus Torvalds 已提交
463
	if (warn)
464
		km_state_expired(x, 0, 0);
L
Linus Torvalds 已提交
465
resched:
466 467 468
	if (next != LONG_MAX){
		tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
	}
469

L
Linus Torvalds 已提交
470 471 472 473 474
	goto out;

expired:
	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
		x->km.state = XFRM_STATE_EXPIRED;
475
		wake_up(&net->xfrm.km_waitq);
L
Linus Torvalds 已提交
476 477 478
		next = 2;
		goto resched;
	}
J
Joy Latten 已提交
479 480 481

	err = __xfrm_state_delete(x);
	if (!err && x->id.spi)
482
		km_state_expired(x, 1, 0);
L
Linus Torvalds 已提交
483

J
Joy Latten 已提交
484
	xfrm_audit_state_delete(x, err ? 0 : 1,
485 486
				audit_get_loginuid(current),
				audit_get_sessionid(current), 0);
J
Joy Latten 已提交
487

L
Linus Torvalds 已提交
488 489
out:
	spin_unlock(&x->lock);
490
	return HRTIMER_NORESTART;
L
Linus Torvalds 已提交
491 492
}

493 494
static void xfrm_replay_timer_handler(unsigned long data);

495
struct xfrm_state *xfrm_state_alloc(struct net *net)
L
Linus Torvalds 已提交
496 497 498
{
	struct xfrm_state *x;

499
	x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
L
Linus Torvalds 已提交
500 501

	if (x) {
502
		write_pnet(&x->xs_net, net);
L
Linus Torvalds 已提交
503 504
		atomic_set(&x->refcnt, 1);
		atomic_set(&x->tunnel_users, 0);
H
Herbert Xu 已提交
505
		INIT_LIST_HEAD(&x->km.all);
506 507 508
		INIT_HLIST_NODE(&x->bydst);
		INIT_HLIST_NODE(&x->bysrc);
		INIT_HLIST_NODE(&x->byspi);
509
		tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
510 511
		setup_timer(&x->rtimer, xfrm_replay_timer_handler,
				(unsigned long)x);
512
		x->curlft.add_time = get_seconds();
L
Linus Torvalds 已提交
513 514 515 516
		x->lft.soft_byte_limit = XFRM_INF;
		x->lft.soft_packet_limit = XFRM_INF;
		x->lft.hard_byte_limit = XFRM_INF;
		x->lft.hard_packet_limit = XFRM_INF;
517 518
		x->replay_maxage = 0;
		x->replay_maxdiff = 0;
519 520
		x->inner_mode = NULL;
		x->inner_mode_iaf = NULL;
L
Linus Torvalds 已提交
521 522 523 524 525 526 527 528
		spin_lock_init(&x->lock);
	}
	return x;
}
EXPORT_SYMBOL(xfrm_state_alloc);

void __xfrm_state_destroy(struct xfrm_state *x)
{
529 530
	struct net *net = xs_net(x);

531
	WARN_ON(x->km.state != XFRM_STATE_DEAD);
L
Linus Torvalds 已提交
532 533

	spin_lock_bh(&xfrm_state_gc_lock);
534
	hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
L
Linus Torvalds 已提交
535
	spin_unlock_bh(&xfrm_state_gc_lock);
536
	schedule_work(&net->xfrm.state_gc_work);
L
Linus Torvalds 已提交
537 538 539
}
EXPORT_SYMBOL(__xfrm_state_destroy);

540
int __xfrm_state_delete(struct xfrm_state *x)
L
Linus Torvalds 已提交
541
{
542
	struct net *net = xs_net(x);
543 544
	int err = -ESRCH;

L
Linus Torvalds 已提交
545 546 547
	if (x->km.state != XFRM_STATE_DEAD) {
		x->km.state = XFRM_STATE_DEAD;
		spin_lock(&xfrm_state_lock);
H
Herbert Xu 已提交
548
		list_del(&x->km.all);
549 550
		hlist_del(&x->bydst);
		hlist_del(&x->bysrc);
551
		if (x->id.spi)
552
			hlist_del(&x->byspi);
553
		net->xfrm.state_num--;
L
Linus Torvalds 已提交
554 555 556 557 558 559
		spin_unlock(&xfrm_state_lock);

		/* All xfrm_state objects are created by xfrm_state_alloc.
		 * The xfrm_state_alloc call gives a reference, and that
		 * is what we are dropping here.
		 */
560
		xfrm_state_put(x);
561
		err = 0;
L
Linus Torvalds 已提交
562
	}
563 564

	return err;
L
Linus Torvalds 已提交
565
}
566
EXPORT_SYMBOL(__xfrm_state_delete);
L
Linus Torvalds 已提交
567

568
int xfrm_state_delete(struct xfrm_state *x)
L
Linus Torvalds 已提交
569
{
570 571
	int err;

L
Linus Torvalds 已提交
572
	spin_lock_bh(&x->lock);
573
	err = __xfrm_state_delete(x);
L
Linus Torvalds 已提交
574
	spin_unlock_bh(&x->lock);
575 576

	return err;
L
Linus Torvalds 已提交
577 578 579
}
EXPORT_SYMBOL(xfrm_state_delete);

580 581
#ifdef CONFIG_SECURITY_NETWORK_XFRM
static inline int
A
Alexey Dobriyan 已提交
582
xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
L
Linus Torvalds 已提交
583
{
584 585
	int i, err = 0;

A
Alexey Dobriyan 已提交
586
	for (i = 0; i <= net->xfrm.state_hmask; i++) {
587 588 589
		struct hlist_node *entry;
		struct xfrm_state *x;

A
Alexey Dobriyan 已提交
590
		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
591 592
			if (xfrm_id_proto_match(x->id.proto, proto) &&
			   (err = security_xfrm_state_delete(x)) != 0) {
J
Joy Latten 已提交
593 594
				xfrm_audit_state_delete(x, 0,
							audit_info->loginuid,
595
							audit_info->sessionid,
J
Joy Latten 已提交
596
							audit_info->secid);
597 598 599 600 601 602 603 604 605
				return err;
			}
		}
	}

	return err;
}
#else
static inline int
A
Alexey Dobriyan 已提交
606
xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
607 608 609 610 611
{
	return 0;
}
#endif

A
Alexey Dobriyan 已提交
612
int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
613
{
614
	int i, err = 0, cnt = 0;
L
Linus Torvalds 已提交
615 616

	spin_lock_bh(&xfrm_state_lock);
A
Alexey Dobriyan 已提交
617
	err = xfrm_state_flush_secctx_check(net, proto, audit_info);
618 619 620
	if (err)
		goto out;

621
	err = -ESRCH;
A
Alexey Dobriyan 已提交
622
	for (i = 0; i <= net->xfrm.state_hmask; i++) {
623 624
		struct hlist_node *entry;
		struct xfrm_state *x;
L
Linus Torvalds 已提交
625
restart:
A
Alexey Dobriyan 已提交
626
		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
L
Linus Torvalds 已提交
627
			if (!xfrm_state_kern(x) &&
628
			    xfrm_id_proto_match(x->id.proto, proto)) {
L
Linus Torvalds 已提交
629 630 631
				xfrm_state_hold(x);
				spin_unlock_bh(&xfrm_state_lock);

