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 int sysctl_xfrm_larval_drop __read_mostly
= 1;
39 #ifdef CONFIG_XFRM_STATISTICS
40 DEFINE_SNMP_STAT(struct linux_xfrm_mib
, xfrm_statistics
) __read_mostly
;
41 EXPORT_SYMBOL(xfrm_statistics
);
44 DEFINE_MUTEX(xfrm_cfg_mutex
);
45 EXPORT_SYMBOL(xfrm_cfg_mutex
);
47 static DEFINE_RWLOCK(xfrm_policy_lock
);
49 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
50 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
52 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
54 static HLIST_HEAD(xfrm_policy_gc_list
);
55 static DEFINE_SPINLOCK(xfrm_policy_gc_lock
);
57 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
58 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
59 static void xfrm_init_pmtu(struct dst_entry
*dst
);
62 __xfrm4_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
64 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
65 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
66 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
67 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
68 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
69 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
73 __xfrm6_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
75 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
76 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
77 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
78 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
79 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
80 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
83 int xfrm_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
,
84 unsigned short family
)
88 return __xfrm4_selector_match(sel
, fl
);
90 return __xfrm6_selector_match(sel
, fl
);
95 static inline struct dst_entry
*__xfrm_dst_lookup(int tos
,
96 xfrm_address_t
*saddr
,
97 xfrm_address_t
*daddr
,
100 struct xfrm_policy_afinfo
*afinfo
;
101 struct dst_entry
*dst
;
103 afinfo
= xfrm_policy_get_afinfo(family
);
104 if (unlikely(afinfo
== NULL
))
105 return ERR_PTR(-EAFNOSUPPORT
);
107 dst
= afinfo
->dst_lookup(tos
, saddr
, daddr
);
109 xfrm_policy_put_afinfo(afinfo
);
114 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
, int tos
,
115 xfrm_address_t
*prev_saddr
,
116 xfrm_address_t
*prev_daddr
,
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(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
);
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);
223 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
227 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
229 struct xfrm_policy
*policy
;
231 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
234 write_pnet(&policy
->xp_net
, net
);
235 INIT_LIST_HEAD(&policy
->walk
.all
);
236 INIT_HLIST_NODE(&policy
->bydst
);
237 INIT_HLIST_NODE(&policy
->byidx
);
238 rwlock_init(&policy
->lock
);
239 atomic_set(&policy
->refcnt
, 1);
240 setup_timer(&policy
->timer
, xfrm_policy_timer
,
241 (unsigned long)policy
);
245 EXPORT_SYMBOL(xfrm_policy_alloc
);
247 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
249 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
251 BUG_ON(!policy
->walk
.dead
);
253 BUG_ON(policy
->bundles
);
255 if (del_timer(&policy
->timer
))
258 security_xfrm_policy_free(policy
->security
);
261 EXPORT_SYMBOL(xfrm_policy_destroy
);
263 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
265 struct dst_entry
*dst
;
267 while ((dst
= policy
->bundles
) != NULL
) {
268 policy
->bundles
= dst
->next
;
272 if (del_timer(&policy
->timer
))
273 atomic_dec(&policy
->refcnt
);
275 if (atomic_read(&policy
->refcnt
) > 1)
278 xfrm_pol_put(policy
);
281 static void xfrm_policy_gc_task(struct work_struct
*work
)
283 struct xfrm_policy
*policy
;
284 struct hlist_node
*entry
, *tmp
;
285 struct hlist_head gc_list
;
287 spin_lock_bh(&xfrm_policy_gc_lock
);
288 gc_list
.first
= xfrm_policy_gc_list
.first
;
289 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
290 spin_unlock_bh(&xfrm_policy_gc_lock
);
292 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
293 xfrm_policy_gc_kill(policy
);
295 static DECLARE_WORK(xfrm_policy_gc_work
, xfrm_policy_gc_task
);
297 /* Rule must be locked. Release descentant resources, announce
298 * entry dead. The rule must be unlinked from lists to the moment.
301 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
305 write_lock_bh(&policy
->lock
);
306 dead
= policy
->walk
.dead
;
307 policy
->walk
.dead
= 1;
308 write_unlock_bh(&policy
->lock
);
310 if (unlikely(dead
)) {
315 spin_lock_bh(&xfrm_policy_gc_lock
);
316 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
317 spin_unlock_bh(&xfrm_policy_gc_lock
);
319 schedule_work(&xfrm_policy_gc_work
);
322 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
324 static inline unsigned int idx_hash(struct net
*net
, u32 index
)
326 return __idx_hash(index
, net
->xfrm
.policy_idx_hmask
);
329 static struct hlist_head
*policy_hash_bysel(struct net
*net
, struct xfrm_selector
*sel
, unsigned short family
, int dir
)
331 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
332 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
334 return (hash
== hmask
+ 1 ?
