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 struct flowi
*fl
, int family
, int strict
);
57 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
61 __xfrm4_selector_match(struct xfrm_selector
*sel
, 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
, 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
, 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
, 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
,
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
, 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
, 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
, struct flowi
*fl
)
995 struct xfrm_policy
*pol
;
997 read_lock_bh(&xfrm_policy_lock
);
998 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
999 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1004 if ((sk
->sk_mark
& pol
->mark
.m
) != pol
->mark
.v
) {
1008 err
= security_xfrm_policy_lookup(pol
->security
,
1010 policy_to_flow_dir(dir
));
1013 else if (err
== -ESRCH
)
1021 read_unlock_bh(&xfrm_policy_lock
);
1025 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1027 struct net
*net
= xp_net(pol
);
1028 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1031 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1032 hlist_add_head(&pol
->bydst
, chain
);
1033 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1034 net
->xfrm
.policy_count
[dir
]++;
1037 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1038 schedule_work(&net
->xfrm
.policy_hash_work
);
1041 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1044 struct net
*net
= xp_net(pol
);
1046 if (hlist_unhashed(&pol
->bydst
))
1049 hlist_del(&pol
->bydst
);
1050 hlist_del(&pol
->byidx
);
1051 list_del(&pol
->walk
.all
);
1052 net
->xfrm
.policy_count
[dir
]--;
1057 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1059 write_lock_bh(&xfrm_policy_lock
);
1060 pol
= __xfrm_policy_unlink(pol
, dir
);
1061 write_unlock_bh(&xfrm_policy_lock
);
1063 xfrm_policy_kill(pol
);
1068 EXPORT_SYMBOL(xfrm_policy_delete
);
1070 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1072 struct net
*net
= xp_net(pol
);
1073 struct xfrm_policy
*old_pol
;
1075 #ifdef CONFIG_XFRM_SUB_POLICY
1076 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1080 write_lock_bh(&xfrm_policy_lock
);
1081 old_pol
= sk
->sk_policy
[dir
];
1082 sk
->sk_policy
[dir
] = pol
;
1084 pol
->curlft
.add_time
= get_seconds();
1085 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1086 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1089 /* Unlinking succeeds always. This is the only function
1090 * allowed to delete or replace socket policy.
1092 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1093 write_unlock_bh(&xfrm_policy_lock
);
1096 xfrm_policy_kill(old_pol
);
1101 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1103 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1106 newp
->selector
= old
->selector
;
1107 if (security_xfrm_policy_clone(old
->security
,
1110 return NULL
; /* ENOMEM */
1112 newp
->lft
= old
->lft
;
1113 newp
->curlft
= old
->curlft
;
1114 newp
->mark
= old
->mark
;
1115 newp
->action
= old
->action
;
1116 newp
->flags
= old
->flags
;
1117 newp
->xfrm_nr
= old
->xfrm_nr
;
1118 newp
->index
= old
->index
;
1119 newp
->type
= old
->type
;
1120 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1121 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1122 write_lock_bh(&xfrm_policy_lock
);
1123 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1124 write_unlock_bh(&xfrm_policy_lock
);
1130 int __xfrm_sk_clone_policy(struct sock
*sk
)
1132 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1133 *p1
= sk
->sk_policy
[1];
1135 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1136 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1138 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1144 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1145 unsigned short family
)
1148 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1150 if (unlikely(afinfo
== NULL
))
1152 err
= afinfo
->get_saddr(net
, local
, remote
);
1153 xfrm_policy_put_afinfo(afinfo
);
1157 /* Resolve list of templates for the flow, given policy. */
1160 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1161 struct xfrm_state
**xfrm
,
1162 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
, struct flowi
*fl
,
1218 struct xfrm_state
**xfrm
,
1219 unsigned short family
)
1221 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1222 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1228 for (i
= 0; i
< npols
; i
++) {
1229 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1234 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1242 /* found states are sorted for outbound processing */
1244 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1249 for (cnx
--; cnx
>=0; cnx
--)
1250 xfrm_state_put(tpp
[cnx
]);
1255 /* Check that the bundle accepts the flow and its components are
1259 static inline int xfrm_get_tos(const struct flowi
*fl
, int family
)
1261 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1267 tos
= afinfo
->get_tos(fl
);
1269 xfrm_policy_put_afinfo(afinfo
);
1274 static struct flow_cache_object
*xfrm_bundle_flo_get(struct flow_cache_object
*flo
)
1276 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1277 struct dst_entry
*dst
= &xdst
->u
.dst
;
1279 if (xdst
->route
== NULL
) {
1280 /* Dummy bundle - if it has xfrms we were not
1281 * able to build bundle as template resolution failed.