J
Joy Latten 已提交
632
				err = xfrm_state_delete(x);
J
Joy Latten 已提交
633 634
				xfrm_audit_state_delete(x, err ? 0 : 1,
							audit_info->loginuid,
635
							audit_info->sessionid,
J
Joy Latten 已提交
636
							audit_info->secid);
L
Linus Torvalds 已提交
637
				xfrm_state_put(x);
638 639
				if (!err)
					cnt++;
L
Linus Torvalds 已提交
640 641 642 643 644 645

				spin_lock_bh(&xfrm_state_lock);
				goto restart;
			}
		}
	}
646 647
	if (cnt)
		err = 0;
648 649

out:
L
Linus Torvalds 已提交
650
	spin_unlock_bh(&xfrm_state_lock);
A
Alexey Dobriyan 已提交
651
	wake_up(&net->xfrm.km_waitq);
652
	return err;
L
Linus Torvalds 已提交
653 654 655
}
EXPORT_SYMBOL(xfrm_state_flush);

656
void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
J
Jamal Hadi Salim 已提交
657 658
{
	spin_lock_bh(&xfrm_state_lock);
659 660
	si->sadcnt = net->xfrm.state_num;
	si->sadhcnt = net->xfrm.state_hmask;
J
Jamal Hadi Salim 已提交
661 662 663 664 665
	si->sadhmcnt = xfrm_state_hashmax;
	spin_unlock_bh(&xfrm_state_lock);
}
EXPORT_SYMBOL(xfrm_sad_getinfo);

L
Linus Torvalds 已提交
666
static int
667
xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
668
		    const struct xfrm_tmpl *tmpl,
669
		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
670
		    unsigned short family)
L
Linus Torvalds 已提交
671 672 673 674
{
	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
	if (!afinfo)
		return -1;
675 676 677 678 679 680 681 682 683
	afinfo->init_tempsel(&x->sel, fl);

	if (family != tmpl->encap_family) {
		xfrm_state_put_afinfo(afinfo);
		afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
		if (!afinfo)
			return -1;
	}
	afinfo->init_temprop(x, tmpl, daddr, saddr);
L
Linus Torvalds 已提交
684 685 686 687
	xfrm_state_put_afinfo(afinfo);
	return 0;
}

688 689 690 691
static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
					      const xfrm_address_t *daddr,
					      __be32 spi, u8 proto,
					      unsigned short family)
692
{
693
	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
694
	struct xfrm_state *x;
695
	struct hlist_node *entry;
696

697
	hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
698 699
		if (x->props.family != family ||
		    x->id.spi       != spi ||
700 701
		    x->id.proto     != proto ||
		    xfrm_addr_cmp(&x->id.daddr, daddr, family))
702 703
			continue;

J
Jamal Hadi Salim 已提交
704 705
		if ((mark & x->mark.m) != x->mark.v)
			continue;
706 707 708 709 710 711 712
		xfrm_state_hold(x);
		return x;
	}

	return NULL;
}

713 714 715 716
static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
						     const xfrm_address_t *daddr,
						     const xfrm_address_t *saddr,
						     u8 proto, unsigned short family)
717
{
718
	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
719
	struct xfrm_state *x;
720
	struct hlist_node *entry;
721

722
	hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
723
		if (x->props.family != family ||
724 725 726
		    x->id.proto     != proto ||
		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
727 728
			continue;

J
Jamal Hadi Salim 已提交
729 730
		if ((mark & x->mark.m) != x->mark.v)
			continue;
731 732 733 734 735 736 737 738 739 740
		xfrm_state_hold(x);
		return x;
	}

	return NULL;
}

static inline struct xfrm_state *
__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
{
741
	struct net *net = xs_net(x);
742
	u32 mark = x->mark.v & x->mark.m;
743

744
	if (use_spi)
745 746
		return __xfrm_state_lookup(net, mark, &x->id.daddr,
					   x->id.spi, x->id.proto, family);
747
	else
748 749
		return __xfrm_state_lookup_byaddr(net, mark,
						  &x->id.daddr,
750 751 752 753
						  &x->props.saddr,
						  x->id.proto, family);
}

754
static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
755 756
{
	if (have_hash_collision &&
757 758 759
	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
	    net->xfrm.state_num > net->xfrm.state_hmask)
		schedule_work(&net->xfrm.state_hash_work);
760 761
}

762
static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
763
			       const struct flowi *fl, unsigned short family,
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
			       struct xfrm_state **best, int *acq_in_progress,
			       int *error)
{
	/* Resolution logic:
	 * 1. There is a valid state with matching selector. Done.
	 * 2. Valid state with inappropriate selector. Skip.
	 *
	 * Entering area of "sysdeps".
	 *
	 * 3. If state is not valid, selector is temporary, it selects
	 *    only session which triggered previous resolution. Key
	 *    manager will do something to install a state with proper
	 *    selector.
	 */
	if (x->km.state == XFRM_STATE_VALID) {
		if ((x->sel.family &&
		     !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
		    !security_xfrm_state_pol_flow_match(x, pol, fl))
			return;

		if (!*best ||
		    (*best)->km.dying > x->km.dying ||
		    ((*best)->km.dying == x->km.dying &&
		     (*best)->curlft.add_time < x->curlft.add_time))
			*best = x;
	} else if (x->km.state == XFRM_STATE_ACQ) {
		*acq_in_progress = 1;
	} else if (x->km.state == XFRM_STATE_ERROR ||
		   x->km.state == XFRM_STATE_EXPIRED) {
		if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
		    security_xfrm_state_pol_flow_match(x, pol, fl))
			*error = -ESRCH;
	}
}

L
Linus Torvalds 已提交
799
struct xfrm_state *
800
xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
801
		const struct flowi *fl, struct xfrm_tmpl *tmpl,
L
Linus Torvalds 已提交
802 803 804
		struct xfrm_policy *pol, int *err,
		unsigned short family)
{
805
	static xfrm_address_t saddr_wildcard = { };
806
	struct net *net = xp_net(pol);
807
	unsigned int h, h_wildcard;
808
	struct hlist_node *entry;
809
	struct xfrm_state *x, *x0, *to_put;
L
Linus Torvalds 已提交
810 811 812
	int acquire_in_progress = 0;
	int error = 0;
	struct xfrm_state *best = NULL;
813
	u32 mark = pol->mark.v & pol->mark.m;
814
	unsigned short encap_family = tmpl->encap_family;
815

816 817
	to_put = NULL;

L
Linus Torvalds 已提交
818
	spin_lock_bh(&xfrm_state_lock);
819
	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
820
	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
821
		if (x->props.family == encap_family &&
L
Linus Torvalds 已提交
822
		    x->props.reqid == tmpl->reqid &&
J
Jamal Hadi Salim 已提交
823
		    (mark & x->mark.m) == x->mark.v &&
824
		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
825
		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
L
Linus Torvalds 已提交
826 827
		    tmpl->mode == x->props.mode &&
		    tmpl->id.proto == x->id.proto &&
828
		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
829
			xfrm_state_look_at(pol, x, fl, encap_family,
830 831 832 833 834
					   &best, &acquire_in_progress, &error);
	}
	if (best)
		goto found;

835
	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
836
	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
837
		if (x->props.family == encap_family &&
838
		    x->props.reqid == tmpl->reqid &&
J
Jamal Hadi Salim 已提交
839
		    (mark & x->mark.m) == x->mark.v &&
840
		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
841
		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
842 843 844
		    tmpl->mode == x->props.mode &&
		    tmpl->id.proto == x->id.proto &&
		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
845
			xfrm_state_look_at(pol, x, fl, encap_family,
846
					   &best, &acquire_in_progress, &error);
L
Linus Torvalds 已提交
847 848
	}

