xref: /linux/net/xfrm/xfrm_state.c (revision 8e3ed5440b0c305dcd1d5fa7419bd8066d22ef42)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * xfrm_state.c
4  *
5  * Changes:
6  *	Mitsuru KANDA @USAGI
7  * 	Kazunori MIYAZAWA @USAGI
8  * 	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9  * 		IPv6 support
10  * 	YOSHIFUJI Hideaki @USAGI
11  * 		Split up af-specific functions
12  *	Derek Atkins <derek@ihtfp.com>
13  *		Add UDP Encapsulation
14  *
15  */
16 
17 #include <linux/compat.h>
18 #include <linux/workqueue.h>
19 #include <net/xfrm.h>
20 #include <linux/pfkeyv2.h>
21 #include <linux/ipsec.h>
22 #include <linux/module.h>
23 #include <linux/cache.h>
24 #include <linux/audit.h>
25 #include <linux/uaccess.h>
26 #include <linux/ktime.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/kernel.h>
30 
31 #include <crypto/aead.h>
32 
33 #include "xfrm_hash.h"
34 
35 #define xfrm_state_deref_prot(table, net) \
36 	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
37 
38 static void xfrm_state_gc_task(struct work_struct *work);
39 
40 /* Each xfrm_state may be linked to two tables:
41 
42    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
43    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
44       destination/tunnel endpoint. (output)
45  */
46 
47 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
48 static struct kmem_cache *xfrm_state_cache __ro_after_init;
49 
50 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
51 static HLIST_HEAD(xfrm_state_gc_list);
52 
53 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54 {
55 	return refcount_inc_not_zero(&x->refcnt);
56 }
57 
58 static inline unsigned int xfrm_dst_hash(struct net *net,
59 					 const xfrm_address_t *daddr,
60 					 const xfrm_address_t *saddr,
61 					 u32 reqid,
62 					 unsigned short family)
63 {
64 	return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
65 }
66 
67 static inline unsigned int xfrm_src_hash(struct net *net,
68 					 const xfrm_address_t *daddr,
69 					 const xfrm_address_t *saddr,
70 					 unsigned short family)
71 {
72 	return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
73 }
74 
75 static inline unsigned int
76 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
77 	      __be32 spi, u8 proto, unsigned short family)
78 {
79 	return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
80 }
81 
82 static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
83 {
84 	return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
85 }
86 
87 #define XFRM_STATE_INSERT(by, _n, _h, _type)                               \
88 	{                                                                  \
89 		struct xfrm_state *_x = NULL;                              \
90 									   \
91 		if (_type != XFRM_DEV_OFFLOAD_PACKET) {                    \
92 			hlist_for_each_entry_rcu(_x, _h, by) {             \
93 				if (_x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
94 					continue;                          \
95 				break;                                     \
96 			}                                                  \
97 		}                                                          \
98 									   \
99 		if (!_x || _x->xso.type == XFRM_DEV_OFFLOAD_PACKET)        \
100 			/* SAD is empty or consist from HW SAs only */     \
101 			hlist_add_head_rcu(_n, _h);                        \
102 		else                                                       \
103 			hlist_add_before_rcu(_n, &_x->by);                 \
104 	}
105 
106 static void xfrm_hash_transfer(struct hlist_head *list,
107 			       struct hlist_head *ndsttable,
108 			       struct hlist_head *nsrctable,
109 			       struct hlist_head *nspitable,
110 			       struct hlist_head *nseqtable,
111 			       unsigned int nhashmask)
112 {
113 	struct hlist_node *tmp;
114 	struct xfrm_state *x;
115 
116 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
117 		unsigned int h;
118 
119 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
120 				    x->props.reqid, x->props.family,
121 				    nhashmask);
122 		XFRM_STATE_INSERT(bydst, &x->bydst, ndsttable + h, x->xso.type);
123 
124 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
125 				    x->props.family,
126 				    nhashmask);
127 		XFRM_STATE_INSERT(bysrc, &x->bysrc, nsrctable + h, x->xso.type);
128 
129 		if (x->id.spi) {
130 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
131 					    x->id.proto, x->props.family,
132 					    nhashmask);
133 			XFRM_STATE_INSERT(byspi, &x->byspi, nspitable + h,
134 					  x->xso.type);
135 		}
136 
137 		if (x->km.seq) {
138 			h = __xfrm_seq_hash(x->km.seq, nhashmask);
139 			XFRM_STATE_INSERT(byseq, &x->byseq, nseqtable + h,
140 					  x->xso.type);
141 		}
142 	}
143 }
144 
145 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
146 {
147 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
148 }
149 
150 static void xfrm_hash_resize(struct work_struct *work)
151 {
152 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
153 	struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
154 	unsigned long nsize, osize;
155 	unsigned int nhashmask, ohashmask;
156 	int i;
157 
158 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
159 	ndst = xfrm_hash_alloc(nsize);
160 	if (!ndst)
161 		return;
162 	nsrc = xfrm_hash_alloc(nsize);
163 	if (!nsrc) {
164 		xfrm_hash_free(ndst, nsize);
165 		return;
166 	}
167 	nspi = xfrm_hash_alloc(nsize);
168 	if (!nspi) {
169 		xfrm_hash_free(ndst, nsize);
170 		xfrm_hash_free(nsrc, nsize);
171 		return;
172 	}
173 	nseq = xfrm_hash_alloc(nsize);
174 	if (!nseq) {
175 		xfrm_hash_free(ndst, nsize);
176 		xfrm_hash_free(nsrc, nsize);
177 		xfrm_hash_free(nspi, nsize);
178 		return;
179 	}
180 
181 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
182 	write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
183 
184 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
185 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
186 	for (i = net->xfrm.state_hmask; i >= 0; i--)
187 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
188 
189 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
190 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
191 	oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
192 	ohashmask = net->xfrm.state_hmask;
193 
194 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
195 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
196 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
197 	rcu_assign_pointer(net->xfrm.state_byseq, nseq);
198 	net->xfrm.state_hmask = nhashmask;
199 
200 	write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
201 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
202 
203 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
204 
205 	synchronize_rcu();
206 
207 	xfrm_hash_free(odst, osize);
208 	xfrm_hash_free(osrc, osize);
209 	xfrm_hash_free(ospi, osize);
210 	xfrm_hash_free(oseq, osize);
211 }
212 
213 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
214 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
215 
216 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
217 
218 int __xfrm_state_delete(struct xfrm_state *x);
219 
220 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
221 static bool km_is_alive(const struct km_event *c);
222 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
223 
224 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
225 {
226 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
227 	int err = 0;
228 
229 	if (!afinfo)
230 		return -EAFNOSUPPORT;
231 
232 #define X(afi, T, name) do {			\
233 		WARN_ON((afi)->type_ ## name);	\
234 		(afi)->type_ ## name = (T);	\
235 	} while (0)
236 
237 	switch (type->proto) {
238 	case IPPROTO_COMP:
239 		X(afinfo, type, comp);
240 		break;
241 	case IPPROTO_AH:
242 		X(afinfo, type, ah);
243 		break;
244 	case IPPROTO_ESP:
245 		X(afinfo, type, esp);
246 		break;
247 	case IPPROTO_IPIP:
248 		X(afinfo, type, ipip);
249 		break;
250 	case IPPROTO_DSTOPTS:
251 		X(afinfo, type, dstopts);
252 		break;
253 	case IPPROTO_ROUTING:
254 		X(afinfo, type, routing);
255 		break;
256 	case IPPROTO_IPV6:
257 		X(afinfo, type, ipip6);
258 		break;
259 	default:
260 		WARN_ON(1);
261 		err = -EPROTONOSUPPORT;
262 		break;
263 	}
264 #undef X
265 	rcu_read_unlock();
266 	return err;
267 }
268 EXPORT_SYMBOL(xfrm_register_type);
269 
270 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
271 {
272 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
273 
274 	if (unlikely(afinfo == NULL))
275 		return;
276 
277 #define X(afi, T, name) do {				\
278 		WARN_ON((afi)->type_ ## name != (T));	\
279 		(afi)->type_ ## name = NULL;		\
280 	} while (0)
281 
282 	switch (type->proto) {
283 	case IPPROTO_COMP:
284 		X(afinfo, type, comp);
285 		break;
286 	case IPPROTO_AH:
287 		X(afinfo, type, ah);
288 		break;
289 	case IPPROTO_ESP:
290 		X(afinfo, type, esp);
291 		break;
292 	case IPPROTO_IPIP:
293 		X(afinfo, type, ipip);
294 		break;
295 	case IPPROTO_DSTOPTS:
296 		X(afinfo, type, dstopts);
297 		break;
298 	case IPPROTO_ROUTING:
299 		X(afinfo, type, routing);
300 		break;
301 	case IPPROTO_IPV6:
302 		X(afinfo, type, ipip6);
303 		break;
304 	default:
305 		WARN_ON(1);
306 		break;
307 	}
308 #undef X
309 	rcu_read_unlock();
310 }
311 EXPORT_SYMBOL(xfrm_unregister_type);
312 
313 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
314 {
315 	const struct xfrm_type *type = NULL;
316 	struct xfrm_state_afinfo *afinfo;
317 	int modload_attempted = 0;
318 
319 retry:
320 	afinfo = xfrm_state_get_afinfo(family);
321 	if (unlikely(afinfo == NULL))
322 		return NULL;
323 
324 	switch (proto) {
325 	case IPPROTO_COMP:
326 		type = afinfo->type_comp;
327 		break;
328 	case IPPROTO_AH:
329 		type = afinfo->type_ah;
330 		break;
331 	case IPPROTO_ESP:
332 		type = afinfo->type_esp;
333 		break;
334 	case IPPROTO_IPIP:
335 		type = afinfo->type_ipip;
336 		break;
337 	case IPPROTO_DSTOPTS:
338 		type = afinfo->type_dstopts;
339 		break;
340 	case IPPROTO_ROUTING:
341 		type = afinfo->type_routing;
342 		break;
343 	case IPPROTO_IPV6:
344 		type = afinfo->type_ipip6;
345 		break;
346 	default:
347 		break;
348 	}
349 
350 	if (unlikely(type && !try_module_get(type->owner)))
351 		type = NULL;
352 
353 	rcu_read_unlock();
354 
355 	if (!type && !modload_attempted) {
356 		request_module("xfrm-type-%d-%d", family, proto);
357 		modload_attempted = 1;
358 		goto retry;
359 	}
360 
361 	return type;
362 }
363 
364 static void xfrm_put_type(const struct xfrm_type *type)
365 {
366 	module_put(type->owner);
367 }
368 
369 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
370 			       unsigned short family)
371 {
372 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
373 	int err = 0;
374 
375 	if (unlikely(afinfo == NULL))
376 		return -EAFNOSUPPORT;
377 
378 	switch (type->proto) {
379 	case IPPROTO_ESP:
380 		WARN_ON(afinfo->type_offload_esp);
381 		afinfo->type_offload_esp = type;
382 		break;
383 	default:
384 		WARN_ON(1);
385 		err = -EPROTONOSUPPORT;
386 		break;
387 	}
388 
389 	rcu_read_unlock();
390 	return err;
391 }
392 EXPORT_SYMBOL(xfrm_register_type_offload);
393 
394 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
395 				  unsigned short family)
396 {
397 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
398 
399 	if (unlikely(afinfo == NULL))
400 		return;
401 
402 	switch (type->proto) {
403 	case IPPROTO_ESP:
404 		WARN_ON(afinfo->type_offload_esp != type);
405 		afinfo->type_offload_esp = NULL;
406 		break;
407 	default:
408 		WARN_ON(1);
409 		break;
410 	}
411 	rcu_read_unlock();
412 }
413 EXPORT_SYMBOL(xfrm_unregister_type_offload);
414 
415 static const struct xfrm_type_offload *
416 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
417 {
418 	const struct xfrm_type_offload *type = NULL;
419 	struct xfrm_state_afinfo *afinfo;
420 
421 retry:
422 	afinfo = xfrm_state_get_afinfo(family);
423 	if (unlikely(afinfo == NULL))
424 		return NULL;
425 
426 	switch (proto) {
427 	case IPPROTO_ESP:
428 		type = afinfo->type_offload_esp;
429 		break;
430 	default:
431 		break;
432 	}
433 
434 	if ((type && !try_module_get(type->owner)))
435 		type = NULL;
436 
437 	rcu_read_unlock();
438 
439 	if (!type && try_load) {
440 		request_module("xfrm-offload-%d-%d", family, proto);
441 		try_load = false;
442 		goto retry;
443 	}
444 
445 	return type;
446 }
447 
448 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
449 {
450 	module_put(type->owner);
451 }
452 
453 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
454 	[XFRM_MODE_BEET] = {
455 		.encap = XFRM_MODE_BEET,
456 		.flags = XFRM_MODE_FLAG_TUNNEL,
457 		.family = AF_INET,
458 	},
459 	[XFRM_MODE_TRANSPORT] = {
460 		.encap = XFRM_MODE_TRANSPORT,
461 		.family = AF_INET,
462 	},
463 	[XFRM_MODE_TUNNEL] = {
464 		.encap = XFRM_MODE_TUNNEL,
465 		.flags = XFRM_MODE_FLAG_TUNNEL,
466 		.family = AF_INET,
467 	},
468 };
469 
470 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
471 	[XFRM_MODE_BEET] = {
472 		.encap = XFRM_MODE_BEET,
473 		.flags = XFRM_MODE_FLAG_TUNNEL,
474 		.family = AF_INET6,
475 	},
476 	[XFRM_MODE_ROUTEOPTIMIZATION] = {
477 		.encap = XFRM_MODE_ROUTEOPTIMIZATION,
478 		.family = AF_INET6,
479 	},
480 	[XFRM_MODE_TRANSPORT] = {
481 		.encap = XFRM_MODE_TRANSPORT,
482 		.family = AF_INET6,
483 	},
484 	[XFRM_MODE_TUNNEL] = {
485 		.encap = XFRM_MODE_TUNNEL,
486 		.flags = XFRM_MODE_FLAG_TUNNEL,
487 		.family = AF_INET6,
488 	},
489 };
490 
491 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
492 {
493 	const struct xfrm_mode *mode;
494 
495 	if (unlikely(encap >= XFRM_MODE_MAX))
496 		return NULL;
497 
498 	switch (family) {
499 	case AF_INET:
500 		mode = &xfrm4_mode_map[encap];
501 		if (mode->family == family)
502 			return mode;
503 		break;
504 	case AF_INET6:
505 		mode = &xfrm6_mode_map[encap];
506 		if (mode->family == family)
507 			return mode;
508 		break;
509 	default:
510 		break;
511 	}
512 
513 	return NULL;
514 }
515 
516 void xfrm_state_free(struct xfrm_state *x)
517 {
518 	kmem_cache_free(xfrm_state_cache, x);
519 }
520 EXPORT_SYMBOL(xfrm_state_free);
521 
522 static void ___xfrm_state_destroy(struct xfrm_state *x)
523 {
524 	hrtimer_cancel(&x->mtimer);
525 	del_timer_sync(&x->rtimer);
526 	kfree(x->aead);
527 	kfree(x->aalg);
528 	kfree(x->ealg);
529 	kfree(x->calg);
530 	kfree(x->encap);
531 	kfree(x->coaddr);
532 	kfree(x->replay_esn);
533 	kfree(x->preplay_esn);
534 	if (x->type_offload)
535 		xfrm_put_type_offload(x->type_offload);
536 	if (x->type) {
537 		x->type->destructor(x);
538 		xfrm_put_type(x->type);
539 	}
540 	if (x->xfrag.page)
541 		put_page(x->xfrag.page);
542 	xfrm_dev_state_free(x);
543 	security_xfrm_state_free(x);
544 	xfrm_state_free(x);
545 }
546 
547 static void xfrm_state_gc_task(struct work_struct *work)
548 {
549 	struct xfrm_state *x;
550 	struct hlist_node *tmp;
551 	struct hlist_head gc_list;
552 
553 	spin_lock_bh(&xfrm_state_gc_lock);
554 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
555 	spin_unlock_bh(&xfrm_state_gc_lock);
556 
557 	synchronize_rcu();
558 
559 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
560 		___xfrm_state_destroy(x);
561 }
562 
563 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
564 {
565 	struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
566 	enum hrtimer_restart ret = HRTIMER_NORESTART;
567 	time64_t now = ktime_get_real_seconds();
568 	time64_t next = TIME64_MAX;
569 	int warn = 0;
570 	int err = 0;
571 
572 	spin_lock(&x->lock);
573 	xfrm_dev_state_update_curlft(x);
574 
575 	if (x->km.state == XFRM_STATE_DEAD)
576 		goto out;
577 	if (x->km.state == XFRM_STATE_EXPIRED)
578 		goto expired;
579 	if (x->lft.hard_add_expires_seconds) {
580 		time64_t tmo = x->lft.hard_add_expires_seconds +
581 			x->curlft.add_time - now;
582 		if (tmo <= 0) {
583 			if (x->xflags & XFRM_SOFT_EXPIRE) {
584 				/* enter hard expire without soft expire first?!
585 				 * setting a new date could trigger this.
586 				 * workaround: fix x->curflt.add_time by below:
587 				 */
588 				x->curlft.add_time = now - x->saved_tmo - 1;
589 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
590 			} else
591 				goto expired;
592 		}
593 		if (tmo < next)
594 			next = tmo;
595 	}
596 	if (x->lft.hard_use_expires_seconds) {
597 		time64_t tmo = x->lft.hard_use_expires_seconds +
598 			(READ_ONCE(x->curlft.use_time) ? : now) - now;
599 		if (tmo <= 0)
600 			goto expired;
601 		if (tmo < next)
602 			next = tmo;
603 	}
604 	if (x->km.dying)
605 		goto resched;
606 	if (x->lft.soft_add_expires_seconds) {
607 		time64_t tmo = x->lft.soft_add_expires_seconds +
608 			x->curlft.add_time - now;
609 		if (tmo <= 0) {
610 			warn = 1;
611 			x->xflags &= ~XFRM_SOFT_EXPIRE;
612 		} else if (tmo < next) {
613 			next = tmo;
614 			x->xflags |= XFRM_SOFT_EXPIRE;
615 			x->saved_tmo = tmo;
616 		}
617 	}
618 	if (x->lft.soft_use_expires_seconds) {
619 		time64_t tmo = x->lft.soft_use_expires_seconds +
620 			(READ_ONCE(x->curlft.use_time) ? : now) - now;
621 		if (tmo <= 0)
622 			warn = 1;
623 		else if (tmo < next)
624 			next = tmo;
625 	}
626 
627 	x->km.dying = warn;
628 	if (warn)
629 		km_state_expired(x, 0, 0);
630 resched:
631 	if (next != TIME64_MAX) {
632 		hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
633 		ret = HRTIMER_RESTART;
634 	}
635 
636 	goto out;
637 
638 expired:
639 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
640 		x->km.state = XFRM_STATE_EXPIRED;
641 
642 	err = __xfrm_state_delete(x);
643 	if (!err)
644 		km_state_expired(x, 1, 0);
645 
646 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
647 
648 out:
649 	spin_unlock(&x->lock);
650 	return ret;
651 }
652 
653 static void xfrm_replay_timer_handler(struct timer_list *t);
654 
655 struct xfrm_state *xfrm_state_alloc(struct net *net)
656 {
657 	struct xfrm_state *x;
658 
659 	x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
660 
661 	if (x) {
662 		write_pnet(&x->xs_net, net);
663 		refcount_set(&x->refcnt, 1);
664 		atomic_set(&x->tunnel_users, 0);
665 		INIT_LIST_HEAD(&x->km.all);
666 		INIT_HLIST_NODE(&x->bydst);
667 		INIT_HLIST_NODE(&x->bysrc);
668 		INIT_HLIST_NODE(&x->byspi);
669 		INIT_HLIST_NODE(&x->byseq);
670 		hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
671 		x->mtimer.function = xfrm_timer_handler;
672 		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
673 		x->curlft.add_time = ktime_get_real_seconds();
674 		x->lft.soft_byte_limit = XFRM_INF;
675 		x->lft.soft_packet_limit = XFRM_INF;
676 		x->lft.hard_byte_limit = XFRM_INF;
677 		x->lft.hard_packet_limit = XFRM_INF;
678 		x->replay_maxage = 0;
679 		x->replay_maxdiff = 0;
680 		spin_lock_init(&x->lock);
681 	}
682 	return x;
683 }
684 EXPORT_SYMBOL(xfrm_state_alloc);
685 
686 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
687 {
688 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
689 
690 	if (sync) {
691 		synchronize_rcu();
692 		___xfrm_state_destroy(x);
693 	} else {
694 		spin_lock_bh(&xfrm_state_gc_lock);
695 		hlist_add_head(&x->gclist, &xfrm_state_gc_list);
696 		spin_unlock_bh(&xfrm_state_gc_lock);
697 		schedule_work(&xfrm_state_gc_work);
698 	}
699 }
700 EXPORT_SYMBOL(__xfrm_state_destroy);
701 
702 int __xfrm_state_delete(struct xfrm_state *x)
703 {
704 	struct net *net = xs_net(x);
705 	int err = -ESRCH;
706 
707 	if (x->km.state != XFRM_STATE_DEAD) {
708 		x->km.state = XFRM_STATE_DEAD;
709 		spin_lock(&net->xfrm.xfrm_state_lock);
710 		list_del(&x->km.all);
711 		hlist_del_rcu(&x->bydst);
712 		hlist_del_rcu(&x->bysrc);
713 		if (x->km.seq)
714 			hlist_del_rcu(&x->byseq);
715 		if (x->id.spi)
716 			hlist_del_rcu(&x->byspi);
717 		net->xfrm.state_num--;
718 		spin_unlock(&net->xfrm.xfrm_state_lock);
719 
720 		if (x->encap_sk)
721 			sock_put(rcu_dereference_raw(x->encap_sk));
722 
723 		xfrm_dev_state_delete(x);
724 
725 		/* All xfrm_state objects are created by xfrm_state_alloc.
726 		 * The xfrm_state_alloc call gives a reference, and that
727 		 * is what we are dropping here.
