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(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(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(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 xfrm_address_t
*saddr
,
96 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
, struct xfrm_selector
*sel
, unsigned short family
, int dir
)
316 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
317 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
319 return (hash
== hmask
+ 1 ?
320 &net
->xfrm
.policy_inexact
[dir
] :
321 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
324 static struct hlist_head
*policy_hash_direct(struct net
*net
, xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
326 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
327 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
329 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
332 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
333 struct hlist_head
*ndsttable
,
334 unsigned int nhashmask
)
336 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
337 struct xfrm_policy
*pol
;
341 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
344 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
345 pol
->family
, nhashmask
);
348 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
354 hlist_add_after(entry0
, &pol
->bydst
);
358 if (!hlist_empty(list
)) {
364 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
365 struct hlist_head
*nidxtable
,
366 unsigned int nhashmask
)
368 struct hlist_node
*entry
, *tmp
;
369 struct xfrm_policy
*pol
;
371 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
374 h
= __idx_hash(pol
->index
, nhashmask
);
375 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
379 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
381 return ((old_hmask
+ 1) << 1) - 1;
384 static void xfrm_bydst_resize(struct net
*net
, int dir
)
386 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
387 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
388 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
389 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
390 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
396 write_lock_bh(&xfrm_policy_lock
);
398 for (i
= hmask
; i
>= 0; i
--)
399 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
401 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
402 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
404 write_unlock_bh(&xfrm_policy_lock
);
406 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
409 static void xfrm_byidx_resize(struct net
*net
, int total
)
411 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
412 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
413 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
414 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
415 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
421 write_lock_bh(&xfrm_policy_lock
);
423 for (i
= hmask
; i
>= 0; i
--)
424 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
426 net
->xfrm
.policy_byidx
= nidx
;
427 net
->xfrm
.policy_idx_hmask
= nhashmask
;
429 write_unlock_bh(&xfrm_policy_lock
);
431 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
434 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
436 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
437 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
442 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
449 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
451 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
453 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
460 void xfrm_spd_getinfo(struct net
*net
, struct xfrmk_spdinfo
*si
)
462 read_lock_bh(&xfrm_policy_lock
);
463 si
->incnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
];
464 si
->outcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
];
465 si
->fwdcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
];
466 si
->inscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
467 si
->outscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
468 si
->fwdscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
469 si
->spdhcnt
= net
->xfrm
.policy_idx_hmask
;
470 si
->spdhmcnt
= xfrm_policy_hashmax
;
471 read_unlock_bh(&xfrm_policy_lock
);
473 EXPORT_SYMBOL(xfrm_spd_getinfo
);
475 static DEFINE_MUTEX(hash_resize_mutex
);
476 static void xfrm_hash_resize(struct work_struct
*work
)
478 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
481 mutex_lock(&hash_resize_mutex
);
484 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
485 if (xfrm_bydst_should_resize(net
, dir
, &total
))
486 xfrm_bydst_resize(net
, dir
);
488 if (xfrm_byidx_should_resize(net
, total
))
489 xfrm_byidx_resize(net
, total
);
491 mutex_unlock(&hash_resize_mutex
);
494 /* Generate new index... KAME seems to generate them ordered by cost
495 * of an absolute inpredictability of ordering of rules. This will not pass. */
496 static u32
xfrm_gen_index(struct net
*net
, int dir
)
498 static u32 idx_generator
;
501 struct hlist_node
*entry
;
502 struct hlist_head
*list
;
503 struct xfrm_policy
*p
;
507 idx
= (idx_generator
| dir
);
511 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
513 hlist_for_each_entry(p
, entry
, list
, byidx
) {
514 if (p
->index
== idx
) {
524 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
526 u32
*p1
= (u32
*) s1
;
527 u32
*p2
= (u32
*) s2
;
528 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
531 for (i
= 0; i
< len
; i
++) {
539 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
541 struct net
*net
= xp_net(policy
);
542 struct xfrm_policy
*pol
;
543 struct xfrm_policy
*delpol
;
544 struct hlist_head
*chain
;
545 struct hlist_node
*entry
, *newpos
;
546 u32 mark
= policy
->mark
.v
& policy
->mark
.m
;
548 write_lock_bh(&xfrm_policy_lock
);
549 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
552 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
553 if (pol
->type
== policy
->type
&&
554 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
555 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
556 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
559 write_unlock_bh(&xfrm_policy_lock
);
563 if (policy
->priority
> pol
->priority
)
565 } else if (policy
->priority
>= pol
->priority
) {
566 newpos
= &pol
->bydst
;
573 hlist_add_after(newpos
, &policy
->bydst
);
575 hlist_add_head(&policy
->bydst
, chain
);
576 xfrm_pol_hold(policy
);
577 net
->xfrm
.policy_count
[dir
]++;
578 atomic_inc(&flow_cache_genid
);
580 __xfrm_policy_unlink(delpol
, dir
);
581 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
582 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
583 policy
->curlft
.add_time
= get_seconds();
584 policy
->curlft
.use_time
= 0;
585 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
586 xfrm_pol_hold(policy
);
587 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
588 write_unlock_bh(&xfrm_policy_lock
);
591 xfrm_policy_kill(delpol
);
592 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
593 schedule_work(&net
->xfrm
.policy_hash_work
);
597 EXPORT_SYMBOL(xfrm_policy_insert
);
599 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u32 mark
, u8 type
,
600 int dir
, struct xfrm_selector
*sel
,
601 struct xfrm_sec_ctx
*ctx
, int delete,
604 struct xfrm_policy
*pol
, *ret
;
605 struct hlist_head
*chain
;
606 struct hlist_node
*entry
;
609 write_lock_bh(&xfrm_policy_lock
);
610 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
612 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
613 if (pol
->type
== type
&&
614 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
615 !selector_cmp(sel
, &pol
->selector
) &&
616 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
619 *err
= security_xfrm_policy_delete(
622 write_unlock_bh(&xfrm_policy_lock
);
625 __xfrm_policy_unlink(pol
, dir
);
631 write_unlock_bh(&xfrm_policy_lock
);
634 xfrm_policy_kill(ret
);
637 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
639 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u32 mark
, u8 type
,
640 int dir
, u32 id
, int delete, int *err
)
642 struct xfrm_policy
*pol
, *ret
;
643 struct hlist_head
*chain
;
644 struct hlist_node
*entry
;
647 if (xfrm_policy_id2dir(id
) != dir
)
651 write_lock_bh(&xfrm_policy_lock
);
652 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
