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_SPINLOCK(xfrm_policy_sk_bundle_lock
);
41 static struct dst_entry
*xfrm_policy_sk_bundles
;
42 static DEFINE_RWLOCK(xfrm_policy_lock
);
44 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
45 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
47 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
49 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
50 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
51 static void xfrm_init_pmtu(struct dst_entry
*dst
);
52 static int stale_bundle(struct dst_entry
*dst
);
53 static int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*xdst
,
54 const struct flowi
*fl
, int family
);
57 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
61 __xfrm4_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
)
63 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
64 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
65 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
66 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
67 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
68 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
72 __xfrm6_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
)
74 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
75 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
76 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
77 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
78 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
79 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
82 int xfrm_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
,
83 unsigned short family
)
87 return __xfrm4_selector_match(sel
, fl
);
89 return __xfrm6_selector_match(sel
, fl
);
94 static inline struct dst_entry
*__xfrm_dst_lookup(struct net
*net
, int tos
,
95 const xfrm_address_t
*saddr
,
96 const xfrm_address_t
*daddr
,
99 struct xfrm_policy_afinfo
*afinfo
;
100 struct dst_entry
*dst
;
102 afinfo
= xfrm_policy_get_afinfo(family
);
103 if (unlikely(afinfo
== NULL
))
104 return ERR_PTR(-EAFNOSUPPORT
);
106 dst
= afinfo
->dst_lookup(net
, tos
, saddr
, daddr
);
108 xfrm_policy_put_afinfo(afinfo
);
113 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
, int tos
,
114 xfrm_address_t
*prev_saddr
,
115 xfrm_address_t
*prev_daddr
,
118 struct net
*net
= xs_net(x
);
119 xfrm_address_t
*saddr
= &x
->props
.saddr
;
120 xfrm_address_t
*daddr
= &x
->id
.daddr
;
121 struct dst_entry
*dst
;
123 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
127 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
132 dst
= __xfrm_dst_lookup(net
, tos
, saddr
, daddr
, family
);
135 if (prev_saddr
!= saddr
)
136 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
137 if (prev_daddr
!= daddr
)
138 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
144 static inline unsigned long make_jiffies(long secs
)
146 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
147 return MAX_SCHEDULE_TIMEOUT
-1;
152 static void xfrm_policy_timer(unsigned long data
)
154 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
155 unsigned long now
= get_seconds();
156 long next
= LONG_MAX
;
160 read_lock(&xp
->lock
);
162 if (unlikely(xp
->walk
.dead
))
165 dir
= xfrm_policy_id2dir(xp
->index
);
167 if (xp
->lft
.hard_add_expires_seconds
) {
168 long tmo
= xp
->lft
.hard_add_expires_seconds
+
169 xp
->curlft
.add_time
- now
;
175 if (xp
->lft
.hard_use_expires_seconds
) {
176 long tmo
= xp
->lft
.hard_use_expires_seconds
+
177 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
183 if (xp
->lft
.soft_add_expires_seconds
) {
184 long tmo
= xp
->lft
.soft_add_expires_seconds
+
185 xp
->curlft
.add_time
- now
;
188 tmo
= XFRM_KM_TIMEOUT
;
193 if (xp
->lft
.soft_use_expires_seconds
) {
194 long tmo
= xp
->lft
.soft_use_expires_seconds
+
195 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
198 tmo
= XFRM_KM_TIMEOUT
;
205 km_policy_expired(xp
, dir
, 0, 0);
206 if (next
!= LONG_MAX
&&
207 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
211 read_unlock(&xp
->lock
);
216 read_unlock(&xp
->lock
);
217 if (!xfrm_policy_delete(xp
, dir
))
218 km_policy_expired(xp
, dir
, 1, 0);
222 static struct flow_cache_object
*xfrm_policy_flo_get(struct flow_cache_object
*flo
)
224 struct xfrm_policy
*pol
= container_of(flo
, struct xfrm_policy
, flo
);
226 if (unlikely(pol
->walk
.dead
))
234 static int xfrm_policy_flo_check(struct flow_cache_object
*flo
)
236 struct xfrm_policy
*pol
= container_of(flo
, struct xfrm_policy
, flo
);
238 return !pol
->walk
.dead
;
241 static void xfrm_policy_flo_delete(struct flow_cache_object
*flo
)
243 xfrm_pol_put(container_of(flo
, struct xfrm_policy
, flo
));
246 static const struct flow_cache_ops xfrm_policy_fc_ops
= {
247 .get
= xfrm_policy_flo_get
,
248 .check
= xfrm_policy_flo_check
,
249 .delete = xfrm_policy_flo_delete
,
252 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
256 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
258 struct xfrm_policy
*policy
;
260 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
263 write_pnet(&policy
->xp_net
, net
);
264 INIT_LIST_HEAD(&policy
->walk
.all
);
265 INIT_HLIST_NODE(&policy
->bydst
);
266 INIT_HLIST_NODE(&policy
->byidx
);
267 rwlock_init(&policy
->lock
);
268 atomic_set(&policy
->refcnt
, 1);
269 setup_timer(&policy
->timer
, xfrm_policy_timer
,
270 (unsigned long)policy
);
271 policy
->flo
.ops
= &xfrm_policy_fc_ops
;
275 EXPORT_SYMBOL(xfrm_policy_alloc
);
277 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
279 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
281 BUG_ON(!policy
->walk
.dead
);
283 if (del_timer(&policy
->timer
))
286 security_xfrm_policy_free(policy
->security
);
289 EXPORT_SYMBOL(xfrm_policy_destroy
);
291 /* Rule must be locked. Release descentant resources, announce
292 * entry dead. The rule must be unlinked from lists to the moment.
295 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
297 policy
->walk
.dead
= 1;
299 atomic_inc(&policy
->genid
);
301 if (del_timer(&policy
->timer
))
302 xfrm_pol_put(policy
);
304 xfrm_pol_put(policy
);
307 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
309 static inline unsigned int idx_hash(struct net
*net
, u32 index
)
311 return __idx_hash(index
, net
->xfrm
.policy_idx_hmask
);
314 static struct hlist_head
*policy_hash_bysel(struct net
*net
,
315 const struct xfrm_selector
*sel
,
316 unsigned short family
, int dir
)
318 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
319 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
321 return (hash
== hmask
+ 1 ?
