• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/workqueue.h>
18 #include <net/xfrm.h>
19 #include <linux/pfkeyv2.h>
20 #include <linux/ipsec.h>
21 #include <linux/module.h>
22 #include <linux/cache.h>
23 #include <linux/audit.h>
24 #include <linux/uaccess.h>
25 #include <linux/ktime.h>
26 #include <linux/slab.h>
27 #include <linux/interrupt.h>
28 #include <linux/kernel.h>
29 
30 #include <crypto/aead.h>
31 
32 #include "xfrm_hash.h"
33 
34 #define xfrm_state_deref_prot(table, net) \
35 	rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
36 
37 static void xfrm_state_gc_task(struct work_struct *work);
38 
39 /* Each xfrm_state may be linked to two tables:
40 
41    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
42    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
43       destination/tunnel endpoint. (output)
44  */
45 
46 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
47 static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
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 
xfrm_state_hold_rcu(struct xfrm_state __rcu * x)53 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54 {
55 	return refcount_inc_not_zero(&x->refcnt);
56 }
57 
xfrm_dst_hash(struct net * net,const xfrm_address_t * daddr,const xfrm_address_t * saddr,u32 reqid,unsigned short family)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 
xfrm_src_hash(struct net * net,const xfrm_address_t * daddr,const xfrm_address_t * saddr,unsigned short family)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
xfrm_spi_hash(struct net * net,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)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 
xfrm_hash_transfer(struct hlist_head * list,struct hlist_head * ndsttable,struct hlist_head * nsrctable,struct hlist_head * nspitable,unsigned int nhashmask)82 static void xfrm_hash_transfer(struct hlist_head *list,
83 			       struct hlist_head *ndsttable,
84 			       struct hlist_head *nsrctable,
85 			       struct hlist_head *nspitable,
86 			       unsigned int nhashmask)
87 {
88 	struct hlist_node *tmp;
89 	struct xfrm_state *x;
90 
91 	hlist_for_each_entry_safe(x, tmp, list, bydst) {
92 		unsigned int h;
93 
94 		h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
95 				    x->props.reqid, x->props.family,
96 				    nhashmask);
97 		hlist_add_head_rcu(&x->bydst, ndsttable + h);
98 
99 		h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
100 				    x->props.family,
101 				    nhashmask);
102 		hlist_add_head_rcu(&x->bysrc, nsrctable + h);
103 
104 		if (x->id.spi) {
105 			h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
106 					    x->id.proto, x->props.family,
107 					    nhashmask);
108 			hlist_add_head_rcu(&x->byspi, nspitable + h);
109 		}
110 	}
111 }
112 
xfrm_hash_new_size(unsigned int state_hmask)113 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
114 {
115 	return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
116 }
117 
xfrm_hash_resize(struct work_struct * work)118 static void xfrm_hash_resize(struct work_struct *work)
119 {
120 	struct net *net = container_of(work, struct net, xfrm.state_hash_work);
121 	struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
122 	unsigned long nsize, osize;
123 	unsigned int nhashmask, ohashmask;
124 	int i;
125 
126 	nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
127 	ndst = xfrm_hash_alloc(nsize);
128 	if (!ndst)
129 		return;
130 	nsrc = xfrm_hash_alloc(nsize);
131 	if (!nsrc) {
132 		xfrm_hash_free(ndst, nsize);
133 		return;
134 	}
135 	nspi = xfrm_hash_alloc(nsize);
136 	if (!nspi) {
137 		xfrm_hash_free(ndst, nsize);
138 		xfrm_hash_free(nsrc, nsize);
139 		return;
140 	}
141 
142 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
143 	write_seqcount_begin(&xfrm_state_hash_generation);
144 
145 	nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
146 	odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
147 	for (i = net->xfrm.state_hmask; i >= 0; i--)
148 		xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
149 
150 	osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
151 	ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
152 	ohashmask = net->xfrm.state_hmask;
153 
154 	rcu_assign_pointer(net->xfrm.state_bydst, ndst);
155 	rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
156 	rcu_assign_pointer(net->xfrm.state_byspi, nspi);
157 	net->xfrm.state_hmask = nhashmask;
158 
159 	write_seqcount_end(&xfrm_state_hash_generation);
160 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
161 
162 	osize = (ohashmask + 1) * sizeof(struct hlist_head);
163 
164 	synchronize_rcu();
165 
166 	xfrm_hash_free(odst, osize);
167 	xfrm_hash_free(osrc, osize);
168 	xfrm_hash_free(ospi, osize);
169 }
170 
171 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
172 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
173 
174 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
175 
176 int __xfrm_state_delete(struct xfrm_state *x);
177 
178 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
179 static bool km_is_alive(const struct km_event *c);
180 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
181 
xfrm_register_type(const struct xfrm_type * type,unsigned short family)182 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
183 {
184 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
185 	int err = 0;
186 
187 	if (!afinfo)
188 		return -EAFNOSUPPORT;
189 
190 #define X(afi, T, name) do {			\
191 		WARN_ON((afi)->type_ ## name);	\
192 		(afi)->type_ ## name = (T);	\
193 	} while (0)
194 
195 	switch (type->proto) {
196 	case IPPROTO_COMP:
197 		X(afinfo, type, comp);
198 		break;
199 	case IPPROTO_AH:
200 		X(afinfo, type, ah);
201 		break;
202 	case IPPROTO_ESP:
203 		X(afinfo, type, esp);
204 		break;
205 	case IPPROTO_IPIP:
206 		X(afinfo, type, ipip);
207 		break;
208 	case IPPROTO_DSTOPTS:
209 		X(afinfo, type, dstopts);
210 		break;
211 	case IPPROTO_ROUTING:
212 		X(afinfo, type, routing);
213 		break;
214 	case IPPROTO_IPV6:
215 		X(afinfo, type, ipip6);
216 		break;
217 	default:
218 		WARN_ON(1);
219 		err = -EPROTONOSUPPORT;
220 		break;
221 	}
222 #undef X
223 	rcu_read_unlock();
224 	return err;
225 }
226 EXPORT_SYMBOL(xfrm_register_type);
227 
xfrm_unregister_type(const struct xfrm_type * type,unsigned short family)228 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
229 {
230 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
231 
232 	if (unlikely(afinfo == NULL))
233 		return;
234 
235 #define X(afi, T, name) do {				\
236 		WARN_ON((afi)->type_ ## name != (T));	\
237 		(afi)->type_ ## name = NULL;		\
238 	} while (0)
239 
240 	switch (type->proto) {
241 	case IPPROTO_COMP:
242 		X(afinfo, type, comp);
243 		break;
244 	case IPPROTO_AH:
245 		X(afinfo, type, ah);
246 		break;
247 	case IPPROTO_ESP:
248 		X(afinfo, type, esp);
249 		break;
250 	case IPPROTO_IPIP:
251 		X(afinfo, type, ipip);
252 		break;
253 	case IPPROTO_DSTOPTS:
254 		X(afinfo, type, dstopts);
255 		break;
256 	case IPPROTO_ROUTING:
257 		X(afinfo, type, routing);
258 		break;
259 	case IPPROTO_IPV6:
260 		X(afinfo, type, ipip6);
261 		break;
262 	default:
263 		WARN_ON(1);
264 		break;
265 	}
266 #undef X
267 	rcu_read_unlock();
268 }
269 EXPORT_SYMBOL(xfrm_unregister_type);
270 
xfrm_get_type(u8 proto,unsigned short family)271 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
272 {
273 	const struct xfrm_type *type = NULL;
274 	struct xfrm_state_afinfo *afinfo;
275 	int modload_attempted = 0;
276 
277 retry:
278 	afinfo = xfrm_state_get_afinfo(family);
279 	if (unlikely(afinfo == NULL))
280 		return NULL;
281 
282 	switch (proto) {
283 	case IPPROTO_COMP:
284 		type = afinfo->type_comp;
285 		break;
286 	case IPPROTO_AH:
287 		type = afinfo->type_ah;
288 		break;
289 	case IPPROTO_ESP:
290 		type = afinfo->type_esp;
291 		break;
292 	case IPPROTO_IPIP:
293 		type = afinfo->type_ipip;
294 		break;
295 	case IPPROTO_DSTOPTS:
296 		type = afinfo->type_dstopts;
297 		break;
298 	case IPPROTO_ROUTING:
299 		type = afinfo->type_routing;
300 		break;
301 	case IPPROTO_IPV6:
302 		type = afinfo->type_ipip6;
303 		break;
304 	default:
305 		break;
306 	}
307 
308 	if (unlikely(type && !try_module_get(type->owner)))
309 		type = NULL;
310 
311 	rcu_read_unlock();
312 
313 	if (!type && !modload_attempted) {
314 		request_module("xfrm-type-%d-%d", family, proto);
315 		modload_attempted = 1;
316 		goto retry;
317 	}
318 
319 	return type;
320 }
321 
xfrm_put_type(const struct xfrm_type * type)322 static void xfrm_put_type(const struct xfrm_type *type)
323 {
324 	module_put(type->owner);
325 }
326 
xfrm_register_type_offload(const struct xfrm_type_offload * type,unsigned short family)327 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
328 			       unsigned short family)
329 {
330 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
331 	int err = 0;
332 
333 	if (unlikely(afinfo == NULL))
334 		return -EAFNOSUPPORT;
335 
336 	switch (type->proto) {
337 	case IPPROTO_ESP:
338 		WARN_ON(afinfo->type_offload_esp);
339 		afinfo->type_offload_esp = type;
340 		break;
341 	default:
342 		WARN_ON(1);
343 		err = -EPROTONOSUPPORT;
344 		break;
345 	}
346 
347 	rcu_read_unlock();
348 	return err;
349 }
