6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * YOSHIFUJI Hideaki @USAGI
10 * Split up af-specific functions
11 * Derek Atkins <derek@ihtfp.com>
12 * Add UDP Encapsulation
16 #include <linux/workqueue.h>
18 #include <linux/pfkeyv2.h>
19 #include <linux/ipsec.h>
20 #include <linux/module.h>
21 #include <linux/cache.h>
22 #include <linux/audit.h>
23 #include <asm/uaccess.h>
24 #include <linux/ktime.h>
25 #include <linux/slab.h>
26 #include <linux/interrupt.h>
27 #include <linux/kernel.h>
29 #include "xfrm_hash.h"
31 /* Each xfrm_state may be linked to two tables:
33 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
34 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
35 destination/tunnel endpoint. (output)
38 static DEFINE_SPINLOCK(xfrm_state_lock
);
40 static unsigned int xfrm_state_hashmax __read_mostly
= 1 * 1024 * 1024;
42 static struct xfrm_state_afinfo
*xfrm_state_get_afinfo(unsigned int family
);
43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo
*afinfo
);
45 static inline unsigned int xfrm_dst_hash(struct net
*net
,
46 const xfrm_address_t
*daddr
,
47 const xfrm_address_t
*saddr
,
49 unsigned short family
)
51 return __xfrm_dst_hash(daddr
, saddr
, reqid
, family
, net
->xfrm
.state_hmask
);
54 static inline unsigned int xfrm_src_hash(struct net
*net
,
55 const xfrm_address_t
*daddr
,
56 const xfrm_address_t
*saddr
,
57 unsigned short family
)
59 return __xfrm_src_hash(daddr
, saddr
, family
, net
->xfrm
.state_hmask
);
62 static inline unsigned int
63 xfrm_spi_hash(struct net
*net
, const xfrm_address_t
*daddr
,
64 __be32 spi
, u8 proto
, unsigned short family
)
66 return __xfrm_spi_hash(daddr
, spi
, proto
, family
, net
->xfrm
.state_hmask
);
69 static void xfrm_hash_transfer(struct hlist_head
*list
,
70 struct hlist_head
*ndsttable
,
71 struct hlist_head
*nsrctable
,
72 struct hlist_head
*nspitable
,
73 unsigned int nhashmask
)
75 struct hlist_node
*tmp
;
78 hlist_for_each_entry_safe(x
, tmp
, list
, bydst
) {
81 h
= __xfrm_dst_hash(&x
->id
.daddr
, &x
->props
.saddr
,
82 x
->props
.reqid
, x
->props
.family
,
84 hlist_add_head(&x
->bydst
, ndsttable
+h
);
86 h
= __xfrm_src_hash(&x
->id
.daddr
, &x
->props
.saddr
,
89 hlist_add_head(&x
->bysrc
, nsrctable
+h
);
92 h
= __xfrm_spi_hash(&x
->id
.daddr
, x
->id
.spi
,
93 x
->id
.proto
, x
->props
.family
,
95 hlist_add_head(&x
->byspi
, nspitable
+h
);
100 static unsigned long xfrm_hash_new_size(unsigned int state_hmask
)
102 return ((state_hmask
+ 1) << 1) * sizeof(struct hlist_head
);
105 static DEFINE_MUTEX(hash_resize_mutex
);
107 static void xfrm_hash_resize(struct work_struct
*work
)
109 struct net
*net
= container_of(work
, struct net
, xfrm
.state_hash_work
);
110 struct hlist_head
*ndst
, *nsrc
, *nspi
, *odst
, *osrc
, *ospi
;
111 unsigned long nsize
, osize
;
112 unsigned int nhashmask
, ohashmask
;
115 mutex_lock(&hash_resize_mutex
);
117 nsize
= xfrm_hash_new_size(net
->xfrm
.state_hmask
);
118 ndst
= xfrm_hash_alloc(nsize
);
121 nsrc
= xfrm_hash_alloc(nsize
);
123 xfrm_hash_free(ndst
, nsize
);
126 nspi
= xfrm_hash_alloc(nsize
);
128 xfrm_hash_free(ndst
, nsize
);
129 xfrm_hash_free(nsrc
, nsize
);
133 spin_lock_bh(&xfrm_state_lock
);
135 nhashmask
= (nsize
/ sizeof(struct hlist_head
)) - 1U;
136 for (i
= net
->xfrm
.state_hmask
; i
>= 0; i
--)
137 xfrm_hash_transfer(net
->xfrm
.state_bydst
+i
, ndst
, nsrc
, nspi
,
140 odst
= net
->xfrm
.state_bydst
;
141 osrc
= net
->xfrm
.state_bysrc
;
142 ospi
= net
->xfrm
.state_byspi
;
143 ohashmask
= net
->xfrm
.state_hmask
;
145 net
->xfrm
.state_bydst
= ndst
;
146 net
->xfrm
.state_bysrc
= nsrc
;
147 net
->xfrm
.state_byspi
= nspi
;
148 net
->xfrm
.state_hmask
= nhashmask
;
150 spin_unlock_bh(&xfrm_state_lock
);
152 osize
= (ohashmask
+ 1) * sizeof(struct hlist_head
);
153 xfrm_hash_free(odst
, osize
);
154 xfrm_hash_free(osrc
, osize
);
155 xfrm_hash_free(ospi
, osize
);
158 mutex_unlock(&hash_resize_mutex
);
161 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock
);
162 static struct xfrm_state_afinfo __rcu
*xfrm_state_afinfo
[NPROTO
];
164 static DEFINE_SPINLOCK(xfrm_state_gc_lock
);
166 int __xfrm_state_delete(struct xfrm_state
*x
);
168 int km_query(struct xfrm_state
*x
, struct xfrm_tmpl
*t
, struct xfrm_policy
*pol
);
169 void km_state_expired(struct xfrm_state
*x
, int hard
, u32 portid
);
171 static DEFINE_SPINLOCK(xfrm_type_lock
);
172 int xfrm_register_type(const struct xfrm_type
*type
, unsigned short family
)
174 struct xfrm_state_afinfo
*afinfo
= xfrm_state_get_afinfo(family
);
175 const struct xfrm_type
**typemap
;
178 if (unlikely(afinfo
== NULL
))
179 return -EAFNOSUPPORT
;
180 typemap
= afinfo
->type_map
;
181 spin_lock_bh(&xfrm_type_lock
);
183 if (likely(typemap
[type
->proto
] == NULL
))
184 typemap
[type
->proto
] = type
;
187 spin_unlock_bh(&xfrm_type_lock
);
188 xfrm_state_put_afinfo(afinfo
);
191 EXPORT_SYMBOL(xfrm_register_type
);
193 int xfrm_unregister_type(const struct xfrm_type
*type
, unsigned short family
)
195 struct xfrm_state_afinfo
*afinfo
= xfrm_state_get_afinfo(family
);
196 const struct xfrm_type
**typemap
;
199 if (unlikely(afinfo
== NULL
))
200 return -EAFNOSUPPORT
;
201 typemap
= afinfo
->type_map
;
202 spin_lock_bh(&xfrm_type_lock
);
204 if (unlikely(typemap
[type
->proto
] != type
))
207 typemap
[type
->proto
] = NULL
;
208 spin_unlock_bh(&xfrm_type_lock
);
209 xfrm_state_put_afinfo(afinfo
);
212 EXPORT_SYMBOL(xfrm_unregister_type
);
214 static const struct xfrm_type
*xfrm_get_type(u8 proto
, unsigned short family
)
216 struct xfrm_state_afinfo
*afinfo
;
217 const struct xfrm_type
**typemap
;
218 const struct xfrm_type
*type
;
219 int modload_attempted
= 0;
222 afinfo
= xfrm_state_get_afinfo(family
);
223 if (unlikely(afinfo
== NULL
))
225 typemap
= afinfo
->type_map
;
227 type
= typemap
[proto
];
228 if (unlikely(type
&& !try_module_get(type
->owner
)))
230 if (!type
&& !modload_attempted
) {
231 xfrm_state_put_afinfo(afinfo
);
232 request_module("xfrm-type-%d-%d", family
, proto
);
233 modload_attempted
= 1;
237 xfrm_state_put_afinfo(afinfo
);
241 static void xfrm_put_type(const struct xfrm_type
*type
)
243 module_put(type
->owner
);
246 static DEFINE_SPINLOCK(xfrm_mode_lock
);
247 int xfrm_register_mode(struct xfrm_mode
*mode
, int family
)
249 struct xfrm_state_afinfo
*afinfo
;
250 struct xfrm_mode
**modemap
;
253 if (unlikely(mode
->encap
>= XFRM_MODE_MAX
))
256 afinfo
= xfrm_state_get_afinfo(family
);
257 if (unlikely(afinfo
== NULL
))
258 return -EAFNOSUPPORT
;
261 modemap
= afinfo
->mode_map
;