849
found:
L
Linus Torvalds 已提交
850 851
	x = best;
	if (!x && !error && !acquire_in_progress) {
852
		if (tmpl->id.spi &&
853
		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
854
					      tmpl->id.proto, encap_family)) != NULL) {
855
			to_put = x0;
L
Linus Torvalds 已提交
856 857 858
			error = -EEXIST;
			goto out;
		}
859
		x = xfrm_state_alloc(net);
L
Linus Torvalds 已提交
860 861 862 863
		if (x == NULL) {
			error = -ENOMEM;
			goto out;
		}
864
		/* Initialize temporary state matching only
L
Linus Torvalds 已提交
865
		 * to current session. */
866
		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
867
		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
L
Linus Torvalds 已提交
868

869
		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
870 871
		if (error) {
			x->km.state = XFRM_STATE_DEAD;
872
			to_put = x;
873 874 875 876
			x = NULL;
			goto out;
		}

L
Linus Torvalds 已提交
877 878
		if (km_query(x, tmpl, pol) == 0) {
			x->km.state = XFRM_STATE_ACQ;
879 880
			list_add(&x->km.all, &net->xfrm.state_all);
			hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
881
			h = xfrm_src_hash(net, daddr, saddr, encap_family);
882
			hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
L
Linus Torvalds 已提交
883
			if (x->id.spi) {
884
				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
885
				hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
L
Linus Torvalds 已提交
886
			}
A
Alexey Dobriyan 已提交
887
			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
888
			tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
889 890
			net->xfrm.state_num++;
			xfrm_hash_grow_check(net, x->bydst.next != NULL);
L
Linus Torvalds 已提交
891 892
		} else {
			x->km.state = XFRM_STATE_DEAD;
893
			to_put = x;
L
Linus Torvalds 已提交
894 895 896 897 898 899 900 901 902 903
			x = NULL;
			error = -ESRCH;
		}
	}
out:
	if (x)
		xfrm_state_hold(x);
	else
		*err = acquire_in_progress ? -EAGAIN : error;
	spin_unlock_bh(&xfrm_state_lock);
904 905
	if (to_put)
		xfrm_state_put(to_put);
L
Linus Torvalds 已提交
906 907 908
	return x;
}

909
struct xfrm_state *
910
xfrm_stateonly_find(struct net *net, u32 mark,
911
		    xfrm_address_t *daddr, xfrm_address_t *saddr,
912 913
		    unsigned short family, u8 mode, u8 proto, u32 reqid)
{
914
	unsigned int h;
915 916 917 918
	struct xfrm_state *rx = NULL, *x = NULL;
	struct hlist_node *entry;

	spin_lock(&xfrm_state_lock);
919 920
	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
921 922
		if (x->props.family == family &&
		    x->props.reqid == reqid &&
J
Jamal Hadi Salim 已提交
923
		    (mark & x->mark.m) == x->mark.v &&
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
		    xfrm_state_addr_check(x, daddr, saddr, family) &&
		    mode == x->props.mode &&
		    proto == x->id.proto &&
		    x->km.state == XFRM_STATE_VALID) {
			rx = x;
			break;
		}
	}

	if (rx)
		xfrm_state_hold(rx);
	spin_unlock(&xfrm_state_lock);


	return rx;
}
EXPORT_SYMBOL(xfrm_stateonly_find);

L
Linus Torvalds 已提交
943 944
static void __xfrm_state_insert(struct xfrm_state *x)
{
945
	struct net *net = xs_net(x);
946
	unsigned int h;
L
Linus Torvalds 已提交
947

948
	list_add(&x->km.all, &net->xfrm.state_all);
949

950
	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
951
			  x->props.reqid, x->props.family);
952
	hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
L
Linus Torvalds 已提交
953

954 955
	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
	hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
L
Linus Torvalds 已提交
956

957
	if (x->id.spi) {
958
		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
959 960
				  x->props.family);

961
		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
962 963
	}

964
	tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
965 966
	if (x->replay_maxage)
		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
967

968
	wake_up(&net->xfrm.km_waitq);
969

970
	net->xfrm.state_num++;
971

972
	xfrm_hash_grow_check(net, x->bydst.next != NULL);
L
Linus Torvalds 已提交
973 974
}

975 976 977
/* xfrm_state_lock is held */
static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
{
978
	struct net *net = xs_net(xnew);
979 980 981 982 983
	unsigned short family = xnew->props.family;
	u32 reqid = xnew->props.reqid;
	struct xfrm_state *x;
	struct hlist_node *entry;
	unsigned int h;
J
Jamal Hadi Salim 已提交
984
	u32 mark = xnew->mark.v & xnew->mark.m;
985

986 987
	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
988 989
		if (x->props.family	== family &&
		    x->props.reqid	== reqid &&
J
Jamal Hadi Salim 已提交
990
		    (mark & x->mark.m) == x->mark.v &&
991 992
		    !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
		    !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
H
Herbert Xu 已提交
993
			x->genid++;
994 995 996
	}
}

L
Linus Torvalds 已提交
997 998 999
void xfrm_state_insert(struct xfrm_state *x)
{
	spin_lock_bh(&xfrm_state_lock);
1000
	__xfrm_state_bump_genids(x);
L
Linus Torvalds 已提交
1001 1002 1003 1004 1005
	__xfrm_state_insert(x);
	spin_unlock_bh(&xfrm_state_lock);
}
EXPORT_SYMBOL(xfrm_state_insert);

1006
/* xfrm_state_lock is held */
1007 1008 1009 1010 1011
static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
					  unsigned short family, u8 mode,
					  u32 reqid, u8 proto,
					  const xfrm_address_t *daddr,
					  const xfrm_address_t *saddr, int create)
1012
{
1013
	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1014
	struct hlist_node *entry;
1015
	struct xfrm_state *x;
J
Jamal Hadi Salim 已提交
1016
	u32 mark = m->v & m->m;
1017

1018
	hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1019 1020 1021 1022
		if (x->props.reqid  != reqid ||
		    x->props.mode   != mode ||
		    x->props.family != family ||
		    x->km.state     != XFRM_STATE_ACQ ||
1023
		    x->id.spi       != 0 ||
1024
		    x->id.proto	    != proto ||
J
Jamal Hadi Salim 已提交
1025
		    (mark & x->mark.m) != x->mark.v ||
1026 1027
		    xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
		    xfrm_addr_cmp(&x->props.saddr, saddr, family))
1028 1029 1030 1031 1032 1033 1034 1035 1036
			continue;

		xfrm_state_hold(x);
		return x;
	}

	if (!create)
		return NULL;

1037
	x = xfrm_state_alloc(net);
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	if (likely(x)) {
		switch (family) {
		case AF_INET:
			x->sel.daddr.a4 = daddr->a4;
			x->sel.saddr.a4 = saddr->a4;
			x->sel.prefixlen_d = 32;
			x->sel.prefixlen_s = 32;
			x->props.saddr.a4 = saddr->a4;
			x->id.daddr.a4 = daddr->a4;
			break;

		case AF_INET6:
A
Alexey Dobriyan 已提交
1050 1051
			*(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
			*(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1052 1053
			x->sel.prefixlen_d = 128;
			x->sel.prefixlen_s = 128;
A
Alexey Dobriyan 已提交
1054 1055
			*(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
			*(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1056
			break;
1057
		}
1058 1059 1060 1061 1062 1063

		x->km.state = XFRM_STATE_ACQ;
		x->id.proto = proto;
		x->props.family = family;
		x->props.mode = mode;
		x->props.reqid = reqid;
1064 1065
		x->mark.v = m->v;
		x->mark.m = m->m;
A
Alexey Dobriyan 已提交
1066
		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1067
		xfrm_state_hold(x);
1068
		tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1069 1070 1071 1072
		list_add(&x->km.all, &net->xfrm.state_all);
		hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
		h = xfrm_src_hash(net, daddr, saddr, family);
		hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1073