335 &net
->xfrm
.policy_inexact
[dir
] :
336 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
339 static struct hlist_head
*policy_hash_direct(struct net
*net
, xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
341 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
342 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
344 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
347 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
348 struct hlist_head
*ndsttable
,
349 unsigned int nhashmask
)
351 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
352 struct xfrm_policy
*pol
;
356 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
359 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
360 pol
->family
, nhashmask
);
363 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
369 hlist_add_after(entry0
, &pol
->bydst
);
373 if (!hlist_empty(list
)) {
379 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
380 struct hlist_head
*nidxtable
,
381 unsigned int nhashmask
)
383 struct hlist_node
*entry
, *tmp
;
384 struct xfrm_policy
*pol
;
386 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
389 h
= __idx_hash(pol
->index
, nhashmask
);
390 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
394 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
396 return ((old_hmask
+ 1) << 1) - 1;
399 static void xfrm_bydst_resize(struct net
*net
, int dir
)
401 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
402 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
403 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
404 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
405 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
411 write_lock_bh(&xfrm_policy_lock
);
413 for (i
= hmask
; i
>= 0; i
--)
414 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
416 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
417 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
419 write_unlock_bh(&xfrm_policy_lock
);
421 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
424 static void xfrm_byidx_resize(struct net
*net
, int total
)
426 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
427 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
428 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
429 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
430 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
436 write_lock_bh(&xfrm_policy_lock
);
438 for (i
= hmask
; i
>= 0; i
--)
439 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
441 net
->xfrm
.policy_byidx
= nidx
;
442 net
->xfrm
.policy_idx_hmask
= nhashmask
;
444 write_unlock_bh(&xfrm_policy_lock
);
446 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
449 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
451 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
452 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
457 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
464 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
466 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
468 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
475 void xfrm_spd_getinfo(struct xfrmk_spdinfo
*si
)
477 read_lock_bh(&xfrm_policy_lock
);
478 si
->incnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
];
479 si
->outcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
];
480 si
->fwdcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
];
481 si
->inscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
482 si
->outscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
483 si
->fwdscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
484 si
->spdhcnt
= init_net
.xfrm
.policy_idx_hmask
;
485 si
->spdhmcnt
= xfrm_policy_hashmax
;
486 read_unlock_bh(&xfrm_policy_lock
);
488 EXPORT_SYMBOL(xfrm_spd_getinfo
);
490 static DEFINE_MUTEX(hash_resize_mutex
);
491 static void xfrm_hash_resize(struct work_struct
*work
)
493 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
496 mutex_lock(&hash_resize_mutex
);
499 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
500 if (xfrm_bydst_should_resize(net
, dir
, &total
))
501 xfrm_bydst_resize(net
, dir
);
503 if (xfrm_byidx_should_resize(net
, total
))
504 xfrm_byidx_resize(net
, total
);
506 mutex_unlock(&hash_resize_mutex
);
509 /* Generate new index... KAME seems to generate them ordered by cost
510 * of an absolute inpredictability of ordering of rules. This will not pass. */
511 static u32
xfrm_gen_index(struct net
*net
, int dir
)
513 static u32 idx_generator
;
516 struct hlist_node
*entry
;
517 struct hlist_head
*list
;
518 struct xfrm_policy
*p
;
522 idx
= (idx_generator
| dir
);
526 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
528 hlist_for_each_entry(p
, entry
, list
, byidx
) {
529 if (p
->index
== idx
) {
539 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
541 u32
*p1
= (u32
*) s1
;
542 u32
*p2
= (u32
*) s2
;
543 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
546 for (i
= 0; i
< len
; i
++) {
554 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
556 struct net
*net
= xp_net(policy
);
557 struct xfrm_policy
*pol
;
558 struct xfrm_policy
*delpol
;
559 struct hlist_head
*chain
;
560 struct hlist_node
*entry
, *newpos
;
561 struct dst_entry
*gc_list
;
563 write_lock_bh(&xfrm_policy_lock
);
564 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
567 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
568 if (pol
->type
== policy
->type
&&
569 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
570 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
573 write_unlock_bh(&xfrm_policy_lock
);
577 if (policy
->priority
> pol
->priority
)
579 } else if (policy
->priority
>= pol
->priority
) {
580 newpos
= &pol
->bydst
;
587 hlist_add_after(newpos
, &policy
->bydst
);
589 hlist_add_head(&policy
->bydst
, chain
);
590 xfrm_pol_hold(policy
);
591 net
->xfrm
.policy_count
[dir
]++;
592 atomic_inc(&flow_cache_genid
);
594 hlist_del(&delpol
->bydst
);
595 hlist_del(&delpol
->byidx
);
596 list_del(&delpol
->walk
.all
);
597 net
->xfrm
.policy_count
[dir
]--;
599 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
600 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
601 policy
->curlft
.add_time
= get_seconds();
602 policy
->curlft
.use_time
= 0;
603 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
604 xfrm_pol_hold(policy
);
605 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
606 write_unlock_bh(&xfrm_policy_lock
);
609 xfrm_policy_kill(delpol
);
610 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
611 schedule_work(&net
->xfrm
.policy_hash_work
);
613 read_lock_bh(&xfrm_policy_lock
);
615 entry
= &policy
->bydst
;
616 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
617 struct dst_entry
*dst
;
619 write_lock(&policy
->lock
);
620 dst
= policy
->bundles
;
622 struct dst_entry
*tail
= dst
;
625 tail
->next
= gc_list
;
628 policy
->bundles
= NULL
;
630 write_unlock(&policy
->lock
);
632 read_unlock_bh(&xfrm_policy_lock
);
635 struct dst_entry
*dst
= gc_list
;
643 EXPORT_SYMBOL(xfrm_policy_insert
);
645 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u8 type
, int dir
,
646 struct xfrm_selector
*sel
,
647 struct xfrm_sec_ctx
*ctx
, int delete,
650 struct xfrm_policy
*pol
, *ret
;
651 struct hlist_head
*chain
;
652 struct hlist_node
*entry
;
655 write_lock_bh(&xfrm_policy_lock
);
656 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
658 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
659 if (pol
->type
== type
&&
660 !selector_cmp(sel
, &pol
->selector
) &&
661 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
664 *err
= security_xfrm_policy_delete(
667 write_unlock_bh(&xfrm_policy_lock
);
670 hlist_del(&pol
->bydst
);
671 hlist_del(&pol
->byidx
);
672 list_del(&pol
->walk
.all
);
673 net
->xfrm
.policy_count
[dir
]--;
679 write_unlock_bh(&xfrm_policy_lock
);
682 atomic_inc(&flow_cache_genid
);
683 xfrm_policy_kill(ret
);