654 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
655 if (pol
->type
== type
&& pol
->index
== id
&&
656 (mark
& pol
->mark
.m
) == pol
->mark
.v
) {
659 *err
= security_xfrm_policy_delete(
662 write_unlock_bh(&xfrm_policy_lock
);
665 __xfrm_policy_unlink(pol
, dir
);
671 write_unlock_bh(&xfrm_policy_lock
);
674 xfrm_policy_kill(ret
);
677 EXPORT_SYMBOL(xfrm_policy_byid
);
679 #ifdef CONFIG_SECURITY_NETWORK_XFRM
681 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
685 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
686 struct xfrm_policy
*pol
;
687 struct hlist_node
*entry
;
690 hlist_for_each_entry(pol
, entry
,
691 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
692 if (pol
->type
!= type
)
694 err
= security_xfrm_policy_delete(pol
->security
);
696 xfrm_audit_policy_delete(pol
, 0,
697 audit_info
->loginuid
,
698 audit_info
->sessionid
,
703 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
704 hlist_for_each_entry(pol
, entry
,
705 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
707 if (pol
->type
!= type
)
709 err
= security_xfrm_policy_delete(
712 xfrm_audit_policy_delete(pol
, 0,
713 audit_info
->loginuid
,
714 audit_info
->sessionid
,
725 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
731 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
733 int dir
, err
= 0, cnt
= 0;
735 write_lock_bh(&xfrm_policy_lock
);
737 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
741 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
742 struct xfrm_policy
*pol
;
743 struct hlist_node
*entry
;
747 hlist_for_each_entry(pol
, entry
,
748 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
749 if (pol
->type
!= type
)
751 __xfrm_policy_unlink(pol
, dir
);
752 write_unlock_bh(&xfrm_policy_lock
);
755 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
756 audit_info
->sessionid
,
759 xfrm_policy_kill(pol
);
761 write_lock_bh(&xfrm_policy_lock
);
765 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
767 hlist_for_each_entry(pol
, entry
,
768 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
770 if (pol
->type
!= type
)
772 __xfrm_policy_unlink(pol
, dir
);
773 write_unlock_bh(&xfrm_policy_lock
);
776 xfrm_audit_policy_delete(pol
, 1,
777 audit_info
->loginuid
,
778 audit_info
->sessionid
,
780 xfrm_policy_kill(pol
);
782 write_lock_bh(&xfrm_policy_lock
);
791 write_unlock_bh(&xfrm_policy_lock
);
794 EXPORT_SYMBOL(xfrm_policy_flush
);
796 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
797 int (*func
)(struct xfrm_policy
*, int, int, void*),
800 struct xfrm_policy
*pol
;
801 struct xfrm_policy_walk_entry
*x
;
804 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
805 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
808 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
811 write_lock_bh(&xfrm_policy_lock
);
812 if (list_empty(&walk
->walk
.all
))
813 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
815 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
816 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
819 pol
= container_of(x
, struct xfrm_policy
, walk
);
820 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
821 walk
->type
!= pol
->type
)
823 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
826 list_move_tail(&walk
->walk
.all
, &x
->all
);
831 if (walk
->seq
== 0) {
835 list_del_init(&walk
->walk
.all
);
837 write_unlock_bh(&xfrm_policy_lock
);
840 EXPORT_SYMBOL(xfrm_policy_walk
);
842 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
844 INIT_LIST_HEAD(&walk
->walk
.all
);
849 EXPORT_SYMBOL(xfrm_policy_walk_init
);
851 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
853 if (list_empty(&walk
->walk
.all
))
856 write_lock_bh(&xfrm_policy_lock
);
857 list_del(&walk
->walk
.all
);
858 write_unlock_bh(&xfrm_policy_lock
);
860 EXPORT_SYMBOL(xfrm_policy_walk_done
);
863 * Find policy to apply to this flow.
865 * Returns 0 if policy found, else an -errno.
867 static int xfrm_policy_match(struct xfrm_policy
*pol
, const struct flowi
*fl
,
868 u8 type
, u16 family
, int dir
)
870 struct xfrm_selector
*sel
= &pol
->selector
;
871 int match
, ret
= -ESRCH
;
873 if (pol
->family
!= family
||
874 (fl
->mark
& pol
->mark
.m
) != pol
->mark
.v
||
878 match
= xfrm_selector_match(sel
, fl
, family
);
880 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
886 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
887 const struct flowi
*fl
,
891 struct xfrm_policy
*pol
, *ret
;
892 xfrm_address_t
*daddr
, *saddr
;
893 struct hlist_node
*entry
;
894 struct hlist_head
*chain
;
897 daddr
= xfrm_flowi_daddr(fl
, family
);
898 saddr
= xfrm_flowi_saddr(fl
, family
);
899 if (unlikely(!daddr
|| !saddr
))
902 read_lock_bh(&xfrm_policy_lock
);
903 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
905 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
906 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
916 priority
= ret
->priority
;
920 chain
= &net
->xfrm
.policy_inexact
[dir
];
921 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
922 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
930 } else if (pol
->priority
< priority
) {
938 read_unlock_bh(&xfrm_policy_lock
);
943 static struct xfrm_policy
*
944 __xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
)
946 #ifdef CONFIG_XFRM_SUB_POLICY
947 struct xfrm_policy
*pol
;
949 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
953 return xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
956 static struct flow_cache_object
*
957 xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
,
958 u8 dir
, struct flow_cache_object
*old_obj
, void *ctx
)
960 struct xfrm_policy
*pol
;
963 xfrm_pol_put(container_of(old_obj
, struct xfrm_policy
, flo
));
965 pol
= __xfrm_policy_lookup(net
, fl
, family
, dir
);
966 if (IS_ERR_OR_NULL(pol
))
967 return ERR_CAST(pol
);
969 /* Resolver returns two references:
970 * one for cache and one for caller of flow_cache_lookup() */
976 static inline int policy_to_flow_dir(int dir
)
978 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
979 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
980 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
986 case XFRM_POLICY_OUT
:
988 case XFRM_POLICY_FWD
:
993 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
,
994 const struct flowi
*fl
)
996 struct xfrm_policy
*pol
;
998 read_lock_bh(&xfrm_policy_lock
);
999 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1000 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1005 if ((sk
->sk_mark
& pol
->mark
.m
) != pol
->mark
.v
) {
1009 err
= security_xfrm_policy_lookup(pol
->security
,
1011 policy_to_flow_dir(dir
));
1014 else if (err
== -ESRCH
)
1022 read_unlock_bh(&xfrm_policy_lock
);
1026 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1028 struct net
*net
= xp_net(pol
);
1029 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1032 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1033 hlist_add_head(&pol
->bydst
, chain
);
1034 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1035 net
->xfrm
.policy_count
[dir
]++;
1038 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1039 schedule_work(&net
->xfrm
.policy_hash_work
);
1042 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1045 struct net
*net
= xp_net(pol
);
1047 if (hlist_unhashed(&pol
->bydst
))
1050 hlist_del(&pol
->bydst
);
1051 hlist_del(&pol
->byidx
);
1052 list_del(&pol
->walk
.all
);
1053 net
->xfrm
.policy_count
[dir
]--;
1058 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1060 write_lock_bh(&xfrm_policy_lock
);
1061 pol
= __xfrm_policy_unlink(pol
, dir
);
1062 write_unlock_bh(&xfrm_policy_lock
);
1064 xfrm_policy_kill(pol
);
1069 EXPORT_SYMBOL(xfrm_policy_delete
);
1071 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1073 struct net
*net
= xp_net(pol
);
1074 struct xfrm_policy
*old_pol
;
1076 #ifdef CONFIG_XFRM_SUB_POLICY
1077 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1081 write_lock_bh(&xfrm_policy_lock
);
1082 old_pol
= sk
->sk_policy
[dir
];
1083 sk
->sk_policy
[dir
] = pol
;
1085 pol
->curlft
.add_time
= get_seconds();
1086 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1087 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1090 /* Unlinking succeeds always. This is the only function
1091 * allowed to delete or replace socket policy.