350 EXPORT_SYMBOL(xfrm_register_type_offload);
351 
xfrm_unregister_type_offload(const struct xfrm_type_offload * type,unsigned short family)352 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
353 				  unsigned short family)
354 {
355 	struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
356 
357 	if (unlikely(afinfo == NULL))
358 		return;
359 
360 	switch (type->proto) {
361 	case IPPROTO_ESP:
362 		WARN_ON(afinfo->type_offload_esp != type);
363 		afinfo->type_offload_esp = NULL;
364 		break;
365 	default:
366 		WARN_ON(1);
367 		break;
368 	}
369 	rcu_read_unlock();
370 }
371 EXPORT_SYMBOL(xfrm_unregister_type_offload);
372 
373 static const struct xfrm_type_offload *
xfrm_get_type_offload(u8 proto,unsigned short family,bool try_load)374 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
375 {
376 	const struct xfrm_type_offload *type = NULL;
377 	struct xfrm_state_afinfo *afinfo;
378 
379 retry:
380 	afinfo = xfrm_state_get_afinfo(family);
381 	if (unlikely(afinfo == NULL))
382 		return NULL;
383 
384 	switch (proto) {
385 	case IPPROTO_ESP:
386 		type = afinfo->type_offload_esp;
387 		break;
388 	default:
389 		break;
390 	}
391 
392 	if ((type && !try_module_get(type->owner)))
393 		type = NULL;
394 
395 	rcu_read_unlock();
396 
397 	if (!type && try_load) {
398 		request_module("xfrm-offload-%d-%d", family, proto);
399 		try_load = false;
400 		goto retry;
401 	}
402 
403 	return type;
404 }
405 
xfrm_put_type_offload(const struct xfrm_type_offload * type)406 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
407 {
408 	module_put(type->owner);
409 }
410 
411 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
412 	[XFRM_MODE_BEET] = {
413 		.encap = XFRM_MODE_BEET,
414 		.flags = XFRM_MODE_FLAG_TUNNEL,
415 		.family = AF_INET,
416 	},
417 	[XFRM_MODE_TRANSPORT] = {
418 		.encap = XFRM_MODE_TRANSPORT,
419 		.family = AF_INET,
420 	},
421 	[XFRM_MODE_TUNNEL] = {
422 		.encap = XFRM_MODE_TUNNEL,
423 		.flags = XFRM_MODE_FLAG_TUNNEL,
424 		.family = AF_INET,
425 	},
426 };
427 
428 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
429 	[XFRM_MODE_BEET] = {
430 		.encap = XFRM_MODE_BEET,
431 		.flags = XFRM_MODE_FLAG_TUNNEL,
432 		.family = AF_INET6,
433 	},
434 	[XFRM_MODE_ROUTEOPTIMIZATION] = {
435 		.encap = XFRM_MODE_ROUTEOPTIMIZATION,
436 		.family = AF_INET6,
437 	},
438 	[XFRM_MODE_TRANSPORT] = {
439 		.encap = XFRM_MODE_TRANSPORT,
440 		.family = AF_INET6,
441 	},
442 	[XFRM_MODE_TUNNEL] = {
443 		.encap = XFRM_MODE_TUNNEL,
444 		.flags = XFRM_MODE_FLAG_TUNNEL,
445 		.family = AF_INET6,
446 	},
447 };
448 
xfrm_get_mode(unsigned int encap,int family)449 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
450 {
451 	const struct xfrm_mode *mode;
452 
453 	if (unlikely(encap >= XFRM_MODE_MAX))
454 		return NULL;
455 
456 	switch (family) {
457 	case AF_INET:
458 		mode = &xfrm4_mode_map[encap];
459 		if (mode->family == family)
460 			return mode;
461 		break;
462 	case AF_INET6:
463 		mode = &xfrm6_mode_map[encap];
464 		if (mode->family == family)
465 			return mode;
466 		break;
467 	default:
468 		break;
469 	}
470 
471 	return NULL;
472 }
473 
xfrm_state_free(struct xfrm_state * x)474 void xfrm_state_free(struct xfrm_state *x)
475 {
476 	kmem_cache_free(xfrm_state_cache, x);
477 }
478 EXPORT_SYMBOL(xfrm_state_free);
479 
___xfrm_state_destroy(struct xfrm_state * x)480 static void ___xfrm_state_destroy(struct xfrm_state *x)
481 {
482 	hrtimer_cancel(&x->mtimer);
483 	del_timer_sync(&x->rtimer);
484 	kfree(x->aead);
485 	kfree(x->aalg);
486 	kfree(x->ealg);
487 	kfree(x->calg);
488 	kfree(x->encap);
489 	kfree(x->coaddr);
490 	kfree(x->replay_esn);
491 	kfree(x->preplay_esn);
492 	if (x->type_offload)
493 		xfrm_put_type_offload(x->type_offload);
494 	if (x->type) {
495 		x->type->destructor(x);
496 		xfrm_put_type(x->type);
497 	}
498 	if (x->xfrag.page)
499 		put_page(x->xfrag.page);
500 	xfrm_dev_state_free(x);
501 	security_xfrm_state_free(x);
502 	xfrm_state_free(x);
503 }
504 
xfrm_state_gc_task(struct work_struct * work)505 static void xfrm_state_gc_task(struct work_struct *work)
506 {
507 	struct xfrm_state *x;
508 	struct hlist_node *tmp;
509 	struct hlist_head gc_list;
510 
511 	spin_lock_bh(&xfrm_state_gc_lock);
512 	hlist_move_list(&xfrm_state_gc_list, &gc_list);
513 	spin_unlock_bh(&xfrm_state_gc_lock);
514 
515 	synchronize_rcu();
516 
517 	hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
518 		___xfrm_state_destroy(x);
519 }
520 
xfrm_timer_handler(struct hrtimer * me)521 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
522 {
523 	struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
524 	enum hrtimer_restart ret = HRTIMER_NORESTART;
525 	time64_t now = ktime_get_real_seconds();
526 	time64_t next = TIME64_MAX;
527 	int warn = 0;
528 	int err = 0;
529 
530 	spin_lock(&x->lock);
531 	if (x->km.state == XFRM_STATE_DEAD)
532 		goto out;
533 	if (x->km.state == XFRM_STATE_EXPIRED)
534 		goto expired;
535 	if (x->lft.hard_add_expires_seconds) {
536 		long tmo = x->lft.hard_add_expires_seconds +
537 			x->curlft.add_time - now;
538 		if (tmo <= 0) {
539 			if (x->xflags & XFRM_SOFT_EXPIRE) {
540 				/* enter hard expire without soft expire first?!
541 				 * setting a new date could trigger this.
542 				 * workaround: fix x->curflt.add_time by below:
543 				 */
544 				x->curlft.add_time = now - x->saved_tmo - 1;
545 				tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
546 			} else
547 				goto expired;
548 		}
549 		if (tmo < next)
550 			next = tmo;
551 	}
552 	if (x->lft.hard_use_expires_seconds) {
553 		long tmo = x->lft.hard_use_expires_seconds +
554 			(x->curlft.use_time ? : now) - now;
555 		if (tmo <= 0)
556 			goto expired;
557 		if (tmo < next)
558 			next = tmo;
559 	}
560 	if (x->km.dying)
561 		goto resched;
562 	if (x->lft.soft_add_expires_seconds) {
563 		long tmo = x->lft.soft_add_expires_seconds +
564 			x->curlft.add_time - now;
565 		if (tmo <= 0) {
566 			warn = 1;
567 			x->xflags &= ~XFRM_SOFT_EXPIRE;
568 		} else if (tmo < next) {
569 			next = tmo;
570 			x->xflags |= XFRM_SOFT_EXPIRE;
571 			x->saved_tmo = tmo;
572 		}
573 	}
574 	if (x->lft.soft_use_expires_seconds) {
575 		long tmo = x->lft.soft_use_expires_seconds +
576 			(x->curlft.use_time ? : now) - now;
577 		if (tmo <= 0)
578 			warn = 1;
579 		else if (tmo < next)
580 			next = tmo;
581 	}
582 
583 	x->km.dying = warn;
584 	if (warn)
585 		km_state_expired(x, 0, 0);
586 resched:
587 	if (next != TIME64_MAX) {
588 		hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
589 		ret = HRTIMER_RESTART;
590 	}
591 
592 	goto out;
593 
594 expired:
595 	if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
596 		x->km.state = XFRM_STATE_EXPIRED;
597 
598 	err = __xfrm_state_delete(x);
599 	if (!err)
600 		km_state_expired(x, 1, 0);
601 
602 	xfrm_audit_state_delete(x, err ? 0 : 1, true);
603 
604 out:
605 	spin_unlock(&x->lock);
606 	return ret;
607 }
608 
609 static void xfrm_replay_timer_handler(struct timer_list *t);
610 
xfrm_state_alloc(struct net * net)611 struct xfrm_state *xfrm_state_alloc(struct net *net)
612 {
613 	struct xfrm_state *x;
614 
615 	x = kmem_cache_alloc(xfrm_state_cache, GFP_ATOMIC | __GFP_ZERO);
616 
617 	if (x) {
618 		write_pnet(&x->xs_net, net);
619 		refcount_set(&x->refcnt, 1);
620 		atomic_set(&x->tunnel_users, 0);
621 		INIT_LIST_HEAD(&x->km.all);
622 		INIT_HLIST_NODE(&x->bydst);
623 		INIT_HLIST_NODE(&x->bysrc);
624 		INIT_HLIST_NODE(&x->byspi);
625 		hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
626 		x->mtimer.function = xfrm_timer_handler;
627 		timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
628 		x->curlft.add_time = ktime_get_real_seconds();
629 		x->lft.soft_byte_limit = XFRM_INF;
630 		x->lft.soft_packet_limit = XFRM_INF;
631 		x->lft.hard_byte_limit = XFRM_INF;
632 		x->lft.hard_packet_limit = XFRM_INF;
633 		x->replay_maxage = 0;
634 		x->replay_maxdiff = 0;
635 		spin_lock_init(&x->lock);
636 	}
637 	return x;
638 }
639 EXPORT_SYMBOL(xfrm_state_alloc);
640 
__xfrm_state_destroy(struct xfrm_state * x,bool sync)641 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
642 {
643 	WARN_ON(x->km.state != XFRM_STATE_DEAD);
644 
645 	if (sync) {
646 		synchronize_rcu();
647 		___xfrm_state_destroy(x);
648 	} else {
649 		spin_lock_bh(&xfrm_state_gc_lock);
650 		hlist_add_head(&x->gclist, &xfrm_state_gc_list);
651 		spin_unlock_bh(&xfrm_state_gc_lock);
652 		schedule_work(&xfrm_state_gc_work);
653 	}
654 }
655 EXPORT_SYMBOL(__xfrm_state_destroy);
656 
__xfrm_state_delete(struct xfrm_state * x)657 int __xfrm_state_delete(struct xfrm_state *x)
658 {
659 	struct net *net = xs_net(x);
660 	int err = -ESRCH;
661 
662 	if (x->km.state != XFRM_STATE_DEAD) {
663 		x->km.state = XFRM_STATE_DEAD;
664 		spin_lock(&net->xfrm.xfrm_state_lock);
665 		list_del(&x->km.all);
666 		hlist_del_rcu(&x->bydst);
667 		hlist_del_rcu(&x->bysrc);
668 		if (x->id.spi)
669 			hlist_del_rcu(&x->byspi);
670 		net->xfrm.state_num--;
671 		spin_unlock(&net->xfrm.xfrm_state_lock);
672 
673 		xfrm_dev_state_delete(x);
674 
675 		/* All xfrm_state objects are created by xfrm_state_alloc.
676 		 * The xfrm_state_alloc call gives a reference, and that
677 		 * is what we are dropping here.
