6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * Kazunori MIYAZAWA @USAGI
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
16 #include <linux/err.h>
17 #include <linux/slab.h>
18 #include <linux/kmod.h>
19 #include <linux/list.h>
20 #include <linux/spinlock.h>
21 #include <linux/workqueue.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/module.h>
26 #include <linux/cache.h>
27 #include <linux/audit.h>
31 #ifdef CONFIG_XFRM_STATISTICS
35 #include "xfrm_hash.h"
37 int sysctl_xfrm_larval_drop __read_mostly
= 1;
39 #ifdef CONFIG_XFRM_STATISTICS
40 DEFINE_SNMP_STAT(struct linux_xfrm_mib
, xfrm_statistics
) __read_mostly
;
41 EXPORT_SYMBOL(xfrm_statistics
);
44 DEFINE_MUTEX(xfrm_cfg_mutex
);
45 EXPORT_SYMBOL(xfrm_cfg_mutex
);
47 static DEFINE_RWLOCK(xfrm_policy_lock
);
49 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
50 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
52 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
54 static HLIST_HEAD(xfrm_policy_gc_list
);
55 static DEFINE_SPINLOCK(xfrm_policy_gc_lock
);
57 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
58 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
59 static void xfrm_init_pmtu(struct dst_entry
*dst
);
62 __xfrm4_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
64 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
65 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
66 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
67 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
68 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
69 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
73 __xfrm6_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
75 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
76 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
77 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
78 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
79 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
80 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
83 int xfrm_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
,
84 unsigned short family
)
88 return __xfrm4_selector_match(sel
, fl
);
90 return __xfrm6_selector_match(sel
, fl
);
95 static inline struct dst_entry
*__xfrm_dst_lookup(int tos
,
96 xfrm_address_t
*saddr
,
97 xfrm_address_t
*daddr
,
100 struct xfrm_policy_afinfo
*afinfo
;
101 struct dst_entry
*dst
;
103 afinfo
= xfrm_policy_get_afinfo(family
);
104 if (unlikely(afinfo
== NULL
))
105 return ERR_PTR(-EAFNOSUPPORT
);
107 dst
= afinfo
->dst_lookup(tos
, saddr
, daddr
);
109 xfrm_policy_put_afinfo(afinfo
);
114 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
, int tos
,
115 xfrm_address_t
*prev_saddr
,
116 xfrm_address_t
*prev_daddr
,
119 xfrm_address_t
*saddr
= &x
->props
.saddr
;
120 xfrm_address_t
*daddr
= &x
->id
.daddr
;
121 struct dst_entry
*dst
;
123 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
127 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
132 dst
= __xfrm_dst_lookup(tos
, saddr
, daddr
, family
);
135 if (prev_saddr
!= saddr
)
136 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
137 if (prev_daddr
!= daddr
)
138 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
144 static inline unsigned long make_jiffies(long secs
)
146 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
147 return MAX_SCHEDULE_TIMEOUT
-1;
152 static void xfrm_policy_timer(unsigned long data
)
154 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
155 unsigned long now
= get_seconds();
156 long next
= LONG_MAX
;
160 read_lock(&xp
->lock
);
165 dir
= xfrm_policy_id2dir(xp
->index
);
167 if (xp
->lft
.hard_add_expires_seconds
) {
168 long tmo
= xp
->lft
.hard_add_expires_seconds
+
169 xp
->curlft
.add_time
- now
;
175 if (xp
->lft
.hard_use_expires_seconds
) {
176 long tmo
= xp
->lft
.hard_use_expires_seconds
+
177 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
183 if (xp
->lft
.soft_add_expires_seconds
) {
184 long tmo
= xp
->lft
.soft_add_expires_seconds
+
185 xp
->curlft
.add_time
- now
;
188 tmo
= XFRM_KM_TIMEOUT
;
193 if (xp
->lft
.soft_use_expires_seconds
) {
194 long tmo
= xp
->lft
.soft_use_expires_seconds
+
195 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
198 tmo
= XFRM_KM_TIMEOUT
;
205 km_policy_expired(xp
, dir
, 0, 0);
206 if (next
!= LONG_MAX
&&
207 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
211 read_unlock(&xp
->lock
);
216 read_unlock(&xp
->lock
);
217 if (!xfrm_policy_delete(xp
, dir
))
218 km_policy_expired(xp
, dir
, 1, 0);
223 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
227 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
229 struct xfrm_policy
*policy
;
231 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
234 write_pnet(&policy
->xp_net
, net
);
235 INIT_LIST_HEAD(&policy
->walk
.all
);
236 INIT_HLIST_NODE(&policy
->bydst
);
237 INIT_HLIST_NODE(&policy
->byidx
);
238 rwlock_init(&policy
->lock
);
239 atomic_set(&policy
->refcnt
, 1);
240 setup_timer(&policy
->timer
, xfrm_policy_timer
,
241 (unsigned long)policy
);
245 EXPORT_SYMBOL(xfrm_policy_alloc
);
247 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
249 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
251 BUG_ON(!policy
->walk
.dead
);
253 BUG_ON(policy
->bundles
);
255 if (del_timer(&policy
->timer
))
258 security_xfrm_policy_free(policy
->security
);
261 EXPORT_SYMBOL(xfrm_policy_destroy
);
263 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
265 struct dst_entry
*dst
;
267 while ((dst
= policy
->bundles
) != NULL
) {
268 policy
->bundles
= dst
->next
;
272 if (del_timer(&policy
->timer
))
273 atomic_dec(&policy
->refcnt
);
275 if (atomic_read(&policy
->refcnt
) > 1)
278 xfrm_pol_put(policy
);
281 static void xfrm_policy_gc_task(struct work_struct
*work
)
283 struct xfrm_policy
*policy
;
284 struct hlist_node
*entry
, *tmp
;
285 struct hlist_head gc_list
;
287 spin_lock_bh(&xfrm_policy_gc_lock
);
288 gc_list
.first
= xfrm_policy_gc_list
.first
;
289 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
290 spin_unlock_bh(&xfrm_policy_gc_lock
);
292 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
293 xfrm_policy_gc_kill(policy
);
295 static DECLARE_WORK(xfrm_policy_gc_work
, xfrm_policy_gc_task
);
297 /* Rule must be locked. Release descentant resources, announce
298 * entry dead. The rule must be unlinked from lists to the moment.
301 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
305 write_lock_bh(&policy
->lock
);
306 dead
= policy
->walk
.dead
;
307 policy
->walk
.dead
= 1;
308 write_unlock_bh(&policy
->lock
);
310 if (unlikely(dead
)) {
315 spin_lock_bh(&xfrm_policy_gc_lock
);
316 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
317 spin_unlock_bh(&xfrm_policy_gc_lock
);
319 schedule_work(&xfrm_policy_gc_work
);
322 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
324 static inline unsigned int idx_hash(u32 index
)
326 return __idx_hash(index
, init_net
.xfrm
.policy_idx_hmask
);
329 static struct hlist_head
*policy_hash_bysel(struct xfrm_selector
*sel
, unsigned short family
, int dir
)
331 unsigned int hmask
= init_net
.xfrm
.policy_bydst
[dir
].hmask
;
332 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
334 return (hash
== hmask
+ 1 ?