1093 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1094 write_unlock_bh(&xfrm_policy_lock
);
1097 xfrm_policy_kill(old_pol
);
1102 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1104 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1107 newp
->selector
= old
->selector
;
1108 if (security_xfrm_policy_clone(old
->security
,
1111 return NULL
; /* ENOMEM */
1113 newp
->lft
= old
->lft
;
1114 newp
->curlft
= old
->curlft
;
1115 newp
->mark
= old
->mark
;
1116 newp
->action
= old
->action
;
1117 newp
->flags
= old
->flags
;
1118 newp
->xfrm_nr
= old
->xfrm_nr
;
1119 newp
->index
= old
->index
;
1120 newp
->type
= old
->type
;
1121 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1122 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1123 write_lock_bh(&xfrm_policy_lock
);
1124 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1125 write_unlock_bh(&xfrm_policy_lock
);
1131 int __xfrm_sk_clone_policy(struct sock
*sk
)
1133 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1134 *p1
= sk
->sk_policy
[1];
1136 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1137 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1139 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1145 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1146 unsigned short family
)
1149 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1151 if (unlikely(afinfo
== NULL
))
1153 err
= afinfo
->get_saddr(net
, local
, remote
);
1154 xfrm_policy_put_afinfo(afinfo
);
1158 /* Resolve list of templates for the flow, given policy. */
1161 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, const struct flowi
*fl
,
1162 struct xfrm_state
**xfrm
, unsigned short family
)
1164 struct net
*net
= xp_net(policy
);
1167 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1168 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1171 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1172 struct xfrm_state
*x
;
1173 xfrm_address_t
*remote
= daddr
;
1174 xfrm_address_t
*local
= saddr
;
1175 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1177 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1178 tmpl
->mode
== XFRM_MODE_BEET
) {
1179 remote
= &tmpl
->id
.daddr
;
1180 local
= &tmpl
->saddr
;
1181 if (xfrm_addr_any(local
, tmpl
->encap_family
)) {
1182 error
= xfrm_get_saddr(net
, &tmp
, remote
, tmpl
->encap_family
);
1189 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1191 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1198 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1202 else if (error
== -ESRCH
)
1205 if (!tmpl
->optional
)
1211 for (nx
--; nx
>=0; nx
--)
1212 xfrm_state_put(xfrm
[nx
]);
1217 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, const struct flowi
*fl
,
1218 struct xfrm_state
**xfrm
, unsigned short family
)
1220 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1221 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1227 for (i
= 0; i
< npols
; i
++) {
1228 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1233 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1241 /* found states are sorted for outbound processing */
1243 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1248 for (cnx
--; cnx
>=0; cnx
--)
1249 xfrm_state_put(tpp
[cnx
]);
1254 /* Check that the bundle accepts the flow and its components are
1258 static inline int xfrm_get_tos(const struct flowi
*fl
, int family
)
1260 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1266 tos
= afinfo
->get_tos(fl
);
1268 xfrm_policy_put_afinfo(afinfo
);
1273 static struct flow_cache_object
*xfrm_bundle_flo_get(struct flow_cache_object
*flo
)
1275 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1276 struct dst_entry
*dst
= &xdst
->u
.dst
;
1278 if (xdst
->route
== NULL
) {
1279 /* Dummy bundle - if it has xfrms we were not
1280 * able to build bundle as template resolution failed.
1281 * It means we need to try again resolving. */
1282 if (xdst
->num_xfrms
> 0)
1286 if (stale_bundle(dst
))
1294 static int xfrm_bundle_flo_check(struct flow_cache_object
*flo
)
1296 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1297 struct dst_entry
*dst
= &xdst
->u
.dst
;
1301 if (stale_bundle(dst
))
1307 static void xfrm_bundle_flo_delete(struct flow_cache_object
*flo
)
1309 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1310 struct dst_entry
*dst
= &xdst
->u
.dst
;
1315 static const struct flow_cache_ops xfrm_bundle_fc_ops
= {
1316 .get
= xfrm_bundle_flo_get
,
1317 .check
= xfrm_bundle_flo_check
,
1318 .delete = xfrm_bundle_flo_delete
,
1321 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1323 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1324 struct dst_ops
*dst_ops
;
1325 struct xfrm_dst
*xdst
;
1328 return ERR_PTR(-EINVAL
);
1332 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1334 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1336 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1342 xdst
= dst_alloc(dst_ops
, 0);
1343 xfrm_policy_put_afinfo(afinfo
);
1346 xdst
->flo
.ops
= &xfrm_bundle_fc_ops
;
1348 xdst
= ERR_PTR(-ENOBUFS
);
1353 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1356 struct xfrm_policy_afinfo
*afinfo
=
1357 xfrm_policy_get_afinfo(dst
->ops
->family
);
1363 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1365 xfrm_policy_put_afinfo(afinfo
);
1370 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
,
1371 const struct flowi
*fl
)
1373 struct xfrm_policy_afinfo
*afinfo
=
1374 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1380 err
= afinfo
->fill_dst(xdst
, dev
, fl
);
1382 xfrm_policy_put_afinfo(afinfo
);
1388 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1389 * all the metrics... Shortly, bundle a bundle.
