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/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/list.h>
19 #include <linux/spinlock.h>
20 #include <linux/workqueue.h>
21 #include <linux/notifier.h>
22 #include <linux/netdevice.h>
23 #include <linux/netfilter.h>
24 #include <linux/module.h>
25 #include <linux/cache.h>
28 #include <linux/audit.h>
30 #include "xfrm_hash.h"
32 DEFINE_MUTEX(xfrm_cfg_mutex
);
33 EXPORT_SYMBOL(xfrm_cfg_mutex
);
35 static DEFINE_RWLOCK(xfrm_policy_lock
);
37 unsigned int xfrm_policy_count
[XFRM_POLICY_MAX
*2];
38 EXPORT_SYMBOL(xfrm_policy_count
);
40 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
41 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
43 static kmem_cache_t
*xfrm_dst_cache __read_mostly
;
45 static struct work_struct xfrm_policy_gc_work
;
46 static HLIST_HEAD(xfrm_policy_gc_list
);
47 static DEFINE_SPINLOCK(xfrm_policy_gc_lock
);
49 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
50 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
51 static struct xfrm_policy_afinfo
*xfrm_policy_lock_afinfo(unsigned int family
);
52 static void xfrm_policy_unlock_afinfo(struct xfrm_policy_afinfo
*afinfo
);
55 __xfrm4_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
57 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
58 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
59 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
60 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
61 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
62 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
66 __xfrm6_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
68 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
69 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
70 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
71 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
72 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
73 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
76 int xfrm_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
,
77 unsigned short family
)
81 return __xfrm4_selector_match(sel
, fl
);
83 return __xfrm6_selector_match(sel
, fl
);
88 int xfrm_register_type(struct xfrm_type
*type
, unsigned short family
)
90 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_lock_afinfo(family
);
91 struct xfrm_type
**typemap
;
94 if (unlikely(afinfo
== NULL
))
96 typemap
= afinfo
->type_map
;
98 if (likely(typemap
[type
->proto
] == NULL
))
99 typemap
[type
->proto
] = type
;
102 xfrm_policy_unlock_afinfo(afinfo
);
105 EXPORT_SYMBOL(xfrm_register_type
);
107 int xfrm_unregister_type(struct xfrm_type
*type
, unsigned short family
)
109 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_lock_afinfo(family
);
110 struct xfrm_type
**typemap
;
113 if (unlikely(afinfo
== NULL
))
114 return -EAFNOSUPPORT
;
115 typemap
= afinfo
->type_map
;
117 if (unlikely(typemap
[type
->proto
] != type
))
120 typemap
[type
->proto
] = NULL
;
121 xfrm_policy_unlock_afinfo(afinfo
);
124 EXPORT_SYMBOL(xfrm_unregister_type
);
126 struct xfrm_type
*xfrm_get_type(u8 proto
, unsigned short family
)
128 struct xfrm_policy_afinfo
*afinfo
;
129 struct xfrm_type
**typemap
;
130 struct xfrm_type
*type
;
131 int modload_attempted
= 0;
134 afinfo
= xfrm_policy_get_afinfo(family
);
135 if (unlikely(afinfo
== NULL
))
137 typemap
= afinfo
->type_map
;
139 type
= typemap
[proto
];
140 if (unlikely(type
&& !try_module_get(type
->owner
)))
142 if (!type
&& !modload_attempted
) {
143 xfrm_policy_put_afinfo(afinfo
);
144 request_module("xfrm-type-%d-%d",
145 (int) family
, (int) proto
);
146 modload_attempted
= 1;
150 xfrm_policy_put_afinfo(afinfo
);
154 int xfrm_dst_lookup(struct xfrm_dst
**dst
, struct flowi
*fl
,
155 unsigned short family
)
157 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
160 if (unlikely(afinfo
== NULL
))
161 return -EAFNOSUPPORT
;
163 if (likely(afinfo
->dst_lookup
!= NULL
))
164 err
= afinfo
->dst_lookup(dst
, fl
);
167 xfrm_policy_put_afinfo(afinfo
);
170 EXPORT_SYMBOL(xfrm_dst_lookup
);
172 void xfrm_put_type(struct xfrm_type
*type
)
174 module_put(type
->owner
);
177 int xfrm_register_mode(struct xfrm_mode
*mode
, int family
)
179 struct xfrm_policy_afinfo
*afinfo
;
180 struct xfrm_mode
**modemap
;
183 if (unlikely(mode
->encap
>= XFRM_MODE_MAX
))
186 afinfo
= xfrm_policy_lock_afinfo(family
);
187 if (unlikely(afinfo
== NULL
))
188 return -EAFNOSUPPORT
;
191 modemap
= afinfo
->mode_map
;
192 if (likely(modemap
[mode
->encap
] == NULL
)) {
193 modemap
[mode
->encap
] = mode
;
197 xfrm_policy_unlock_afinfo(afinfo
);
200 EXPORT_SYMBOL(xfrm_register_mode
);
202 int xfrm_unregister_mode(struct xfrm_mode
*mode
, int family
)
204 struct xfrm_policy_afinfo
*afinfo
;
205 struct xfrm_mode
**modemap
;
208 if (unlikely(mode
->encap
>= XFRM_MODE_MAX
))
211 afinfo
= xfrm_policy_lock_afinfo(family
);
212 if (unlikely(afinfo
== NULL
))
213 return -EAFNOSUPPORT
;
216 modemap
= afinfo
->mode_map
;