335 &init_net
.xfrm
.policy_inexact
[dir
] :
336 init_net
.xfrm
.policy_bydst
[dir
].table
+ hash
);
339 static struct hlist_head
*policy_hash_direct(xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
341 unsigned int hmask
= init_net
.xfrm
.policy_bydst
[dir
].hmask
;
342 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
344 return init_net
.xfrm
.policy_bydst
[dir
].table
+ hash
;
347 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
348 struct hlist_head
*ndsttable
,
349 unsigned int nhashmask
)
351 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
352 struct xfrm_policy
*pol
;
356 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
359 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
360 pol
->family
, nhashmask
);
363 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
369 hlist_add_after(entry0
, &pol
->bydst
);
373 if (!hlist_empty(list
)) {
379 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
380 struct hlist_head
*nidxtable
,
381 unsigned int nhashmask
)
383 struct hlist_node
*entry
, *tmp
;
384 struct xfrm_policy
*pol
;
386 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
389 h
= __idx_hash(pol
->index
, nhashmask
);
390 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
394 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
396 return ((old_hmask
+ 1) << 1) - 1;
399 static void xfrm_bydst_resize(int dir
)
401 unsigned int hmask
= init_net
.xfrm
.policy_bydst
[dir
].hmask
;
402 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
403 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
404 struct hlist_head
*odst
= init_net
.xfrm
.policy_bydst
[dir
].table
;
405 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
411 write_lock_bh(&xfrm_policy_lock
);
413 for (i
= hmask
; i
>= 0; i
--)
414 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
416 init_net
.xfrm
.policy_bydst
[dir
].table
= ndst
;
417 init_net
.xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
419 write_unlock_bh(&xfrm_policy_lock
);
421 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
424 static void xfrm_byidx_resize(int total
)
426 unsigned int hmask
= init_net
.xfrm
.policy_idx_hmask
;
427 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
428 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
429 struct hlist_head
*oidx
= init_net
.xfrm
.policy_byidx
;
430 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
436 write_lock_bh(&xfrm_policy_lock
);
438 for (i
= hmask
; i
>= 0; i
--)
439 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
441 init_net
.xfrm
.policy_byidx
= nidx
;
442 init_net
.xfrm
.policy_idx_hmask
= nhashmask
;
444 write_unlock_bh(&xfrm_policy_lock
);
446 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
449 static inline int xfrm_bydst_should_resize(int dir
, int *total
)
451 unsigned int cnt
= init_net
.xfrm
.policy_count
[dir
];
452 unsigned int hmask
= init_net
.xfrm
.policy_bydst
[dir
].hmask
;
457 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
464 static inline int xfrm_byidx_should_resize(int total
)
466 unsigned int hmask
= init_net
.xfrm
.policy_idx_hmask
;
468 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
475 void xfrm_spd_getinfo(struct xfrmk_spdinfo
*si
)
477 read_lock_bh(&xfrm_policy_lock
);
478 si
->incnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
];
479 si
->outcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
];
480 si
->fwdcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
];
481 si
->inscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
482 si
->outscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
483 si
->fwdscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
484 si
->spdhcnt
= init_net
.xfrm
.policy_idx_hmask
;
485 si
->spdhmcnt
= xfrm_policy_hashmax
;
486 read_unlock_bh(&xfrm_policy_lock
);
488 EXPORT_SYMBOL(xfrm_spd_getinfo
);
490 static DEFINE_MUTEX(hash_resize_mutex
);
491 static void xfrm_hash_resize(struct work_struct
*__unused
)
495 mutex_lock(&hash_resize_mutex
);
498 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
499 if (xfrm_bydst_should_resize(dir
, &total
))
500 xfrm_bydst_resize(dir
);
502 if (xfrm_byidx_should_resize(total
))
503 xfrm_byidx_resize(total
);
505 mutex_unlock(&hash_resize_mutex
);
508 static DECLARE_WORK(xfrm_hash_work
, xfrm_hash_resize
);
510 /* Generate new index... KAME seems to generate them ordered by cost
511 * of an absolute inpredictability of ordering of rules. This will not pass. */
512 static u32
xfrm_gen_index(int dir
)
514 static u32 idx_generator
;
517 struct hlist_node
*entry
;
518 struct hlist_head
*list
;
519 struct xfrm_policy
*p
;
523 idx
= (idx_generator
| dir
);
527 list
= init_net
.xfrm
.policy_byidx
+ idx_hash(idx
);
529 hlist_for_each_entry(p
, entry
, list
, byidx
) {
530 if (p
->index
== idx
) {
540 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
542 u32
*p1
= (u32
*) s1
;
543 u32
*p2
= (u32
*) s2
;
544 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
547 for (i
= 0; i
< len
; i
++) {
555 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
557 struct xfrm_policy
*pol
;
558 struct xfrm_policy
*delpol
;
559 struct hlist_head
*chain
;
560 struct hlist_node
*entry
, *newpos
;
561 struct dst_entry
*gc_list
;
563 write_lock_bh(&xfrm_policy_lock
);
564 chain
= policy_hash_bysel(&policy
->selector
, policy
->family
, dir
);
567 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
568 if (pol
->type
== policy
->type
&&
569 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
570 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
573 write_unlock_bh(&xfrm_policy_lock
);
577 if (policy
->priority
> pol
->priority
)
579 } else if (policy
->priority
>= pol
->priority
) {
580 newpos
= &pol
->bydst
;
587 hlist_add_after(newpos
, &policy
->bydst
);
589 hlist_add_head(&policy
->bydst
, chain
);
590 xfrm_pol_hold(policy
);
591 init_net
.xfrm
.policy_count
[dir
]++;
592 atomic_inc(&flow_cache_genid
);
594 hlist_del(&delpol
->bydst
);
595 hlist_del(&delpol
->byidx
);
596 list_del(&delpol
->walk
.all
);
597 init_net
.xfrm
.policy_count
[dir
]--;
599 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(dir
);
600 hlist_add_head(&policy
->byidx
, init_net
.xfrm
.policy_byidx
+idx_hash(policy
->index
));
601 policy
->curlft
.add_time
= get_seconds();