262 spin_lock_bh(&xfrm_mode_lock
);
263 if (modemap
[mode
->encap
])
267 if (!try_module_get(afinfo
->owner
))
270 mode
->afinfo
= afinfo
;
271 modemap
[mode
->encap
] = mode
;
275 spin_unlock_bh(&xfrm_mode_lock
);
276 xfrm_state_put_afinfo(afinfo
);
279 EXPORT_SYMBOL(xfrm_register_mode
);
281 int xfrm_unregister_mode(struct xfrm_mode
*mode
, int family
)
283 struct xfrm_state_afinfo
*afinfo
;
284 struct xfrm_mode
**modemap
;
287 if (unlikely(mode
->encap
>= XFRM_MODE_MAX
))
290 afinfo
= xfrm_state_get_afinfo(family
);
291 if (unlikely(afinfo
== NULL
))
292 return -EAFNOSUPPORT
;
295 modemap
= afinfo
->mode_map
;
296 spin_lock_bh(&xfrm_mode_lock
);
297 if (likely(modemap
[mode
->encap
] == mode
)) {
298 modemap
[mode
->encap
] = NULL
;
299 module_put(mode
->afinfo
->owner
);
303 spin_unlock_bh(&xfrm_mode_lock
);
304 xfrm_state_put_afinfo(afinfo
);
307 EXPORT_SYMBOL(xfrm_unregister_mode
);
309 static struct xfrm_mode
*xfrm_get_mode(unsigned int encap
, int family
)
311 struct xfrm_state_afinfo
*afinfo
;
312 struct xfrm_mode
*mode
;
313 int modload_attempted
= 0;
315 if (unlikely(encap
>= XFRM_MODE_MAX
))
319 afinfo
= xfrm_state_get_afinfo(family
);
320 if (unlikely(afinfo
== NULL
))
323 mode
= afinfo
->mode_map
[encap
];
324 if (unlikely(mode
&& !try_module_get(mode
->owner
)))
326 if (!mode
&& !modload_attempted
) {
327 xfrm_state_put_afinfo(afinfo
);
328 request_module("xfrm-mode-%d-%d", family
, encap
);
329 modload_attempted
= 1;
333 xfrm_state_put_afinfo(afinfo
);
337 static void xfrm_put_mode(struct xfrm_mode
*mode
)
339 module_put(mode
->owner
);
342 static void xfrm_state_gc_destroy(struct xfrm_state
*x
)
344 tasklet_hrtimer_cancel(&x
->mtimer
);
345 del_timer_sync(&x
->rtimer
);
351 kfree(x
->replay_esn
);
352 kfree(x
->preplay_esn
);
354 xfrm_put_mode(x
->inner_mode
);
355 if (x
->inner_mode_iaf
)
356 xfrm_put_mode(x
->inner_mode_iaf
);
358 xfrm_put_mode(x
->outer_mode
);
360 x
->type
->destructor(x
);
361 xfrm_put_type(x
->type
);
363 security_xfrm_state_free(x
);
367 static void xfrm_state_gc_task(struct work_struct
*work
)
369 struct net
*net
= container_of(work
, struct net
, xfrm
.state_gc_work
);
370 struct xfrm_state
*x
;
371 struct hlist_node
*tmp
;
372 struct hlist_head gc_list
;
374 spin_lock_bh(&xfrm_state_gc_lock
);
375 hlist_move_list(&net
->xfrm
.state_gc_list
, &gc_list
);
376 spin_unlock_bh(&xfrm_state_gc_lock
);
378 hlist_for_each_entry_safe(x
, tmp
, &gc_list
, gclist
)
379 xfrm_state_gc_destroy(x
);
381 wake_up(&net
->xfrm
.km_waitq
);
384 static inline unsigned long make_jiffies(long secs
)
386 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
387 return MAX_SCHEDULE_TIMEOUT
-1;
392 static enum hrtimer_restart
xfrm_timer_handler(struct hrtimer
* me
)
394 struct tasklet_hrtimer
*thr
= container_of(me
, struct tasklet_hrtimer
, timer
);
395 struct xfrm_state
*x
= container_of(thr
, struct xfrm_state
, mtimer
);
396 struct net
*net
= xs_net(x
);
397 unsigned long now
= get_seconds();
398 long next
= LONG_MAX
;
403 if (x
->km
.state
== XFRM_STATE_DEAD
)
405 if (x
->km
.state
== XFRM_STATE_EXPIRED
)
407 if (x
->lft
.hard_add_expires_seconds
) {
408 long tmo
= x
->lft
.hard_add_expires_seconds
+
409 x
->curlft
.add_time
- now
;
411 if (x
->xflags
& XFRM_SOFT_EXPIRE
) {
412 /* enter hard expire without soft expire first?!
413 * setting a new date could trigger this.
414 * workarbound: fix x->curflt.add_time by below:
416 x
->curlft
.add_time
= now
- x
->saved_tmo
- 1;
417 tmo
= x
->lft
.hard_add_expires_seconds
- x
->saved_tmo
;
424 if (x
->lft
.hard_use_expires_seconds
) {
425 long tmo
= x
->lft
.hard_use_expires_seconds
+
426 (x
->curlft
.use_time
? : now
) - now
;
434 if (x
->lft
.soft_add_expires_seconds
) {
435 long tmo
= x
->lft
.soft_add_expires_seconds
+
436 x
->curlft
.add_time
- now
;
439 x
->xflags
&= ~XFRM_SOFT_EXPIRE
;
440 } else if (tmo
< next
) {
442 x
->xflags
|= XFRM_SOFT_EXPIRE
;
446 if (x
->lft
.soft_use_expires_seconds
) {
447 long tmo
= x
->lft
.soft_use_expires_seconds
+
448 (x
->curlft
.use_time
? : now
) - now
;
457 km_state_expired(x
, 0, 0);
459 if (next
!= LONG_MAX
){
460 tasklet_hrtimer_start(&x
->mtimer
, ktime_set(next
, 0), HRTIMER_MODE_REL
);
466 if (x
->km
.state
== XFRM_STATE_ACQ
&& x
->id
.spi
== 0) {
467 x
->km
.state
= XFRM_STATE_EXPIRED
;
468 wake_up(&net
->xfrm
.km_waitq
);
473 err
= __xfrm_state_delete(x
);
474 if (!err
&& x
->id
.spi
)
475 km_state_expired(x
, 1, 0);
477 xfrm_audit_state_delete(x
, err
? 0 : 1,
478 audit_get_loginuid(current
),
479 audit_get_sessionid(current
), 0);
482 spin_unlock(&x
->lock
);
483 return HRTIMER_NORESTART
;
486 static void xfrm_replay_timer_handler(unsigned long data
);
488 struct xfrm_state
*xfrm_state_alloc(struct net
*net
)
490 struct xfrm_state
*x
;
492 x
= kzalloc(sizeof(struct xfrm_state
), GFP_ATOMIC
);
495 write_pnet(&x
->xs_net
, net
);
496 atomic_set(&x
->refcnt
, 1);
497 atomic_set(&x
->tunnel_users
, 0);
498 INIT_LIST_HEAD(&x
->km
.all
);
499 INIT_HLIST_NODE(&x
->bydst
);
500 INIT_HLIST_NODE(&x
->bysrc
);
501 INIT_HLIST_NODE(&x
->byspi
);
502 tasklet_hrtimer_init(&x
->mtimer
, xfrm_timer_handler
, CLOCK_REALTIME
, HRTIMER_MODE_ABS
);
503 setup_timer(&x
->rtimer
, xfrm_replay_timer_handler
,
505 x
->curlft
.add_time
= get_seconds();
506 x
->lft
.soft_byte_limit
= XFRM_INF
;
507 x
->lft
.soft_packet_limit
= XFRM_INF
;
508 x
->lft
.hard_byte_limit
= XFRM_INF
;
509 x
->lft
.hard_packet_limit
= XFRM_INF
;
510 x
->replay_maxage
= 0;
511 x
->replay_maxdiff
= 0;
512 x
->inner_mode
= NULL
;
513 x
->inner_mode_iaf
= NULL
;
514 spin_lock_init(&x
->lock
);
518 EXPORT_SYMBOL(xfrm_state_alloc
);
520 void __xfrm_state_destroy(struct xfrm_state
*x
)
522 struct net
*net
= xs_net(x
);
524 WARN_ON(x
->km
.state
!= XFRM_STATE_DEAD
);
526 spin_lock_bh(&xfrm_state_gc_lock
);
527 hlist_add_head(&x
->gclist
, &net
->xfrm
.state_gc_list
);
528 spin_unlock_bh(&xfrm_state_gc_lock
);
529 schedule_work(&net
->xfrm
.state_gc_work
);
531 EXPORT_SYMBOL(__xfrm_state_destroy
);
533 int __xfrm_state_delete(struct xfrm_state
*x
)
535 struct net
*net
= xs_net(x
);
538 if (x
->km
.state
!= XFRM_STATE_DEAD
) {
539 x
->km
.state
= XFRM_STATE_DEAD
;
540 spin_lock(&xfrm_state_lock
);
541 list_del(&x
->km
.all
);
542 hlist_del(&x
->bydst
);
543 hlist_del(&x
->bysrc
);
545 hlist_del(&x
->byspi
);
546 net
->xfrm
.state_num
--;
547 spin_unlock(&xfrm_state_lock
);
549 /* All xfrm_state objects are created by xfrm_state_alloc.
550 * The xfrm_state_alloc call gives a reference, and that
551 * is what we are dropping here.