217 if (likely(modemap
[mode
->encap
] == mode
)) {
218 modemap
[mode
->encap
] = NULL
;
222 xfrm_policy_unlock_afinfo(afinfo
);
225 EXPORT_SYMBOL(xfrm_unregister_mode
);
227 struct xfrm_mode
*xfrm_get_mode(unsigned int encap
, int family
)
229 struct xfrm_policy_afinfo
*afinfo
;
230 struct xfrm_mode
*mode
;
231 int modload_attempted
= 0;
233 if (unlikely(encap
>= XFRM_MODE_MAX
))
237 afinfo
= xfrm_policy_get_afinfo(family
);
238 if (unlikely(afinfo
== NULL
))
241 mode
= afinfo
->mode_map
[encap
];
242 if (unlikely(mode
&& !try_module_get(mode
->owner
)))
244 if (!mode
&& !modload_attempted
) {
245 xfrm_policy_put_afinfo(afinfo
);
246 request_module("xfrm-mode-%d-%d", family
, encap
);
247 modload_attempted
= 1;
251 xfrm_policy_put_afinfo(afinfo
);
255 void xfrm_put_mode(struct xfrm_mode
*mode
)
257 module_put(mode
->owner
);
260 static inline unsigned long make_jiffies(long secs
)
262 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
263 return MAX_SCHEDULE_TIMEOUT
-1;
268 static void xfrm_policy_timer(unsigned long data
)
270 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
271 unsigned long now
= (unsigned long)xtime
.tv_sec
;
272 long next
= LONG_MAX
;
276 read_lock(&xp
->lock
);
281 dir
= xfrm_policy_id2dir(xp
->index
);
283 if (xp
->lft
.hard_add_expires_seconds
) {
284 long tmo
= xp
->lft
.hard_add_expires_seconds
+
285 xp
->curlft
.add_time
- now
;
291 if (xp
->lft
.hard_use_expires_seconds
) {
292 long tmo
= xp
->lft
.hard_use_expires_seconds
+
293 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
299 if (xp
->lft
.soft_add_expires_seconds
) {
300 long tmo
= xp
->lft
.soft_add_expires_seconds
+
301 xp
->curlft
.add_time
- now
;
304 tmo
= XFRM_KM_TIMEOUT
;
309 if (xp
->lft
.soft_use_expires_seconds
) {
310 long tmo
= xp
->lft
.soft_use_expires_seconds
+
311 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
314 tmo
= XFRM_KM_TIMEOUT
;
321 km_policy_expired(xp
, dir
, 0, 0);
322 if (next
!= LONG_MAX
&&
323 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
327 read_unlock(&xp
->lock
);
332 read_unlock(&xp
->lock
);
333 if (!xfrm_policy_delete(xp
, dir
))
334 km_policy_expired(xp
, dir
, 1, 0);
339 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
343 struct xfrm_policy
*xfrm_policy_alloc(gfp_t gfp
)
345 struct xfrm_policy
*policy
;
347 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
350 INIT_HLIST_NODE(&policy
->bydst
);
351 INIT_HLIST_NODE(&policy
->byidx
);
352 rwlock_init(&policy
->lock
);
353 atomic_set(&policy
->refcnt
, 1);
354 init_timer(&policy
->timer
);
355 policy
->timer
.data
= (unsigned long)policy
;
356 policy
->timer
.function
= xfrm_policy_timer
;
360 EXPORT_SYMBOL(xfrm_policy_alloc
);
362 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
364 void __xfrm_policy_destroy(struct xfrm_policy
*policy
)
366 BUG_ON(!policy
->dead
);
368 BUG_ON(policy
->bundles
);
370 if (del_timer(&policy
->timer
))
373 security_xfrm_policy_free(policy
);
376 EXPORT_SYMBOL(__xfrm_policy_destroy
);
378 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
380 struct dst_entry
*dst
;
382 while ((dst
= policy
->bundles
) != NULL
) {
383 policy
->bundles
= dst
->next
;
387 if (del_timer(&policy
->timer
))
388 atomic_dec(&policy
->refcnt
);
390 if (atomic_read(&policy
->refcnt
) > 1)
393 xfrm_pol_put(policy
);
396 static void xfrm_policy_gc_task(struct work_struct
*work
)
398 struct xfrm_policy
*policy
;
399 struct hlist_node
*entry
, *tmp
;
400 struct hlist_head gc_list
;
402 spin_lock_bh(&xfrm_policy_gc_lock
);
403 gc_list
.first
= xfrm_policy_gc_list
.first
;
404 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
405 spin_unlock_bh(&xfrm_policy_gc_lock
);
407 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
408 xfrm_policy_gc_kill(policy
);
411 /* Rule must be locked. Release descentant resources, announce
412 * entry dead. The rule must be unlinked from lists to the moment.
415 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
419 write_lock_bh(&policy
->lock
);
422 write_unlock_bh(&policy
->lock
);
424 if (unlikely(dead
)) {
429 spin_lock(&xfrm_policy_gc_lock
);
430 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
431 spin_unlock(&xfrm_policy_gc_lock
);
433 schedule_work(&xfrm_policy_gc_work
);
436 struct xfrm_policy_hash
{
437 struct hlist_head
*table
;
441 static struct hlist_head xfrm_policy_inexact
[XFRM_POLICY_MAX
*2];
442 static struct xfrm_policy_hash xfrm_policy_bydst
[XFRM_POLICY_MAX
*2] __read_mostly
;
443 static struct hlist_head
*xfrm_policy_byidx __read_mostly
;
444 static unsigned int xfrm_idx_hmask __read_mostly
;
445 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
447 static inline unsigned int idx_hash(u32 index
)
449 return __idx_hash(index
, xfrm_idx_hmask
);
452 static struct hlist_head
*policy_hash_bysel(struct xfrm_selector
*sel
, unsigned short family
, int dir
)
454 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
455 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
457 return (hash
== hmask
+ 1 ?