322 &net
->xfrm
.policy_inexact
[dir
] :
323 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
326 static struct hlist_head
*policy_hash_direct(struct net
*net
,
327 const xfrm_address_t
*daddr
,
328 const xfrm_address_t
*saddr
,
329 unsigned short family
, int dir
)
331 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
332 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
334 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
337 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
338 struct hlist_head
*ndsttable
,
339 unsigned int nhashmask
)
341 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
342 struct xfrm_policy
*pol
;
346 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
349 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
350 pol
->family
, nhashmask
);
353 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
359 hlist_add_after(entry0
, &pol
->bydst
);
363 if (!hlist_empty(list
)) {
369 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
370 struct hlist_head
*nidxtable
,
371 unsigned int nhashmask
)
373 struct hlist_node
*entry
, *tmp
;
374 struct xfrm_policy
*pol
;
376 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
379 h
= __idx_hash(pol
->index
, nhashmask
);
380 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
384 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
386 return ((old_hmask
+ 1) << 1) - 1;
389 static void xfrm_bydst_resize(struct net
*net
, int dir
)
391 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
392 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
393 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
394 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
395 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
401 write_lock_bh(&xfrm_policy_lock
);
403 for (i
= hmask
; i
>= 0; i
--)
404 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
406 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
407 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
409 write_unlock_bh(&xfrm_policy_lock
);
411 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
414 static void xfrm_byidx_resize(struct net
*net
, int total
)
416 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
417 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
418 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
419 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
420 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
426 write_lock_bh(&xfrm_policy_lock
);
428 for (i
= hmask
; i
>= 0; i
--)
429 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
431 net
->xfrm
.policy_byidx
= nidx
;
432 net
->xfrm
.policy_idx_hmask
= nhashmask
;
434 write_unlock_bh(&xfrm_policy_lock
);
436 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
439 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
441 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
442 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
447 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
454 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
456 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
458 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
465 void xfrm_spd_getinfo(struct net
*net
, struct xfrmk_spdinfo
*si
)
467 read_lock_bh(&xfrm_policy_lock
);
468 si
->incnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
];
469 si
->outcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
];
470 si
->fwdcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
];
471 si
->inscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
472 si
->outscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
473 si
->fwdscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
474 si
->spdhcnt
= net
->xfrm
.policy_idx_hmask
;
475 si
->spdhmcnt
= xfrm_policy_hashmax
;
476 read_unlock_bh(&xfrm_policy_lock
);
478 EXPORT_SYMBOL(xfrm_spd_getinfo
);
480 static DEFINE_MUTEX(hash_resize_mutex
);
481 static void xfrm_hash_resize(struct work_struct
*work
)
483 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
486 mutex_lock(&hash_resize_mutex
);
489 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
490 if (xfrm_bydst_should_resize(net
, dir
, &total
))
491 xfrm_bydst_resize(net
, dir
);
493 if (xfrm_byidx_should_resize(net
, total
))
494 xfrm_byidx_resize(net
, total
);
496 mutex_unlock(&hash_resize_mutex
);
499 /* Generate new index... KAME seems to generate them ordered by cost
500 * of an absolute inpredictability of ordering of rules. This will not pass. */
501 static u32
xfrm_gen_index(struct net
*net
, int dir
)
503 static u32 idx_generator
;
506 struct hlist_node
*entry
;
507 struct hlist_head
*list
;
508 struct xfrm_policy
*p
;
512 idx
= (idx_generator
| dir
);
516 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
518 hlist_for_each_entry(p
, entry
, list
, byidx
) {
519 if (p
->index
== idx
) {
529 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
531 u32
*p1
= (u32
*) s1
;
532 u32
*p2
= (u32
*) s2
;
533 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
536 for (i
= 0; i
< len
; i
++) {
544 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
546 struct net
*net
= xp_net(policy
);
547 struct xfrm_policy
*pol
;
548 struct xfrm_policy
*delpol
;
549 struct hlist_head
*chain
;
550 struct hlist_node
*entry
, *newpos
;
551 u32 mark
= policy
->mark
.v
& policy
->mark
.m
;
553 write_lock_bh(&xfrm_policy_lock
);
554 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
557 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
558 if (pol
->type
== policy
->type
&&
559 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
560 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
561 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
564 write_unlock_bh(&xfrm_policy_lock
);
568 if (policy
->priority
> pol
->priority
)
570 } else if (policy
->priority
>= pol
->priority
) {
571 newpos
= &pol
->bydst
;
578 hlist_add_after(newpos
, &policy
->bydst
);
580 hlist_add_head(&policy
->bydst
, chain
);
581 xfrm_pol_hold(policy
);
582 net
->xfrm
.policy_count
[dir
]++;
583 atomic_inc(&flow_cache_genid
);
585 __xfrm_policy_unlink(delpol
, dir
);
586 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
587 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
588 policy
->curlft
.add_time
= get_seconds();
589 policy
->curlft
.use_time
= 0;
590 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
591 xfrm_pol_hold(policy
);
592 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
593 write_unlock_bh(&xfrm_policy_lock
);
596 xfrm_policy_kill(delpol
);
597 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
598 schedule_work(&net
->xfrm
.policy_hash_work
);
602 EXPORT_SYMBOL(xfrm_policy_insert
);
604 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u32 mark
, u8 type
,
605 int dir
, struct xfrm_selector
*sel
,
606 struct xfrm_sec_ctx
*ctx
, int delete,
609 struct xfrm_policy
*pol
, *ret
;
610 struct hlist_head
*chain
;
611 struct hlist_node
*entry
;
614 write_lock_bh(&xfrm_policy_lock
);
615 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
617 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
618 if (pol
->type
== type
&&
619 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
620 !selector_cmp(sel
, &pol
->selector
) &&
621 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
624 *err
= security_xfrm_policy_delete(
627 write_unlock_bh(&xfrm_policy_lock
);
630 __xfrm_policy_unlink(pol
, dir
);
636 write_unlock_bh(&xfrm_policy_lock
);
639 xfrm_policy_kill(ret
);
642 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
644 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u32 mark
, u8 type
,
645 int dir
, u32 id
, int delete, int *err
)
647 struct xfrm_policy
*pol
, *ret
;
648 struct hlist_head
*chain
;
649 struct hlist_node
*entry
;
652 if (xfrm_policy_id2dir(id
) != dir
)
656 write_lock_bh(&xfrm_policy_lock
);
657 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
659 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
660 if (pol
->type
== type
&& pol
->index
== id
&&
661 (mark
& pol
->mark
.m
) == pol
->mark
.v
) {
664 *err
= security_xfrm_policy_delete(
667 write_unlock_bh(&xfrm_policy_lock
);
670 __xfrm_policy_unlink(pol
, dir
);
676 write_unlock_bh(&xfrm_policy_lock
);
679 xfrm_policy_kill(ret
);
682 EXPORT_SYMBOL(xfrm_policy_byid
);
684 #ifdef CONFIG_SECURITY_NETWORK_XFRM
686 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
690 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
691 struct xfrm_policy
*pol
;
692 struct hlist_node
*entry
;
695 hlist_for_each_entry(pol
, entry
,
696 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
697 if (pol
->type
!= type
)
699 err
= security_xfrm_policy_delete(pol
->security
);
701 xfrm_audit_policy_delete(pol
, 0,
702 audit_info
->loginuid
,
703 audit_info
->sessionid
,
708 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
709 hlist_for_each_entry(pol
, entry
,
710 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
712 if (pol
->type
!= type
)
714 err
= security_xfrm_policy_delete(
717 xfrm_audit_policy_delete(pol
, 0,
718 audit_info
->loginuid
,
719 audit_info
->sessionid
,
730 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
736 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
738 int dir
, err
= 0, cnt
= 0;
740 write_lock_bh(&xfrm_policy_lock
);
742 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
746 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
747 struct xfrm_policy
*pol
;
748 struct hlist_node
*entry
;
752 hlist_for_each_entry(pol
, entry
,
753 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
754 if (pol
->type
!= type
)
756 __xfrm_policy_unlink(pol
, dir
);
757 write_unlock_bh(&xfrm_policy_lock
);
760 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
761 audit_info
->sessionid
,
764 xfrm_policy_kill(pol
);
766 write_lock_bh(&xfrm_policy_lock
);
770 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
772 hlist_for_each_entry(pol
, entry
,
773 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
775 if (pol
->type
!= type
)
777 __xfrm_policy_unlink(pol
, dir
);
778 write_unlock_bh(&xfrm_policy_lock
);
781 xfrm_audit_policy_delete(pol
, 1,
782 audit_info
->loginuid
,
783 audit_info
->sessionid
,
785 xfrm_policy_kill(pol
);
787 write_lock_bh(&xfrm_policy_lock
);
796 write_unlock_bh(&xfrm_policy_lock
);
799 EXPORT_SYMBOL(xfrm_policy_flush
);
801 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
802 int (*func
)(struct xfrm_policy
*, int, int, void*),
805 struct xfrm_policy
*pol
;
806 struct xfrm_policy_walk_entry
*x
;
809 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
810 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
813 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
816 write_lock_bh(&xfrm_policy_lock
);
817 if (list_empty(&walk
->walk
.all
))
818 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
820 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
821 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
824 pol
= container_of(x
, struct xfrm_policy
, walk
);
825 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
826 walk
->type
!= pol
->type
)
828 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
831 list_move_tail(&walk
->walk
.all
, &x
->all
);
836 if (walk
->seq
== 0) {
840 list_del_init(&walk
->walk
.all
);
842 write_unlock_bh(&xfrm_policy_lock
);
845 EXPORT_SYMBOL(xfrm_policy_walk
);
847 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
849 INIT_LIST_HEAD(&walk
->walk
.all
);
854 EXPORT_SYMBOL(xfrm_policy_walk_init
);
856 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
858 if (list_empty(&walk
->walk
.all
))
861 write_lock_bh(&xfrm_policy_lock
);
862 list_del(&walk
->walk
.all
);
863 write_unlock_bh(&xfrm_policy_lock
);
865 EXPORT_SYMBOL(xfrm_policy_walk_done
);
868 * Find policy to apply to this flow.