602 policy
->curlft
.use_time
= 0;
603 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
604 xfrm_pol_hold(policy
);
605 list_add(&policy
->walk
.all
, &init_net
.xfrm
.policy_all
);
606 write_unlock_bh(&xfrm_policy_lock
);
609 xfrm_policy_kill(delpol
);
610 else if (xfrm_bydst_should_resize(dir
, NULL
))
611 schedule_work(&xfrm_hash_work
);
613 read_lock_bh(&xfrm_policy_lock
);
615 entry
= &policy
->bydst
;
616 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
617 struct dst_entry
*dst
;
619 write_lock(&policy
->lock
);
620 dst
= policy
->bundles
;
622 struct dst_entry
*tail
= dst
;
625 tail
->next
= gc_list
;
628 policy
->bundles
= NULL
;
630 write_unlock(&policy
->lock
);
632 read_unlock_bh(&xfrm_policy_lock
);
635 struct dst_entry
*dst
= gc_list
;
643 EXPORT_SYMBOL(xfrm_policy_insert
);
645 struct xfrm_policy
*xfrm_policy_bysel_ctx(u8 type
, int dir
,
646 struct xfrm_selector
*sel
,
647 struct xfrm_sec_ctx
*ctx
, int delete,
650 struct xfrm_policy
*pol
, *ret
;
651 struct hlist_head
*chain
;
652 struct hlist_node
*entry
;
655 write_lock_bh(&xfrm_policy_lock
);
656 chain
= policy_hash_bysel(sel
, sel
->family
, dir
);
658 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
659 if (pol
->type
== type
&&
660 !selector_cmp(sel
, &pol
->selector
) &&
661 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
664 *err
= security_xfrm_policy_delete(
667 write_unlock_bh(&xfrm_policy_lock
);
670 hlist_del(&pol
->bydst
);
671 hlist_del(&pol
->byidx
);
672 list_del(&pol
->walk
.all
);
673 init_net
.xfrm
.policy_count
[dir
]--;
679 write_unlock_bh(&xfrm_policy_lock
);
682 atomic_inc(&flow_cache_genid
);
683 xfrm_policy_kill(ret
);
687 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
689 struct xfrm_policy
*xfrm_policy_byid(u8 type
, int dir
, u32 id
, int delete,
692 struct xfrm_policy
*pol
, *ret
;
693 struct hlist_head
*chain
;
694 struct hlist_node
*entry
;
697 if (xfrm_policy_id2dir(id
) != dir
)
701 write_lock_bh(&xfrm_policy_lock
);
702 chain
= init_net
.xfrm
.policy_byidx
+ idx_hash(id
);
704 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
705 if (pol
->type
== type
&& pol
->index
== id
) {
708 *err
= security_xfrm_policy_delete(
711 write_unlock_bh(&xfrm_policy_lock
);
714 hlist_del(&pol
->bydst
);
715 hlist_del(&pol
->byidx
);
716 list_del(&pol
->walk
.all
);
717 init_net
.xfrm
.policy_count
[dir
]--;
723 write_unlock_bh(&xfrm_policy_lock
);
726 atomic_inc(&flow_cache_genid
);
727 xfrm_policy_kill(ret
);
731 EXPORT_SYMBOL(xfrm_policy_byid
);
733 #ifdef CONFIG_SECURITY_NETWORK_XFRM
735 xfrm_policy_flush_secctx_check(u8 type
, struct xfrm_audit
*audit_info
)
739 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
740 struct xfrm_policy
*pol
;
741 struct hlist_node
*entry
;
744 hlist_for_each_entry(pol
, entry
,
745 &init_net
.xfrm
.policy_inexact
[dir
], bydst
) {
746 if (pol
->type
!= type
)
748 err
= security_xfrm_policy_delete(pol
->security
);
750 xfrm_audit_policy_delete(pol
, 0,
751 audit_info
->loginuid
,
752 audit_info
->sessionid
,
757 for (i
= init_net
.xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
758 hlist_for_each_entry(pol
, entry
,
759 init_net
.xfrm
.policy_bydst
[dir
].table
+ i
,
761 if (pol
->type
!= type
)
763 err
= security_xfrm_policy_delete(
766 xfrm_audit_policy_delete(pol
, 0,
767 audit_info
->loginuid
,
768 audit_info
->sessionid
,
779 xfrm_policy_flush_secctx_check(u8 type
, struct xfrm_audit
*audit_info
)
785 int xfrm_policy_flush(u8 type
, struct xfrm_audit
*audit_info
)
789 write_lock_bh(&xfrm_policy_lock
);
791 err
= xfrm_policy_flush_secctx_check(type
, audit_info
);
795 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
796 struct xfrm_policy
*pol
;
797 struct hlist_node
*entry
;
802 hlist_for_each_entry(pol
, entry
,
803 &init_net
.xfrm
.policy_inexact
[dir
], bydst
) {
804 if (pol
->type
!= type
)
806 hlist_del(&pol
->bydst
);
807 hlist_del(&pol
->byidx
);
808 write_unlock_bh(&xfrm_policy_lock
);
810 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
811 audit_info
->sessionid
,
814 xfrm_policy_kill(pol
);
817 write_lock_bh(&xfrm_policy_lock
);
821 for (i
= init_net
.xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
823 hlist_for_each_entry(pol
, entry
,
824 init_net
.xfrm
.policy_bydst
[dir
].table
+ i
,
826 if (pol
->type
!= type
)
828 hlist_del(&pol
->bydst
);
829 hlist_del(&pol
->byidx
);
830 list_del(&pol
->walk
.all
);
831 write_unlock_bh(&xfrm_policy_lock
);
833 xfrm_audit_policy_delete(pol
, 1,
834 audit_info
->loginuid
,
835 audit_info
->sessionid
,
837 xfrm_policy_kill(pol
);
840 write_lock_bh(&xfrm_policy_lock
);
845 init_net
.xfrm
.policy_count
[dir
] -= killed
;
847 atomic_inc(&flow_cache_genid
);
849 write_unlock_bh(&xfrm_policy_lock
);
852 EXPORT_SYMBOL(xfrm_policy_flush
);
854 int xfrm_policy_walk(struct xfrm_policy_walk
*walk
,
855 int (*func
)(struct xfrm_policy
*, int, int, void*),
858 struct xfrm_policy
*pol
;
859 struct xfrm_policy_walk_entry
*x
;
862 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
863 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
866 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
869 write_lock_bh(&xfrm_policy_lock
);
870 if (list_empty(&walk
->walk
.all
))
871 x
= list_first_entry(&init_net
.xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
873 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
874 list_for_each_entry_from(x
, &init_net
.xfrm
.policy_all
, all
) {
877 pol
= container_of(x
, struct xfrm_policy
, walk
);
878 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
879 walk
->type
!= pol
->type
)
881 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
884 list_move_tail(&walk
->walk
.all
, &x
->all
);
889 if (walk
->seq
== 0) {
893 list_del_init(&walk
->walk
.all
);
895 write_unlock_bh(&xfrm_policy_lock
);
898 EXPORT_SYMBOL(xfrm_policy_walk
);
900 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
902 INIT_LIST_HEAD(&walk
->walk
.all
);
907 EXPORT_SYMBOL(xfrm_policy_walk_init
);
909 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
911 if (list_empty(&walk
->walk
.all
))
914 write_lock_bh(&xfrm_policy_lock
);
915 list_del(&walk
->walk
.all
);
916 write_unlock_bh(&xfrm_policy_lock
);
918 EXPORT_SYMBOL(xfrm_policy_walk_done
);
921 * Find policy to apply to this flow.