458 &xfrm_policy_inexact
[dir
] :
459 xfrm_policy_bydst
[dir
].table
+ hash
);
462 static struct hlist_head
*policy_hash_direct(xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
464 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
465 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
467 return xfrm_policy_bydst
[dir
].table
+ hash
;
470 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
471 struct hlist_head
*ndsttable
,
472 unsigned int nhashmask
)
474 struct hlist_node
*entry
, *tmp
;
475 struct xfrm_policy
*pol
;
477 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
480 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
481 pol
->family
, nhashmask
);
482 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
486 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
487 struct hlist_head
*nidxtable
,
488 unsigned int nhashmask
)
490 struct hlist_node
*entry
, *tmp
;
491 struct xfrm_policy
*pol
;
493 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
496 h
= __idx_hash(pol
->index
, nhashmask
);
497 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
501 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
503 return ((old_hmask
+ 1) << 1) - 1;
506 static void xfrm_bydst_resize(int dir
)
508 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
509 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
510 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
511 struct hlist_head
*odst
= xfrm_policy_bydst
[dir
].table
;
512 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
518 write_lock_bh(&xfrm_policy_lock
);
520 for (i
= hmask
; i
>= 0; i
--)
521 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
523 xfrm_policy_bydst
[dir
].table
= ndst
;
524 xfrm_policy_bydst
[dir
].hmask
= nhashmask
;
526 write_unlock_bh(&xfrm_policy_lock
);
528 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
531 static void xfrm_byidx_resize(int total
)
533 unsigned int hmask
= xfrm_idx_hmask
;
534 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
535 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
536 struct hlist_head
*oidx
= xfrm_policy_byidx
;
537 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
543 write_lock_bh(&xfrm_policy_lock
);
545 for (i
= hmask
; i
>= 0; i
--)
546 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
548 xfrm_policy_byidx
= nidx
;
549 xfrm_idx_hmask
= nhashmask
;
551 write_unlock_bh(&xfrm_policy_lock
);
553 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
556 static inline int xfrm_bydst_should_resize(int dir
, int *total
)
558 unsigned int cnt
= xfrm_policy_count
[dir
];
559 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
564 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
571 static inline int xfrm_byidx_should_resize(int total
)
573 unsigned int hmask
= xfrm_idx_hmask
;
575 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
582 static DEFINE_MUTEX(hash_resize_mutex
);
584 static void xfrm_hash_resize(struct work_struct
*__unused
)
588 mutex_lock(&hash_resize_mutex
);
591 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
592 if (xfrm_bydst_should_resize(dir
, &total
))
593 xfrm_bydst_resize(dir
);
595 if (xfrm_byidx_should_resize(total
))
596 xfrm_byidx_resize(total
);
598 mutex_unlock(&hash_resize_mutex
);
601 static DECLARE_WORK(xfrm_hash_work
, xfrm_hash_resize
);
603 /* Generate new index... KAME seems to generate them ordered by cost
604 * of an absolute inpredictability of ordering of rules. This will not pass. */
605 static u32
xfrm_gen_index(u8 type
, int dir
)
607 static u32 idx_generator
;
610 struct hlist_node
*entry
;
611 struct hlist_head
*list
;
612 struct xfrm_policy
*p
;
616 idx
= (idx_generator
| dir
);
620 list
= xfrm_policy_byidx
+ idx_hash(idx
);
622 hlist_for_each_entry(p
, entry
, list
, byidx
) {
623 if (p
->index
== idx
) {
633 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
635 u32
*p1
= (u32
*) s1
;
636 u32
*p2
= (u32
*) s2
;
637 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
640 for (i
= 0; i
< len
; i
++) {
648 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
650 struct xfrm_policy
*pol
;
651 struct xfrm_policy
*delpol
;
652 struct hlist_head
*chain
;
653 struct hlist_node
*entry
, *newpos
, *last
;
654 struct dst_entry
*gc_list
;
656 write_lock_bh(&xfrm_policy_lock
);
657 chain
= policy_hash_bysel(&policy
->selector
, policy
->family
, dir
);
661 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
663 pol
->type
== policy
->type
&&
664 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
665 xfrm_sec_ctx_match(pol
->security
, policy
->security
)) {
667 write_unlock_bh(&xfrm_policy_lock
);
671 if (policy
->priority
> pol
->priority
)
673 } else if (policy
->priority
>= pol
->priority
) {
678 newpos
= &pol
->bydst
;
686 hlist_add_after(newpos
, &policy
->bydst
);
688 hlist_add_head(&policy
->bydst
, chain
);
689 xfrm_pol_hold(policy
);
690 xfrm_policy_count
[dir
]++;
691 atomic_inc(&flow_cache_genid
);
693 hlist_del(&delpol
->bydst
);
694 hlist_del(&delpol
->byidx
);
695 xfrm_policy_count
[dir
]--;
697 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(policy
->type
, dir
);
698 hlist_add_head(&policy
->byidx
, xfrm_policy_byidx
+idx_hash(policy
->index
));
699 policy
->curlft
.add_time
= (unsigned long)xtime
.tv_sec
;
700 policy
->curlft
.use_time
= 0;
701 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
702 xfrm_pol_hold(policy
);
703 write_unlock_bh(&xfrm_policy_lock
);
706 xfrm_policy_kill(delpol
);
707 else if (xfrm_bydst_should_resize(dir
, NULL
))
708 schedule_work(&xfrm_hash_work
);
710 read_lock_bh(&xfrm_policy_lock
);
712 entry
= &policy
->bydst
;
713 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
714 struct dst_entry
*dst
;
716 write_lock(&policy
->lock
);
717 dst
= policy
->bundles
;
719 struct dst_entry
*tail
= dst
;
722 tail
->next
= gc_list
;
725 policy
->bundles
= NULL
;
727 write_unlock(&policy
->lock
);
729 read_unlock_bh(&xfrm_policy_lock
);
732 struct dst_entry
*dst
= gc_list
;
740 EXPORT_SYMBOL(xfrm_policy_insert
);
742 struct xfrm_policy
*xfrm_policy_bysel_ctx(u8 type
, int dir
,
743 struct xfrm_selector
*sel
,
744 struct xfrm_sec_ctx
*ctx
, int delete)
746 struct xfrm_policy
*pol
, *ret
;
747 struct hlist_head
