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
;
559 u32 mark
= policy
->mark
.v
& policy
->mark
.m
;
561 write_lock_bh(&xfrm_policy_lock
);
562 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
565 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
566 if (pol
->type
== policy
->type
&&
567 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
568 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
569 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
572 write_unlock_bh(&xfrm_policy_lock
);
576 if (policy
->priority
> pol
->priority
)
578 } else if (policy
->priority
>= pol
->priority
) {
579 newpos
= &pol
->bydst
;
586 hlist_add_after(newpos
, &policy
->bydst
);
588 hlist_add_head(&policy
->bydst
, chain
);
589 xfrm_pol_hold(policy
);
590 net
->xfrm
.policy_count
[dir
]++;
591 atomic_inc(&flow_cache_genid
);
593 __xfrm_policy_unlink(delpol
, dir
);
594 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
595 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
596 policy
->curlft
.add_time
= get_seconds();
597 policy
->curlft
.use_time
= 0;
598 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
599 xfrm_pol_hold(policy
);
600 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
601 write_unlock_bh(&xfrm_policy_lock
);
604 xfrm_policy_kill(delpol
);
605 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
606 schedule_work(&net
->xfrm
.policy_hash_work
);
608 read_lock_bh(&xfrm_policy_lock
);
610 entry
= &policy
->bydst
;
611 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
612 struct dst_entry
*dst
;
614 write_lock(&policy
->lock
);
615 dst
= policy
->bundles
;
617 struct dst_entry
*tail
= dst
;
620 tail
->next
= gc_list
;
623 policy
->bundles
= NULL
;
625 write_unlock(&policy
->lock
);
627 read_unlock_bh(&xfrm_policy_lock
);
630 struct dst_entry
*dst
= gc_list
;
638 EXPORT_SYMBOL(xfrm_policy_insert
);
640 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u32 mark
, u8 type
,
641 int dir
, struct xfrm_selector
*sel
,
642 struct xfrm_sec_ctx
*ctx
, int delete,
645 struct xfrm_policy
*pol
, *ret
;
646 struct hlist_head
*chain
;
647 struct hlist_node
*entry
;
650 write_lock_bh(&xfrm_policy_lock
);
651 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
653 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
654 if (pol
->type
== type
&&
655 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
656 !selector_cmp(sel
, &pol
->selector
) &&
657 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
660 *err
= security_xfrm_policy_delete(
663 write_unlock_bh(&xfrm_policy_lock
);
666 __xfrm_policy_unlink(pol
, dir
);
672 write_unlock_bh(&xfrm_policy_lock
);
675 atomic_inc(&flow_cache_genid
);
676 xfrm_policy_kill(ret
);
680 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
682 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u32 mark
, u8 type
,
683 int dir
, u32 id
, int delete, int *err
)
685 struct xfrm_policy
*pol
, *ret
;
686 struct hlist_head
*chain
;
687 struct hlist_node
*entry
;
690 if (xfrm_policy_id2dir(id
) != dir
)
694 write_lock_bh(&xfrm_policy_lock
);
695 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
697 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
698 if (pol
->type
== type
&& pol
->index
== id
&&
699 (mark
& pol
->mark
.m
) == pol
->mark
.v
) {
702 *err
= security_xfrm_policy_delete(
705 write_unlock_bh(&xfrm_policy_lock
);
708 __xfrm_policy_unlink(pol
, dir
);
714 write_unlock_bh(&xfrm_policy_lock
);
717 atomic_inc(&flow_cache_genid
);
718 xfrm_policy_kill(ret
);
722 EXPORT_SYMBOL(xfrm_policy_byid
);
724 #ifdef CONFIG_SECURITY_NETWORK_XFRM
726 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
730 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
731 struct xfrm_policy
*pol
;
732 struct hlist_node
*entry
;
735 hlist_for_each_entry(pol
, entry
,
736 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
737 if (pol
->type
!= type
)
739 err
= security_xfrm_policy_delete(pol
->security
);
741 xfrm_audit_policy_delete(pol
, 0,
742 audit_info
->loginuid
,
743 audit_info
->sessionid
,
748 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
749 hlist_for_each_entry(pol
, entry
,
750 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
752 if (pol
->type
!= type
)
754 err
= security_xfrm_policy_delete(
757 xfrm_audit_policy_delete(pol
, 0,
758 audit_info
->loginuid
,
759 audit_info
->sessionid
,
770 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
776 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
778 int dir
, err
= 0, cnt
= 0;
779 struct xfrm_policy
*dp
;
781 write_lock_bh(&xfrm_policy_lock
);
783 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
787 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
788 struct xfrm_policy
*pol
;
789 struct hlist_node
*entry
;
793 hlist_for_each_entry(pol
, entry
,
794 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
795 if (pol
->type
!= type
)
797 dp
= __xfrm_policy_unlink(pol
, dir
);
798 write_unlock_bh(&xfrm_policy_lock
);
802 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
803 audit_info
->sessionid
,
806 xfrm_policy_kill(pol
);
808 write_lock_bh(&xfrm_policy_lock
);
812 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
814 hlist_for_each_entry(pol
, entry
,
815 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
817 if (pol
->type
!= type
)
819 dp
= __xfrm_policy_unlink(pol
, dir
);
820 write_unlock_bh(&xfrm_policy_lock
);
824 xfrm_audit_policy_delete(pol
, 1,
825 audit_info
->loginuid
,
826 audit_info
->sessionid
,
828 xfrm_policy_kill(pol
);
830 write_lock_bh(&xfrm_policy_lock
);
838 atomic_inc(&flow_cache_genid
);
840 write_unlock_bh(&xfrm_policy_lock
);
843 EXPORT_SYMBOL(xfrm_policy_flush
);
845 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
846 int (*func
)(struct xfrm_policy
*, int, int, void*),
849 struct xfrm_policy
*pol
;
850 struct xfrm_policy_walk_entry
*x
;
853 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
854 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
857 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
860 write_lock_bh(&xfrm_policy_lock
);
861 if (list_empty(&walk
->walk
.all
))
862 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
864 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
865 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
868 pol
= container_of(x
, struct xfrm_policy
, walk
);
869 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
870 walk
->type
!= pol
->type
)
872 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
875 list_move_tail(&walk
->walk
.all
, &x
->all
);
880 if (walk
->seq
== 0) {
884 list_del_init(&walk
->walk
.all
);
886 write_unlock_bh(&xfrm_policy_lock
);
889 EXPORT_SYMBOL(xfrm_policy_walk
);
891 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
893 INIT_LIST_HEAD(&walk
->walk
.all
);
898 EXPORT_SYMBOL(xfrm_policy_walk_init
);
900 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
902 if (list_empty(&walk
->walk
.all
))
905 write_lock_bh(&xfrm_policy_lock
);
906 list_del(&walk
->walk
.all
);
907 write_unlock_bh(&xfrm_policy_lock
);
909 EXPORT_SYMBOL(xfrm_policy_walk_done
);
912 * Find policy to apply to this flow.