678 		 */
679 		xfrm_state_put(x);
680 		err = 0;
681 	}
682 
683 	return err;
684 }
685 EXPORT_SYMBOL(__xfrm_state_delete);
686 
xfrm_state_delete(struct xfrm_state * x)687 int xfrm_state_delete(struct xfrm_state *x)
688 {
689 	int err;
690 
691 	spin_lock_bh(&x->lock);
692 	err = __xfrm_state_delete(x);
693 	spin_unlock_bh(&x->lock);
694 
695 	return err;
696 }
697 EXPORT_SYMBOL(xfrm_state_delete);
698 
699 #ifdef CONFIG_SECURITY_NETWORK_XFRM
700 static inline int
xfrm_state_flush_secctx_check(struct net * net,u8 proto,bool task_valid)701 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
702 {
703 	int i, err = 0;
704 
705 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
706 		struct xfrm_state *x;
707 
708 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
709 			if (xfrm_id_proto_match(x->id.proto, proto) &&
710 			   (err = security_xfrm_state_delete(x)) != 0) {
711 				xfrm_audit_state_delete(x, 0, task_valid);
712 				return err;
713 			}
714 		}
715 	}
716 
717 	return err;
718 }
719 
720 static inline int
xfrm_dev_state_flush_secctx_check(struct net * net,struct net_device * dev,bool task_valid)721 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
722 {
723 	int i, err = 0;
724 
725 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
726 		struct xfrm_state *x;
727 		struct xfrm_state_offload *xso;
728 
729 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
730 			xso = &x->xso;
731 
732 			if (xso->dev == dev &&
733 			   (err = security_xfrm_state_delete(x)) != 0) {
734 				xfrm_audit_state_delete(x, 0, task_valid);
735 				return err;
736 			}
737 		}
738 	}
739 
740 	return err;
741 }
742 #else
743 static inline int
xfrm_state_flush_secctx_check(struct net * net,u8 proto,bool task_valid)744 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
745 {
746 	return 0;
747 }
748 
749 static inline int
xfrm_dev_state_flush_secctx_check(struct net * net,struct net_device * dev,bool task_valid)750 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
751 {
752 	return 0;
753 }
754 #endif
755 
xfrm_state_flush(struct net * net,u8 proto,bool task_valid,bool sync)756 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
757 {
758 	int i, err = 0, cnt = 0;
759 
760 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
761 	err = xfrm_state_flush_secctx_check(net, proto, task_valid);
762 	if (err)
763 		goto out;
764 
765 	err = -ESRCH;
766 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
767 		struct xfrm_state *x;
768 restart:
769 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
770 			if (!xfrm_state_kern(x) &&
771 			    xfrm_id_proto_match(x->id.proto, proto)) {
772 				xfrm_state_hold(x);
773 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
774 
775 				err = xfrm_state_delete(x);
776 				xfrm_audit_state_delete(x, err ? 0 : 1,
777 							task_valid);
778 				if (sync)
779 					xfrm_state_put_sync(x);
780 				else
781 					xfrm_state_put(x);
782 				if (!err)
783 					cnt++;
784 
785 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
786 				goto restart;
787 			}
788 		}
789 	}
790 out:
791 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
792 	if (cnt)
793 		err = 0;
794 
795 	return err;
796 }
797 EXPORT_SYMBOL(xfrm_state_flush);
798 
xfrm_dev_state_flush(struct net * net,struct net_device * dev,bool task_valid)799 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
800 {
801 	int i, err = 0, cnt = 0;
802 
803 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
804 	err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
805 	if (err)
806 		goto out;
807 
808 	err = -ESRCH;
809 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
810 		struct xfrm_state *x;
811 		struct xfrm_state_offload *xso;
812 restart:
813 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
814 			xso = &x->xso;
815 
816 			if (!xfrm_state_kern(x) && xso->dev == dev) {
817 				xfrm_state_hold(x);
818 				spin_unlock_bh(&net->xfrm.xfrm_state_lock);
819 
820 				err = xfrm_state_delete(x);
821 				xfrm_audit_state_delete(x, err ? 0 : 1,
822 							task_valid);
823 				xfrm_state_put(x);
824 				if (!err)
825 					cnt++;
826 
827 				spin_lock_bh(&net->xfrm.xfrm_state_lock);
828 				goto restart;
829 			}
830 		}
831 	}
832 	if (cnt)
833 		err = 0;
834 
835 out:
836 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
837 	return err;
838 }
839 EXPORT_SYMBOL(xfrm_dev_state_flush);
840 
xfrm_sad_getinfo(struct net * net,struct xfrmk_sadinfo * si)841 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
842 {
843 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
844 	si->sadcnt = net->xfrm.state_num;
845 	si->sadhcnt = net->xfrm.state_hmask + 1;
846 	si->sadhmcnt = xfrm_state_hashmax;
847 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
848 }
849 EXPORT_SYMBOL(xfrm_sad_getinfo);
850 
851 static void
__xfrm4_init_tempsel(struct xfrm_selector * sel,const struct flowi * fl)852 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
853 {
854 	const struct flowi4 *fl4 = &fl->u.ip4;
855 
856 	sel->daddr.a4 = fl4->daddr;
857 	sel->saddr.a4 = fl4->saddr;
858 	sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
859 	sel->dport_mask = htons(0xffff);
860 	sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
861 	sel->sport_mask = htons(0xffff);
862 	sel->family = AF_INET;
863 	sel->prefixlen_d = 32;
864 	sel->prefixlen_s = 32;
865 	sel->proto = fl4->flowi4_proto;
866 	sel->ifindex = fl4->flowi4_oif;
867 }
868 
869 static void
__xfrm6_init_tempsel(struct xfrm_selector * sel,const struct flowi * fl)870 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
871 {
872 	const struct flowi6 *fl6 = &fl->u.ip6;
873 
874 	/* Initialize temporary selector matching only to current session. */
875 	*(struct in6_addr *)&sel->daddr = fl6->daddr;
876 	*(struct in6_addr *)&sel->saddr = fl6->saddr;
877 	sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
878 	sel->dport_mask = htons(0xffff);
879 	sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
880 	sel->sport_mask = htons(0xffff);
881 	sel->family = AF_INET6;
882 	sel->prefixlen_d = 128;
883 	sel->prefixlen_s = 128;
884 	sel->proto = fl6->flowi6_proto;
885 	sel->ifindex = fl6->flowi6_oif;
886 }
887 
888 static void
xfrm_init_tempstate(struct xfrm_state * x,const struct flowi * fl,const struct xfrm_tmpl * tmpl,const xfrm_address_t * daddr,const xfrm_address_t * saddr,unsigned short family)889 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
890 		    const struct xfrm_tmpl *tmpl,
891 		    const xfrm_address_t *daddr, const xfrm_address_t *saddr,
892 		    unsigned short family)
893 {
894 	switch (family) {
895 	case AF_INET:
896 		__xfrm4_init_tempsel(&x->sel, fl);
897 		break;
898 	case AF_INET6:
899 		__xfrm6_init_tempsel(&x->sel, fl);
900 		break;
901 	}
902 
903 	x->id = tmpl->id;
904 
905 	switch (tmpl->encap_family) {
906 	case AF_INET:
907 		if (x->id.daddr.a4 == 0)
908 			x->id.daddr.a4 = daddr->a4;
909 		x->props.saddr = tmpl->saddr;
910 		if (x->props.saddr.a4 == 0)
911 			x->props.saddr.a4 = saddr->a4;
912 		break;
913 	case AF_INET6:
914 		if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
915 			memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
916 		memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
917 		if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
918 			memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
919 		break;
920 	}
921 
922 	x->props.mode = tmpl->mode;
923 	x->props.reqid = tmpl->reqid;
924 	x->props.family = tmpl->encap_family;
925 }
926 
__xfrm_state_lookup(struct net * net,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)927 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
928 					      const xfrm_address_t *daddr,
929 					      __be32 spi, u8 proto,
930 					      unsigned short family)
931 {
932 	unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
933 	struct xfrm_state *x;
934 
935 	hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
936 		if (x->props.family != family ||
937 		    x->id.spi       != spi ||
938 		    x->id.proto     != proto ||
939 		    !xfrm_addr_equal(&x->id.daddr, daddr, family))
940 			continue;
941 
942 		if ((mark & x->mark.m) != x->mark.v)
943 			continue;
944 		if (!xfrm_state_hold_rcu(x))
945 			continue;
946 		return x;
947 	}
948 
949 	return NULL;
950 }
951 
__xfrm_state_lookup_byaddr(struct net * net,u32 mark,const xfrm_address_t * daddr,const xfrm_address_t * saddr,u8 proto,unsigned short family)952 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
953 						     const xfrm_address_t *daddr,
954 						     const xfrm_address_t *saddr,
955 						     u8 proto, unsigned short family)
956 {
957 	unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
958 	struct xfrm_state *x;
959 
960 	hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
961 		if (x->props.family != family ||
962 		    x->id.proto     != proto ||
963 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
964 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
965 			continue;
966 
967 		if ((mark & x->mark.m) != x->mark.v)
968 			continue;
969 		if (!xfrm_state_hold_rcu(x))
970 			continue;
971 		return x;
972 	}
973 
974 	return NULL;
975 }
976 
977 static inline struct xfrm_state *
__xfrm_state_locate(struct xfrm_state * x,int use_spi,int family)978 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
979 {
980 	struct net *net = xs_net(x);
981 	u32 mark = x->mark.v & x->mark.m;
982 
983 	if (use_spi)
984 		return __xfrm_state_lookup(net, mark, &x->id.daddr,
985 					   x->id.spi, x->id.proto, family);
986 	else
987 		return __xfrm_state_lookup_byaddr(net, mark,
988 						  &x->id.daddr,
989 						  &x->props.saddr,
990 						  x->id.proto, family);
991 }
992 
xfrm_hash_grow_check(struct net * net,int have_hash_collision)993 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
994 {
995 	if (have_hash_collision &&
996 	    (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
997 	    net->xfrm.state_num > net->xfrm.state_hmask)
998 		schedule_work(&net->xfrm.state_hash_work);
999 }
1000 
xfrm_state_look_at(struct xfrm_policy * pol,struct xfrm_state * x,const struct flowi * fl,unsigned short family,struct xfrm_state ** best,int * acq_in_progress,int * error)1001 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1002 			       const struct flowi *fl, unsigned short family,
1003 			       struct xfrm_state **best, int *acq_in_progress,
1004 			       int *error)
1005 {
1006 	/* Resolution logic:
1007 	 * 1. There is a valid state with matching selector. Done.
1008 	 * 2. Valid state with inappropriate selector. Skip.
1009 	 *
1010 	 * Entering area of "sysdeps".
1011 	 *
1012 	 * 3. If state is not valid, selector is temporary, it selects
1013 	 *    only session which triggered previous resolution. Key
1014 	 *    manager will do something to install a state with proper
1015 	 *    selector.
1016 	 */
1017 	if (x->km.state == XFRM_STATE_VALID) {
1018 		if ((x->sel.family &&
1019 		     (x->sel.family != family ||
1020 		      !xfrm_selector_match(&x->sel, fl, family))) ||
1021 		    !security_xfrm_state_pol_flow_match(x, pol, fl))
1022 			return;
1023 
1024 		if (!*best ||
1025 		    (*best)->km.dying > x->km.dying ||
1026 		    ((*best)->km.dying == x->km.dying &&
1027 		     (*best)->curlft.add_time < x->curlft.add_time))
1028 			*best = x;
1029 	} else if (x->km.state == XFRM_STATE_ACQ) {
1030 		*acq_in_progress = 1;
1031 	} else if (x->km.state == XFRM_STATE_ERROR ||
1032 		   x->km.state == XFRM_STATE_EXPIRED) {
1033 		if ((!x->sel.family ||
1034 		     (x->sel.family == family &&
1035 		      xfrm_selector_match(&x->sel, fl, family))) &&
1036 		    security_xfrm_state_pol_flow_match(x, pol, fl))
1037 			*error = -ESRCH;
1038 	}
1039 }
1040 
1041 struct xfrm_state *
xfrm_state_find(const xfrm_address_t * daddr,const xfrm_address_t * saddr,const struct flowi * fl,struct xfrm_tmpl * tmpl,struct xfrm_policy * pol,int * err,unsigned short family,u32 if_id)1042 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1043 		const struct flowi *fl, struct xfrm_tmpl *tmpl,
1044 		struct xfrm_policy *pol, int *err,
1045 		unsigned short family, u32 if_id)
1046 {
1047 	static xfrm_address_t saddr_wildcard = { };
1048 	struct net *net = xp_net(pol);
1049 	unsigned int h, h_wildcard;
1050 	struct xfrm_state *x, *x0, *to_put;
1051 	int acquire_in_progress = 0;
1052 	int error = 0;
1053 	struct xfrm_state *best = NULL;
1054 	u32 mark = pol->mark.v & pol->mark.m;
1055 	unsigned short encap_family = tmpl->encap_family;
1056 	unsigned int sequence;
1057 	struct km_event c;
1058 
1059 	to_put = NULL;
1060 
1061 	sequence = read_seqcount_begin(&xfrm_state_hash_generation);
1062 
1063 	rcu_read_lock();
1064 	h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1065 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1066 		if (x->props.family == encap_family &&
1067 		    x->props.reqid == tmpl->reqid &&
1068 		    (mark & x->mark.m) == x->mark.v &&
1069 		    x->if_id == if_id &&
1070 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1071 		    xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1072 		    tmpl->mode == x->props.mode &&
1073 		    tmpl->id.proto == x->id.proto &&
1074 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1075 			xfrm_state_look_at(pol, x, fl, family,
1076 					   &best, &acquire_in_progress, &error);
1077 	}
1078 	if (best || acquire_in_progress)
1079 		goto found;
1080 
1081 	h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1082 	hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1083 		if (x->props.family == encap_family &&
1084 		    x->props.reqid == tmpl->reqid &&
1085 		    (mark & x->mark.m) == x->mark.v &&
1086 		    x->if_id == if_id &&
1087 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1088 		    xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1089 		    tmpl->mode == x->props.mode &&
1090 		    tmpl->id.proto == x->id.proto &&
1091 		    (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1092 			xfrm_state_look_at(pol, x, fl, family,
1093 					   &best, &acquire_in_progress, &error);
1094 	}
1095 
1096 found:
1097 	x = best;
1098 	if (!x && !error && !acquire_in_progress) {
1099 		if (tmpl->id.spi &&
1100 		    (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1101 					      tmpl->id.proto, encap_family)) != NULL) {
1102 			to_put = x0;
1103 			error = -EEXIST;
1104 			goto out;
1105 		}
1106 
1107 		c.net = net;
1108 		/* If the KMs have no listeners (yet...), avoid allocating an SA
1109 		 * for each and every packet - garbage collection might not
1110 		 * handle the flood.