870 * Returns 0 if policy found, else an -errno.
872 static int xfrm_policy_match(const struct xfrm_policy
*pol
,
873 const struct flowi
*fl
,
874 u8 type
, u16 family
, int dir
)
876 const struct xfrm_selector
*sel
= &pol
->selector
;
877 int match
, ret
= -ESRCH
;
879 if (pol
->family
!= family
||
880 (fl
->mark
& pol
->mark
.m
) != pol
->mark
.v
||
884 match
= xfrm_selector_match(sel
, fl
, family
);
886 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
892 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
893 const struct flowi
*fl
,
897 struct xfrm_policy
*pol
, *ret
;
898 const xfrm_address_t
*daddr
, *saddr
;
899 struct hlist_node
*entry
;
900 struct hlist_head
*chain
;
903 daddr
= xfrm_flowi_daddr(fl
, family
);
904 saddr
= xfrm_flowi_saddr(fl
, family
);
905 if (unlikely(!daddr
|| !saddr
))
908 read_lock_bh(&xfrm_policy_lock
);
909 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
911 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
912 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
922 priority
= ret
->priority
;
926 chain
= &net
->xfrm
.policy_inexact
[dir
];
927 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
928 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
936 } else if (pol
->priority
< priority
) {
944 read_unlock_bh(&xfrm_policy_lock
);
949 static struct xfrm_policy
*
950 __xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
)
952 #ifdef CONFIG_XFRM_SUB_POLICY
953 struct xfrm_policy
*pol
;
955 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
959 return xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
962 static struct flow_cache_object
*
963 xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
,
964 u8 dir
, struct flow_cache_object
*old_obj
, void *ctx
)
966 struct xfrm_policy
*pol
;
969 xfrm_pol_put(container_of(old_obj
, struct xfrm_policy
, flo
));
971 pol
= __xfrm_policy_lookup(net
, fl
, family
, dir
);
972 if (IS_ERR_OR_NULL(pol
))
973 return ERR_CAST(pol
);
975 /* Resolver returns two references:
976 * one for cache and one for caller of flow_cache_lookup() */
982 static inline int policy_to_flow_dir(int dir
)
984 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
985 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
986 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
992 case XFRM_POLICY_OUT
:
994 case XFRM_POLICY_FWD
:
999 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
,
1000 const struct flowi
*fl
)
1002 struct xfrm_policy
*pol
;
1004 read_lock_bh(&xfrm_policy_lock
);
1005 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1006 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1011 if ((sk
->sk_mark
& pol
->mark
.m
) != pol
->mark
.v
) {
1015 err
= security_xfrm_policy_lookup(pol
->security
,
1017 policy_to_flow_dir(dir
));
1020 else if (err
== -ESRCH
)
1028 read_unlock_bh(&xfrm_policy_lock
);
1032 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1034 struct net
*net
= xp_net(pol
);
1035 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1038 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1039 hlist_add_head(&pol
->bydst
, chain
);
1040 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1041 net
->xfrm
.policy_count
[dir
]++;
1044 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1045 schedule_work(&net
->xfrm
.policy_hash_work
);
1048 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1051 struct net
*net
= xp_net(pol
);
1053 if (hlist_unhashed(&pol
->bydst
))
1056 hlist_del(&pol
->bydst
);
1057 hlist_del(&pol
->byidx
);
1058 list_del(&pol
->walk
.all
);
1059 net
->xfrm
.policy_count
[dir
]--;
1064 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1066 write_lock_bh(&xfrm_policy_lock
);
1067 pol
= __xfrm_policy_unlink(pol
, dir
);
1068 write_unlock_bh(&xfrm_policy_lock
);
1070 xfrm_policy_kill(pol
);
1075 EXPORT_SYMBOL(xfrm_policy_delete
);
1077 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1079 struct net
*net
= xp_net(pol
);
1080 struct xfrm_policy
*old_pol
;
1082 #ifdef CONFIG_XFRM_SUB_POLICY
1083 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1087 write_lock_bh(&xfrm_policy_lock
);
1088 old_pol
= sk
->sk_policy
[dir
];
1089 sk
->sk_policy
[dir
] = pol
;
1091 pol
->curlft
.add_time
= get_seconds();
1092 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1093 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1096 /* Unlinking succeeds always. This is the only function
1097 * allowed to delete or replace socket policy.