914 * Returns 0 if policy found, else an -errno.
916 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
917 u8 type
, u16 family
, int dir
)
919 struct xfrm_selector
*sel
= &pol
->selector
;
920 int match
, ret
= -ESRCH
;
922 if (pol
->family
!= family
||
923 (fl
->mark
& pol
->mark
.m
) != pol
->mark
.v
||
927 match
= xfrm_selector_match(sel
, fl
, family
);
929 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
935 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
940 struct xfrm_policy
*pol
, *ret
;
941 xfrm_address_t
*daddr
, *saddr
;
942 struct hlist_node
*entry
;
943 struct hlist_head
*chain
;
946 daddr
= xfrm_flowi_daddr(fl
, family
);
947 saddr
= xfrm_flowi_saddr(fl
, family
);
948 if (unlikely(!daddr
|| !saddr
))
951 read_lock_bh(&xfrm_policy_lock
);
952 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
954 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
955 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
965 priority
= ret
->priority
;
969 chain
= &net
->xfrm
.policy_inexact
[dir
];
970 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
971 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
979 } else if (pol
->priority
< priority
) {
987 read_unlock_bh(&xfrm_policy_lock
);
992 static int xfrm_policy_lookup(struct net
*net
, struct flowi
*fl
, u16 family
,
993 u8 dir
, void **objp
, atomic_t
**obj_refp
)
995 struct xfrm_policy
*pol
;
998 #ifdef CONFIG_XFRM_SUB_POLICY
999 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1007 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1012 #ifdef CONFIG_XFRM_SUB_POLICY
1015 if ((*objp
= (void *) pol
) != NULL
)
1016 *obj_refp
= &pol
->refcnt
;
1020 static inline int policy_to_flow_dir(int dir
)
1022 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1023 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1024 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1028 case XFRM_POLICY_IN
:
1030 case XFRM_POLICY_OUT
:
1031 return FLOW_DIR_OUT
;
1032 case XFRM_POLICY_FWD
:
1033 return FLOW_DIR_FWD
;
1037 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1039 struct xfrm_policy
*pol
;
1041 read_lock_bh(&xfrm_policy_lock
);
1042 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1043 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1048 if ((sk
->sk_mark
& pol
->mark
.m
) != pol
->mark
.v
) {
1052 err
= security_xfrm_policy_lookup(pol
->security
,
1054 policy_to_flow_dir(dir
));
1057 else if (err
== -ESRCH
)
1065 read_unlock_bh(&xfrm_policy_lock
);
1069 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1071 struct net
*net
= xp_net(pol
);
1072 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1075 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1076 hlist_add_head(&pol
->bydst
, chain
);
1077 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1078 net
->xfrm
.policy_count
[dir
]++;
1081 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1082 schedule_work(&net
->xfrm
.policy_hash_work
);
1085 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1088 struct net
*net
= xp_net(pol
);
1090 if (hlist_unhashed(&pol
->bydst
))
1093 hlist_del(&pol
->bydst
);
1094 hlist_del(&pol
->byidx
);
1095 list_del(&pol
->walk
.all
);
1096 net
->xfrm
.policy_count
[dir
]--;
1101 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1103 write_lock_bh(&xfrm_policy_lock
);
1104 pol
= __xfrm_policy_unlink(pol
, dir
);
1105 write_unlock_bh(&xfrm_policy_lock
);
1107 if (dir
< XFRM_POLICY_MAX
)
1108 atomic_inc(&flow_cache_genid
);
1109 xfrm_policy_kill(pol
);
1114 EXPORT_SYMBOL(xfrm_policy_delete
);
1116 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1118 struct net
*net
= xp_net(pol
);
1119 struct xfrm_policy
*old_pol
;
1121 #ifdef CONFIG_XFRM_SUB_POLICY
1122 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1126 write_lock_bh(&xfrm_policy_lock
);
1127 old_pol
= sk
->sk_policy
[dir
];
1128 sk
->sk_policy
[dir
] = pol
;
1130 pol
->curlft
.add_time
= get_seconds();
1131 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1132 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1135 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1136 write_unlock_bh(&xfrm_policy_lock
);
1139 xfrm_policy_kill(old_pol
);
1144 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1146 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1149 newp
->selector
= old
->selector
;
1150 if (security_xfrm_policy_clone(old
->security
,
1153 return NULL
; /* ENOMEM */
1155 newp
->lft
= old
->lft
;
1156 newp
->curlft