1111 		 */
1112 		if (!km_is_alive(&c)) {
1113 			error = -ESRCH;
1114 			goto out;
1115 		}
1116 
1117 		x = xfrm_state_alloc(net);
1118 		if (x == NULL) {
1119 			error = -ENOMEM;
1120 			goto out;
1121 		}
1122 		/* Initialize temporary state matching only
1123 		 * to current session. */
1124 		xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1125 		memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1126 		x->if_id = if_id;
1127 
1128 		error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1129 		if (error) {
1130 			x->km.state = XFRM_STATE_DEAD;
1131 			to_put = x;
1132 			x = NULL;
1133 			goto out;
1134 		}
1135 
1136 		if (km_query(x, tmpl, pol) == 0) {
1137 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1138 			x->km.state = XFRM_STATE_ACQ;
1139 			list_add(&x->km.all, &net->xfrm.state_all);
1140 			hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1141 			h = xfrm_src_hash(net, daddr, saddr, encap_family);
1142 			hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1143 			if (x->id.spi) {
1144 				h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1145 				hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1146 			}
1147 			x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1148 			hrtimer_start(&x->mtimer,
1149 				      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1150 				      HRTIMER_MODE_REL_SOFT);
1151 			net->xfrm.state_num++;
1152 			xfrm_hash_grow_check(net, x->bydst.next != NULL);
1153 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1154 		} else {
1155 			x->km.state = XFRM_STATE_DEAD;
1156 			to_put = x;
1157 			x = NULL;
1158 			error = -ESRCH;
1159 		}
1160 	}
1161 out:
1162 	if (x) {
1163 		if (!xfrm_state_hold_rcu(x)) {
1164 			*err = -EAGAIN;
1165 			x = NULL;
1166 		}
1167 	} else {
1168 		*err = acquire_in_progress ? -EAGAIN : error;
1169 	}
1170 	rcu_read_unlock();
1171 	if (to_put)
1172 		xfrm_state_put(to_put);
1173 
1174 	if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1175 		*err = -EAGAIN;
1176 		if (x) {
1177 			xfrm_state_put(x);
1178 			x = NULL;
1179 		}
1180 	}
1181 
1182 	return x;
1183 }
1184 
1185 struct xfrm_state *
xfrm_stateonly_find(struct net * net,u32 mark,u32 if_id,xfrm_address_t * daddr,xfrm_address_t * saddr,unsigned short family,u8 mode,u8 proto,u32 reqid)1186 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1187 		    xfrm_address_t *daddr, xfrm_address_t *saddr,
1188 		    unsigned short family, u8 mode, u8 proto, u32 reqid)
1189 {
1190 	unsigned int h;
1191 	struct xfrm_state *rx = NULL, *x = NULL;
1192 
1193 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1194 	h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1195 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1196 		if (x->props.family == family &&
1197 		    x->props.reqid == reqid &&
1198 		    (mark & x->mark.m) == x->mark.v &&
1199 		    x->if_id == if_id &&
1200 		    !(x->props.flags & XFRM_STATE_WILDRECV) &&
1201 		    xfrm_state_addr_check(x, daddr, saddr, family) &&
1202 		    mode == x->props.mode &&
1203 		    proto == x->id.proto &&
1204 		    x->km.state == XFRM_STATE_VALID) {
1205 			rx = x;
1206 			break;
1207 		}
1208 	}
1209 
1210 	if (rx)
1211 		xfrm_state_hold(rx);
1212 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1213 
1214 
1215 	return rx;
1216 }
1217 EXPORT_SYMBOL(xfrm_stateonly_find);
1218 
xfrm_state_lookup_byspi(struct net * net,__be32 spi,unsigned short family)1219 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1220 					      unsigned short family)
1221 {
1222 	struct xfrm_state *x;
1223 	struct xfrm_state_walk *w;
1224 
1225 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1226 	list_for_each_entry(w, &net->xfrm.state_all, all) {
1227 		x = container_of(w, struct xfrm_state, km);
1228 		if (x->props.family != family ||
1229 			x->id.spi != spi)
1230 			continue;
1231 
1232 		xfrm_state_hold(x);
1233 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1234 		return x;
1235 	}
1236 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1237 	return NULL;
1238 }
1239 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1240 
__xfrm_state_insert(struct xfrm_state * x)1241 static void __xfrm_state_insert(struct xfrm_state *x)
1242 {
1243 	struct net *net = xs_net(x);
1244 	unsigned int h;
1245 
1246 	list_add(&x->km.all, &net->xfrm.state_all);
1247 
1248 	h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1249 			  x->props.reqid, x->props.family);
1250 	hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1251 
1252 	h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1253 	hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1254 
1255 	if (x->id.spi) {
1256 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1257 				  x->props.family);
1258 
1259 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1260 	}
1261 
1262 	hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1263 	if (x->replay_maxage)
1264 		mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1265 
1266 	net->xfrm.state_num++;
1267 
1268 	xfrm_hash_grow_check(net, x->bydst.next != NULL);
1269 }
1270 
1271 /* net->xfrm.xfrm_state_lock is held */
__xfrm_state_bump_genids(struct xfrm_state * xnew)1272 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1273 {
1274 	struct net *net = xs_net(xnew);
1275 	unsigned short family = xnew->props.family;
1276 	u32 reqid = xnew->props.reqid;
1277 	struct xfrm_state *x;
1278 	unsigned int h;
1279 	u32 mark = xnew->mark.v & xnew->mark.m;
1280 	u32 if_id = xnew->if_id;
1281 
1282 	h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1283 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1284 		if (x->props.family	== family &&
1285 		    x->props.reqid	== reqid &&
1286 		    x->if_id		== if_id &&
1287 		    (mark & x->mark.m) == x->mark.v &&
1288 		    xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1289 		    xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1290 			x->genid++;
1291 	}
1292 }
1293 
xfrm_state_insert(struct xfrm_state * x)1294 void xfrm_state_insert(struct xfrm_state *x)
1295 {
1296 	struct net *net = xs_net(x);
1297 
1298 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1299 	__xfrm_state_bump_genids(x);
1300 	__xfrm_state_insert(x);
1301 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1302 }
1303 EXPORT_SYMBOL(xfrm_state_insert);
1304 
1305 /* net->xfrm.xfrm_state_lock is held */
__find_acq_core(struct net * net,const struct xfrm_mark * m,unsigned short family,u8 mode,u32 reqid,u32 if_id,u8 proto,const xfrm_address_t * daddr,const xfrm_address_t * saddr,int create)1306 static struct xfrm_state *__find_acq_core(struct net *net,
1307 					  const struct xfrm_mark *m,
1308 					  unsigned short family, u8 mode,
1309 					  u32 reqid, u32 if_id, u8 proto,
1310 					  const xfrm_address_t *daddr,
1311 					  const xfrm_address_t *saddr,
1312 					  int create)
1313 {
1314 	unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1315 	struct xfrm_state *x;
1316 	u32 mark = m->v & m->m;
1317 
1318 	hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1319 		if (x->props.reqid  != reqid ||
1320 		    x->props.mode   != mode ||
1321 		    x->props.family != family ||
1322 		    x->km.state     != XFRM_STATE_ACQ ||
1323 		    x->id.spi       != 0 ||
1324 		    x->id.proto	    != proto ||
1325 		    (mark & x->mark.m) != x->mark.v ||
1326 		    !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1327 		    !xfrm_addr_equal(&x->props.saddr, saddr, family))
1328 			continue;
1329 
1330 		xfrm_state_hold(x);
1331 		return x;
1332 	}
1333 
1334 	if (!create)
1335 		return NULL;
1336 
1337 	x = xfrm_state_alloc(net);
1338 	if (likely(x)) {
1339 		switch (family) {
1340 		case AF_INET:
1341 			x->sel.daddr.a4 = daddr->a4;
1342 			x->sel.saddr.a4 = saddr->a4;
1343 			x->sel.prefixlen_d = 32;
1344 			x->sel.prefixlen_s = 32;
1345 			x->props.saddr.a4 = saddr->a4;
1346 			x->id.daddr.a4 = daddr->a4;
1347 			break;
1348 
1349 		case AF_INET6:
1350 			x->sel.daddr.in6 = daddr->in6;
1351 			x->sel.saddr.in6 = saddr->in6;
1352 			x->sel.prefixlen_d = 128;
1353 			x->sel.prefixlen_s = 128;
1354 			x->props.saddr.in6 = saddr->in6;
1355 			x->id.daddr.in6 = daddr->in6;
1356 			break;
1357 		}
1358 
1359 		x->km.state = XFRM_STATE_ACQ;
1360 		x->id.proto = proto;
1361 		x->props.family = family;
1362 		x->props.mode = mode;
1363 		x->props.reqid = reqid;
1364 		x->if_id = if_id;
1365 		x->mark.v = m->v;
1366 		x->mark.m = m->m;
1367 		x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1368 		xfrm_state_hold(x);
1369 		hrtimer_start(&x->mtimer,
1370 			      ktime_set(net->xfrm.sysctl_acq_expires, 0),
1371 			      HRTIMER_MODE_REL_SOFT);
1372 		list_add(&x->km.all, &net->xfrm.state_all);
1373 		hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1374 		h = xfrm_src_hash(net, daddr, saddr, family);
1375 		hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1376 
1377 		net->xfrm.state_num++;
1378 
1379 		xfrm_hash_grow_check(net, x->bydst.next != NULL);
1380 	}
1381 
1382 	return x;
1383 }
1384 
1385 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1386 
xfrm_state_add(struct xfrm_state * x)1387 int xfrm_state_add(struct xfrm_state *x)
1388 {
1389 	struct net *net = xs_net(x);
1390 	struct xfrm_state *x1, *to_put;
1391 	int family;
1392 	int err;
1393 	u32 mark = x->mark.v & x->mark.m;
1394 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1395 
1396 	family = x->props.family;
1397 
1398 	to_put = NULL;
1399 
1400 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1401 
1402 	x1 = __xfrm_state_locate(x, use_spi, family);
1403 	if (x1) {
1404 		to_put = x1;
1405 		x1 = NULL;
1406 		err = -EEXIST;
1407 		goto out;
1408 	}
1409 
1410 	if (use_spi && x->km.seq) {
1411 		x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1412 		if (x1 && ((x1->id.proto != x->id.proto) ||
1413 		    !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1414 			to_put = x1;
1415 			x1 = NULL;
1416 		}
1417 	}
1418 
1419 	if (use_spi && !x1)
1420 		x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1421 				     x->props.reqid, x->if_id, x->id.proto,
1422 				     &x->id.daddr, &x->props.saddr, 0);
1423 
1424 	__xfrm_state_bump_genids(x);
1425 	__xfrm_state_insert(x);
1426 	err = 0;
1427 
1428 out:
1429 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1430 
1431 	if (x1) {
1432 		xfrm_state_delete(x1);
1433 		xfrm_state_put(x1);
1434 	}
1435 
1436 	if (to_put)
1437 		xfrm_state_put(to_put);
1438 
1439 	return err;
1440 }
1441 EXPORT_SYMBOL(xfrm_state_add);
1442 
1443 #ifdef CONFIG_XFRM_MIGRATE
clone_security(struct xfrm_state * x,struct xfrm_sec_ctx * security)1444 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1445 {
1446 	struct xfrm_user_sec_ctx *uctx;
1447 	int size = sizeof(*uctx) + security->ctx_len;