1099 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1100 write_unlock_bh(&xfrm_policy_lock
);
1103 xfrm_policy_kill(old_pol
);
1108 static struct xfrm_policy
*clone_policy(const struct xfrm_policy
*old
, int dir
)
1110 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1113 newp
->selector
= old
->selector
;
1114 if (security_xfrm_policy_clone(old
->security
,
1117 return NULL
; /* ENOMEM */
1119 newp
->lft
= old
->lft
;
1120 newp
->curlft
= old
->curlft
;
1121 newp
->mark
= old
->mark
;
1122 newp
->action
= old
->action
;
1123 newp
->flags
= old
->flags
;
1124 newp
->xfrm_nr
= old
->xfrm_nr
;
1125 newp
->index
= old
->index
;
1126 newp
->type
= old
->type
;
1127 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1128 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1129 write_lock_bh(&xfrm_policy_lock
);
1130 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1131 write_unlock_bh(&xfrm_policy_lock
);
1137 int __xfrm_sk_clone_policy(struct sock
*sk
)
1139 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1140 *p1
= sk
->sk_policy
[1];
1142 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1143 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1145 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1151 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1152 unsigned short family
)
1155 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1157 if (unlikely(afinfo
== NULL
))
1159 err
= afinfo
->get_saddr(net
, local
, remote
);
1160 xfrm_policy_put_afinfo(afinfo
);
1164 /* Resolve list of templates for the flow, given policy. */
1167 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, const struct flowi
*fl
,
1168 struct xfrm_state
**xfrm
, unsigned short family
)
1170 struct net
*net
= xp_net(policy
);
1173 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1174 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1177 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1178 struct xfrm_state
*x
;
1179 xfrm_address_t
*remote
= daddr
;
1180 xfrm_address_t
*local
= saddr
;
1181 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1183 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1184 tmpl
->mode
== XFRM_MODE_BEET
) {
1185 remote
= &tmpl
->id
.daddr
;
1186 local
= &tmpl
->saddr
;
1187 if (xfrm_addr_any(local
, tmpl
->encap_family
)) {
1188 error
= xfrm_get_saddr(net
, &tmp
, remote
, tmpl
->encap_family
);
1195 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1197 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1204 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1208 else if (error
== -ESRCH
)
1211 if (!tmpl
->optional
)
1217 for (nx
--; nx
>=0; nx
--)
1218 xfrm_state_put(xfrm
[nx
]);
1223 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, const struct flowi
*fl
,
1224 struct xfrm_state
**xfrm
, unsigned short family
)
1226 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1227 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1233 for (i
= 0; i
< npols
; i
++) {
1234 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1239 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1247 /* found states are sorted for outbound processing */
1249 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1254 for (cnx
--; cnx
>=0; cnx
--)
1255 xfrm_state_put(tpp
[cnx
]);
1260 /* Check that the bundle accepts the flow and its components are
1264 static inline int xfrm_get_tos(const struct flowi
*fl
, int family
)
1266 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1272 tos
= afinfo
->get_tos(fl
);
1274 xfrm_policy_put_afinfo(afinfo
);
1279 static struct flow_cache_object
*xfrm_bundle_flo_get(struct flow_cache_object
*flo
)
1281 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1282 struct dst_entry
*dst
= &xdst
->u
.dst
;
1284 if (xdst
->route
== NULL
) {
1285 /* Dummy bundle - if it has xfrms we were not
1286 * able to build bundle as template resolution failed.
1287 * It means we need to try again resolving. */
1288 if (xdst
->num_xfrms
> 0)
1292 if (stale_bundle(dst
))
1300 static int xfrm_bundle_flo_check(struct flow_cache_object
*flo
)
1302 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1303 struct dst_entry
*dst
= &xdst
->u
.dst
;
1307 if (stale_bundle(dst
))
1313 static void xfrm_bundle_flo_delete(struct flow_cache_object
*flo
)
1315 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1316 struct dst_entry
*dst
= &xdst
->u
.dst
;
1321 static const struct flow_cache_ops xfrm_bundle_fc_ops
= {
1322 .get
= xfrm_bundle_flo_get
,
1323 .check
= xfrm_bundle_flo_check
,
1324 .delete = xfrm_bundle_flo_delete
,
1327 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1329 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1330 struct dst_ops
*dst_ops
;
1331 struct xfrm_dst
*xdst
;
1334 return ERR_PTR(-EINVAL
);
1338 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1340 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1342 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1348 xdst
= dst_alloc(dst_ops
, 0);
1349 xfrm_policy_put_afinfo(afinfo
);
1352 xdst
->flo
.ops
= &xfrm_bundle_fc_ops
;
1354 xdst
= ERR_PTR(-ENOBUFS
);
1359 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1362 struct xfrm_policy_afinfo
*afinfo
=
1363 xfrm_policy_get_afinfo(dst
->ops
->family
);
1369 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1371 xfrm_policy_put_afinfo(afinfo
);
1376 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
,
1377 const struct flowi
*fl
)
1379 struct xfrm_policy_afinfo
*afinfo
=
1380 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1386 err
= afinfo
->fill_dst(xdst
, dev
, fl
);
1388 xfrm_policy_put_afinfo(afinfo
);
1394 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1395 * all the metrics... Shortly, bundle a bundle.
1398 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1399 struct xfrm_state
**xfrm
, int nx
,
1400 const struct flowi
*fl
,
1401 struct dst_entry
*dst
)
1403 struct net
*net
= xp_net(policy
);
1404 unsigned long now
= jiffies
;
1405 struct net_device
*dev
;
1406 struct dst_entry
*dst_prev
= NULL
;
1407 struct dst_entry
*dst0
= NULL
;
1411 int nfheader_len
= 0;
1412 int trailer_len
= 0;
1414 int family
= policy
->selector
.family
;
1415 xfrm_address_t saddr
, daddr
;
1417 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1419 tos
= xfrm_get_tos(fl
, family
);
1426 for (; i
< nx
; i
++) {
1427 struct xfrm_dst
*xdst
= xfrm_alloc_dst(net
, family
);
1428 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1430 err
= PTR_ERR(xdst
);
1439 dst_prev
->child
= dst_clone(dst1
);
1440 dst1
->flags
|= DST_NOHASH
;
1444 dst_copy_metrics(dst1
, dst
);
1446 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1447 family
= xfrm
[i
]->props
.family
;
1448 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1456 dst1
->xfrm
= xfrm
[i
];
1457 xdst
->xfrm_genid
= xfrm
[i
]->genid
;
1459 dst1
->obsolete
= -1;
1460 dst1
->flags
|= DST_HOST
;
1461 dst1
->lastuse
= now
;
1463 dst1
->input
= dst_discard
;
1464 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1466 dst1
->next
= dst_prev
;
1469 header_len
+= xfrm
[i
]->props
.header_len
;
1470 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1471 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1472 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1475 dst_prev
->child
= dst
;
1483 /* Copy neighbour for reachability confirmation */
1484 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1486 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1487 xfrm_init_pmtu(dst_prev
);
1489 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1490 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1492 err
= xfrm_fill_dst(xdst
, dev
, fl
);
1496 dst_prev
->header_len
= header_len
;
1497 dst_prev
->trailer_len
= trailer_len
;
1498 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1499 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1507 xfrm_state_put(xfrm
[i
]);
1511 dst0
= ERR_PTR(err
);
1516 xfrm_dst_alloc_copy(void **target
, const void *src
, int size
)
1519 *target
= kmalloc(size
, GFP_ATOMIC
);
1523 memcpy(*target
, src
, size
);
1528 xfrm_dst_update_parent(struct dst_entry
*dst
, const struct xfrm_selector
*sel
)
1530 #ifdef CONFIG_XFRM_SUB_POLICY
1531 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1532 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1540 xfrm_dst_update_origin(struct dst_entry
*dst
, const struct flowi
*fl
)
1542 #ifdef CONFIG_XFRM_SUB_POLICY
1543 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1544 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1550 static int xfrm_expand_policies(const struct flowi
