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 DEFINE_MUTEX(xfrm_cfg_mutex
);
38 EXPORT_SYMBOL(xfrm_cfg_mutex
);
40 static DEFINE_RWLOCK(xfrm_policy_lock
);
42 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
43 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
45 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
47 static HLIST_HEAD(xfrm_policy_gc_list
);
48 static DEFINE_SPINLOCK(xfrm_policy_gc_lock
);
50 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
51 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
52 static void xfrm_init_pmtu(struct dst_entry
*dst
);
54 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
58 __xfrm4_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
60 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
61 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
62 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
63 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
64 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
65 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
69 __xfrm6_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
71 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
72 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
73 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
74 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
75 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
76 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
79 int xfrm_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
,
80 unsigned short family
)
84 return __xfrm4_selector_match(sel
, fl
);
86 return __xfrm6_selector_match(sel
, fl
);
91 static inline struct dst_entry
*__xfrm_dst_lookup(struct net
*net
, int tos
,
92 xfrm_address_t
*saddr
,
93 xfrm_address_t
*daddr
,
96 struct xfrm_policy_afinfo
*afinfo
;
97 struct dst_entry
*dst
;
99 afinfo
= xfrm_policy_get_afinfo(family
);
100 if (unlikely(afinfo
== NULL
))
101 return ERR_PTR(-EAFNOSUPPORT
);
103 dst
= afinfo
->dst_lookup(net
, tos
, saddr
, daddr
);
105 xfrm_policy_put_afinfo(afinfo
);
110 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
, int tos
,
111 xfrm_address_t
*prev_saddr
,
112 xfrm_address_t
*prev_daddr
,
115 struct net
*net
= xs_net(x
);
116 xfrm_address_t
*saddr
= &x
->props
.saddr
;
117 xfrm_address_t
*daddr
= &x
->id
.daddr
;
118 struct dst_entry
*dst
;
120 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
124 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
129 dst
= __xfrm_dst_lookup(net
, tos
, saddr
, daddr
, family
);
132 if (prev_saddr
!= saddr
)
133 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
134 if (prev_daddr
!= daddr
)
135 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
141 static inline unsigned long make_jiffies(long secs
)
143 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
144 return MAX_SCHEDULE_TIMEOUT
-1;
149 static void xfrm_policy_timer(unsigned long data
)
151 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
152 unsigned long now
= get_seconds();
153 long next
= LONG_MAX
;
157 read_lock(&xp
->lock
);
162 dir
= xfrm_policy_id2dir(xp
->index
);
164 if (xp
->lft
.hard_add_expires_seconds
) {
165 long tmo
= xp
->lft
.hard_add_expires_seconds
+
166 xp
->curlft
.add_time
- now
;
172 if (xp
->lft
.hard_use_expires_seconds
) {
173 long tmo
= xp
->lft
.hard_use_expires_seconds
+
174 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
180 if (xp
->lft
.soft_add_expires_seconds
) {
181 long tmo
= xp
->lft
.soft_add_expires_seconds
+
182 xp
->curlft
.add_time
- now
;
185 tmo
= XFRM_KM_TIMEOUT
;
190 if (xp
->lft
.soft_use_expires_seconds
) {
191 long tmo
= xp
->lft
.soft_use_expires_seconds
+
192 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
195 tmo
= XFRM_KM_TIMEOUT
;
202 km_policy_expired(xp
, dir
, 0, 0);
203 if (next
!= LONG_MAX
&&
204 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
208 read_unlock(&xp
->lock
);
213 read_unlock(&xp
->lock
);
214 if (!xfrm_policy_delete(xp
, dir
))
215 km_policy_expired(xp
, dir
, 1, 0);
220 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
224 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
226 struct xfrm_policy
*policy
;
228 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
231 write_pnet(&policy
->xp_net
, net
);
232 INIT_LIST_HEAD(&policy
->walk
.all
);
233 INIT_HLIST_NODE(&policy
->bydst
);
234 INIT_HLIST_NODE(&policy
->byidx
);
235 rwlock_init(&policy
->lock
);
236 atomic_set(&policy
->refcnt
, 1);
237 setup_timer(&policy
->timer
, xfrm_policy_timer
,
238 (unsigned long)policy
);
242 EXPORT_SYMBOL(xfrm_policy_alloc
);
244 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
246 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
248 BUG_ON(!policy
->walk
.dead
);
250 BUG_ON(policy
->bundles
);
252 if (del_timer(&policy
->timer
))
255 security_xfrm_policy_free(policy
->security
);
258 EXPORT_SYMBOL(xfrm_policy_destroy
);
260 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
262 struct dst_entry
*dst
;
264 while ((dst
= policy
->bundles
) != NULL
) {
265 policy
->bundles
= dst
->next
;
269 if (del_timer(&policy
->timer
))
270 atomic_dec(&policy
->refcnt
);
272 if (atomic_read(&policy
->refcnt
) > 1)
275 xfrm_pol_put(policy
);
278 static void xfrm_policy_gc_task(struct work_struct
*work
)
280 struct xfrm_policy
*policy
;
281 struct hlist_node
*entry
, *tmp
;
282 struct hlist_head gc_list
;
284 spin_lock_bh(&xfrm_policy_gc_lock
);
285 gc_list
.first
= xfrm_policy_gc_list
.first
;
286 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
287 spin_unlock_bh(&xfrm_policy_gc_lock
);
289 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
290 xfrm_policy_gc_kill(policy
);
292 static DECLARE_WORK(xfrm_policy_gc_work
, xfrm_policy_gc_task
);
294 /* Rule must be locked. Release descentant resources, announce
295 * entry dead. The rule must be unlinked from lists to the moment.
298 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
302 write_lock_bh(&policy
->lock
);
303 dead
= policy
->walk
.dead
;
304 policy
->walk
.dead
= 1;
305 write_unlock_bh(&policy
->lock
);
307 if (unlikely(dead
)) {
312 spin_lock_bh(&xfrm_policy_gc_lock
);
313 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
314 spin_unlock_bh(&xfrm_policy_gc_lock
);
316 schedule_work(&xfrm_policy_gc_work
);
319 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
321 static inline unsigned int idx_hash(struct net
*net
, u32 index
)
323 return __idx_hash(index
, net
->xfrm
.policy_idx_hmask
);
326 static struct hlist_head
*policy_hash_bysel(struct net
*net
, struct xfrm_selector
*sel
, unsigned short family
, int dir
)
328 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
329 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
331 return (hash
== hmask
+ 1 ?