*chain
;
748 struct hlist_node
*entry
;
750 write_lock_bh(&xfrm_policy_lock
);
751 chain
= policy_hash_bysel(sel
, sel
->family
, dir
);
753 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
754 if (pol
->type
== type
&&
755 !selector_cmp(sel
, &pol
->selector
) &&
756 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
759 hlist_del(&pol
->bydst
);
760 hlist_del(&pol
->byidx
);
761 xfrm_policy_count
[dir
]--;
767 write_unlock_bh(&xfrm_policy_lock
);
770 atomic_inc(&flow_cache_genid
);
771 xfrm_policy_kill(ret
);
775 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
777 struct xfrm_policy
*xfrm_policy_byid(u8 type
, int dir
, u32 id
, int delete)
779 struct xfrm_policy
*pol
, *ret
;
780 struct hlist_head
*chain
;
781 struct hlist_node
*entry
;
783 write_lock_bh(&xfrm_policy_lock
);
784 chain
= xfrm_policy_byidx
+ idx_hash(id
);
786 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
787 if (pol
->type
== type
&& pol
->index
== id
) {
790 hlist_del(&pol
->bydst
);
791 hlist_del(&pol
->byidx
);
792 xfrm_policy_count
[dir
]--;
798 write_unlock_bh(&xfrm_policy_lock
);
801 atomic_inc(&flow_cache_genid
);
802 xfrm_policy_kill(ret
);
806 EXPORT_SYMBOL(xfrm_policy_byid
);
808 void xfrm_policy_flush(u8 type
, struct xfrm_audit
*audit_info
)
812 write_lock_bh(&xfrm_policy_lock
);
813 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
814 struct xfrm_policy
*pol
;
815 struct hlist_node
*entry
;
820 hlist_for_each_entry(pol
, entry
,
821 &xfrm_policy_inexact
[dir
], bydst
) {
822 if (pol
->type
!= type
)
824 hlist_del(&pol
->bydst
);
825 hlist_del(&pol
->byidx
);
826 write_unlock_bh(&xfrm_policy_lock
);
828 xfrm_audit_log(audit_info
->loginuid
, audit_info
->secid
,
829 AUDIT_MAC_IPSEC_DELSPD
, 1, pol
, NULL
);
831 xfrm_policy_kill(pol
);
834 write_lock_bh(&xfrm_policy_lock
);
838 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
840 hlist_for_each_entry(pol
, entry
,
841 xfrm_policy_bydst
[dir
].table
+ i
,
843 if (pol
->type
!= type
)
845 hlist_del(&pol
->bydst
);
846 hlist_del(&pol
->byidx
);
847 write_unlock_bh(&xfrm_policy_lock
);
849 xfrm_audit_log(audit_info
->loginuid
,
851 AUDIT_MAC_IPSEC_DELSPD
, 1,
854 xfrm_policy_kill(pol
);
857 write_lock_bh(&xfrm_policy_lock
);
862 xfrm_policy_count
[dir
] -= killed
;
864 atomic_inc(&flow_cache_genid
);
865 write_unlock_bh(&xfrm_policy_lock
);
867 EXPORT_SYMBOL(xfrm_policy_flush
);
869 int xfrm_policy_walk(u8 type
, int (*func
)(struct xfrm_policy
*, int, int, void*),
872 struct xfrm_policy
*pol
, *last
= NULL
;
873 struct hlist_node
*entry
;
874 int dir
, last_dir
= 0, count
, error
;
876 read_lock_bh(&xfrm_policy_lock
);
879 for (dir
= 0; dir
< 2*XFRM_POLICY_MAX
; dir
++) {
880 struct hlist_head
*table
= xfrm_policy_bydst
[dir
].table
;
883 hlist_for_each_entry(pol
, entry
,
884 &xfrm_policy_inexact
[dir
], bydst
) {
885 if (pol
->type
!= type
)
888 error
= func(last
, last_dir
% XFRM_POLICY_MAX
,
897 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
898 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
) {
899 if (pol
->type
!= type
)
902 error
= func(last
, last_dir
% XFRM_POLICY_MAX
,
917 error
= func(last
, last_dir
% XFRM_POLICY_MAX
, 0, data
);
919 read_unlock_bh(&xfrm_policy_lock
);
922 EXPORT_SYMBOL(xfrm_policy_walk
);
925 * Find policy to apply to this flow.
927 * Returns 0 if policy found, else an -errno.
929 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
930 u8 type
, u16 family
, int dir
)
932 struct xfrm_selector
*sel
= &pol
->selector
;
933 int match
, ret
= -ESRCH
;
935 if (pol
->family
!= family
||
939 match
= xfrm_selector_match(sel
, fl
, family
);
941 ret
= security_xfrm_policy_lookup(pol
, fl
->secid
, dir
);
946 static struct xfrm_policy
*xfrm_policy_lookup_bytype(u8 type
, struct flowi
*fl
,
950 struct xfrm_policy
*pol
, *ret
;
951 xfrm_address_t
*daddr
, *saddr
;
952 struct hlist_node
*entry
;
953 struct hlist_head
*chain
;
956 daddr
= xfrm_flowi_daddr(fl
, family
);
957 saddr
= xfrm_flowi_saddr(fl
, family
);
958 if (unlikely(!daddr
|| !saddr
))
961 read_lock_bh(&xfrm_policy_lock
);
962 chain
= policy_hash_direct(daddr
, saddr
, family
, dir
);
964 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
965 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
975 priority
= ret
->priority
;
979 chain
= &xfrm_policy_inexact
[dir
];
980 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
981 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
989 } else if (pol
->priority
< priority
) {
997 read_unlock_bh(&xfrm_policy_lock
);
1002 static int xfrm_policy_lookup(struct flowi
*fl
, u16 family
, u8 dir
,
1003 void **objp
, atomic_t
**obj_refp
)
1005 struct xfrm_policy
*pol
;
1008 #ifdef CONFIG_XFRM_SUB_POLICY
1009 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1017 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1022 #ifdef CONFIG_XFRM_SUB_POLICY
1025 if ((*objp
= (void *) pol
) != NULL
)
1026 *obj_refp
= &pol
->refcnt
;
1030 static inline int policy_to_flow_dir(int dir
)
1032 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1033 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1034 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1038 case XFRM_POLICY_IN
:
1040 case XFRM_POLICY_OUT
:
1041 return FLOW_DIR_OUT
;
1042 case XFRM_POLICY_FWD
:
1043 return FLOW_DIR_FWD
;
1047 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1049 struct xfrm_policy
*pol
;
1051 read_lock_bh(&xfrm_policy_lock
);
1052 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1053 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1058 err
= security_xfrm_policy_lookup(pol
, fl
->secid
,
1059 policy_to_flow_dir(dir
));
1062 else if (err
== -ESRCH
)
1069 read_unlock_bh(&xfrm_policy_lock
);
1073 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1075 struct hlist_head
*chain
= policy_hash_bysel(&pol
->selector
,
1078 hlist_add_head(&pol
->bydst
, chain
);
1079 hlist_add_head(&pol
->byidx
, xfrm_policy_byidx
+idx_hash(pol
->index
));
1080 xfrm_policy_count
[dir
]++;
1083 if (xfrm_bydst_should_resize(dir
, NULL
))
1084 schedule_work(&xfrm_hash_work
);
1087 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1090 if (hlist_unhashed(&pol
->bydst
))
1093 hlist_del(&pol
->bydst
);
1094 hlist_del(&pol