1282 * It means we need to try again resolving. */
1283 if (xdst
->num_xfrms
> 0)
1287 if (stale_bundle(dst
))
1295 static int xfrm_bundle_flo_check(struct flow_cache_object
*flo
)
1297 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1298 struct dst_entry
*dst
= &xdst
->u
.dst
;
1302 if (stale_bundle(dst
))
1308 static void xfrm_bundle_flo_delete(struct flow_cache_object
*flo
)
1310 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1311 struct dst_entry
*dst
= &xdst
->u
.dst
;
1316 static const struct flow_cache_ops xfrm_bundle_fc_ops
= {
1317 .get
= xfrm_bundle_flo_get
,
1318 .check
= xfrm_bundle_flo_check
,
1319 .delete = xfrm_bundle_flo_delete
,
1322 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1324 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1325 struct dst_ops
*dst_ops
;
1326 struct xfrm_dst
*xdst
;
1329 return ERR_PTR(-EINVAL
);
1333 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1335 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1337 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1343 xdst
= dst_alloc(dst_ops
, 0);
1344 xfrm_policy_put_afinfo(afinfo
);
1347 xdst
->flo
.ops
= &xfrm_bundle_fc_ops
;
1349 xdst
= ERR_PTR(-ENOBUFS
);
1354 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1357 struct xfrm_policy_afinfo
*afinfo
=
1358 xfrm_policy_get_afinfo(dst
->ops
->family
);
1364 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1366 xfrm_policy_put_afinfo(afinfo
);
1371 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
,
1374 struct xfrm_policy_afinfo
*afinfo
=
1375 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1381 err
= afinfo
->fill_dst(xdst
, dev
, fl
);
1383 xfrm_policy_put_afinfo(afinfo
);
1389 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1390 * all the metrics... Shortly, bundle a bundle.
1393 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1394 struct xfrm_state
**xfrm
, int nx
,
1396 struct dst_entry
*dst
)
1398 struct net
*net
= xp_net(policy
);
1399 unsigned long now
= jiffies
;
1400 struct net_device
*dev
;
1401 struct dst_entry
*dst_prev
= NULL
;
1402 struct dst_entry
*dst0
= NULL
;
1406 int nfheader_len
= 0;
1407 int trailer_len
= 0;
1409 int family
= policy
->selector
.family
;
1410 xfrm_address_t saddr
, daddr
;
1412 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1414 tos
= xfrm_get_tos(fl
, family
);
1421 for (; i
< nx
; i
++) {
1422 struct xfrm_dst
*xdst
= xfrm_alloc_dst(net
, family
);
1423 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1425 err
= PTR_ERR(xdst
);
1434 dst_prev
->child
= dst_clone(dst1
);
1435 dst1
->flags
|= DST_NOHASH
;
1439 dst_copy_metrics(dst1
, dst
);
1441 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1442 family
= xfrm
[i
]->props
.family
;
1443 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1451 dst1
->xfrm
= xfrm
[i
];
1452 xdst
->xfrm_genid
= xfrm
[i
]->genid
;
1454 dst1
->obsolete
= -1;
1455 dst1
->flags
|= DST_HOST
;
1456 dst1
->lastuse
= now
;
1458 dst1
->input
= dst_discard
;
1459 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1461 dst1
->next
= dst_prev
;
1464 header_len
+= xfrm
[i
]->props
.header_len
;
1465 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1466 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1467 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1470 dst_prev
->child
= dst
;
1478 /* Copy neighbour for reachability confirmation */
1479 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1481 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1482 xfrm_init_pmtu(dst_prev
);
1484 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1485 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1487 err
= xfrm_fill_dst(xdst
, dev
, fl
);
1491 dst_prev
->header_len
= header_len
;
1492 dst_prev
->trailer_len
= trailer_len
;
1493 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1494 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1502 xfrm_state_put(xfrm
[i
]);
1506 dst0
= ERR_PTR(err
);
1511 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1514 *target
= kmalloc(size
, GFP_ATOMIC
);
1518 memcpy(*target
, src
, size
);
1523 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1525 #ifdef CONFIG_XFRM_SUB_POLICY
1526 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1527 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1535 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1537 #ifdef CONFIG_XFRM_SUB_POLICY