332 &net
->xfrm
.policy_inexact
[dir
] :
333 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
336 static struct hlist_head
*policy_hash_direct(struct net
*net
, xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
338 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
339 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
341 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
344 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
345 struct hlist_head
*ndsttable
,
346 unsigned int nhashmask
)
348 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
349 struct xfrm_policy
*pol
;
353 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
356 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
357 pol
->family
, nhashmask
);
360 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
366 hlist_add_after(entry0
, &pol
->bydst
);
370 if (!hlist_empty(list
)) {
376 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
377 struct hlist_head
*nidxtable
,
378 unsigned int nhashmask
)
380 struct hlist_node
*entry
, *tmp
;
381 struct xfrm_policy
*pol
;
383 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
386 h
= __idx_hash(pol
->index
, nhashmask
);
387 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
391 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
393 return ((old_hmask
+ 1) << 1) - 1;
396 static void xfrm_bydst_resize(struct net
*net
, int dir
)
398 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
399 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
400 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
401 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
402 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
408 write_lock_bh(&xfrm_policy_lock
);
410 for (i
= hmask
; i
>= 0; i
--)
411 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
413 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
414 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
416 write_unlock_bh(&xfrm_policy_lock
);
418 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
421 static void xfrm_byidx_resize(struct net
*net
, int total
)
423 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
424 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
425 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
426 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
427 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
433 write_lock_bh(&xfrm_policy_lock
);
435 for (i
= hmask
; i
>= 0; i
--)
436 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
438 net
->xfrm
.policy_byidx
= nidx
;
439 net
->xfrm
.policy_idx_hmask
= nhashmask
;
441 write_unlock_bh(&xfrm_policy_lock
);
443 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
446 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
448 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
449 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
454 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
461 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
463 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
465 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
472 void xfrm_spd_getinfo(struct net
*net
, struct xfrmk_spdinfo
*si
)
474 read_lock_bh(&xfrm_policy_lock
);
475 si
->incnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
];
476 si
->outcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
];
477 si
->fwdcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
];
478 si
->inscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
479 si
->outscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
480 si
->fwdscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
481 si
->spdhcnt
= net
->xfrm
.policy_idx_hmask
;
482 si
->spdhmcnt
= xfrm_policy_hashmax
;
483 read_unlock_bh(&xfrm_policy_lock
);
485 EXPORT_SYMBOL(xfrm_spd_getinfo
);
487 static DEFINE_MUTEX(hash_resize_mutex
);
488 static void xfrm_hash_resize(struct work_struct
*work
)
490 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
493 mutex_lock(&hash_resize_mutex
);
496 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
497 if (xfrm_bydst_should_resize(net
, dir
, &total
))
498 xfrm_bydst_resize(net
, dir
);
500 if (xfrm_byidx_should_resize(net
, total
))
501 xfrm_byidx_resize(net
, total
);
503 mutex_unlock(&hash_resize_mutex
);
506 /* Generate new index... KAME seems to generate them ordered by cost
507 * of an absolute inpredictability of ordering of rules. This will not pass. */
508 static u32
xfrm_gen_index(struct net
*net
, int dir
)
510 static u32 idx_generator
;
513 struct hlist_node
*entry
;
514 struct hlist_head
*list
;
515 struct xfrm_policy
*p
;
519 idx
= (idx_generator
| dir
);
523 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
525 hlist_for_each_entry(p
, entry
, list
, byidx
) {
526 if (p
->index
== idx
) {
536 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
538 u32
*p1
= (u32
*) s1
;
539 u32
*p2
= (u32
*) s2
;
540 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
543 for (i
= 0; i
< len
; i
++) {
551 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
553 struct net
*net
= xp_net(policy
);
554 struct xfrm_policy
*pol
;
555 struct xfrm_policy
*delpol
;
556 struct hlist_head
*chain
;
557 struct hlist_node
*entry
, *newpos
;
558 struct dst_entry
*gc_list
;
560 write_lock_bh(&xfrm_policy_lock
);
561 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
564 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
565 if (pol
->type
== policy
->type
&&
566 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
567 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
570 write_unlock_bh(&xfrm_policy_lock
);
574 if (policy
->priority
> pol
->priority
)
576 } else if (policy
->priority
>= pol
->priority
) {
577 newpos
= &pol
->bydst
;
584 hlist_add_after(newpos
, &policy
->bydst
);
586 hlist_add_head(&policy
->bydst
, chain
);
587 xfrm_pol_hold(policy
);
588 net
->xfrm
.policy_count
[dir
]++;
589 atomic_inc(&flow_cache_genid
);
591 __xfrm_policy_unlink(delpol
, dir
);
592 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