->byidx
);
1095 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
= (unsigned long)xtime
.tv_sec
;
1129 pol
->index
= xfrm_gen_index(pol
->type
, 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(GFP_ATOMIC
);
1147 newp
->selector
= old
->selector
;
1148 if (security_xfrm_policy_clone(old
, newp
)) {
1150 return NULL
; /* ENOMEM */
1152 newp
->lft
= old
->lft
;
1153 newp
->curlft
= old
->curlft
;
1154 newp
->action
= old
->action
;
1155 newp
->flags
= old
->flags
;
1156 newp
->xfrm_nr
= old
->xfrm_nr
;
1157 newp
->index
= old
->index
;
1158 newp
->type
= old
->type
;
1159 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1160 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1161 write_lock_bh(&xfrm_policy_lock
);
1162 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1163 write_unlock_bh(&xfrm_policy_lock
);
1169 int __xfrm_sk_clone_policy(struct sock
*sk
)
1171 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1172 *p1
= sk
->sk_policy
[1];
1174 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1175 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1177 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1183 xfrm_get_saddr(xfrm_address_t
*local
, xfrm_address_t
*remote
,
1184 unsigned short family
)
1187 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1189 if (unlikely(afinfo
== NULL
))
1191 err
= afinfo
->get_saddr(local
, remote
);
1192 xfrm_policy_put_afinfo(afinfo
);
1196 /* Resolve list of templates for the flow, given policy. */
1199 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1200 struct xfrm_state
**xfrm
,
1201 unsigned short family
)
1205 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1206 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1209 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1210 struct xfrm_state
*x
;
1211 xfrm_address_t
*remote
= daddr
;
1212 xfrm_address_t
*local
= saddr
;
1213 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1215 if (tmpl
->mode
== XFRM_MODE_TUNNEL
) {
1216 remote
= &tmpl
->id
.daddr
;
1217 local
= &tmpl
->saddr
;
1218 family
= tmpl
->encap_family
;
1219 if (xfrm_addr_any(local
, family
)) {
1220 error
= xfrm_get_saddr(&tmp
, remote
, family
);
1227 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1229 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1236 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1241 if (!tmpl
->optional
)
1247 for (nx
--; nx
>=0; nx
--)
1248 xfrm_state_put(xfrm
[nx
]);
1253 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1254 struct xfrm_state
**xfrm
,
1255 unsigned short family
)
1257 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1258 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1264 for (i
= 0; i
< npols
; i
++) {
1265 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1270 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1278 /* found states are sorted for outbound processing */
1280 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1285 for (cnx
--; cnx
>=0; cnx
--)
1286 xfrm_state_put(tpp
[cnx
]);
1291 /* Check that the bundle accepts the flow and its components are
1295 static struct dst_entry
*
1296 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1298 struct dst_entry
*x
;
1299 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1300 if (unlikely(afinfo
== NULL
))
1301 return ERR_PTR(-EINVAL
);
1302 x
= afinfo
->find_bundle(fl
, policy
);
1303 xfrm_policy_put_afinfo(afinfo
);
1307 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1308 * all the metrics... Shortly, bundle a bundle.
1312 xfrm_bundle_create(struct xfrm_policy
*policy
, struct xfrm_state
**xfrm
, int nx
,
1313 struct flowi
*fl
, struct dst_entry
**dst_p
,
1314 unsigned short family
)
1317 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1318 if (unlikely(afinfo
== NULL
))
1320 err
= afinfo
->bundle_create(policy
, xfrm
, nx
, fl
, dst_p
);
1321 xfrm_policy_put_afinfo(afinfo
);
1326 static int stale_bundle(struct dst_entry
*dst
);
1328 /* Main function: finds/creates a bundle for given flow.
1330 * At the moment we eat a raw IP route. Mostly to speed up lookups
1331 * on interfaces with disabled IPsec.
1333 int xfrm_lookup(struct dst_entry
**dst_p
, struct flowi
*fl
,
1334 struct sock
*sk
, int flags
)
1336 struct xfrm_policy
*policy
;
1337 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1342 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1343 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1348 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1351 genid
= atomic_read(&flow_cache_genid
);
1353 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1359 if (sk
&& sk
->sk_policy
[1]) {
1360 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1362 return PTR_ERR(policy
);
1366 /* To accelerate a bit... */
1367 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1368 !xfrm_policy_count
[XFRM_POLICY_OUT
])
1371 policy
= flow_cache_lookup(fl
, dst_orig
->ops
->family
,
1372 dir
, xfrm_policy_lookup
);
1374 return PTR_ERR(policy
);
1380 family
= dst_orig
->ops
->family
;
1381 policy
->curlft
.use_time
= (unsigned long)xtime
.tv_sec
;
1384 xfrm_nr
+= pols
[0]->xfrm_nr
;
1386 switch (policy
->action
) {
1387 case XFRM_POLICY_BLOCK
:
1388 /* Prohibit the flow */
1392 case XFRM_POLICY_ALLOW
:
1393 #ifndef CONFIG_XFRM_SUB_POLICY
1394 if (policy
->xfrm_nr
== 0) {
1395 /* Flow passes not transformed. */
1396 xfrm_pol_put(policy
);
1401 /* Try to find matching bundle.
1403 * LATER: help from flow cache. It is optional, this
1404 * is required only for output policy.
1406 dst
= xfrm_find_bundle(fl
, policy
, family
);
1415 #ifdef CONFIG_XFRM_SUB_POLICY
1416 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1417 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1421 if (IS_ERR(pols
[1])) {
1422 err
= PTR_ERR(pols
[1]);
1425 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1430 xfrm_nr
+= pols
[1]->xfrm_nr
;
1435 * Because neither flowi nor bundle information knows about
1436 * transformation template size. On more than one policy usage
1437 * we can realize whether all of them is bypass or not after
1438 * they are searched. See above not-transformed bypass
1439 * is surrounded by non-sub policy configuration, too.