= old
->curlft
;
1157 newp
->mark
= old
->mark
;
1158 newp
->action
= old
->action
;
1159 newp
->flags
= old
->flags
;
1160 newp
->xfrm_nr
= old
->xfrm_nr
;
1161 newp
->index
= old
->index
;
1162 newp
->type
= old
->type
;
1163 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1164 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1165 write_lock_bh(&xfrm_policy_lock
);
1166 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1167 write_unlock_bh(&xfrm_policy_lock
);
1173 int __xfrm_sk_clone_policy(struct sock
*sk
)
1175 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1176 *p1
= sk
->sk_policy
[1];
1178 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1179 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1181 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1187 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1188 unsigned short family
)
1191 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1193 if (unlikely(afinfo
== NULL
))
1195 err
= afinfo
->get_saddr(net
, local
, remote
);
1196 xfrm_policy_put_afinfo(afinfo
);
1200 /* Resolve list of templates for the flow, given policy. */
1203 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1204 struct xfrm_state
**xfrm
,
1205 unsigned short family
)
1207 struct net
*net
= xp_net(policy
);
1210 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1211 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1214 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1215 struct xfrm_state
*x
;
1216 xfrm_address_t
*remote
= daddr
;
1217 xfrm_address_t
*local
= saddr
;
1218 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1220 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1221 tmpl
->mode
== XFRM_MODE_BEET
) {
1222 remote
= &tmpl
->id
.daddr
;
1223 local
= &tmpl
->saddr
;
1224 family
= tmpl
->encap_family
;
1225 if (xfrm_addr_any(local
, family
)) {
1226 error
= xfrm_get_saddr(net
, &tmp
, remote
, family
);
1233 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1235 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1242 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1246 else if (error
== -ESRCH
)
1249 if (!tmpl
->optional
)
1255 for (nx
--; nx
>=0; nx
--)
1256 xfrm_state_put(xfrm
[nx
]);
1261 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1262 struct xfrm_state
**xfrm
,
1263 unsigned short family
)
1265 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1266 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1272 for (i
= 0; i
< npols
; i
++) {
1273 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1278 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1286 /* found states are sorted for outbound processing */
1288 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1293 for (cnx
--; cnx
>=0; cnx
--)
1294 xfrm_state_put(tpp
[cnx
]);
1299 /* Check that the bundle accepts the flow and its components are
1303 static struct dst_entry
*
1304 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1306 struct dst_entry
*x
;
1307 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1308 if (unlikely(afinfo
== NULL
))
1309 return ERR_PTR(-EINVAL
);
1310 x
= afinfo
->find_bundle(fl
, policy
);
1311 xfrm_policy_put_afinfo(afinfo
);
1315 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1317 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1323 tos
= afinfo
->get_tos(fl
);
1325 xfrm_policy_put_afinfo(afinfo
);
1330 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1332 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1333 struct dst_ops
*dst_ops
;
1334 struct xfrm_dst
*xdst
;
1337 return ERR_PTR(-EINVAL
);
1341 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1343 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1345 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1351 xdst
= dst_alloc(dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1353 xfrm_policy_put_afinfo(afinfo
);
1358 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1361 struct xfrm_policy_afinfo
*afinfo
=
1362 xfrm_policy_get_afinfo(dst
->ops
->family
);
1368 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1370 xfrm_policy_put_afinfo(afinfo
);
1375 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1377 struct xfrm_policy_afinfo
*afinfo
=
1378 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1384 err
= afinfo
->fill_dst(xdst
, dev
);
1386 xfrm_policy_put_afinfo(afinfo
);
1391 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1392 * all the metrics... Shortly, bundle a bundle.