593 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
594 policy
->curlft
.add_time
= get_seconds();
595 policy
->curlft
.use_time
= 0;
596 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
597 xfrm_pol_hold(policy
);
598 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
599 write_unlock_bh(&xfrm_policy_lock
);
602 xfrm_policy_kill(delpol
);
603 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
604 schedule_work(&net
->xfrm
.policy_hash_work
);
606 read_lock_bh(&xfrm_policy_lock
);
608 entry
= &policy
->bydst
;
609 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
610 struct dst_entry
*dst
;
612 write_lock(&policy
->lock
);
613 dst
= policy
->bundles
;
615 struct dst_entry
*tail
= dst
;
618 tail
->next
= gc_list
;
621 policy
->bundles
= NULL
;
623 write_unlock(&policy
->lock
);
625 read_unlock_bh(&xfrm_policy_lock
);
628 struct dst_entry
*dst
= gc_list
;
636 EXPORT_SYMBOL(xfrm_policy_insert
);
638 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u8 type
, int dir
,
639 struct xfrm_selector
*sel
,
640 struct xfrm_sec_ctx
*ctx
, int delete,
643 struct xfrm_policy
*pol
, *ret
;
644 struct hlist_head
*chain
;
645 struct hlist_node
*entry
;
648 write_lock_bh(&xfrm_policy_lock
);
649 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
651 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
652 if (pol
->type
== type
&&
653 !selector_cmp(sel
, &pol
->selector
) &&
654 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
657 *err
= security_xfrm_policy_delete(
660 write_unlock_bh(&xfrm_policy_lock
);
663 __xfrm_policy_unlink(pol
, dir
);
669 write_unlock_bh(&xfrm_policy_lock
);
672 atomic_inc(&flow_cache_genid
);
673 xfrm_policy_kill(ret
);
677 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
679 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u8 type
, int dir
, u32 id
,
680 int delete, int *err
)
682 struct xfrm_policy
*pol
, *ret
;
683 struct hlist_head
*chain
;
684 struct hlist_node
*entry
;
687 if (xfrm_policy_id2dir(id
) != dir
)
691 write_lock_bh(&xfrm_policy_lock
);
692 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
694 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
695 if (pol
->type
== type
&& pol
->index
== id
) {
698 *err
= security_xfrm_policy_delete(
701 write_unlock_bh(&xfrm_policy_lock
);
704 __xfrm_policy_unlink(pol
, dir
);
710 write_unlock_bh(&xfrm_policy_lock
);
713 atomic_inc(&flow_cache_genid
);
714 xfrm_policy_kill(ret
);
718 EXPORT_SYMBOL(xfrm_policy_byid
);
720 #ifdef CONFIG_SECURITY_NETWORK_XFRM
722 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
726 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
727 struct xfrm_policy
*pol
;
728 struct hlist_node
*entry
;
731 hlist_for_each_entry(pol
, entry
,
732 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
733 if (pol
->type
!= type
)
735 err
= security_xfrm_policy_delete(pol
->security
);
737 xfrm_audit_policy_delete(pol
, 0,
738 audit_info
->loginuid
,
739 audit_info
->sessionid
,
744 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
745 hlist_for_each_entry(pol
, entry
,
746 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
748 if (pol
->type
!= type
)
750 err
= security_xfrm_policy_delete(
753 xfrm_audit_policy_delete(pol
, 0,
754 audit_info
->loginuid
,
755 audit_info
->sessionid
,
766 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
772 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
776 write_lock_bh(&xfrm_policy_lock
);
778 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
782 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
783 struct xfrm_policy
*pol
;
784 struct hlist_node
*entry
;
788 hlist_for_each_entry(pol
, entry
,
789 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
790 if (pol
->type
!= type
)
792 __xfrm_policy_unlink(pol
, dir
);
793 write_unlock_bh(&xfrm_policy_lock
);
795 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
796 audit_info
->sessionid
,
799 xfrm_policy_kill(pol
);
801 write_lock_bh(&xfrm_policy_lock
);
805 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
807 hlist_for_each_entry(pol
, entry
,
808 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
810 if (pol
->type
!= type
)
812 __xfrm_policy_unlink(pol
, dir
);
813 write_unlock_bh(&xfrm_policy_lock
);
815 xfrm_audit_policy_delete(pol
, 1,
816 audit_info
->loginuid
,
817 audit_info
->sessionid
,
819 xfrm_policy_kill(pol
);
821 write_lock_bh(&xfrm_policy_lock
);
827 atomic_inc(&flow_cache_genid
);
829 write_unlock_bh(&xfrm_policy_lock
);
832 EXPORT_SYMBOL(xfrm_policy_flush
);
834 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
835 int (*func
)(struct xfrm_policy
*, int, int, void*),
838 struct xfrm_policy
*pol
;
839 struct xfrm_policy_walk_entry
*x
;
842 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
843 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
846 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
849 write_lock_bh(&xfrm_policy_lock
);
850 if (list_empty(&walk
->walk
.all
))
851 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
853 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
854 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
857 pol
= container_of(x
, struct xfrm_policy
, walk
);
858 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
859 walk
->type
!= pol
->type
)
861 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
864 list_move_tail(&walk
->walk
.all
, &x
->all
);
869 if (walk
->seq
== 0) {
873 list_del_init(&walk
->walk
.all
);
875 write_unlock_bh(&xfrm_policy_lock
);
878 EXPORT_SYMBOL(xfrm_policy_walk
);
880 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
882 INIT_LIST_HEAD(&walk
->walk
.all
);
887 EXPORT_SYMBOL(xfrm_policy_walk_init
);
889 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
891 if (list_empty(&walk
->walk
.all
))
894 write_lock_bh(&xfrm_policy_lock
);
895 list_del(&walk
->walk
.all
);
896 write_unlock_bh(&xfrm_policy_lock
);
898 EXPORT_SYMBOL(xfrm_policy_walk_done
);
901 * Find policy to apply to this flow.
903 * Returns 0 if policy found, else an -errno.