687 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
689 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u8 type
, int dir
, u32 id
,
690 int delete, int *err
)
692 struct xfrm_policy
*pol
, *ret
;
693 struct hlist_head
*chain
;
694 struct hlist_node
*entry
;
697 if (xfrm_policy_id2dir(id
) != dir
)
701 write_lock_bh(&xfrm_policy_lock
);
702 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
704 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
705 if (pol
->type
== type
&& pol
->index
== id
) {
708 *err
= security_xfrm_policy_delete(
711 write_unlock_bh(&xfrm_policy_lock
);
714 hlist_del(&pol
->bydst
);
715 hlist_del(&pol
->byidx
);
716 list_del(&pol
->walk
.all
);
717 net
->xfrm
.policy_count
[dir
]--;
723 write_unlock_bh(&xfrm_policy_lock
);
726 atomic_inc(&flow_cache_genid
);
727 xfrm_policy_kill(ret
);
731 EXPORT_SYMBOL(xfrm_policy_byid
);
733 #ifdef CONFIG_SECURITY_NETWORK_XFRM
735 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
739 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
740 struct xfrm_policy
*pol
;
741 struct hlist_node
*entry
;
744 hlist_for_each_entry(pol
, entry
,
745 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
746 if (pol
->type
!= type
)
748 err
= security_xfrm_policy_delete(pol
->security
);
750 xfrm_audit_policy_delete(pol
, 0,
751 audit_info
->loginuid
,
752 audit_info
->sessionid
,
757 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
758 hlist_for_each_entry(pol
, entry
,
759 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
761 if (pol
->type
!= type
)
763 err
= security_xfrm_policy_delete(
766 xfrm_audit_policy_delete(pol
, 0,
767 audit_info
->loginuid
,
768 audit_info
->sessionid
,
779 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
785 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
789 write_lock_bh(&xfrm_policy_lock
);
791 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
795 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
796 struct xfrm_policy
*pol
;
797 struct hlist_node
*entry
;
802 hlist_for_each_entry(pol
, entry
,
803 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
804 if (pol
->type
!= type
)
806 hlist_del(&pol
->bydst
);
807 hlist_del(&pol
->byidx
);
808 write_unlock_bh(&xfrm_policy_lock
);
810 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
811 audit_info
->sessionid
,
814 xfrm_policy_kill(pol
);
817 write_lock_bh(&xfrm_policy_lock
);
821 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
823 hlist_for_each_entry(pol
, entry
,
824 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
826 if (pol
->type
!= type
)
828 hlist_del(&pol
->bydst
);
829 hlist_del(&pol
->byidx
);
830 list_del(&pol
->walk
.all
);
831 write_unlock_bh(&xfrm_policy_lock
);
833 xfrm_audit_policy_delete(pol
, 1,
834 audit_info
->loginuid
,
835 audit_info
->sessionid
,
837 xfrm_policy_kill(pol
);
840 write_lock_bh(&xfrm_policy_lock
);
845 net
->xfrm
.policy_count
[dir
] -= killed
;
847 atomic_inc(&flow_cache_genid
);
849 write_unlock_bh(&xfrm_policy_lock
);
852 EXPORT_SYMBOL(xfrm_policy_flush
);
854 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
855 int (*func
)(struct xfrm_policy
*, int, int, void*),
858 struct xfrm_policy
*pol
;
859 struct xfrm_policy_walk_entry
*x
;
862 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
863 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
866 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
869 write_lock_bh(&xfrm_policy_lock
);
870 if (list_empty(&walk
->walk
.all
))
871 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
873 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
874 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
877 pol
= container_of(x
, struct xfrm_policy
, walk
);
878 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
879 walk
->type
!= pol
->type
)
881 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
884 list_move_tail(&walk
->walk
.all
, &x
->all
);
889 if (walk
->seq
== 0) {
893 list_del_init(&walk
->walk
.all
);
895 write_unlock_bh(&xfrm_policy_lock
);
898 EXPORT_SYMBOL(xfrm_policy_walk
);
900 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
902 INIT_LIST_HEAD(&walk
->walk
.all
);
907 EXPORT_SYMBOL(xfrm_policy_walk_init
);
909 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
911 if (list_empty(&walk
->walk
.all
))
914 write_lock_bh(&xfrm_policy_lock
);
915 list_del(&walk
->walk
.all
);
916 write_unlock_bh(&xfrm_policy_lock
);
918 EXPORT_SYMBOL(xfrm_policy_walk_done
);
921 * Find policy to apply to this flow.
923 * Returns 0 if policy found, else an -errno.
925 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
926 u8 type
, u16 family
, int dir
)
928 struct xfrm_selector
*sel
= &pol
->selector
;
929 int match
, ret
= -ESRCH
;
931 if (pol
->family
!= family
||
935 match
= xfrm_selector_match(sel
, fl
, family
);
937 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
943 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
948 struct xfrm_policy
*pol
, *ret
;
949 xfrm_address_t
*daddr
, *saddr
;
950 struct hlist_node
*entry
;
951 struct hlist_head
*chain
;
954 daddr
= xfrm_flowi_daddr(fl
, family
);
955 saddr
= xfrm_flowi_saddr(fl
, family
);
956 if (unlikely(!daddr
|| !saddr
))
959 read_lock_bh(&xfrm_policy_lock
);
960 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
962 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
963 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
973 priority
= ret
->priority
;
977 chain
= &net
->xfrm
.policy_inexact
[dir
];
978 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
979 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
987 } else if (pol
->priority
< priority
) {
995 read_unlock_bh(&xfrm_policy_lock
);
1000 static int xfrm_policy_lookup(struct net
*net
, struct flowi
*fl
, u16 family
,
1001 u8 dir
, void **objp
, atomic_t
**obj_refp
)
1003 struct xfrm_policy
*pol
;
1006 #ifdef CONFIG_XFRM_SUB_POLICY
1007 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1015 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1020 #ifdef CONFIG_XFRM_SUB_POLICY
1023 if ((*objp
= (void *) pol
) != NULL
)
1024 *obj_refp
= &pol
->refcnt
;
1028 static inline int policy_to_flow_dir(int dir
)
1030 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1031 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1032 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1036 case XFRM_POLICY_IN
:
1038 case XFRM_POLICY_OUT
:
1039 return FLOW_DIR_OUT
;
1040 case XFRM_POLICY_FWD
:
1041 return FLOW_DIR_FWD
;
1045 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1047 struct xfrm_policy
*pol
;
1049 read_lock_bh(&xfrm_policy_lock
);
1050 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1051 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1056 err
= security_xfrm_policy_lookup(pol
->security
,
1058 policy_to_flow_dir(dir
));
1061 else if (err
== -ESRCH
)
1068 read_unlock_bh(&xfrm_policy_lock
);
1072 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1074 struct net
*net
= xp_net(pol
);
1075 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1078 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1079 hlist_add_head(&pol
->bydst
, chain
);
1080 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1081 net
->xfrm
.policy_count
[dir
]++;
1084 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1085 schedule_work(&net
->xfrm
.policy_hash_work
);
1088 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1091 struct net
*net
= xp_net(pol
);
1093 if (hlist_unhashed(&pol
->bydst
))
1096 hlist_del(&pol
->bydst
);
1097 hlist_del(&pol
->byidx
);
1098 list_del(&pol
->walk
.all
);
1099 net
->xfrm
.policy_count
[dir
]--;
1104 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1106 write_lock_bh(&xfrm_policy_lock
);
1107 pol
= __xfrm_policy_unlink(pol
, dir
);
1108 write_unlock_bh(&xfrm_policy_lock
);
1110 if (dir
< XFRM_POLICY_MAX
)
1111 atomic_inc(&flow_cache_genid
);
1112 xfrm_policy_kill(pol
);
1117 EXPORT_SYMBOL(xfrm_policy_delete
);