1074
		net->xfrm.state_num++;
1075

1076
		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1077 1078 1079 1080 1081
	}

	return x;
}

1082
static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
L
Linus Torvalds 已提交
1083 1084 1085

int xfrm_state_add(struct xfrm_state *x)
{
1086
	struct net *net = xs_net(x);
1087
	struct xfrm_state *x1, *to_put;
L
Linus Torvalds 已提交
1088 1089
	int family;
	int err;
1090
	u32 mark = x->mark.v & x->mark.m;
1091
	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
L
Linus Torvalds 已提交
1092 1093 1094

	family = x->props.family;

1095 1096
	to_put = NULL;

L
Linus Torvalds 已提交
1097 1098
	spin_lock_bh(&xfrm_state_lock);

1099
	x1 = __xfrm_state_locate(x, use_spi, family);
L
Linus Torvalds 已提交
1100
	if (x1) {
1101
		to_put = x1;
L
Linus Torvalds 已提交
1102 1103 1104 1105 1106
		x1 = NULL;
		err = -EEXIST;
		goto out;
	}

1107
	if (use_spi && x->km.seq) {
1108
		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1109 1110
		if (x1 && ((x1->id.proto != x->id.proto) ||
		    xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1111
			to_put = x1;
L
Linus Torvalds 已提交
1112 1113 1114 1115
			x1 = NULL;
		}
	}

1116
	if (use_spi && !x1)
1117 1118
		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
				     x->props.reqid, x->id.proto,
1119
				     &x->id.daddr, &x->props.saddr, 0);
L
Linus Torvalds 已提交
1120

1121
	__xfrm_state_bump_genids(x);
L
Linus Torvalds 已提交
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
	__xfrm_state_insert(x);
	err = 0;

out:
	spin_unlock_bh(&xfrm_state_lock);

	if (x1) {
		xfrm_state_delete(x1);
		xfrm_state_put(x1);
	}

1133 1134 1135
	if (to_put)
		xfrm_state_put(to_put);

L
Linus Torvalds 已提交
1136 1137 1138 1139
	return err;
}
EXPORT_SYMBOL(xfrm_state_add);

1140
#ifdef CONFIG_XFRM_MIGRATE
1141
static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1142
{
1143
	struct net *net = xs_net(orig);
1144
	int err = -ENOMEM;
1145
	struct xfrm_state *x = xfrm_state_alloc(net);
1146
	if (!x)
H
Herbert Xu 已提交
1147
		goto out;
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158

	memcpy(&x->id, &orig->id, sizeof(x->id));
	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
	x->props.mode = orig->props.mode;
	x->props.replay_window = orig->props.replay_window;
	x->props.reqid = orig->props.reqid;
	x->props.family = orig->props.family;
	x->props.saddr = orig->props.saddr;

	if (orig->aalg) {
1159
		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
		if (!x->aalg)
			goto error;
	}
	x->props.aalgo = orig->props.aalgo;

	if (orig->ealg) {
		x->ealg = xfrm_algo_clone(orig->ealg);
		if (!x->ealg)
			goto error;
	}
	x->props.ealgo = orig->props.ealgo;

	if (orig->calg) {
		x->calg = xfrm_algo_clone(orig->calg);
		if (!x->calg)
			goto error;
	}
	x->props.calgo = orig->props.calgo;

1179
	if (orig->encap) {
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
		x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
		if (!x->encap)
			goto error;
	}

	if (orig->coaddr) {
		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
				    GFP_KERNEL);
		if (!x->coaddr)
			goto error;
	}

1192 1193 1194 1195 1196 1197
	if (orig->replay_esn) {
		err = xfrm_replay_clone(x, orig);
		if (err)
			goto error;
	}

1198 1199
	memcpy(&x->mark, &orig->mark, sizeof(x->mark));

1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
	err = xfrm_init_state(x);
	if (err)
		goto error;

	x->props.flags = orig->props.flags;

	x->curlft.add_time = orig->curlft.add_time;
	x->km.state = orig->km.state;
	x->km.seq = orig->km.seq;

	return x;

 error:
H
Herbert Xu 已提交
1213 1214
	xfrm_state_put(x);
out:
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
	if (errp)
		*errp = err;
	return NULL;
}

/* xfrm_state_lock is held */
struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
{
	unsigned int h;
	struct xfrm_state *x;
	struct hlist_node *entry;

	if (m->reqid) {
1228
		h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1229
				  m->reqid, m->old_family);
1230
		hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
			if (x->props.mode != m->mode ||
			    x->id.proto != m->proto)
				continue;
			if (m->reqid && x->props.reqid != m->reqid)
				continue;
			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
					  m->old_family) ||
			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
					  m->old_family))
				continue;
			xfrm_state_hold(x);
			return x;
		}
	} else {
1245
		h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1246
				  m->old_family);
1247
		hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
			if (x->props.mode != m->mode ||
			    x->id.proto != m->proto)
				continue;
			if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
					  m->old_family) ||
			    xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
					  m->old_family))
				continue;
			xfrm_state_hold(x);
			return x;
		}
	}

1261
	return NULL;
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289
}
EXPORT_SYMBOL(xfrm_migrate_state_find);

struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
				       struct xfrm_migrate *m)
{
	struct xfrm_state *xc;
	int err;

	xc = xfrm_state_clone(x, &err);
	if (!xc)
		return NULL;

	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));

	/* add state */
	if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
		/* a care is needed when the destination address of the
		   state is to be updated as it is a part of triplet */
		xfrm_state_insert(xc);
	} else {
		if ((err = xfrm_state_add(xc)) < 0)
			goto error;
	}

	return xc;
error:
T
Thomas Egerer 已提交
1290
	xfrm_state_put(xc);
1291 1292 1293 1294 1295
	return NULL;
}
EXPORT_SYMBOL(xfrm_state_migrate);
#endif

L
Linus Torvalds 已提交
1296 1297
int xfrm_state_update(struct xfrm_state *x)
{
1298
	struct xfrm_state *x1, *to_put;
L
Linus Torvalds 已提交
1299
	int err;
1300
	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
L
Linus Torvalds 已提交
1301

1302 1303
	to_put = NULL;

L
Linus Torvalds 已提交
1304
	spin_lock_bh(&xfrm_state_lock);
1305
	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
L
Linus Torvalds 已提交
1306 1307 1308 1309 1310 1311

	err = -ESRCH;
	if (!x1)
		goto out;

	if (xfrm_state_kern(x1)) {
1312
		to_put = x1;
L
Linus Torvalds 已提交
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
		err = -EEXIST;
		goto out;
	}

	if (x1->km.state == XFRM_STATE_ACQ) {
		__xfrm_state_insert(x);
		x = NULL;
	}
	err = 0;

out:
	spin_unlock_bh(&xfrm_state_lock);

1326 1327 1328
	if (to_put)
		xfrm_state_put(to_put);

L
Linus Torvalds 已提交
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
	if (err)
		return err;

	if (!x) {
		xfrm_state_delete(x1);
		xfrm_state_put(x1);
		return 0;
	}

	err = -EINVAL;
	spin_lock_bh(&x1->lock);
	if (likely(x1->km.state == XFRM_STATE_VALID)) {
		if (x->encap && x1->encap)
			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1343 1344 1345 1346 1347
		if (x->coaddr && x1->coaddr) {
			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
		}
		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
L
Linus Torvalds 已提交
1348 1349 1350
		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
		x1->km.dying = 0;