1448 	int err;
1449 
1450 	uctx = kmalloc(size, GFP_KERNEL);
1451 	if (!uctx)
1452 		return -ENOMEM;
1453 
1454 	uctx->exttype = XFRMA_SEC_CTX;
1455 	uctx->len = size;
1456 	uctx->ctx_doi = security->ctx_doi;
1457 	uctx->ctx_alg = security->ctx_alg;
1458 	uctx->ctx_len = security->ctx_len;
1459 	memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1460 	err = security_xfrm_state_alloc(x, uctx);
1461 	kfree(uctx);
1462 	if (err)
1463 		return err;
1464 
1465 	return 0;
1466 }
1467 
xfrm_state_clone(struct xfrm_state * orig,struct xfrm_encap_tmpl * encap)1468 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1469 					   struct xfrm_encap_tmpl *encap)
1470 {
1471 	struct net *net = xs_net(orig);
1472 	struct xfrm_state *x = xfrm_state_alloc(net);
1473 	if (!x)
1474 		goto out;
1475 
1476 	memcpy(&x->id, &orig->id, sizeof(x->id));
1477 	memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1478 	memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1479 	x->props.mode = orig->props.mode;
1480 	x->props.replay_window = orig->props.replay_window;
1481 	x->props.reqid = orig->props.reqid;
1482 	x->props.family = orig->props.family;
1483 	x->props.saddr = orig->props.saddr;
1484 
1485 	if (orig->aalg) {
1486 		x->aalg = xfrm_algo_auth_clone(orig->aalg);
1487 		if (!x->aalg)
1488 			goto error;
1489 	}
1490 	x->props.aalgo = orig->props.aalgo;
1491 
1492 	if (orig->aead) {
1493 		x->aead = xfrm_algo_aead_clone(orig->aead);
1494 		x->geniv = orig->geniv;
1495 		if (!x->aead)
1496 			goto error;
1497 	}
1498 	if (orig->ealg) {
1499 		x->ealg = xfrm_algo_clone(orig->ealg);
1500 		if (!x->ealg)
1501 			goto error;
1502 	}
1503 	x->props.ealgo = orig->props.ealgo;
1504 
1505 	if (orig->calg) {
1506 		x->calg = xfrm_algo_clone(orig->calg);
1507 		if (!x->calg)
1508 			goto error;
1509 	}
1510 	x->props.calgo = orig->props.calgo;
1511 
1512 	if (encap || orig->encap) {
1513 		if (encap)
1514 			x->encap = kmemdup(encap, sizeof(*x->encap),
1515 					GFP_KERNEL);
1516 		else
1517 			x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1518 					GFP_KERNEL);
1519 
1520 		if (!x->encap)
1521 			goto error;
1522 	}
1523 
1524 	if (orig->security)
1525 		if (clone_security(x, orig->security))
1526 			goto error;
1527 
1528 	if (orig->coaddr) {
1529 		x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1530 				    GFP_KERNEL);
1531 		if (!x->coaddr)
1532 			goto error;
1533 	}
1534 
1535 	if (orig->replay_esn) {
1536 		if (xfrm_replay_clone(x, orig))
1537 			goto error;
1538 	}
1539 
1540 	memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1541 	memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1542 
1543 	x->props.flags = orig->props.flags;
1544 	x->props.extra_flags = orig->props.extra_flags;
1545 
1546 	x->if_id = orig->if_id;
1547 	x->tfcpad = orig->tfcpad;
1548 	x->replay_maxdiff = orig->replay_maxdiff;
1549 	x->replay_maxage = orig->replay_maxage;
1550 	memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1551 	x->km.state = orig->km.state;
1552 	x->km.seq = orig->km.seq;
1553 	x->replay = orig->replay;
1554 	x->preplay = orig->preplay;
1555 
1556 	return x;
1557 
1558  error:
1559 	xfrm_state_put(x);
1560 out:
1561 	return NULL;
1562 }
1563 
xfrm_migrate_state_find(struct xfrm_migrate * m,struct net * net,u32 if_id)1564 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
1565 						u32 if_id)
1566 {
1567 	unsigned int h;
1568 	struct xfrm_state *x = NULL;
1569 
1570 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1571 
1572 	if (m->reqid) {
1573 		h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1574 				  m->reqid, m->old_family);
1575 		hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1576 			if (x->props.mode != m->mode ||
1577 			    x->id.proto != m->proto)
1578 				continue;
1579 			if (m->reqid && x->props.reqid != m->reqid)
1580 				continue;
1581 			if (if_id != 0 && x->if_id != if_id)
1582 				continue;
1583 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1584 					     m->old_family) ||
1585 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1586 					     m->old_family))
1587 				continue;
1588 			xfrm_state_hold(x);
1589 			break;
1590 		}
1591 	} else {
1592 		h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1593 				  m->old_family);
1594 		hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1595 			if (x->props.mode != m->mode ||
1596 			    x->id.proto != m->proto)
1597 				continue;
1598 			if (if_id != 0 && x->if_id != if_id)
1599 				continue;
1600 			if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1601 					     m->old_family) ||
1602 			    !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1603 					     m->old_family))
1604 				continue;
1605 			xfrm_state_hold(x);
1606 			break;
1607 		}
1608 	}
1609 
1610 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1611 
1612 	return x;
1613 }
1614 EXPORT_SYMBOL(xfrm_migrate_state_find);
1615 
xfrm_state_migrate(struct xfrm_state * x,struct xfrm_migrate * m,struct xfrm_encap_tmpl * encap)1616 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1617 				      struct xfrm_migrate *m,
1618 				      struct xfrm_encap_tmpl *encap)
1619 {
1620 	struct xfrm_state *xc;
1621 
1622 	xc = xfrm_state_clone(x, encap);
1623 	if (!xc)
1624 		return NULL;
1625 
1626 	xc->props.family = m->new_family;
1627 
1628 	if (xfrm_init_state(xc) < 0)
1629 		goto error;
1630 
1631 	memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1632 	memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1633 
1634 	/* add state */
1635 	if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1636 		/* a care is needed when the destination address of the
1637 		   state is to be updated as it is a part of triplet */
1638 		xfrm_state_insert(xc);
1639 	} else {
1640 		if (xfrm_state_add(xc) < 0)
1641 			goto error;
1642 	}
1643 
1644 	return xc;
1645 error:
1646 	xfrm_state_put(xc);
1647 	return NULL;
1648 }
1649 EXPORT_SYMBOL(xfrm_state_migrate);
1650 #endif
1651 
xfrm_state_update(struct xfrm_state * x)1652 int xfrm_state_update(struct xfrm_state *x)
1653 {
1654 	struct xfrm_state *x1, *to_put;
1655 	int err;
1656 	int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1657 	struct net *net = xs_net(x);
1658 
1659 	to_put = NULL;
1660 
1661 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1662 	x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1663 
1664 	err = -ESRCH;
1665 	if (!x1)
1666 		goto out;
1667 
1668 	if (xfrm_state_kern(x1)) {
1669 		to_put = x1;
1670 		err = -EEXIST;
1671 		goto out;
1672 	}
1673 
1674 	if (x1->km.state == XFRM_STATE_ACQ) {
1675 		__xfrm_state_insert(x);
1676 		x = NULL;
1677 	}
1678 	err = 0;
1679 
1680 out:
1681 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1682 
1683 	if (to_put)
1684 		xfrm_state_put(to_put);
1685 
1686 	if (err)
1687 		return err;
1688 
1689 	if (!x) {
1690 		xfrm_state_delete(x1);
1691 		xfrm_state_put(x1);
1692 		return 0;
1693 	}
1694 
1695 	err = -EINVAL;
1696 	spin_lock_bh(&x1->lock);
1697 	if (likely(x1->km.state == XFRM_STATE_VALID)) {
1698 		if (x->encap && x1->encap &&
1699 		    x->encap->encap_type == x1->encap->encap_type)
1700 			memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1701 		else if (x->encap || x1->encap)
1702 			goto fail;
1703 
1704 		if (x->coaddr && x1->coaddr) {
1705 			memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1706 		}
1707 		if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1708 			memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1709 		memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1710 		x1->km.dying = 0;
1711 
1712 		hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1713 			      HRTIMER_MODE_REL_SOFT);
1714 		if (x1->curlft.use_time)
1715 			xfrm_state_check_expire(x1);
1716 
1717 		if (x->props.smark.m || x->props.smark.v || x->if_id) {
1718 			spin_lock_bh(&net->xfrm.xfrm_state_lock);
1719 
1720 			if (x->props.smark.m || x->props.smark.v)
1721 				x1->props.smark = x->props.smark;
1722 
1723 			if (x->if_id)
1724 				x1->if_id = x->if_id;
1725 
1726 			__xfrm_state_bump_genids(x1);
1727 			spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1728 		}
1729 
1730 		err = 0;
1731 		x->km.state = XFRM_STATE_DEAD;
1732 		__xfrm_state_put(x);
1733 	}
1734 
1735 fail:
1736 	spin_unlock_bh(&x1->lock);
1737 
1738 	xfrm_state_put(x1);
1739 
1740 	return err;
1741 }
1742 EXPORT_SYMBOL(xfrm_state_update);
1743 
xfrm_state_check_expire(struct xfrm_state * x)1744 int xfrm_state_check_expire(struct xfrm_state *x)
1745 {
1746 	if (!x->curlft.use_time)
1747 		x->curlft.use_time = ktime_get_real_seconds();
1748 
1749 	if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1750 	    x->curlft.packets >= x->lft.hard_packet_limit) {
1751 		x->km.state = XFRM_STATE_EXPIRED;
1752 		hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1753 		return -EINVAL;
1754 	}
1755 
1756 	if (!x->km.dying &&
1757 	    (x->curlft.bytes >= x->lft.soft_byte_limit ||
1758 	     x->curlft.packets >= x->lft.soft_packet_limit)) {
1759 		x->km.dying = 1;
1760 		km_state_expired(x, 0, 0);
1761 	}
1762 	return 0;
1763 }
1764 EXPORT_SYMBOL(xfrm_state_check_expire);
1765 
1766 struct xfrm_state *
xfrm_state_lookup(struct net * net,u32 mark,const xfrm_address_t * daddr,__be32 spi,u8 proto,unsigned short family)1767 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1768 		  u8 proto, unsigned short family)
1769 {
1770 	struct xfrm_state *x;
1771 
1772 	rcu_read_lock();
1773 	x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1774 	rcu_read_unlock();
1775 	return x;
1776 }
1777 EXPORT_SYMBOL(xfrm_state_lookup);
1778 
1779 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)1780 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1781 			 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1782 			 u8 proto, unsigned short family)
1783 {
1784 	struct xfrm_state *x;
1785 
1786 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1787 	x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1788 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1789 	return x;
1790 }
1791 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1792 
1793 struct xfrm_state *
xfrm_find_acq(struct net * net,const struct xfrm_mark * mark,u8 mode,u32 reqid,u32 if_id,u8 proto,const xfrm_address_t * daddr,const xfrm_address_t * saddr,int create,unsigned short family)1794 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1795 	      u32 if_id, u8 proto, const xfrm_address_t *daddr,
1796 	      const xfrm_address_t *saddr, int create, unsigned short family)
1797 {
1798 	struct xfrm_state *x;
1799 
1800 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1801 	x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1802 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1803 