1119 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1121 struct net
*net
= xp_net(pol
);
1122 struct xfrm_policy
*old_pol
;
1124 #ifdef CONFIG_XFRM_SUB_POLICY
1125 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1129 write_lock_bh(&xfrm_policy_lock
);
1130 old_pol
= sk
->sk_policy
[dir
];
1131 sk
->sk_policy
[dir
] = pol
;
1133 pol
->curlft
.add_time
= get_seconds();
1134 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1135 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1138 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1139 write_unlock_bh(&xfrm_policy_lock
);
1142 xfrm_policy_kill(old_pol
);
1147 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1149 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1152 newp
->selector
= old
->selector
;
1153 if (security_xfrm_policy_clone(old
->security
,
1156 return NULL
; /* ENOMEM */
1158 newp
->lft
= old
->lft
;
1159 newp
->curlft
= old
->curlft
;
1160 newp
->action
= old
->action
;
1161 newp
->flags
= old
->flags
;
1162 newp
->xfrm_nr
= old
->xfrm_nr
;
1163 newp
->index
= old
->index
;
1164 newp
->type
= old
->type
;
1165 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1166 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1167 write_lock_bh(&xfrm_policy_lock
);
1168 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1169 write_unlock_bh(&xfrm_policy_lock
);
1175 int __xfrm_sk_clone_policy(struct sock
*sk
)
1177 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1178 *p1
= sk
->sk_policy
[1];
1180 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1181 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1183 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1189 xfrm_get_saddr(xfrm_address_t
*local
, xfrm_address_t
*remote
,
1190 unsigned short family
)
1193 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1195 if (unlikely(afinfo
== NULL
))
1197 err
= afinfo
->get_saddr(local
, remote
);
1198 xfrm_policy_put_afinfo(afinfo
);
1202 /* Resolve list of templates for the flow, given policy. */
1205 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1206 struct xfrm_state
**xfrm
,
1207 unsigned short family
)
1211 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1212 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1215 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1216 struct xfrm_state
*x
;
1217 xfrm_address_t
*remote
= daddr
;
1218 xfrm_address_t
*local
= saddr
;
1219 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1221 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1222 tmpl
->mode
== XFRM_MODE_BEET
) {
1223 remote
= &tmpl
->id
.daddr
;
1224 local
= &tmpl
->saddr
;
1225 family
= tmpl
->encap_family
;
1226 if (xfrm_addr_any(local
, family
)) {
1227 error
= xfrm_get_saddr(&tmp
, remote
, family
);
1234 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1236 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1243 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1247 else if (error
== -ESRCH
)
1250 if (!tmpl
->optional
)
1256 for (nx
--; nx
>=0; nx
--)
1257 xfrm_state_put(xfrm
[nx
]);
1262 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1263 struct xfrm_state
**xfrm
,
1264 unsigned short family
)
1266 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1267 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1273 for (i
= 0; i
< npols
; i
++) {
1274 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1279 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1287 /* found states are sorted for outbound processing */
1289 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1294 for (cnx
--; cnx
>=0; cnx
--)
1295 xfrm_state_put(tpp
[cnx
]);
1300 /* Check that the bundle accepts the flow and its components are
1304 static struct dst_entry
*
1305 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1307 struct dst_entry
*x
;
1308 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1309 if (unlikely(afinfo
== NULL
))
1310 return ERR_PTR(-EINVAL
);
1311 x
= afinfo
->find_bundle(fl
, policy
);
1312 xfrm_policy_put_afinfo(afinfo
);
1316 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1318 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1324 tos
= afinfo
->get_tos(fl
);
1326 xfrm_policy_put_afinfo(afinfo
);
1331 static inline struct xfrm_dst
*xfrm_alloc_dst(int family
)
1333 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1334 struct xfrm_dst
*xdst
;
1337 return ERR_PTR(-EINVAL
);
1339 xdst
= dst_alloc(afinfo
->dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1341 xfrm_policy_put_afinfo(afinfo
);
1346 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1349 struct xfrm_policy_afinfo
*afinfo
=
1350 xfrm_policy_get_afinfo(dst
->ops
->family
);
1356 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1358 xfrm_policy_put_afinfo(afinfo
);
1363 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1365 struct xfrm_policy_afinfo
*afinfo
=
1366 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1372 err
= afinfo
->fill_dst(xdst
, dev
);
1374 xfrm_policy_put_afinfo(afinfo
);
1379 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1380 * all the metrics... Shortly, bundle a bundle.
1383 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1384 struct xfrm_state
**xfrm
, int nx
,
1386 struct dst_entry
*dst
)
1388 unsigned long now
= jiffies
;
1389 struct net_device
*dev
;
1390 struct dst_entry
*dst_prev
= NULL
;
1391 struct dst_entry
*dst0
= NULL
;
1395 int nfheader_len
= 0;
1396 int trailer_len
= 0;
1398 int family
= policy
->selector
.family
;
1399 xfrm_address_t saddr
, daddr
;
1401 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1403 tos
= xfrm_get_tos(fl
, family
);
1410 for (; i
< nx
; i
++) {
1411 struct xfrm_dst
*xdst
= xfrm_alloc_dst(family
);
1412 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1414 err
= PTR_ERR(xdst
);
1423 dst_prev
->child
= dst_clone(dst1
);
1424 dst1
->flags
|= DST_NOHASH
;
1428 memcpy(&dst1
->metrics
, &dst
->metrics
, sizeof(dst
->metrics
));
1430 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1431 family
= xfrm
[i
]->props
.family
;
1432 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1440 dst1
->xfrm
= xfrm
[i
];
1441 xdst
->genid
= xfrm
[i
]->genid
;
1443 dst1
->obsolete
= -1;
1444 dst1
->flags
|= DST_HOST
;
1445 dst1
->lastuse
= now
;
1447 dst1
->input
= dst_discard
;
1448 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1450 dst1
->next
= dst_prev
;
1453 header_len
+= xfrm
[i
]->props
.header_len
;
1454 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1455 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1456 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1459 dst_prev
->child
= dst
;
1467 /* Copy neighbout for reachability confirmation */
1468 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1470 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1471 xfrm_init_pmtu(dst_prev
);
1473 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1474 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1476 err
= xfrm_fill_dst(xdst
, dev
);
1480 dst_prev
->header_len
= header_len
;
1481 dst_prev
->trailer_len
= trailer_len
;
1482 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1483 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1491 xfrm_state_put(xfrm
[i
]);
1495 dst0
= ERR_PTR(err
);
1500 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1503 *target
= kmalloc(size
, GFP_ATOMIC
);
1507 memcpy(*target
, src
, size
);
1512 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1514 #ifdef CONFIG_XFRM_SUB_POLICY
1515 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1516 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1524 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1526 #ifdef CONFIG_XFRM_SUB_POLICY
1527 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1528 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1534 static int stale_bundle(struct dst_entry
*dst
);
1536 /* Main function: finds/creates a bundle for given flow.