1395 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1396 struct xfrm_state
**xfrm
, int nx
,
1398 struct dst_entry
*dst
)
1400 struct net
*net
= xp_net(policy
);
1401 unsigned long now
= jiffies
;
1402 struct net_device
*dev
;
1403 struct dst_entry
*dst_prev
= NULL
;
1404 struct dst_entry
*dst0
= NULL
;
1408 int nfheader_len
= 0;
1409 int trailer_len
= 0;
1411 int family
= policy
->selector
.family
;
1412 xfrm_address_t saddr
, daddr
;
1414 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1416 tos
= xfrm_get_tos(fl
, family
);
1423 for (; i
< nx
; i
++) {
1424 struct xfrm_dst
*xdst
= xfrm_alloc_dst(net
, family
);
1425 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1427 err
= PTR_ERR(xdst
);
1436 dst_prev
->child
= dst_clone(dst1
);
1437 dst1
->flags
|= DST_NOHASH
;
1441 memcpy(&dst1
->metrics
, &dst
->metrics
, sizeof(dst
->metrics
));
1443 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1444 family
= xfrm
[i
]->props
.family
;
1445 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1453 dst1
->xfrm
= xfrm
[i
];
1454 xdst
->genid
= xfrm
[i
]->genid
;
1456 dst1
->obsolete
= -1;
1457 dst1
->flags
|= DST_HOST
;
1458 dst1
->lastuse
= now
;
1460 dst1
->input
= dst_discard
;
1461 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1463 dst1
->next
= dst_prev
;
1466 header_len
+= xfrm
[i
]->props
.header_len
;
1467 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1468 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1469 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1472 dst_prev
->child
= dst
;
1480 /* Copy neighbour for reachability confirmation */
1481 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1483 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1484 xfrm_init_pmtu(dst_prev
);
1486 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1487 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1489 err
= xfrm_fill_dst(xdst
, dev
);
1493 dst_prev
->header_len
= header_len
;
1494 dst_prev
->trailer_len
= trailer_len
;
1495 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1496 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1504 xfrm_state_put(xfrm
[i
]);
1508 dst0
= ERR_PTR(err
);
1513 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1516 *target
= kmalloc(size
, GFP_ATOMIC
);
1520 memcpy(*target
, src
, size
);
1525 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1527 #ifdef CONFIG_XFRM_SUB_POLICY
1528 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1529 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1537 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1539 #ifdef CONFIG_XFRM_SUB_POLICY
1540 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1541 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1547 static int stale_bundle(struct dst_entry
*dst
);
1549 /* Main function: finds/creates a bundle for given flow.
1551 * At the moment we eat a raw IP route. Mostly to speed up lookups
1552 * on interfaces with disabled IPsec.
1554 int __xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1555 struct sock
*sk
, int flags
)
1557 struct xfrm_policy
*policy
;
1558 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1563 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1564 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1569 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1572 genid
= atomic_read(&flow_cache_genid
);
1574 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1580 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1581 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1582 err
= PTR_ERR(policy
);
1583 if (IS_ERR(policy
)) {
1584 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1590 /* To accelerate a bit... */
1591 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1592 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1595 policy
= flow_cache_lookup(net
, fl
, dst_orig
->ops
->family
,
1596 dir
, xfrm_policy_lookup
);
1597 err
= PTR_ERR(policy
);
1598 if (IS_ERR(policy
)) {
1599 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1607 family
= dst_orig
->ops
->family
;
1610 xfrm_nr
+= pols
[0]->xfrm_nr
;
1613 if ((flags
& XFRM_LOOKUP_ICMP
) && !(policy
->flags
& XFRM_POLICY_ICMP
))
1616 policy
->curlft
.use_time
= get_seconds();
1618 switch (policy
->action
) {
1620 case XFRM_POLICY_BLOCK
:
1621 /* Prohibit the flow */
1622 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
1626 case XFRM_POLICY_ALLOW
:
1627 #ifndef CONFIG_XFRM_SUB_POLICY
1628 if (policy
->xfrm_nr
== 0) {
1629 /* Flow passes not transformed. */
1630 xfrm_pol_put(policy
);
1635 /* Try to find matching bundle.
1637 * LATER: help from flow cache. It is optional, this
1638 * is required only for output policy.
1640 dst
= xfrm_find_bundle(fl
, policy
, family
);
1642 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1650 #ifdef CONFIG_XFRM_SUB_POLICY
1651 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1652 pols
[1] = xfrm_policy_lookup_bytype(net
,
1653 XFRM_POLICY_TYPE_MAIN
,
1657 if (IS_ERR(pols
[1])) {
1658 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1659 err
= PTR_ERR(pols
[1]);
1662 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1663 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
1668 xfrm_nr
+= pols
[1]->xfrm_nr
;
1673 * Because neither flowi nor bundle information knows about
1674 * transformation template size. On more than one policy usage
1675 * we can realize whether all of them is bypass or not after
1676 * they are searched. See above not-transformed bypass
1677 * is surrounded by non-sub policy configuration, too.
1680 /* Flow passes not transformed. */
1681 xfrm_pols_put(pols
, npols
);
1686 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1688 if (unlikely(nx
<0)) {
1690 if (err
== -EAGAIN
&& net
->xfrm
.sysctl_larval_drop
) {
1691 /* EREMOTE tells the caller to generate
1692 * a one-shot blackhole route.
1694 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1695 xfrm_pol_put(policy
);
1698 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1699 DECLARE_WAITQUEUE(wait
, current
);
1701 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1702 set_current_state(TASK_INTERRUPTIBLE
);
1704 set_current_state(TASK_RUNNING
);
1705 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1707 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1709 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1710 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1714 if (nx
== -EAGAIN
||
1715 genid
!= atomic_read(&flow_cache_genid
)) {
1716 xfrm_pols_put(pols
, npols
);
1722 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1727 /* Flow passes not transformed. */
1728 xfrm_pols_put(pols
, npols
);
1732 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1735 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1739 for (pi
= 0; pi
< npols
; pi
++) {
1740 read_lock_bh(&pols
[pi
]->lock
);
1741 pol_dead
|= pols
[pi
]->walk
.dead
;
1742 read_unlock_bh(&pols
[pi
]->lock
);
1745 write_lock_bh(&policy
->lock
);
1746 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1747 /* Wow! While we worked on resolving, this
1748 * policy has gone. Retry. It is not paranoia,
1749 * we just cannot enlist new bundle to dead object.
1750 * We can't enlist stable bundles either.