*fl
, u16 family
,
1551 struct xfrm_policy
**pols
,
1552 int *num_pols
, int *num_xfrms
)
1556 if (*num_pols
== 0 || !pols
[0]) {
1561 if (IS_ERR(pols
[0]))
1562 return PTR_ERR(pols
[0]);
1564 *num_xfrms
= pols
[0]->xfrm_nr
;
1566 #ifdef CONFIG_XFRM_SUB_POLICY
1567 if (pols
[0] && pols
[0]->action
== XFRM_POLICY_ALLOW
&&
1568 pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1569 pols
[1] = xfrm_policy_lookup_bytype(xp_net(pols
[0]),
1570 XFRM_POLICY_TYPE_MAIN
,
1574 if (IS_ERR(pols
[1])) {
1575 xfrm_pols_put(pols
, *num_pols
);
1576 return PTR_ERR(pols
[1]);
1579 (*num_xfrms
) += pols
[1]->xfrm_nr
;
1583 for (i
= 0; i
< *num_pols
; i
++) {
1584 if (pols
[i
]->action
!= XFRM_POLICY_ALLOW
) {
1594 static struct xfrm_dst
*
1595 xfrm_resolve_and_create_bundle(struct xfrm_policy
**pols
, int num_pols
,
1596 const struct flowi
*fl
, u16 family
,
1597 struct dst_entry
*dst_orig
)
1599 struct net
*net
= xp_net(pols
[0]);
1600 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1601 struct dst_entry
*dst
;
1602 struct xfrm_dst
*xdst
;
1605 /* Try to instantiate a bundle */
1606 err
= xfrm_tmpl_resolve(pols
, num_pols
, fl
, xfrm
, family
);
1608 if (err
!= 0 && err
!= -EAGAIN
)
1609 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1610 return ERR_PTR(err
);
1613 dst
= xfrm_bundle_create(pols
[0], xfrm
, err
, fl
, dst_orig
);
1615 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1616 return ERR_CAST(dst
);
1619 xdst
= (struct xfrm_dst
*)dst
;
1620 xdst
->num_xfrms
= err
;
1622 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1624 err
= xfrm_dst_update_origin(dst
, fl
);
1625 if (unlikely(err
)) {
1627 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1628 return ERR_PTR(err
);
1631 xdst
->num_pols
= num_pols
;
1632 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1633 xdst
->policy_genid
= atomic_read(&pols
[0]->genid
);
1638 static struct flow_cache_object
*
1639 xfrm_bundle_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
,
1640 struct flow_cache_object
*oldflo
, void *ctx
)
1642 struct dst_entry
*dst_orig
= (struct dst_entry
*)ctx
;
1643 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1644 struct xfrm_dst
*xdst
, *new_xdst
;
1645 int num_pols
= 0, num_xfrms
= 0, i
, err
, pol_dead
;
1647 /* Check if the policies from old bundle are usable */
1650 xdst
= container_of(oldflo
, struct xfrm_dst
, flo
);
1651 num_pols
= xdst
->num_pols
;
1652 num_xfrms
= xdst
->num_xfrms
;
1654 for (i
= 0; i
< num_pols
; i
++) {
1655 pols
[i
] = xdst
->pols
[i
];
1656 pol_dead
|= pols
[i
]->walk
.dead
;
1659 dst_free(&xdst
->u
.dst
);
1667 /* Resolve policies to use if we couldn't get them from
1668 * previous cache entry */
1671 pols
[0] = __xfrm_policy_lookup(net
, fl
, family
, dir
);
1672 err
= xfrm_expand_policies(fl
, family
, pols
,
1673 &num_pols
, &num_xfrms
);
1679 goto make_dummy_bundle
;
1682 new_xdst
= xfrm_resolve_and_create_bundle(pols
, num_pols
, fl
, family
, dst_orig
);
1683 if (IS_ERR(new_xdst
)) {
1684 err
= PTR_ERR(new_xdst
);
1688 goto make_dummy_bundle
;
1689 dst_hold(&xdst
->u
.dst
);
1691 } else if (new_xdst
== NULL
) {
1694 goto make_dummy_bundle
;
1695 xdst
->num_xfrms
= 0;
1696 dst_hold(&xdst
->u
.dst
);
1700 /* Kill the previous bundle */
1702 /* The policies were stolen for newly generated bundle */
1704 dst_free(&xdst
->u
.dst
);
1707 /* Flow cache does not have reference, it dst_free()'s,
1708 * but we do need to return one reference for original caller */
1709 dst_hold(&new_xdst
->u
.dst
);
1710 return &new_xdst
->flo
;
1713 /* We found policies, but there's no bundles to instantiate:
1714 * either because the policy blocks, has no transformations or
1715 * we could not build template (no xfrm_states).*/
1716 xdst
= xfrm_alloc_dst(net
, family
);
1718 xfrm_pols_put(pols
, num_pols
);
1719 return ERR_CAST(xdst
);
1721 xdst
->num_pols
= num_pols
;
1722 xdst
->num_xfrms
= num_xfrms
;
1723 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1725 dst_hold(&xdst
->u
.dst
);
1729 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1732 dst_free(&xdst
->u
.dst
);
1734 xfrm_pols_put(pols
, num_pols
);
1735 return ERR_PTR(err
);
1738 static struct dst_entry
*make_blackhole(struct net
*net
, u16 family
,
1739 struct dst_entry
*dst_orig
)
1741 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1742 struct dst_entry
*ret
;
1745 dst_release(dst_orig
);
1746 ret
= ERR_PTR(-EINVAL
);
1748 ret
= afinfo
->blackhole_route(net
, dst_orig
);
1750 xfrm_policy_put_afinfo(afinfo
);
1755 /* Main function: finds/creates a bundle for given flow.
1757 * At the moment we eat a raw IP route. Mostly to speed up lookups
1758 * on interfaces with disabled IPsec.
1760 struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
1761 const struct flowi
*fl
,
1762 struct sock
*sk
, int flags
)
1764 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1765 struct flow_cache_object
*flo
;
1766 struct xfrm_dst
*xdst
;
1767 struct dst_entry
*dst
, *route
;
1768 u16 family
= dst_orig
->ops
->family
;
1769 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1770 int i
, err
, num_pols
, num_xfrms
= 0, drop_pols
= 0;
1777 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1779 pols
[0] = xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1780 err
= xfrm_expand_policies(fl
, family
, pols
,
1781 &num_pols
, &num_xfrms
);
1786 if (num_xfrms
<= 0) {
1787 drop_pols
= num_pols
;
1791 xdst
= xfrm_resolve_and_create_bundle(
1795 xfrm_pols_put(pols
, num_pols
);
1796 err
= PTR_ERR(xdst
);
1798 } else if (xdst
== NULL
) {
1800 drop_pols
= num_pols
;
1804 spin_lock_bh(&xfrm_policy_sk_bundle_lock
);
1805 xdst
->u
.dst
.next
= xfrm_policy_sk_bundles
;
1806 xfrm_policy_sk_bundles
= &xdst
->u
.dst
;
1807 spin_unlock_bh(&xfrm_policy_sk_bundle_lock
);
1809 route
= xdst
->route
;
1814 /* To accelerate a bit... */
1815 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1816 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1819 flo
= flow_cache_lookup(net
, fl
, family
, dir
,
1820 xfrm_bundle_lookup
, dst_orig
);
1827 xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1829 num_pols
= xdst
->num_pols
;
1830 num_xfrms
= xdst
->num_xfrms
;
1831 memcpy(pols
, xdst
->pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1832 route
= xdst
->route
;
1836 if (route
== NULL
&& num_xfrms
> 0) {
1837 /* The only case when xfrm_bundle_lookup() returns a
1838 * bundle with null route, is when the template could
1839 * not be resolved. It means policies are there, but
1840 * bundle could not be created, since we don't yet
1841 * have the xfrm_state's. We need to wait for KM to
1842 * negotiate new SA's or bail out with error.*/
1843 if (net
->xfrm
.sysctl_larval_drop
) {
1844 /* EREMOTE tells the caller to generate
1845 * a one-shot blackhole route. */
1847 xfrm_pols_put(pols
, drop_pols
);
1848 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1850 return make_blackhole(net
, family
, dst_orig
);
1852 if (fl
->flags
& FLOWI_FLAG_CAN_SLEEP
) {
1853 DECLARE_WAITQUEUE(wait
, current
);
1855 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1856 set_current_state(TASK_INTERRUPTIBLE
);
1858 set_current_state(TASK_RUNNING
);
1859 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1861 if (!signal_pending(current
)) {
1870 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1878 if ((flags
& XFRM_LOOKUP_ICMP
) &&
1879 !(pols
[0]->flags
& XFRM_POLICY_ICMP
)) {
1884 for (i
= 0; i
< num_pols
; i
++)
1885 pols
[i
]->curlft
.use_time
= get_seconds();
1887 if (num_xfrms
< 0) {
1888 /* Prohibit the flow */
1889 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
1892 } else if (num_xfrms
> 0) {
1893 /* Flow transformed */
1894 dst_release(dst_orig
);
1896 /* Flow passes untransformed */
1901 xfrm_pols_put(pols
, drop_pols
);
1905 if (!(flags
& XFRM_LOOKUP_ICMP
)) {
1913 dst_release(dst_orig
);
1914 xfrm_pols_put(pols
, drop_pols
);
1915 return ERR_PTR(err
);
1917 EXPORT_SYMBOL(xfrm_lookup
);
1920 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, const struct flowi
*fl
)
1922 struct xfrm_state
*x
;
1924 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1926 x
= skb
->sp
->xvec
[idx
];
1927 if (!x
->type
->reject
)
1929 return x
->type
->reject(x
, skb
, fl
);
1932 /* When skb is transformed back to its "native" form, we have to
1933 * check policy restrictions. At the moment we make this in maximally
1934 * stupid way. Shame on me. :-) Of course, connected sockets must
1935 * have policy cached at them.