1351
		tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
L
Linus Torvalds 已提交
1352 1353 1354 1355
		if (x1->curlft.use_time)
			xfrm_state_check_expire(x1);

		err = 0;
1356 1357
		x->km.state = XFRM_STATE_DEAD;
		__xfrm_state_put(x);
L
Linus Torvalds 已提交
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
	}
	spin_unlock_bh(&x1->lock);

	xfrm_state_put(x1);

	return err;
}
EXPORT_SYMBOL(xfrm_state_update);

int xfrm_state_check_expire(struct xfrm_state *x)
{
	if (!x->curlft.use_time)
1370
		x->curlft.use_time = get_seconds();
L
Linus Torvalds 已提交
1371 1372 1373

	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
	    x->curlft.packets >= x->lft.hard_packet_limit) {
1374
		x->km.state = XFRM_STATE_EXPIRED;
1375
		tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
L
Linus Torvalds 已提交
1376 1377 1378 1379 1380
		return -EINVAL;
	}

	if (!x->km.dying &&
	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1381 1382
	     x->curlft.packets >= x->lft.soft_packet_limit)) {
		x->km.dying = 1;
1383
		km_state_expired(x, 0, 0);
1384
	}
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389
	return 0;
}
EXPORT_SYMBOL(xfrm_state_check_expire);

struct xfrm_state *
1390
xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1391
		  u8 proto, unsigned short family)
L
Linus Torvalds 已提交
1392 1393 1394 1395
{
	struct xfrm_state *x;

	spin_lock_bh(&xfrm_state_lock);
1396
	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
L
Linus Torvalds 已提交
1397 1398 1399 1400 1401 1402
	spin_unlock_bh(&xfrm_state_lock);
	return x;
}
EXPORT_SYMBOL(xfrm_state_lookup);

struct xfrm_state *
1403
xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1404
			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1405 1406 1407 1408 1409
			 u8 proto, unsigned short family)
{
	struct xfrm_state *x;

	spin_lock_bh(&xfrm_state_lock);
1410
	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1411 1412 1413 1414 1415 1416
	spin_unlock_bh(&xfrm_state_lock);
	return x;
}
EXPORT_SYMBOL(xfrm_state_lookup_byaddr);

struct xfrm_state *
1417
xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1418
	      const xfrm_address_t *daddr, const xfrm_address_t *saddr,
L
Linus Torvalds 已提交
1419 1420 1421 1422 1423
	      int create, unsigned short family)
{
	struct xfrm_state *x;

	spin_lock_bh(&xfrm_state_lock);
1424
	x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
L
Linus Torvalds 已提交
1425
	spin_unlock_bh(&xfrm_state_lock);
1426

L
Linus Torvalds 已提交
1427 1428 1429 1430
	return x;
}
EXPORT_SYMBOL(xfrm_find_acq);

1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
#ifdef CONFIG_XFRM_SUB_POLICY
int
xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
	       unsigned short family)
{
	int err = 0;
	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
	if (!afinfo)
		return -EAFNOSUPPORT;

	spin_lock_bh(&xfrm_state_lock);
	if (afinfo->tmpl_sort)
		err = afinfo->tmpl_sort(dst, src, n);
	spin_unlock_bh(&xfrm_state_lock);
	xfrm_state_put_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_tmpl_sort);

int
xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
		unsigned short family)
{
	int err = 0;
	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
	if (!afinfo)
		return -EAFNOSUPPORT;

	spin_lock_bh(&xfrm_state_lock);
	if (afinfo->state_sort)
		err = afinfo->state_sort(dst, src, n);
	spin_unlock_bh(&xfrm_state_lock);
	xfrm_state_put_afinfo(afinfo);
	return err;
}
EXPORT_SYMBOL(xfrm_state_sort);
#endif

L
Linus Torvalds 已提交
1469 1470
/* Silly enough, but I'm lazy to build resolution list */

1471
static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
L
Linus Torvalds 已提交
1472 1473 1474
{
	int i;

1475
	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1476 1477 1478
		struct hlist_node *entry;
		struct xfrm_state *x;

1479
		hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1480
			if (x->km.seq == seq &&
J
Jamal Hadi Salim 已提交
1481
			    (mark & x->mark.m) == x->mark.v &&
1482
			    x->km.state == XFRM_STATE_ACQ) {
L
Linus Torvalds 已提交
1483 1484 1485 1486 1487 1488 1489 1490
				xfrm_state_hold(x);
				return x;
			}
		}
	}
	return NULL;
}

1491
struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
L
Linus Torvalds 已提交
1492 1493 1494 1495
{
	struct xfrm_state *x;

	spin_lock_bh(&xfrm_state_lock);
1496
	x = __xfrm_find_acq_byseq(net, mark, seq);
L
Linus Torvalds 已提交
1497 1498 1499 1500 1501 1502 1503 1504
	spin_unlock_bh(&xfrm_state_lock);
	return x;
}
EXPORT_SYMBOL(xfrm_find_acq_byseq);

u32 xfrm_get_acqseq(void)
{
	u32 res;
1505 1506 1507 1508 1509
	static atomic_t acqseq;

	do {
		res = atomic_inc_return(&acqseq);
	} while (!res);
L
Linus Torvalds 已提交
1510 1511 1512 1513 1514

	return res;
}
EXPORT_SYMBOL(xfrm_get_acqseq);

1515
int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
L
Linus Torvalds 已提交
1516
{
1517
	struct net *net = xs_net(x);
1518
	unsigned int h;
L
Linus Torvalds 已提交
1519
	struct xfrm_state *x0;
1520 1521 1522
	int err = -ENOENT;
	__be32 minspi = htonl(low);
	__be32 maxspi = htonl(high);
1523
	u32 mark = x->mark.v & x->mark.m;
L
Linus Torvalds 已提交
1524

1525 1526 1527 1528 1529
	spin_lock_bh(&x->lock);
	if (x->km.state == XFRM_STATE_DEAD)
		goto unlock;

	err = 0;
L
Linus Torvalds 已提交
1530
	if (x->id.spi)
1531 1532 1533
		goto unlock;

	err = -ENOENT;
L
Linus Torvalds 已提交
1534 1535

	if (minspi == maxspi) {
1536
		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
L
Linus Torvalds 已提交
1537 1538
		if (x0) {
			xfrm_state_put(x0);
1539
			goto unlock;
L
Linus Torvalds 已提交
1540 1541 1542 1543
		}
		x->id.spi = minspi;
	} else {
		u32 spi = 0;
A
Al Viro 已提交
1544 1545
		for (h=0; h<high-low+1; h++) {
			spi = low + net_random()%(high-low+1);
1546
			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
L
Linus Torvalds 已提交
1547 1548 1549 1550 1551 1552 1553 1554 1555
			if (x0 == NULL) {
				x->id.spi = htonl(spi);
				break;
			}
			xfrm_state_put(x0);
		}
	}
	if (x->id.spi) {
		spin_lock_bh(&xfrm_state_lock);
1556 1557
		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
		hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
L
Linus Torvalds 已提交
1558
		spin_unlock_bh(&xfrm_state_lock);
1559 1560

		err = 0;
L
Linus Torvalds 已提交
1561
	}
1562 1563 1564 1565 1566

unlock:
	spin_unlock_bh(&x->lock);

	return err;
L
Linus Torvalds 已提交
1567 1568 1569
}
EXPORT_SYMBOL(xfrm_alloc_spi);

1570
int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1571
		    int (*func)(struct xfrm_state *, int, void*),
L
Linus Torvalds 已提交
1572 1573
		    void *data)
{
H
Herbert Xu 已提交
1574 1575
	struct xfrm_state *state;
	struct xfrm_state_walk *x;
L
Linus Torvalds 已提交
1576 1577
	int err = 0;