1752 write_unlock_bh(&policy
->lock
);
1756 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLDEAD
);
1758 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1759 err
= -EHOSTUNREACH
;
1764 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1766 err
= xfrm_dst_update_origin(dst
, fl
);
1767 if (unlikely(err
)) {
1768 write_unlock_bh(&policy
->lock
);
1770 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1774 dst
->next
= policy
->bundles
;
1775 policy
->bundles
= dst
;
1777 write_unlock_bh(&policy
->lock
);
1780 dst_release(dst_orig
);
1781 xfrm_pols_put(pols
, npols
);
1785 xfrm_pols_put(pols
, npols
);
1787 dst_release(dst_orig
);
1793 if (flags
& XFRM_LOOKUP_ICMP
)
1797 EXPORT_SYMBOL(__xfrm_lookup
);
1799 int xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1800 struct sock
*sk
, int flags
)
1802 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1804 if (err
== -EREMOTE
) {
1805 dst_release(*dst_p
);
1812 EXPORT_SYMBOL(xfrm_lookup
);
1815 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1817 struct xfrm_state
*x
;
1819 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1821 x
= skb
->sp
->xvec
[idx
];
1822 if (!x
->type
->reject
)
1824 return x
->type
->reject(x
, skb
, fl
);
1827 /* When skb is transformed back to its "native" form, we have to
1828 * check policy restrictions. At the moment we make this in maximally
1829 * stupid way. Shame on me. :-) Of course, connected sockets must
1830 * have policy cached at them.
1834 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1835 unsigned short family
)
1837 if (xfrm_state_kern(x
))
1838 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1839 return x
->id
.proto
== tmpl
->id
.proto
&&
1840 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1841 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1842 x
->props
.mode
== tmpl
->mode
&&
1843 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1844 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1845 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1846 xfrm_state_addr_cmp(tmpl
, x
, family
));
1850 * 0 or more than 0 is returned when validation is succeeded (either bypass
1851 * because of optional transport mode, or next index of the mathced secpath
1852 * state with the template.
1853 * -1 is returned when no matching template is found.
1854 * Otherwise "-2 - errored_index" is returned.
1857 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1858 unsigned short family
)
1862 if (tmpl
->optional
) {
1863 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1867 for (; idx
< sp
->len
; idx
++) {
1868 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1870 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1879 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1880 unsigned int family
, int reverse
)
1882 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1885 if (unlikely(afinfo
== NULL
))
1886 return -EAFNOSUPPORT
;
1888 afinfo
->decode_session(skb
, fl
, reverse
);
1889 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1890 xfrm_policy_put_afinfo(afinfo
);
1893 EXPORT_SYMBOL(__xfrm_decode_session
);
1895 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1897 for (; k
< sp
->len
; k
++) {
1898 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1907 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1908 unsigned short family
)
1910 struct net
*net
= dev_net(skb
->dev
);
1911 struct xfrm_policy
*pol
;
1912 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1921 reverse
= dir
& ~XFRM_POLICY_MASK
;
1922 dir
&= XFRM_POLICY_MASK
;
1923 fl_dir
= policy_to_flow_dir(dir
);
1925 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1926 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
1930 nf_nat_decode_session(skb
, &fl
, family
);
1932 /* First, check used SA against their selectors. */
1936 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1937 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1938 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1939 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
1946 if (sk
&& sk
->sk_policy
[dir
]) {
1947 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1949 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1955 pol
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
1956 xfrm_policy_lookup
);
1959 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1964 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1965 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1966 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
1972 pol
->curlft
.use_time
= get_seconds();
1976 #ifdef CONFIG_XFRM_SUB_POLICY
1977 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1978 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
1982 if (IS_ERR(pols
[1])) {
1983 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
1986 pols
[1]->curlft
.use_time
= get_seconds();
1992 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1993 struct sec_path
*sp
;
1994 static struct sec_path dummy
;
1995 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1996 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1997 struct xfrm_tmpl
**tpp
= tp
;
2001 if ((sp
= skb
->sp
) == NULL
)
2004 for (pi
= 0; pi
< npols
; pi
++) {
2005 if (pols
[pi
] != pol
&&
2006 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
2007 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2010 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2011 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
2014 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2015 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2019 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2023 /* For each tunnel xfrm, find the first matching tmpl.
2024 * For each tmpl before that, find corresponding xfrm.
2025 * Order is _important_. Later we will implement
2026 * some barriers, but at the moment barriers
2027 * are implied between each two transformations.
2029 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2030 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2033 /* "-2 - errored_index" returned */
2035 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2040 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2041 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2045 xfrm_pols_put(pols
, npols
);
2048 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2051 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2053 xfrm_pols_put(pols
, npols
);
2056 EXPORT_SYMBOL(__xfrm_policy_check
);
2058 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2060 struct net
*net
= dev_net(skb
->dev
);
2062 struct dst_entry
*dst
;
2065 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2066 XFRM_INC_STATS(net
, LINUX_MIB_XFRMFWDHDRERROR
);
2072 res
= xfrm_lookup(net
, &dst
, &fl
, NULL
, 0) == 0;
2073 skb_dst_set(skb
, dst
);
2076 EXPORT_SYMBOL(__xfrm_route_forward
);
2078 /* Optimize later using cookies and generation ids. */
2080 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2082 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2083 * to "-1" to force all XFRM destinations to get validated by
2084 * dst_ops->check on every use. We do this because when a
2085 * normal route referenced by an XFRM dst is obsoleted we do
2086 * not go looking around for all parent referencing XFRM dsts
2087 * so that we can invalidate them. It is just too much work.