923 * Returns 0 if policy found, else an -errno.
925 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
926 u8 type
, u16 family
, int dir
)
928 struct xfrm_selector
*sel
= &pol
->selector
;
929 int match
, ret
= -ESRCH
;
931 if (pol
->family
!= family
||
935 match
= xfrm_selector_match(sel
, fl
, family
);
937 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
943 static struct xfrm_policy
*xfrm_policy_lookup_bytype(u8 type
, struct flowi
*fl
,
947 struct xfrm_policy
*pol
, *ret
;
948 xfrm_address_t
*daddr
, *saddr
;
949 struct hlist_node
*entry
;
950 struct hlist_head
*chain
;
953 daddr
= xfrm_flowi_daddr(fl
, family
);
954 saddr
= xfrm_flowi_saddr(fl
, family
);
955 if (unlikely(!daddr
|| !saddr
))
958 read_lock_bh(&xfrm_policy_lock
);
959 chain
= policy_hash_direct(daddr
, saddr
, family
, dir
);
961 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
962 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
972 priority
= ret
->priority
;
976 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
977 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
978 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
986 } else if (pol
->priority
< priority
) {
994 read_unlock_bh(&xfrm_policy_lock
);
999 static int xfrm_policy_lookup(struct flowi
*fl
, u16 family
, u8 dir
,
1000 void **objp
, atomic_t
**obj_refp
)
1002 struct xfrm_policy
*pol
;
1005 #ifdef CONFIG_XFRM_SUB_POLICY
1006 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1014 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1019 #ifdef CONFIG_XFRM_SUB_POLICY
1022 if ((*objp
= (void *) pol
) != NULL
)
1023 *obj_refp
= &pol
->refcnt
;
1027 static inline int policy_to_flow_dir(int dir
)
1029 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1030 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1031 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1035 case XFRM_POLICY_IN
:
1037 case XFRM_POLICY_OUT
:
1038 return FLOW_DIR_OUT
;
1039 case XFRM_POLICY_FWD
:
1040 return FLOW_DIR_FWD
;
1044 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1046 struct xfrm_policy
*pol
;
1048 read_lock_bh(&xfrm_policy_lock
);
1049 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1050 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1055 err
= security_xfrm_policy_lookup(pol
->security
,
1057 policy_to_flow_dir(dir
));
1060 else if (err
== -ESRCH
)
1067 read_unlock_bh(&xfrm_policy_lock
);
1071 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1073 struct hlist_head
*chain
= policy_hash_bysel(&pol
->selector
,
1076 list_add(&pol
->walk
.all
, &init_net
.xfrm
.policy_all
);
1077 hlist_add_head(&pol
->bydst
, chain
);
1078 hlist_add_head(&pol
->byidx
, init_net
.xfrm
.policy_byidx
+idx_hash(pol
->index
));
1079 init_net
.xfrm
.policy_count
[dir
]++;
1082 if (xfrm_bydst_should_resize(dir
, NULL
))
1083 schedule_work(&xfrm_hash_work
);
1086 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1089 if (hlist_unhashed(&pol
->bydst
))
1092 hlist_del(&pol
->bydst
);
1093 hlist_del(&pol
->byidx
);
1094 list_del(&pol
->walk
.all
);
1095 init_net
.xfrm
.policy_count
[dir
]--;
1100 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1102 write_lock_bh(&xfrm_policy_lock
);
1103 pol
= __xfrm_policy_unlink(pol
, dir
);
1104 write_unlock_bh(&xfrm_policy_lock
);
1106 if (dir
< XFRM_POLICY_MAX
)
1107 atomic_inc(&flow_cache_genid
);
1108 xfrm_policy_kill(pol
);
1113 EXPORT_SYMBOL(xfrm_policy_delete
);
1115 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1117 struct xfrm_policy
*old_pol
;
1119 #ifdef CONFIG_XFRM_SUB_POLICY
1120 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1124 write_lock_bh(&xfrm_policy_lock
);
1125 old_pol
= sk
->sk_policy
[dir
];
1126 sk
->sk_policy
[dir
] = pol
;
1128 pol
->curlft
.add_time
= get_seconds();
1129 pol
->index
= xfrm_gen_index(XFRM_POLICY_MAX
+dir
);
1130 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1133 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1134 write_unlock_bh(&xfrm_policy_lock
);
1137 xfrm_policy_kill(old_pol
);
1142 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1144 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1147 newp
->selector
= old
->selector
;
1148 if (security_xfrm_policy_clone(old
->security
,
1151 return NULL
; /* ENOMEM */
1153 newp
->lft
= old
->lft
;
1154 newp
->curlft
= old
->curlft
;
1155 newp
->action
= old
->action
;
1156 newp
->flags
= old
->flags
;
1157 newp
->xfrm_nr
= old
->xfrm_nr
;
1158 newp
->index
= old
->index
;
1159 newp
->type
= old
->type
;
1160 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1161 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1162 write_lock_bh(&xfrm_policy_lock
);
1163 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1164 write_unlock_bh(&xfrm_policy_lock
);
1170 int __xfrm_sk_clone_policy(struct sock
*sk
)
1172 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1173 *p1
= sk
->sk_policy
[1];
1175 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1176 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1178 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1184 xfrm_get_saddr(xfrm_address_t
*local
, xfrm_address_t
*remote
,
1185 unsigned short family
)
1188 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1190 if (unlikely(afinfo
== NULL
))
1192 err
= afinfo
->get_saddr(local
, remote
);
1193 xfrm_policy_put_afinfo(afinfo
);
1197 /* Resolve list of templates for the flow, given policy. */
1200 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1201 struct xfrm_state
**xfrm
,
1202 unsigned short family
)
1206 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1207 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1210 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1211 struct xfrm_state
*x
;
1212 xfrm_address_t
*remote
= daddr
;
1213 xfrm_address_t
*local
= saddr
;
1214 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1216 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1217 tmpl
->mode
== XFRM_MODE_BEET
) {
1218 remote
= &tmpl
->id
.daddr
;
1219 local
= &tmpl
->saddr
;
1220 family
= tmpl
->encap_family
;
1221 if (xfrm_addr_any(local
, family
)) {
1222 error
= xfrm_get_saddr(&tmp
, remote
, family
);
1229 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1231 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1238 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1242 else if (error
== -ESRCH
)
1245 if (!tmpl
->optional
)
1251 for (nx
--; nx
>=0; nx
--)
1252 xfrm_state_put(xfrm
[nx
]);
1257 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1258 struct xfrm_state
**xfrm
,
1259 unsigned short family
)
1261 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1262 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1268 for (i
= 0; i
< npols
; i
++) {
1269 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1274 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1282 /* found states are sorted for outbound processing */
1284 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1289 for (cnx
--; cnx
>=0; cnx
--)
1290 xfrm_state_put(tpp
[cnx
]);
1295 /* Check that the bundle accepts the flow and its components are
1299 static struct dst_entry
*
1300 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1302 struct dst_entry
*x
;
1303 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1304 if (unlikely(afinfo
== NULL
))
1305 return ERR_PTR(-EINVAL
);
1306 x
= afinfo
->find_bundle(fl
, policy
);
1307 xfrm_policy_put_afinfo(afinfo
);
1311 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1313 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1319 tos
= afinfo
->get_tos(fl
);
1321 xfrm_policy_put_afinfo(afinfo
);
1326 static inline struct xfrm_dst
*xfrm_alloc_dst(int family
)
1328 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1329 struct xfrm_dst
*xdst
;
1332 return ERR_PTR(-EINVAL
);
1334 xdst
= dst_alloc(afinfo
->dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1336 xfrm_policy_put_afinfo(afinfo
);
1341 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1344 struct xfrm_policy_afinfo
*afinfo
=
1345 xfrm_policy_get_afinfo(dst
->ops
->family
);
1351 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1353 xfrm_policy_put_afinfo(afinfo
);
1358 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1360 struct xfrm_policy_afinfo
*afinfo
=
1361 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1367 err
= afinfo
->fill_dst(xdst
, dev
);
1369 xfrm_policy_put_afinfo(afinfo
);
1374 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1375 * all the metrics... Shortly, bundle a bundle.