1538 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1539 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1545 static int xfrm_expand_policies(struct flowi
*fl
, u16 family
,
1546 struct xfrm_policy
**pols
,
1547 int *num_pols
, int *num_xfrms
)
1551 if (*num_pols
== 0 || !pols
[0]) {
1556 if (IS_ERR(pols
[0]))
1557 return PTR_ERR(pols
[0]);
1559 *num_xfrms
= pols
[0]->xfrm_nr
;
1561 #ifdef CONFIG_XFRM_SUB_POLICY
1562 if (pols
[0] && pols
[0]->action
== XFRM_POLICY_ALLOW
&&
1563 pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1564 pols
[1] = xfrm_policy_lookup_bytype(xp_net(pols
[0]),
1565 XFRM_POLICY_TYPE_MAIN
,
1569 if (IS_ERR(pols
[1])) {
1570 xfrm_pols_put(pols
, *num_pols
);
1571 return PTR_ERR(pols
[1]);
1574 (*num_xfrms
) += pols
[1]->xfrm_nr
;
1578 for (i
= 0; i
< *num_pols
; i
++) {
1579 if (pols
[i
]->action
!= XFRM_POLICY_ALLOW
) {
1589 static struct xfrm_dst
*
1590 xfrm_resolve_and_create_bundle(struct xfrm_policy
**pols
, int num_pols
,
1591 struct flowi
*fl
, u16 family
,
1592 struct dst_entry
*dst_orig
)
1594 struct net
*net
= xp_net(pols
[0]);
1595 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1596 struct dst_entry
*dst
;
1597 struct xfrm_dst
*xdst
;
1600 /* Try to instantiate a bundle */
1601 err
= xfrm_tmpl_resolve(pols
, num_pols
, fl
, xfrm
, family
);
1603 if (err
!= 0 && err
!= -EAGAIN
)
1604 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1605 return ERR_PTR(err
);
1608 dst
= xfrm_bundle_create(pols
[0], xfrm
, err
, fl
, dst_orig
);
1610 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1611 return ERR_CAST(dst
);
1614 xdst
= (struct xfrm_dst
*)dst
;
1615 xdst
->num_xfrms
= err
;
1617 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1619 err
= xfrm_dst_update_origin(dst
, fl
);
1620 if (unlikely(err
)) {
1622 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1623 return ERR_PTR(err
);
1626 xdst
->num_pols
= num_pols
;
1627 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1628 xdst
->policy_genid
= atomic_read(&pols
[0]->genid
);
1633 static struct flow_cache_object
*
1634 xfrm_bundle_lookup(struct net
*net
, struct flowi
*fl
, u16 family
, u8 dir
,
1635 struct flow_cache_object
*oldflo
, void *ctx
)
1637 struct dst_entry
*dst_orig
= (struct dst_entry
*)ctx
;
1638 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1639 struct xfrm_dst
*xdst
, *new_xdst
;
1640 int num_pols
= 0, num_xfrms
= 0, i
, err
, pol_dead
;
1642 /* Check if the policies from old bundle are usable */
1645 xdst
= container_of(oldflo
, struct xfrm_dst
, flo
);
1646 num_pols
= xdst
->num_pols
;
1647 num_xfrms
= xdst
->num_xfrms
;
1649 for (i
= 0; i
< num_pols
; i
++) {
1650 pols
[i
] = xdst
->pols
[i
];
1651 pol_dead
|= pols
[i
]->walk
.dead
;
1654 dst_free(&xdst
->u
.dst
);
1662 /* Resolve policies to use if we couldn't get them from
1663 * previous cache entry */
1666 pols
[0] = __xfrm_policy_lookup(net
, fl
, family
, dir
);
1667 err
= xfrm_expand_policies(fl
, family
, pols
,
1668 &num_pols
, &num_xfrms
);
1674 goto make_dummy_bundle
;
1677 new_xdst
= xfrm_resolve_and_create_bundle(pols
, num_pols
, fl
, family
, dst_orig
);
1678 if (IS_ERR(new_xdst
)) {
1679 err
= PTR_ERR(new_xdst
);
1683 goto make_dummy_bundle
;
1684 dst_hold(&xdst
->u
.dst
);
1686 } else if (new_xdst
== NULL
) {
1689 goto make_dummy_bundle
;
1690 xdst
->num_xfrms
= 0;
1691 dst_hold(&xdst
->u
.dst
);
1695 /* Kill the previous bundle */
1697 /* The policies were stolen for newly generated bundle */
1699 dst_free(&xdst
->u
.dst
);
1702 /* Flow cache does not have reference, it dst_free()'s,
1703 * but we do need to return one reference for original caller */
1704 dst_hold(&new_xdst
->u
.dst
);
1705 return &new_xdst
->flo
;
1708 /* We found policies, but there's no bundles to instantiate:
1709 * either because the policy blocks, has no transformations or
1710 * we could not build template (no xfrm_states).*/
1711 xdst
= xfrm_alloc_dst(net
, family
);
1713 xfrm_pols_put(pols
, num_pols
);
1714 return ERR_CAST(xdst
);
1716 xdst
->num_pols
= num_pols
;
1717 xdst
->num_xfrms
= num_xfrms
;
1718 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1720 dst_hold(&xdst
->u
.dst
);
1724 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1727 dst_free(&xdst
->u
.dst
);
1729 xfrm_pols_put(pols
, num_pols
);
1730 return ERR_PTR(err
);
1733 /* Main function: finds/creates a bundle for given flow.