1939 xfrm_state_ok(const struct xfrm_tmpl
*tmpl
, const struct xfrm_state
*x
,
1940 unsigned short family
)
1942 if (xfrm_state_kern(x
))
1943 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1944 return x
->id
.proto
== tmpl
->id
.proto
&&
1945 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1946 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1947 x
->props
.mode
== tmpl
->mode
&&
1948 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1949 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1950 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1951 xfrm_state_addr_cmp(tmpl
, x
, family
));
1955 * 0 or more than 0 is returned when validation is succeeded (either bypass
1956 * because of optional transport mode, or next index of the mathced secpath
1957 * state with the template.
1958 * -1 is returned when no matching template is found.
1959 * Otherwise "-2 - errored_index" is returned.
1962 xfrm_policy_ok(const struct xfrm_tmpl
*tmpl
, const struct sec_path
*sp
, int start
,
1963 unsigned short family
)
1967 if (tmpl
->optional
) {
1968 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1972 for (; idx
< sp
->len
; idx
++) {
1973 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1975 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1984 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1985 unsigned int family
, int reverse
)
1987 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1990 if (unlikely(afinfo
== NULL
))
1991 return -EAFNOSUPPORT
;
1993 afinfo
->decode_session(skb
, fl
, reverse
);
1994 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1995 xfrm_policy_put_afinfo(afinfo
);
1998 EXPORT_SYMBOL(__xfrm_decode_session
);
2000 static inline int secpath_has_nontransport(const struct sec_path
*sp
, int k
, int *idxp
)
2002 for (; k
< sp
->len
; k
++) {
2003 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
2012 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
2013 unsigned short family
)
2015 struct net
*net
= dev_net(skb
->dev
);
2016 struct xfrm_policy
*pol
;
2017 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2026 reverse
= dir
& ~XFRM_POLICY_MASK
;
2027 dir
&= XFRM_POLICY_MASK
;
2028 fl_dir
= policy_to_flow_dir(dir
);
2030 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
2031 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
2035 nf_nat_decode_session(skb
, &fl
, family
);
2037 /* First, check used SA against their selectors. */
2041 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
2042 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
2043 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
2044 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
2051 if (sk
&& sk
->sk_policy
[dir
]) {
2052 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
2054 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2060 struct flow_cache_object
*flo
;
2062 flo
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
2063 xfrm_policy_lookup
, NULL
);
2064 if (IS_ERR_OR_NULL(flo
))
2065 pol
= ERR_CAST(flo
);
2067 pol
= container_of(flo
, struct xfrm_policy
, flo
);
2071 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2076 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
2077 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2078 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
2084 pol
->curlft
.use_time
= get_seconds();
2088 #ifdef CONFIG_XFRM_SUB_POLICY
2089 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
2090 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
2094 if (IS_ERR(pols
[1])) {
2095 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2098 pols
[1]->curlft
.use_time
= get_seconds();
2104 if (pol
->action
== XFRM_POLICY_ALLOW
) {
2105 struct sec_path
*sp
;
2106 static struct sec_path dummy
;
2107 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
2108 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
2109 struct xfrm_tmpl
**tpp
= tp
;
2113 if ((sp
= skb
->sp
) == NULL
)
2116 for (pi
= 0; pi
< npols
; pi
++) {
2117 if (pols
[pi
] != pol
&&
2118 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
2119 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2122 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2123 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
2126 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2127 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2131 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2135 /* For each tunnel xfrm, find the first matching tmpl.
2136 * For each tmpl before that, find corresponding xfrm.
2137 * Order is _important_. Later we will implement
2138 * some barriers, but at the moment barriers
2139 * are implied between each two transformations.
2141 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2142 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2145 /* "-2 - errored_index" returned */
2147 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2152 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2153 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2157 xfrm_pols_put(pols
, npols
);
2160 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2163 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2165 xfrm_pols_put(pols
, npols
);
2168 EXPORT_SYMBOL(__xfrm_policy_check
);
2170 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2172 struct net
*net
= dev_net(skb
->dev
);
2174 struct dst_entry
*dst
;
2177 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2178 XFRM_INC_STATS(net
, LINUX_MIB_XFRMFWDHDRERROR
);
2184 dst
= xfrm_lookup(net
, skb_dst(skb
), &fl
, NULL
, 0);
2189 skb_dst_set(skb
, dst
);
2192 EXPORT_SYMBOL(__xfrm_route_forward
);
2194 /* Optimize later using cookies and generation ids. */
2196 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2198 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2199 * to "-1" to force all XFRM destinations to get validated by
2200 * dst_ops->check on every use. We do this because when a
2201 * normal route referenced by an XFRM dst is obsoleted we do
2202 * not go looking around for all parent referencing XFRM dsts
2203 * so that we can invalidate them. It is just too much work.
2204 * Instead we make the checks here on every use. For example:
2206 * XFRM dst A --> IPv4 dst X
2208 * X is the "xdst->route" of A (X is also the "dst->path" of A
2209 * in this example). If X is marked obsolete, "A" will not
2210 * notice. That's what we are validating here via the
2211 * stale_bundle() check.
2213 * When a policy's bundle is pruned, we dst_free() the XFRM
2214 * dst which causes it's ->obsolete field to be set to a
2215 * positive non-zero integer. If an XFRM dst has been pruned
2216 * like this, we want to force a new route lookup.