1538 * At the moment we eat a raw IP route. Mostly to speed up lookups
1539 * on interfaces with disabled IPsec.
1541 int __xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1542 struct sock
*sk
, int flags
)
1544 struct xfrm_policy
*policy
;
1545 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1550 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1551 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1556 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1559 genid
= atomic_read(&flow_cache_genid
);
1561 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1567 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1568 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1569 err
= PTR_ERR(policy
);
1570 if (IS_ERR(policy
)) {
1571 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1577 /* To accelerate a bit... */
1578 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1579 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1582 policy
= flow_cache_lookup(net
, fl
, dst_orig
->ops
->family
,
1583 dir
, xfrm_policy_lookup
);
1584 err
= PTR_ERR(policy
);
1585 if (IS_ERR(policy
)) {
1586 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1594 family
= dst_orig
->ops
->family
;
1597 xfrm_nr
+= pols
[0]->xfrm_nr
;
1600 if ((flags
& XFRM_LOOKUP_ICMP
) && !(policy
->flags
& XFRM_POLICY_ICMP
))
1603 policy
->curlft
.use_time
= get_seconds();
1605 switch (policy
->action
) {
1607 case XFRM_POLICY_BLOCK
:
1608 /* Prohibit the flow */
1609 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1613 case XFRM_POLICY_ALLOW
:
1614 #ifndef CONFIG_XFRM_SUB_POLICY
1615 if (policy
->xfrm_nr
== 0) {
1616 /* Flow passes not transformed. */
1617 xfrm_pol_put(policy
);
1622 /* Try to find matching bundle.
1624 * LATER: help from flow cache. It is optional, this
1625 * is required only for output policy.
1627 dst
= xfrm_find_bundle(fl
, policy
, family
);
1629 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1637 #ifdef CONFIG_XFRM_SUB_POLICY
1638 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1639 pols
[1] = xfrm_policy_lookup_bytype(net
,
1640 XFRM_POLICY_TYPE_MAIN
,
1644 if (IS_ERR(pols
[1])) {
1645 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1646 err
= PTR_ERR(pols
[1]);
1649 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1650 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1655 xfrm_nr
+= pols
[1]->xfrm_nr
;
1660 * Because neither flowi nor bundle information knows about
1661 * transformation template size. On more than one policy usage
1662 * we can realize whether all of them is bypass or not after
1663 * they are searched. See above not-transformed bypass
1664 * is surrounded by non-sub policy configuration, too.
1667 /* Flow passes not transformed. */
1668 xfrm_pols_put(pols
, npols
);
1673 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1675 if (unlikely(nx
<0)) {
1677 if (err
== -EAGAIN
&& sysctl_xfrm_larval_drop
) {
1678 /* EREMOTE tells the caller to generate
1679 * a one-shot blackhole route.
1681 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1682 xfrm_pol_put(policy
);
1685 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1686 DECLARE_WAITQUEUE(wait
, current
);
1688 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1689 set_current_state(TASK_INTERRUPTIBLE
);
1691 set_current_state(TASK_RUNNING
);
1692 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1694 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1696 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1697 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1701 if (nx
== -EAGAIN
||
1702 genid
!= atomic_read(&flow_cache_genid
)) {
1703 xfrm_pols_put(pols
, npols
);
1709 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1714 /* Flow passes not transformed. */
1715 xfrm_pols_put(pols
, npols
);
1719 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1722 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1726 for (pi
= 0; pi
< npols
; pi
++) {
1727 read_lock_bh(&pols
[pi
]->lock
);
1728 pol_dead
|= pols
[pi
]->walk
.dead
;
1729 read_unlock_bh(&pols
[pi
]->lock
);
1732 write_lock_bh(&policy
->lock
);
1733 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1734 /* Wow! While we worked on resolving, this
1735 * policy has gone. Retry. It is not paranoia,
1736 * we just cannot enlist new bundle to dead object.
1737 * We can't enlist stable bundles either.
1739 write_unlock_bh(&policy
->lock
);
1743 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD
);
1745 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1746 err
= -EHOSTUNREACH
;
1751 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1753 err
= xfrm_dst_update_origin(dst
, fl
);
1754 if (unlikely(err
)) {
1755 write_unlock_bh(&policy
->lock
);
1757 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1761 dst
->next
= policy
->bundles
;
1762 policy
->bundles
= dst
;
1764 write_unlock_bh(&policy
->lock
);
1767 dst_release(dst_orig
);
1768 xfrm_pols_put(pols
, npols
);
1772 xfrm_pols_put(pols
, npols
);
1774 dst_release(dst_orig
);
1780 if (flags
& XFRM_LOOKUP_ICMP
)
1784 EXPORT_SYMBOL(__xfrm_lookup
);
1786 int xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1787 struct sock
*sk
, int flags
)
1789 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1791 if (err
== -EREMOTE
) {
1792 dst_release(*dst_p
);
1799 EXPORT_SYMBOL(xfrm_lookup
);
1802 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1804 struct xfrm_state
*x
;
1806 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1808 x
= skb
->sp
->xvec
[idx
];
1809 if (!x
->type
->reject
)
1811 return x
->type
->reject(x
, skb
, fl
);
1814 /* When skb is transformed back to its "native" form, we have to
1815 * check policy restrictions. At the moment we make this in maximally
1816 * stupid way. Shame on me. :-) Of course, connected sockets must
1817 * have policy cached at them.
