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 struct list_head xfrm_policy_all
;
50 unsigned int xfrm_policy_count
[XFRM_POLICY_MAX
*2];
51 EXPORT_SYMBOL(xfrm_policy_count
);
53 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock
);
54 static struct xfrm_policy_afinfo
*xfrm_policy_afinfo
[NPROTO
];
56 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
58 static struct work_struct xfrm_policy_gc_work
;
59 static HLIST_HEAD(xfrm_policy_gc_list
);
60 static DEFINE_SPINLOCK(xfrm_policy_gc_lock
);
62 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
);
63 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
);
64 static void xfrm_init_pmtu(struct dst_entry
*dst
);
67 __xfrm4_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
69 return addr_match(&fl
->fl4_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
70 addr_match(&fl
->fl4_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
71 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
72 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
73 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
74 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
78 __xfrm6_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
)
80 return addr_match(&fl
->fl6_dst
, &sel
->daddr
, sel
->prefixlen_d
) &&
81 addr_match(&fl
->fl6_src
, &sel
->saddr
, sel
->prefixlen_s
) &&
82 !((xfrm_flowi_dport(fl
) ^ sel
->dport
) & sel
->dport_mask
) &&
83 !((xfrm_flowi_sport(fl
) ^ sel
->sport
) & sel
->sport_mask
) &&
84 (fl
->proto
== sel
->proto
|| !sel
->proto
) &&
85 (fl
->oif
== sel
->ifindex
|| !sel
->ifindex
);
88 int xfrm_selector_match(struct xfrm_selector
*sel
, struct flowi
*fl
,
89 unsigned short family
)
93 return __xfrm4_selector_match(sel
, fl
);
95 return __xfrm6_selector_match(sel
, fl
);
100 static inline struct dst_entry
*__xfrm_dst_lookup(int tos
,
101 xfrm_address_t
*saddr
,
102 xfrm_address_t
*daddr
,
105 struct xfrm_policy_afinfo
*afinfo
;
106 struct dst_entry
*dst
;
108 afinfo
= xfrm_policy_get_afinfo(family
);
109 if (unlikely(afinfo
== NULL
))
110 return ERR_PTR(-EAFNOSUPPORT
);
112 dst
= afinfo
->dst_lookup(tos
, saddr
, daddr
);
114 xfrm_policy_put_afinfo(afinfo
);
119 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
, int tos
,
120 xfrm_address_t
*prev_saddr
,
121 xfrm_address_t
*prev_daddr
,
124 xfrm_address_t
*saddr
= &x
->props
.saddr
;
125 xfrm_address_t
*daddr
= &x
->id
.daddr
;
126 struct dst_entry
*dst
;
128 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
132 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
137 dst
= __xfrm_dst_lookup(tos
, saddr
, daddr
, family
);
140 if (prev_saddr
!= saddr
)
141 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
142 if (prev_daddr
!= daddr
)
143 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
149 static inline unsigned long make_jiffies(long secs
)
151 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
152 return MAX_SCHEDULE_TIMEOUT
-1;
157 static void xfrm_policy_timer(unsigned long data
)
159 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
160 unsigned long now
= get_seconds();
161 long next
= LONG_MAX
;
165 read_lock(&xp
->lock
);
170 dir
= xfrm_policy_id2dir(xp
->index
);
172 if (xp
->lft
.hard_add_expires_seconds
) {
173 long tmo
= xp
->lft
.hard_add_expires_seconds
+
174 xp
->curlft
.add_time
- now
;
180 if (xp
->lft
.hard_use_expires_seconds
) {
181 long tmo
= xp
->lft
.hard_use_expires_seconds
+
182 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
188 if (xp