728 		 */
729 		xfrm_state_put(x);
730 		err = 0;
731 	}
732 
733 	return err;
734 }
735 EXPORT_SYMBOL(__xfrm_state_delete);
736 
737 int xfrm_state_delete(struct xfrm_state *x)
738 {
739 	int err;
740 
741 	spin_lock_bh(&x->lock);
742 	err = __xfrm_state_delete(x);
743 	spin_unlock_bh(&x->lock);
744 
745 	return err;
746 }
747 EXPORT_SYMBOL(xfrm_state_delete);
748 
749 #ifdef CONFIG_SECURITY_NETWORK_XFRM
750 static inline int
751 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
752 {
753 	int i, err = 0;
754 
755 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
756 		struct xfrm_state *x;
757 
758 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
759 			if (xfrm_id_proto_match(x->id.proto, proto) &&
760 			   (err = security_xfrm_state_delete(x)) != 0) {
761 				xfrm_audit_state_delete(x, 0, task_valid);
762 				return err;
763 			}
764 		}
765 	}
766 
767 	return err;
768 }
769 
770 static inline int
771 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
772 {
773 	int i, err = 0;
774 
775 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
776 		struct xfrm_state *x;
777 		struct xfrm_dev_offload *xso;
778 
779 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
780 			xso = &x->xso;
781 
782 			if (xso->dev == dev &&
783 			   (err = security_xfrm_state_delete(x)) != 0) {
784 				xfrm_audit_state_delete(x, 0, task_valid);
785 				return err;
786 			}
787 		}
788 	}
789 
790 	return err;
791 }
792 #else
793 static inline int
794 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
795 {
796 	return 0;
797 }
798 
799 static inline int
800 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
801 {
802 	return 0;
803 }
804 #endif
805 
806 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
807 {
808 	int i, err = 0, cnt = 0;
809 
810 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
811 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
812 	if (err)
813 		goto out;
814 
815 	err = -ESRCH;
816 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
817 		struct xfrm_state *x;
818 restart:
819 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
820 			if (!xfrm_state_kern(x) &&
821 			    xfrm_id_proto_match(x->id.proto, proto)) {
822 				xfrm_state_hold(x);
823 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
824 
825 				err = xfrm_state_delete(x);
826 				xfrm_audit_state_delete(x, err ? 0 : 1,
827 							task_valid);
828 				if (sync)
829 					xfrm_state_put_sync(x);
830 				else
831 					xfrm_state_put(x);
832 				if (!err)
833 					cnt++;
834 
835 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
836 				goto restart;
837 			}
838 		}
839 	}
840 out:
841 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
842 	if (cnt)
843 		err = 0;
844 
845 	return err;
846 }
847 EXPORT_SYMBOL(xfrm_state_flush);
848 
849 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
850 {
851 	int i, err = 0, cnt = 0;
852 
853 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
854 	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
855 	if (err)
856 		goto out;
857 
858 	err = -ESRCH;
859 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
860 		struct xfrm_state *x;
861 		struct xfrm_dev_offload *xso;
862 restart:
863 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
864 			xso = &x->xso;
865 
866 			if (!xfrm_state_kern(x) && xso->dev == dev) {
867 				xfrm_state_hold(x);
868 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
869 
870 				err = xfrm_state_delete(x);
871 				xfrm_audit_state_delete(x, err ? 0 : 1,
872 							task_valid);
873 				xfrm_state_put(x);
874 				if (!err)
875 					cnt++;
876 
877 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
878 				goto restart;
879 			}
880 		}
881 	}
882 	if (cnt)
883 		err = 0;
884 
885 out:
886 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
887 	return err;
888 }
889 EXPORT_SYMBOL(xfrm_dev_state_flush);
890 
891 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
892 {
893 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
894 	si->sadcnt = net->xfrm.state_num;
895 	si->sadhcnt = net->xfrm.state_hmask + 1;
896 	si->sadhmcnt = xfrm_state_hashmax;
897 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
898 }
899 EXPORT_SYMBOL(xfrm_sad_getinfo);
900 
901 static void
902 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
903 {
904 	const struct flowi4 *fl4 = &fl->u.ip4;
905 
906 	sel->daddr.a4 = fl4->daddr;
907 	sel->saddr.a4 = fl4->saddr;
908 	sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
909 	sel->dport_mask = htons(0xffff);
910 	sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
911 	sel->sport_mask = htons(0xffff);
912 	sel->family = AF_INET;
913 	sel->prefixlen_d = 32;
914 	sel->prefixlen_s = 32;
915 	sel->proto = fl4->flowi4_proto;
916 	sel->ifindex = fl4->flowi4_oif;
917 }
918 
919 static void
920 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
921 {
922 	const struct flowi6 *fl6 = &fl->u.ip6;
923 
924 	/* Initialize temporary selector matching only to current session. */
925 	*(struct in6_addr *)&sel->daddr = fl6->daddr;
926 	*(struct in6_addr *)&sel->saddr = fl6->saddr;
927 	sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
928 	sel->dport_mask = htons(0xffff);
929 	sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
930 	sel->sport_mask = htons(0xffff);
931 	sel->family = AF_INET6;
932 	sel->prefixlen_d = 128;
933 	sel->prefixlen_s = 128;
934 	sel->proto = fl6->flowi6_proto;
935 	sel->ifindex = fl6->flowi6_oif;
936 }
937 
938 static void
939 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
940 		    const struct xfrm_tmpl *tmpl,
941 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
942 		    unsigned short family)
943 {
944 	switch (family) {
945 	case AF_INET:
946 		__xfrm4_init_tempsel(&x->sel, fl);
947 		break;
948 	case AF_INET6:
949 		__xfrm6_init_tempsel(&x->sel, fl);
950 		break;
951 	}
952 
953 	x->id = tmpl->id;
954 
955 	switch (tmpl->encap_family) {
956 	case AF_INET:
957 		if (x->id.daddr.a4 == 0)
958 			x->id.daddr.a4 = daddr->a4;
959 		x->props.saddr = tmpl->saddr;
960 		if (x->props.saddr.a4 == 0)
961 			x->props.saddr.a4 = saddr->a4;
962 		break;
963 	case AF_INET6:
964 		if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
965 			memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
966 		memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
967 		if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
968 			memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
969 		break;
970 	}
971 
972 	x->props.mode = tmpl->mode;
973 	x->props.reqid = tmpl->reqid;
974 	x->props.family = tmpl->encap_family;
975 }
976 
977 static struct xfrm_state *__xfrm_state_lookup_all(struct net *net, u32 mark,
978 						  const xfrm_address_t *daddr,
979 						  __be32 spi, u8 proto,
980 						  unsigned short family,
981 						  struct xfrm_dev_offload *xdo)
982 {
983 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
984 	struct xfrm_state *x;
985 
986 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
987 #ifdef CONFIG_XFRM_OFFLOAD
988 		if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) {
989 			if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
990 				/* HW states are in the head of list, there is
991 				 * no need to iterate further.
992 				 */
993 				break;
994 
995 			/* Packet offload: both policy and SA should
996 			 * have same device.
997 			 */
998 			if (xdo->dev != x->xso.dev)
999 				continue;
1000 		} else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1001 			/* Skip HW policy for SW lookups */
1002 			continue;
1003 #endif
1004 		if (x->props.family != family ||
1005 		    x->id.spi       != spi ||
1006 		    x->id.proto     != proto ||
1007 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
1008 			continue;
1009 
1010 		if ((mark & x->mark.m) != x->mark.v)
1011 			continue;
1012 		if (!xfrm_state_hold_rcu(x))
1013 			continue;
1014 		return x;
1015 	}
1016 
1017 	return NULL;
1018 }
1019 
1020 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
1021 					      const xfrm_address_t *daddr,
1022 					      __be32 spi, u8 proto,
1023 					      unsigned short family)
1024 {
1025 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
1026 	struct xfrm_state *x;
1027 
1028 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
1029 		if (x->props.family != family ||
1030 		    x->id.spi       != spi ||
1031 		    x->id.proto     != proto ||
1032 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
1033 			continue;
1034 
1035 		if ((mark & x->mark.m) != x->mark.v)
1036 			continue;
1037 		if (!xfrm_state_hold_rcu(x))
1038 			continue;
1039 		return x;
1040 	}
1041 
1042 	return NULL;
1043 }
1044 
1045 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1046 						     const xfrm_address_t *daddr,
1047 						     const xfrm_address_t *saddr,
1048 						     u8 proto, unsigned short family)
1049 {
1050 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
1051 	struct xfrm_state *x;
1052 
1053 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
1054 		if (x->props.family != family ||
1055 		    x->id.proto     != proto ||
1056 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1057 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1058 			continue;
1059 
1060 		if ((mark & x->mark.m) != x->mark.v)
1061 			continue;
1062 		if (!xfrm_state_hold_rcu(x))
1063 			continue;
1064 		return x;
1065 	}
1066 
1067 	return NULL;
1068 }
1069 
1070 static inline struct xfrm_state *
1071 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1072 {
1073 	struct net *net = xs_net(x);
1074 	u32 mark = x->mark.v & x->mark.m;
1075 
1076 	if (use_spi)
1077 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
1078 					   x->id.spi, x->id.proto, family);
1079 	else
1080 		return __xfrm_state_lookup_byaddr(net, mark,
1081 						  &x->id.daddr,
1082 						  &x->props.saddr,
1083 						  x->id.proto, family);
1084 }
1085 
1086 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1087 {
1088 	if (have_hash_collision &&
1089 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1090 	    net->xfrm.state_num > net->xfrm.state_hmask)
1091 		schedule_work(&net->xfrm.state_hash_work);
1092 }
1093 
1094 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1095 			       const struct flowi *fl, unsigned short family,
1096 			       struct xfrm_state **best, int *acq_in_progress,
1097 			       int *error)
1098 {
1099 	/* Resolution logic:
1100 	 * 1. There is a valid state with matching selector. Done.
1101 	 * 2. Valid state with inappropriate selector. Skip.
1102 	 *
1103 	 * Entering area of "sysdeps".
1104 	 *
1105 	 * 3. If state is not valid, selector is temporary, it selects
1106 	 *    only session which triggered previous resolution. Key
1107 	 *    manager will do something to install a state with proper
1108 	 *    selector.
1109 	 */
1110 	if (x->km.state == XFRM_STATE_VALID) {
1111 		if ((x->sel.family &&
1112 		     (x->sel.family != family ||
1113 		      !xfrm_selector_match(&x->sel, fl, family))) ||
1114 		    !security_xfrm_state_pol_flow_match(x, pol,
1115 							&fl->u.__fl_common))
1116 			return;
1117 
1118 		if (!*best ||
1119 		    (*best)->km.dying > x->km.dying ||
1120 		    ((*best)->km.dying == x->km.dying &&
1121 		     (*best)->curlft.add_time < x->curlft.add_time))
1122 			*best = x;
1123 	} else if (x->km.state == XFRM_STATE_ACQ) {
1124 		*acq_in_progress = 1;
1125 	} else if (x->km.state == XFRM_STATE_ERROR ||
1126 		   x->km.state == XFRM_STATE_EXPIRED) {
1127 		if ((!x->sel.family ||
1128 		     (x->sel.family == family &&
1129 		      xfrm_selector_match(&x->sel, fl, family))) &&
1130 		    security_xfrm_state_pol_flow_match(x, pol,
1131 						       &fl->u.__fl_common))
1132 			*error = -ESRCH;
1133 	}
1134 }
1135 
1136 struct xfrm_state *
1137 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1138 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1139 		struct xfrm_policy *pol, int *err,
1140 		unsigned short family, u32 if_id)
1141 {
1142 	static xfrm_address_t saddr_wildcard = { };
1143 	struct net *net = xp_net(pol);
1144 	unsigned int h, h_wildcard;
1145 	struct xfrm_state *x, *x0, *to_put;
1146 	int acquire_in_progress = 0;
1147 	int error = 0;
1148 	struct xfrm_state *best = NULL;
1149 	u32 mark = pol->mark.v & pol->mark.m;
1150 	unsigned short encap_family = tmpl->encap_family;
1151 	unsigned int sequence;
1152 	struct km_event c;
1153 
1154 	to_put = NULL;
1155 
1156 	sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1157 
1158 	rcu_read_lock();
1159 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1160 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1161 #ifdef CONFIG_XFRM_OFFLOAD
1162 		if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1163 			if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1164 				/* HW states are in the head of list, there is
1165 				 * no need to iterate further.