1392 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1393 struct xfrm_state
**xfrm
, int nx
,
1394 const struct flowi
*fl
,
1395 struct dst_entry
*dst
)
1397 struct net
*net
= xp_net(policy
);
1398 unsigned long now
= jiffies
;
1399 struct net_device
*dev
;
1400 struct dst_entry
*dst_prev
= NULL
;
1401 struct dst_entry
*dst0
= NULL
;
1405 int nfheader_len
= 0;
1406 int trailer_len
= 0;
1408 int family
= policy
->selector
.family
;
1409 xfrm_address_t saddr
, daddr
;
1411 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1413 tos
= xfrm_get_tos(fl
, family
);
1420 for (; i
< nx
; i
++) {
1421 struct xfrm_dst
*xdst
= xfrm_alloc_dst(net
, family
);
1422 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1424 err
= PTR_ERR(xdst
);
1433 dst_prev
->child
= dst_clone(dst1
);
1434 dst1
->flags
|= DST_NOHASH
;
1438 dst_copy_metrics(dst1
, dst
);
1440 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1441 family
= xfrm
[i
]->props
.family
;
1442 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1450 dst1
->xfrm
= xfrm
[i
];
1451 xdst
->xfrm_genid
= xfrm
[i
]->genid
;
1453 dst1
->obsolete
= -1;
1454 dst1
->flags
|= DST_HOST
;
1455 dst1
->lastuse
= now
;
1457 dst1
->input
= dst_discard
;
1458 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1460 dst1
->next
= dst_prev
;
1463 header_len
+= xfrm
[i
]->props
.header_len
;
1464 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1465 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1466 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1469 dst_prev
->child
= dst
;
1477 /* Copy neighbour for reachability confirmation */
1478 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1480 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1481 xfrm_init_pmtu(dst_prev
);
1483 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1484 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1486 err
= xfrm_fill_dst(xdst
, dev
, fl
);
1490 dst_prev
->header_len
= header_len
;
1491 dst_prev
->trailer_len
= trailer_len
;
1492 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1493 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1501 xfrm_state_put(xfrm
[i
]);
1505 dst0
= ERR_PTR(err
);
1510 xfrm_dst_alloc_copy(void **target
, const void *src
, int size
)
1513 *target
= kmalloc(size
, GFP_ATOMIC
);
1517 memcpy(*target
, src
, size
);
1522 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1524 #ifdef CONFIG_XFRM_SUB_POLICY
1525 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1526 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1534 xfrm_dst_update_origin(struct dst_entry
*dst
, const struct flowi
*fl
)
1536 #ifdef CONFIG_XFRM_SUB_POLICY
1537 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1538 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1544 static int xfrm_expand_policies(const struct flowi
*fl
, u16 family
,
1545 struct xfrm_policy
**pols
,
1546 int *num_pols
, int *num_xfrms
)
1550 if (*num_pols
== 0 || !pols
[0]) {
1555 if (IS_ERR(pols
[0]))
1556 return PTR_ERR(pols
[0]);
1558 *num_xfrms
= pols
[0]->xfrm_nr
;
1560 #ifdef CONFIG_XFRM_SUB_POLICY
1561 if (pols
[0] && pols
[0]->action
== XFRM_POLICY_ALLOW
&&
1562 pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1563 pols
[1] = xfrm_policy_lookup_bytype(xp_net(pols
[0]),
1564 XFRM_POLICY_TYPE_MAIN
,
1568 if (IS_ERR(pols
[1])) {
1569 xfrm_pols_put(pols
, *num_pols
);
1570 return PTR_ERR(pols
[1]);
1573 (*num_xfrms
) += pols
[1]->xfrm_nr
;
1577 for (i
= 0; i
< *num_pols
; i
++) {
1578 if (pols
[i
]->action
!= XFRM_POLICY_ALLOW
) {
1588 static struct xfrm_dst
*
1589 xfrm_resolve_and_create_bundle(struct xfrm_policy
**pols
, int num_pols
,
1590 const struct flowi
*fl
, u16 family
,
1591 struct dst_entry
*dst_orig
)
1593 struct net
*net
= xp_net(pols
[0]);
1594 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1595 struct dst_entry
*dst
;
1596 struct xfrm_dst
*xdst
;
1599 /* Try to instantiate a bundle */
1600 err
= xfrm_tmpl_resolve(pols
, num_pols
, fl
, xfrm
, family
);
1602 if (err
!= 0 && err
!= -EAGAIN
)
1603 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1604 return ERR_PTR(err
);
1607 dst
= xfrm_bundle_create(pols
[0], xfrm
, err
, fl
, dst_orig
);
1609 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1610 return ERR_CAST(dst
);
1613 xdst
= (struct xfrm_dst
*)dst
;
1614 xdst
->num_xfrms
= err
;
1616 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1618 err
= xfrm_dst_update_origin(dst
, fl
);
1619 if (unlikely(err
)) {
1621 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1622 return ERR_PTR(err
);
1625 xdst
->num_pols
= num_pols
;
1626 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1627 xdst
->policy_genid
= atomic_read(&pols
[0]->genid
);
1632 static struct flow_cache_object
*
1633 xfrm_bundle_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
,
1634 struct flow_cache_object
*oldflo
, void *ctx
)
1636 struct dst_entry
*dst_orig
= (struct dst_entry
*)ctx
;
1637 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1638 struct xfrm_dst
*xdst
, *new_xdst
;
1639 int num_pols
= 0, num_xfrms
= 0, i
, err
, pol_dead
;
1641 /* Check if the policies from old bundle are usable */
1644 xdst
= container_of(oldflo
, struct xfrm_dst
, flo
);
1645 num_pols
= xdst
->num_pols
;
1646 num_xfrms
= xdst
->num_xfrms
;
1648 for (i
= 0; i
< num_pols
; i
++) {
1649 pols
[i
] = xdst
->pols
[i
];
1650 pol_dead
|= pols
[i
]->walk
.dead
;
1653 dst_free(&xdst
->u
.dst
);
1661 /* Resolve policies to use if we couldn't get them from
1662 * previous cache entry */
1665 pols
[0] = __xfrm_policy_lookup(net
, fl
, family
, dir
);
1666 err
= xfrm_expand_policies(fl
, family
, pols
,
1667 &num_pols
, &num_xfrms
);
1673 goto make_dummy_bundle
;
1676 new_xdst
= xfrm_resolve_and_create_bundle(pols
, num_pols
, fl
, family
, dst_orig
);
1677 if (IS_ERR(new_xdst
)) {
1678 err
= PTR_ERR(new_xdst
);
1682 goto make_dummy_bundle
;
1683 dst_hold(&xdst
->u
.dst
);
1685 } else if (new_xdst
== NULL
) {
1688 goto make_dummy_bundle
;
1689 xdst
->num_xfrms
= 0;
1690 dst_hold(&xdst
->u
.dst
);
1694 /* Kill the previous bundle */
1696 /* The policies were stolen for newly generated bundle */
1698 dst_free(&xdst
->u
.dst
);
1701 /* Flow cache does not have reference, it dst_free()'s,
1702 * but we do need to return one reference for original caller */
1703 dst_hold(&new_xdst
->u
.dst
);
1704 return &new_xdst
->flo
;
1707 /* We found policies, but there's no bundles to instantiate:
1708 * either because the policy blocks, has no transformations or
1709 * we could not build template (no xfrm_states).*/
1710 xdst
= xfrm_alloc_dst(net
, family
);
1712 xfrm_pols_put(pols
, num_pols
);
1713 return ERR_CAST(xdst
);
1715 xdst
->num_pols
= num_pols
;
1716 xdst
->num_xfrms
= num_xfrms
;
1717 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1719 dst_hold(&xdst
->u
.dst
);
1723 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1726 dst_free(&xdst
->u
.dst
);
1728 xfrm_pols_put(pols
, num_pols
);
1729 return ERR_PTR(err
);
1732 /* Main function: finds/creates a bundle for given flow.