H
Herbert Xu 已提交
1578
	if (walk->seq != 0 && list_empty(&walk->all))
1579 1580
		return 0;

L
Linus Torvalds 已提交
1581
	spin_lock_bh(&xfrm_state_lock);
H
Herbert Xu 已提交
1582
	if (list_empty(&walk->all))
1583
		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
H
Herbert Xu 已提交
1584 1585
	else
		x = list_entry(&walk->all, struct xfrm_state_walk, all);
1586
	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
H
Herbert Xu 已提交
1587
		if (x->state == XFRM_STATE_DEAD)
1588
			continue;
H
Herbert Xu 已提交
1589 1590
		state = container_of(x, struct xfrm_state, km);
		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1591
			continue;
H
Herbert Xu 已提交
1592 1593 1594 1595
		err = func(state, walk->seq, data);
		if (err) {
			list_move_tail(&walk->all, &x->all);
			goto out;
L
Linus Torvalds 已提交
1596
		}
H
Herbert Xu 已提交
1597
		walk->seq++;
L
Linus Torvalds 已提交
1598
	}
H
Herbert Xu 已提交
1599
	if (walk->seq == 0) {
L
Linus Torvalds 已提交
1600 1601 1602
		err = -ENOENT;
		goto out;
	}
H
Herbert Xu 已提交
1603
	list_del_init(&walk->all);
L
Linus Torvalds 已提交
1604 1605 1606 1607 1608 1609
out:
	spin_unlock_bh(&xfrm_state_lock);
	return err;
}
EXPORT_SYMBOL(xfrm_state_walk);

H
Herbert Xu 已提交
1610 1611
void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
{
H
Herbert Xu 已提交
1612
	INIT_LIST_HEAD(&walk->all);
H
Herbert Xu 已提交
1613
	walk->proto = proto;
H
Herbert Xu 已提交
1614 1615
	walk->state = XFRM_STATE_DEAD;
	walk->seq = 0;
H
Herbert Xu 已提交
1616 1617 1618
}
EXPORT_SYMBOL(xfrm_state_walk_init);

1619 1620
void xfrm_state_walk_done(struct xfrm_state_walk *walk)
{
H
Herbert Xu 已提交
1621
	if (list_empty(&walk->all))
H
Herbert Xu 已提交
1622 1623
		return;

H
Herbert Xu 已提交
1624 1625
	spin_lock_bh(&xfrm_state_lock);
	list_del(&walk->all);
1626
	spin_unlock_bh(&xfrm_state_lock);
1627 1628 1629
}
EXPORT_SYMBOL(xfrm_state_walk_done);

1630 1631 1632 1633 1634 1635
static void xfrm_replay_timer_handler(unsigned long data)
{
	struct xfrm_state *x = (struct xfrm_state*)data;

	spin_lock(&x->lock);

J
Jamal Hadi Salim 已提交
1636
	if (x->km.state == XFRM_STATE_VALID) {
1637
		if (xfrm_aevent_is_on(xs_net(x)))
1638
			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
J
Jamal Hadi Salim 已提交
1639 1640 1641
		else
			x->xflags |= XFRM_TIME_DEFER;
	}
1642 1643 1644 1645

	spin_unlock(&x->lock);
}

1646
static LIST_HEAD(xfrm_km_list);
L
Linus Torvalds 已提交
1647

1648
void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
L
Linus Torvalds 已提交
1649 1650 1651
{
	struct xfrm_mgr *km;

1652 1653
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1654 1655
		if (km->notify_policy)
			km->notify_policy(xp, dir, c);
1656
	rcu_read_unlock();
1657
}
L
Linus Torvalds 已提交
1658

1659
void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1660 1661
{
	struct xfrm_mgr *km;
1662 1663
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list)
1664 1665
		if (km->notify)
			km->notify(x, c);
1666
	rcu_read_unlock();
1667 1668 1669 1670 1671
}

EXPORT_SYMBOL(km_policy_notify);
EXPORT_SYMBOL(km_state_notify);

1672
void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1673
{
1674
	struct net *net = xs_net(x);
1675 1676
	struct km_event c;

1677
	c.data.hard = hard;
1678
	c.portid = portid;
1679
	c.event = XFRM_MSG_EXPIRE;
1680
	km_state_notify(x, &c);
L
Linus Torvalds 已提交
1681 1682

	if (hard)
1683
		wake_up(&net->xfrm.km_waitq);
L
Linus Torvalds 已提交
1684 1685
}

1686
EXPORT_SYMBOL(km_state_expired);
1687 1688 1689 1690
/*
 * We send to all registered managers regardless of failure
 * We are happy with one success
*/
1691
int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
L
Linus Torvalds 已提交
1692
{
1693
	int err = -EINVAL, acqret;
L
Linus Torvalds 已提交
1694 1695
	struct xfrm_mgr *km;

1696 1697
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1698
		acqret = km->acquire(x, t, pol);
1699 1700
		if (!acqret)
			err = acqret;
L
Linus Torvalds 已提交
1701
	}
1702
	rcu_read_unlock();
L
Linus Torvalds 已提交
1703 1704
	return err;
}
1705
EXPORT_SYMBOL(km_query);
L
Linus Torvalds 已提交
1706

A
Al Viro 已提交
1707
int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
L
Linus Torvalds 已提交
1708 1709 1710 1711
{
	int err = -EINVAL;
	struct xfrm_mgr *km;

1712 1713
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
L
Linus Torvalds 已提交
1714 1715 1716 1717 1718
		if (km->new_mapping)
			err = km->new_mapping(x, ipaddr, sport);
		if (!err)
			break;
	}
1719
	rcu_read_unlock();
L
Linus Torvalds 已提交
1720 1721 1722 1723
	return err;
}
EXPORT_SYMBOL(km_new_mapping);

1724
void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
L
Linus Torvalds 已提交
1725
{
1726
	struct net *net = xp_net(pol);
1727
	struct km_event c;
L
Linus Torvalds 已提交
1728

1729
	c.data.hard = hard;
1730
	c.portid = portid;
1731
	c.event = XFRM_MSG_POLEXPIRE;
1732
	km_policy_notify(pol, dir, &c);
L
Linus Torvalds 已提交
1733 1734

	if (hard)
1735
		wake_up(&net->xfrm.km_waitq);
L
Linus Torvalds 已提交
1736
}
1737
EXPORT_SYMBOL(km_policy_expired);
L
Linus Torvalds 已提交
1738

1739
#ifdef CONFIG_XFRM_MIGRATE
1740 1741 1742
int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
	       const struct xfrm_migrate *m, int num_migrate,
	       const struct xfrm_kmaddress *k)
1743 1744 1745 1746 1747
{
	int err = -EINVAL;
	int ret;
	struct xfrm_mgr *km;

1748 1749
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1750
		if (km->migrate) {
1751
			ret = km->migrate(sel, dir, type, m, num_migrate, k);
1752 1753 1754 1755
			if (!ret)
				err = ret;
		}
	}
1756
	rcu_read_unlock();
1757 1758 1759
	return err;
}
EXPORT_SYMBOL(km_migrate);
1760
#endif
1761

1762
int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1763 1764 1765 1766 1767
{
	int err = -EINVAL;
	int ret;
	struct xfrm_mgr *km;

1768 1769
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1770
		if (km->report) {
1771
			ret = km->report(net, proto, sel, addr);
1772 1773 1774 1775
			if (!ret)
				err = ret;
		}
	}
1776
	rcu_read_unlock();
1777 1778 1779 1780
	return err;
}
EXPORT_SYMBOL(km_report);