1378 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1379 struct xfrm_state
**xfrm
, int nx
,
1381 struct dst_entry
*dst
)
1383 unsigned long now
= jiffies
;
1384 struct net_device
*dev
;
1385 struct dst_entry
*dst_prev
= NULL
;
1386 struct dst_entry
*dst0
= NULL
;
1390 int nfheader_len
= 0;
1391 int trailer_len
= 0;
1393 int family
= policy
->selector
.family
;
1394 xfrm_address_t saddr
, daddr
;
1396 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1398 tos
= xfrm_get_tos(fl
, family
);
1405 for (; i
< nx
; i
++) {
1406 struct xfrm_dst
*xdst
= xfrm_alloc_dst(family
);
1407 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1409 err
= PTR_ERR(xdst
);
1418 dst_prev
->child
= dst_clone(dst1
);
1419 dst1
->flags
|= DST_NOHASH
;
1423 memcpy(&dst1
->metrics
, &dst
->metrics
, sizeof(dst
->metrics
));
1425 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1426 family
= xfrm
[i
]->props
.family
;
1427 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1435 dst1
->xfrm
= xfrm
[i
];
1436 xdst
->genid
= xfrm
[i
]->genid
;
1438 dst1
->obsolete
= -1;
1439 dst1
->flags
|= DST_HOST
;
1440 dst1
->lastuse
= now
;
1442 dst1
->input
= dst_discard
;
1443 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1445 dst1
->next
= dst_prev
;
1448 header_len
+= xfrm
[i
]->props
.header_len
;
1449 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1450 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1451 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1454 dst_prev
->child
= dst
;
1462 /* Copy neighbout for reachability confirmation */
1463 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1465 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1466 xfrm_init_pmtu(dst_prev
);
1468 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1469 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1471 err
= xfrm_fill_dst(xdst
, dev
);
1475 dst_prev
->header_len
= header_len
;
1476 dst_prev
->trailer_len
= trailer_len
;
1477 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1478 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1486 xfrm_state_put(xfrm
[i
]);
1490 dst0
= ERR_PTR(err
);
1495 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1498 *target
= kmalloc(size
, GFP_ATOMIC
);
1502 memcpy(*target
, src
, size
);
1507 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1509 #ifdef CONFIG_XFRM_SUB_POLICY
1510 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1511 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1519 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1521 #ifdef CONFIG_XFRM_SUB_POLICY
1522 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1523 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1529 static int stale_bundle(struct dst_entry
*dst
);
1531 /* Main function: finds/creates a bundle for given flow.
1533 * At the moment we eat a raw IP route. Mostly to speed up lookups
1534 * on interfaces with disabled IPsec.
1536 int __xfrm_lookup(struct dst_entry
**dst_p
, struct flowi
*fl
,
1537 struct sock
*sk
, int flags
)
1539 struct xfrm_policy
*policy
;
1540 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1545 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1546 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1551 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1554 genid
= atomic_read(&flow_cache_genid
);
1556 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1562 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1563 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1564 err
= PTR_ERR(policy
);
1565 if (IS_ERR(policy
)) {
1566 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1572 /* To accelerate a bit... */
1573 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1574 !init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
])
1577 policy
= flow_cache_lookup(fl
, dst_orig
->ops
->family
,
1578 dir
, xfrm_policy_lookup
);
1579 err
= PTR_ERR(policy
);
1580 if (IS_ERR(policy
)) {
1581 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1589 family
= dst_orig
->ops
->family
;
1592 xfrm_nr
+= pols
[0]->xfrm_nr
;
1595 if ((flags
& XFRM_LOOKUP_ICMP
) && !(policy
->flags
& XFRM_POLICY_ICMP
))
1598 policy
->curlft
.use_time
= get_seconds();
1600 switch (policy
->action
) {
1602 case XFRM_POLICY_BLOCK
:
1603 /* Prohibit the flow */
1604 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1608 case XFRM_POLICY_ALLOW
:
1609 #ifndef CONFIG_XFRM_SUB_POLICY
1610 if (policy
->xfrm_nr
== 0) {
1611 /* Flow passes not transformed. */
1612 xfrm_pol_put(policy
);
1617 /* Try to find matching bundle.
1619 * LATER: help from flow cache. It is optional, this
1620 * is required only for output policy.
1622 dst
= xfrm_find_bundle(fl
, policy
, family
);
1624 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1632 #ifdef CONFIG_XFRM_SUB_POLICY
1633 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1634 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1638 if (IS_ERR(pols
[1])) {
1639 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1640 err
= PTR_ERR(pols
[1]);
1643 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1644 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1649 xfrm_nr
+= pols
[1]->xfrm_nr
;
1654 * Because neither flowi nor bundle information knows about
1655 * transformation template size. On more than one policy usage
1656 * we can realize whether all of them is bypass or not after
1657 * they are searched. See above not-transformed bypass
1658 * is surrounded by non-sub policy configuration, too.
1661 /* Flow passes not transformed. */
1662 xfrm_pols_put(pols
, npols
);
1667 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1669 if (unlikely(nx
<0)) {
1671 if (err
== -EAGAIN
&& sysctl_xfrm_larval_drop
) {
1672 /* EREMOTE tells the caller to generate
1673 * a one-shot blackhole route.