1821 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1822 unsigned short family
)
1824 if (xfrm_state_kern(x
))
1825 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1826 return x
->id
.proto
== tmpl
->id
.proto
&&
1827 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1828 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1829 x
->props
.mode
== tmpl
->mode
&&
1830 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1831 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1832 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1833 xfrm_state_addr_cmp(tmpl
, x
, family
));
1837 * 0 or more than 0 is returned when validation is succeeded (either bypass
1838 * because of optional transport mode, or next index of the mathced secpath
1839 * state with the template.
1840 * -1 is returned when no matching template is found.
1841 * Otherwise "-2 - errored_index" is returned.
1844 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1845 unsigned short family
)
1849 if (tmpl
->optional
) {
1850 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1854 for (; idx
< sp
->len
; idx
++) {
1855 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1857 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1866 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1867 unsigned int family
, int reverse
)
1869 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1872 if (unlikely(afinfo
== NULL
))
1873 return -EAFNOSUPPORT
;
1875 afinfo
->decode_session(skb
, fl
, reverse
);
1876 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1877 xfrm_policy_put_afinfo(afinfo
);
1880 EXPORT_SYMBOL(__xfrm_decode_session
);
1882 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1884 for (; k
< sp
->len
; k
++) {
1885 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1894 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1895 unsigned short family
)
1897 struct net
*net
= dev_net(skb
->dev
);
1898 struct xfrm_policy
*pol
;
1899 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1908 reverse
= dir
& ~XFRM_POLICY_MASK
;
1909 dir
&= XFRM_POLICY_MASK
;
1910 fl_dir
= policy_to_flow_dir(dir
);
1912 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1913 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
1917 nf_nat_decode_session(skb
, &fl
, family
);
1919 /* First, check used SA against their selectors. */
1923 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1924 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1925 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1926 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH
);
1933 if (sk
&& sk
->sk_policy
[dir
]) {
1934 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1936 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1942 pol
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
1943 xfrm_policy_lookup
);
1946 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1951 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1952 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1953 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS
);
1959 pol
->curlft
.use_time
= get_seconds();
1963 #ifdef CONFIG_XFRM_SUB_POLICY
1964 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1965 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
1969 if (IS_ERR(pols
[1])) {
1970 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1973 pols
[1]->curlft
.use_time
= get_seconds();
1979 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1980 struct sec_path
*sp
;
1981 static struct sec_path dummy
;
1982 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1983 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1984 struct xfrm_tmpl
**tpp
= tp
;
1988 if ((sp
= skb
->sp
) == NULL
)
1991 for (pi
= 0; pi
< npols
; pi
++) {
1992 if (pols
[pi
] != pol
&&
1993 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
1994 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
1997 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1998 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR
);
2001 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2002 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2006 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2010 /* For each tunnel xfrm, find the first matching tmpl.
2011 * For each tmpl before that, find corresponding xfrm.
2012 * Order is _important_. Later we will implement
2013 * some barriers, but at the moment barriers
2014 * are implied between each two transformations.
2016 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2017 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2020 /* "-2 - errored_index" returned */
2022 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2027 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2028 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2032 xfrm_pols_put(pols
, npols
);
2035 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
2038 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2040 xfrm_pols_put(pols
, npols
);
2043 EXPORT_SYMBOL(__xfrm_policy_check
);
2045 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2047 struct net
*net
= dev_net(skb
->dev
);
2050 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2051 /* XXX: we should have something like FWDHDRERROR here. */
2052 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
2056 return xfrm_lookup(net
, &skb
->dst
, &fl
, NULL
, 0) == 0;
2058 EXPORT_SYMBOL(__xfrm_route_forward
);
2060 /* Optimize later using cookies and generation ids. */
2062 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2064 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2065 * to "-1" to force all XFRM destinations to get validated by
2066 * dst_ops->check on every use. We do this because when a
2067 * normal route referenced by an XFRM dst is obsoleted we do
2068 * not go looking around for all parent referencing XFRM dsts
2069 * so that we can invalidate them. It is just too much work.
2070 * Instead we make the checks here on every use. For example:
2072 * XFRM dst A --> IPv4 dst X
2074 * X is the "xdst->route" of A (X is also the "dst->path" of A
2075 * in this example). If X is marked obsolete, "A" will not
2076 * notice. That's what we are validating here via the
2077 * stale_bundle() check.
2079 * When a policy's bundle is pruned, we dst_free() the XFRM
2080 * dst which causes it's ->obsolete field to be set to a
2081 * positive non-zero integer. If an XFRM dst has been pruned
2082 * like this, we want to force a new route lookup.