1735 * At the moment we eat a raw IP route. Mostly to speed up lookups
1736 * on interfaces with disabled IPsec.
1738 int __xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, 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
, struct flowi
*fl
,
1895 struct sock
*sk
, int flags
)
1897 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1899 if (err
== -EREMOTE
) {
1900 dst_release(*dst_p
);
1907 EXPORT_SYMBOL(xfrm_lookup
);
1910 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1912 struct xfrm_state
*x
;
1914 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1916 x
= skb
->sp
->xvec
[idx
];
1917 if (!x
->type
->reject
)
1919 return x
->type
->reject(x
, skb
, fl
);
1922 /* When skb is transformed back to its "native" form, we have to
1923 * check policy restrictions. At the moment we make this in maximally
1924 * stupid way. Shame on me. :-) Of course, connected sockets must
1925 * have policy cached at them.
1929 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1930 unsigned short family
)
1932 if (xfrm_state_kern(x
))
1933 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1934 return x
->id
.proto
== tmpl
->id
.proto
&&
1935 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1936 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1937 x
->props
.mode
== tmpl
->mode
&&
1938 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1939 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1940 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1941 xfrm_state_addr_cmp(tmpl
, x
, family
));
1945 * 0 or more than 0 is returned when validation is succeeded (either bypass
1946 * because of optional transport mode, or next index of the mathced secpath
1947 * state with the template.
1948 * -1 is returned when no matching template is found.
1949 * Otherwise "-2 - errored_index" is returned.
1952 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1953 unsigned short family
)
1957 if (tmpl
->optional
) {
1958 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1962 for (; idx
< sp
->len
; idx
++) {
1963 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1965 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1974 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1975 unsigned int family
, int reverse
)
1977 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1980 if (unlikely(afinfo
== NULL
))
1981 return -EAFNOSUPPORT
;
1983 afinfo
->decode_session(skb
, fl
, reverse
);
1984 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1985 xfrm_policy_put_afinfo(afinfo
);
1988 EXPORT_SYMBOL(__xfrm_decode_session
);
1990 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1992 for (; k
< sp
->len
; k
++) {
1993 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
2002 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
2003 unsigned short family
)
2005 struct net
*net
= dev_net(skb
->dev
);
2006 struct xfrm_policy
*pol
;
2007 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2016 reverse
= dir
& ~XFRM_POLICY_MASK
;
2017 dir
&= XFRM_POLICY_MASK
;
2018 fl_dir
= policy_to_flow_dir(dir
);
2020 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
2021 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
2025 nf_nat_decode_session(skb
, &fl
, family
);
2027 /* First, check used SA against their selectors. */
2031 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
2032 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
2033 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
2034 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
2041 if (sk
&& sk
->sk_policy
[dir
]) {
2042 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
2044 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2050 struct flow_cache_object
*flo
;
2052 flo
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
2053 xfrm_policy_lookup
, NULL
);
2054 if (IS_ERR_OR_NULL(flo
))
2055 pol
= ERR_CAST(flo
);
2057 pol
= container_of(flo
, struct xfrm_policy
, flo
);
2061 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2066 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
2067 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2068 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
2074 pol
->curlft
.use_time
= get_seconds();
2078 #ifdef CONFIG_XFRM_SUB_POLICY
2079 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
2080 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
2084 if (IS_ERR(pols
[1])) {
2085 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2088 pols
[1]->curlft
.use_time
= get_seconds();
2094 if (pol
->action
== XFRM_POLICY_ALLOW
) {
2095 struct sec_path
*sp
;
2096 static struct sec_path dummy
;
2097 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
2098 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
2099 struct xfrm_tmpl
**tpp
= tp
;
2103 if ((sp
= skb
->sp
) == NULL
)
2106 for (pi
= 0; pi
< npols
; pi
++) {
2107 if (pols
[pi
] != pol
&&
2108 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
2109 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2112 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2113 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
2116 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2117 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2121 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2125 /* For each tunnel xfrm, find the first matching tmpl.