2218 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2224 static int stale_bundle(struct dst_entry
*dst
)
2226 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
);
2229 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2231 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2232 dst
->dev
= dev_net(dev
)->loopback_dev
;
2237 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2239 static void xfrm_link_failure(struct sk_buff
*skb
)
2241 /* Impossible. Such dst must be popped before reaches point of failure. */
2244 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2247 if (dst
->obsolete
) {
2255 static void __xfrm_garbage_collect(struct net
*net
)
2257 struct dst_entry
*head
, *next
;
2261 spin_lock_bh(&xfrm_policy_sk_bundle_lock
);
2262 head
= xfrm_policy_sk_bundles
;
2263 xfrm_policy_sk_bundles
= NULL
;
2264 spin_unlock_bh(&xfrm_policy_sk_bundle_lock
);
2273 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2276 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2277 u32 pmtu
, route_mtu_cached
;
2279 pmtu
= dst_mtu(dst
->child
);
2280 xdst
->child_mtu_cached
= pmtu
;
2282 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2284 route_mtu_cached
= dst_mtu(xdst
->route
);
2285 xdst
->route_mtu_cached
= route_mtu_cached
;
2287 if (pmtu
> route_mtu_cached
)
2288 pmtu
= route_mtu_cached
;
2290 dst_metric_set(dst
, RTAX_MTU
, pmtu
);
2291 } while ((dst
= dst
->next
));
2294 /* Check that the bundle accepts the flow and its components are
2298 static int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2299 const struct flowi
*fl
, int family
)
2301 struct dst_entry
*dst
= &first
->u
.dst
;
2302 struct xfrm_dst
*last
;
2305 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2306 (dst
->dev
&& !netif_running(dst
->dev
)))
2308 #ifdef CONFIG_XFRM_SUB_POLICY
2310 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2312 if (first
->partner
&&
2313 !xfrm_selector_match(first
->partner
, fl
, family
))
2321 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2323 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2326 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2328 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2330 if (xdst
->xfrm_genid
!= dst
->xfrm
->genid
)
2332 if (xdst
->num_pols
> 0 &&
2333 xdst
->policy_genid
!= atomic_read(&xdst
->pols
[0]->genid
))
2336 mtu
= dst_mtu(dst
->child
);
2337 if (xdst
->child_mtu_cached
!= mtu
) {
2339 xdst
->child_mtu_cached
= mtu
;
2342 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2344 mtu
= dst_mtu(xdst
->route
);
2345 if (xdst
->route_mtu_cached
!= mtu
) {
2347 xdst
->route_mtu_cached
= mtu
;
2351 } while (dst
->xfrm
);
2356 mtu
= last
->child_mtu_cached
;
2360 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2361 if (mtu
> last
->route_mtu_cached
)
2362 mtu
= last
->route_mtu_cached
;
2363 dst_metric_set(dst
, RTAX_MTU
, mtu
);
2368 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2369 last
->child_mtu_cached
= mtu
;
2375 static unsigned int xfrm_default_advmss(const struct dst_entry
*dst
)
2377 return dst_metric_advmss(dst
->path
);
2380 static unsigned int xfrm_default_mtu(const struct dst_entry
*dst
)
2382 return dst_mtu(dst
->path
);
2385 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2389 if (unlikely(afinfo
== NULL
))
2391 if (unlikely(afinfo
->family
>= NPROTO
))
2392 return -EAFNOSUPPORT
;
2393 write_lock_bh(&xfrm_policy_afinfo_lock
);
2394 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2397 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2398 if (likely(dst_ops
->kmem_cachep
== NULL
))
2399 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2400 if (likely(dst_ops
->check
== NULL
))
2401 dst_ops
->check
= xfrm_dst_check
;
2402 if (likely(dst_ops
->default_advmss
== NULL
))
2403 dst_ops
->default_advmss
= xfrm_default_advmss
;
2404 if (likely(dst_ops
->default_mtu
== NULL
))
2405 dst_ops
->default_mtu
= xfrm_default_mtu
;
2406 if (likely(dst_ops
->negative_advice
== NULL
))
2407 dst_ops
->negative_advice
= xfrm_negative_advice
;
2408 if (likely(dst_ops
->link_failure
== NULL
))
2409 dst_ops
->link_failure
= xfrm_link_failure
;
2410 if (likely(afinfo
->garbage_collect
== NULL
))
2411 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2412 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2414 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2418 struct dst_ops
*xfrm_dst_ops
;
2420 switch (afinfo
->family
) {
2422 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2424 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2426 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2432 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2438 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2440 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2443 if (unlikely(afinfo
== NULL
))
2445 if (unlikely(afinfo
->family
>= NPROTO
))
2446 return -EAFNOSUPPORT
;
2447 write_lock_bh(&xfrm_policy_afinfo_lock
);
2448 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2449 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2452 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2453 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2454 dst_ops
->kmem_cachep
= NULL
;
2455 dst_ops
->check
= NULL
;
2456 dst_ops
->negative_advice
= NULL
;
2457 dst_ops
->link_failure
= NULL
;
2458 afinfo
->garbage_collect
= NULL
;
2461 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2464 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2466 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2468 struct xfrm_policy_afinfo
*afinfo
;
2470 read_lock_bh(&xfrm_policy_afinfo_lock
);
2471 afinfo
= xfrm_policy_afinfo
[AF_INET
];
2473 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2474 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2475 afinfo
= xfrm_policy_afinfo
[AF_INET6
];
2477 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2479 read_unlock_bh(&xfrm_policy_afinfo_lock
);
2482 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2484 struct xfrm_policy_afinfo
*afinfo
;
2485 if (unlikely(family
>= NPROTO
))
2487 read_lock(&xfrm_policy_afinfo_lock
);
2488 afinfo
= xfrm_policy_afinfo
[family
];
2489 if (unlikely(!afinfo
))
2490 read_unlock(&xfrm_policy_afinfo_lock
);
2494 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2496 read_unlock(&xfrm_policy_afinfo_lock
);
2499 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2501 struct net_device
*dev
= ptr
;
2505 __xfrm_garbage_collect(dev_net(dev
));
2510 static struct notifier_block xfrm_dev_notifier
= {
2511 .notifier_call
= xfrm_dev_event
,
2514 #ifdef CONFIG_XFRM_STATISTICS
2515 static int __net_init
xfrm_statistics_init(struct net
*net
)
2519 if (snmp_mib_init((void __percpu
**)net
->mib
.xfrm_statistics
,
2520 sizeof(struct linux_xfrm_mib
),
2521 __alignof__(struct linux_xfrm_mib
)) < 0)
2523 rv
= xfrm_proc_init(net
);
2525 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2529 static void xfrm_statistics_fini(struct net
*net
)
2531 xfrm_proc_fini(net
);
2532 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2535 static int __net_init
xfrm_statistics_init(struct net
*net
)
2540 static void xfrm_statistics_fini(struct net
*net
)
2545 static int __net_init
xfrm_policy_init(struct net
*net
)
2547 unsigned int hmask
, sz
;
2550 if (net_eq(net
, &init_net
))
2551 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2552 sizeof(struct xfrm_dst
),
2553 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2557 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2559 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2560 if (!net
->xfrm
.policy_byidx
)
2562 net
->xfrm
.policy_idx_hmask
= hmask
;
2564 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2565 struct xfrm_policy_hash
*htab
;
2567 net
->xfrm
.policy_count
[dir
] = 0;
2568 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2570 htab
= &net
->xfrm
.policy_bydst
[dir
];
2571 htab
->table
= xfrm_hash_alloc(sz
);
2574 htab
->hmask
= hmask
;
2577 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2578 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2579 if (net_eq(net
, &init_net
))
2580 register_netdevice_notifier(&xfrm_dev_notifier
);
2584 for (dir
--; dir
>= 0; dir
--) {
2585 struct xfrm_policy_hash
*htab
;
2587 htab
= &net
->xfrm
.policy_bydst
[dir
];
2588 xfrm_hash_free(htab
->table
, sz
);
2590 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2595 static void xfrm_policy_fini(struct net
*net
)
2597 struct xfrm_audit audit_info
;
2601 flush_work(&net
->xfrm
.policy_hash_work
);
2602 #ifdef CONFIG_XFRM_SUB_POLICY
2603 audit_info
.loginuid