2088 * Instead we make the checks here on every use. For example:
2090 * XFRM dst A --> IPv4 dst X
2092 * X is the "xdst->route" of A (X is also the "dst->path" of A
2093 * in this example). If X is marked obsolete, "A" will not
2094 * notice. That's what we are validating here via the
2095 * stale_bundle() check.
2097 * When a policy's bundle is pruned, we dst_free() the XFRM
2098 * dst which causes it's ->obsolete field to be set to a
2099 * positive non-zero integer. If an XFRM dst has been pruned
2100 * like this, we want to force a new route lookup.
2102 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2108 static int stale_bundle(struct dst_entry
*dst
)
2110 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2113 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2115 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2116 dst
->dev
= dev_net(dev
)->loopback_dev
;
2121 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2123 static void xfrm_link_failure(struct sk_buff
*skb
)
2125 /* Impossible. Such dst must be popped before reaches point of failure. */
2129 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2132 if (dst
->obsolete
) {
2140 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
2142 struct dst_entry
*dst
, **dstp
;
2144 write_lock(&pol
->lock
);
2145 dstp
= &pol
->bundles
;
2146 while ((dst
=*dstp
) != NULL
) {
2149 dst
->next
= *gc_list_p
;
2155 write_unlock(&pol
->lock
);
2158 static void xfrm_prune_bundles(struct net
*net
, int (*func
)(struct dst_entry
*))
2160 struct dst_entry
*gc_list
= NULL
;
2163 read_lock_bh(&xfrm_policy_lock
);
2164 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2165 struct xfrm_policy
*pol
;
2166 struct hlist_node
*entry
;
2167 struct hlist_head
*table
;
2170 hlist_for_each_entry(pol
, entry
,
2171 &net
->xfrm
.policy_inexact
[dir
], bydst
)
2172 prune_one_bundle(pol
, func
, &gc_list
);
2174 table
= net
->xfrm
.policy_bydst
[dir
].table
;
2175 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
2176 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
2177 prune_one_bundle(pol
, func
, &gc_list
);
2180 read_unlock_bh(&xfrm_policy_lock
);
2183 struct dst_entry
*dst
= gc_list
;
2184 gc_list
= dst
->next
;
2189 static int unused_bundle(struct dst_entry
*dst
)
2191 return !atomic_read(&dst
->__refcnt
);
2194 static void __xfrm_garbage_collect(struct net
*net
)
2196 xfrm_prune_bundles(net
, unused_bundle
);
2199 static int xfrm_flush_bundles(struct net
*net
)
2201 xfrm_prune_bundles(net
, stale_bundle
);
2205 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2208 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2209 u32 pmtu
, route_mtu_cached
;
2211 pmtu
= dst_mtu(dst
->child
);
2212 xdst
->child_mtu_cached
= pmtu
;
2214 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2216 route_mtu_cached
= dst_mtu(xdst
->route
);
2217 xdst
->route_mtu_cached
= route_mtu_cached
;
2219 if (pmtu
> route_mtu_cached
)
2220 pmtu
= route_mtu_cached
;
2222 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
2223 } while ((dst
= dst
->next
));
2226 /* Check that the bundle accepts the flow and its components are
2230 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2231 struct flowi
*fl
, int family
, int strict
)
2233 struct dst_entry
*dst
= &first
->u
.dst
;
2234 struct xfrm_dst
*last
;
2237 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2238 (dst
->dev
&& !netif_running(dst
->dev
)))
2240 #ifdef CONFIG_XFRM_SUB_POLICY
2242 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2244 if (first
->partner
&&
2245 !xfrm_selector_match(first
->partner
, fl
, family
))
2253 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2255 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2258 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2260 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2262 if (xdst
->genid
!= dst
->xfrm
->genid
)
2266 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2267 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2270 mtu
= dst_mtu(dst
->child
);
2271 if (xdst
->child_mtu_cached
!= mtu
) {
2273 xdst
->child_mtu_cached
= mtu
;
2276 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2278 mtu
= dst_mtu(xdst
->route
);
2279 if (xdst
->route_mtu_cached
!= mtu
) {
2281 xdst
->route_mtu_cached
= mtu
;
2285 } while (dst
->xfrm
);
2290 mtu
= last
->child_mtu_cached
;
2294 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2295 if (mtu
> last
->route_mtu_cached
)
2296 mtu
= last
->route_mtu_cached
;
2297 dst
->metrics
[RTAX_MTU
-1] = mtu
;
2302 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2303 last
->child_mtu_cached
= mtu
;
2309 EXPORT_SYMBOL(xfrm_bundle_ok
);
2311 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2315 if (unlikely(afinfo
== NULL
))
2317 if (unlikely(afinfo
->family
>= NPROTO
))
2318 return -EAFNOSUPPORT
;
2319 write_lock_bh(&xfrm_policy_afinfo_lock
);
2320 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2323 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2324 if (likely(dst_ops
->kmem_cachep
== NULL
))
2325 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2326 if (likely(dst_ops
->check
== NULL
))
2327 dst_ops
->check
= xfrm_dst_check
;
2328 if (likely(dst_ops
->negative_advice
== NULL
))
2329 dst_ops
->negative_advice
= xfrm_negative_advice
;
2330 if (likely(dst_ops
->link_failure
== NULL
))
2331 dst_ops
->link_failure
= xfrm_link_failure
;
2332 if (likely(afinfo
->garbage_collect
== NULL
))
2333 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2334 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2336 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2340 struct dst_ops
*xfrm_dst_ops
;
2342 switch (afinfo
->family
) {
2344 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2346 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2348 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2354 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2360 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2362 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2365 if (unlikely(afinfo
== NULL
))
2367 if (unlikely(afinfo
->family
>= NPROTO