1734 * At the moment we eat a raw IP route. Mostly to speed up lookups
1735 * on interfaces with disabled IPsec.
1737 int __xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
,
1738 const struct flowi
*fl
,
1739 struct sock
*sk
, int flags
)
1741 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1742 struct flow_cache_object
*flo
;
1743 struct xfrm_dst
*xdst
;
1744 struct dst_entry
*dst
, *dst_orig
= *dst_p
, *route
;
1745 u16 family
= dst_orig
->ops
->family
;
1746 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1747 int i
, err
, num_pols
, num_xfrms
= 0, drop_pols
= 0;
1754 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1756 pols
[0] = xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1757 err
= xfrm_expand_policies(fl
, family
, pols
,
1758 &num_pols
, &num_xfrms
);
1763 if (num_xfrms
<= 0) {
1764 drop_pols
= num_pols
;
1768 xdst
= xfrm_resolve_and_create_bundle(
1772 xfrm_pols_put(pols
, num_pols
);
1773 err
= PTR_ERR(xdst
);
1775 } else if (xdst
== NULL
) {
1777 drop_pols
= num_pols
;
1781 spin_lock_bh(&xfrm_policy_sk_bundle_lock
);
1782 xdst
->u
.dst
.next
= xfrm_policy_sk_bundles
;
1783 xfrm_policy_sk_bundles
= &xdst
->u
.dst
;
1784 spin_unlock_bh(&xfrm_policy_sk_bundle_lock
);
1786 route
= xdst
->route
;
1791 /* To accelerate a bit... */
1792 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1793 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1796 flo
= flow_cache_lookup(net
, fl
, family
, dir
,
1797 xfrm_bundle_lookup
, dst_orig
);
1804 xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1806 num_pols
= xdst
->num_pols
;
1807 num_xfrms
= xdst
->num_xfrms
;
1808 memcpy(pols
, xdst
->pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1809 route
= xdst
->route
;
1813 if (route
== NULL
&& num_xfrms
> 0) {
1814 /* The only case when xfrm_bundle_lookup() returns a
1815 * bundle with null route, is when the template could
1816 * not be resolved. It means policies are there, but
1817 * bundle could not be created, since we don't yet
1818 * have the xfrm_state's. We need to wait for KM to
1819 * negotiate new SA's or bail out with error.*/
1820 if (net
->xfrm
.sysctl_larval_drop
) {
1821 /* EREMOTE tells the caller to generate
1822 * a one-shot blackhole route. */
1824 xfrm_pols_put(pols
, drop_pols
);
1825 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1828 if (flags
& XFRM_LOOKUP_WAIT
) {
1829 DECLARE_WAITQUEUE(wait
, current
);
1831 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1832 set_current_state(TASK_INTERRUPTIBLE
);
1834 set_current_state(TASK_RUNNING
);
1835 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1837 if (!signal_pending(current
)) {
1846 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
1854 if ((flags
& XFRM_LOOKUP_ICMP
) &&
1855 !(pols
[0]->flags
& XFRM_POLICY_ICMP
)) {
1860 for (i
= 0; i
< num_pols
; i
++)
1861 pols
[i
]->curlft
.use_time
= get_seconds();
1863 if (num_xfrms
< 0) {
1864 /* Prohibit the flow */
1865 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
1868 } else if (num_xfrms
> 0) {
1869 /* Flow transformed */
1871 dst_release(dst_orig
);
1873 /* Flow passes untransformed */
1877 xfrm_pols_put(pols
, drop_pols
);
1881 if (!(flags
& XFRM_LOOKUP_ICMP
))
1887 dst_release(dst_orig
);
1889 xfrm_pols_put(pols
, drop_pols
);
1892 EXPORT_SYMBOL(__xfrm_lookup
);
1894 int xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
,
1895 const struct flowi
*fl
,
1896 struct sock
*sk
, int flags
)
1898 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1900 if (err
== -EREMOTE
) {
1901 dst_release(*dst_p
);
1908 EXPORT_SYMBOL(xfrm_lookup
);
1911 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, const struct flowi
*fl
)
1913 struct xfrm_state
*x
;
1915 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1917 x
= skb
->sp
->xvec
[idx
];
1918 if (!x
->type
->reject
)
1920 return x
->type
->reject(x
, skb
, fl
);
1923 /* When skb is transformed back to its "native" form, we have to
1924 * check policy restrictions. At the moment we make this in maximally
1925 * stupid way. Shame on me. :-) Of course, connected sockets must
1926 * have policy cached at them.
1930 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1931 unsigned short family
)
1933 if (xfrm_state_kern(x
))
1934 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1935 return x
->id
.proto
== tmpl
->id
.proto
&&
1936 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1937 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1938 x
->props
.mode
== tmpl
->mode
&&
1939 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1940 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1941 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1942 xfrm_state_addr_cmp(tmpl
, x
, family
));
1946 * 0 or more than 0 is returned when validation is succeeded (either bypass
1947 * because of optional transport mode, or next index of the mathced secpath
1948 * state with the template.