559 EXPORT_SYMBOL(__xfrm_state_delete
);
561 int xfrm_state_delete(struct xfrm_state
*x
)
565 spin_lock_bh(&x
->lock
);
566 err
= __xfrm_state_delete(x
);
567 spin_unlock_bh(&x
->lock
);
571 EXPORT_SYMBOL(xfrm_state_delete
);
573 #ifdef CONFIG_SECURITY_NETWORK_XFRM
575 xfrm_state_flush_secctx_check(struct net
*net
, u8 proto
, struct xfrm_audit
*audit_info
)
579 for (i
= 0; i
<= net
->xfrm
.state_hmask
; i
++) {
580 struct xfrm_state
*x
;
582 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+i
, bydst
) {
583 if (xfrm_id_proto_match(x
->id
.proto
, proto
) &&
584 (err
= security_xfrm_state_delete(x
)) != 0) {
585 xfrm_audit_state_delete(x
, 0,
586 audit_info
->loginuid
,
587 audit_info
->sessionid
,
598 xfrm_state_flush_secctx_check(struct net
*net
, u8 proto
, struct xfrm_audit
*audit_info
)
604 int xfrm_state_flush(struct net
*net
, u8 proto
, struct xfrm_audit
*audit_info
)
606 int i
, err
= 0, cnt
= 0;
608 spin_lock_bh(&xfrm_state_lock
);
609 err
= xfrm_state_flush_secctx_check(net
, proto
, audit_info
);
614 for (i
= 0; i
<= net
->xfrm
.state_hmask
; i
++) {
615 struct xfrm_state
*x
;
617 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+i
, bydst
) {
618 if (!xfrm_state_kern(x
) &&
619 xfrm_id_proto_match(x
->id
.proto
, proto
)) {
621 spin_unlock_bh(&xfrm_state_lock
);
623 err
= xfrm_state_delete(x
);
624 xfrm_audit_state_delete(x
, err
? 0 : 1,
625 audit_info
->loginuid
,
626 audit_info
->sessionid
,
632 spin_lock_bh(&xfrm_state_lock
);
641 spin_unlock_bh(&xfrm_state_lock
);
642 wake_up(&net
->xfrm
.km_waitq
);
645 EXPORT_SYMBOL(xfrm_state_flush
);
647 void xfrm_sad_getinfo(struct net
*net
, struct xfrmk_sadinfo
*si
)
649 spin_lock_bh(&xfrm_state_lock
);
650 si
->sadcnt
= net
->xfrm
.state_num
;
651 si
->sadhcnt
= net
->xfrm
.state_hmask
;
652 si
->sadhmcnt
= xfrm_state_hashmax
;
653 spin_unlock_bh(&xfrm_state_lock
);
655 EXPORT_SYMBOL(xfrm_sad_getinfo
);
658 xfrm_init_tempstate(struct xfrm_state
*x
, const struct flowi
*fl
,
659 const struct xfrm_tmpl
*tmpl
,
660 const xfrm_address_t
*daddr
, const xfrm_address_t
*saddr
,
661 unsigned short family
)
663 struct xfrm_state_afinfo
*afinfo
= xfrm_state_get_afinfo(family
);
666 afinfo
->init_tempsel(&x
->sel
, fl
);
668 if (family
!= tmpl
->encap_family
) {
669 xfrm_state_put_afinfo(afinfo
);
670 afinfo
= xfrm_state_get_afinfo(tmpl
->encap_family
);
674 afinfo
->init_temprop(x
, tmpl
, daddr
, saddr
);
675 xfrm_state_put_afinfo(afinfo
);
679 static struct xfrm_state
*__xfrm_state_lookup(struct net
*net
, u32 mark
,
680 const xfrm_address_t
*daddr
,
681 __be32 spi
, u8 proto
,
682 unsigned short family
)
684 unsigned int h
= xfrm_spi_hash(net
, daddr
, spi
, proto
, family
);
685 struct xfrm_state
*x
;
687 hlist_for_each_entry(x
, net
->xfrm
.state_byspi
+h
, byspi
) {
688 if (x
->props
.family
!= family
||
690 x
->id
.proto
!= proto
||
691 !xfrm_addr_equal(&x
->id
.daddr
, daddr
, family
))
694 if ((mark
& x
->mark
.m
) != x
->mark
.v
)
703 static struct xfrm_state
*__xfrm_state_lookup_byaddr(struct net
*net
, u32 mark
,
704 const xfrm_address_t
*daddr
,
705 const xfrm_address_t
*saddr
,
706 u8 proto
, unsigned short family
)
708 unsigned int h
= xfrm_src_hash(net
, daddr
, saddr
, family
);
709 struct xfrm_state
*x
;
711 hlist_for_each_entry(x
, net
->xfrm
.state_bysrc
+h
, bysrc
) {
712 if (x
->props
.family
!= family
||
713 x
->id
.proto
!= proto
||
714 !xfrm_addr_equal(&x
->id
.daddr
, daddr
, family
) ||
715 !xfrm_addr_equal(&x
->props
.saddr
, saddr
, family
))
718 if ((mark
& x
->mark
.m
) != x
->mark
.v
)
727 static inline struct xfrm_state
*
728 __xfrm_state_locate(struct xfrm_state
*x
, int use_spi
, int family
)
730 struct net
*net
= xs_net(x
);
731 u32 mark
= x
->mark
.v
& x
->mark
.m
;
734 return __xfrm_state_lookup(net
, mark
, &x
->id
.daddr
,
735 x
->id
.spi
, x
->id
.proto
, family
);
737 return __xfrm_state_lookup_byaddr(net
, mark
,
740 x
->id
.proto
, family
);
743 static void xfrm_hash_grow_check(struct net
*net
, int have_hash_collision
)
745 if (have_hash_collision
&&
746 (net
->xfrm
.state_hmask
+ 1) < xfrm_state_hashmax
&&
747 net
->xfrm
.state_num
> net
->xfrm
.state_hmask
)
748 schedule_work(&net
->xfrm
.state_hash_work
);
751 static void xfrm_state_look_at(struct xfrm_policy
*pol
, struct xfrm_state
*x
,
752 const struct flowi
*fl
, unsigned short family
,
753 struct xfrm_state
**best
, int *acq_in_progress
,
757 * 1. There is a valid state with matching selector. Done.
758 * 2. Valid state with inappropriate selector. Skip.
760 * Entering area of "sysdeps".
762 * 3. If state is not valid, selector is temporary, it selects
763 * only session which triggered previous resolution. Key
764 * manager will do something to install a state with proper
767 if (x
->km
.state
== XFRM_STATE_VALID
) {
768 if ((x
->sel
.family
&&
769 !xfrm_selector_match(&x
->sel
, fl
, x
->sel
.family
)) ||
770 !security_xfrm_state_pol_flow_match(x
, pol
, fl
))
774 (*best
)->km
.dying
> x
->km
.dying
||
775 ((*best
)->km
.dying
== x
->km
.dying
&&
776 (*best
)->curlft
.add_time
< x
->curlft
.add_time
))
778 } else if (x
->km
.state
== XFRM_STATE_ACQ
) {
779 *acq_in_progress
= 1;
780 } else if (x
->km
.state
== XFRM_STATE_ERROR
||
781 x
->km
.state
== XFRM_STATE_EXPIRED
) {
782 if (xfrm_selector_match(&x
->sel
, fl
, x
->sel
.family
) &&
783 security_xfrm_state_pol_flow_match(x
, pol
, fl
))
789 xfrm_state_find(const xfrm_address_t
*daddr
, const xfrm_address_t
*saddr
,
790 const struct flowi
*fl
, struct xfrm_tmpl
*tmpl
,
791 struct xfrm_policy
*pol
, int *err
,
792 unsigned short family
)
794 static xfrm_address_t saddr_wildcard
= { };
795 struct net
*net
= xp_net(pol
);
796 unsigned int h
, h_wildcard
;
797 struct xfrm_state
*x
, *x0
, *to_put
;
798 int acquire_in_progress
= 0;
800 struct xfrm_state
*best
= NULL
;
801 u32 mark
= pol
->mark
.v
& pol
->mark
.m
;
802 unsigned short encap_family
= tmpl
->encap_family
;
806 spin_lock_bh(&xfrm_state_lock
);
807 h
= xfrm_dst_hash(net
, daddr
, saddr
, tmpl
->reqid
, encap_family
);
808 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+h
, bydst
) {
809 if (x
->props
.family
== encap_family
&&
810 x
->props
.reqid
== tmpl
->reqid
&&
811 (mark
& x
->mark
.m
) == x
->mark
.v
&&
812 !(x
->props
.flags
& XFRM_STATE_WILDRECV
) &&
813 xfrm_state_addr_check(x
, daddr
, saddr
, encap_family
) &&
814 tmpl
->mode
== x
->props
.mode