1804 	return x;
1805 }
1806 EXPORT_SYMBOL(xfrm_find_acq);
1807 
1808 #ifdef CONFIG_XFRM_SUB_POLICY
1809 #if IS_ENABLED(CONFIG_IPV6)
1810 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1811 static void
__xfrm6_sort(void ** dst,void ** src,int n,int (* cmp)(const void * p),int maxclass)1812 __xfrm6_sort(void **dst, void **src, int n,
1813 	     int (*cmp)(const void *p), int maxclass)
1814 {
1815 	int count[XFRM_MAX_DEPTH] = { };
1816 	int class[XFRM_MAX_DEPTH];
1817 	int i;
1818 
1819 	for (i = 0; i < n; i++) {
1820 		int c = cmp(src[i]);
1821 
1822 		class[i] = c;
1823 		count[c]++;
1824 	}
1825 
1826 	for (i = 2; i < maxclass; i++)
1827 		count[i] += count[i - 1];
1828 
1829 	for (i = 0; i < n; i++) {
1830 		dst[count[class[i] - 1]++] = src[i];
1831 		src[i] = NULL;
1832 	}
1833 }
1834 
1835 /* Rule for xfrm_state:
1836  *
1837  * rule 1: select IPsec transport except AH
1838  * rule 2: select MIPv6 RO or inbound trigger
1839  * rule 3: select IPsec transport AH
1840  * rule 4: select IPsec tunnel
1841  * rule 5: others
1842  */
__xfrm6_state_sort_cmp(const void * p)1843 static int __xfrm6_state_sort_cmp(const void *p)
1844 {
1845 	const struct xfrm_state *v = p;
1846 
1847 	switch (v->props.mode) {
1848 	case XFRM_MODE_TRANSPORT:
1849 		if (v->id.proto != IPPROTO_AH)
1850 			return 1;
1851 		else
1852 			return 3;
1853 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1854 	case XFRM_MODE_ROUTEOPTIMIZATION:
1855 	case XFRM_MODE_IN_TRIGGER:
1856 		return 2;
1857 #endif
1858 	case XFRM_MODE_TUNNEL:
1859 	case XFRM_MODE_BEET:
1860 		return 4;
1861 	}
1862 	return 5;
1863 }
1864 
1865 /* Rule for xfrm_tmpl:
1866  *
1867  * rule 1: select IPsec transport
1868  * rule 2: select MIPv6 RO or inbound trigger
1869  * rule 3: select IPsec tunnel
1870  * rule 4: others
1871  */
__xfrm6_tmpl_sort_cmp(const void * p)1872 static int __xfrm6_tmpl_sort_cmp(const void *p)
1873 {
1874 	const struct xfrm_tmpl *v = p;
1875 
1876 	switch (v->mode) {
1877 	case XFRM_MODE_TRANSPORT:
1878 		return 1;
1879 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1880 	case XFRM_MODE_ROUTEOPTIMIZATION:
1881 	case XFRM_MODE_IN_TRIGGER:
1882 		return 2;
1883 #endif
1884 	case XFRM_MODE_TUNNEL:
1885 	case XFRM_MODE_BEET:
1886 		return 3;
1887 	}
1888 	return 4;
1889 }
1890 #else
__xfrm6_state_sort_cmp(const void * p)1891 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
__xfrm6_tmpl_sort_cmp(const void * p)1892 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1893 
1894 static inline void
__xfrm6_sort(void ** dst,void ** src,int n,int (* cmp)(const void * p),int maxclass)1895 __xfrm6_sort(void **dst, void **src, int n,
1896 	     int (*cmp)(const void *p), int maxclass)
1897 {
1898 	int i;
1899 
1900 	for (i = 0; i < n; i++)
1901 		dst[i] = src[i];
1902 }
1903 #endif /* CONFIG_IPV6 */
1904 
1905 void
xfrm_tmpl_sort(struct xfrm_tmpl ** dst,struct xfrm_tmpl ** src,int n,unsigned short family)1906 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1907 	       unsigned short family)
1908 {
1909 	int i;
1910 
1911 	if (family == AF_INET6)
1912 		__xfrm6_sort((void **)dst, (void **)src, n,
1913 			     __xfrm6_tmpl_sort_cmp, 5);
1914 	else
1915 		for (i = 0; i < n; i++)
1916 			dst[i] = src[i];
1917 }
1918 
1919 void
xfrm_state_sort(struct xfrm_state ** dst,struct xfrm_state ** src,int n,unsigned short family)1920 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1921 		unsigned short family)
1922 {
1923 	int i;
1924 
1925 	if (family == AF_INET6)
1926 		__xfrm6_sort((void **)dst, (void **)src, n,
1927 			     __xfrm6_state_sort_cmp, 6);
1928 	else
1929 		for (i = 0; i < n; i++)
1930 			dst[i] = src[i];
1931 }
1932 #endif
1933 
1934 /* Silly enough, but I'm lazy to build resolution list */
1935 
__xfrm_find_acq_byseq(struct net * net,u32 mark,u32 seq)1936 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1937 {
1938 	int i;
1939 
1940 	for (i = 0; i <= net->xfrm.state_hmask; i++) {
1941 		struct xfrm_state *x;
1942 
1943 		hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1944 			if (x->km.seq == seq &&
1945 			    (mark & x->mark.m) == x->mark.v &&
1946 			    x->km.state == XFRM_STATE_ACQ) {
1947 				xfrm_state_hold(x);
1948 				return x;
1949 			}
1950 		}
1951 	}
1952 	return NULL;
1953 }
1954 
xfrm_find_acq_byseq(struct net * net,u32 mark,u32 seq)1955 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1956 {
1957 	struct xfrm_state *x;
1958 
1959 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
1960 	x = __xfrm_find_acq_byseq(net, mark, seq);
1961 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1962 	return x;
1963 }
1964 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1965 
xfrm_get_acqseq(void)1966 u32 xfrm_get_acqseq(void)
1967 {
1968 	u32 res;
1969 	static atomic_t acqseq;
1970 
1971 	do {
1972 		res = atomic_inc_return(&acqseq);
1973 	} while (!res);
1974 
1975 	return res;
1976 }
1977 EXPORT_SYMBOL(xfrm_get_acqseq);
1978 
verify_spi_info(u8 proto,u32 min,u32 max)1979 int verify_spi_info(u8 proto, u32 min, u32 max)
1980 {
1981 	switch (proto) {
1982 	case IPPROTO_AH:
1983 	case IPPROTO_ESP:
1984 		break;
1985 
1986 	case IPPROTO_COMP:
1987 		/* IPCOMP spi is 16-bits. */
1988 		if (max >= 0x10000)
1989 			return -EINVAL;
1990 		break;
1991 
1992 	default:
1993 		return -EINVAL;
1994 	}
1995 
1996 	if (min > max)
1997 		return -EINVAL;
1998 
1999 	return 0;
2000 }
2001 EXPORT_SYMBOL(verify_spi_info);
2002 
xfrm_alloc_spi(struct xfrm_state * x,u32 low,u32 high)2003 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2004 {
2005 	struct net *net = xs_net(x);
2006 	unsigned int h;
2007 	struct xfrm_state *x0;
2008 	int err = -ENOENT;
2009 	__be32 minspi = htonl(low);
2010 	__be32 maxspi = htonl(high);
2011 	__be32 newspi = 0;
2012 	u32 mark = x->mark.v & x->mark.m;
2013 
2014 	spin_lock_bh(&x->lock);
2015 	if (x->km.state == XFRM_STATE_DEAD)
2016 		goto unlock;
2017 
2018 	err = 0;
2019 	if (x->id.spi)
2020 		goto unlock;
2021 
2022 	err = -ENOENT;
2023 
2024 	if (minspi == maxspi) {
2025 		x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2026 		if (x0) {
2027 			xfrm_state_put(x0);
2028 			goto unlock;
2029 		}
2030 		newspi = minspi;
2031 	} else {
2032 		u32 spi = 0;
2033 		for (h = 0; h < high-low+1; h++) {
2034 			spi = low + prandom_u32()%(high-low+1);
2035 			x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2036 			if (x0 == NULL) {
2037 				newspi = htonl(spi);
2038 				break;
2039 			}
2040 			xfrm_state_put(x0);
2041 		}
2042 	}
2043 	if (newspi) {
2044 		spin_lock_bh(&net->xfrm.xfrm_state_lock);
2045 		x->id.spi = newspi;
2046 		h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2047 		hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2048 		spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2049 
2050 		err = 0;
2051 	}
2052 
2053 unlock:
2054 	spin_unlock_bh(&x->lock);
2055 
2056 	return err;
2057 }
2058 EXPORT_SYMBOL(xfrm_alloc_spi);
2059 
__xfrm_state_filter_match(struct xfrm_state * x,struct xfrm_address_filter * filter)2060 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2061 				      struct xfrm_address_filter *filter)
2062 {
2063 	if (filter) {
2064 		if ((filter->family == AF_INET ||
2065 		     filter->family == AF_INET6) &&
2066 		    x->props.family != filter->family)
2067 			return false;
2068 
2069 		return addr_match(&x->props.saddr, &filter->saddr,
2070 				  filter->splen) &&
2071 		       addr_match(&x->id.daddr, &filter->daddr,
2072 				  filter->dplen);
2073 	}
2074 	return true;
2075 }
2076 
xfrm_state_walk(struct net * net,struct xfrm_state_walk * walk,int (* func)(struct xfrm_state *,int,void *),void * data)2077 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2078 		    int (*func)(struct xfrm_state *, int, void*),
2079 		    void *data)
2080 {
2081 	struct xfrm_state *state;
2082 	struct xfrm_state_walk *x;
2083 	int err = 0;
2084 
2085 	if (walk->seq != 0 && list_empty(&walk->all))
2086 		return 0;
2087 
2088 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2089 	if (list_empty(&walk->all))
2090 		x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2091 	else
2092 		x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2093 	list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2094 		if (x->state == XFRM_STATE_DEAD)
2095 			continue;
2096 		state = container_of(x, struct xfrm_state, km);
2097 		if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2098 			continue;
2099 		if (!__xfrm_state_filter_match(state, walk->filter))
2100 			continue;
2101 		err = func(state, walk->seq, data);
2102 		if (err) {
2103 			list_move_tail(&walk->all, &x->all);
2104 			goto out;
2105 		}
2106 		walk->seq++;
2107 	}
2108 	if (walk->seq == 0) {
2109 		err = -ENOENT;
2110 		goto out;
2111 	}
2112 	list_del_init(&walk->all);
2113 out:
2114 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2115 	return err;
2116 }
2117 EXPORT_SYMBOL(xfrm_state_walk);
2118 
xfrm_state_walk_init(struct xfrm_state_walk * walk,u8 proto,struct xfrm_address_filter * filter)2119 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2120 			  struct xfrm_address_filter *filter)
2121 {
2122 	INIT_LIST_HEAD(&walk->all);
2123 	walk->proto = proto;
2124 	walk->state = XFRM_STATE_DEAD;
2125 	walk->seq = 0;
2126 	walk->filter = filter;
2127 }
2128 EXPORT_SYMBOL(xfrm_state_walk_init);
2129 
xfrm_state_walk_done(struct xfrm_state_walk * walk,struct net * net)2130 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2131 {
2132 	kfree(walk->filter);
2133 
2134 	if (list_empty(&walk->all))
2135 		return;
2136 
2137 	spin_lock_bh(&net->xfrm.xfrm_state_lock);
2138 	list_del(&walk->all);
2139 	spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2140 }
2141 EXPORT_SYMBOL(xfrm_state_walk_done);
2142 
xfrm_replay_timer_handler(struct timer_list * t)2143 static void xfrm_replay_timer_handler(struct timer_list *t)
2144 {
2145 	struct xfrm_state *x = from_timer(x, t, rtimer);
2146 
2147 	spin_lock(&x->lock);
2148 
2149 	if (x->km.state == XFRM_STATE_VALID) {
2150 		if (xfrm_aevent_is_on(xs_net(x)))
2151 			x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2152 		else
2153 			x->xflags |= XFRM_TIME_DEFER;
2154 	}
2155 
2156 	spin_unlock(&x->lock);
2157 }
2158 
2159 static LIST_HEAD(xfrm_km_list);
2160 
km_policy_notify(struct xfrm_policy * xp,int dir,const struct km_event * c)2161 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2162 {
2163 	struct xfrm_mgr *km;
2164 
2165 	rcu_read_lock();
2166 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2167 		if (km->notify_policy)
2168 			km->notify_policy(xp, dir, c);
2169 	rcu_read_unlock();
2170 }
2171 
km_state_notify(struct xfrm_state * x,const struct km_event * c)2172 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2173 {