1442 /* Flow passes not transformed. */
1443 xfrm_pols_put(pols
, npols
);
1448 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1450 if (unlikely(nx
<0)) {
1452 if (err
== -EAGAIN
&& flags
) {
1453 DECLARE_WAITQUEUE(wait
, current
);
1455 add_wait_queue(&km_waitq
, &wait
);
1456 set_current_state(TASK_INTERRUPTIBLE
);
1458 set_current_state(TASK_RUNNING
);
1459 remove_wait_queue(&km_waitq
, &wait
);
1461 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1463 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1467 if (nx
== -EAGAIN
||
1468 genid
!= atomic_read(&flow_cache_genid
)) {
1469 xfrm_pols_put(pols
, npols
);
1478 /* Flow passes not transformed. */
1479 xfrm_pols_put(pols
, npols
);
1484 err
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, &dst
, family
);
1486 if (unlikely(err
)) {
1488 for (i
=0; i
<nx
; i
++)
1489 xfrm_state_put(xfrm
[i
]);
1493 for (pi
= 0; pi
< npols
; pi
++) {
1494 read_lock_bh(&pols
[pi
]->lock
);
1495 pol_dead
|= pols
[pi
]->dead
;
1496 read_unlock_bh(&pols
[pi
]->lock
);
1499 write_lock_bh(&policy
->lock
);
1500 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1501 /* Wow! While we worked on resolving, this
1502 * policy has gone. Retry. It is not paranoia,
1503 * we just cannot enlist new bundle to dead object.
1504 * We can't enlist stable bundles either.
1506 write_unlock_bh(&policy
->lock
);
1510 err
= -EHOSTUNREACH
;
1513 dst
->next
= policy
->bundles
;
1514 policy
->bundles
= dst
;
1516 write_unlock_bh(&policy
->lock
);
1519 dst_release(dst_orig
);
1520 xfrm_pols_put(pols
, npols
);
1524 dst_release(dst_orig
);
1525 xfrm_pols_put(pols
, npols
);
1529 EXPORT_SYMBOL(xfrm_lookup
);
1532 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1534 struct xfrm_state
*x
;
1537 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1539 x
= skb
->sp
->xvec
[idx
];
1540 if (!x
->type
->reject
)
1543 err
= x
->type
->reject(x
, skb
, fl
);
1548 /* When skb is transformed back to its "native" form, we have to
1549 * check policy restrictions. At the moment we make this in maximally
1550 * stupid way. Shame on me. :-) Of course, connected sockets must
1551 * have policy cached at them.
1555 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1556 unsigned short family
)
1558 if (xfrm_state_kern(x
))
1559 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, family
);
1560 return x
->id
.proto
== tmpl
->id
.proto
&&
1561 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1562 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1563 x
->props
.mode
== tmpl
->mode
&&
1564 ((tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1565 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1566 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1567 xfrm_state_addr_cmp(tmpl
, x
, family
));
1571 * 0 or more than 0 is returned when validation is succeeded (either bypass
1572 * because of optional transport mode, or next index of the mathced secpath
1573 * state with the template.
1574 * -1 is returned when no matching template is found.
1575 * Otherwise "-2 - errored_index" is returned.
1578 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1579 unsigned short family
)
1583 if (tmpl
->optional
) {
1584 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1588 for (; idx
< sp
->len
; idx
++) {
1589 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1591 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1601 xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
, unsigned short family
)
1603 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1606 if (unlikely(afinfo
== NULL
))
1607 return -EAFNOSUPPORT
;
1609 afinfo
->decode_session(skb
, fl
);
1610 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1611 xfrm_policy_put_afinfo(afinfo
);
1614 EXPORT_SYMBOL(xfrm_decode_session
);
1616 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1618 for (; k
< sp
->len
; k
++) {
1619 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1628 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1629 unsigned short family
)
1631 struct xfrm_policy
*pol
;
1632 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1637 u8 fl_dir
= policy_to_flow_dir(dir
);
1640 if (xfrm_decode_session(skb
, &fl
, family
) < 0)
1642 nf_nat_decode_session(skb
, &fl
, family
);
1644 /* First, check used SA against their selectors. */
1648 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1649 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1650 if (!xfrm_selector_match(&x
->sel
, &fl
, family
))
1656 if (sk
&& sk
->sk_policy
[dir
]) {
1657 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1663 pol
= flow_cache_lookup(&fl
, family
, fl_dir
,
1664 xfrm_policy_lookup
);
1670 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1671 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1677 pol
->curlft
.use_time
= (unsigned long)xtime
.tv_sec
;
1681 #ifdef CONFIG_XFRM_SUB_POLICY
1682 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1683 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1687 if (IS_ERR(pols
[1]))
1689 pols
[1]->curlft
.use_time
= (unsigned long)xtime
.tv_sec
;
1695 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1696 struct sec_path
*sp
;
1697 static struct sec_path dummy
;
1698 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1699 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1700 struct xfrm_tmpl
**tpp
= tp
;
1704 if ((sp
= skb
->sp
) == NULL
)
1707 for (pi
= 0; pi
< npols
; pi
++) {
1708 if (pols
[pi
] != pol
&&
1709 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
)
1711 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
)
1713 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
1714 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
1718 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
1722 /* For each tunnel xfrm, find the first matching tmpl.
1723 * For each tmpl before that, find corresponding xfrm.
1724 * Order is _important_. Later we will implement
1725 * some barriers, but at the moment barriers
1726 * are implied between each two transformations.
1728 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
1729 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
1732 /* "-2 - errored_index" returned */
1738 if (secpath_has_nontransport(sp
, k
, &xerr_idx
))
1741 xfrm_pols_put(pols
, npols
);
1746 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1748 xfrm_pols_put(pols
, npols
);
1751 EXPORT_SYMBOL(__xfrm_policy_check
);
1753 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
1757 if (xfrm_decode_session(skb
, &fl
, family
) < 0)
1760 return xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) == 0;
1762 EXPORT_SYMBOL(__xfrm_route_forward
);
1764 /* Optimize later using cookies and generation ids. */
1766 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
1768 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
1769 * to "-1" to force all XFRM destinations to get validated by
1770 * dst_ops->check on every use. We do this because when a
1771 * normal route referenced by an XFRM dst is obsoleted we do
1772 * not go looking around for all parent referencing XFRM dsts
1773 * so that we can invalidate them. It is just too much work.