&&
815 tmpl
->id
.proto
== x
->id
.proto
&&
816 (tmpl
->id
.spi
== x
->id
.spi
|| !tmpl
->id
.spi
))
817 xfrm_state_look_at(pol
, x
, fl
, encap_family
,
818 &best
, &acquire_in_progress
, &error
);
823 h_wildcard
= xfrm_dst_hash(net
, daddr
, &saddr_wildcard
, tmpl
->reqid
, encap_family
);
824 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+h_wildcard
, bydst
) {
825 if (x
->props
.family
== encap_family
&&
826 x
->props
.reqid
== tmpl
->reqid
&&
827 (mark
& x
->mark
.m
) == x
->mark
.v
&&
828 !(x
->props
.flags
& XFRM_STATE_WILDRECV
) &&
829 xfrm_state_addr_check(x
, daddr
, saddr
, encap_family
) &&
830 tmpl
->mode
== x
->props
.mode
&&
831 tmpl
->id
.proto
== x
->id
.proto
&&
832 (tmpl
->id
.spi
== x
->id
.spi
|| !tmpl
->id
.spi
))
833 xfrm_state_look_at(pol
, x
, fl
, encap_family
,
834 &best
, &acquire_in_progress
, &error
);
839 if (!x
&& !error
&& !acquire_in_progress
) {
841 (x0
= __xfrm_state_lookup(net
, mark
, daddr
, tmpl
->id
.spi
,
842 tmpl
->id
.proto
, encap_family
)) != NULL
) {
847 x
= xfrm_state_alloc(net
);
852 /* Initialize temporary state matching only
853 * to current session. */
854 xfrm_init_tempstate(x
, fl
, tmpl
, daddr
, saddr
, family
);
855 memcpy(&x
->mark
, &pol
->mark
, sizeof(x
->mark
));
857 error
= security_xfrm_state_alloc_acquire(x
, pol
->security
, fl
->flowi_secid
);
859 x
->km
.state
= XFRM_STATE_DEAD
;
865 if (km_query(x
, tmpl
, pol
) == 0) {
866 x
->km
.state
= XFRM_STATE_ACQ
;
867 list_add(&x
->km
.all
, &net
->xfrm
.state_all
);
868 hlist_add_head(&x
->bydst
, net
->xfrm
.state_bydst
+h
);
869 h
= xfrm_src_hash(net
, daddr
, saddr
, encap_family
);
870 hlist_add_head(&x
->bysrc
, net
->xfrm
.state_bysrc
+h
);
872 h
= xfrm_spi_hash(net
, &x
->id
.daddr
, x
->id
.spi
, x
->id
.proto
, encap_family
);
873 hlist_add_head(&x
->byspi
, net
->xfrm
.state_byspi
+h
);
875 x
->lft
.hard_add_expires_seconds
= net
->xfrm
.sysctl_acq_expires
;
876 tasklet_hrtimer_start(&x
->mtimer
, ktime_set(net
->xfrm
.sysctl_acq_expires
, 0), HRTIMER_MODE_REL
);
877 net
->xfrm
.state_num
++;
878 xfrm_hash_grow_check(net
, x
->bydst
.next
!= NULL
);
880 x
->km
.state
= XFRM_STATE_DEAD
;
890 *err
= acquire_in_progress
? -EAGAIN
: error
;
891 spin_unlock_bh(&xfrm_state_lock
);
893 xfrm_state_put(to_put
);
898 xfrm_stateonly_find(struct net
*net
, u32 mark
,
899 xfrm_address_t
*daddr
, xfrm_address_t
*saddr
,
900 unsigned short family
, u8 mode
, u8 proto
, u32 reqid
)
903 struct xfrm_state
*rx
= NULL
, *x
= NULL
;
905 spin_lock(&xfrm_state_lock
);
906 h
= xfrm_dst_hash(net
, daddr
, saddr
, reqid
, family
);
907 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+h
, bydst
) {
908 if (x
->props
.family
== family
&&
909 x
->props
.reqid
== reqid
&&
910 (mark
& x
->mark
.m
) == x
->mark
.v
&&
911 !(x
->props
.flags
& XFRM_STATE_WILDRECV
) &&
912 xfrm_state_addr_check(x
, daddr
, saddr
, family
) &&
913 mode
== x
->props
.mode
&&
914 proto
== x
->id
.proto
&&
915 x
->km
.state
== XFRM_STATE_VALID
) {
923 spin_unlock(&xfrm_state_lock
);
928 EXPORT_SYMBOL(xfrm_stateonly_find
);
930 static void __xfrm_state_insert(struct xfrm_state
*x
)
932 struct net
*net
= xs_net(x
);
935 list_add(&x
->km
.all
, &net
->xfrm
.state_all
);
937 h
= xfrm_dst_hash(net
, &x
->id
.daddr
, &x
->props
.saddr
,
938 x
->props
.reqid
, x
->props
.family
);
939 hlist_add_head(&x
->bydst
, net
->xfrm
.state_bydst
+h
);
941 h
= xfrm_src_hash(net
, &x
->id
.daddr
, &x
->props
.saddr
, x
->props
.family
);
942 hlist_add_head(&x
->bysrc
, net
->xfrm
.state_bysrc
+h
);
945 h
= xfrm_spi_hash(net
, &x
->id
.daddr
, x
->id
.spi
, x
->id
.proto
,
948 hlist_add_head(&x
->byspi
, net
->xfrm
.state_byspi
+h
);
951 tasklet_hrtimer_start(&x
->mtimer
, ktime_set(1, 0), HRTIMER_MODE_REL
);
952 if (x
->replay_maxage
)
953 mod_timer(&x
->rtimer
, jiffies
+ x
->replay_maxage
);
955 wake_up(&net
->xfrm
.km_waitq
);
957 net
->xfrm
.state_num
++;
959 xfrm_hash_grow_check(net
, x
->bydst
.next
!= NULL
);
962 /* xfrm_state_lock is held */
963 static void __xfrm_state_bump_genids(struct xfrm_state
*xnew
)
965 struct net
*net
= xs_net(xnew
);
966 unsigned short family
= xnew
->props
.family
;
967 u32 reqid
= xnew
->props
.reqid
;
968 struct xfrm_state
*x
;
970 u32 mark
= xnew
->mark
.v
& xnew
->mark
.m
;
972 h
= xfrm_dst_hash(net
, &xnew
->id
.daddr
, &xnew
->props
.saddr
, reqid
, family
);
973 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+h
, bydst
) {
974 if (x
->props
.family
== family
&&
975 x
->props
.reqid
== reqid
&&
976 (mark
& x
->mark
.m
) == x
->mark
.v
&&
977 xfrm_addr_equal(&x
->id
.daddr
, &xnew
->id
.daddr
, family
) &&
978 xfrm_addr_equal(&x
->props
.saddr
, &xnew
->props
.saddr
, family
))
983 void xfrm_state_insert(struct xfrm_state
*x
)
985 spin_lock_bh(&xfrm_state_lock
);
986 __xfrm_state_bump_genids(x
);
987 __xfrm_state_insert(x
);
988 spin_unlock_bh(&xfrm_state_lock
);
990 EXPORT_SYMBOL(xfrm_state_insert
);
992 /* xfrm_state_lock is held */
993 static struct xfrm_state
*__find_acq_core(struct net
*net
, struct xfrm_mark
*m
,
994 unsigned short family
, u8 mode
,
996 const xfrm_address_t
*daddr
,
997 const xfrm_address_t
*saddr
, int create
)
999 unsigned int h
= xfrm_dst_hash(net
, daddr
, saddr
, reqid
, family
);
1000 struct xfrm_state
*x
;
1001 u32 mark
= m
->v
& m
->m
;
1003 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+h
, bydst
) {
1004 if (x
->props
.reqid
!= reqid
||
1005 x
->props
.mode
!= mode
||
1006 x
->props
.family
!= family
||
1007 x
->km
.state
!= XFRM_STATE_ACQ
||
1009 x
->id
.proto
!= proto
||
1010 (mark
& x
->mark
.m
) != x
->mark
.v
||
1011 !xfrm_addr_equal(&x
->id
.daddr
, daddr
, family
) ||
1012 !xfrm_addr_equal(&x
->props
.saddr
, saddr
, family
))
1022 x
= xfrm_state_alloc(net
);
1026 x
->sel
.daddr
.a4
= daddr
->a4
;
1027 x
->sel
.saddr
.a4
= saddr
->a4
;
1028 x
->sel
.prefixlen_d
= 32;
1029 x
->sel
.prefixlen_s
= 32;
1030 x
->props
.saddr
.a4
= saddr
->a4
;
1031 x
->id
.daddr
.a4
= daddr
->a4
;
1035 *(struct in6_addr
*)x
->sel
.daddr
.a6
= *(struct in6_addr
*)daddr
;
1036 *(struct in6_addr
*)x
->sel
.saddr
.a6
= *(struct in6_addr
*)saddr
;
1037 x
->sel
.prefixlen_d
= 128;
1038 x
->sel
.prefixlen_s
= 128;
1039 *(struct in6_addr
*)x
->props
.saddr
.a6
= *(struct in6_addr
*)saddr
;
1040 *(struct in6_addr
*)x
->id
.daddr
.a6
= *(struct in6_addr
*)daddr
;
1044 x
->km
.state
= XFRM_STATE_ACQ
;
1045 x