905 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
906 u8 type
, u16 family
, int dir
)
908 struct xfrm_selector
*sel
= &pol
->selector
;
909 int match
, ret
= -ESRCH
;
911 if (pol
->family
!= family
||
915 match
= xfrm_selector_match(sel
, fl
, family
);
917 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
923 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
928 struct xfrm_policy
*pol
, *ret
;
929 xfrm_address_t
*daddr
, *saddr
;
930 struct hlist_node
*entry
;
931 struct hlist_head
*chain
;
934 daddr
= xfrm_flowi_daddr(fl
, family
);
935 saddr
= xfrm_flowi_saddr(fl
, family
);
936 if (unlikely(!daddr
|| !saddr
))
939 read_lock_bh(&xfrm_policy_lock
);
940 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
942 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
943 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
953 priority
= ret
->priority
;
957 chain
= &net
->xfrm
.policy_inexact
[dir
];
958 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
959 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
967 } else if (pol
->priority
< priority
) {
975 read_unlock_bh(&xfrm_policy_lock
);
980 static int xfrm_policy_lookup(struct net
*net
, struct flowi
*fl
, u16 family
,
981 u8 dir
, void **objp
, atomic_t
**obj_refp
)
983 struct xfrm_policy
*pol
;
986 #ifdef CONFIG_XFRM_SUB_POLICY
987 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
995 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1000 #ifdef CONFIG_XFRM_SUB_POLICY
1003 if ((*objp
= (void *) pol
) != NULL
)
1004 *obj_refp
= &pol
->refcnt
;
1008 static inline int policy_to_flow_dir(int dir
)
1010 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1011 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1012 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1016 case XFRM_POLICY_IN
:
1018 case XFRM_POLICY_OUT
:
1019 return FLOW_DIR_OUT
;
1020 case XFRM_POLICY_FWD
:
1021 return FLOW_DIR_FWD
;
1025 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1027 struct xfrm_policy
*pol
;
1029 read_lock_bh(&xfrm_policy_lock
);
1030 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1031 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1036 err
= security_xfrm_policy_lookup(pol
->security
,
1038 policy_to_flow_dir(dir
));
1041 else if (err
== -ESRCH
)
1048 read_unlock_bh(&xfrm_policy_lock
);
1052 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1054 struct net
*net
= xp_net(pol
);
1055 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1058 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1059 hlist_add_head(&pol
->bydst
, chain
);
1060 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1061 net
->xfrm
.policy_count
[dir
]++;
1064 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1065 schedule_work(&net
->xfrm
.policy_hash_work
);
1068 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1071 struct net
*net
= xp_net(pol
);
1073 if (hlist_unhashed(&pol
->bydst
))
1076 hlist_del(&pol
->bydst
);
1077 hlist_del(&pol
->byidx
);
1078 list_del(&pol
->walk
.all
);
1079 net
->xfrm
.policy_count
[dir
]--;
1084 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1086 write_lock_bh(&xfrm_policy_lock
);
1087 pol
= __xfrm_policy_unlink(pol
, dir
);
1088 write_unlock_bh(&xfrm_policy_lock
);
1090 if (dir
< XFRM_POLICY_MAX
)
1091 atomic_inc(&flow_cache_genid
);
1092 xfrm_policy_kill(pol
);
1097 EXPORT_SYMBOL(xfrm_policy_delete
);
1099 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1101 struct net
*net
= xp_net(pol
);
1102 struct xfrm_policy
*old_pol
;
1104 #ifdef CONFIG_XFRM_SUB_POLICY
1105 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1109 write_lock_bh(&xfrm_policy_lock
);
1110 old_pol
= sk
->sk_policy
[dir
];
1111 sk
->sk_policy
[dir
] = pol
;
1113 pol
->curlft
.add_time
= get_seconds();
1114 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1115 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1118 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1119 write_unlock_bh(&xfrm_policy_lock
);
1122 xfrm_policy_kill(old_pol
);
1127 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1129 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1132 newp
->selector
= old
->selector
;
1133 if (security_xfrm_policy_clone(old
->security
,
1136 return NULL
; /* ENOMEM */
1138 newp
->lft
= old
->lft
;
1139 newp
->curlft
= old
->curlft
;
1140 newp
->action
= old
->action
;
1141 newp
->flags
= old
->flags
;
1142 newp
->xfrm_nr
= old
->xfrm_nr
;
1143 newp
->index
= old
->index
;
1144 newp
->type
= old
->type
;
1145 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1146 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1147 write_lock_bh(&xfrm_policy_lock
);
1148 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1149 write_unlock_bh(&xfrm_policy_lock
);
1155 int __xfrm_sk_clone_policy(struct sock
*sk
)
1157 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1158 *p1
= sk
->sk_policy
[1];
1160 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1161 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1163 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1169 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1170 unsigned short family
)
1173 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1175 if (unlikely(afinfo
== NULL
))
1177 err
= afinfo
->get_saddr(net
, local
, remote
);
1178 xfrm_policy_put_afinfo(afinfo
);
1182 /* Resolve list of templates for the flow, given policy. */
1185 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1186 struct xfrm_state
**xfrm
,
1187 unsigned short family
)
1189 struct net
*net
= xp_net(policy
);
1192 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1193 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1196 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1197 struct xfrm_state
*x
;
1198 xfrm_address_t
*remote
= daddr
;
1199 xfrm_address_t
*local
= saddr
;
1200 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1202 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1203 tmpl
->mode
== XFRM_MODE_BEET
) {
1204 remote
= &tmpl
->id
.daddr
;
1205 local
= &tmpl
->saddr
;
1206 family
= tmpl
->encap_family
;
1207 if (xfrm_addr_any(local
, family
)) {
1208 error
= xfrm_get_saddr(net
, &tmp
, remote
, family
);
1215 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1217 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1224 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1228 else if (error
== -ESRCH
)
1231 if (!tmpl
->optional
)
1237 for (nx
--; nx
>=0; nx
--)
1238 xfrm_state_put(xfrm
[nx
]);
1243 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1244 struct xfrm_state
**xfrm
,
1245 unsigned short family
)
1247 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1248 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1254 for (i
= 0; i
< npols
; i
++) {
1255 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1260 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1268 /* found states are sorted for outbound processing */
1270 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1275 for (cnx
--; cnx
>=0; cnx
--)
1276 xfrm_state_put(tpp
[cnx
]);
1281 /* Check that the bundle accepts the flow and its components are
1285 static struct dst_entry
*
1286 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1288 struct dst_entry
*x
;
1289 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1290 if (unlikely(afinfo
== NULL
))
1291 return ERR_PTR(-EINVAL
);
1292 x
= afinfo
->find_bundle(fl
, policy
);
1293 xfrm_policy_put_afinfo(afinfo
);
1297 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1299 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1305 tos
= afinfo
->get_tos(fl
);
1307 xfrm_policy_put_afinfo(afinfo
);
1312 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1314 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1315 struct dst_ops
*dst_ops
;
1316 struct xfrm_dst
*xdst
;
1319 return ERR_PTR(-EINVAL
);
1323 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1325 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1327 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1333 xdst
= dst_alloc(dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1335 xfrm_policy_put_afinfo(afinfo
);
1340 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1343 struct xfrm_policy_afinfo
*afinfo
=
1344 xfrm_policy_get_afinfo(dst
->ops
->family
);
1350 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1352 xfrm_policy_put_afinfo(afinfo
);
1357 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1359 struct xfrm_policy_afinfo
*afinfo
=
1360 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1366 err
= afinfo
->fill_dst(xdst
, dev
);
1368 xfrm_policy_put_afinfo(afinfo
);
1373 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1374 * all the metrics... Shortly, bundle a bundle.