1166 				 */
1167 				break;
1168 
1169 			/* Packet offload: both policy and SA should
1170 			 * have same device.
1171 			 */
1172 			if (pol->xdo.dev != x->xso.dev)
1173 				continue;
1174 		} else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1175 			/* Skip HW policy for SW lookups */
1176 			continue;
1177 #endif
1178 		if (x->props.family == encap_family &&
1179 		    x->props.reqid == tmpl->reqid &&
1180 		    (mark & x->mark.m) == x->mark.v &&
1181 		    x->if_id == if_id &&
1182 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1183 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1184 		    tmpl->mode == x->props.mode &&
1185 		    tmpl->id.proto == x->id.proto &&
1186 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1187 			xfrm_state_look_at(pol, x, fl, family,
1188 					   &best, &acquire_in_progress, &error);
1189 	}
1190 	if (best || acquire_in_progress)
1191 		goto found;
1192 
1193 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1194 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1195 #ifdef CONFIG_XFRM_OFFLOAD
1196 		if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1197 			if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1198 				/* HW states are in the head of list, there is
1199 				 * no need to iterate further.
1200 				 */
1201 				break;
1202 
1203 			/* Packet offload: both policy and SA should
1204 			 * have same device.
1205 			 */
1206 			if (pol->xdo.dev != x->xso.dev)
1207 				continue;
1208 		} else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1209 			/* Skip HW policy for SW lookups */
1210 			continue;
1211 #endif
1212 		if (x->props.family == encap_family &&
1213 		    x->props.reqid == tmpl->reqid &&
1214 		    (mark & x->mark.m) == x->mark.v &&
1215 		    x->if_id == if_id &&
1216 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1217 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1218 		    tmpl->mode == x->props.mode &&
1219 		    tmpl->id.proto == x->id.proto &&
1220 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1221 			xfrm_state_look_at(pol, x, fl, family,
1222 					   &best, &acquire_in_progress, &error);
1223 	}
1224 
1225 found:
1226 	x = best;
1227 	if (!x && !error && !acquire_in_progress) {
1228 		if (tmpl->id.spi &&
1229 		    (x0 = __xfrm_state_lookup_all(net, mark, daddr,
1230 						  tmpl->id.spi, tmpl->id.proto,
1231 						  encap_family,
1232 						  &pol->xdo)) != NULL) {
1233 			to_put = x0;
1234 			error = -EEXIST;
1235 			goto out;
1236 		}
1237 
1238 		c.net = net;
1239 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1240 		 * for each and every packet - garbage collection might not
1241 		 * handle the flood.
1242 		 */
1243 		if (!km_is_alive(&c)) {
1244 			error = -ESRCH;
1245 			goto out;
1246 		}
1247 
1248 		x = xfrm_state_alloc(net);
1249 		if (x == NULL) {
1250 			error = -ENOMEM;
1251 			goto out;
1252 		}
1253 		/* Initialize temporary state matching only
1254 		 * to current session. */
1255 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1256 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1257 		x->if_id = if_id;
1258 
1259 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1260 		if (error) {
1261 			x->km.state = XFRM_STATE_DEAD;
1262 			to_put = x;
1263 			x = NULL;
1264 			goto out;
1265 		}
1266 #ifdef CONFIG_XFRM_OFFLOAD
1267 		if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1268 			struct xfrm_dev_offload *xdo = &pol->xdo;
1269 			struct xfrm_dev_offload *xso = &x->xso;
1270 
1271 			xso->type = XFRM_DEV_OFFLOAD_PACKET;
1272 			xso->dir = xdo->dir;
1273 			xso->dev = xdo->dev;
1274 			xso->real_dev = xdo->real_dev;
1275 			xso->flags = XFRM_DEV_OFFLOAD_FLAG_ACQ;
1276 			netdev_tracker_alloc(xso->dev, &xso->dev_tracker,
1277 					     GFP_ATOMIC);
1278 			error = xso->dev->xfrmdev_ops->xdo_dev_state_add(x, NULL);
1279 			if (error) {
1280 				xso->dir = 0;
1281 				netdev_put(xso->dev, &xso->dev_tracker);
1282 				xso->dev = NULL;
1283 				xso->real_dev = NULL;
1284 				xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
1285 				x->km.state = XFRM_STATE_DEAD;
1286 				to_put = x;
1287 				x = NULL;
1288 				goto out;
1289 			}
1290 		}
1291 #endif
1292 		if (km_query(x, tmpl, pol) == 0) {
1293 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1294 			x->km.state = XFRM_STATE_ACQ;
1295 			list_add(&x->km.all, &net->xfrm.state_all);
1296 			XFRM_STATE_INSERT(bydst, &x->bydst,
1297 					  net->xfrm.state_bydst + h,
1298 					  x->xso.type);
1299 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1300 			XFRM_STATE_INSERT(bysrc, &x->bysrc,
1301 					  net->xfrm.state_bysrc + h,
1302 					  x->xso.type);
1303 			if (x->id.spi) {
1304 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1305 				XFRM_STATE_INSERT(byspi, &x->byspi,
1306 						  net->xfrm.state_byspi + h,
1307 						  x->xso.type);
1308 			}
1309 			if (x->km.seq) {
1310 				h = xfrm_seq_hash(net, x->km.seq);
1311 				XFRM_STATE_INSERT(byseq, &x->byseq,
1312 						  net->xfrm.state_byseq + h,
1313 						  x->xso.type);
1314 			}
1315 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1316 			hrtimer_start(&x->mtimer,
1317 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1318 				      HRTIMER_MODE_REL_SOFT);
1319 			net->xfrm.state_num++;
1320 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1321 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1322 		} else {
1323 #ifdef CONFIG_XFRM_OFFLOAD
1324 			struct xfrm_dev_offload *xso = &x->xso;
1325 
1326 			if (xso->type == XFRM_DEV_OFFLOAD_PACKET) {
1327 				xfrm_dev_state_delete(x);
1328 				xfrm_dev_state_free(x);
1329 			}
1330 #endif
1331 			x->km.state = XFRM_STATE_DEAD;
1332 			to_put = x;
1333 			x = NULL;
1334 			error = -ESRCH;
1335 		}
1336 	}
1337 out:
1338 	if (x) {
1339 		if (!xfrm_state_hold_rcu(x)) {
1340 			*err = -EAGAIN;
1341 			x = NULL;
1342 		}
1343 	} else {
1344 		*err = acquire_in_progress ? -EAGAIN : error;
1345 	}
1346 	rcu_read_unlock();
1347 	if (to_put)
1348 		xfrm_state_put(to_put);
1349 
1350 	if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1351 		*err = -EAGAIN;
1352 		if (x) {
1353 			xfrm_state_put(x);
1354 			x = NULL;
1355 		}
1356 	}
1357 
1358 	return x;
1359 }
1360 
1361 struct xfrm_state *
1362 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1363 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1364 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1365 {
1366 	unsigned int h;
1367 	struct xfrm_state *rx = NULL, *x = NULL;
1368 
1369 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1370 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1371 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1372 		if (x->props.family == family &&
1373 		    x->props.reqid == reqid &&
1374 		    (mark & x->mark.m) == x->mark.v &&
1375 		    x->if_id == if_id &&
1376 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1377 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1378 		    mode == x->props.mode &&
1379 		    proto == x->id.proto &&
1380 		    x->km.state == XFRM_STATE_VALID) {
1381 			rx = x;
1382 			break;
1383 		}
1384 	}
1385 
1386 	if (rx)
1387 		xfrm_state_hold(rx);
1388 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1389 
1390 
1391 	return rx;
1392 }
1393 EXPORT_SYMBOL(xfrm_stateonly_find);
1394 
1395 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1396 					      unsigned short family)
1397 {
1398 	struct xfrm_state *x;
1399 	struct xfrm_state_walk *w;
1400 
1401 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1402 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1403 		x = container_of(w, struct xfrm_state, km);
1404 		if (x->props.family != family ||
1405 			x->id.spi != spi)
1406 			continue;
1407 
1408 		xfrm_state_hold(x);
1409 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1410 		return x;
1411 	}
1412 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1413 	return NULL;
1414 }
1415 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1416 
1417 static void __xfrm_state_insert(struct xfrm_state *x)
1418 {
1419 	struct net *net = xs_net(x);
1420 	unsigned int h;
1421 
1422 	list_add(&x->km.all, &net->xfrm.state_all);
1423 
1424 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1425 			  x->props.reqid, x->props.family);
1426 	XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1427 			  x->xso.type);
1428 
1429 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1430 	XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1431 			  x->xso.type);
1432 
1433 	if (x->id.spi) {
1434 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1435 				  x->props.family);
1436 
1437 		XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
1438 				  x->xso.type);
1439 	}
1440 
1441 	if (x->km.seq) {
1442 		h = xfrm_seq_hash(net, x->km.seq);
1443 
1444 		XFRM_STATE_INSERT(byseq, &x->byseq, net->xfrm.state_byseq + h,
1445 				  x->xso.type);
1446 	}
1447 
1448 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1449 	if (x->replay_maxage)
1450 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1451 
1452 	net->xfrm.state_num++;
1453 
1454 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1455 }
1456 
1457 /* net->xfrm.xfrm_state_lock is held */
1458 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1459 {
1460 	struct net *net = xs_net(xnew);
1461 	unsigned short family = xnew->props.family;
1462 	u32 reqid = xnew->props.reqid;
1463 	struct xfrm_state *x;
1464 	unsigned int h;
1465 	u32 mark = xnew->mark.v & xnew->mark.m;
1466 	u32 if_id = xnew->if_id;
1467 
1468 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1469 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1470 		if (x->props.family	== family &&
1471 		    x->props.reqid	== reqid &&
1472 		    x->if_id		== if_id &&
1473 		    (mark & x->mark.m) == x->mark.v &&
1474 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1475 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1476 			x->genid++;
1477 	}
1478 }
1479 
1480 void xfrm_state_insert(struct xfrm_state *x)
1481 {
1482 	struct net *net = xs_net(x);
1483 
1484 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1485 	__xfrm_state_bump_genids(x);
1486 	__xfrm_state_insert(x);
1487 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1488 }
1489 EXPORT_SYMBOL(xfrm_state_insert);
1490 
1491 /* net->xfrm.xfrm_state_lock is held */
1492 static struct xfrm_state *__find_acq_core(struct net *net,
1493 					  const struct xfrm_mark *m,
1494 					  unsigned short family, u8 mode,
1495 					  u32 reqid, u32 if_id, u8 proto,
1496 					  const xfrm_address_t *daddr,
1497 					  const xfrm_address_t *saddr,
1498 					  int create)
1499 {
1500 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1501 	struct xfrm_state *x;
1502 	u32 mark = m->v & m->m;
1503 
1504 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1505 		if (x->props.reqid  != reqid ||
1506 		    x->props.mode   != mode ||
1507 		    x->props.family != family ||
1508 		    x->km.state     != XFRM_STATE_ACQ ||
1509 		    x->id.spi       != 0 ||