))
2368 return -EAFNOSUPPORT
;
2369 write_lock_bh(&xfrm_policy_afinfo_lock
);
2370 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2371 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2374 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2375 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2376 dst_ops
->kmem_cachep
= NULL
;
2377 dst_ops
->check
= NULL
;
2378 dst_ops
->negative_advice
= NULL
;
2379 dst_ops
->link_failure
= NULL
;
2380 afinfo
->garbage_collect
= NULL
;
2383 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2386 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2388 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2390 struct xfrm_policy_afinfo
*afinfo
;
2392 read_lock_bh(&xfrm_policy_afinfo_lock
);
2393 afinfo
= xfrm_policy_afinfo
[AF_INET
];
2395 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2396 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2397 afinfo
= xfrm_policy_afinfo
[AF_INET6
];
2399 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2401 read_unlock_bh(&xfrm_policy_afinfo_lock
);
2404 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2406 struct xfrm_policy_afinfo
*afinfo
;
2407 if (unlikely(family
>= NPROTO
))
2409 read_lock(&xfrm_policy_afinfo_lock
);
2410 afinfo
= xfrm_policy_afinfo
[family
];
2411 if (unlikely(!afinfo
))
2412 read_unlock(&xfrm_policy_afinfo_lock
);
2416 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2418 read_unlock(&xfrm_policy_afinfo_lock
);
2421 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2423 struct net_device
*dev
= ptr
;
2427 xfrm_flush_bundles(dev_net(dev
));
2432 static struct notifier_block xfrm_dev_notifier
= {
2433 .notifier_call
= xfrm_dev_event
,
2436 #ifdef CONFIG_XFRM_STATISTICS
2437 static int __net_init
xfrm_statistics_init(struct net
*net
)
2441 if (snmp_mib_init((void __percpu
**)net
->mib
.xfrm_statistics
,
2442 sizeof(struct linux_xfrm_mib
)) < 0)
2444 rv
= xfrm_proc_init(net
);
2446 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2450 static void xfrm_statistics_fini(struct net
*net
)
2452 xfrm_proc_fini(net
);
2453 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2456 static int __net_init
xfrm_statistics_init(struct net
*net
)
2461 static void xfrm_statistics_fini(struct net
*net
)
2466 static int __net_init
xfrm_policy_init(struct net
*net
)
2468 unsigned int hmask
, sz
;
2471 if (net_eq(net
, &init_net
))
2472 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2473 sizeof(struct xfrm_dst
),
2474 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2478 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2480 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2481 if (!net
->xfrm
.policy_byidx
)
2483 net
->xfrm
.policy_idx_hmask
= hmask
;
2485 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2486 struct xfrm_policy_hash
*htab
;
2488 net
->xfrm
.policy_count
[dir
] = 0;
2489 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2491 htab
= &net
->xfrm
.policy_bydst
[dir
];
2492 htab
->table
= xfrm_hash_alloc(sz
);
2495 htab
->hmask
= hmask
;
2498 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2499 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2500 if (net_eq(net
, &init_net
))
2501 register_netdevice_notifier(&xfrm_dev_notifier
);
2505 for (dir
--; dir
>= 0; dir
--) {
2506 struct xfrm_policy_hash
*htab
;
2508 htab
= &net
->xfrm
.policy_bydst
[dir
];
2509 xfrm_hash_free(htab
->table
, sz
);
2511 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2516 static void xfrm_policy_fini(struct net
*net
)
2518 struct xfrm_audit audit_info
;
2522 flush_work(&net
->xfrm
.policy_hash_work
);
2523 #ifdef CONFIG_XFRM_SUB_POLICY
2524 audit_info
.loginuid
= -1;
2525 audit_info
.sessionid
= -1;
2526 audit_info
.secid
= 0;
2527 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2529 audit_info
.loginuid
= -1;
2530 audit_info
.sessionid
= -1;
2531 audit_info
.secid
= 0;
2532 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2533 flush_work(&xfrm_policy_gc_work
);
2535 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2537 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2538 struct xfrm_policy_hash
*htab
;
2540 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2542 htab
= &net
->xfrm
.policy_bydst
[dir
];
2543 sz
= (htab
->hmask
+ 1);
2544 WARN_ON(!hlist_empty(htab
->table
));
2545 xfrm_hash_free(htab
->table
, sz
);
2548 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2549 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2550 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2553 static int __net_init
xfrm_net_init(struct net
*net
)
2557 rv
= xfrm_statistics_init(net
);
2559 goto out_statistics
;
2560 rv
= xfrm_state_init(net
);
2563 rv
= xfrm_policy_init(net
);
2566 xfrm_dst_ops_init(net
);
2567 rv
= xfrm_sysctl_init(net
);
2573 xfrm_policy_fini(net
);
2575 xfrm_state_fini(net
);
2577 xfrm_statistics_fini(net
);
2582 static void __net_exit
xfrm_net_exit(struct net
*net
)
2584 xfrm_sysctl_fini(net
);
2585 xfrm_policy_fini(net
);
2586 xfrm_state_fini(net
);
2587 xfrm_statistics_fini(net
);
2590 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2591 .init
= xfrm_net_init
,
2592 .exit
= xfrm_net_exit
,
2595 void __init
xfrm_init(void)
2597 register_pernet_subsys(&xfrm_net_ops
);
2601 #ifdef CONFIG_AUDITSYSCALL
2602 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2603 struct audit_buffer
*audit_buf
)
2605 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2606 struct xfrm_selector
*sel
= &xp
->selector
;
2609 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2610 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2612 switch(sel
->family
) {
2614 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2615 if (sel
->prefixlen_s
!= 32)
2616 audit_log_format(audit_buf
, " src_prefixlen=%d",
2618 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2619 if (sel
->prefixlen_d
!= 32)
2620 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2624 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2625 if (sel
->prefixlen_s
!= 128)
2626 audit_log_format(audit_buf
, " src_prefixlen=%d",