1377 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1378 struct xfrm_state
**xfrm
, int nx
,
1380 struct dst_entry
*dst
)
1382 struct net
*net
= xp_net(policy
);
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(net
, 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 neighbour 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 net
*net
, 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(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1572 /* To accelerate a bit... */
1573 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1574 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1577 policy
= flow_cache_lookup(net
, fl
, dst_orig
->ops
->family
,
1578 dir
, xfrm_policy_lookup
);
1579 err
= PTR_ERR(policy
);
1580 if (IS_ERR(policy
)) {
1581 XFRM_INC_STATS(net
, 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(net
, 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(net
, 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(net
,
1635 XFRM_POLICY_TYPE_MAIN
,
1639 if (IS_ERR(pols
[1])) {
1640 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1641 err
= PTR_ERR(pols
[1]);
1644 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1645 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
1650 xfrm_nr
+= pols
[1]->xfrm_nr
;
1655 * Because neither flowi nor bundle information knows about
1656 * transformation template size. On more than one policy usage
1657 * we can realize whether all of them is bypass or not after
1658 * they are searched. See above not-transformed bypass
1659 * is surrounded by non-sub policy configuration, too.
1662 /* Flow passes not transformed. */
1663 xfrm_pols_put(pols
, npols
);
1668 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1670 if (unlikely(nx
<0)) {
1672 if (err
== -EAGAIN
&& net
->xfrm
.sysctl_larval_drop
) {
1673 /* EREMOTE tells the caller to generate
1674 * a one-shot blackhole route.
1676 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1677 xfrm_pol_put(policy
);
1680 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1681 DECLARE_WAITQUEUE(wait
, current
);
1683 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1684 set_current_state(TASK_INTERRUPTIBLE
);
1686 set_current_state(TASK_RUNNING
);
1687 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1689 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1691 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1692 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1696 if (nx
== -EAGAIN
||
1697 genid
!= atomic_read(&flow_cache_genid
)) {
1698 xfrm_pols_put(pols
, npols
);
1704 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1709 /* Flow passes not transformed. */
1710 xfrm_pols_put(pols
, npols
);
1714 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1717 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1721 for (pi
= 0; pi
< npols
; pi
++) {
1722 read_lock_bh(&pols
[pi
]->lock
);
1723 pol_dead
|= pols
[pi
]->walk
.dead
;
1724 read_unlock_bh(&pols
[pi
]->lock
);
1727 write_lock_bh(&policy
->lock
);
1728 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1729 /* Wow! While we worked on resolving, this
1730 * policy has gone. Retry. It is not paranoia,
1731 * we just cannot enlist new bundle to dead object.
1732 * We can't enlist stable bundles either.
1734 write_unlock_bh(&policy
->lock
);
1738 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLDEAD
);
1740 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1741 err
= -EHOSTUNREACH
;
1746 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1748 err
= xfrm_dst_update_origin(dst
, fl
);
1749 if (unlikely(err
)) {
1750 write_unlock_bh(&policy
->lock
);
1752 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1756 dst
->next
= policy
->bundles
;
1757 policy
->bundles
= dst
;
1759 write_unlock_bh(&policy
->lock
);
1762 dst_release(dst_orig
);
1763 xfrm_pols_put(pols
, npols
);
1767 xfrm_pols_put(pols
, npols
);
1769 dst_release(dst_orig
);
1775 if (flags
& XFRM_LOOKUP_ICMP
)
1779 EXPORT_SYMBOL(__xfrm_lookup
);
1781 int xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1782 struct sock
*sk
, int flags
)
1784 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1786 if (err
== -EREMOTE
) {
1787 dst_release(*dst_p
);
1794 EXPORT_SYMBOL(xfrm_lookup
);
1797 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1799 struct xfrm_state
*x
;
1801 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1803 x
= skb
->sp
->xvec
[idx
];
1804 if (!x
->type
->reject
)
1806 return x
->type
->reject(x
, skb
, fl
);
1809 /* When skb is transformed back to its "native" form, we have to
1810 * check policy restrictions. At the moment we make this in maximally
1811 * stupid way. Shame on me. :-) Of course, connected sockets must
1812 * have policy cached at them.
1816 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1817 unsigned short family
)
1819 if (xfrm_state_kern(x
))
1820 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1821 return x
->id
.proto
== tmpl
->id
.proto
&&
1822 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1823 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1824 x
->props
.mode
== tmpl
->mode
&&
1825 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1826 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1827 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1828 xfrm_state_addr_cmp(tmpl
, x
, family
));
1832 * 0 or more than 0 is returned when validation is succeeded (either bypass
1833 * because of optional transport mode, or next index of the mathced secpath
1834 * state with the template.
1835 * -1 is returned when no matching template is found.
1836 * Otherwise "-2 - errored_index" is returned.
1839 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1840 unsigned short family
)
1844 if (tmpl
->optional
) {
1845 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1849 for (; idx
< sp
->len
; idx
++) {
1850 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1852 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1861 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1862 unsigned int family
, int reverse
)
1864 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1867 if (unlikely(afinfo
== NULL
))
1868 return -EAFNOSUPPORT
;
1870 afinfo
->decode_session(skb
, fl
, reverse
);
1871 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1872 xfrm_policy_put_afinfo(afinfo
);
1875 EXPORT_SYMBOL(__xfrm_decode_session
);
1877 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1879 for (; k
< sp
->len
; k
++) {
1880 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1889 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1890 unsigned short family
)
1892 struct net
*net
= dev_net(skb
->dev
);
1893 struct xfrm_policy
*pol
;
1894 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1903 reverse
= dir
& ~XFRM_POLICY_MASK
;
1904 dir
&= XFRM_POLICY_MASK
;
1905 fl_dir
= policy_to_flow_dir(dir
);
1907 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1908 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
1912 nf_nat_decode_session(skb
, &fl
, family
);
1914 /* First, check used SA against their selectors. */
1918 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1919 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1920 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1921 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
1928 if (sk
&& sk
->sk_policy
[dir
]) {
1929 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1931 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1937 pol
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
1938 xfrm_policy_lookup
);
1941 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1946 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1947 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1948 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
1954 pol
->curlft
.use_time
= get_seconds();
1958 #ifdef CONFIG_XFRM_SUB_POLICY
1959 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1960 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
1964 if (IS_ERR(pols
[1])) {
1965 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1968 pols
[1]->curlft
.use_time
= get_seconds();
1974 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1975 struct sec_path
*sp
;
1976 static struct sec_path dummy
;
1977 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1978 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1979 struct xfrm_tmpl
**tpp
= tp
;
1983 if ((sp
= skb
->sp
) == NULL
)
1986 for (pi
= 0; pi
< npols
; pi
++) {
1987 if (pols
[pi
] != pol
&&
1988 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
1989 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
1992 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1993 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
1996 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
1997 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2001 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2005 /* For each tunnel xfrm, find the first matching tmpl.