1949 * -1 is returned when no matching template is found.
1950 * Otherwise "-2 - errored_index" is returned.
1953 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1954 unsigned short family
)
1958 if (tmpl
->optional
) {
1959 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1963 for (; idx
< sp
->len
; idx
++) {
1964 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1966 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1975 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1976 unsigned int family
, int reverse
)
1978 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1981 if (unlikely(afinfo
== NULL
))
1982 return -EAFNOSUPPORT
;
1984 afinfo
->decode_session(skb
, fl
, reverse
);
1985 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1986 xfrm_policy_put_afinfo(afinfo
);
1989 EXPORT_SYMBOL(__xfrm_decode_session
);
1991 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1993 for (; k
< sp
->len
; k
++) {
1994 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
2003 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
2004 unsigned short family
)
2006 struct net
*net
= dev_net(skb
->dev
);
2007 struct xfrm_policy
*pol
;
2008 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2017 reverse
= dir
& ~XFRM_POLICY_MASK
;
2018 dir
&= XFRM_POLICY_MASK
;
2019 fl_dir
= policy_to_flow_dir(dir
);
2021 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
2022 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
2026 nf_nat_decode_session(skb
, &fl
, family
);
2028 /* First, check used SA against their selectors. */
2032 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
2033 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
2034 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
2035 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
2042 if (sk
&& sk
->sk_policy
[dir
]) {
2043 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
2045 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2051 struct flow_cache_object
*flo
;
2053 flo
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
2054 xfrm_policy_lookup
, NULL
);
2055 if (IS_ERR_OR_NULL(flo
))
2056 pol
= ERR_CAST(flo
);
2058 pol
= container_of(flo
, struct xfrm_policy
, flo
);
2062 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2067 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
2068 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2069 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
2075 pol
->curlft
.use_time
= get_seconds();
2079 #ifdef CONFIG_XFRM_SUB_POLICY
2080 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
2081 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
2085 if (IS_ERR(pols
[1])) {
2086 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2089 pols
[1]->curlft
.use_time
= get_seconds();
2095 if (pol
->action
== XFRM_POLICY_ALLOW
) {
2096 struct sec_path
*sp
;
2097 static struct sec_path dummy
;
2098 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
2099 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
2100 struct xfrm_tmpl
**tpp
= tp
;
2104 if ((sp
= skb
->sp
) == NULL
)
2107 for (pi
= 0; pi
< npols
; pi
++) {
2108 if (pols
[pi
] != pol
&&
2109 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
2110 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2113 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2114 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
2117 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2118 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2122 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2126 /* For each tunnel xfrm, find the first matching tmpl.
2127 * For each tmpl before that, find corresponding xfrm.
2128 * Order is _important_. Later we will implement
2129 * some barriers, but at the moment barriers
2130 * are implied between each two transformations.
2132 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2133 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2136 /* "-2 - errored_index" returned */
2138 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2143 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2144 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2148 xfrm_pols_put(pols
, npols
);
2151 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2154 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2156 xfrm_pols_put(pols
, npols
);
2159 EXPORT_SYMBOL(__xfrm_policy_check
);
2161 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2163 struct net
*net
= dev_net(skb
->dev
);
2165 struct dst_entry
*dst
;
2168 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2169 XFRM_INC_STATS(net
, LINUX_MIB_XFRMFWDHDRERROR
);
2176 res
= xfrm_lookup(net
, &dst
, &fl
, NULL
, 0) == 0;
2177 skb_dst_set(skb
, dst
);
2180 EXPORT_SYMBOL(__xfrm_route_forward
);
2182 /* Optimize later using cookies and generation ids. */
2184 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2186 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2187 * to "-1" to force all XFRM destinations to get validated by
2188 * dst_ops->check on every use. We do this because when a
2189 * normal route referenced by an XFRM dst is obsoleted we do
2190 * not go looking around for all parent referencing XFRM dsts
2191 * so that we can invalidate them. It is just too much work.
2192 * Instead we make the checks here on every use. For example:
2194 * XFRM dst A --> IPv4 dst X
2196 * X is the "xdst->route" of A (X is also the "dst->path" of A
2197 * in this example). If X is marked obsolete, "A" will not
2198 * notice. That's what we are validating here via the
2199 * stale_bundle() check.
2201 * When a policy's bundle is pruned, we dst_free() the XFRM
2202 * dst which causes it's ->obsolete field to be set to a
2203 * positive non-zero integer. If an XFRM dst has been pruned
2204 * like this, we want to force a new route lookup.
2206 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2212 static int stale_bundle(struct dst_entry
*dst
)
2214 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
);
2217 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2219 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2220 dst
->dev
= dev_net(dev
)->loopback_dev
;
2225 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2227 static void xfrm_link_failure(struct sk_buff
*skb
)
2229 /* Impossible. Such dst must be popped before reaches point of failure. */
2232 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2235 if (dst
->obsolete
) {
2243 static void __xfrm_garbage_collect(struct net
*net
)
2245 struct dst_entry
*head
, *next
;
2249 spin_lock_bh(&xfrm_policy_sk_bundle_lock
);
2250 head
= xfrm_policy_sk_bundles
;
2251 xfrm_policy_sk_bundles
= NULL
;
2252 spin_unlock_bh(&xfrm_policy_sk_bundle_lock
);
2261 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2264 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2265 u32 pmtu
, route_mtu_cached
;
2267 pmtu
= dst_mtu(dst
->child
);
2268 xdst
->child_mtu_cached
= pmtu
;
2270 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2272 route_mtu_cached
= dst_mtu(xdst
->route
);
2273 xdst
->route_mtu_cached
= route_mtu_cached
;
2275 if (pmtu
> route_mtu_cached
)
2276 pmtu
= route_mtu_cached
;
2278 dst_metric_set(dst
, RTAX_MTU
, pmtu
);
2279 } while ((dst
= dst
->next
));
2282 /* Check that the bundle accepts the flow and its components are
2286 static int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2287 const struct flowi
*fl
, int family
)
2289 struct dst_entry
*dst
= &first
->u
.dst
;
2290 struct xfrm_dst
*last
;
2293 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2294 (dst
->dev
&& !netif_running(dst
->dev
)))