1774 * Instead we make the checks here on every use. For example:
1776 * XFRM dst A --> IPv4 dst X
1778 * X is the "xdst->route" of A (X is also the "dst->path" of A
1779 * in this example). If X is marked obsolete, "A" will not
1780 * notice. That's what we are validating here via the
1781 * stale_bundle() check.
1783 * When a policy's bundle is pruned, we dst_free() the XFRM
1784 * dst which causes it's ->obsolete field to be set to a
1785 * positive non-zero integer. If an XFRM dst has been pruned
1786 * like this, we want to force a new route lookup.
1788 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
1794 static int stale_bundle(struct dst_entry
*dst
)
1796 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
1799 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
1801 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
1802 dst
->dev
= &loopback_dev
;
1803 dev_hold(&loopback_dev
);
1807 EXPORT_SYMBOL(xfrm_dst_ifdown
);
1809 static void xfrm_link_failure(struct sk_buff
*skb
)
1811 /* Impossible. Such dst must be popped before reaches point of failure. */
1815 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
1818 if (dst
->obsolete
) {
1826 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
1828 struct dst_entry
*dst
, **dstp
;
1830 write_lock(&pol
->lock
);
1831 dstp
= &pol
->bundles
;
1832 while ((dst
=*dstp
) != NULL
) {
1835 dst
->next
= *gc_list_p
;
1841 write_unlock(&pol
->lock
);
1844 static void xfrm_prune_bundles(int (*func
)(struct dst_entry
*))
1846 struct dst_entry
*gc_list
= NULL
;
1849 read_lock_bh(&xfrm_policy_lock
);
1850 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
1851 struct xfrm_policy
*pol
;
1852 struct hlist_node
*entry
;
1853 struct hlist_head
*table
;
1856 hlist_for_each_entry(pol
, entry
,
1857 &xfrm_policy_inexact
[dir
], bydst
)
1858 prune_one_bundle(pol
, func
, &gc_list
);
1860 table
= xfrm_policy_bydst
[dir
].table
;
1861 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
1862 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
1863 prune_one_bundle(pol
, func
, &gc_list
);
1866 read_unlock_bh(&xfrm_policy_lock
);
1869 struct dst_entry
*dst
= gc_list
;
1870 gc_list
= dst
->next
;
1875 static int unused_bundle(struct dst_entry
*dst
)
1877 return !atomic_read(&dst
->__refcnt
);
1880 static void __xfrm_garbage_collect(void)
1882 xfrm_prune_bundles(unused_bundle
);
1885 static int xfrm_flush_bundles(void)
1887 xfrm_prune_bundles(stale_bundle
);
1891 void xfrm_init_pmtu(struct dst_entry
*dst
)
1894 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1895 u32 pmtu
, route_mtu_cached
;
1897 pmtu
= dst_mtu(dst
->child
);
1898 xdst
->child_mtu_cached
= pmtu
;
1900 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
1902 route_mtu_cached
= dst_mtu(xdst
->route
);
1903 xdst
->route_mtu_cached
= route_mtu_cached
;
1905 if (pmtu
> route_mtu_cached
)
1906 pmtu
= route_mtu_cached
;
1908 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
1909 } while ((dst
= dst
->next
));
1912 EXPORT_SYMBOL(xfrm_init_pmtu
);
1914 /* Check that the bundle accepts the flow and its components are
1918 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
1919 struct flowi
*fl
, int family
, int strict
)
1921 struct dst_entry
*dst
= &first
->u
.dst
;
1922 struct xfrm_dst
*last
;
1925 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
1926 (dst
->dev
&& !netif_running(dst
->dev
)))
1932 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1934 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
1937 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
1939 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
1941 if (xdst
->genid
!= dst
->xfrm
->genid
)
1944 if (strict
&& fl
&& dst
->xfrm
->props
.mode
!= XFRM_MODE_TUNNEL
&&
1945 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
1948 mtu
= dst_mtu(dst
->child
);
1949 if (xdst
->child_mtu_cached
!= mtu
) {
1951 xdst
->child_mtu_cached
= mtu
;
1954 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
1956 mtu
= dst_mtu(xdst
->route
);
1957 if (xdst
->route_mtu_cached
!= mtu
) {
1959 xdst
->route_mtu_cached
= mtu
;
1963 } while (dst
->xfrm
);
1968 mtu
= last
->child_mtu_cached
;
1972 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
1973 if (mtu
> last
->route_mtu_cached
)
1974 mtu
= last
->route_mtu_cached
;
1975 dst
->metrics
[RTAX_MTU
-1] = mtu
;
1980 last
= last
->u
.next
;
1981 last
->child_mtu_cached
= mtu
;
1987 EXPORT_SYMBOL(xfrm_bundle_ok
);
1989 /* Audit addition and deletion of SAs and ipsec policy */
1991 void xfrm_audit_log(uid_t auid
, u32 sid
, int type
, int result
,
1992 struct xfrm_policy
*xp
, struct xfrm_state
*x
)
1997 struct xfrm_sec_ctx
*sctx
= NULL
;
1998 struct audit_buffer
*audit_buf
;
2000 extern int audit_enabled
;
2002 if (audit_enabled
== 0)
2005 audit_buf
= audit_log_start(current
->audit_context
, GFP_ATOMIC
, type
);
2006 if (audit_buf
== NULL
)
2010 case AUDIT_MAC_IPSEC_ADDSA
:
2011 audit_log_format(audit_buf
, "SAD add: auid=%u", auid
);
2013 case AUDIT_MAC_IPSEC_DELSA
:
2014 audit_log_format(audit_buf
, "SAD delete: auid=%u", auid
);
2016 case AUDIT_MAC_IPSEC_ADDSPD
:
2017 audit_log_format(audit_buf
, "SPD add: auid=%u", auid
);
2019 case AUDIT_MAC_IPSEC_DELSPD
:
2020 audit_log_format(audit_buf
, "SPD delete: auid=%u", auid