1675 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1676 xfrm_pol_put(policy
);
1679 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1680 DECLARE_WAITQUEUE(wait
, current
);
1682 add_wait_queue(&init_net
.xfrm
.km_waitq
, &wait
);
1683 set_current_state(TASK_INTERRUPTIBLE
);
1685 set_current_state(TASK_RUNNING
);
1686 remove_wait_queue(&init_net
.xfrm
.km_waitq
, &wait
);
1688 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1690 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1691 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1695 if (nx
== -EAGAIN
||
1696 genid
!= atomic_read(&flow_cache_genid
)) {
1697 xfrm_pols_put(pols
, npols
);
1703 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1708 /* Flow passes not transformed. */
1709 xfrm_pols_put(pols
, npols
);
1713 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1716 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1720 for (pi
= 0; pi
< npols
; pi
++) {
1721 read_lock_bh(&pols
[pi
]->lock
);
1722 pol_dead
|= pols
[pi
]->walk
.dead
;
1723 read_unlock_bh(&pols
[pi
]->lock
);
1726 write_lock_bh(&policy
->lock
);
1727 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1728 /* Wow! While we worked on resolving, this
1729 * policy has gone. Retry. It is not paranoia,
1730 * we just cannot enlist new bundle to dead object.
1731 * We can't enlist stable bundles either.
1733 write_unlock_bh(&policy
->lock
);
1737 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD
);
1739 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1740 err
= -EHOSTUNREACH
;
1745 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1747 err
= xfrm_dst_update_origin(dst
, fl
);
1748 if (unlikely(err
)) {
1749 write_unlock_bh(&policy
->lock
);
1751 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1755 dst
->next
= policy
->bundles
;
1756 policy
->bundles
= dst
;
1758 write_unlock_bh(&policy
->lock
);
1761 dst_release(dst_orig
);
1762 xfrm_pols_put(pols
, npols
);
1766 xfrm_pols_put(pols
, npols
);
1768 dst_release(dst_orig
);
1774 if (flags
& XFRM_LOOKUP_ICMP
)
1778 EXPORT_SYMBOL(__xfrm_lookup
);
1780 int xfrm_lookup(struct dst_entry
**dst_p
, struct flowi
*fl
,
1781 struct sock
*sk
, int flags
)
1783 int err
= __xfrm_lookup(dst_p
, fl
, sk
, flags
);
1785 if (err
== -EREMOTE
) {
1786 dst_release(*dst_p
);
1793 EXPORT_SYMBOL(xfrm_lookup
);
1796 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1798 struct xfrm_state
*x
;
1800 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1802 x
= skb
->sp
->xvec
[idx
];
1803 if (!x
->type
->reject
)
1805 return x
->type
->reject(x
, skb
, fl
);
1808 /* When skb is transformed back to its "native" form, we have to
1809 * check policy restrictions. At the moment we make this in maximally
1810 * stupid way. Shame on me. :-) Of course, connected sockets must
1811 * have policy cached at them.
1815 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1816 unsigned short family
)
1818 if (xfrm_state_kern(x
))
1819 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1820 return x
->id
.proto
== tmpl
->id
.proto
&&
1821 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1822 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1823 x
->props
.mode
== tmpl
->mode
&&
1824 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1825 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1826 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1827 xfrm_state_addr_cmp(tmpl
, x
, family
));
1831 * 0 or more than 0 is returned when validation is succeeded (either bypass
1832 * because of optional transport mode, or next index of the mathced secpath
1833 * state with the template.
1834 * -1 is returned when no matching template is found.
1835 * Otherwise "-2 - errored_index" is returned.
1838 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1839 unsigned short family
)
1843 if (tmpl
->optional
) {
1844 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1848 for (; idx
< sp
->len
; idx
++) {
1849 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1851 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1860 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1861 unsigned int family
, int reverse
)
1863 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1866 if (unlikely(afinfo
== NULL
))
1867 return -EAFNOSUPPORT
;
1869 afinfo
->decode_session(skb
, fl
, reverse
);
1870 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1871 xfrm_policy_put_afinfo(afinfo
);
1874 EXPORT_SYMBOL(__xfrm_decode_session
);
1876 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1878 for (; k
< sp
->len
; k
++) {
1879 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1888 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1889 unsigned short family
)
1891 struct xfrm_policy
*pol
;
1892 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1901 reverse
= dir
& ~XFRM_POLICY_MASK
;
1902 dir
&= XFRM_POLICY_MASK
;
1903 fl_dir
= policy_to_flow_dir(dir
);
1905 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1906 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
1910 nf_nat_decode_session(skb
, &fl
, family
);
1912 /* First, check used SA against their selectors. */
1916 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1917 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1918 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1919 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH
);
1926 if (sk
&& sk
->sk_policy
[dir
]) {
1927 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1929 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1935 pol
= flow_cache_lookup(&fl
, family
, fl_dir
,
1936 xfrm_policy_lookup
);
1939 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1944 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1945 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1946 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS
);
1952 pol
->curlft
.use_time
= get_seconds();
1956 #ifdef CONFIG_XFRM_SUB_POLICY
1957 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1958 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1962 if (IS_ERR(pols
[1])) {
1963 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1966 pols
[1]->curlft
.use_time
= get_seconds();
1972 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1973 struct sec_path
*sp
;
1974 static struct sec_path dummy
;
1975 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1976 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1977 struct xfrm_tmpl
**tpp
= tp
;
1981 if ((sp
= skb
->sp
) == NULL
)
1984 for (pi
= 0; pi
< npols
; pi
++) {
1985 if (pols
[pi
] != pol
&&
1986 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
1987 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
1990 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1991 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR
);
1994 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
1995 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
1999 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2003 /* For each tunnel xfrm, find the first matching tmpl.
2004 * For each tmpl before that, find corresponding xfrm.
2005 * Order is _important_. Later we will implement
2006 * some barriers, but at the moment barriers
2007 * are implied between each two transformations.
2009 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2010 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2013 /* "-2 - errored_index" returned */
2015 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2020 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2021 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2025 xfrm_pols_put(pols
, npols
);
2028 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
2031 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2033 xfrm_pols_put(pols
, npols
);
2036 EXPORT_SYMBOL(__xfrm_policy_check
);
2038 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2042 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2043 /* XXX: we should have something like FWDHDRERROR here. */
2044 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
2048 return xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) == 0;
2050 EXPORT_SYMBOL(__xfrm_route_forward
);
2052 /* Optimize later using cookies and generation ids. */
2054 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2056 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2057 * to "-1" to force all XFRM destinations to get validated by
2058 * dst_ops->check on every use. We do this because when a
2059 * normal route referenced by an XFRM dst is obsoleted we do
2060 * not go looking around for all parent referencing XFRM dsts
2061 * so that we can invalidate them. It is just too much work.
2062 * Instead we make the checks here on every use. For example:
2064 * XFRM dst A --> IPv4 dst X
2066 * X is the "xdst->route" of A (X is also the "dst->path" of A
2067 * in this example). If X is marked obsolete, "A" will not
2068 * notice. That's what we are validating here via the
2069 * stale_bundle() check.
2071 * When a policy's bundle is pruned, we dst_free() the XFRM
2072 * dst which causes it's ->obsolete field to be set to a
2073 * positive non-zero integer. If an XFRM dst has been pruned
2074 * like this, we want to force a new route lookup.