2126 * For each tmpl before that, find corresponding xfrm.
2127 * Order is _important_. Later we will implement
2128 * some barriers, but at the moment barriers
2129 * are implied between each two transformations.
2131 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2132 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2135 /* "-2 - errored_index" returned */
2137 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2142 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2143 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2147 xfrm_pols_put(pols
, npols
);
2150 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2153 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2155 xfrm_pols_put(pols
, npols
);
2158 EXPORT_SYMBOL(__xfrm_policy_check
);
2160 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2162 struct net
*net
= dev_net(skb
->dev
);
2164 struct dst_entry
*dst
;
2167 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2168 XFRM_INC_STATS(net
, LINUX_MIB_XFRMFWDHDRERROR
);
2175 res
= xfrm_lookup(net
, &dst
, &fl
, NULL
, 0) == 0;
2176 skb_dst_set(skb
, dst
);
2179 EXPORT_SYMBOL(__xfrm_route_forward
);
2181 /* Optimize later using cookies and generation ids. */
2183 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2185 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2186 * to "-1" to force all XFRM destinations to get validated by
2187 * dst_ops->check on every use. We do this because when a
2188 * normal route referenced by an XFRM dst is obsoleted we do
2189 * not go looking around for all parent referencing XFRM dsts
2190 * so that we can invalidate them. It is just too much work.
2191 * Instead we make the checks here on every use. For example:
2193 * XFRM dst A --> IPv4 dst X
2195 * X is the "xdst->route" of A (X is also the "dst->path" of A
2196 * in this example). If X is marked obsolete, "A" will not
2197 * notice. That's what we are validating here via the
2198 * stale_bundle() check.
2200 * When a policy's bundle is pruned, we dst_free() the XFRM
2201 * dst which causes it's ->obsolete field to be set to a
2202 * positive non-zero integer. If an XFRM dst has been pruned
2203 * like this, we want to force a new route lookup.
2205 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2211 static int stale_bundle(struct dst_entry
*dst
)
2213 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2216 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2218 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2219 dst
->dev
= dev_net(dev
)->loopback_dev
;
2224 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2226 static void xfrm_link_failure(struct sk_buff
*skb
)
2228 /* Impossible. Such dst must be popped before reaches point of failure. */
2231 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2234 if (dst
->obsolete
) {
2242 static void __xfrm_garbage_collect(struct net
*net
)
2244 struct dst_entry
*head
, *next
;
2248 spin_lock_bh(&xfrm_policy_sk_bundle_lock
);
2249 head
= xfrm_policy_sk_bundles
;
2250 xfrm_policy_sk_bundles
= NULL
;
2251 spin_unlock_bh(&xfrm_policy_sk_bundle_lock
);
2260 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2263 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2264 u32 pmtu
, route_mtu_cached
;
2266 pmtu
= dst_mtu(dst
->child
);
2267 xdst
->child_mtu_cached
= pmtu
;
2269 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2271 route_mtu_cached
= dst_mtu(xdst
->route
);
2272 xdst
->route_mtu_cached
= route_mtu_cached
;
2274 if (pmtu
> route_mtu_cached
)
2275 pmtu
= route_mtu_cached
;
2277 dst_metric_set(dst
, RTAX_MTU
, pmtu
);
2278 } while ((dst
= dst
->next
));
2281 /* Check that the bundle accepts the flow and its components are
2285 static int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2286 struct flowi
*fl
, int family
, int strict
)
2288 struct dst_entry
*dst
= &first
->u
.dst
;
2289 struct xfrm_dst
*last
;
2292 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2293 (dst
->dev
&& !netif_running(dst
->dev
)))
2295 #ifdef CONFIG_XFRM_SUB_POLICY
2297 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2299 if (first
->partner
&&
2300 !xfrm_selector_match(first
->partner
, fl
, family
))
2308 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2310 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2313 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2315 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2317 if (xdst
->xfrm_genid
!= dst
->xfrm
->genid
)
2319 if (xdst
->num_pols
> 0 &&
2320 xdst
->policy_genid
!= atomic_read(&xdst
->pols
[0]->genid
))
2324 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2325 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2328 mtu
= dst_mtu(dst
->child
);
2329 if (xdst
->child_mtu_cached
!= mtu
) {
2331 xdst
->child_mtu_cached
= mtu
;
2334 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2336 mtu