2628 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2629 if (sel
->prefixlen_d
!= 128)
2630 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2636 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2637 uid_t auid
, u32 sessionid
, u32 secid
)
2639 struct audit_buffer
*audit_buf
;
2641 audit_buf
= xfrm_audit_start("SPD-add");
2642 if (audit_buf
== NULL
)
2644 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2645 audit_log_format(audit_buf
, " res=%u", result
);
2646 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2647 audit_log_end(audit_buf
);
2649 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2651 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2652 uid_t auid
, u32 sessionid
, u32 secid
)
2654 struct audit_buffer
*audit_buf
;
2656 audit_buf
= xfrm_audit_start("SPD-delete");
2657 if (audit_buf
== NULL
)
2659 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2660 audit_log_format(audit_buf
, " res=%u", result
);
2661 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2662 audit_log_end(audit_buf
);
2664 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2667 #ifdef CONFIG_XFRM_MIGRATE
2668 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2669 struct xfrm_selector
*sel_tgt
)
2671 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2672 if (sel_tgt
->family
== sel_cmp
->family
&&
2673 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2674 sel_cmp
->family
) == 0 &&
2675 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2676 sel_cmp
->family
) == 0 &&
2677 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2678 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2682 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2689 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2692 struct xfrm_policy
*pol
, *ret
= NULL
;
2693 struct hlist_node
*entry
;
2694 struct hlist_head
*chain
;
2697 read_lock_bh(&xfrm_policy_lock
);
2698 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2699 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2700 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2701 pol
->type
== type
) {
2703 priority
= ret
->priority
;
2707 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2708 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2709 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2710 pol
->type
== type
&&
2711 pol
->priority
< priority
) {
2720 read_unlock_bh(&xfrm_policy_lock
);
2725 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2729 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2730 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2732 case XFRM_MODE_TUNNEL
:
2733 case XFRM_MODE_BEET
:
2734 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2735 m
->old_family
) == 0 &&
2736 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2737 m
->old_family
) == 0) {
2741 case XFRM_MODE_TRANSPORT
:
2742 /* in case of transport mode, template does not store
2743 any IP addresses, hence we just compare mode and
2754 /* update endpoint address(es) of template(s) */
2755 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2756 struct xfrm_migrate
*m
, int num_migrate
)
2758 struct xfrm_migrate
*mp
;
2759 struct dst_entry
*dst
;
2762 write_lock_bh(&pol
->lock
);
2763 if (unlikely(pol
->walk
.dead
)) {
2764 /* target policy has been deleted */
2765 write_unlock_bh(&pol
->lock
);
2769 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2770 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2771 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2774 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2775 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2777 /* update endpoints */
2778 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2779 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2780 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2781 sizeof(pol
->xfrm_vec
[i
].saddr
));
2782 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2784 while ((dst
= pol
->bundles
) != NULL
) {
2785 pol
->bundles
= dst
->next
;
2791 write_unlock_bh(&pol
->lock
);
2799 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2803 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2806 for (i
= 0; i
< num_migrate
; i
++) {
2807 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2808 m
[i
].old_family
) == 0) &&
2809 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2810 m
[i
].old_family
) == 0))
2812 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2813 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2816 /* check if there is any duplicated entry */
2817 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2818 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2819 sizeof(m
[i
].old_daddr
)) &&
2820 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2821 sizeof(m
[i
].old_saddr
)) &&
2822 m
[i
].proto
== m
[j
].proto
&&
2823 m
[i
].mode
== m
[j
].mode
&&
2824 m
[i
].reqid
== m
[j
].reqid
&&
2825 m
[i
].old_family
== m
[j
].old_family
)
2833 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2834 struct xfrm_migrate
*m
, int num_migrate
,
2835 struct xfrm_kmaddress
*k
)
2837 int i
, err
, nx_cur
= 0, nx_new
= 0;
2838 struct xfrm_policy
*pol
= NULL
;
2839 struct xfrm_state
*x
, *xc
;
2840 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2841 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2842 struct xfrm_migrate
*mp
;
2844 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2847 /* Stage 1 - find policy */
2848 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2853 /* Stage 2 - find and update state(s) */
2854 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2855 if ((x
= xfrm_migrate_state_find(mp
))) {
2858 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2868 /* Stage 3 - update policy */
2869 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2872 /* Stage 4 - delete old state(s) */
2874 xfrm_states_put(x_cur
, nx_cur
);
2875 xfrm_states_delete(x_cur
, nx_cur
);
2878 /* Stage 5 - announce */
2879 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2891 xfrm_states_put(x_cur
, nx_cur
);
2893 xfrm_states_delete(x_new
, nx_new
);
2897 EXPORT_SYMBOL(xfrm_migrate
);