2076 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2082 static int stale_bundle(struct dst_entry
*dst
)
2084 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2087 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2089 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2090 dst
->dev
= dev_net(dev
)->loopback_dev
;
2095 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2097 static void xfrm_link_failure(struct sk_buff
*skb
)
2099 /* Impossible. Such dst must be popped before reaches point of failure. */
2103 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2106 if (dst
->obsolete
) {
2114 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
2116 struct dst_entry
*dst
, **dstp
;
2118 write_lock(&pol
->lock
);
2119 dstp
= &pol
->bundles
;
2120 while ((dst
=*dstp
) != NULL
) {
2123 dst
->next
= *gc_list_p
;
2129 write_unlock(&pol
->lock
);
2132 static void xfrm_prune_bundles(int (*func
)(struct dst_entry
*))
2134 struct dst_entry
*gc_list
= NULL
;
2137 read_lock_bh(&xfrm_policy_lock
);
2138 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2139 struct xfrm_policy
*pol
;
2140 struct hlist_node
*entry
;
2141 struct hlist_head
*table
;
2144 hlist_for_each_entry(pol
, entry
,
2145 &init_net
.xfrm
.policy_inexact
[dir
], bydst
)
2146 prune_one_bundle(pol
, func
, &gc_list
);
2148 table
= init_net
.xfrm
.policy_bydst
[dir
].table
;
2149 for (i
= init_net
.xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
2150 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
2151 prune_one_bundle(pol
, func
, &gc_list
);
2154 read_unlock_bh(&xfrm_policy_lock
);
2157 struct dst_entry
*dst
= gc_list
;
2158 gc_list
= dst
->next
;
2163 static int unused_bundle(struct dst_entry
*dst
)
2165 return !atomic_read(&dst
->__refcnt
);
2168 static void __xfrm_garbage_collect(void)
2170 xfrm_prune_bundles(unused_bundle
);
2173 static int xfrm_flush_bundles(void)
2175 xfrm_prune_bundles(stale_bundle
);
2179 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2182 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2183 u32 pmtu
, route_mtu_cached
;
2185 pmtu
= dst_mtu(dst
->child
);
2186 xdst
->child_mtu_cached
= pmtu
;
2188 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2190 route_mtu_cached
= dst_mtu(xdst
->route
);
2191 xdst
->route_mtu_cached
= route_mtu_cached
;
2193 if (pmtu
> route_mtu_cached
)
2194 pmtu
= route_mtu_cached
;
2196 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
2197 } while ((dst
= dst
->next
));
2200 /* Check that the bundle accepts the flow and its components are
2204 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2205 struct flowi
*fl
, int family
, int strict
)
2207 struct dst_entry
*dst
= &first
->u
.dst
;
2208 struct xfrm_dst
*last
;
2211 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2212 (dst
->dev
&& !netif_running(dst
->dev
)))
2214 #ifdef CONFIG_XFRM_SUB_POLICY
2216 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2218 if (first
->partner
&&
2219 !xfrm_selector_match(first
->partner
, fl
, family
))
2227 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2229 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2232 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2234 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2236 if (xdst
->genid
!= dst
->xfrm
->genid
)
2240 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2241 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2244 mtu
= dst_mtu(dst
->child
);
2245 if (xdst
->child_mtu_cached
!= mtu
) {
2247 xdst
->child_mtu_cached
= mtu
;
2250 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2252 mtu
= dst_mtu(xdst
->route
);
2253 if (xdst
->route_mtu_cached
!= mtu
) {
2255 xdst
->route_mtu_cached
= mtu
;
2259 } while (dst
->xfrm
);
2264 mtu
= last
->child_mtu_cached
;
2268 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2269 if (mtu
> last
->route_mtu_cached
)
2270 mtu
= last
->route_mtu_cached
;
2271 dst
->metrics
[RTAX_MTU
-1] = mtu
;
2276 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2277 last
->child_mtu_cached
= mtu
;
2283 EXPORT_SYMBOL(xfrm_bundle_ok
);
2285 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2288 if (unlikely(afinfo
== NULL
))
2290 if (unlikely(afinfo
->family
>= NPROTO
))
2291 return -EAFNOSUPPORT
;
2292 write_lock_bh(&xfrm_policy_afinfo_lock
);
2293 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2296 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2297 if (likely(dst_ops
->kmem_cachep
== NULL
))
2298 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2299 if (likely(dst_ops
->check
== NULL
))
2300 dst_ops
->check
= xfrm_dst_check
;
2301 if (likely(dst_ops
->negative_advice
== NULL
))
2302 dst_ops
->negative_advice
= xfrm_negative_advice
;
2303 if (likely(dst_ops
->link_failure
== NULL
))
2304 dst_ops
->link_failure
= xfrm_link_failure
;
2305 if (likely(afinfo
->garbage_collect
== NULL
))
2306 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2307 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2309 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2312 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2314 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2317 if (unlikely(afinfo
== NULL
))
2319 if (unlikely(afinfo
->family
>= NPROTO
))
2320 return -EAFNOSUPPORT
;
2321 write_lock_bh(&xfrm_policy_afinfo_lock
);
2322 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2323 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2326 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2327 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2328 dst_ops
->kmem_cachep
= NULL
;
2329 dst_ops
->check
= NULL
;
2330 dst_ops
->negative_advice
= NULL
;
2331 dst_ops
->link_failure
= NULL
;
2332 afinfo
->garbage_collect
= NULL
;
2335 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2338 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2340 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2342 struct xfrm_policy_afinfo
*afinfo
;
2343 if (unlikely(family
>= NPROTO
))
2345 read_lock(&xfrm_policy_afinfo_lock
);
2346 afinfo
= xfrm_policy_afinfo
[family
];
2347 if (unlikely(!afinfo
))
2348 read_unlock(&xfrm_policy_afinfo_lock
);
2352 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2354 read_unlock(&xfrm_policy_afinfo_lock
);
2357 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2359 struct net_device
*dev
= ptr
;
2361 if (!net_eq(dev_net(dev
), &init_net
))
2366 xfrm_flush_bundles();
2371 static struct notifier_block xfrm_dev_notifier
= {
2372 .notifier_call
= xfrm_dev_event
,
2375 #ifdef CONFIG_XFRM_STATISTICS
2376 static int __init
xfrm_statistics_init(void)
2378 if (snmp_mib_init((void **)xfrm_statistics
,
2379 sizeof(struct linux_xfrm_mib
)) < 0)
2385 static int __net_init
xfrm_policy_init(struct net
*net
)
2387 unsigned int hmask
, sz
;
2390 if (net_eq(net
, &init_net
))
2391 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2392 sizeof(struct xfrm_dst
),
2393 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2397 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2399 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2400 if (!net
->xfrm
.policy_byidx
)
2402 net
->xfrm
.policy_idx_hmask
= hmask
;
2404 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2405 struct xfrm_policy_hash
*htab
;
2407 net
->xfrm
.policy_count
[dir
] = 0;
2408 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2410 htab
= &net
->xfrm
.policy_bydst
[dir
];
2411 htab
->table
= xfrm_hash_alloc(sz
);
2414 htab
->hmask
= hmask
;
2417 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2418 if (net_eq(net
, &init_net
))
2419 register_netdevice_notifier(&xfrm_dev_notifier
);
2423 for (dir
--; dir
>= 0; dir
--) {
2424 struct xfrm_policy_hash
*htab
;
2426 htab
= &net
->xfrm
.policy_bydst
[dir
];
2427 xfrm_hash_free(htab
->table
, sz
);
2429 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2434 static void xfrm_policy_fini(struct net
*net
)
2439 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2441 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2442 struct xfrm_policy_hash