1510 		    x->id.proto	    != proto ||
1511 		    (mark & x->mark.m) != x->mark.v ||
1512 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1513 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1514 			continue;
1515 
1516 		xfrm_state_hold(x);
1517 		return x;
1518 	}
1519 
1520 	if (!create)
1521 		return NULL;
1522 
1523 	x = xfrm_state_alloc(net);
1524 	if (likely(x)) {
1525 		switch (family) {
1526 		case AF_INET:
1527 			x->sel.daddr.a4 = daddr->a4;
1528 			x->sel.saddr.a4 = saddr->a4;
1529 			x->sel.prefixlen_d = 32;
1530 			x->sel.prefixlen_s = 32;
1531 			x->props.saddr.a4 = saddr->a4;
1532 			x->id.daddr.a4 = daddr->a4;
1533 			break;
1534 
1535 		case AF_INET6:
1536 			x->sel.daddr.in6 = daddr->in6;
1537 			x->sel.saddr.in6 = saddr->in6;
1538 			x->sel.prefixlen_d = 128;
1539 			x->sel.prefixlen_s = 128;
1540 			x->props.saddr.in6 = saddr->in6;
1541 			x->id.daddr.in6 = daddr->in6;
1542 			break;
1543 		}
1544 
1545 		x->km.state = XFRM_STATE_ACQ;
1546 		x->id.proto = proto;
1547 		x->props.family = family;
1548 		x->props.mode = mode;
1549 		x->props.reqid = reqid;
1550 		x->if_id = if_id;
1551 		x->mark.v = m->v;
1552 		x->mark.m = m->m;
1553 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1554 		xfrm_state_hold(x);
1555 		hrtimer_start(&x->mtimer,
1556 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1557 			      HRTIMER_MODE_REL_SOFT);
1558 		list_add(&x->km.all, &net->xfrm.state_all);
1559 		XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1560 				  x->xso.type);
1561 		h = xfrm_src_hash(net, daddr, saddr, family);
1562 		XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1563 				  x->xso.type);
1564 
1565 		net->xfrm.state_num++;
1566 
1567 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1568 	}
1569 
1570 	return x;
1571 }
1572 
1573 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1574 
1575 int xfrm_state_add(struct xfrm_state *x)
1576 {
1577 	struct net *net = xs_net(x);
1578 	struct xfrm_state *x1, *to_put;
1579 	int family;
1580 	int err;
1581 	u32 mark = x->mark.v & x->mark.m;
1582 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1583 
1584 	family = x->props.family;
1585 
1586 	to_put = NULL;
1587 
1588 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1589 
1590 	x1 = __xfrm_state_locate(x, use_spi, family);
1591 	if (x1) {
1592 		to_put = x1;
1593 		x1 = NULL;
1594 		err = -EEXIST;
1595 		goto out;
1596 	}
1597 
1598 	if (use_spi && x->km.seq) {
1599 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1600 		if (x1 && ((x1->id.proto != x->id.proto) ||
1601 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1602 			to_put = x1;
1603 			x1 = NULL;
1604 		}
1605 	}
1606 
1607 	if (use_spi && !x1)
1608 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1609 				     x->props.reqid, x->if_id, x->id.proto,
1610 				     &x->id.daddr, &x->props.saddr, 0);
1611 
1612 	__xfrm_state_bump_genids(x);
1613 	__xfrm_state_insert(x);
1614 	err = 0;
1615 
1616 out:
1617 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1618 
1619 	if (x1) {
1620 		xfrm_state_delete(x1);
1621 		xfrm_state_put(x1);
1622 	}
1623 
1624 	if (to_put)
1625 		xfrm_state_put(to_put);
1626 
1627 	return err;
1628 }
1629 EXPORT_SYMBOL(xfrm_state_add);
1630 
1631 #ifdef CONFIG_XFRM_MIGRATE
1632 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1633 {
1634 	struct xfrm_user_sec_ctx *uctx;
1635 	int size = sizeof(*uctx) + security->ctx_len;
1636 	int err;
1637 
1638 	uctx = kmalloc(size, GFP_KERNEL);
1639 	if (!uctx)
1640 		return -ENOMEM;
1641 
1642 	uctx->exttype = XFRMA_SEC_CTX;
1643 	uctx->len = size;
1644 	uctx->ctx_doi = security->ctx_doi;
1645 	uctx->ctx_alg = security->ctx_alg;
1646 	uctx->ctx_len = security->ctx_len;
1647 	memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1648 	err = security_xfrm_state_alloc(x, uctx);
1649 	kfree(uctx);
1650 	if (err)
1651 		return err;
1652 
1653 	return 0;
1654 }
1655 
1656 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1657 					   struct xfrm_encap_tmpl *encap)
1658 {
1659 	struct net *net = xs_net(orig);
1660 	struct xfrm_state *x = xfrm_state_alloc(net);
1661 	if (!x)
1662 		goto out;
1663 
1664 	memcpy(&x->id, &orig->id, sizeof(x->id));
1665 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1666 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1667 	x->props.mode = orig->props.mode;
1668 	x->props.replay_window = orig->props.replay_window;
1669 	x->props.reqid = orig->props.reqid;
1670 	x->props.family = orig->props.family;
1671 	x->props.saddr = orig->props.saddr;
1672 
1673 	if (orig->aalg) {
1674 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1675 		if (!x->aalg)
1676 			goto error;
1677 	}
1678 	x->props.aalgo = orig->props.aalgo;
1679 
1680 	if (orig->aead) {
1681 		x->aead = xfrm_algo_aead_clone(orig->aead);
1682 		x->geniv = orig->geniv;
1683 		if (!x->aead)
1684 			goto error;
1685 	}
1686 	if (orig->ealg) {
1687 		x->ealg = xfrm_algo_clone(orig->ealg);
1688 		if (!x->ealg)
1689 			goto error;
1690 	}
1691 	x->props.ealgo = orig->props.ealgo;
1692 
1693 	if (orig->calg) {
1694 		x->calg = xfrm_algo_clone(orig->calg);
1695 		if (!x->calg)
1696 			goto error;
1697 	}
1698 	x->props.calgo = orig->props.calgo;
1699 
1700 	if (encap || orig->encap) {
1701 		if (encap)
1702 			x->encap = kmemdup(encap, sizeof(*x->encap),
1703 					GFP_KERNEL);
1704 		else
1705 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1706 					GFP_KERNEL);
1707 
1708 		if (!x->encap)
1709 			goto error;
1710 	}
1711 
1712 	if (orig->security)
1713 		if (clone_security(x, orig->security))
1714 			goto error;
1715 
1716 	if (orig->coaddr) {
1717 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1718 				    GFP_KERNEL);
1719 		if (!x->coaddr)
1720 			goto error;
1721 	}
1722 
1723 	if (orig->replay_esn) {
1724 		if (xfrm_replay_clone(x, orig))
1725 			goto error;
1726 	}
1727 
1728 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1729 	memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1730 
1731 	x->props.flags = orig->props.flags;
1732 	x->props.extra_flags = orig->props.extra_flags;
1733 
1734 	x->if_id = orig->if_id;
1735 	x->tfcpad = orig->tfcpad;
1736 	x->replay_maxdiff = orig->replay_maxdiff;
1737 	x->replay_maxage = orig->replay_maxage;
1738 	memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1739 	x->km.state = orig->km.state;
1740 	x->km.seq = orig->km.seq;
1741 	x->replay = orig->replay;
1742 	x->preplay = orig->preplay;
1743 	x->mapping_maxage = orig->mapping_maxage;
1744 	x->lastused = orig->lastused;
1745 	x->new_mapping = 0;
1746 	x->new_mapping_sport = 0;
1747 
1748 	return x;
1749 
1750  error:
1751 	xfrm_state_put(x);
1752 out:
1753 	return NULL;
1754 }
1755 
1756 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
1757 						u32 if_id)
1758 {
1759 	unsigned int h;
1760 	struct xfrm_state *x = NULL;
1761 
1762 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1763 
1764 	if (m->reqid) {
1765 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1766 				  m->reqid, m->old_family);
1767 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1768 			if (x->props.mode != m->mode ||
1769 			    x->id.proto != m->proto)
1770 				continue;
1771 			if (m->reqid && x->props.reqid != m->reqid)
1772 				continue;
1773 			if (if_id != 0 && x->if_id != if_id)
1774 				continue;
1775 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1776 					     m->old_family) ||
1777 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1778 					     m->old_family))
1779 				continue;
1780 			xfrm_state_hold(x);
1781 			break;
1782 		}
1783 	} else {
1784 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1785 				  m->old_family);
1786 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1787 			if (x->props.mode != m->mode ||
1788 			    x->id.proto != m->proto)
1789 				continue;
1790 			if (if_id != 0 && x->if_id != if_id)
1791 				continue;
1792 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1793 					     m->old_family) ||
1794 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1795 					     m->old_family))
1796 				continue;
1797 			xfrm_state_hold(x);
1798 			break;
1799 		}
1800 	}
1801 
1802 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1803 
1804 	return x;
1805 }
1806 EXPORT_SYMBOL(xfrm_migrate_state_find);
1807 
1808 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1809 				      struct xfrm_migrate *m,
1810 				      struct xfrm_encap_tmpl *encap)
1811 {
1812 	struct xfrm_state *xc;
1813 
1814 	xc = xfrm_state_clone(x, encap);
1815 	if (!xc)
1816 		return NULL;
1817 
1818 	xc->props.family = m->new_family;
1819 
1820 	if (xfrm_init_state(xc) < 0)
1821 		goto error;
1822 
1823 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1824 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1825 
1826 	/* add state */
1827 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1828 		/* a care is needed when the destination address of the
1829 		   state is to be updated as it is a part of triplet */
1830 		xfrm_state_insert(xc);
1831 	} else {
1832 		if (xfrm_state_add(xc) < 0)
1833 			goto error;
1834 	}
1835 
1836 	return xc;
1837 error:
1838 	xfrm_state_put(xc);
1839 	return NULL;
1840 }
1841 EXPORT_SYMBOL(xfrm_state_migrate);
1842 #endif
1843 
1844 int xfrm_state_update(struct xfrm_state *x)
1845 {
1846 	struct xfrm_state *x1, *to_put;
1847 	int err;
1848 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1849 	struct net *net = xs_net(x);
1850 
1851 	to_put = NULL;
1852 
1853 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1854 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1855 
1856 	err = -ESRCH;
1857 	if (!x1)
1858 		goto out;
1859 
1860 	if (xfrm_state_kern(x1)) {
1861 		to_put = x1;
1862 		err = -EEXIST;
1863 		goto out;
1864 	}
1865 
1866 	if (x1->km.state == XFRM_STATE_ACQ) {
1867 		__xfrm_state_insert(x);
1868 		x = NULL;
1869 	}
1870 	err = 0;
1871 
1872 out:
1873 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1874 
1875 	if (to_put)
1876 		xfrm_state_put(to_put);
1877 
1878 	if (err)
1879 		return err;
1880 
1881 	if (!x) {
1882 		xfrm_state_delete(x1);
1883 		xfrm_state_put(x1);
1884 		return 0;
1885 	}
1886 
1887 	err = -EINVAL;
1888 	spin_lock_bh(&x1->lock);
1889 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1890 		if (x->encap && x1->encap &&
1891 		    x->encap->encap_type == x1->encap->encap_type)
1892 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1893 		else if (x->encap || x1->encap)
1894 			goto fail;
1895 
1896 		if (x->coaddr && x1->coaddr) {
1897 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1898 		}
1899 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1900 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1901 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1902 		x1->km.dying = 0;
1903 
1904 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1905 			      HRTIMER_MODE_REL_SOFT);
1906 		if (READ_ONCE(x1->curlft.use_time))
1907 			xfrm_state_check_expire(x1);
1908 