L
Linus Torvalds 已提交
1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799
int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
{
	int err;
	u8 *data;
	struct xfrm_mgr *km;
	struct xfrm_policy *pol = NULL;

	if (optlen <= 0 || optlen > PAGE_SIZE)
		return -EMSGSIZE;

	data = kmalloc(optlen, GFP_KERNEL);
	if (!data)
		return -ENOMEM;

	err = -EFAULT;
	if (copy_from_user(data, optval, optlen))
		goto out;

	err = -EINVAL;
1800 1801
	rcu_read_lock();
	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1802
		pol = km->compile_policy(sk, optname, data,
L
Linus Torvalds 已提交
1803 1804 1805 1806
					 optlen, &err);
		if (err >= 0)
			break;
	}
1807
	rcu_read_unlock();
L
Linus Torvalds 已提交
1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820

	if (err >= 0) {
		xfrm_sk_policy_insert(sk, err, pol);
		xfrm_pol_put(pol);
		err = 0;
	}

out:
	kfree(data);
	return err;
}
EXPORT_SYMBOL(xfrm_user_policy);

1821 1822
static DEFINE_SPINLOCK(xfrm_km_lock);

L
Linus Torvalds 已提交
1823 1824
int xfrm_register_km(struct xfrm_mgr *km)
{
1825 1826 1827
	spin_lock_bh(&xfrm_km_lock);
	list_add_tail_rcu(&km->list, &xfrm_km_list);
	spin_unlock_bh(&xfrm_km_lock);
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833
	return 0;
}
EXPORT_SYMBOL(xfrm_register_km);

int xfrm_unregister_km(struct xfrm_mgr *km)
{
1834 1835 1836 1837
	spin_lock_bh(&xfrm_km_lock);
	list_del_rcu(&km->list);
	spin_unlock_bh(&xfrm_km_lock);
	synchronize_rcu();
L
Linus Torvalds 已提交
1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
	return 0;
}
EXPORT_SYMBOL(xfrm_unregister_km);

int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
{
	int err = 0;
	if (unlikely(afinfo == NULL))
		return -EINVAL;
	if (unlikely(afinfo->family >= NPROTO))
		return -EAFNOSUPPORT;
1849
	spin_lock_bh(&xfrm_state_afinfo_lock);
L
Linus Torvalds 已提交
1850 1851
	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
		err = -ENOBUFS;
1852
	else
1853 1854
		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
	spin_unlock_bh(&xfrm_state_afinfo_lock);
L
Linus Torvalds 已提交
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865
	return err;
}
EXPORT_SYMBOL(xfrm_state_register_afinfo);

int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
{
	int err = 0;
	if (unlikely(afinfo == NULL))
		return -EINVAL;
	if (unlikely(afinfo->family >= NPROTO))
		return -EAFNOSUPPORT;
1866
	spin_lock_bh(&xfrm_state_afinfo_lock);
L
Linus Torvalds 已提交
1867 1868 1869
	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
		if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
			err = -EINVAL;
1870
		else
1871
			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
L
Linus Torvalds 已提交
1872
	}
1873 1874
	spin_unlock_bh(&xfrm_state_afinfo_lock);
	synchronize_rcu();
L
Linus Torvalds 已提交
1875 1876 1877 1878
	return err;
}
EXPORT_SYMBOL(xfrm_state_unregister_afinfo);

1879
static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
L
Linus Torvalds 已提交
1880 1881 1882 1883
{
	struct xfrm_state_afinfo *afinfo;
	if (unlikely(family >= NPROTO))
		return NULL;
1884 1885
	rcu_read_lock();
	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1886
	if (unlikely(!afinfo))
1887
		rcu_read_unlock();
L
Linus Torvalds 已提交
1888 1889 1890
	return afinfo;
}

1891
static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
L
Linus Torvalds 已提交
1892
{
1893
	rcu_read_unlock();
L
Linus Torvalds 已提交
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
}

/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
void xfrm_state_delete_tunnel(struct xfrm_state *x)
{
	if (x->tunnel) {
		struct xfrm_state *t = x->tunnel;

		if (atomic_read(&t->tunnel_users) == 2)
			xfrm_state_delete(t);
		atomic_dec(&t->tunnel_users);
		xfrm_state_put(t);
		x->tunnel = NULL;
	}
}
EXPORT_SYMBOL(xfrm_state_delete_tunnel);

int xfrm_state_mtu(struct xfrm_state *x, int mtu)
{
1913
	int res;
L
Linus Torvalds 已提交
1914

1915 1916 1917 1918 1919
	spin_lock_bh(&x->lock);
	if (x->km.state == XFRM_STATE_VALID &&
	    x->type && x->type->get_mtu)
		res = x->type->get_mtu(x, mtu);
	else
1920
		res = mtu - x->props.header_len;
1921
	spin_unlock_bh(&x->lock);
L
Linus Torvalds 已提交
1922 1923 1924
	return res;
}

1925
int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
H
Herbert Xu 已提交
1926
{
1927
	struct xfrm_state_afinfo *afinfo;
1928
	struct xfrm_mode *inner_mode;
1929
	int family = x->props.family;
H
Herbert Xu 已提交
1930 1931
	int err;

1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
	err = -EAFNOSUPPORT;
	afinfo = xfrm_state_get_afinfo(family);
	if (!afinfo)
		goto error;

	err = 0;
	if (afinfo->init_flags)
		err = afinfo->init_flags(x);

	xfrm_state_put_afinfo(afinfo);

	if (err)
		goto error;

	err = -EPROTONOSUPPORT;
1947

1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
	if (x->sel.family != AF_UNSPEC) {
		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
		if (inner_mode == NULL)
			goto error;

		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
		    family != x->sel.family) {
			xfrm_put_mode(inner_mode);
			goto error;
		}

		x->inner_mode = inner_mode;
	} else {
		struct xfrm_mode *inner_mode_iaf;
1962
		int iafamily = AF_INET;
1963

1964
		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1965 1966 1967 1968 1969 1970 1971
		if (inner_mode == NULL)
			goto error;

		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
			xfrm_put_mode(inner_mode);
			goto error;
		}
1972
		x->inner_mode = inner_mode;
1973

1974 1975
		if (x->props.family == AF_INET)
			iafamily = AF_INET6;
1976

1977 1978 1979 1980 1981 1982
		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
		if (inner_mode_iaf) {
			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
				x->inner_mode_iaf = inner_mode_iaf;
			else
				xfrm_put_mode(inner_mode_iaf);
1983 1984
		}
	}
1985

1986
	x->type = xfrm_get_type(x->id.proto, family);
H
Herbert Xu 已提交
1987 1988 1989 1990 1991 1992 1993
	if (x->type == NULL)
		goto error;

	err = x->type->init_state(x);
	if (err)
		goto error;

1994
	x->outer_mode = xfrm_get_mode(x->props.mode, family);
1995 1996
	if (x->outer_mode == NULL) {
		err = -EPROTONOSUPPORT;
1997
		goto error;
1998
	}
1999

2000 2001 2002 2003 2004 2005
	if (init_replay) {
		err = xfrm_init_replay(x);
		if (err)
			goto error;
	}

H
Herbert Xu 已提交
2006 2007 2008 2009 2010 2011
	x->km.state = XFRM_STATE_VALID;

error:
	return err;
}

2012 2013 2014 2015 2016 2017 2018
EXPORT_SYMBOL(__xfrm_init_state);

int xfrm_init_state(struct xfrm_state *x)
{
	return __xfrm_init_state(x, true);
}