2084 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2090 static int stale_bundle(struct dst_entry
*dst
)
2092 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2095 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2097 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2098 dst
->dev
= dev_net(dev
)->loopback_dev
;
2103 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2105 static void xfrm_link_failure(struct sk_buff
*skb
)
2107 /* Impossible. Such dst must be popped before reaches point of failure. */
2111 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2114 if (dst
->obsolete
) {
2122 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
2124 struct dst_entry
*dst
, **dstp
;
2126 write_lock(&pol
->lock
);
2127 dstp
= &pol
->bundles
;
2128 while ((dst
=*dstp
) != NULL
) {
2131 dst
->next
= *gc_list_p
;
2137 write_unlock(&pol
->lock
);
2140 static void xfrm_prune_bundles(struct net
*net
, int (*func
)(struct dst_entry
*))
2142 struct dst_entry
*gc_list
= NULL
;
2145 read_lock_bh(&xfrm_policy_lock
);
2146 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2147 struct xfrm_policy
*pol
;
2148 struct hlist_node
*entry
;
2149 struct hlist_head
*table
;
2152 hlist_for_each_entry(pol
, entry
,
2153 &net
->xfrm
.policy_inexact
[dir
], bydst
)
2154 prune_one_bundle(pol
, func
, &gc_list
);
2156 table
= net
->xfrm
.policy_bydst
[dir
].table
;
2157 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
2158 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
2159 prune_one_bundle(pol
, func
, &gc_list
);
2162 read_unlock_bh(&xfrm_policy_lock
);
2165 struct dst_entry
*dst
= gc_list
;
2166 gc_list
= dst
->next
;
2171 static int unused_bundle(struct dst_entry
*dst
)
2173 return !atomic_read(&dst
->__refcnt
);
2176 static void __xfrm_garbage_collect(void)
2178 xfrm_prune_bundles(&init_net
, unused_bundle
);
2181 static int xfrm_flush_bundles(struct net
*net
)
2183 xfrm_prune_bundles(net
, stale_bundle
);
2187 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2190 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2191 u32 pmtu
, route_mtu_cached
;
2193 pmtu
= dst_mtu(dst
->child
);
2194 xdst
->child_mtu_cached
= pmtu
;
2196 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2198 route_mtu_cached
= dst_mtu(xdst
->route
);
2199 xdst
->route_mtu_cached
= route_mtu_cached
;
2201 if (pmtu
> route_mtu_cached
)
2202 pmtu
= route_mtu_cached
;
2204 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
2205 } while ((dst
= dst
->next
));
2208 /* Check that the bundle accepts the flow and its components are
2212 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2213 struct flowi
*fl
, int family
, int strict
)
2215 struct dst_entry
*dst
= &first
->u
.dst
;
2216 struct xfrm_dst
*last
;
2219 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2220 (dst
->dev
&& !netif_running(dst
->dev
)))
2222 #ifdef CONFIG_XFRM_SUB_POLICY
2224 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2226 if (first
->partner
&&
2227 !xfrm_selector_match(first
->partner
, fl
, family
))
2235 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2237 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2240 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2242 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2244 if (xdst
->genid
!= dst
->xfrm
->genid
)
2248 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2249 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2252 mtu
= dst_mtu(dst
->child
);
2253 if (xdst
->child_mtu_cached
!= mtu
) {
2255 xdst
->child_mtu_cached
= mtu
;
2258 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2260 mtu
= dst_mtu(xdst
->route
);
2261 if (xdst
->route_mtu_cached
!= mtu
) {
2263 xdst
->route_mtu_cached
= mtu
;
2267 } while (dst
->xfrm
);
2272 mtu
= last
->child_mtu_cached
;
2276 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2277 if (mtu
> last
->route_mtu_cached
)
2278 mtu
= last
->route_mtu_cached
;
2279 dst
->metrics
[RTAX_MTU
-1] = mtu
;
2284 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2285 last
->child_mtu_cached
= mtu
;
2291 EXPORT_SYMBOL(xfrm_bundle_ok
);
2293 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2296 if (unlikely(afinfo
== NULL
))
2298 if (unlikely(afinfo
->family
>= NPROTO
))
2299 return -EAFNOSUPPORT
;
2300 write_lock_bh(&xfrm_policy_afinfo_lock
);
2301 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2304 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2305 if (likely(dst_ops
->kmem_cachep
== NULL
))
2306 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2307 if (likely(dst_ops
->check
== NULL
))
2308 dst_ops
->check
= xfrm_dst_check
;
2309 if (likely(dst_ops
->negative_advice
== NULL
))
2310 dst_ops
->negative_advice
= xfrm_negative_advice
;
2311 if (likely(dst_ops
->link_failure
== NULL
))
2312 dst_ops
->link_failure
= xfrm_link_failure
;
2313 if (likely(afinfo
->garbage_collect
== NULL
))
2314 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2315 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2317 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2320 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2322 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2325 if (unlikely(afinfo
== NULL
))
2327 if (unlikely(afinfo
->family
>= NPROTO
))
2328 return -EAFNOSUPPORT
;
2329 write_lock_bh(&xfrm_policy_afinfo_lock
);
2330 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2331 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2334 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2335 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2336 dst_ops
->kmem_cachep
= NULL
;
2337 dst_ops
->check
= NULL
;
2338 dst_ops
->negative_advice
= NULL
;
2339 dst_ops
->link_failure
= NULL
;
2340 afinfo
->garbage_collect
= NULL
;
2343 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2346 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2348 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2350 struct xfrm_policy_afinfo
*afinfo
;
2351 if (unlikely(family
>= NPROTO
))
2353 read_lock(&xfrm_policy_afinfo_lock
);
2354 afinfo
= xfrm_policy_afinfo
[family
];
2355 if (unlikely(!afinfo
))
2356 read_unlock(&xfrm_policy_afinfo_lock
);
2360 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2362 read_unlock(&xfrm_policy_afinfo_lock
);
2365 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2367 struct net_device
*dev
= ptr
;
2371 xfrm_flush_bundles(dev_net(dev
));
2376 static struct notifier_block xfrm_dev_notifier
= {
2377 .notifier_call
= xfrm_dev_event
,
2380 #ifdef CONFIG_XFRM_STATISTICS
2381 static int __init
xfrm_statistics_init(void)
2383 if (snmp_mib_init((void **)xfrm_statistics
,
2384 sizeof(struct linux_xfrm_mib
)) < 0)
2390 static int __net_init
xfrm_policy_init(struct net
*net
)
2392 unsigned int hmask
, sz
;
2395 if (net_eq(net
, &init_net
))
2396 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2397 sizeof(struct xfrm_dst
),
2398 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2402 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2404 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2405 if (!net
->xfrm
.policy_byidx
)
2407 net
->xfrm
.policy_idx_hmask
= hmask
;
2409 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2410 struct xfrm_policy_hash
*htab
;
2412 net
->xfrm
.policy_count
[dir
] = 0;
2413 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2415 htab
= &net
->xfrm
.policy_bydst
[dir
];
2416 htab
->table
= xfrm_hash_alloc(sz
);
2419 htab
->hmask
= hmask
;
2422 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2423 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2424 if (net_eq(net
, &init_net
))
2425 register_netdevice_notifier(&xfrm_dev_notifier
);
2429 for (dir
--; dir
>= 0; dir
--) {
2430 struct xfrm_policy_hash
*htab
;
2432 htab
= &net
->xfrm
.policy_bydst
[dir
];
2433 xfrm_hash_free(htab
->table
, sz
);
2435 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2440 static void xfrm_policy_fini(struct net
*net
)
2445 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2447 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2448 struct xfrm_policy_hash