->lft
.soft_add_expires_seconds
) {
189 long tmo
= xp
->lft
.soft_add_expires_seconds
+
190 xp
->curlft
.add_time
- now
;
193 tmo
= XFRM_KM_TIMEOUT
;
198 if (xp
->lft
.soft_use_expires_seconds
) {
199 long tmo
= xp
->lft
.soft_use_expires_seconds
+
200 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
203 tmo
= XFRM_KM_TIMEOUT
;
210 km_policy_expired(xp
, dir
, 0, 0);
211 if (next
!= LONG_MAX
&&
212 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
216 read_unlock(&xp
->lock
);
221 read_unlock(&xp
->lock
);
222 if (!xfrm_policy_delete(xp
, dir
))
223 km_policy_expired(xp
, dir
, 1, 0);
228 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
232 struct xfrm_policy
*xfrm_policy_alloc(gfp_t gfp
)
234 struct xfrm_policy
*policy
;
236 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
239 INIT_LIST_HEAD(&policy
->walk
.all
);
240 INIT_HLIST_NODE(&policy
->bydst
);
241 INIT_HLIST_NODE(&policy
->byidx
);
242 rwlock_init(&policy
->lock
);
243 atomic_set(&policy
->refcnt
, 1);
244 setup_timer(&policy
->timer
, xfrm_policy_timer
,
245 (unsigned long)policy
);
249 EXPORT_SYMBOL(xfrm_policy_alloc
);
251 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
253 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
255 BUG_ON(!policy
->walk
.dead
);
257 BUG_ON(policy
->bundles
);
259 if (del_timer(&policy
->timer
))
262 security_xfrm_policy_free(policy
->security
);
265 EXPORT_SYMBOL(xfrm_policy_destroy
);
267 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
269 struct dst_entry
*dst
;
271 while ((dst
= policy
->bundles
) != NULL
) {
272 policy
->bundles
= dst
->next
;
276 if (del_timer(&policy
->timer
))
277 atomic_dec(&policy
->refcnt
);
279 if (atomic_read(&policy
->refcnt
) > 1)
282 xfrm_pol_put(policy
);
285 static void xfrm_policy_gc_task(struct work_struct
*work
)
287 struct xfrm_policy
*policy
;
288 struct hlist_node
*entry
, *tmp
;
289 struct hlist_head gc_list
;
291 spin_lock_bh(&xfrm_policy_gc_lock
);
292 gc_list
.first
= xfrm_policy_gc_list
.first
;
293 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
294 spin_unlock_bh(&xfrm_policy_gc_lock
);
296 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
297 xfrm_policy_gc_kill(policy
);
300 /* Rule must be locked. Release descentant resources, announce
301 * entry dead. The rule must be unlinked from lists to the moment.
304 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
308 write_lock_bh(&policy
->lock
);
309 dead
= policy
->walk
.dead
;
310 policy
->walk
.dead
= 1;
311 write_unlock_bh(&policy
->lock
);
313 if (unlikely(dead
)) {
318 spin_lock(&xfrm_policy_gc_lock
);
319 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
320 spin_unlock(&xfrm_policy_gc_lock
);
322 schedule_work(&xfrm_policy_gc_work
);
325 struct xfrm_policy_hash
{
326 struct hlist_head
*table
;
330 static struct hlist_head xfrm_policy_inexact
[XFRM_POLICY_MAX
*2];
331 static struct xfrm_policy_hash xfrm_policy_bydst
[XFRM_POLICY_MAX
*2] __read_mostly
;
332 static struct hlist_head
*xfrm_policy_byidx __read_mostly
;
333 static unsigned int xfrm_idx_hmask __read_mostly
;
334 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
336 static inline unsigned int idx_hash(u32 index
)
338 return __idx_hash(index
, xfrm_idx_hmask
);
341 static struct hlist_head
*policy_hash_bysel(struct xfrm_selector
*sel
, unsigned short family
, int dir
)
343 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
344 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
346 return (hash
== hmask
+ 1 ?