2296 #ifdef CONFIG_XFRM_SUB_POLICY
2298 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2300 if (first
->partner
&&
2301 !xfrm_selector_match(first
->partner
, fl
, family
))
2309 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2311 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2314 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2316 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2318 if (xdst
->xfrm_genid
!= dst
->xfrm
->genid
)
2320 if (xdst
->num_pols
> 0 &&
2321 xdst
->policy_genid
!= atomic_read(&xdst
->pols
[0]->genid
))
2324 mtu
= dst_mtu(dst
->child
);
2325 if (xdst
->child_mtu_cached
!= mtu
) {
2327 xdst
->child_mtu_cached
= mtu
;
2330 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2332 mtu
= dst_mtu(xdst
->route
);
2333 if (xdst
->route_mtu_cached
!= mtu
) {
2335 xdst
->route_mtu_cached
= mtu
;
2339 } while (dst
->xfrm
);
2344 mtu
= last
->child_mtu_cached
;
2348 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2349 if (mtu
> last
->route_mtu_cached
)
2350 mtu
= last
->route_mtu_cached
;
2351 dst_metric_set(dst
, RTAX_MTU
, mtu
);
2356 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2357 last
->child_mtu_cached
= mtu
;
2363 static unsigned int xfrm_default_advmss(const struct dst_entry
*dst
)
2365 return dst_metric_advmss(dst
->path
);
2368 static unsigned int xfrm_default_mtu(const struct dst_entry
*dst
)
2370 return dst_mtu(dst
->path
);
2373 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2377 if (unlikely(afinfo
== NULL
))
2379 if (unlikely(afinfo
->family
>= NPROTO
))
2380 return -EAFNOSUPPORT
;
2381 write_lock_bh(&xfrm_policy_afinfo_lock
);
2382 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2385 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2386 if (likely(dst_ops
->kmem_cachep
== NULL
))
2387 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2388 if (likely(dst_ops
->check
== NULL
))
2389 dst_ops
->check
= xfrm_dst_check
;
2390 if (likely(dst_ops
->default_advmss
== NULL
))
2391 dst_ops
->default_advmss
= xfrm_default_advmss
;
2392 if (likely(dst_ops
->default_mtu
== NULL
))
2393 dst_ops
->default_mtu
= xfrm_default_mtu
;
2394 if (likely(dst_ops
->negative_advice
== NULL
))
2395 dst_ops
->negative_advice
= xfrm_negative_advice
;
2396 if (likely(dst_ops
->link_failure
== NULL
))
2397 dst_ops
->link_failure
= xfrm_link_failure
;
2398 if (likely(afinfo
->garbage_collect
== NULL
))
2399 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2400 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2402 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2406 struct dst_ops
*xfrm_dst_ops
;
2408 switch (afinfo
->family
) {
2410 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2412 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2414 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2420 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2426 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2428 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2431 if (unlikely(afinfo
== NULL
))
2433 if (unlikely(afinfo
->family
>= NPROTO
))
2434 return -EAFNOSUPPORT
;
2435 write_lock_bh(&xfrm_policy_afinfo_lock
);
2436 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2437 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2440 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2441 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2442 dst_ops
->kmem_cachep
= NULL
;
2443 dst_ops
->check
= NULL
;
2444 dst_ops
->negative_advice
= NULL
;
2445 dst_ops
->link_failure
= NULL
;
2446 afinfo
->garbage_collect
= NULL
;
2449 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2452 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2454 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2456 struct xfrm_policy_afinfo
*afinfo
;
2458 read_lock_bh(&xfrm_policy_afinfo_lock
);
2459 afinfo
= xfrm_policy_afinfo
[AF_INET
];
2461 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2462 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2463 afinfo
= xfrm_policy_afinfo
[AF_INET6
];
2465 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2467 read_unlock_bh(&xfrm_policy_afinfo_lock
);
2470 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2472 struct xfrm_policy_afinfo
*afinfo
;
2473 if (unlikely(family
>= NPROTO
))
2475 read_lock(&xfrm_policy_afinfo_lock
);
2476 afinfo
= xfrm_policy_afinfo
[family
];
2477 if (unlikely(!afinfo
))
2478 read_unlock(&xfrm_policy_afinfo_lock
);
2482 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2484 read_unlock(&xfrm_policy_afinfo_lock
);
2487 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2489 struct net_device
*dev
= ptr
;
2493 __xfrm_garbage_collect(dev_net(dev
));
2498 static struct notifier_block xfrm_dev_notifier
= {
2499 .notifier_call
= xfrm_dev_event
,
2502 #ifdef CONFIG_XFRM_STATISTICS
2503 static int __net_init
xfrm_statistics_init(struct net
*net
)
2507 if (snmp_mib_init((void __percpu
**)net
->mib
.xfrm_statistics
,
2508 sizeof(struct linux_xfrm_mib
),
2509 __alignof__(struct linux_xfrm_mib
)) < 0)
2511 rv
= xfrm_proc_init(net
);
2513 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2517 static void xfrm_statistics_fini(struct net
*net
)
2519 xfrm_proc_fini(net
);
2520 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2523 static int __net_init
xfrm_statistics_init(struct net
*net
)
2528 static void xfrm_statistics_fini(struct net
*net
)
2533 static int __net_init
xfrm_policy_init(struct net
*net
)
2535 unsigned int hmask
, sz
;
2538 if (net_eq(net
, &init_net
))
2539 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2540 sizeof(struct xfrm_dst
),
2541 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2545 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2547 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2548 if (!net
->xfrm
.policy_byidx
)
2550 net
->xfrm
.policy_idx_hmask
= hmask
;
2552 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2553 struct xfrm_policy_hash
*htab
;
2555 net
->xfrm
.policy_count
[dir
] = 0;
2556 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2558 htab
= &net
->xfrm
.policy_bydst
[dir
];
2559 htab
->table
= xfrm_hash_alloc(sz
);
2562 htab
->hmask
= hmask
;
2565 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2566 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2567 if (net_eq(net
, &init_net
))
2568 register_netdevice_notifier(&xfrm_dev_notifier
);
2572 for (dir
--; dir
>= 0; dir
--) {
2573 struct xfrm_policy_hash
*htab
;
2575 htab
= &net
->xfrm
.policy_bydst
[dir
];
2576 xfrm_hash_free(htab
->table
, sz
);
2578 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2583 static void xfrm_policy_fini(struct net
*net
)
2585 struct xfrm_audit audit_info
;
2589 flush_work(&net
->xfrm
.policy_hash_work
);
2590 #ifdef CONFIG_XFRM_SUB_POLICY
2591 audit_info
.loginuid
= -1;
2592 audit_info
.sessionid
= -1;
2593 audit_info
.secid
= 0;
2594 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2596 audit_info
.loginuid
= -1;
2597 audit_info
.sessionid
= -1;
2598 audit_info
.secid
= 0;
2599 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2601 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2603 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2604 struct xfrm_policy_hash
*htab
;
2606 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2608 htab
= &net
->xfrm
.policy_bydst
[dir
];
2609 sz
= (htab
->hmask
+ 1);
2610 WARN_ON(!hlist_empty(htab
->table
));
2611 xfrm_hash_free(htab
->table
, sz
);
2614 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2615 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2616 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2619 static int __net_init