= -1;
2604 audit_info
.sessionid
= -1;
2605 audit_info
.secid
= 0;
2606 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2608 audit_info
.loginuid
= -1;
2609 audit_info
.sessionid
= -1;
2610 audit_info
.secid
= 0;
2611 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2613 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2615 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2616 struct xfrm_policy_hash
*htab
;
2618 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2620 htab
= &net
->xfrm
.policy_bydst
[dir
];
2621 sz
= (htab
->hmask
+ 1);
2622 WARN_ON(!hlist_empty(htab
->table
));
2623 xfrm_hash_free(htab
->table
, sz
);
2626 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2627 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2628 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2631 static int __net_init
xfrm_net_init(struct net
*net
)
2635 rv
= xfrm_statistics_init(net
);
2637 goto out_statistics
;
2638 rv
= xfrm_state_init(net
);
2641 rv
= xfrm_policy_init(net
);
2644 xfrm_dst_ops_init(net
);
2645 rv
= xfrm_sysctl_init(net
);
2651 xfrm_policy_fini(net
);
2653 xfrm_state_fini(net
);
2655 xfrm_statistics_fini(net
);
2660 static void __net_exit
xfrm_net_exit(struct net
*net
)
2662 xfrm_sysctl_fini(net
);
2663 xfrm_policy_fini(net
);
2664 xfrm_state_fini(net
);
2665 xfrm_statistics_fini(net
);
2668 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2669 .init
= xfrm_net_init
,
2670 .exit
= xfrm_net_exit
,
2673 void __init
xfrm_init(void)
2675 register_pernet_subsys(&xfrm_net_ops
);
2679 #ifdef CONFIG_AUDITSYSCALL
2680 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2681 struct audit_buffer
*audit_buf
)
2683 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2684 struct xfrm_selector
*sel
= &xp
->selector
;
2687 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2688 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2690 switch(sel
->family
) {
2692 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2693 if (sel
->prefixlen_s
!= 32)
2694 audit_log_format(audit_buf
, " src_prefixlen=%d",
2696 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2697 if (sel
->prefixlen_d
!= 32)
2698 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2702 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2703 if (sel
->prefixlen_s
!= 128)
2704 audit_log_format(audit_buf
, " src_prefixlen=%d",
2706 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2707 if (sel
->prefixlen_d
!= 128)
2708 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2714 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2715 uid_t auid
, u32 sessionid
, u32 secid
)
2717 struct audit_buffer
*audit_buf
;
2719 audit_buf
= xfrm_audit_start("SPD-add");
2720 if (audit_buf
== NULL
)
2722 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2723 audit_log_format(audit_buf
, " res=%u", result
);
2724 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2725 audit_log_end(audit_buf
);
2727 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2729 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2730 uid_t auid
, u32 sessionid
, u32 secid
)
2732 struct audit_buffer
*audit_buf
;
2734 audit_buf
= xfrm_audit_start("SPD-delete");
2735 if (audit_buf
== NULL
)
2737 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2738 audit_log_format(audit_buf
, " res=%u", result
);
2739 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2740 audit_log_end(audit_buf
);
2742 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2745 #ifdef CONFIG_XFRM_MIGRATE
2746 static int xfrm_migrate_selector_match(const struct xfrm_selector
*sel_cmp
,
2747 const struct xfrm_selector
*sel_tgt
)
2749 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2750 if (sel_tgt
->family
== sel_cmp
->family
&&
2751 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2752 sel_cmp
->family
) == 0 &&
2753 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2754 sel_cmp
->family
) == 0 &&
2755 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2756 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2760 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2767 static struct xfrm_policy
* xfrm_migrate_policy_find(const struct xfrm_selector
*sel
,
2770 struct xfrm_policy
*pol
, *ret
= NULL
;
2771 struct hlist_node
*entry
;
2772 struct hlist_head
*chain
;
2775 read_lock_bh(&xfrm_policy_lock
);
2776 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2777 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2778 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2779 pol
->type
== type
) {
2781 priority
= ret
->priority
;
2785 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2786 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2787 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2788 pol
->type
== type
&&
2789 pol
->priority
< priority
) {
2798 read_unlock_bh(&xfrm_policy_lock
);
2803 static int migrate_tmpl_match(const struct xfrm_migrate
*m
, const struct xfrm_tmpl
*t
)
2807 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2808 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2810 case XFRM_MODE_TUNNEL
:
2811 case XFRM_MODE_BEET
:
2812 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2813 m
->old_family
) == 0 &&
2814 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2815 m
->old_family
) == 0) {
2819 case XFRM_MODE_TRANSPORT
:
2820 /* in case of transport mode, template does not store
2821 any IP addresses, hence we just compare mode and
2832 /* update endpoint address(es) of template(s) */
2833 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2834 struct xfrm_migrate
*m
, int num_migrate
)
2836 struct xfrm_migrate
*mp
;
2839 write_lock_bh(&pol
->lock
);
2840 if (unlikely(pol
->walk
.dead
)) {
2841 /* target policy has been deleted */
2842 write_unlock_bh(&pol
->lock
);
2846 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2847 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2848 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2851 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2852 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2854 /* update endpoints */
2855 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2856 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2857 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2858 sizeof(pol
->xfrm_vec
[i
].saddr
));
2859 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2861 atomic_inc(&pol
->genid
);
2865 write_unlock_bh(&pol
->lock
);
2873 static int xfrm_migrate_check(const struct xfrm_migrate
*m
, int num_migrate
)
2877 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2880 for (i
= 0; i
< num_migrate
; i
++) {
2881 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2882 m
[i
].old_family
) == 0) &&
2883 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2884 m
[i
].old_family
) == 0))
2886 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2887 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2890 /* check if there is any duplicated entry */
2891 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2892 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2893 sizeof(m
[i
].old_daddr
)) &&
2894 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2895 sizeof(m
[i
].old_saddr
)) &&
2896 m
[i
].proto
== m
[j
].proto
&&
2897 m
[i
].mode
== m
[j
].mode
&&
2898 m
[i
].reqid
== m
[j
].reqid
&&
2899 m
[i
].old_family
== m
[j
].old_family
)
2907 int xfrm_migrate(const struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2908 struct xfrm_migrate
*m
, int num_migrate
,
2909 struct xfrm_kmaddress
*k
)
2911 int i
, err
, nx_cur
= 0, nx_new
= 0;
2912 struct xfrm_policy
*pol
= NULL
;
2913 struct xfrm_state
*x
, *xc
;
2914 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2915 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2916 struct xfrm_migrate
*mp
;
2918 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2921 /* Stage 1 - find policy */
2922 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2927 /* Stage 2 - find and update state(s) */
2928 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2929 if ((x
= xfrm_migrate_state_find(mp
))) {
2932 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2942 /* Stage 3 - update policy */
2943 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2946 /* Stage 4 - delete old state(s) */
2948 xfrm_states_put(x_cur
, nx_cur
);
2949 xfrm_states_delete(x_cur
, nx_cur
);
2952 /* Stage 5 - announce */
2953 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2965 xfrm_states_put(x_cur
, nx_cur
);
2967 xfrm_states_delete(x_new
, nx_new
);
2971 EXPORT_SYMBOL(xfrm_migrate
);