2006 * For each tmpl before that, find corresponding xfrm.
2007 * Order is _important_. Later we will implement
2008 * some barriers, but at the moment barriers
2009 * are implied between each two transformations.
2011 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2012 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2015 /* "-2 - errored_index" returned */
2017 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2022 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2023 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2027 xfrm_pols_put(pols
, npols
);
2030 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2033 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2035 xfrm_pols_put(pols
, npols
);
2038 EXPORT_SYMBOL(__xfrm_policy_check
);
2040 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2042 struct net
*net
= dev_net(skb
->dev
);
2044 struct dst_entry
*dst
;
2047 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2048 /* XXX: we should have something like FWDHDRERROR here. */
2049 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
2055 res
= xfrm_lookup(net
, &dst
, &fl
, NULL
, 0) == 0;
2056 skb_dst_set(skb
, dst
);
2059 EXPORT_SYMBOL(__xfrm_route_forward
);
2061 /* Optimize later using cookies and generation ids. */
2063 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2065 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2066 * to "-1" to force all XFRM destinations to get validated by
2067 * dst_ops->check on every use. We do this because when a
2068 * normal route referenced by an XFRM dst is obsoleted we do
2069 * not go looking around for all parent referencing XFRM dsts
2070 * so that we can invalidate them. It is just too much work.
2071 * Instead we make the checks here on every use. For example:
2073 * XFRM dst A --> IPv4 dst X
2075 * X is the "xdst->route" of A (X is also the "dst->path" of A
2076 * in this example). If X is marked obsolete, "A" will not
2077 * notice. That's what we are validating here via the
2078 * stale_bundle() check.
2080 * When a policy's bundle is pruned, we dst_free() the XFRM
2081 * dst which causes it's ->obsolete field to be set to a
2082 * positive non-zero integer. If an XFRM dst has been pruned
2083 * like this, we want to force a new route lookup.
2085 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2091 static int stale_bundle(struct dst_entry
*dst
)
2093 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2096 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2098 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2099 dst
->dev
= dev_net(dev
)->loopback_dev
;
2104 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2106 static void xfrm_link_failure(struct sk_buff
*skb
)
2108 /* Impossible. Such dst must be popped before reaches point of failure. */
2112 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2115 if (dst
->obsolete
) {
2123 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
2125 struct dst_entry
*dst
, **dstp
;
2127 write_lock(&pol
->lock
);
2128 dstp
= &pol
->bundles
;
2129 while ((dst
=*dstp
) != NULL
) {
2132 dst
->next
= *gc_list_p
;
2138 write_unlock(&pol
->lock
);
2141 static void xfrm_prune_bundles(struct net
*net
, int (*func
)(struct dst_entry
*))
2143 struct dst_entry
*gc_list
= NULL
;
2146 read_lock_bh(&xfrm_policy_lock
);
2147 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2148 struct xfrm_policy
*pol
;
2149 struct hlist_node
*entry
;
2150 struct hlist_head
*table
;
2153 hlist_for_each_entry(pol
, entry
,
2154 &net
->xfrm
.policy_inexact
[dir
], bydst
)
2155 prune_one_bundle(pol
, func
, &gc_list
);
2157 table
= net
->xfrm
.policy_bydst
[dir
].table
;
2158 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
2159 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
2160 prune_one_bundle(pol
, func
, &gc_list
);
2163 read_unlock_bh(&xfrm_policy_lock
);
2166 struct dst_entry
*dst
= gc_list
;
2167 gc_list
= dst
->next
;
2172 static int unused_bundle(struct dst_entry
*dst
)
2174 return !atomic_read(&dst
->__refcnt
);
2177 static void __xfrm_garbage_collect(struct net
*net
)
2179 xfrm_prune_bundles(net
, unused_bundle
);
2182 static int xfrm_flush_bundles(struct net
*net
)
2184 xfrm_prune_bundles(net
, stale_bundle
);
2188 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2191 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2192 u32 pmtu
, route_mtu_cached
;
2194 pmtu
= dst_mtu(dst
->child
);
2195 xdst
->child_mtu_cached
= pmtu
;
2197 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2199 route_mtu_cached
= dst_mtu(xdst
->route
);
2200 xdst
->route_mtu_cached
= route_mtu_cached
;
2202 if (pmtu
> route_mtu_cached
)
2203 pmtu
= route_mtu_cached
;
2205 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
2206 } while ((dst
= dst
->next
));
2209 /* Check that the bundle accepts the flow and its components are
2213 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2214 struct flowi
*fl
, int family
, int strict
)
2216 struct dst_entry
*dst
= &first
->u
.dst
;
2217 struct xfrm_dst
*last
;
2220 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2221 (dst
->dev
&& !netif_running(dst
->dev
)))
2223 #ifdef CONFIG_XFRM_SUB_POLICY
2225 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2227 if (first
->partner
&&
2228 !xfrm_selector_match(first
->partner
, fl
, family
))
2236 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2238 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2241 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2243 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2245 if (xdst
->genid
!= dst
->xfrm
->genid
)
2249 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2250 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2253 mtu
= dst_mtu(dst
->child
);
2254 if (xdst
->child_mtu_cached
!= mtu
) {
2256 xdst
->child_mtu_cached
= mtu
;
2259 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2261 mtu
= dst_mtu(xdst
->route
);
2262 if (xdst
->route_mtu_cached
!= mtu
) {
2264 xdst
->route_mtu_cached
= mtu
;
2268 } while (dst
->xfrm
);
2273 mtu
= last
->child_mtu_cached
;
2277 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2278 if (mtu
> last
->route_mtu_cached
)
2279 mtu
= last
->route_mtu_cached
;
2280 dst
->metrics
[RTAX_MTU
-1] = mtu
;
2285 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2286 last
->child_mtu_cached
= mtu
;
2292 EXPORT_SYMBOL(xfrm_bundle_ok
);
2294 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2298 if (unlikely(afinfo
== NULL
))
2300 if (unlikely(afinfo
->family
>= NPROTO
))
2301 return -EAFNOSUPPORT
;
2302 write_lock_bh(&xfrm_policy_afinfo_lock
);
2303 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2306 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2307 if (likely(dst_ops