xfrm_net_init(struct net
*net
)
2623 rv
= xfrm_statistics_init(net
);
2625 goto out_statistics
;
2626 rv
= xfrm_state_init(net
);
2629 rv
= xfrm_policy_init(net
);
2632 xfrm_dst_ops_init(net
);
2633 rv
= xfrm_sysctl_init(net
);
2639 xfrm_policy_fini(net
);
2641 xfrm_state_fini(net
);
2643 xfrm_statistics_fini(net
);
2648 static void __net_exit
xfrm_net_exit(struct net
*net
)
2650 xfrm_sysctl_fini(net
);
2651 xfrm_policy_fini(net
);
2652 xfrm_state_fini(net
);
2653 xfrm_statistics_fini(net
);
2656 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2657 .init
= xfrm_net_init
,
2658 .exit
= xfrm_net_exit
,
2661 void __init
xfrm_init(void)
2663 register_pernet_subsys(&xfrm_net_ops
);
2667 #ifdef CONFIG_AUDITSYSCALL
2668 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2669 struct audit_buffer
*audit_buf
)
2671 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2672 struct xfrm_selector
*sel
= &xp
->selector
;
2675 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2676 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2678 switch(sel
->family
) {
2680 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2681 if (sel
->prefixlen_s
!= 32)
2682 audit_log_format(audit_buf
, " src_prefixlen=%d",
2684 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2685 if (sel
->prefixlen_d
!= 32)
2686 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2690 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2691 if (sel
->prefixlen_s
!= 128)
2692 audit_log_format(audit_buf
, " src_prefixlen=%d",
2694 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2695 if (sel
->prefixlen_d
!= 128)
2696 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2702 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2703 uid_t auid
, u32 sessionid
, u32 secid
)
2705 struct audit_buffer
*audit_buf
;
2707 audit_buf
= xfrm_audit_start("SPD-add");
2708 if (audit_buf
== NULL
)
2710 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2711 audit_log_format(audit_buf
, " res=%u", result
);
2712 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2713 audit_log_end(audit_buf
);
2715 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2717 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2718 uid_t auid
, u32 sessionid
, u32 secid
)
2720 struct audit_buffer
*audit_buf
;
2722 audit_buf
= xfrm_audit_start("SPD-delete");
2723 if (audit_buf
== NULL
)
2725 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2726 audit_log_format(audit_buf
, " res=%u", result
);
2727 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2728 audit_log_end(audit_buf
);
2730 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2733 #ifdef CONFIG_XFRM_MIGRATE
2734 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2735 struct xfrm_selector
*sel_tgt
)
2737 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2738 if (sel_tgt
->family
== sel_cmp
->family
&&
2739 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2740 sel_cmp
->family
) == 0 &&
2741 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2742 sel_cmp
->family
) == 0 &&
2743 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2744 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2748 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2755 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2758 struct xfrm_policy
*pol
, *ret
= NULL
;
2759 struct hlist_node
*entry
;
2760 struct hlist_head
*chain
;
2763 read_lock_bh(&xfrm_policy_lock
);
2764 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2765 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2766 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2767 pol
->type
== type
) {
2769 priority
= ret
->priority
;
2773 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2774 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2775 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2776 pol
->type
== type
&&
2777 pol
->priority
< priority
) {
2786 read_unlock_bh(&xfrm_policy_lock
);
2791 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2795 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2796 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2798 case XFRM_MODE_TUNNEL
:
2799 case XFRM_MODE_BEET
:
2800 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2801 m
->old_family
) == 0 &&
2802 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2803 m
->old_family
) == 0) {
2807 case XFRM_MODE_TRANSPORT
:
2808 /* in case of transport mode, template does not store
2809 any IP addresses, hence we just compare mode and
2820 /* update endpoint address(es) of template(s) */
2821 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2822 struct xfrm_migrate
*m
, int num_migrate
)
2824 struct xfrm_migrate
*mp
;
2827 write_lock_bh(&pol
->lock
);
2828 if (unlikely(pol
->walk
.dead
)) {
2829 /* target policy has been deleted */
2830 write_unlock_bh(&pol
->lock
);
2834 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2835 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2836 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2839 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2840 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2842 /* update endpoints */
2843 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2844 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2845 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2846 sizeof(pol
->xfrm_vec
[i
].saddr
));
2847 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2849 atomic_inc(&pol
->genid
);
2853 write_unlock_bh(&pol
->lock
);
2861 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2865 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2868 for (i
= 0; i
< num_migrate
; i
++) {
2869 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2870 m
[i
].old_family
) == 0) &&
2871 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2872 m
[i
].old_family
) == 0))
2874 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2875 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2878 /* check if there is any duplicated entry */
2879 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2880 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2881 sizeof(m
[i
].old_daddr
)) &&
2882 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2883 sizeof(m
[i
].old_saddr
)) &&
2884 m
[i
].proto
== m
[j
].proto
&&
2885 m
[i
].mode
== m
[j
].mode
&&
2886 m
[i
].reqid
== m
[j
].reqid
&&
2887 m
[i
].old_family
== m
[j
].old_family
)
2895 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2896 struct xfrm_migrate
*m
, int num_migrate
,
2897 struct xfrm_kmaddress
*k
)
2899 int i
, err
, nx_cur
= 0, nx_new
= 0;
2900 struct xfrm_policy
*pol
= NULL
;
2901 struct xfrm_state
*x
, *xc
;
2902 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2903 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2904 struct xfrm_migrate
*mp
;
2906 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2909 /* Stage 1 - find policy */
2910 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2915 /* Stage 2 - find and update state(s) */
2916 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2917 if ((x
= xfrm_migrate_state_find(mp
))) {
2920 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2930 /* Stage 3 - update policy */
2931 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2934 /* Stage 4 - delete old state(s) */
2936 xfrm_states_put(x_cur
, nx_cur
);
2937 xfrm_states_delete(x_cur
, nx_cur
);
2940 /* Stage 5 - announce */
2941 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2953 xfrm_states_put(x_cur
, nx_cur
);
2955 xfrm_states_delete(x_new
, nx_new
);
2959 EXPORT_SYMBOL(xfrm_migrate
);