*htab
;
2444 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2446 htab
= &net
->xfrm
.policy_bydst
[dir
];
2447 sz
= (htab
->hmask
+ 1);
2448 WARN_ON(!hlist_empty(htab
->table
));
2449 xfrm_hash_free(htab
->table
, sz
);
2452 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2453 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2454 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2457 static int __net_init
xfrm_net_init(struct net
*net
)
2461 rv
= xfrm_state_init(net
);
2464 rv
= xfrm_policy_init(net
);
2470 xfrm_state_fini(net
);
2475 static void __net_exit
xfrm_net_exit(struct net
*net
)
2477 xfrm_policy_fini(net
);
2478 xfrm_state_fini(net
);
2481 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2482 .init
= xfrm_net_init
,
2483 .exit
= xfrm_net_exit
,
2486 void __init
xfrm_init(void)
2488 register_pernet_subsys(&xfrm_net_ops
);
2489 #ifdef CONFIG_XFRM_STATISTICS
2490 xfrm_statistics_init();
2493 #ifdef CONFIG_XFRM_STATISTICS
2498 #ifdef CONFIG_AUDITSYSCALL
2499 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2500 struct audit_buffer
*audit_buf
)
2502 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2503 struct xfrm_selector
*sel
= &xp
->selector
;
2506 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2507 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2509 switch(sel
->family
) {
2511 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2512 if (sel
->prefixlen_s
!= 32)
2513 audit_log_format(audit_buf
, " src_prefixlen=%d",
2515 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2516 if (sel
->prefixlen_d
!= 32)
2517 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2521 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2522 if (sel
->prefixlen_s
!= 128)
2523 audit_log_format(audit_buf
, " src_prefixlen=%d",
2525 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2526 if (sel
->prefixlen_d
!= 128)
2527 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2533 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2534 uid_t auid
, u32 sessionid
, u32 secid
)
2536 struct audit_buffer
*audit_buf
;
2538 audit_buf
= xfrm_audit_start("SPD-add");
2539 if (audit_buf
== NULL
)
2541 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2542 audit_log_format(audit_buf
, " res=%u", result
);
2543 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2544 audit_log_end(audit_buf
);
2546 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2548 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2549 uid_t auid
, u32 sessionid
, u32 secid
)
2551 struct audit_buffer
*audit_buf
;
2553 audit_buf
= xfrm_audit_start("SPD-delete");
2554 if (audit_buf
== NULL
)
2556 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2557 audit_log_format(audit_buf
, " res=%u", result
);
2558 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2559 audit_log_end(audit_buf
);
2561 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2564 #ifdef CONFIG_XFRM_MIGRATE
2565 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2566 struct xfrm_selector
*sel_tgt
)
2568 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2569 if (sel_tgt
->family
== sel_cmp
->family
&&
2570 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2571 sel_cmp
->family
) == 0 &&
2572 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2573 sel_cmp
->family
) == 0 &&
2574 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2575 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2579 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2586 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2589 struct xfrm_policy
*pol
, *ret
= NULL
;
2590 struct hlist_node
*entry
;
2591 struct hlist_head
*chain
;
2594 read_lock_bh(&xfrm_policy_lock
);
2595 chain
= policy_hash_direct(&sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2596 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2597 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2598 pol
->type
== type
) {
2600 priority
= ret
->priority
;
2604 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2605 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2606 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2607 pol
->type
== type
&&
2608 pol
->priority
< priority
) {
2617 read_unlock_bh(&xfrm_policy_lock
);
2622 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2626 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2627 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2629 case XFRM_MODE_TUNNEL
:
2630 case XFRM_MODE_BEET
:
2631 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2632 m
->old_family
) == 0 &&
2633 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2634 m
->old_family
) == 0) {
2638 case XFRM_MODE_TRANSPORT
:
2639 /* in case of transport mode, template does not store
2640 any IP addresses, hence we just compare mode and
2651 /* update endpoint address(es) of template(s) */
2652 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2653 struct xfrm_migrate
*m
, int num_migrate
)
2655 struct xfrm_migrate
*mp
;
2656 struct dst_entry
*dst
;
2659 write_lock_bh(&pol
->lock
);
2660 if (unlikely(pol
->walk
.dead
)) {
2661 /* target policy has been deleted */
2662 write_unlock_bh(&pol
->lock
);
2666 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2667 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2668 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2671 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2672 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2674 /* update endpoints */
2675 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2676 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2677 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2678 sizeof(pol
->xfrm_vec
[i
].saddr
));
2679 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2681 while ((dst
= pol
->bundles
) != NULL
) {
2682 pol
->bundles
= dst
->next
;
2688 write_unlock_bh(&pol
->lock
);
2696 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2700 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2703 for (i
= 0; i
< num_migrate
; i
++) {
2704 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2705 m
[i
].old_family
) == 0) &&
2706 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2707 m
[i
].old_family
) == 0))
2709 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2710 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2713 /* check if there is any duplicated entry */
2714 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2715 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2716 sizeof(m
[i
].old_daddr
)) &&
2717 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2718 sizeof(m
[i
].old_saddr
)) &&
2719 m
[i
].proto
== m
[j
].proto
&&
2720 m
[i
].mode
== m
[j
].mode
&&
2721 m
[i
].reqid
== m
[j
].reqid
&&
2722 m
[i
].old_family
== m
[j
].old_family
)
2730 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2731 struct xfrm_migrate
*m
, int num_migrate
,
2732 struct xfrm_kmaddress
*k
)
2734 int i
, err
, nx_cur
= 0, nx_new
= 0;
2735 struct xfrm_policy
*pol
= NULL
;
2736 struct xfrm_state
*x
, *xc
;
2737 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2738 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2739 struct xfrm_migrate
*mp
;
2741 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2744 /* Stage 1 - find policy */
2745 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2750 /* Stage 2 - find and update state(s) */
2751 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2752 if ((x
= xfrm_migrate_state_find(mp
))) {
2755 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2765 /* Stage 3 - update policy */
2766 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2769 /* Stage 4 - delete old state(s) */
2771 xfrm_states_put(x_cur
, nx_cur
);
2772 xfrm_states_delete(x_cur
, nx_cur
);
2775 /* Stage 5 - announce */
2776 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2788 xfrm_states_put(x_cur
, nx_cur
);
2790 xfrm_states_delete(x_new
, nx_new
);
2794 EXPORT_SYMBOL(xfrm_migrate
);