->id
.proto
= proto
;
1046 x
->props
.family
= family
;
1047 x
->props
.mode
= mode
;
1048 x
->props
.reqid
= reqid
;
1051 x
->lft
.hard_add_expires_seconds
= net
->xfrm
.sysctl_acq_expires
;
1053 tasklet_hrtimer_start(&x
->mtimer
, ktime_set(net
->xfrm
.sysctl_acq_expires
, 0), HRTIMER_MODE_REL
);
1054 list_add(&x
->km
.all
, &net
->xfrm
.state_all
);
1055 hlist_add_head(&x
->bydst
, net
->xfrm
.state_bydst
+h
);
1056 h
= xfrm_src_hash(net
, daddr
, saddr
, family
);
1057 hlist_add_head(&x
->bysrc
, net
->xfrm
.state_bysrc
+h
);
1059 net
->xfrm
.state_num
++;
1061 xfrm_hash_grow_check(net
, x
->bydst
.next
!= NULL
);
1067 static struct xfrm_state
*__xfrm_find_acq_byseq(struct net
*net
, u32 mark
, u32 seq
);
1069 int xfrm_state_add(struct xfrm_state
*x
)
1071 struct net
*net
= xs_net(x
);
1072 struct xfrm_state
*x1
, *to_put
;
1075 u32 mark
= x
->mark
.v
& x
->mark
.m
;
1076 int use_spi
= xfrm_id_proto_match(x
->id
.proto
, IPSEC_PROTO_ANY
);
1078 family
= x
->props
.family
;
1082 spin_lock_bh(&xfrm_state_lock
);
1084 x1
= __xfrm_state_locate(x
, use_spi
, family
);
1092 if (use_spi
&& x
->km
.seq
) {
1093 x1
= __xfrm_find_acq_byseq(net
, mark
, x
->km
.seq
);
1094 if (x1
&& ((x1
->id
.proto
!= x
->id
.proto
) ||
1095 !xfrm_addr_equal(&x1
->id
.daddr
, &x
->id
.daddr
, family
))) {
1102 x1
= __find_acq_core(net
, &x
->mark
, family
, x
->props
.mode
,
1103 x
->props
.reqid
, x
->id
.proto
,
1104 &x
->id
.daddr
, &x
->props
.saddr
, 0);
1106 __xfrm_state_bump_genids(x
);
1107 __xfrm_state_insert(x
);
1111 spin_unlock_bh(&xfrm_state_lock
);
1114 xfrm_state_delete(x1
);
1119 xfrm_state_put(to_put
);
1123 EXPORT_SYMBOL(xfrm_state_add
);
1125 #ifdef CONFIG_XFRM_MIGRATE
1126 static struct xfrm_state
*xfrm_state_clone(struct xfrm_state
*orig
, int *errp
)
1128 struct net
*net
= xs_net(orig
);
1130 struct xfrm_state
*x
= xfrm_state_alloc(net
);
1134 memcpy(&x
->id
, &orig
->id
, sizeof(x
->id
));
1135 memcpy(&x
->sel
, &orig
->sel
, sizeof(x
->sel
));
1136 memcpy(&x
->lft
, &orig
->lft
, sizeof(x
->lft
));
1137 x
->props
.mode
= orig
->props
.mode
;
1138 x
->props
.replay_window
= orig
->props
.replay_window
;
1139 x
->props
.reqid
= orig
->props
.reqid
;
1140 x
->props
.family
= orig
->props
.family
;
1141 x
->props
.saddr
= orig
->props
.saddr
;
1144 x
->aalg
= xfrm_algo_auth_clone(orig
->aalg
);
1148 x
->props
.aalgo
= orig
->props
.aalgo
;
1151 x
->ealg
= xfrm_algo_clone(orig
->ealg
);
1155 x
->props
.ealgo
= orig
->props
.ealgo
;
1158 x
->calg
= xfrm_algo_clone(orig
->calg
);
1162 x
->props
.calgo
= orig
->props
.calgo
;
1165 x
->encap
= kmemdup(orig
->encap
, sizeof(*x
->encap
), GFP_KERNEL
);
1171 x
->coaddr
= kmemdup(orig
->coaddr
, sizeof(*x
->coaddr
),
1177 if (orig
->replay_esn
) {
1178 err
= xfrm_replay_clone(x
, orig
);
1183 memcpy(&x
->mark
, &orig
->mark
, sizeof(x
->mark
));
1185 err
= xfrm_init_state(x
);
1189 x
->props
.flags
= orig
->props
.flags
;
1191 x
->curlft
.add_time
= orig
->curlft
.add_time
;
1192 x
->km
.state
= orig
->km
.state
;
1193 x
->km
.seq
= orig
->km
.seq
;
1205 /* xfrm_state_lock is held */
1206 struct xfrm_state
* xfrm_migrate_state_find(struct xfrm_migrate
*m
)
1209 struct xfrm_state
*x
;
1212 h
= xfrm_dst_hash(&init_net
, &m
->old_daddr
, &m
->old_saddr
,
1213 m
->reqid
, m
->old_family
);
1214 hlist_for_each_entry(x
, init_net
.xfrm
.state_bydst
+h
, bydst
) {
1215 if (x
->props
.mode
!= m
->mode
||
1216 x
->id
.proto
!= m
->proto
)
1218 if (m
->reqid
&& x
->props
.reqid
!= m
->reqid
)
1220 if (!xfrm_addr_equal(&x
->id
.daddr
, &m
->old_daddr
,
1222 !xfrm_addr_equal(&x
->props
.saddr
, &m
->old_saddr
,
1229 h
= xfrm_src_hash(&init_net
, &m
->old_daddr
, &m
->old_saddr
,
1231 hlist_for_each_entry(x
, init_net
.xfrm
.state_bysrc
+h
, bysrc
) {
1232 if (x
->props
.mode
!= m
->mode
||
1233 x
->id
.proto
!= m
->proto
)
1235 if (!xfrm_addr_equal(&x
->id
.daddr
, &m
->old_daddr
,
1237 !xfrm_addr_equal(&x
->props
.saddr
, &m
->old_saddr
,
1247 EXPORT_SYMBOL(xfrm_migrate_state_find
);
1249 struct xfrm_state
* xfrm_state_migrate(struct xfrm_state
*x
,
1250 struct xfrm_migrate
*m
)
1252 struct xfrm_state
*xc
;
1255 xc
= xfrm_state_clone(x
, &err
);
1259 memcpy(&xc
->id
.daddr
, &m
->new_daddr
, sizeof(xc
->id
.daddr
));
1260 memcpy(&xc
->props
.saddr
, &m
->new_saddr
, sizeof(xc
->props
.saddr
));
1263 if (xfrm_addr_equal(&x
->id
.daddr
, &m
->new_daddr
, m
->new_family
)) {
1264 /* a care is needed when the destination address of the
1265 state is to be updated as it is a part of triplet */
1266 xfrm_state_insert(xc
);
1268 if ((err
= xfrm_state_add(xc
)) < 0)
1277 EXPORT_SYMBOL(xfrm_state_migrate
);
1280 int xfrm_state_update(struct xfrm_state
*x
)
1282 struct xfrm_state
*x1
, *to_put
;
1284 int use_spi
= xfrm_id_proto_match(x
->id
.proto
, IPSEC_PROTO_ANY
);
1288 spin_lock_bh(&xfrm_state_lock
);
1289 x1
= __xfrm_state_locate(x
, use_spi
, x
->props
.family
);
1295 if (xfrm_state_kern(x1
)) {
1301 if (x1
->km
.state
== XFRM_STATE_ACQ
) {
1302 __xfrm_state_insert(x
);
1308 spin_unlock_bh(&xfrm_state_lock
);
1311 xfrm_state_put(to_put
);
1317 xfrm_state_delete(x1
);
1323 spin_lock_bh(&x1
->lock
);
1324 if (likely(x1
->km
.state
== XFRM_STATE_VALID
)) {
1325 if (x
->encap
&& x1
->encap
)
1326 memcpy(x1
->encap
, x
->encap
, sizeof(*x1
->encap
));
1327 if (x
->coaddr
&& x1
->coaddr
) {
1328 memcpy(x1
->coaddr
, x
->coaddr
, sizeof(*x1
->coaddr
));
1330 if (!use_spi
&& memcmp(&x1
->sel
, &x
->sel
, sizeof(x1
->sel
)))
1331 memcpy(&x1
->sel
, &x
->sel
, sizeof(x1
->sel
));
1332 memcpy(&x1
->lft
, &x
->lft
, sizeof(x1
->lft
));
1335 tasklet_hrtimer_start(&x1
->mtimer
, ktime_set(1, 0), HRTIMER_MODE_REL
);
1336 if (x1
->curlft
.use_time
)
1337 xfrm_state_check_expire(x1
);
1340 x
->km
.state
= XFRM_STATE_DEAD
;
1341 __xfrm_state_put(x
);
1343 spin_unlock_bh(&x1
->lock
);
1349 EXPORT_SYMBOL(xfrm_state_update
);
1351 int xfrm_state_check_expire(struct xfrm_state
*x
)
1353 if (!x
->curlft
.use_time
)
1354 x
->curlft
.use_time
= get_seconds();
1356 if (x
->curlft
.bytes
>= x
->lft
.hard_byte_limit
||
1357 x
->curlft
.packets
>= x
->lft
.hard_packet_limit
) {
1358 x
->km
.state
= XFRM_STATE_EXPIRED
;
1359 tasklet_hrtimer_start(&x
->mtimer
, ktime_set(0,0), HRTIMER_MODE_REL
);
1364 (x
->curlft
.bytes
>= x
->lft
.soft_byte_limit
||
1365 x
->curlft
.packets
>= x
->lft
.soft_packet_limit
)) {
1367 km_state_expired(x
, 0, 0);