1909 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1910 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1911 
1912 			if (x->props.smark.m || x->props.smark.v)
1913 				x1->props.smark = x->props.smark;
1914 
1915 			if (x->if_id)
1916 				x1->if_id = x->if_id;
1917 
1918 			__xfrm_state_bump_genids(x1);
1919 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1920 		}
1921 
1922 		err = 0;
1923 		x->km.state = XFRM_STATE_DEAD;
1924 		__xfrm_state_put(x);
1925 	}
1926 
1927 fail:
1928 	spin_unlock_bh(&x1->lock);
1929 
1930 	xfrm_state_put(x1);
1931 
1932 	return err;
1933 }
1934 EXPORT_SYMBOL(xfrm_state_update);
1935 
1936 int xfrm_state_check_expire(struct xfrm_state *x)
1937 {
1938 	xfrm_dev_state_update_curlft(x);
1939 
1940 	if (!READ_ONCE(x->curlft.use_time))
1941 		WRITE_ONCE(x->curlft.use_time, ktime_get_real_seconds());
1942 
1943 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1944 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1945 		x->km.state = XFRM_STATE_EXPIRED;
1946 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1947 		return -EINVAL;
1948 	}
1949 
1950 	if (!x->km.dying &&
1951 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1952 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1953 		x->km.dying = 1;
1954 		km_state_expired(x, 0, 0);
1955 	}
1956 	return 0;
1957 }
1958 EXPORT_SYMBOL(xfrm_state_check_expire);
1959 
1960 struct xfrm_state *
1961 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1962 		  u8 proto, unsigned short family)
1963 {
1964 	struct xfrm_state *x;
1965 
1966 	rcu_read_lock();
1967 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1968 	rcu_read_unlock();
1969 	return x;
1970 }
1971 EXPORT_SYMBOL(xfrm_state_lookup);
1972 
1973 struct xfrm_state *
1974 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1975 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1976 			 u8 proto, unsigned short family)
1977 {
1978 	struct xfrm_state *x;
1979 
1980 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1981 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1982 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1983 	return x;
1984 }
1985 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1986 
1987 struct xfrm_state *
1988 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1989 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1990 	      const xfrm_address_t *saddr, int create, unsigned short family)
1991 {
1992 	struct xfrm_state *x;
1993 
1994 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1995 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1996 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1997 
1998 	return x;
1999 }
2000 EXPORT_SYMBOL(xfrm_find_acq);
2001 
2002 #ifdef CONFIG_XFRM_SUB_POLICY
2003 #if IS_ENABLED(CONFIG_IPV6)
2004 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
2005 static void
2006 __xfrm6_sort(void **dst, void **src, int n,
2007 	     int (*cmp)(const void *p), int maxclass)
2008 {
2009 	int count[XFRM_MAX_DEPTH] = { };
2010 	int class[XFRM_MAX_DEPTH];
2011 	int i;
2012 
2013 	for (i = 0; i < n; i++) {
2014 		int c = cmp(src[i]);
2015 
2016 		class[i] = c;
2017 		count[c]++;
2018 	}
2019 
2020 	for (i = 2; i < maxclass; i++)
2021 		count[i] += count[i - 1];
2022 
2023 	for (i = 0; i < n; i++) {
2024 		dst[count[class[i] - 1]++] = src[i];
2025 		src[i] = NULL;
2026 	}
2027 }
2028 
2029 /* Rule for xfrm_state:
2030  *
2031  * rule 1: select IPsec transport except AH
2032  * rule 2: select MIPv6 RO or inbound trigger
2033  * rule 3: select IPsec transport AH
2034  * rule 4: select IPsec tunnel
2035  * rule 5: others
2036  */
2037 static int __xfrm6_state_sort_cmp(const void *p)
2038 {
2039 	const struct xfrm_state *v = p;
2040 
2041 	switch (v->props.mode) {
2042 	case XFRM_MODE_TRANSPORT:
2043 		if (v->id.proto != IPPROTO_AH)
2044 			return 1;
2045 		else
2046 			return 3;
2047 #if IS_ENABLED(CONFIG_IPV6_MIP6)
2048 	case XFRM_MODE_ROUTEOPTIMIZATION:
2049 	case XFRM_MODE_IN_TRIGGER:
2050 		return 2;
2051 #endif
2052 	case XFRM_MODE_TUNNEL:
2053 	case XFRM_MODE_BEET:
2054 		return 4;
2055 	}
2056 	return 5;
2057 }
2058 
2059 /* Rule for xfrm_tmpl:
2060  *
2061  * rule 1: select IPsec transport
2062  * rule 2: select MIPv6 RO or inbound trigger
2063  * rule 3: select IPsec tunnel
2064  * rule 4: others
2065  */
2066 static int __xfrm6_tmpl_sort_cmp(const void *p)
2067 {
2068 	const struct xfrm_tmpl *v = p;
2069 
2070 	switch (v->mode) {
2071 	case XFRM_MODE_TRANSPORT:
2072 		return 1;
2073 #if IS_ENABLED(CONFIG_IPV6_MIP6)
2074 	case XFRM_MODE_ROUTEOPTIMIZATION:
2075 	case XFRM_MODE_IN_TRIGGER:
2076 		return 2;
2077 #endif
2078 	case XFRM_MODE_TUNNEL:
2079 	case XFRM_MODE_BEET:
2080 		return 3;
2081 	}
2082 	return 4;
2083 }
2084 #else
2085 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
2086 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
2087 
2088 static inline void
2089 __xfrm6_sort(void **dst, void **src, int n,
2090 	     int (*cmp)(const void *p), int maxclass)
2091 {
2092 	int i;
2093 
2094 	for (i = 0; i < n; i++)
2095 		dst[i] = src[i];
2096 }
2097 #endif /* CONFIG_IPV6 */
2098 
2099 void
2100 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
2101 	       unsigned short family)
2102 {
2103 	int i;
2104 
2105 	if (family == AF_INET6)
2106 		__xfrm6_sort((void **)dst, (void **)src, n,
2107 			     __xfrm6_tmpl_sort_cmp, 5);
2108 	else
2109 		for (i = 0; i < n; i++)
2110 			dst[i] = src[i];
2111 }
2112 
2113 void
2114 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
2115 		unsigned short family)
2116 {
2117 	int i;
2118 
2119 	if (family == AF_INET6)
2120 		__xfrm6_sort((void **)dst, (void **)src, n,
2121 			     __xfrm6_state_sort_cmp, 6);
2122 	else
2123 		for (i = 0; i < n; i++)
2124 			dst[i] = src[i];
2125 }
2126 #endif
2127 
2128 /* Silly enough, but I'm lazy to build resolution list */
2129 
2130 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
2131 {
2132 	unsigned int h = xfrm_seq_hash(net, seq);
2133 	struct xfrm_state *x;
2134 
2135 	hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
2136 		if (x->km.seq == seq &&
2137 		    (mark & x->mark.m) == x->mark.v &&
2138 		    x->km.state == XFRM_STATE_ACQ) {
2139 			xfrm_state_hold(x);
2140 			return x;
2141 		}
2142 	}
2143 
2144 	return NULL;
2145 }
2146 
2147 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
2148 {
2149 	struct xfrm_state *x;
2150 
2151 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2152 	x = __xfrm_find_acq_byseq(net, mark, seq);
2153 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2154 	return x;
2155 }
2156 EXPORT_SYMBOL(xfrm_find_acq_byseq);
2157 
2158 u32 xfrm_get_acqseq(void)
2159 {
2160 	u32 res;
2161 	static atomic_t acqseq;
2162 
2163 	do {
2164 		res = atomic_inc_return(&acqseq);
2165 	} while (!res);
2166 
2167 	return res;
2168 }
2169 EXPORT_SYMBOL(xfrm_get_acqseq);
2170 
2171 int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack)
2172 {
2173 	switch (proto) {
2174 	case IPPROTO_AH:
2175 	case IPPROTO_ESP:
2176 		break;
2177 
2178 	case IPPROTO_COMP:
2179 		/* IPCOMP spi is 16-bits. */
2180 		if (max >= 0x10000) {
2181 			NL_SET_ERR_MSG(extack, "IPCOMP SPI must be <= 65535");
2182 			return -EINVAL;
2183 		}
2184 		break;
2185 
2186 	default:
2187 		NL_SET_ERR_MSG(extack, "Invalid protocol, must be one of AH, ESP, IPCOMP");
2188 		return -EINVAL;
2189 	}
2190 
2191 	if (min > max) {
2192 		NL_SET_ERR_MSG(extack, "Invalid SPI range: min > max");
2193 		return -EINVAL;
2194 	}
2195 
2196 	return 0;
2197 }
2198 EXPORT_SYMBOL(verify_spi_info);
2199 
2200 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high,
2201 		   struct netlink_ext_ack *extack)
2202 {
2203 	struct net *net = xs_net(x);
2204 	unsigned int h;
2205 	struct xfrm_state *x0;
2206 	int err = -ENOENT;
2207 	__be32 minspi = htonl(low);
2208 	__be32 maxspi = htonl(high);
2209 	__be32 newspi = 0;
2210 	u32 mark = x->mark.v & x->mark.m;
2211 
2212 	spin_lock_bh(&x->lock);
2213 	if (x->km.state == XFRM_STATE_DEAD) {
2214 		NL_SET_ERR_MSG(extack, "Target ACQUIRE is in DEAD state");
2215 		goto unlock;
2216 	}
2217 
2218 	err = 0;
2219 	if (x->id.spi)
2220 		goto unlock;
2221 
2222 	err = -ENOENT;
2223 
2224 	if (minspi == maxspi) {
2225 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2226 		if (x0) {
2227 			NL_SET_ERR_MSG(extack, "Requested SPI is already in use");
2228 			xfrm_state_put(x0);
2229 			goto unlock;
2230 		}
2231 		newspi = minspi;
2232 	} else {
2233 		u32 spi = 0;
2234 		for (h = 0; h < high-low+1; h++) {
2235 			spi = get_random_u32_inclusive(low, high);
2236 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2237 			if (x0 == NULL) {
2238 				newspi = htonl(spi);
2239 				break;
2240 			}
2241 			xfrm_state_put(x0);
2242 		}
2243 	}
2244 	if (newspi) {
2245 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2246 		x->id.spi = newspi;
2247 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2248 		XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
2249 				  x->xso.type);
2250 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2251 
2252 		err = 0;
2253 	} else {
2254 		NL_SET_ERR_MSG(extack, "No SPI available in the requested range");
2255 	}
2256 
2257 unlock:
2258 	spin_unlock_bh(&x->lock);
2259 
2260 	return err;
2261 }
2262 EXPORT_SYMBOL(xfrm_alloc_spi);
2263 
2264 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2265 				      struct xfrm_address_filter *filter)
2266 {
2267 	if (filter) {
2268 		if ((filter->family == AF_INET ||
2269 		     filter->family == AF_INET6) &&
2270 		    x->props.family != filter->family)
2271 			return false;
2272 
2273 		return addr_match(&x->props.saddr, &filter->saddr,
2274 				  filter->splen) &&
2275 		       addr_match(&x->id.daddr, &filter->daddr,
2276 				  filter->dplen);
2277 	}
2278 	return true;
2279 }
2280 
2281 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2282 		    int (*func)(struct xfrm_state *, int, void*),
2283 		    void *data)
2284 {
2285 	struct xfrm_state *state;
2286 	struct xfrm_state_walk *x;
2287 	int err = 0;
2288 
2289 	if (walk->seq != 0 && list_empty(&walk->all))
2290 		return 0;
2291 
2292 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2293 	if (list_empty(&walk->all))
2294 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2295 	else
2296 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2297 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2298 		if (x->state == XFRM_STATE_DEAD)
2299 			continue;
2300 		state = container_of(x, struct xfrm_state, km);
2301 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2302 			continue;
2303 		if (!__xfrm_state_filter_match(state, walk->filter))
2304 			continue;
2305 		err = func(state, walk->seq, data);
2306 		if (err) {
2307 			list_move_tail(&walk->all, &x->all);
2308 			goto out;
2309 		}
2310 		walk->seq++;
2311 	}
2312 	if (walk->seq == 0) {
2313 		err = -ENOENT;
2314 		goto out;
2315 	}
2316 	list_del_init(&walk->all);
2317 out:
2318 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2319 	return err;