2174 	struct xfrm_mgr *km;
2175 	rcu_read_lock();
2176 	list_for_each_entry_rcu(km, &xfrm_km_list, list)
2177 		if (km->notify)
2178 			km->notify(x, c);
2179 	rcu_read_unlock();
2180 }
2181 
2182 EXPORT_SYMBOL(km_policy_notify);
2183 EXPORT_SYMBOL(km_state_notify);
2184 
km_state_expired(struct xfrm_state * x,int hard,u32 portid)2185 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2186 {
2187 	struct km_event c;
2188 
2189 	c.data.hard = hard;
2190 	c.portid = portid;
2191 	c.event = XFRM_MSG_EXPIRE;
2192 	km_state_notify(x, &c);
2193 }
2194 
2195 EXPORT_SYMBOL(km_state_expired);
2196 /*
2197  * We send to all registered managers regardless of failure
2198  * We are happy with one success
2199 */
km_query(struct xfrm_state * x,struct xfrm_tmpl * t,struct xfrm_policy * pol)2200 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2201 {
2202 	int err = -EINVAL, acqret;
2203 	struct xfrm_mgr *km;
2204 
2205 	rcu_read_lock();
2206 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2207 		acqret = km->acquire(x, t, pol);
2208 		if (!acqret)
2209 			err = acqret;
2210 	}
2211 	rcu_read_unlock();
2212 	return err;
2213 }
2214 EXPORT_SYMBOL(km_query);
2215 
km_new_mapping(struct xfrm_state * x,xfrm_address_t * ipaddr,__be16 sport)2216 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2217 {
2218 	int err = -EINVAL;
2219 	struct xfrm_mgr *km;
2220 
2221 	rcu_read_lock();
2222 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2223 		if (km->new_mapping)
2224 			err = km->new_mapping(x, ipaddr, sport);
2225 		if (!err)
2226 			break;
2227 	}
2228 	rcu_read_unlock();
2229 	return err;
2230 }
2231 EXPORT_SYMBOL(km_new_mapping);
2232 
km_policy_expired(struct xfrm_policy * pol,int dir,int hard,u32 portid)2233 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2234 {
2235 	struct km_event c;
2236 
2237 	c.data.hard = hard;
2238 	c.portid = portid;
2239 	c.event = XFRM_MSG_POLEXPIRE;
2240 	km_policy_notify(pol, dir, &c);
2241 }
2242 EXPORT_SYMBOL(km_policy_expired);
2243 
2244 #ifdef CONFIG_XFRM_MIGRATE
km_migrate(const struct xfrm_selector * sel,u8 dir,u8 type,const struct xfrm_migrate * m,int num_migrate,const struct xfrm_kmaddress * k,const struct xfrm_encap_tmpl * encap)2245 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2246 	       const struct xfrm_migrate *m, int num_migrate,
2247 	       const struct xfrm_kmaddress *k,
2248 	       const struct xfrm_encap_tmpl *encap)
2249 {
2250 	int err = -EINVAL;
2251 	int ret;
2252 	struct xfrm_mgr *km;
2253 
2254 	rcu_read_lock();
2255 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2256 		if (km->migrate) {
2257 			ret = km->migrate(sel, dir, type, m, num_migrate, k,
2258 					  encap);
2259 			if (!ret)
2260 				err = ret;
2261 		}
2262 	}
2263 	rcu_read_unlock();
2264 	return err;
2265 }
2266 EXPORT_SYMBOL(km_migrate);
2267 #endif
2268 
km_report(struct net * net,u8 proto,struct xfrm_selector * sel,xfrm_address_t * addr)2269 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2270 {
2271 	int err = -EINVAL;
2272 	int ret;
2273 	struct xfrm_mgr *km;
2274 
2275 	rcu_read_lock();
2276 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2277 		if (km->report) {
2278 			ret = km->report(net, proto, sel, addr);
2279 			if (!ret)
2280 				err = ret;
2281 		}
2282 	}
2283 	rcu_read_unlock();
2284 	return err;
2285 }
2286 EXPORT_SYMBOL(km_report);
2287 
km_is_alive(const struct km_event * c)2288 static bool km_is_alive(const struct km_event *c)
2289 {
2290 	struct xfrm_mgr *km;
2291 	bool is_alive = false;
2292 
2293 	rcu_read_lock();
2294 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2295 		if (km->is_alive && km->is_alive(c)) {
2296 			is_alive = true;
2297 			break;
2298 		}
2299 	}
2300 	rcu_read_unlock();
2301 
2302 	return is_alive;
2303 }
2304 
2305 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2306 static DEFINE_SPINLOCK(xfrm_translator_lock);
2307 static struct xfrm_translator __rcu *xfrm_translator;
2308 
xfrm_get_translator(void)2309 struct xfrm_translator *xfrm_get_translator(void)
2310 {
2311 	struct xfrm_translator *xtr;
2312 
2313 	rcu_read_lock();
2314 	xtr = rcu_dereference(xfrm_translator);
2315 	if (unlikely(!xtr))
2316 		goto out;
2317 	if (!try_module_get(xtr->owner))
2318 		xtr = NULL;
2319 out:
2320 	rcu_read_unlock();
2321 	return xtr;
2322 }
2323 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2324 
xfrm_put_translator(struct xfrm_translator * xtr)2325 void xfrm_put_translator(struct xfrm_translator *xtr)
2326 {
2327 	module_put(xtr->owner);
2328 }
2329 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2330 
xfrm_register_translator(struct xfrm_translator * xtr)2331 int xfrm_register_translator(struct xfrm_translator *xtr)
2332 {
2333 	int err = 0;
2334 
2335 	spin_lock_bh(&xfrm_translator_lock);
2336 	if (unlikely(xfrm_translator != NULL))
2337 		err = -EEXIST;
2338 	else
2339 		rcu_assign_pointer(xfrm_translator, xtr);
2340 	spin_unlock_bh(&xfrm_translator_lock);
2341 
2342 	return err;
2343 }
2344 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2345 
xfrm_unregister_translator(struct xfrm_translator * xtr)2346 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2347 {
2348 	int err = 0;
2349 
2350 	spin_lock_bh(&xfrm_translator_lock);
2351 	if (likely(xfrm_translator != NULL)) {
2352 		if (rcu_access_pointer(xfrm_translator) != xtr)
2353 			err = -EINVAL;
2354 		else
2355 			RCU_INIT_POINTER(xfrm_translator, NULL);
2356 	}
2357 	spin_unlock_bh(&xfrm_translator_lock);
2358 	synchronize_rcu();
2359 
2360 	return err;
2361 }
2362 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2363 #endif
2364 
xfrm_user_policy(struct sock * sk,int optname,u8 __user * optval,int optlen)2365 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2366 {
2367 	int err;
2368 	u8 *data;
2369 	struct xfrm_mgr *km;
2370 	struct xfrm_policy *pol = NULL;
2371 
2372 	if (!optval && !optlen) {
2373 		xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2374 		xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2375 		__sk_dst_reset(sk);
2376 		return 0;
2377 	}
2378 
2379 	if (optlen <= 0 || optlen > PAGE_SIZE)
2380 		return -EMSGSIZE;
2381 
2382 	data = memdup_user(optval, optlen);
2383 	if (IS_ERR(data))
2384 		return PTR_ERR(data);
2385 
2386 	/* Use the 64-bit / untranslated format on Android, even for compat */
2387 	if (!IS_ENABLED(CONFIG_ANDROID) || IS_ENABLED(CONFIG_XFRM_USER_COMPAT)) {
2388 		if (in_compat_syscall()) {
2389 			struct xfrm_translator *xtr = xfrm_get_translator();
2390 
2391 			if (!xtr) {
2392 				kfree(data);
2393 				return -EOPNOTSUPP;
2394 			}
2395 
2396 			err = xtr->xlate_user_policy_sockptr(&data, optlen);
2397 			xfrm_put_translator(xtr);
2398 			if (err) {
2399 				kfree(data);
2400 				return err;
2401 			}
2402 		}
2403 	}
2404 
2405 	err = -EINVAL;
2406 	rcu_read_lock();
2407 	list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2408 		pol = km->compile_policy(sk, optname, data,
2409 					 optlen, &err);
2410 		if (err >= 0)
2411 			break;
2412 	}
2413 	rcu_read_unlock();
2414 
2415 	if (err >= 0) {
2416 		xfrm_sk_policy_insert(sk, err, pol);
2417 		xfrm_pol_put(pol);
2418 		__sk_dst_reset(sk);
2419 		err = 0;
2420 	}
2421 
2422 	kfree(data);
2423 	return err;
2424 }
2425 EXPORT_SYMBOL(xfrm_user_policy);
2426 
2427 static DEFINE_SPINLOCK(xfrm_km_lock);
2428 
xfrm_register_km(struct xfrm_mgr * km)2429 int xfrm_register_km(struct xfrm_mgr *km)
2430 {
2431 	spin_lock_bh(&xfrm_km_lock);
2432 	list_add_tail_rcu(&km->list, &xfrm_km_list);
2433 	spin_unlock_bh(&xfrm_km_lock);
2434 	return 0;
2435 }
2436 EXPORT_SYMBOL(xfrm_register_km);
2437 
xfrm_unregister_km(struct xfrm_mgr * km)2438 int xfrm_unregister_km(struct xfrm_mgr *km)
2439 {
2440 	spin_lock_bh(&xfrm_km_lock);
2441 	list_del_rcu(&km->list);
2442 	spin_unlock_bh(&xfrm_km_lock);
2443 	synchronize_rcu();
2444 	return 0;
2445 }
2446 EXPORT_SYMBOL(xfrm_unregister_km);
2447 
xfrm_state_register_afinfo(struct xfrm_state_afinfo * afinfo)2448 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2449 {
2450 	int err = 0;
2451 
2452 	if (WARN_ON(afinfo->family >= NPROTO))
2453 		return -EAFNOSUPPORT;
2454 
2455 	spin_lock_bh(&xfrm_state_afinfo_lock);
2456 	if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2457 		err = -EEXIST;
2458 	else
2459 		rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2460 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2461 	return err;
2462 }
2463 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2464 
xfrm_state_unregister_afinfo(struct xfrm_state_afinfo * afinfo)2465 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2466 {
2467 	int err = 0, family = afinfo->family;
2468 
2469 	if (WARN_ON(family >= NPROTO))
2470 		return -EAFNOSUPPORT;
2471 
2472 	spin_lock_bh(&xfrm_state_afinfo_lock);
2473 	if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2474 		if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2475 			err = -EINVAL;
2476 		else
2477 			RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2478 	}
2479 	spin_unlock_bh(&xfrm_state_afinfo_lock);
2480 	synchronize_rcu();
2481 	return err;
2482 }
2483 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2484 
xfrm_state_afinfo_get_rcu(unsigned int family)2485 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2486 {
2487 	if (unlikely(family >= NPROTO))
2488 		return NULL;
2489 
2490 	return rcu_dereference(xfrm_state_afinfo[family]);
2491 }
2492 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2493 
xfrm_state_get_afinfo(unsigned int family)2494 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2495 {
2496 	struct xfrm_state_afinfo *afinfo;
2497 	if (unlikely(family >= NPROTO))
2498 		return NULL;
2499 	rcu_read_lock();
2500 	afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2501 	if (unlikely(!afinfo))
2502 		rcu_read_unlock();
2503 	return afinfo;
2504 }
2505 
xfrm_flush_gc(void)2506 void xfrm_flush_gc(void)
2507 {
2508 	flush_work(&xfrm_state_gc_work);
2509 }
2510 EXPORT_SYMBOL(xfrm_flush_gc);
2511 
2512 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
xfrm_state_delete_tunnel(struct xfrm_state * x)2513 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2514 {
2515 	if (x->tunnel) {
2516 		struct xfrm_state *t = x->tunnel;
2517 
2518 		if (atomic_read(&t->tunnel_users) == 2)
2519 			xfrm_state_delete(t);
2520 		atomic_dec(&t->tunnel_users);
2521 		xfrm_state_put_sync(t);
2522 		x->tunnel = NULL;
2523 	}
2524 }
2525 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2526 
xfrm_state_mtu(struct xfrm_state * x,int mtu)2527 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2528 {
2529 	const struct xfrm_type *type = READ_ONCE(x->type);
2530 	struct crypto_aead *aead;
2531 	u32 blksize, net_adj = 0;
2532 
2533 	if (x->km.state != XFRM_STATE_VALID ||