1371 EXPORT_SYMBOL(xfrm_state_check_expire
);
1374 xfrm_state_lookup(struct net
*net
, u32 mark
, const xfrm_address_t
*daddr
, __be32 spi
,
1375 u8 proto
, unsigned short family
)
1377 struct xfrm_state
*x
;
1379 spin_lock_bh(&xfrm_state_lock
);
1380 x
= __xfrm_state_lookup(net
, mark
, daddr
, spi
, proto
, family
);
1381 spin_unlock_bh(&xfrm_state_lock
);
1384 EXPORT_SYMBOL(xfrm_state_lookup
);
1387 xfrm_state_lookup_byaddr(struct net
*net
, u32 mark
,
1388 const xfrm_address_t
*daddr
, const xfrm_address_t
*saddr
,
1389 u8 proto
, unsigned short family
)
1391 struct xfrm_state
*x
;
1393 spin_lock_bh(&xfrm_state_lock
);
1394 x
= __xfrm_state_lookup_byaddr(net
, mark
, daddr
, saddr
, proto
, family
);
1395 spin_unlock_bh(&xfrm_state_lock
);
1398 EXPORT_SYMBOL(xfrm_state_lookup_byaddr
);
1401 xfrm_find_acq(struct net
*net
, struct xfrm_mark
*mark
, u8 mode
, u32 reqid
, u8 proto
,
1402 const xfrm_address_t
*daddr
, const xfrm_address_t
*saddr
,
1403 int create
, unsigned short family
)
1405 struct xfrm_state
*x
;
1407 spin_lock_bh(&xfrm_state_lock
);
1408 x
= __find_acq_core(net
, mark
, family
, mode
, reqid
, proto
, daddr
, saddr
, create
);
1409 spin_unlock_bh(&xfrm_state_lock
);
1413 EXPORT_SYMBOL(xfrm_find_acq
);
1415 #ifdef CONFIG_XFRM_SUB_POLICY
1417 xfrm_tmpl_sort(struct xfrm_tmpl
**dst
, struct xfrm_tmpl
**src
, int n
,
1418 unsigned short family
)
1421 struct xfrm_state_afinfo
*afinfo
= xfrm_state_get_afinfo(family
);
1423 return -EAFNOSUPPORT
;
1425 spin_lock_bh(&xfrm_state_lock
);
1426 if (afinfo
->tmpl_sort
)
1427 err
= afinfo
->tmpl_sort(dst
, src
, n
);
1428 spin_unlock_bh(&xfrm_state_lock
);
1429 xfrm_state_put_afinfo(afinfo
);
1432 EXPORT_SYMBOL(xfrm_tmpl_sort
);
1435 xfrm_state_sort(struct xfrm_state
**dst
, struct xfrm_state
**src
, int n
,
1436 unsigned short family
)
1439 struct xfrm_state_afinfo
*afinfo
= xfrm_state_get_afinfo(family
);
1441 return -EAFNOSUPPORT
;
1443 spin_lock_bh(&xfrm_state_lock
);
1444 if (afinfo
->state_sort
)
1445 err
= afinfo
->state_sort(dst
, src
, n
);
1446 spin_unlock_bh(&xfrm_state_lock
);
1447 xfrm_state_put_afinfo(afinfo
);
1450 EXPORT_SYMBOL(xfrm_state_sort
);
1453 /* Silly enough, but I'm lazy to build resolution list */
1455 static struct xfrm_state
*__xfrm_find_acq_byseq(struct net
*net
, u32 mark
, u32 seq
)
1459 for (i
= 0; i
<= net
->xfrm
.state_hmask
; i
++) {
1460 struct xfrm_state
*x
;
1462 hlist_for_each_entry(x
, net
->xfrm
.state_bydst
+i
, bydst
) {
1463 if (x
->km
.seq
== seq
&&
1464 (mark
& x
->mark
.m
) == x
->mark
.v
&&
1465 x
->km
.state
== XFRM_STATE_ACQ
) {
1474 struct xfrm_state
*xfrm_find_acq_byseq(struct net
*net
, u32 mark
, u32 seq
)
1476 struct xfrm_state
*x
;
1478 spin_lock_bh(&xfrm_state_lock
);
1479 x
= __xfrm_find_acq_byseq(net
, mark
, seq
);
1480 spin_unlock_bh(&xfrm_state_lock
);
1483 EXPORT_SYMBOL(xfrm_find_acq_byseq
);
1485 u32
xfrm_get_acqseq(void)
1488 static atomic_t acqseq
;
1491 res
= atomic_inc_return(&acqseq
);
1496 EXPORT_SYMBOL(xfrm_get_acqseq
);
1498 int xfrm_alloc_spi(struct xfrm_state
*x
, u32 low
, u32 high
)
1500 struct net
*net
= xs_net(x
);
1502 struct xfrm_state
*x0
;
1504 __be32 minspi
= htonl(low
);
1505 __be32 maxspi
= htonl(high
);
1506 u32 mark
= x
->mark
.v
& x
->mark
.m
;
1508 spin_lock_bh(&x
->lock
);
1509 if (x
->km
.state
== XFRM_STATE_DEAD
)
1518 if (minspi
== maxspi
) {
1519 x0
= xfrm_state_lookup(net
, mark
, &x
->id
.daddr
, minspi
, x
->id
.proto
, x
->props
.family
);
1527 for (h
=0; h
<high
-low
+1; h
++) {
1528 spi
= low
+ net_random()%(high
-low
+1);
1529 x0
= xfrm_state_lookup(net
, mark
, &x
->id
.daddr
, htonl(spi
), x
->id
.proto
, x
->props
.family
);
1531 x
->id
.spi
= htonl(spi
);
1538 spin_lock_bh(&xfrm_state_lock
);
1539 h
= xfrm_spi_hash(net
, &x
->id
.daddr
, x
->id
.spi
, x
->id
.proto
, x
->props
.family
);
1540 hlist_add_head(&x
->byspi
, net
->xfrm
.state_byspi
+h
);
1541 spin_unlock_bh(&xfrm_state_lock
);
1547 spin_unlock_bh(&x
->lock
);
1551 EXPORT_SYMBOL(xfrm_alloc_spi
);
1553 int xfrm_state_walk(struct net
*net
, struct xfrm_state_walk
*walk
,
1554 int (*func
)(struct xfrm_state
*, int, void*),
1557 struct xfrm_state
*state
;
1558 struct xfrm_state_walk
*x
;
1561 if (walk
->seq
!= 0 && list_empty(&walk
->all
))
1564 spin_lock_bh(&xfrm_state_lock
);
1565 if (list_empty(&walk
->all
))
1566 x
= list_first_entry(&net
->xfrm
.state_all
, struct xfrm_state_walk
, all
);
1568 x
= list_entry(&walk
->all
, struct xfrm_state_walk
, all
);
1569 list_for_each_entry_from(x
, &net
->xfrm
.state_all
, all
) {
1570 if (x
->state
== XFRM_STATE_DEAD
)
1572 state
= container_of(x
, struct xfrm_state
, km
);
1573 if (!xfrm_id_proto_match(state
->id
.proto
, walk
->proto
))
1575 err
= func(state
, walk
->seq
, data
);
1577 list_move_tail(&walk
->all
, &x
->all
);
1582 if (walk
->seq
== 0) {
1586 list_del_init(&walk
->all
);
1588 spin_unlock_bh(&xfrm_state_lock
);
1591 EXPORT_SYMBOL(xfrm_state_walk
);
1593 void xfrm_state_walk_init(struct xfrm_state_walk
*walk
, u8 proto
)
1595 INIT_LIST_HEAD(&walk
->all
);
1596 walk
->proto
= proto
;
1597 walk
->state
= XFRM_STATE_DEAD
;
1600 EXPORT_SYMBOL(xfrm_state_walk_init
);
1602 void xfrm_state_walk_done(struct xfrm_state_walk
*walk
)
1604 if (list_empty(&walk
->all
))
1607 spin_lock_bh(&xfrm_state_lock
);
1608 list_del(&walk
->all
);
1609 spin_unlock_bh(&xfrm_state_lock
);
1611 EXPORT_SYMBOL(xfrm_state_walk_done
);
1613 static void xfrm_replay_timer_handler(unsigned long data
)
1615 struct xfrm_state
*x
= (struct xfrm_state
*)data
;
1617 spin_lock(&x
->lock
);
1619 if (x
->km
.state
== XFRM_STATE_VALID
) {
1620 if (xfrm_aevent_is_on(xs_net(x
)))
1621 x
->repl
->notify(x
, XFRM_REPLAY_TIMEOUT
);
1623 x
->xflags
|= XFRM_TIME_DEFER
;
1626 spin_unlock(&x
->lock
);
1629 static LIST_HEAD(xfrm_km_list
);
1631 void km_policy_notify(struct xfrm_policy
*xp
, int dir
, const struct km_event
*c
)
1633 struct xfrm_mgr
*km
;
1636 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
)
1637 if (km
->notify_policy
)
1638 km
->notify_policy(xp
, dir
, c
);
1642 void km_state_notify(struct xfrm_state
*x
, const struct km_event
*c
)
1644 struct xfrm_mgr
*km
;
1646 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
)
1652 EXPORT_SYMBOL(km_policy_notify
);
1653 EXPORT_SYMBOL(km_state_notify
);
1655 void km_state_expired(struct xfrm_state
*x
, int hard
, u32 portid
)
1657 struct net