= dst_mtu(xdst
->route
);
2337 if (xdst
->route_mtu_cached
!= mtu
) {
2339 xdst
->route_mtu_cached
= mtu
;
2343 } while (dst
->xfrm
);
2348 mtu
= last
->child_mtu_cached
;
2352 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2353 if (mtu
> last
->route_mtu_cached
)
2354 mtu
= last
->route_mtu_cached
;
2355 dst_metric_set(dst
, RTAX_MTU
, mtu
);
2360 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2361 last
->child_mtu_cached
= mtu
;
2367 static unsigned int xfrm_default_advmss(const struct dst_entry
*dst
)
2369 return dst_metric_advmss(dst
->path
);
2372 static unsigned int xfrm_default_mtu(const struct dst_entry
*dst
)
2374 return dst_mtu(dst
->path
);
2377 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2381 if (unlikely(afinfo
== NULL
))
2383 if (unlikely(afinfo
->family
>= NPROTO
))
2384 return -EAFNOSUPPORT
;
2385 write_lock_bh(&xfrm_policy_afinfo_lock
);
2386 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2389 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2390 if (likely(dst_ops
->kmem_cachep
== NULL
))
2391 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2392 if (likely(dst_ops
->check
== NULL
))
2393 dst_ops
->check
= xfrm_dst_check
;
2394 if (likely(dst_ops
->default_advmss
== NULL
))
2395 dst_ops
->default_advmss
= xfrm_default_advmss
;
2396 if (likely(dst_ops
->default_mtu
== NULL
))
2397 dst_ops
->default_mtu
= xfrm_default_mtu
;
2398 if (likely(dst_ops
->negative_advice
== NULL
))
2399 dst_ops
->negative_advice
= xfrm_negative_advice
;
2400 if (likely(dst_ops
->link_failure
== NULL
))
2401 dst_ops
->link_failure
= xfrm_link_failure
;
2402 if (likely(afinfo
->garbage_collect
== NULL
))
2403 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2404 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2406 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2410 struct dst_ops
*xfrm_dst_ops
;
2412 switch (afinfo
->family
) {
2414 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2416 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2418 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2424 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2430 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2432 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2435 if (unlikely(afinfo
== NULL
))
2437 if (unlikely(afinfo
->family
>= NPROTO
))
2438 return -EAFNOSUPPORT
;
2439 write_lock_bh(&xfrm_policy_afinfo_lock
);
2440 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2441 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2444 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2445 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2446 dst_ops
->kmem_cachep
= NULL
;
2447 dst_ops
->check
= NULL
;
2448 dst_ops
->negative_advice
= NULL
;
2449 dst_ops
->link_failure
= NULL
;
2450 afinfo
->garbage_collect
= NULL
;
2453 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2456 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2458 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2460 struct xfrm_policy_afinfo
*afinfo
;
2462 read_lock_bh(&xfrm_policy_afinfo_lock
);
2463 afinfo
= xfrm_policy_afinfo
[AF_INET
];
2465 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2466 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2467 afinfo
= xfrm_policy_afinfo
[AF_INET6
];
2469 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2471 read_unlock_bh(&xfrm_policy_afinfo_lock
);
2474 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2476 struct xfrm_policy_afinfo
*afinfo
;
2477 if (unlikely(family
>= NPROTO
))
2479 read_lock(&xfrm_policy_afinfo_lock
);
2480 afinfo
= xfrm_policy_afinfo
[family
];
2481 if (unlikely(!afinfo
))
2482 read_unlock(&xfrm_policy_afinfo_lock
);
2486 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2488 read_unlock(&xfrm_policy_afinfo_lock
);
2491 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2493 struct net_device
*dev
= ptr
;
2497 __xfrm_garbage_collect(dev_net(dev
));
2502 static struct notifier_block xfrm_dev_notifier
= {
2503 .notifier_call
= xfrm_dev_event
,
2506 #ifdef CONFIG_XFRM_STATISTICS
2507 static int __net_init
xfrm_statistics_init(struct net
*net
)
2511 if (snmp_mib_init((void __percpu
**)net
->mib
.xfrm_statistics
,
2512 sizeof(struct linux_xfrm_mib
),
2513 __alignof__(struct linux_xfrm_mib
)) < 0)
2515 rv
= xfrm_proc_init(net
);
2517 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2521 static void xfrm_statistics_fini(struct net
*net
)
2523 xfrm_proc_fini(net
);
2524 snmp_mib_free((void __percpu
**)net
->mib
.xfrm_statistics
);
2527 static int __net_init
xfrm_statistics_init(struct net
*net
)
2532 static void xfrm_statistics_fini(struct net
*net
)
2537 static int __net_init
xfrm_policy_init(struct net
*net
)
2539 unsigned int hmask
, sz
;
2542 if (net_eq(net
, &init_net
))