2534 	    !type || type->proto != IPPROTO_ESP)
2535 		return mtu - x->props.header_len;
2536 
2537 	aead = x->data;
2538 	blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2539 
2540 	switch (x->props.mode) {
2541 	case XFRM_MODE_TRANSPORT:
2542 	case XFRM_MODE_BEET:
2543 		if (x->props.family == AF_INET)
2544 			net_adj = sizeof(struct iphdr);
2545 		else if (x->props.family == AF_INET6)
2546 			net_adj = sizeof(struct ipv6hdr);
2547 		break;
2548 	case XFRM_MODE_TUNNEL:
2549 		break;
2550 	default:
2551 		WARN_ON_ONCE(1);
2552 		break;
2553 	}
2554 
2555 	return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2556 		 net_adj) & ~(blksize - 1)) + net_adj - 2;
2557 }
2558 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2559 
__xfrm_init_state(struct xfrm_state * x,bool init_replay,bool offload)2560 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2561 {
2562 	const struct xfrm_mode *inner_mode;
2563 	const struct xfrm_mode *outer_mode;
2564 	int family = x->props.family;
2565 	int err;
2566 
2567 	if (family == AF_INET &&
2568 	    READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
2569 		x->props.flags |= XFRM_STATE_NOPMTUDISC;
2570 
2571 	err = -EPROTONOSUPPORT;
2572 
2573 	if (x->sel.family != AF_UNSPEC) {
2574 		inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2575 		if (inner_mode == NULL)
2576 			goto error;
2577 
2578 		if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2579 		    family != x->sel.family)
2580 			goto error;
2581 
2582 		x->inner_mode = *inner_mode;
2583 	} else {
2584 		const struct xfrm_mode *inner_mode_iaf;
2585 		int iafamily = AF_INET;
2586 
2587 		inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2588 		if (inner_mode == NULL)
2589 			goto error;
2590 
2591 		x->inner_mode = *inner_mode;
2592 
2593 		if (x->props.family == AF_INET)
2594 			iafamily = AF_INET6;
2595 
2596 		inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2597 		if (inner_mode_iaf) {
2598 			if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2599 				x->inner_mode_iaf = *inner_mode_iaf;
2600 		}
2601 	}
2602 
2603 	x->type = xfrm_get_type(x->id.proto, family);
2604 	if (x->type == NULL)
2605 		goto error;
2606 
2607 	x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2608 
2609 	err = x->type->init_state(x);
2610 	if (err)
2611 		goto error;
2612 
2613 	outer_mode = xfrm_get_mode(x->props.mode, family);
2614 	if (!outer_mode) {
2615 		err = -EPROTONOSUPPORT;
2616 		goto error;
2617 	}
2618 
2619 	x->outer_mode = *outer_mode;
2620 	if (init_replay) {
2621 		err = xfrm_init_replay(x);
2622 		if (err)
2623 			goto error;
2624 	}
2625 
2626 error:
2627 	return err;
2628 }
2629 
2630 EXPORT_SYMBOL(__xfrm_init_state);
2631 
xfrm_init_state(struct xfrm_state * x)2632 int xfrm_init_state(struct xfrm_state *x)
2633 {
2634 	int err;
2635 
2636 	err = __xfrm_init_state(x, true, false);
2637 	if (!err)
2638 		x->km.state = XFRM_STATE_VALID;
2639 
2640 	return err;
2641 }
2642 
2643 EXPORT_SYMBOL(xfrm_init_state);
2644 
xfrm_state_init(struct net * net)2645 int __net_init xfrm_state_init(struct net *net)
2646 {
2647 	unsigned int sz;
2648 
2649 	if (net_eq(net, &init_net))
2650 		xfrm_state_cache = KMEM_CACHE(xfrm_state,
2651 					      SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2652 
2653 	INIT_LIST_HEAD(&net->xfrm.state_all);
2654 
2655 	sz = sizeof(struct hlist_head) * 8;
2656 
2657 	net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2658 	if (!net->xfrm.state_bydst)
2659 		goto out_bydst;
2660 	net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2661 	if (!net->xfrm.state_bysrc)
2662 		goto out_bysrc;
2663 	net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2664 	if (!net->xfrm.state_byspi)
2665 		goto out_byspi;
2666 	net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2667 
2668 	net->xfrm.state_num = 0;
2669 	INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2670 	spin_lock_init(&net->xfrm.xfrm_state_lock);
2671 	return 0;
2672 
2673 out_byspi:
2674 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2675 out_bysrc:
2676 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2677 out_bydst:
2678 	return -ENOMEM;
2679 }
2680 
xfrm_state_fini(struct net * net)2681 void xfrm_state_fini(struct net *net)
2682 {
2683 	unsigned int sz;
2684 
2685 	flush_work(&net->xfrm.state_hash_work);
2686 	flush_work(&xfrm_state_gc_work);
2687 	xfrm_state_flush(net, 0, false, true);
2688 
2689 	WARN_ON(!list_empty(&net->xfrm.state_all));
2690 
2691 	sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2692 	WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2693 	xfrm_hash_free(net->xfrm.state_byspi, sz);
2694 	WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2695 	xfrm_hash_free(net->xfrm.state_bysrc, sz);
2696 	WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2697 	xfrm_hash_free(net->xfrm.state_bydst, sz);
2698 }
2699 
2700 #ifdef CONFIG_AUDITSYSCALL
xfrm_audit_helper_sainfo(struct xfrm_state * x,struct audit_buffer * audit_buf)2701 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2702 				     struct audit_buffer *audit_buf)
2703 {
2704 	struct xfrm_sec_ctx *ctx = x->security;
2705 	u32 spi = ntohl(x->id.spi);
2706 
2707 	if (ctx)
2708 		audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2709 				 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2710 
2711 	switch (x->props.family) {
2712 	case AF_INET:
2713 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2714 				 &x->props.saddr.a4, &x->id.daddr.a4);
2715 		break;
2716 	case AF_INET6:
2717 		audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2718 				 x->props.saddr.a6, x->id.daddr.a6);
2719 		break;
2720 	}
2721 
2722 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2723 }
2724 
xfrm_audit_helper_pktinfo(struct sk_buff * skb,u16 family,struct audit_buffer * audit_buf)2725 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2726 				      struct audit_buffer *audit_buf)
2727 {
2728 	const struct iphdr *iph4;
2729 	const struct ipv6hdr *iph6;
2730 
2731 	switch (family) {
2732 	case AF_INET:
2733 		iph4 = ip_hdr(skb);
2734 		audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2735 				 &iph4->saddr, &iph4->daddr);
2736 		break;
2737 	case AF_INET6:
2738 		iph6 = ipv6_hdr(skb);
2739 		audit_log_format(audit_buf,
2740 				 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2741 				 &iph6->saddr, &iph6->daddr,
2742 				 iph6->flow_lbl[0] & 0x0f,
2743 				 iph6->flow_lbl[1],
2744 				 iph6->flow_lbl[2]);
2745 		break;
2746 	}
2747 }
2748 
xfrm_audit_state_add(struct xfrm_state * x,int result,bool task_valid)2749 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2750 {
2751 	struct audit_buffer *audit_buf;
2752 
2753 	audit_buf = xfrm_audit_start("SAD-add");
2754 	if (audit_buf == NULL)
2755 		return;
2756 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2757 	xfrm_audit_helper_sainfo(x, audit_buf);
2758 	audit_log_format(audit_buf, " res=%u", result);
2759 	audit_log_end(audit_buf);
2760 }
2761 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2762 
xfrm_audit_state_delete(struct xfrm_state * x,int result,bool task_valid)2763 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2764 {
2765 	struct audit_buffer *audit_buf;
2766 
2767 	audit_buf = xfrm_audit_start("SAD-delete");
2768 	if (audit_buf == NULL)
2769 		return;
2770 	xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2771 	xfrm_audit_helper_sainfo(x, audit_buf);
2772 	audit_log_format(audit_buf, " res=%u", result);
2773 	audit_log_end(audit_buf);
2774 }
2775 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2776 
xfrm_audit_state_replay_overflow(struct xfrm_state * x,struct sk_buff * skb)2777 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2778 				      struct sk_buff *skb)
2779 {
2780 	struct audit_buffer *audit_buf;
2781 	u32 spi;
2782 
2783 	audit_buf = xfrm_audit_start("SA-replay-overflow");
2784 	if (audit_buf == NULL)
2785 		return;
2786 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2787 	/* don't record the sequence number because it's inherent in this kind
2788 	 * of audit message */
2789 	spi = ntohl(x->id.spi);
2790 	audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2791 	audit_log_end(audit_buf);
2792 }
2793 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2794 
xfrm_audit_state_replay(struct xfrm_state * x,struct sk_buff * skb,__be32 net_seq)2795 void xfrm_audit_state_replay(struct xfrm_state *x,
2796 			     struct sk_buff *skb, __be32 net_seq)
2797 {
2798 	struct audit_buffer *audit_buf;
2799 	u32 spi;
2800 
2801 	audit_buf = xfrm_audit_start("SA-replayed-pkt");
2802 	if (audit_buf == NULL)
2803 		return;
2804 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2805 	spi = ntohl(x->id.spi);
2806 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2807 			 spi, spi, ntohl(net_seq));
2808 	audit_log_end(audit_buf);
2809 }
2810 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2811 
xfrm_audit_state_notfound_simple(struct sk_buff * skb,u16 family)2812 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2813 {
2814 	struct audit_buffer *audit_buf;
2815 
2816 	audit_buf = xfrm_audit_start("SA-notfound");
2817 	if (audit_buf == NULL)
2818 		return;
2819 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2820 	audit_log_end(audit_buf);
2821 }
2822 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2823 
xfrm_audit_state_notfound(struct sk_buff * skb,u16 family,__be32 net_spi,__be32 net_seq)2824 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2825 			       __be32 net_spi, __be32 net_seq)
2826 {
2827 	struct audit_buffer *audit_buf;
2828 	u32 spi;
2829 
2830 	audit_buf = xfrm_audit_start("SA-notfound");
2831 	if (audit_buf == NULL)
2832 		return;
2833 	xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2834 	spi = ntohl(net_spi);
2835 	audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2836 			 spi, spi, ntohl(net_seq));
2837 	audit_log_end(audit_buf);
2838 }
2839 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2840 
xfrm_audit_state_icvfail(struct xfrm_state * x,struct sk_buff * skb,u8 proto)2841 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2842 			      struct sk_buff *skb, u8 proto)
2843 {
2844 	struct audit_buffer *audit_buf;
2845 	__be32 net_spi;
2846 	__be32 net_seq;
2847 
2848 	audit_buf = xfrm_audit_start("SA-icv-failure");
2849 	if (audit_buf == NULL)
2850 		return;
2851 	xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2852 	if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2853 		u32 spi = ntohl(net_spi);
2854 		audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2855 				 spi, spi, ntohl(net_seq));
2856 	}
2857 	audit_log_end(audit_buf);
2858 }
2859 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2860 #endif /* CONFIG_AUDITSYSCALL */
2861