->kmem_cachep
== NULL
))
2308 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2309 if (likely(dst_ops
->check
== NULL
))
2310 dst_ops
->check
= xfrm_dst_check
;
2311 if (likely(dst_ops
->negative_advice
== NULL
))
2312 dst_ops
->negative_advice
= xfrm_negative_advice
;
2313 if (likely(dst_ops
->link_failure
== NULL
))
2314 dst_ops
->link_failure
= xfrm_link_failure
;
2315 if (likely(afinfo
->garbage_collect
== NULL
))
2316 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2317 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2319 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2323 struct dst_ops
*xfrm_dst_ops
;
2325 switch (afinfo
->family
) {
2327 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2329 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2331 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2337 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2343 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2345 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2348 if (unlikely(afinfo
== NULL
))
2350 if (unlikely(afinfo
->family
>= NPROTO
))
2351 return -EAFNOSUPPORT
;
2352 write_lock_bh(&xfrm_policy_afinfo_lock
);
2353 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2354 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2357 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2358 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2359 dst_ops
->kmem_cachep
= NULL
;
2360 dst_ops
->check
= NULL
;
2361 dst_ops
->negative_advice
= NULL
;
2362 dst_ops
->link_failure
= NULL
;
2363 afinfo
->garbage_collect
= NULL
;
2366 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2369 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2371 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2373 struct xfrm_policy_afinfo
*afinfo
;
2375 read_lock_bh(&xfrm_policy_afinfo_lock
);
2376 afinfo
= xfrm_policy_afinfo
[AF_INET
];
2378 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2379 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2380 afinfo
= xfrm_policy_afinfo
[AF_INET6
];
2382 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2384 read_unlock_bh(&xfrm_policy_afinfo_lock
);
2387 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2389 struct xfrm_policy_afinfo
*afinfo
;
2390 if (unlikely(family
>= NPROTO
))
2392 read_lock(&xfrm_policy_afinfo_lock
);
2393 afinfo
= xfrm_policy_afinfo
[family
];
2394 if (unlikely(!afinfo
))
2395 read_unlock(&xfrm_policy_afinfo_lock
);
2399 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2401 read_unlock(&xfrm_policy_afinfo_lock
);
2404 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2406 struct net_device
*dev
= ptr
;
2410 xfrm_flush_bundles(dev_net(dev
));
2415 static struct notifier_block xfrm_dev_notifier
= {
2416 .notifier_call
= xfrm_dev_event
,
2419 #ifdef CONFIG_XFRM_STATISTICS
2420 static int __net_init
xfrm_statistics_init(struct net
*net
)
2424 if (snmp_mib_init((void **)net
->mib
.xfrm_statistics
,
2425 sizeof(struct linux_xfrm_mib
)) < 0)
2427 rv
= xfrm_proc_init(net
);
2429 snmp_mib_free((void **)net
->mib
.xfrm_statistics
);
2433 static void xfrm_statistics_fini(struct net
*net
)
2435 xfrm_proc_fini(net
);
2436 snmp_mib_free((void **)net
->mib
.xfrm_statistics
);
2439 static int __net_init
xfrm_statistics_init(struct net
*net
)
2444 static void xfrm_statistics_fini(struct net
*net
)
2449 static int __net_init
xfrm_policy_init(struct net
*net
)
2451 unsigned int hmask
, sz
;
2454 if (net_eq(net
, &init_net
))
2455 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2456 sizeof(struct xfrm_dst
),
2457 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2461 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2463 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2464 if (!net
->xfrm
.policy_byidx
)
2466 net
->xfrm
.policy_idx_hmask
= hmask
;
2468 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2469 struct xfrm_policy_hash
*htab
;
2471 net
->xfrm
.policy_count
[dir
] = 0;
2472 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2474 htab
= &net
->xfrm
.policy_bydst
[dir
];
2475 htab
->table
= xfrm_hash_alloc(sz
);
2478 htab
->hmask
= hmask
;
2481 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2482 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2483 if (net_eq(net
, &init_net
))
2484 register_netdevice_notifier(&xfrm_dev_notifier
);
2488 for (dir
--; dir
>= 0; dir
--) {
2489 struct xfrm_policy_hash
*htab
;
2491 htab
= &net
->xfrm
.policy_bydst
[dir
];
2492 xfrm_hash_free(htab
->table
, sz
);
2494 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2499 static void xfrm_policy_fini(struct net
*net
)
2501 struct xfrm_audit audit_info
;
2505 flush_work(&net
->xfrm
.policy_hash_work
);
2506 #ifdef CONFIG_XFRM_SUB_POLICY
2507 audit_info
.loginuid
= -1;
2508 audit_info
.sessionid
= -1;
2509 audit_info
.secid
= 0;
2510 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2512 audit_info
.loginuid
= -1;
2513 audit_info
.sessionid
= -1;
2514 audit_info
.secid
= 0;
2515 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2516 flush_work(&xfrm_policy_gc_work
);
2518 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2520 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2521 struct xfrm_policy_hash
*htab
;
2523 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2525 htab
= &net
->xfrm
.policy_bydst
[dir
];
2526 sz
= (htab
->hmask
+ 1);
2527 WARN_ON(!hlist_empty(htab
->table
));
2528 xfrm_hash_free(htab
->table
, sz
);
2531 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2532 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2533 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2536 static int __net_init
xfrm_net_init(struct net
*net
)
2540 rv
= xfrm_statistics_init(net
);
2542 goto out_statistics
;
2543 rv
= xfrm_state_init(net
);
2546 rv
= xfrm_policy_init(net
);
2549 xfrm_dst_ops_init(net
);
2550 rv
= xfrm_sysctl_init(net
);
2556 xfrm_policy_fini(net
);
2558 xfrm_state_fini(net
);
2560 xfrm_statistics_fini(net
);
2565 static void __net_exit
xfrm_net_exit(struct net
*net
)
2567 xfrm_sysctl_fini(net
);
2568 xfrm_policy_fini(net
);
2569 xfrm_state_fini(net
);
2570 xfrm_statistics_fini(net
);
2573 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2574 .init
= xfrm_net_init
,
2575 .exit
= xfrm_net_exit
,
2578 void __init
xfrm_init(void)
2580 register_pernet_subsys(&xfrm_net_ops
);
2584 #ifdef CONFIG_AUDITSYSCALL
2585 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2586 struct audit_buffer
*audit_buf
)
2588 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2589 struct xfrm_selector
*sel
= &xp
->selector
;