2320 }
2321 EXPORT_SYMBOL(xfrm_state_walk);
2322 
2323 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2324 			  struct xfrm_address_filter *filter)
2325 {
2326 	INIT_LIST_HEAD(&walk->all);
2327 	walk->proto = proto;
2328 	walk->state = XFRM_STATE_DEAD;
2329 	walk->seq = 0;
2330 	walk->filter = filter;
2331 }
2332 EXPORT_SYMBOL(xfrm_state_walk_init);
2333 
2334 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2335 {
2336 	kfree(walk->filter);
2337 
2338 	if (list_empty(&walk->all))
2339 		return;
2340 
2341 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2342 	list_del(&walk->all);
2343 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2344 }
2345 EXPORT_SYMBOL(xfrm_state_walk_done);
2346 
2347 static void xfrm_replay_timer_handler(struct timer_list *t)
2348 {
2349 	struct xfrm_state *x = from_timer(x, t, rtimer);
2350 
2351 	spin_lock(&x->lock);
2352 
2353 	if (x->km.state == XFRM_STATE_VALID) {
2354 		if (xfrm_aevent_is_on(xs_net(x)))
2355 			xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2356 		else
2357 			x->xflags |= XFRM_TIME_DEFER;
2358 	}
2359 
2360 	spin_unlock(&x->lock);
2361 }
2362 
2363 static LIST_HEAD(xfrm_km_list);
2364 
2365 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2366 {
2367 	struct xfrm_mgr *km;
2368 
2369 	rcu_read_lock();
2370 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2371 		if (km->notify_policy)
2372 			km->notify_policy(xp, dir, c);
2373 	rcu_read_unlock();
2374 }
2375 
2376 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2377 {
2378 	struct xfrm_mgr *km;
2379 	rcu_read_lock();
2380 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2381 		if (km->notify)
2382 			km->notify(x, c);
2383 	rcu_read_unlock();
2384 }
2385 
2386 EXPORT_SYMBOL(km_policy_notify);
2387 EXPORT_SYMBOL(km_state_notify);
2388 
2389 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2390 {
2391 	struct km_event c;
2392 
2393 	c.data.hard = hard;
2394 	c.portid = portid;
2395 	c.event = XFRM_MSG_EXPIRE;
2396 	km_state_notify(x, &c);
2397 }
2398 
2399 EXPORT_SYMBOL(km_state_expired);
2400 /*
2401  * We send to all registered managers regardless of failure
2402  * We are happy with one success
2403 */
2404 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2405 {
2406 	int err = -EINVAL, acqret;
2407 	struct xfrm_mgr *km;
2408 
2409 	rcu_read_lock();
2410 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2411 		acqret = km->acquire(x, t, pol);
2412 		if (!acqret)
2413 			err = acqret;
2414 	}
2415 	rcu_read_unlock();
2416 	return err;
2417 }
2418 EXPORT_SYMBOL(km_query);
2419 
2420 static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2421 {
2422 	int err = -EINVAL;
2423 	struct xfrm_mgr *km;
2424 
2425 	rcu_read_lock();
2426 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2427 		if (km->new_mapping)
2428 			err = km->new_mapping(x, ipaddr, sport);
2429 		if (!err)
2430 			break;
2431 	}
2432 	rcu_read_unlock();
2433 	return err;
2434 }
2435 
2436 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2437 {
2438 	int ret = 0;
2439 
2440 	if (x->mapping_maxage) {
2441 		if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage ||
2442 		    x->new_mapping_sport != sport) {
2443 			x->new_mapping_sport = sport;
2444 			x->new_mapping = jiffies / HZ;
2445 			ret = __km_new_mapping(x, ipaddr, sport);
2446 		}
2447 	} else {
2448 		ret = __km_new_mapping(x, ipaddr, sport);
2449 	}
2450 
2451 	return ret;
2452 }
2453 EXPORT_SYMBOL(km_new_mapping);
2454 
2455 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2456 {
2457 	struct km_event c;
2458 
2459 	c.data.hard = hard;
2460 	c.portid = portid;
2461 	c.event = XFRM_MSG_POLEXPIRE;
2462 	km_policy_notify(pol, dir, &c);
2463 }
2464 EXPORT_SYMBOL(km_policy_expired);
2465 
2466 #ifdef CONFIG_XFRM_MIGRATE
2467 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2468 	       const struct xfrm_migrate *m, int num_migrate,
2469 	       const struct xfrm_kmaddress *k,
2470 	       const struct xfrm_encap_tmpl *encap)
2471 {
2472 	int err = -EINVAL;
2473 	int ret;
2474 	struct xfrm_mgr *km;
2475 
2476 	rcu_read_lock();
2477 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2478 		if (km->migrate) {
2479 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2480 					  encap);
2481 			if (!ret)
2482 				err = ret;
2483 		}
2484 	}
2485 	rcu_read_unlock();
2486 	return err;
2487 }
2488 EXPORT_SYMBOL(km_migrate);
2489 #endif
2490 
2491 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2492 {
2493 	int err = -EINVAL;
2494 	int ret;
2495 	struct xfrm_mgr *km;
2496 
2497 	rcu_read_lock();
2498 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2499 		if (km->report) {
2500 			ret = km->report(net, proto, sel, addr);
2501 			if (!ret)
2502 				err = ret;
2503 		}
2504 	}
2505 	rcu_read_unlock();
2506 	return err;
2507 }
2508 EXPORT_SYMBOL(km_report);
2509 
2510 static bool km_is_alive(const struct km_event *c)
2511 {
2512 	struct xfrm_mgr *km;
2513 	bool is_alive = false;
2514 
2515 	rcu_read_lock();
2516 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2517 		if (km->is_alive && km->is_alive(c)) {
2518 			is_alive = true;
2519 			break;
2520 		}
2521 	}
2522 	rcu_read_unlock();
2523 
2524 	return is_alive;
2525 }
2526 
2527 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2528 static DEFINE_SPINLOCK(xfrm_translator_lock);
2529 static struct xfrm_translator __rcu *xfrm_translator;
2530 
2531 struct xfrm_translator *xfrm_get_translator(void)
2532 {
2533 	struct xfrm_translator *xtr;
2534 
2535 	rcu_read_lock();
2536 	xtr = rcu_dereference(xfrm_translator);
2537 	if (unlikely(!xtr))
2538 		goto out;
2539 	if (!try_module_get(xtr->owner))
2540 		xtr = NULL;
2541 out:
2542 	rcu_read_unlock();
2543 	return xtr;
2544 }
2545 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2546 
2547 void xfrm_put_translator(struct xfrm_translator *xtr)
2548 {
2549 	module_put(xtr->owner);
2550 }
2551 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2552 
2553 int xfrm_register_translator(struct xfrm_translator *xtr)
2554 {
2555 	int err = 0;
2556 
2557 	spin_lock_bh(&xfrm_translator_lock);
2558 	if (unlikely(xfrm_translator != NULL))
2559 		err = -EEXIST;
2560 	else
2561 		rcu_assign_pointer(xfrm_translator, xtr);
2562 	spin_unlock_bh(&xfrm_translator_lock);
2563 
2564 	return err;
2565 }
2566 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2567 
2568 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2569 {
2570 	int err = 0;
2571 
2572 	spin_lock_bh(&xfrm_translator_lock);
2573 	if (likely(xfrm_translator != NULL)) {
2574 		if (rcu_access_pointer(xfrm_translator) != xtr)
2575 			err = -EINVAL;
2576 		else
2577 			RCU_INIT_POINTER(xfrm_translator, NULL);
2578 	}
2579 	spin_unlock_bh(&xfrm_translator_lock);
2580 	synchronize_rcu();
2581 
2582 	return err;
2583 }
2584 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2585 #endif
2586 
2587 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2588 {
2589 	int err;
2590 	u8 *data;
2591 	struct xfrm_mgr *km;
2592 	struct xfrm_policy *pol = NULL;
2593 
2594 	if (sockptr_is_null(optval) && !optlen) {
2595 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2596 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2597 		__sk_dst_reset(sk);
2598 		return 0;
2599 	}
2600 
2601 	if (optlen <= 0 || optlen > PAGE_SIZE)
2602 		return -EMSGSIZE;
2603 
2604 	data = memdup_sockptr(optval, optlen);
2605 	if (IS_ERR(data))
2606 		return PTR_ERR(data);
2607 
2608 	if (in_compat_syscall()) {
2609 		struct xfrm_translator *xtr = xfrm_get_translator();
2610 
2611 		if (!xtr) {
2612 			kfree(data);
2613 			return -EOPNOTSUPP;
2614 		}
2615 
2616 		err = xtr->xlate_user_policy_sockptr(&data, optlen);
2617 		xfrm_put_translator(xtr);
2618 		if (err) {
2619 			kfree(data);
2620 			return err;
2621 		}
2622 	}
2623 
2624 	err = -EINVAL;
2625 	rcu_read_lock();
2626 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2627 		pol = km->compile_policy(sk, optname, data,
2628 					 optlen, &err);
2629 		if (err >= 0)
2630 			break;
2631 	}
2632 	rcu_read_unlock();
2633 
2634 	if (err >= 0) {
2635 		xfrm_sk_policy_insert(sk, err, pol);
2636 		xfrm_pol_put(pol);
2637 		__sk_dst_reset(sk);
2638 		err = 0;
2639 	}
2640 
2641 	kfree(data);
2642 	return err;
2643 }
2644 EXPORT_SYMBOL(xfrm_user_policy);
2645 
2646 static DEFINE_SPINLOCK(xfrm_km_lock);
2647 
2648 void xfrm_register_km(struct xfrm_mgr *km)
2649 {
2650 	spin_lock_bh(&xfrm_km_lock);
2651 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2652 	spin_unlock_bh(&xfrm_km_lock);
2653 }
2654 EXPORT_SYMBOL(xfrm_register_km);
2655 
2656 void xfrm_unregister_km(struct xfrm_mgr *km)
2657 {
2658 	spin_lock_bh(&xfrm_km_lock);
2659 	list_del_rcu(&km->list);
2660 	spin_unlock_bh(&xfrm_km_lock);
2661 	synchronize_rcu();
2662 }
2663 EXPORT_SYMBOL(xfrm_unregister_km);
2664 
2665 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2666 {
2667 	int err = 0;
2668 
2669 	if (WARN_ON(afinfo->family >= NPROTO))
2670 		return -EAFNOSUPPORT;
2671 
2672 	spin_lock_bh(&xfrm_state_afinfo_lock);
2673 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2674 		err = -EEXIST;
2675 	else
2676 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2677 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2678 	return err;
2679 }
2680 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2681 
2682 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2683 {
2684 	int err = 0, family = afinfo->family;
2685 
2686 	if (WARN_ON(family >= NPROTO))
2687 		return -EAFNOSUPPORT;
2688 
2689 	spin_lock_bh(&xfrm_state_afinfo_lock);
2690 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2691 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2692 			err = -EINVAL;
2693 		else
2694 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2695 	}
2696 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2697 	synchronize_rcu();
2698 	return err;
2699 }
2700 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2701 
2702 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2703 {
2704 	if (unlikely(family >= NPROTO))
2705 		return NULL;
2706 
2707 	return rcu_dereference(xfrm_state_afinfo[family]);
2708 }
2709 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2710 
2711 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2712 {
2713 	struct xfrm_state_afinfo *afinfo;
2714 	if (unlikely(family >= NPROTO))
2715 		return NULL;
2716 	rcu_read_lock();
2717 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2718 	if (unlikely(!afinfo))
2719 		rcu_read_unlock();
2720 	return afinfo;
2721 }
2722 
2723 void xfrm_flush_gc(void)
2724 {
2725 	flush_work(&xfrm_state_gc_work);
2726 }
2727 EXPORT_SYMBOL(xfrm_flush_gc);
2728 
2729 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2730 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2731 {
2732 	if (x->tunnel) {
2733 		struct xfrm_state *t = x->tunnel;
2734 
2735 		if (atomic_read(&t->tunnel_users) == 2)
2736 			xfrm_state_delete(t);
2737 		atomic_dec(&t->tunnel_users);
2738 		xfrm_state_put_sync(t);
2739 		x->tunnel = NULL;