H
Herbert Xu 已提交
2019
EXPORT_SYMBOL(xfrm_init_state);
2020

2021
int __net_init xfrm_state_init(struct net *net)
L
Linus Torvalds 已提交
2022
{
2023 2024
	unsigned int sz;

2025 2026
	INIT_LIST_HEAD(&net->xfrm.state_all);

2027 2028
	sz = sizeof(struct hlist_head) * 8;

2029 2030 2031
	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
	if (!net->xfrm.state_bydst)
		goto out_bydst;
2032 2033 2034
	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
	if (!net->xfrm.state_bysrc)
		goto out_bysrc;
2035 2036 2037
	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
	if (!net->xfrm.state_byspi)
		goto out_byspi;
2038
	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
L
Linus Torvalds 已提交
2039

2040
	net->xfrm.state_num = 0;
2041
	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2042
	INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2043
	INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
A
Alexey Dobriyan 已提交
2044
	init_waitqueue_head(&net->xfrm.km_waitq);
2045
	return 0;
2046

2047 2048
out_byspi:
	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2049 2050
out_bysrc:
	xfrm_hash_free(net->xfrm.state_bydst, sz);
2051 2052
out_bydst:
	return -ENOMEM;
2053 2054 2055 2056
}

void xfrm_state_fini(struct net *net)
{
2057
	struct xfrm_audit audit_info;
2058 2059
	unsigned int sz;

2060
	flush_work(&net->xfrm.state_hash_work);
2061
	audit_info.loginuid = INVALID_UID;
2062 2063 2064 2065 2066
	audit_info.sessionid = -1;
	audit_info.secid = 0;
	xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
	flush_work(&net->xfrm.state_gc_work);

2067
	WARN_ON(!list_empty(&net->xfrm.state_all));
2068

2069
	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2070 2071
	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
	xfrm_hash_free(net->xfrm.state_byspi, sz);
2072 2073
	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2074 2075
	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
	xfrm_hash_free(net->xfrm.state_bydst, sz);
L
Linus Torvalds 已提交
2076 2077
}

J
Joy Latten 已提交
2078
#ifdef CONFIG_AUDITSYSCALL
I
Ilpo Järvinen 已提交
2079 2080
static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
				     struct audit_buffer *audit_buf)
J
Joy Latten 已提交
2081
{
P
Paul Moore 已提交
2082 2083 2084 2085
	struct xfrm_sec_ctx *ctx = x->security;
	u32 spi = ntohl(x->id.spi);

	if (ctx)
J
Joy Latten 已提交
2086
		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
P
Paul Moore 已提交
2087
				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
J
Joy Latten 已提交
2088 2089 2090

	switch(x->props.family) {
	case AF_INET:
H
Harvey Harrison 已提交
2091 2092
		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
				 &x->props.saddr.a4, &x->id.daddr.a4);
J
Joy Latten 已提交
2093 2094
		break;
	case AF_INET6:
H
Harvey Harrison 已提交
2095
		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2096
				 x->props.saddr.a6, x->id.daddr.a6);
J
Joy Latten 已提交
2097 2098
		break;
	}
P
Paul Moore 已提交
2099 2100

	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
J
Joy Latten 已提交
2101 2102
}

I
Ilpo Järvinen 已提交
2103 2104
static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
				      struct audit_buffer *audit_buf)
P
Paul Moore 已提交
2105
{
2106 2107
	const struct iphdr *iph4;
	const struct ipv6hdr *iph6;
P
Paul Moore 已提交
2108 2109 2110 2111

	switch (family) {
	case AF_INET:
		iph4 = ip_hdr(skb);
H
Harvey Harrison 已提交
2112 2113
		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
				 &iph4->saddr, &iph4->daddr);
P
Paul Moore 已提交
2114 2115 2116 2117
		break;
	case AF_INET6:
		iph6 = ipv6_hdr(skb);
		audit_log_format(audit_buf,
H
Harvey Harrison 已提交
2118
				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2119
				 &iph6->saddr,&iph6->daddr,
P
Paul Moore 已提交
2120 2121 2122 2123 2124 2125 2126
				 iph6->flow_lbl[0] & 0x0f,
				 iph6->flow_lbl[1],
				 iph6->flow_lbl[2]);
		break;
	}
}

P
Paul Moore 已提交
2127
void xfrm_audit_state_add(struct xfrm_state *x, int result,
2128
			  kuid_t auid, u32 sessionid, u32 secid)
J
Joy Latten 已提交
2129 2130 2131
{
	struct audit_buffer *audit_buf;

P
Paul Moore 已提交
2132
	audit_buf = xfrm_audit_start("SAD-add");
J
Joy Latten 已提交
2133 2134
	if (audit_buf == NULL)
		return;
2135
	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
P
Paul Moore 已提交
2136 2137
	xfrm_audit_helper_sainfo(x, audit_buf);
	audit_log_format(audit_buf, " res=%u", result);
J
Joy Latten 已提交
2138 2139 2140 2141
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_add);

P
Paul Moore 已提交
2142
void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2143
			     kuid_t auid, u32 sessionid, u32 secid)
J
Joy Latten 已提交
2144 2145 2146
{
	struct audit_buffer *audit_buf;

P
Paul Moore 已提交
2147
	audit_buf = xfrm_audit_start("SAD-delete");
J
Joy Latten 已提交
2148 2149
	if (audit_buf == NULL)
		return;
2150
	xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
P
Paul Moore 已提交
2151 2152
	xfrm_audit_helper_sainfo(x, audit_buf);
	audit_log_format(audit_buf, " res=%u", result);
J
Joy Latten 已提交
2153 2154 2155
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
P
Paul Moore 已提交
2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174

void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
				      struct sk_buff *skb)
{
	struct audit_buffer *audit_buf;
	u32 spi;

	audit_buf = xfrm_audit_start("SA-replay-overflow");
	if (audit_buf == NULL)
		return;
	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
	/* don't record the sequence number because it's inherent in this kind
	 * of audit message */
	spi = ntohl(x->id.spi);
	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);

2175
void xfrm_audit_state_replay(struct xfrm_state *x,
P
Paul Moore 已提交
2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
			     struct sk_buff *skb, __be32 net_seq)
{
	struct audit_buffer *audit_buf;
	u32 spi;

	audit_buf = xfrm_audit_start("SA-replayed-pkt");
	if (audit_buf == NULL)
		return;
	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
	spi = ntohl(x->id.spi);
	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
			 spi, spi, ntohl(net_seq));
	audit_log_end(audit_buf);
}
2190
EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
P
Paul Moore 已提交
2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239

void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
{
	struct audit_buffer *audit_buf;

	audit_buf = xfrm_audit_start("SA-notfound");
	if (audit_buf == NULL)
		return;
	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);

void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
			       __be32 net_spi, __be32 net_seq)
{
	struct audit_buffer *audit_buf;
	u32 spi;

	audit_buf = xfrm_audit_start("SA-notfound");
	if (audit_buf == NULL)
		return;
	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
	spi = ntohl(net_spi);
	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
			 spi, spi, ntohl(net_seq));
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);

void xfrm_audit_state_icvfail(struct xfrm_state *x,
			      struct sk_buff *skb, u8 proto)
{
	struct audit_buffer *audit_buf;
	__be32 net_spi;
	__be32 net_seq;

	audit_buf = xfrm_audit_start("SA-icv-failure");
	if (audit_buf == NULL)
		return;
	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
		u32 spi = ntohl(net_spi);
		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
				 spi, spi, ntohl(net_seq));
	}
	audit_log_end(audit_buf);
}
EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
J
Joy Latten 已提交
2240
#endif /* CONFIG_AUDITSYSCALL */