2592 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2593 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2595 switch(sel
->family
) {
2597 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2598 if (sel
->prefixlen_s
!= 32)
2599 audit_log_format(audit_buf
, " src_prefixlen=%d",
2601 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2602 if (sel
->prefixlen_d
!= 32)
2603 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2607 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2608 if (sel
->prefixlen_s
!= 128)
2609 audit_log_format(audit_buf
, " src_prefixlen=%d",
2611 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2612 if (sel
->prefixlen_d
!= 128)
2613 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2619 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2620 uid_t auid
, u32 sessionid
, u32 secid
)
2622 struct audit_buffer
*audit_buf
;
2624 audit_buf
= xfrm_audit_start("SPD-add");
2625 if (audit_buf
== NULL
)
2627 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2628 audit_log_format(audit_buf
, " res=%u", result
);
2629 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2630 audit_log_end(audit_buf
);
2632 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2634 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2635 uid_t auid
, u32 sessionid
, u32 secid
)
2637 struct audit_buffer
*audit_buf
;
2639 audit_buf
= xfrm_audit_start("SPD-delete");
2640 if (audit_buf
== NULL
)
2642 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2643 audit_log_format(audit_buf
, " res=%u", result
);
2644 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2645 audit_log_end(audit_buf
);
2647 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2650 #ifdef CONFIG_XFRM_MIGRATE
2651 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2652 struct xfrm_selector
*sel_tgt
)
2654 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2655 if (sel_tgt
->family
== sel_cmp
->family
&&
2656 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2657 sel_cmp
->family
) == 0 &&
2658 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2659 sel_cmp
->family
) == 0 &&
2660 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2661 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2665 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2672 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2675 struct xfrm_policy
*pol
, *ret
= NULL
;
2676 struct hlist_node
*entry
;
2677 struct hlist_head
*chain
;
2680 read_lock_bh(&xfrm_policy_lock
);
2681 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2682 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2683 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2684 pol
->type
== type
) {
2686 priority
= ret
->priority
;
2690 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2691 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2692 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2693 pol
->type
== type
&&
2694 pol
->priority
< priority
) {
2703 read_unlock_bh(&xfrm_policy_lock
);
2708 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2712 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2713 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2715 case XFRM_MODE_TUNNEL
:
2716 case XFRM_MODE_BEET
:
2717 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2718 m
->old_family
) == 0 &&
2719 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2720 m
->old_family
) == 0) {
2724 case XFRM_MODE_TRANSPORT
:
2725 /* in case of transport mode, template does not store
2726 any IP addresses, hence we just compare mode and
2737 /* update endpoint address(es) of template(s) */
2738 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2739 struct xfrm_migrate
*m
, int num_migrate
)
2741 struct xfrm_migrate
*mp
;
2742 struct dst_entry
*dst
;
2745 write_lock_bh(&pol
->lock
);
2746 if (unlikely(pol
->walk
.dead
)) {
2747 /* target policy has been deleted */
2748 write_unlock_bh(&pol
->lock
);
2752 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2753 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2754 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2757 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2758 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2760 /* update endpoints */
2761 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2762 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2763 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2764 sizeof(pol
->xfrm_vec
[i
].saddr
));
2765 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2767 while ((dst
= pol
->bundles
) != NULL
) {
2768 pol
->bundles
= dst
->next
;
2774 write_unlock_bh(&pol
->lock
);
2782 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2786 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2789 for (i
= 0; i
< num_migrate
; i
++) {
2790 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2791 m
[i
].old_family
) == 0) &&
2792 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2793 m
[i
].old_family
) == 0))
2795 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2796 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2799 /* check if there is any duplicated entry */
2800 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2801 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2802 sizeof(m
[i
].old_daddr
)) &&
2803 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2804 sizeof(m
[i
].old_saddr
)) &&
2805 m
[i
].proto
== m
[j
].proto
&&
2806 m
[i
].mode
== m
[j
].mode
&&
2807 m
[i
].reqid
== m
[j
].reqid
&&
2808 m
[i
].old_family
== m
[j
].old_family
)
2816 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2817 struct xfrm_migrate
*m
, int num_migrate
,
2818 struct xfrm_kmaddress
*k
)
2820 int i
, err
, nx_cur
= 0, nx_new
= 0;
2821 struct xfrm_policy
*pol
= NULL
;
2822 struct xfrm_state
*x
, *xc
;
2823 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2824 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2825 struct xfrm_migrate
*mp
;
2827 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2830 /* Stage 1 - find policy */
2831 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2836 /* Stage 2 - find and update state(s) */
2837 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2838 if ((x
= xfrm_migrate_state_find(mp
))) {
2841 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2851 /* Stage 3 - update policy */
2852 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2855 /* Stage 4 - delete old state(s) */
2857 xfrm_states_put(x_cur
, nx_cur
);
2858 xfrm_states_delete(x_cur
, nx_cur
);
2861 /* Stage 5 - announce */
2862 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2874 xfrm_states_put(x_cur
, nx_cur
);
2876 xfrm_states_delete(x_new
, nx_new
);
2880 EXPORT_SYMBOL(xfrm_migrate
);