*net
= xs_net(x
);
1662 c
.event
= XFRM_MSG_EXPIRE
;
1663 km_state_notify(x
, &c
);
1666 wake_up(&net
->xfrm
.km_waitq
);
1669 EXPORT_SYMBOL(km_state_expired
);
1671 * We send to all registered managers regardless of failure
1672 * We are happy with one success
1674 int km_query(struct xfrm_state
*x
, struct xfrm_tmpl
*t
, struct xfrm_policy
*pol
)
1676 int err
= -EINVAL
, acqret
;
1677 struct xfrm_mgr
*km
;
1680 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
) {
1681 acqret
= km
->acquire(x
, t
, pol
);
1688 EXPORT_SYMBOL(km_query
);
1690 int km_new_mapping(struct xfrm_state
*x
, xfrm_address_t
*ipaddr
, __be16 sport
)
1693 struct xfrm_mgr
*km
;
1696 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
) {
1697 if (km
->new_mapping
)
1698 err
= km
->new_mapping(x
, ipaddr
, sport
);
1705 EXPORT_SYMBOL(km_new_mapping
);
1707 void km_policy_expired(struct xfrm_policy
*pol
, int dir
, int hard
, u32 portid
)
1709 struct net
*net
= xp_net(pol
);
1714 c
.event
= XFRM_MSG_POLEXPIRE
;
1715 km_policy_notify(pol
, dir
, &c
);
1718 wake_up(&net
->xfrm
.km_waitq
);
1720 EXPORT_SYMBOL(km_policy_expired
);
1722 #ifdef CONFIG_XFRM_MIGRATE
1723 int km_migrate(const struct xfrm_selector
*sel
, u8 dir
, u8 type
,
1724 const struct xfrm_migrate
*m
, int num_migrate
,
1725 const struct xfrm_kmaddress
*k
)
1729 struct xfrm_mgr
*km
;
1732 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
) {
1734 ret
= km
->migrate(sel
, dir
, type
, m
, num_migrate
, k
);
1742 EXPORT_SYMBOL(km_migrate
);
1745 int km_report(struct net
*net
, u8 proto
, struct xfrm_selector
*sel
, xfrm_address_t
*addr
)
1749 struct xfrm_mgr
*km
;
1752 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
) {
1754 ret
= km
->report(net
, proto
, sel
, addr
);
1762 EXPORT_SYMBOL(km_report
);
1764 int xfrm_user_policy(struct sock
*sk
, int optname
, u8 __user
*optval
, int optlen
)
1768 struct xfrm_mgr
*km
;
1769 struct xfrm_policy
*pol
= NULL
;
1771 if (optlen
<= 0 || optlen
> PAGE_SIZE
)
1774 data
= kmalloc(optlen
, GFP_KERNEL
);
1779 if (copy_from_user(data
, optval
, optlen
))
1784 list_for_each_entry_rcu(km
, &xfrm_km_list
, list
) {
1785 pol
= km
->compile_policy(sk
, optname
, data
,
1793 xfrm_sk_policy_insert(sk
, err
, pol
);
1802 EXPORT_SYMBOL(xfrm_user_policy
);
1804 static DEFINE_SPINLOCK(xfrm_km_lock
);
1806 int xfrm_register_km(struct xfrm_mgr
*km
)
1808 spin_lock_bh(&xfrm_km_lock
);
1809 list_add_tail_rcu(&km
->list
, &xfrm_km_list
);
1810 spin_unlock_bh(&xfrm_km_lock
);
1813 EXPORT_SYMBOL(xfrm_register_km
);
1815 int xfrm_unregister_km(struct xfrm_mgr
*km
)
1817 spin_lock_bh(&xfrm_km_lock
);
1818 list_del_rcu(&km
->list
);
1819 spin_unlock_bh(&xfrm_km_lock
);
1823 EXPORT_SYMBOL(xfrm_unregister_km
);
1825 int xfrm_state_register_afinfo(struct xfrm_state_afinfo
*afinfo
)
1828 if (unlikely(afinfo
== NULL
))
1830 if (unlikely(afinfo
->family
>= NPROTO
))
1831 return -EAFNOSUPPORT
;
1832 spin_lock_bh(&xfrm_state_afinfo_lock
);
1833 if (unlikely(xfrm_state_afinfo
[afinfo
->family
] != NULL
))
1836 rcu_assign_pointer(xfrm_state_afinfo
[afinfo
->family
], afinfo
);
1837 spin_unlock_bh(&xfrm_state_afinfo_lock
);
1840 EXPORT_SYMBOL(xfrm_state_register_afinfo
);
1842 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo
*afinfo
)
1845 if (unlikely(afinfo
== NULL
))
1847 if (unlikely(afinfo
->family
>= NPROTO
))
1848 return -EAFNOSUPPORT
;
1849 spin_lock_bh(&xfrm_state_afinfo_lock
);
1850 if (likely(xfrm_state_afinfo
[afinfo
->family
] != NULL
)) {
1851 if (unlikely(xfrm_state_afinfo
[afinfo
->family
] != afinfo
))
1854 RCU_INIT_POINTER(xfrm_state_afinfo
[afinfo
->family
], NULL
);
1856 spin_unlock_bh(&xfrm_state_afinfo_lock
);
1860 EXPORT_SYMBOL(xfrm_state_unregister_afinfo
);
1862 static struct xfrm_state_afinfo
*xfrm_state_get_afinfo(unsigned int family
)
1864 struct xfrm_state_afinfo
*afinfo
;
1865 if (unlikely(family
>= NPROTO
))
1868 afinfo
= rcu_dereference(xfrm_state_afinfo
[family
]);
1869 if (unlikely(!afinfo
))
1874 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo
*afinfo
)
1879 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1880 void xfrm_state_delete_tunnel(struct xfrm_state
*x
)
1883 struct xfrm_state
*t
= x
->tunnel
;
1885 if (atomic_read(&t
->tunnel_users
) == 2)
1886 xfrm_state_delete(t
);
1887 atomic_dec(&t
->tunnel_users
);
1892 EXPORT_SYMBOL(xfrm_state_delete_tunnel
);
1894 int xfrm_state_mtu(struct xfrm_state
*x
, int mtu
)
1898 spin_lock_bh(&x
->lock
);
1899 if (x
->km
.state
== XFRM_STATE_VALID
&&
1900 x
->type
&& x
->type
->get_mtu
)
1901 res
= x
->type
->get_mtu(x
, mtu
);
1903 res
= mtu
- x
->props
.header_len
;
1904 spin_unlock_bh(&x
->lock
);
1908 int __xfrm_init_state(struct xfrm_state
*x
, bool init_replay
)
1910 struct xfrm_state_afinfo
*afinfo
;
1911 struct xfrm_mode
*inner_mode
;
1912 int family
= x
->props
.family
;
1915 err
= -EAFNOSUPPORT
;
1916 afinfo
= xfrm_state_get_afinfo(family
);
1921 if (afinfo
->init_flags
)
1922 err
= afinfo
->init_flags(x
);
1924 xfrm_state_put_afinfo(afinfo
);
1929 err
= -EPROTONOSUPPORT
;
1931 if (x
->sel
.family
!= AF_UNSPEC
) {
1932 inner_mode
= xfrm_get_mode(x
->props
.mode
, x
->sel
.family
);
1933 if (inner_mode
== NULL
)
1936 if (!(inner_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
1937 family
!= x
->sel
.family
) {
1938 xfrm_put_mode(inner_mode
);
1942 x
->inner_mode
= inner_mode
;
1944 struct xfrm_mode
*inner_mode_iaf
;
1945 int iafamily
= AF_INET
;
1947 inner_mode
= xfrm_get_mode(x
->props
.mode
, x
->props
.family
);
1948 if (inner_mode
== NULL
)
1951 if (!(inner_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
)) {
1952 xfrm_put_mode(inner_mode
);
1955 x
->inner_mode
= inner_mode
;
1957 if (x
->props
.family
== AF_INET
)
1958 iafamily
= AF_INET6
;
1960 inner_mode_iaf
= xfrm_get_mode(x
->props
.mode
, iafamily
);
1961 if (inner_mode_iaf
) {
1962 if (inner_mode_iaf
->flags
& XFRM_MODE_FLAG_TUNNEL
)
1963 x
->inner_mode_iaf
= inner_mode_iaf
;
1965 xfrm_put_mode(inner_mode_iaf
);
1969 x
->type
= xfrm_get_type(x
->id
.proto
, family
);
1970 if (x
->type
== NULL
)
1973 err
= x
->type
->init_state(x
);
1977 x
->outer_mode
= xfrm_get_mode(x
->props
.mode
, family
);
1978 if (x
->outer_mode
== NULL
) {
1979 err
= -EPROTONOSUPPORT
;
1984 err
= xfrm_init_replay(x
);
1989 x
->km
.state
= XFRM_STATE_VALID
;
1995 EXPORT_SYMBOL(__xfrm_init_state
);
1997 int xfrm_init_state(struct xfrm_state