*htab
;
2450 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2452 htab
= &net
->xfrm
.policy_bydst
[dir
];
2453 sz
= (htab
->hmask
+ 1);
2454 WARN_ON(!hlist_empty(htab
->table
));
2455 xfrm_hash_free(htab
->table
, sz
);
2458 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2459 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2460 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2463 static int __net_init
xfrm_net_init(struct net
*net
)
2467 rv
= xfrm_state_init(net
);
2470 rv
= xfrm_policy_init(net
);
2476 xfrm_state_fini(net
);
2481 static void __net_exit
xfrm_net_exit(struct net
*net
)
2483 xfrm_policy_fini(net
);
2484 xfrm_state_fini(net
);
2487 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2488 .init
= xfrm_net_init
,
2489 .exit
= xfrm_net_exit
,
2492 void __init
xfrm_init(void)
2494 register_pernet_subsys(&xfrm_net_ops
);
2495 #ifdef CONFIG_XFRM_STATISTICS
2496 xfrm_statistics_init();
2499 #ifdef CONFIG_XFRM_STATISTICS
2504 #ifdef CONFIG_AUDITSYSCALL
2505 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2506 struct audit_buffer
*audit_buf
)
2508 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2509 struct xfrm_selector
*sel
= &xp
->selector
;
2512 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2513 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2515 switch(sel
->family
) {
2517 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2518 if (sel
->prefixlen_s
!= 32)
2519 audit_log_format(audit_buf
, " src_prefixlen=%d",
2521 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2522 if (sel
->prefixlen_d
!= 32)
2523 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2527 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2528 if (sel
->prefixlen_s
!= 128)
2529 audit_log_format(audit_buf
, " src_prefixlen=%d",
2531 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2532 if (sel
->prefixlen_d
!= 128)
2533 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2539 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2540 uid_t auid
, u32 sessionid
, u32 secid
)
2542 struct audit_buffer
*audit_buf
;
2544 audit_buf
= xfrm_audit_start("SPD-add");
2545 if (audit_buf
== NULL
)
2547 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2548 audit_log_format(audit_buf
, " res=%u", result
);
2549 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2550 audit_log_end(audit_buf
);
2552 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2554 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2555 uid_t auid
, u32 sessionid
, u32 secid
)
2557 struct audit_buffer
*audit_buf
;
2559 audit_buf
= xfrm_audit_start("SPD-delete");
2560 if (audit_buf
== NULL
)
2562 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2563 audit_log_format(audit_buf
, " res=%u", result
);
2564 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2565 audit_log_end(audit_buf
);
2567 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2570 #ifdef CONFIG_XFRM_MIGRATE
2571 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2572 struct xfrm_selector
*sel_tgt
)
2574 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2575 if (sel_tgt
->family
== sel_cmp
->family
&&
2576 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2577 sel_cmp
->family
) == 0 &&
2578 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2579 sel_cmp
->family
) == 0 &&
2580 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2581 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2585 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2592 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2595 struct xfrm_policy
*pol
, *ret
= NULL
;
2596 struct hlist_node
*entry
;
2597 struct hlist_head
*chain
;
2600 read_lock_bh(&xfrm_policy_lock
);
2601 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2602 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2603 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2604 pol
->type
== type
) {
2606 priority
= ret
->priority
;
2610 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2611 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2612 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2613 pol
->type
== type
&&
2614 pol
->priority
< priority
) {
2623 read_unlock_bh(&xfrm_policy_lock
);
2628 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2632 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2633 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2635 case XFRM_MODE_TUNNEL
:
2636 case XFRM_MODE_BEET
:
2637 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2638 m
->old_family
) == 0 &&
2639 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2640 m
->old_family
) == 0) {
2644 case XFRM_MODE_TRANSPORT
:
2645 /* in case of transport mode, template does not store
2646 any IP addresses, hence we just compare mode and
2657 /* update endpoint address(es) of template(s) */
2658 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2659 struct xfrm_migrate
*m
, int num_migrate
)
2661 struct xfrm_migrate
*mp
;
2662 struct dst_entry
*dst
;
2665 write_lock_bh(&pol
->lock
);
2666 if (unlikely(pol
->walk
.dead
)) {
2667 /* target policy has been deleted */
2668 write_unlock_bh(&pol
->lock
);
2672 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2673 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2674 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2677 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2678 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2680 /* update endpoints */
2681 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2682 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2683 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2684 sizeof(pol
->xfrm_vec
[i
].saddr
));
2685 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2687 while ((dst
= pol
->bundles
) != NULL
) {
2688 pol
->bundles
= dst
->next
;
2694 write_unlock_bh(&pol
->lock
);
2702 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2706 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2709 for (i
= 0; i
< num_migrate
; i
++) {
2710 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2711 m
[i
].old_family
) == 0) &&
2712 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2713 m
[i
].old_family
) == 0))
2715 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2716 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2719 /* check if there is any duplicated entry */
2720 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2721 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2722 sizeof(m
[i
].old_daddr
)) &&
2723 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2724 sizeof(m
[i
].old_saddr
)) &&
2725 m
[i
].proto
== m
[j
].proto
&&
2726 m
[i
].mode
== m
[j
].mode
&&
2727 m
[i
].reqid
== m
[j
].reqid
&&
2728 m
[i
].old_family
== m
[j
].old_family
)
2736 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2737 struct xfrm_migrate
*m
, int num_migrate
,
2738 struct xfrm_kmaddress
*k
)
2740 int i
, err
, nx_cur
= 0, nx_new
= 0;
2741 struct xfrm_policy
*pol
= NULL
;
2742 struct xfrm_state
*x
, *xc
;
2743 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2744 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2745 struct xfrm_migrate
*mp
;
2747 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2750 /* Stage 1 - find policy */
2751 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2756 /* Stage 2 - find and update state(s) */
2757 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2758 if ((x
= xfrm_migrate_state_find(mp
))) {
2761 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2771 /* Stage 3 - update policy */
2772 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2775 /* Stage 4 - delete old state(s) */
2777 xfrm_states_put(x_cur
, nx_cur
);
2778 xfrm_states_delete(x_cur
, nx_cur
);
2781 /* Stage 5 - announce */
2782 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2794 xfrm_states_put(x_cur
, nx_cur
);
2796 xfrm_states_delete(x_new
, nx_new
);
2800 EXPORT_SYMBOL(xfrm_migrate
);