2543 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2544 sizeof(struct xfrm_dst
),
2545 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2549 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2551 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2552 if (!net
->xfrm
.policy_byidx
)
2554 net
->xfrm
.policy_idx_hmask
= hmask
;
2556 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2557 struct xfrm_policy_hash
*htab
;
2559 net
->xfrm
.policy_count
[dir
] = 0;
2560 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2562 htab
= &net
->xfrm
.policy_bydst
[dir
];
2563 htab
->table
= xfrm_hash_alloc(sz
);
2566 htab
->hmask
= hmask
;
2569 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2570 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2571 if (net_eq(net
, &init_net
))
2572 register_netdevice_notifier(&xfrm_dev_notifier
);
2576 for (dir
--; dir
>= 0; dir
--) {
2577 struct xfrm_policy_hash
*htab
;
2579 htab
= &net
->xfrm
.policy_bydst
[dir
];
2580 xfrm_hash_free(htab
->table
, sz
);
2582 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2587 static void xfrm_policy_fini(struct net
*net
)
2589 struct xfrm_audit audit_info
;
2593 flush_work(&net
->xfrm
.policy_hash_work
);
2594 #ifdef CONFIG_XFRM_SUB_POLICY
2595 audit_info
.loginuid
= -1;
2596 audit_info
.sessionid
= -1;
2597 audit_info
.secid
= 0;
2598 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2600 audit_info
.loginuid
= -1;
2601 audit_info
.sessionid
= -1;
2602 audit_info
.secid
= 0;
2603 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2605 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2607 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2608 struct xfrm_policy_hash
*htab
;
2610 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2612 htab
= &net
->xfrm
.policy_bydst
[dir
];
2613 sz
= (htab
->hmask
+ 1);
2614 WARN_ON(!hlist_empty(htab
->table
));
2615 xfrm_hash_free(htab
->table
, sz
);
2618 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2619 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2620 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2623 static int __net_init
xfrm_net_init(struct net
*net
)
2627 rv
= xfrm_statistics_init(net
);
2629 goto out_statistics
;
2630 rv
= xfrm_state_init(net
);
2633 rv
= xfrm_policy_init(net
);
2636 xfrm_dst_ops_init(net
);
2637 rv
= xfrm_sysctl_init(net
);
2643 xfrm_policy_fini(net
);
2645 xfrm_state_fini(net
);
2647 xfrm_statistics_fini(net
);
2652 static void __net_exit
xfrm_net_exit(struct net
*net
)
2654 xfrm_sysctl_fini(net
);
2655 xfrm_policy_fini(net
);
2656 xfrm_state_fini(net
);
2657 xfrm_statistics_fini(net
);
2660 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2661 .init
= xfrm_net_init
,
2662 .exit
= xfrm_net_exit
,
2665 void __init
xfrm_init(void)
2667 register_pernet_subsys(&xfrm_net_ops
);
2671 #ifdef CONFIG_AUDITSYSCALL
2672 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2673 struct audit_buffer
*audit_buf
)
2675 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2676 struct xfrm_selector
*sel
= &xp
->selector
;
2679 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2680 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2682 switch(sel
->family
) {
2684 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2685 if (sel
->prefixlen_s
!= 32)
2686 audit_log_format(audit_buf
, " src_prefixlen=%d",
2688 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2689 if (sel
->prefixlen_d
!= 32)
2690 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2694 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2695 if (sel
->prefixlen_s
!= 128)
2696 audit_log_format(audit_buf
, " src_prefixlen=%d",
2698 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2699 if (sel
->prefixlen_d
!= 128)
2700 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2706 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2707 uid_t auid
, u32 sessionid
, u32 secid
)
2709 struct audit_buffer
*audit_buf
;
2711 audit_buf
= xfrm_audit_start("SPD-add");
2712 if (audit_buf
== NULL
)
2714 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2715 audit_log_format(audit_buf
, " res=%u", result
);
2716 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2717 audit_log_end(audit_buf
);
2719 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2721 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2722 uid_t auid
, u32 sessionid
, u32 secid
)
2724 struct audit_buffer
*audit_buf
;
2726 audit_buf
= xfrm_audit_start("SPD-delete");
2727 if (audit_buf
== NULL
)
2729 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2730 audit_log_format(audit_buf
, " res=%u", result
);
2731 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2732 audit_log_end(audit_buf
);
2734 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2737 #ifdef CONFIG_XFRM_MIGRATE
2738 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2739 struct xfrm_selector