2740 	}
2741 }
2742 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2743 
2744 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2745 {
2746 	const struct xfrm_type *type = READ_ONCE(x->type);
2747 	struct crypto_aead *aead;
2748 	u32 blksize, net_adj = 0;
2749 
2750 	if (x->km.state != XFRM_STATE_VALID ||
2751 	    !type || type->proto != IPPROTO_ESP)
2752 		return mtu - x->props.header_len;
2753 
2754 	aead = x->data;
2755 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2756 
2757 	switch (x->props.mode) {
2758 	case XFRM_MODE_TRANSPORT:
2759 	case XFRM_MODE_BEET:
2760 		if (x->props.family == AF_INET)
2761 			net_adj = sizeof(struct iphdr);
2762 		else if (x->props.family == AF_INET6)
2763 			net_adj = sizeof(struct ipv6hdr);
2764 		break;
2765 	case XFRM_MODE_TUNNEL:
2766 		break;
2767 	default:
2768 		WARN_ON_ONCE(1);
2769 		break;
2770 	}
2771 
2772 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2773 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2774 }
2775 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2776 
2777 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload,
2778 		      struct netlink_ext_ack *extack)
2779 {
2780 	const struct xfrm_mode *inner_mode;
2781 	const struct xfrm_mode *outer_mode;
2782 	int family = x->props.family;
2783 	int err;
2784 
2785 	if (family == AF_INET &&
2786 	    READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
2787 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2788 
2789 	err = -EPROTONOSUPPORT;
2790 
2791 	if (x->sel.family != AF_UNSPEC) {
2792 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2793 		if (inner_mode == NULL) {
2794 			NL_SET_ERR_MSG(extack, "Requested mode not found");
2795 			goto error;
2796 		}
2797 
2798 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2799 		    family != x->sel.family) {
2800 			NL_SET_ERR_MSG(extack, "Only tunnel modes can accommodate a change of family");
2801 			goto error;
2802 		}
2803 
2804 		x->inner_mode = *inner_mode;
2805 	} else {
2806 		const struct xfrm_mode *inner_mode_iaf;
2807 		int iafamily = AF_INET;
2808 
2809 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2810 		if (inner_mode == NULL) {
2811 			NL_SET_ERR_MSG(extack, "Requested mode not found");
2812 			goto error;
2813 		}
2814 
2815 		x->inner_mode = *inner_mode;
2816 
2817 		if (x->props.family == AF_INET)
2818 			iafamily = AF_INET6;
2819 
2820 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2821 		if (inner_mode_iaf) {
2822 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2823 				x->inner_mode_iaf = *inner_mode_iaf;
2824 		}
2825 	}
2826 
2827 	x->type = xfrm_get_type(x->id.proto, family);
2828 	if (x->type == NULL) {
2829 		NL_SET_ERR_MSG(extack, "Requested type not found");
2830 		goto error;
2831 	}
2832 
2833 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2834 
2835 	err = x->type->init_state(x, extack);
2836 	if (err)
2837 		goto error;
2838 
2839 	outer_mode = xfrm_get_mode(x->props.mode, family);
2840 	if (!outer_mode) {
2841 		NL_SET_ERR_MSG(extack, "Requested mode not found");
2842 		err = -EPROTONOSUPPORT;
2843 		goto error;
2844 	}
2845 
2846 	x->outer_mode = *outer_mode;
2847 	if (init_replay) {
2848 		err = xfrm_init_replay(x, extack);
2849 		if (err)
2850 			goto error;
2851 	}
2852 
2853 error:
2854 	return err;
2855 }
2856 
2857 EXPORT_SYMBOL(__xfrm_init_state);
2858 
2859 int xfrm_init_state(struct xfrm_state *x)
2860 {
2861 	int err;
2862 
2863 	err = __xfrm_init_state(x, true, false, NULL);
2864 	if (!err)
2865 		x->km.state = XFRM_STATE_VALID;
2866 
2867 	return err;
2868 }
2869 
2870 EXPORT_SYMBOL(xfrm_init_state);
2871 
2872 int __net_init xfrm_state_init(struct net *net)
2873 {
2874 	unsigned int sz;
2875 
2876 	if (net_eq(net, &init_net))
2877 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2878 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2879 
2880 	INIT_LIST_HEAD(&net->xfrm.state_all);
2881 
2882 	sz = sizeof(struct hlist_head) * 8;
2883 
2884 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2885 	if (!net->xfrm.state_bydst)
2886 		goto out_bydst;
2887 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2888 	if (!net->xfrm.state_bysrc)
2889 		goto out_bysrc;
2890 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2891 	if (!net->xfrm.state_byspi)
2892 		goto out_byspi;
2893 	net->xfrm.state_byseq = xfrm_hash_alloc(sz);
2894 	if (!net->xfrm.state_byseq)
2895 		goto out_byseq;
2896 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2897 
2898 	net->xfrm.state_num = 0;
2899 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2900 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2901 	seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2902 			       &net->xfrm.xfrm_state_lock);
2903 	return 0;
2904 
2905 out_byseq:
2906 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2907 out_byspi:
2908 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2909 out_bysrc:
2910 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2911 out_bydst:
2912 	return -ENOMEM;
2913 }
2914 
2915 void xfrm_state_fini(struct net *net)
2916 {
2917 	unsigned int sz;
2918 
2919 	flush_work(&net->xfrm.state_hash_work);
2920 	flush_work(&xfrm_state_gc_work);
2921 	xfrm_state_flush(net, 0, false, true);
2922 
2923 	WARN_ON(!list_empty(&net->xfrm.state_all));
2924 
2925 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2926 	WARN_ON(!hlist_empty(net->xfrm.state_byseq));
2927 	xfrm_hash_free(net->xfrm.state_byseq, sz);
2928 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2929 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2930 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2931 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2932 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2933 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2934 }
2935 
2936 #ifdef CONFIG_AUDITSYSCALL
2937 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2938 				     struct audit_buffer *audit_buf)
2939 {
2940 	struct xfrm_sec_ctx *ctx = x->security;
2941 	u32 spi = ntohl(x->id.spi);
2942 
2943 	if (ctx)
2944 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2945 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2946 
2947 	switch (x->props.family) {
2948 	case AF_INET:
2949 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2950 				 &x->props.saddr.a4, &x->id.daddr.a4);
2951 		break;
2952 	case AF_INET6:
2953 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2954 				 x->props.saddr.a6, x->id.daddr.a6);
2955 		break;
2956 	}
2957 
2958 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2959 }
2960 
2961 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2962 				      struct audit_buffer *audit_buf)
2963 {
2964 	const struct iphdr *iph4;
2965 	const struct ipv6hdr *iph6;
2966 
2967 	switch (family) {
2968 	case AF_INET:
2969 		iph4 = ip_hdr(skb);
2970 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2971 				 &iph4->saddr, &iph4->daddr);
2972 		break;
2973 	case AF_INET6:
2974 		iph6 = ipv6_hdr(skb);
2975 		audit_log_format(audit_buf,
2976 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2977 				 &iph6->saddr, &iph6->daddr,
2978 				 iph6->flow_lbl[0] & 0x0f,
2979 				 iph6->flow_lbl[1],
2980 				 iph6->flow_lbl[2]);
2981 		break;
2982 	}
2983 }
2984 
2985 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2986 {
2987 	struct audit_buffer *audit_buf;
2988 
2989 	audit_buf = xfrm_audit_start("SAD-add");
2990 	if (audit_buf == NULL)
2991 		return;
2992 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2993 	xfrm_audit_helper_sainfo(x, audit_buf);
2994 	audit_log_format(audit_buf, " res=%u", result);
2995 	audit_log_end(audit_buf);
2996 }
2997 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2998 
2999 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
3000 {
3001 	struct audit_buffer *audit_buf;
3002 
3003 	audit_buf = xfrm_audit_start("SAD-delete");
3004 	if (audit_buf == NULL)
3005 		return;
3006 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3007 	xfrm_audit_helper_sainfo(x, audit_buf);
3008 	audit_log_format(audit_buf, " res=%u", result);
3009 	audit_log_end(audit_buf);
3010 }
3011 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
3012 
3013 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
3014 				      struct sk_buff *skb)
3015 {
3016 	struct audit_buffer *audit_buf;
3017 	u32 spi;
3018 
3019 	audit_buf = xfrm_audit_start("SA-replay-overflow");
3020 	if (audit_buf == NULL)
3021 		return;
3022 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3023 	/* don't record the sequence number because it's inherent in this kind
3024 	 * of audit message */
3025 	spi = ntohl(x->id.spi);
3026 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3027 	audit_log_end(audit_buf);
3028 }
3029 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
3030 
3031 void xfrm_audit_state_replay(struct xfrm_state *x,
3032 			     struct sk_buff *skb, __be32 net_seq)
3033 {
3034 	struct audit_buffer *audit_buf;
3035 	u32 spi;
3036 
3037 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
3038 	if (audit_buf == NULL)
3039 		return;
3040 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3041 	spi = ntohl(x->id.spi);
3042 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3043 			 spi, spi, ntohl(net_seq));
3044 	audit_log_end(audit_buf);
3045 }
3046 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
3047 
3048 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
3049 {
3050 	struct audit_buffer *audit_buf;
3051 
3052 	audit_buf = xfrm_audit_start("SA-notfound");
3053 	if (audit_buf == NULL)
3054 		return;
3055 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3056 	audit_log_end(audit_buf);
3057 }
3058 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
3059 
3060 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
3061 			       __be32 net_spi, __be32 net_seq)
3062 {
3063 	struct audit_buffer *audit_buf;
3064 	u32 spi;
3065 
3066 	audit_buf = xfrm_audit_start("SA-notfound");
3067 	if (audit_buf == NULL)
3068 		return;
3069 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3070 	spi = ntohl(net_spi);
3071 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3072 			 spi, spi, ntohl(net_seq));
3073 	audit_log_end(audit_buf);
3074 }
3075 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
3076 
3077 void xfrm_audit_state_icvfail(struct xfrm_state *x,
3078 			      struct sk_buff *skb, u8 proto)
3079 {
3080 	struct audit_buffer *audit_buf;
3081 	__be32 net_spi;
3082 	__be32 net_seq;
3083 
3084 	audit_buf = xfrm_audit_start("SA-icv-failure");
3085 	if (audit_buf == NULL)
3086 		return;
3087 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3088 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
3089 		u32 spi = ntohl(net_spi);
3090 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3091 				 spi, spi, ntohl(net_seq));
3092 	}
3093 	audit_log_end(audit_buf);
3094 }
3095 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
3096 #endif /* CONFIG_AUDITSYSCALL */
3097