347 &xfrm_policy_inexact
[dir
] :
348 xfrm_policy_bydst
[dir
].table
+ hash
);
351 static struct hlist_head
*policy_hash_direct(xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
353 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
354 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
356 return xfrm_policy_bydst
[dir
].table
+ hash
;
359 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
360 struct hlist_head
*ndsttable
,
361 unsigned int nhashmask
)
363 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
364 struct xfrm_policy
*pol
;
368 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
371 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
372 pol
->family
, nhashmask
);
375 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
381 hlist_add_after(entry0
, &pol
->bydst
);
385 if (!hlist_empty(list
)) {
391 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
392 struct hlist_head
*nidxtable
,
393 unsigned int nhashmask
)
395 struct hlist_node
*entry
, *tmp
;
396 struct xfrm_policy
*pol
;
398 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
401 h
= __idx_hash(pol
->index
, nhashmask
);
402 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
406 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
408 return ((old_hmask
+ 1) << 1) - 1;
411 static void xfrm_bydst_resize(int dir
)
413 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
414 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
415 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
416 struct hlist_head
*odst
= xfrm_policy_bydst
[dir
].table
;
417 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
423 write_lock_bh(&xfrm_policy_lock
);
425 for (i
= hmask
; i
>= 0; i
--)
426 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
428 xfrm_policy_bydst
[dir
].table
= ndst
;
429 xfrm_policy_bydst
[dir
].hmask
= nhashmask
;
431 write_unlock_bh(&xfrm_policy_lock
);
433 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
436 static void xfrm_byidx_resize(int total
)
438 unsigned int hmask
= xfrm_idx_hmask
;
439 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
440 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
441 struct hlist_head
*oidx
= xfrm_policy_byidx
;
442 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
448 write_lock_bh(&xfrm_policy_lock
);
450 for (i
= hmask
; i
>= 0; i
--)
451 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
453 xfrm_policy_byidx
= nidx
;
454 xfrm_idx_hmask
= nhashmask
;
456 write_unlock_bh(&xfrm_policy_lock
);
458 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
461 static inline int xfrm_bydst_should_resize(int dir
, int *total
)
463 unsigned int cnt
= xfrm_policy_count
[dir
];
464 unsigned int hmask
= xfrm_policy_bydst
[dir
].hmask
;
469 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
476 static inline int xfrm_byidx_should_resize(int total
)
478 unsigned int hmask
= xfrm_idx_hmask
;
480 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
487 void xfrm_spd_getinfo(struct xfrmk_spdinfo
*si
)
489 read_lock_bh(&xfrm_policy_lock
);
490 si
->incnt
= xfrm_policy_count
[XFRM_POLICY_IN
];
491 si
->outcnt
= xfrm_policy_count
[XFRM_POLICY_OUT
];
492 si
->fwdcnt
= xfrm_policy_count
[XFRM_POLICY_FWD
];
493 si
->inscnt
= xfrm_policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
494 si
->outscnt
= xfrm_policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
495 si
->fwdscnt
= xfrm_policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
496 si
->spdhcnt
= xfrm_idx_hmask
;
497 si
->spdhmcnt
= xfrm_policy_hashmax
;
498 read_unlock_bh(&xfrm_policy_lock
);
500 EXPORT_SYMBOL(xfrm_spd_getinfo
);
502 static DEFINE_MUTEX(hash_resize_mutex
);
503 static void xfrm_hash_resize(struct work_struct
*__unused
)
507 mutex_lock(&hash_resize_mutex