*sel_tgt
)
2741 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2742 if (sel_tgt
->family
== sel_cmp
->family
&&
2743 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2744 sel_cmp
->family
) == 0 &&
2745 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2746 sel_cmp
->family
) == 0 &&
2747 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2748 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2752 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2759 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2762 struct xfrm_policy
*pol
, *ret
= NULL
;
2763 struct hlist_node
*entry
;
2764 struct hlist_head
*chain
;
2767 read_lock_bh(&xfrm_policy_lock
);
2768 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2769 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2770 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2771 pol
->type
== type
) {
2773 priority
= ret
->priority
;
2777 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2778 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2779 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2780 pol
->type
== type
&&
2781 pol
->priority
< priority
) {
2790 read_unlock_bh(&xfrm_policy_lock
);
2795 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2799 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2800 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2802 case XFRM_MODE_TUNNEL
:
2803 case XFRM_MODE_BEET
:
2804 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2805 m
->old_family
) == 0 &&
2806 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2807 m
->old_family
) == 0) {
2811 case XFRM_MODE_TRANSPORT
:
2812 /* in case of transport mode, template does not store
2813 any IP addresses, hence we just compare mode and
2824 /* update endpoint address(es) of template(s) */
2825 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2826 struct xfrm_migrate
*m
, int num_migrate
)
2828 struct xfrm_migrate
*mp
;
2831 write_lock_bh(&pol
->lock
);
2832 if (unlikely(pol
->walk
.dead
)) {
2833 /* target policy has been deleted */
2834 write_unlock_bh(&pol
->lock
);
2838 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2839 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2840 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2843 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2844 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2846 /* update endpoints */
2847 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2848 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2849 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2850 sizeof(pol
->xfrm_vec
[i
].saddr
));
2851 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2853 atomic_inc(&pol
->genid
);
2857 write_unlock_bh(&pol
->lock
);
2865 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2869 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2872 for (i
= 0; i
< num_migrate
; i
++) {
2873 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2874 m
[i
].old_family
) == 0) &&
2875 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2876 m
[i
].old_family
) == 0))
2878 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2879 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2882 /* check if there is any duplicated entry */
2883 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2884 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2885 sizeof(m
[i
].old_daddr
)) &&
2886 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2887 sizeof(m
[i
].old_saddr
)) &&
2888 m
[i
].proto
== m
[j
].proto
&&
2889 m
[i
].mode
== m
[j
].mode
&&
2890 m
[i
].reqid
== m
[j
].reqid
&&
2891 m
[i
].old_family
== m
[j
].old_family
)
2899 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2900 struct xfrm_migrate
*m
, int num_migrate
,
2901 struct xfrm_kmaddress
*k
)
2903 int i
, err
, nx_cur
= 0, nx_new
= 0;
2904 struct xfrm_policy
*pol
= NULL
;
2905 struct xfrm_state
*x
, *xc
;
2906 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2907 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2908 struct xfrm_migrate
*mp
;
2910 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2913 /* Stage 1 - find policy */
2914 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2919 /* Stage 2 - find and update state(s) */
2920 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2921 if ((x
= xfrm_migrate_state_find(mp
))) {
2924 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2934 /* Stage 3 - update policy */
2935 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2938 /* Stage 4 - delete old state(s) */
2940 xfrm_states_put(x_cur
, nx_cur
);
2941 xfrm_states_delete(x_cur
, nx_cur
);
2944 /* Stage 5 - announce */
2945 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2957 xfrm_states_put(x_cur
, nx_cur
);
2959 xfrm_states_delete(x_new
, nx_new
);
2963 EXPORT_SYMBOL(xfrm_migrate
);