);
2027 security_secid_to_secctx(sid
, &secctx
, &secctx_len
) == 0)
2028 audit_log_format(audit_buf
, " subj=%s", secctx
);
2030 audit_log_task_context(audit_buf
);
2033 family
= xp
->selector
.family
;
2035 sctx
= xp
->security
;
2037 family
= x
->props
.family
;
2043 audit_log_format(audit_buf
,
2044 " sec_alg=%u sec_doi=%u sec_obj=%s",
2045 sctx
->ctx_alg
, sctx
->ctx_doi
, sctx
->ctx_str
);
2050 struct in_addr saddr
, daddr
;
2052 saddr
.s_addr
= xp
->selector
.saddr
.a4
;
2053 daddr
.s_addr
= xp
->selector
.daddr
.a4
;
2055 saddr
.s_addr
= x
->props
.saddr
.a4
;
2056 daddr
.s_addr
= x
->id
.daddr
.a4
;
2058 audit_log_format(audit_buf
,
2059 " src=%u.%u.%u.%u dst=%u.%u.%u.%u",
2060 NIPQUAD(saddr
), NIPQUAD(daddr
));
2065 struct in6_addr saddr6
, daddr6
;
2067 memcpy(&saddr6
, xp
->selector
.saddr
.a6
,
2068 sizeof(struct in6_addr
));
2069 memcpy(&daddr6
, xp
->selector
.daddr
.a6
,
2070 sizeof(struct in6_addr
));
2072 memcpy(&saddr6
, x
->props
.saddr
.a6
,
2073 sizeof(struct in6_addr
));
2074 memcpy(&daddr6
, x
->id
.daddr
.a6
,
2075 sizeof(struct in6_addr
));
2077 audit_log_format(audit_buf
,
2078 " src=" NIP6_FMT
"dst=" NIP6_FMT
,
2079 NIP6(saddr6
), NIP6(daddr6
));
2085 audit_log_format(audit_buf
, " spi=%lu(0x%lx) protocol=%s",
2086 (unsigned long)ntohl(x
->id
.spi
),
2087 (unsigned long)ntohl(x
->id
.spi
),
2088 x
->id
.proto
== IPPROTO_AH
? "AH" :
2089 (x
->id
.proto
== IPPROTO_ESP
?
2092 audit_log_format(audit_buf
, " res=%u", result
);
2093 audit_log_end(audit_buf
);
2096 EXPORT_SYMBOL(xfrm_audit_log
);
2098 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2101 if (unlikely(afinfo
== NULL
))
2103 if (unlikely(afinfo
->family
>= NPROTO
))
2104 return -EAFNOSUPPORT
;
2105 write_lock_bh(&xfrm_policy_afinfo_lock
);
2106 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2109 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2110 if (likely(dst_ops
->kmem_cachep
== NULL
))
2111 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2112 if (likely(dst_ops
->check
== NULL
))
2113 dst_ops
->check
= xfrm_dst_check
;
2114 if (likely(dst_ops
->negative_advice
== NULL
))
2115 dst_ops
->negative_advice
= xfrm_negative_advice
;
2116 if (likely(dst_ops
->link_failure
== NULL
))
2117 dst_ops
->link_failure
= xfrm_link_failure
;
2118 if (likely(afinfo
->garbage_collect
== NULL
))
2119 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2120 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2122 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2125 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2127 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2130 if (unlikely(afinfo
== NULL
))
2132 if (unlikely(afinfo
->family
>= NPROTO
))
2133 return -EAFNOSUPPORT
;
2134 write_lock_bh(&xfrm_policy_afinfo_lock
);
2135 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2136 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2139 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2140 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2141 dst_ops
->kmem_cachep
= NULL
;
2142 dst_ops
->check
= NULL
;
2143 dst_ops
->negative_advice
= NULL
;
2144 dst_ops
->link_failure
= NULL
;
2145 afinfo
->garbage_collect
= NULL
;
2148 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2151 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2153 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2155 struct xfrm_policy_afinfo
*afinfo
;
2156 if (unlikely(family
>= NPROTO
))
2158 read_lock(&xfrm_policy_afinfo_lock
);
2159 afinfo
= xfrm_policy_afinfo
[family
];
2160 if (unlikely(!afinfo
))
2161 read_unlock(&xfrm_policy_afinfo_lock
);
2165 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2167 read_unlock(&xfrm_policy_afinfo_lock
);
2170 static struct xfrm_policy_afinfo
*xfrm_policy_lock_afinfo(unsigned int family
)
2172 struct xfrm_policy_afinfo
*afinfo
;
2173 if (unlikely(family
>= NPROTO
))
2175 write_lock_bh(&xfrm_policy_afinfo_lock
);
2176 afinfo
= xfrm_policy_afinfo
[family
];
2177 if (unlikely(!afinfo
))
2178 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2182 static void xfrm_policy_unlock_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2184 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2187 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2191 xfrm_flush_bundles();
2196 static struct notifier_block xfrm_dev_notifier
= {
2202 static void __init
xfrm_policy_init(void)
2204 unsigned int hmask
, sz
;
2207 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2208 sizeof(struct xfrm_dst
),
2209 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2213 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2215 xfrm_policy_byidx
= xfrm_hash_alloc(sz
);
2216 xfrm_idx_hmask
= hmask
;
2217 if (!xfrm_policy_byidx
)
2218 panic("XFRM: failed to allocate byidx hash\n");
2220 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2221 struct xfrm_policy_hash
*htab
;
2223 INIT_HLIST_HEAD(&xfrm_policy_inexact
[dir
]);
2225 htab
= &xfrm_policy_bydst
[dir
];
2226 htab
->table
= xfrm_hash_alloc(sz
);
2227 htab
->hmask
= hmask
;
2229 panic("XFRM: failed to allocate bydst hash\n");
2232 INIT_WORK(&xfrm_policy_gc_work
, xfrm_policy_gc_task
);
2233 register_netdevice_notifier(&xfrm_dev_notifier
);
2236 void __init
xfrm_init(void)