);
510 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
511 if (xfrm_bydst_should_resize(dir
, &total
))
512 xfrm_bydst_resize(dir
);
514 if (xfrm_byidx_should_resize(total
))
515 xfrm_byidx_resize(total
);
517 mutex_unlock(&hash_resize_mutex
);
520 static DECLARE_WORK(xfrm_hash_work
, xfrm_hash_resize
);
522 /* Generate new index... KAME seems to generate them ordered by cost
523 * of an absolute inpredictability of ordering of rules. This will not pass. */
524 static u32
xfrm_gen_index(u8 type
, int dir
)
526 static u32 idx_generator
;
529 struct hlist_node
*entry
;
530 struct hlist_head
*list
;
531 struct xfrm_policy
*p
;
535 idx
= (idx_generator
| dir
);
539 list
= xfrm_policy_byidx
+ idx_hash(idx
);
541 hlist_for_each_entry(p
, entry
, list
, byidx
) {
542 if (p
->index
== idx
) {
552 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
554 u32
*p1
= (u32
*) s1
;
555 u32
*p2
= (u32
*) s2
;
556 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
559 for (i
= 0; i
< len
; i
++) {
567 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
569 struct xfrm_policy
*pol
;
570 struct xfrm_policy
*delpol
;
571 struct hlist_head
*chain
;
572 struct hlist_node
*entry
, *newpos
;
573 struct dst_entry
*gc_list
;
575 write_lock_bh(&xfrm_policy_lock
);
576 chain
= policy_hash_bysel(&policy
->selector
, policy
->family
, dir
);
579 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
580 if (pol
->type
== policy
->type
&&
581 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
582 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
585 write_unlock_bh(&xfrm_policy_lock
);
589 if (policy
->priority
> pol
->priority
)
591 } else if (policy
->priority
>= pol
->priority
) {
592 newpos
= &pol
->bydst
;
599 hlist_add_after(newpos
, &policy
->bydst
);
601 hlist_add_head(&policy
->bydst
, chain
);
602 xfrm_pol_hold(policy
);
603 xfrm_policy_count
[dir
]++;
604 atomic_inc(&flow_cache_genid
);
606 hlist_del(&delpol
->bydst
);
607 hlist_del(&delpol
->byidx
);
608 list_del(&delpol
->walk
.all
);
609 xfrm_policy_count
[dir
]--;
611 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(policy
->type
, dir
);
612 hlist_add_head(&policy
->byidx
, xfrm_policy_byidx
+idx_hash(policy
->index
));
613 policy
->curlft
.add_time
= get_seconds();
614 policy
->curlft
.use_time
= 0;
615 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
616 xfrm_pol_hold(policy
);
617 list_add(&policy
->walk
.all
, &xfrm_policy_all
);
618 write_unlock_bh(&xfrm_policy_lock
);
621 xfrm_policy_kill(delpol
);
622 else if (xfrm_bydst_should_resize(dir
, NULL
))
623 schedule_work(&xfrm_hash_work
);
625 read_lock_bh(&xfrm_policy_lock
);
627 entry
= &policy
->bydst
;
628 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
629 struct dst_entry
*dst
;
631 write_lock(&policy
->lock
);
632 dst
= policy
->bundles
;
634 struct dst_entry
*tail
= dst
;
637 tail
->next
= gc_list
;
640 policy
->bundles
= NULL
;
642 write_unlock(&policy
->lock
);
644 read_unlock_bh(&xfrm_policy_lock
);
647 struct dst_entry
*dst
= gc_list
;
655 EXPORT_SYMBOL(xfrm_policy_insert
);
657 struct xfrm_policy
*xfrm_policy_bysel_ctx(u8 type
, int dir
,
658 struct xfrm_selector
*sel
,
659 struct xfrm_sec_ctx
*ctx
, int delete,
662 struct xfrm_policy
*pol
, *ret
;
663 struct hlist_head
*chain
;
664 struct hlist_node
*entry
;
667 write_lock_bh(&xfrm_policy_lock
);
668 chain
= policy_hash_bysel(sel
, sel
->family
, dir
);
670 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
671 if (pol
->type
== type
&&
672 !selector_cmp(sel
, &pol
->selector
) &&
673 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
676 *err
= security_xfrm_policy_delete(
679 write_unlock_bh(&xfrm_policy_lock
);
682 hlist_del(&pol
->bydst
);
683 hlist_del(&pol
->byidx
);