*x
)
1999 return __xfrm_init_state(x
, true);
2002 EXPORT_SYMBOL(xfrm_init_state
);
2004 int __net_init
xfrm_state_init(struct net
*net
)
2008 INIT_LIST_HEAD(&net
->xfrm
.state_all
);
2010 sz
= sizeof(struct hlist_head
) * 8;
2012 net
->xfrm
.state_bydst
= xfrm_hash_alloc(sz
);
2013 if (!net
->xfrm
.state_bydst
)
2015 net
->xfrm
.state_bysrc
= xfrm_hash_alloc(sz
);
2016 if (!net
->xfrm
.state_bysrc
)
2018 net
->xfrm
.state_byspi
= xfrm_hash_alloc(sz
);
2019 if (!net
->xfrm
.state_byspi
)
2021 net
->xfrm
.state_hmask
= ((sz
/ sizeof(struct hlist_head
)) - 1);
2023 net
->xfrm
.state_num
= 0;
2024 INIT_WORK(&net
->xfrm
.state_hash_work
, xfrm_hash_resize
);
2025 INIT_HLIST_HEAD(&net
->xfrm
.state_gc_list
);
2026 INIT_WORK(&net
->xfrm
.state_gc_work
, xfrm_state_gc_task
);
2027 init_waitqueue_head(&net
->xfrm
.km_waitq
);
2031 xfrm_hash_free(net
->xfrm
.state_bysrc
, sz
);
2033 xfrm_hash_free(net
->xfrm
.state_bydst
, sz
);
2038 void xfrm_state_fini(struct net
*net
)
2040 struct xfrm_audit audit_info
;
2043 flush_work(&net
->xfrm
.state_hash_work
);
2044 audit_info
.loginuid
= INVALID_UID
;
2045 audit_info
.sessionid
= -1;
2046 audit_info
.secid
= 0;
2047 xfrm_state_flush(net
, IPSEC_PROTO_ANY
, &audit_info
);
2048 flush_work(&net
->xfrm
.state_gc_work
);
2050 WARN_ON(!list_empty(&net
->xfrm
.state_all
));
2052 sz
= (net
->xfrm
.state_hmask
+ 1) * sizeof(struct hlist_head
);
2053 WARN_ON(!hlist_empty(net
->xfrm
.state_byspi
));
2054 xfrm_hash_free(net
->xfrm
.state_byspi
, sz
);
2055 WARN_ON(!hlist_empty(net
->xfrm
.state_bysrc
));
2056 xfrm_hash_free(net
->xfrm
.state_bysrc
, sz
);
2057 WARN_ON(!hlist_empty(net
->xfrm
.state_bydst
));
2058 xfrm_hash_free(net
->xfrm
.state_bydst
, sz
);
2061 #ifdef CONFIG_AUDITSYSCALL
2062 static void xfrm_audit_helper_sainfo(struct xfrm_state
*x
,
2063 struct audit_buffer
*audit_buf
)
2065 struct xfrm_sec_ctx
*ctx
= x
->security
;
2066 u32 spi
= ntohl(x
->id
.spi
);
2069 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2070 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2072 switch(x
->props
.family
) {
2074 audit_log_format(audit_buf
, " src=%pI4 dst=%pI4",
2075 &x
->props
.saddr
.a4
, &x
->id
.daddr
.a4
);
2078 audit_log_format(audit_buf
, " src=%pI6 dst=%pI6",
2079 x
->props
.saddr
.a6
, x
->id
.daddr
.a6
);
2083 audit_log_format(audit_buf
, " spi=%u(0x%x)", spi
, spi
);
2086 static void xfrm_audit_helper_pktinfo(struct sk_buff
*skb
, u16 family
,
2087 struct audit_buffer
*audit_buf
)
2089 const struct iphdr
*iph4
;
2090 const struct ipv6hdr
*iph6
;
2095 audit_log_format(audit_buf
, " src=%pI4 dst=%pI4",
2096 &iph4
->saddr
, &iph4
->daddr
);
2099 iph6
= ipv6_hdr(skb
);
2100 audit_log_format(audit_buf
,
2101 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2102 &iph6
->saddr
,&iph6
->daddr
,
2103 iph6
->flow_lbl
[0] & 0x0f,
2110 void xfrm_audit_state_add(struct xfrm_state
*x
, int result
,
2111 kuid_t auid
, u32 sessionid
, u32 secid
)
2113 struct audit_buffer
*audit_buf
;
2115 audit_buf
= xfrm_audit_start("SAD-add");
2116 if (audit_buf
== NULL
)
2118 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2119 xfrm_audit_helper_sainfo(x
, audit_buf
);
2120 audit_log_format(audit_buf
, " res=%u", result
);
2121 audit_log_end(audit_buf
);
2123 EXPORT_SYMBOL_GPL(xfrm_audit_state_add
);
2125 void xfrm_audit_state_delete(struct xfrm_state
*x
, int result
,
2126 kuid_t auid
, u32 sessionid
, u32 secid
)
2128 struct audit_buffer
*audit_buf
;
2130 audit_buf
= xfrm_audit_start("SAD-delete");
2131 if (audit_buf
== NULL
)
2133 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2134 xfrm_audit_helper_sainfo(x
, audit_buf
);
2135 audit_log_format(audit_buf
, " res=%u", result
);
2136 audit_log_end(audit_buf
);
2138 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete
);
2140 void xfrm_audit_state_replay_overflow(struct xfrm_state
*x
,
2141 struct sk_buff
*skb
)
2143 struct audit_buffer
*audit_buf
;
2146 audit_buf
= xfrm_audit_start("SA-replay-overflow");
2147 if (audit_buf
== NULL
)
2149 xfrm_audit_helper_pktinfo(skb
, x
->props
.family
, audit_buf
);
2150 /* don't record the sequence number because it's inherent in this kind
2151 * of audit message */
2152 spi
= ntohl(x
->id
.spi
);
2153 audit_log_format(audit_buf
, " spi=%u(0x%x)", spi
, spi
);
2154 audit_log_end(audit_buf
);
2156 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow
);
2158 void xfrm_audit_state_replay(struct xfrm_state
*x
,
2159 struct sk_buff
*skb
, __be32 net_seq
)
2161 struct audit_buffer
*audit_buf
;
2164 audit_buf
= xfrm_audit_start("SA-replayed-pkt");
2165 if (audit_buf
== NULL
)
2167 xfrm_audit_helper_pktinfo(skb
, x
->props
.family
, audit_buf
);
2168 spi
= ntohl(x
->id
.spi
);
2169 audit_log_format(audit_buf
, " spi=%u(0x%x) seqno=%u",
2170 spi
, spi
, ntohl(net_seq
));
2171 audit_log_end(audit_buf
);
2173 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay
);
2175 void xfrm_audit_state_notfound_simple(struct sk_buff
*skb
, u16 family
)
2177 struct audit_buffer
*audit_buf
;
2179 audit_buf
= xfrm_audit_start("SA-notfound");
2180 if (audit_buf
== NULL
)
2182 xfrm_audit_helper_pktinfo(skb
, family
, audit_buf
);
2183 audit_log_end(audit_buf
);
2185 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple
);
2187 void xfrm_audit_state_notfound(struct sk_buff
*skb
, u16 family
,
2188 __be32 net_spi
, __be32 net_seq
)
2190 struct audit_buffer
*audit_buf
;
2193 audit_buf
= xfrm_audit_start("SA-notfound");
2194 if (audit_buf
== NULL
)
2196 xfrm_audit_helper_pktinfo(skb
, family
, audit_buf
);
2197 spi
= ntohl(net_spi
);
2198 audit_log_format(audit_buf
, " spi=%u(0x%x) seqno=%u",
2199 spi
, spi
, ntohl(net_seq
));
2200 audit_log_end(audit_buf
);
2202 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound
);
2204 void xfrm_audit_state_icvfail(struct xfrm_state
*x
,
2205 struct sk_buff
*skb
, u8 proto
)
2207 struct audit_buffer
*audit_buf
;
2211 audit_buf
= xfrm_audit_start("SA-icv-failure");
2212 if (audit_buf
== NULL
)
2214 xfrm_audit_helper_pktinfo(skb
, x
->props
.family
, audit_buf
);
2215 if (xfrm_parse_spi(skb
, proto
, &net_spi
, &net_seq
) == 0) {
2216 u32 spi
= ntohl(net_spi
);
2217 audit_log_format(audit_buf
, " spi=%u(0x%x) seqno=%u",
2218 spi
, spi
, ntohl(net_seq
));
2220 audit_log_end(audit_buf
);
2222 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail
);
2223 #endif /* CONFIG_AUDITSYSCALL */