684 list_del(&pol
->walk
.all
);
685 xfrm_policy_count
[dir
]--;
691 write_unlock_bh(&xfrm_policy_lock
);
694 atomic_inc(&flow_cache_genid
);
695 xfrm_policy_kill(ret
);
699 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
701 struct xfrm_policy
*xfrm_policy_byid(u8 type
, int dir
, u32 id
, int delete,
704 struct xfrm_policy
*pol
, *ret
;
705 struct hlist_head
*chain
;
706 struct hlist_node
*entry
;
709 if (xfrm_policy_id2dir(id
) != dir
)
713 write_lock_bh(&xfrm_policy_lock
);
714 chain
= xfrm_policy_byidx
+ idx_hash(id
);
716 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
717 if (pol
->type
== type
&& pol
->index
== id
) {
720 *err
= security_xfrm_policy_delete(
723 write_unlock_bh(&xfrm_policy_lock
);
726 hlist_del(&pol
->bydst
);
727 hlist_del(&pol
->byidx
);
728 list_del(&pol
->walk
.all
);
729 xfrm_policy_count
[dir
]--;
735 write_unlock_bh(&xfrm_policy_lock
);
738 atomic_inc(&flow_cache_genid
);
739 xfrm_policy_kill(ret
);
743 EXPORT_SYMBOL(xfrm_policy_byid
);
745 #ifdef CONFIG_SECURITY_NETWORK_XFRM
747 xfrm_policy_flush_secctx_check(u8 type
, struct xfrm_audit
*audit_info
)
751 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
752 struct xfrm_policy
*pol
;
753 struct hlist_node
*entry
;
756 hlist_for_each_entry(pol
, entry
,
757 &xfrm_policy_inexact
[dir
], bydst
) {
758 if (pol
->type
!= type
)
760 err
= security_xfrm_policy_delete(pol
->security
);
762 xfrm_audit_policy_delete(pol
, 0,
763 audit_info
->loginuid
,
764 audit_info
->sessionid
,
769 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
770 hlist_for_each_entry(pol
, entry
,
771 xfrm_policy_bydst
[dir
].table
+ i
,
773 if (pol
->type
!= type
)
775 err
= security_xfrm_policy_delete(
778 xfrm_audit_policy_delete(pol
, 0,
779 audit_info
->loginuid
,
780 audit_info
->sessionid
,
791 xfrm_policy_flush_secctx_check(u8 type
, struct xfrm_audit
*audit_info
)
797 int xfrm_policy_flush(u8 type
, struct xfrm_audit
*audit_info
)
801 write_lock_bh(&xfrm_policy_lock
);
803 err
= xfrm_policy_flush_secctx_check(type
, audit_info
);
807 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
808 struct xfrm_policy
*pol
;
809 struct hlist_node
*entry
;
814 hlist_for_each_entry(pol
, entry
,
815 &xfrm_policy_inexact
[dir
], bydst
) {
816 if (pol
->type
!= type
)
818 hlist_del(&pol
->bydst
);
819 hlist_del(&pol
->byidx
);
820 write_unlock_bh(&xfrm_policy_lock
);
822 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
823 audit_info
->sessionid
,
826 xfrm_policy_kill(pol
);
829 write_lock_bh(&xfrm_policy_lock
);
833 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
835 hlist_for_each_entry(pol
, entry
,
836 xfrm_policy_bydst
[dir
].table
+ i
,
838 if (pol
->type
!= type
)
840 hlist_del(&pol
->bydst
);
841 hlist_del(&pol
->byidx
);
842 list_del(&pol
->walk
.all
);
843 write_unlock_bh(&xfrm_policy_lock
);
845 xfrm_audit_policy_delete(pol
, 1,
846 audit_info
->loginuid
,
847 audit_info
->sessionid
,
849 xfrm_policy_kill(pol
);
852 write_lock_bh(&xfrm_policy_lock
);
857 xfrm_policy_count
[dir
] -= killed
;
859 atomic_inc(&flow_cache_genid
);
861 write_unlock_bh(&xfrm_policy_lock
);
864 EXPORT_SYMBOL(xfrm_policy_flush
);
866 int xfrm_policy_walk(struct xfrm_policy_walk
*walk
,
867 int (*func
)(struct xfrm_policy
*, int, int, void*),
870 struct xfrm_policy
*pol
;
871 struct xfrm_policy_walk_entry
*x
;
874 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
875 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
878 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
881 write_lock_bh(&xfrm_policy_lock
);
882 if (list_empty(&walk
->walk
.all
))
883 x
= list_first_entry(&xfrm_policy_all
, struct xfrm_policy_walk_entry
, all
);
885 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
886 list_for_each_entry_from(x
, &xfrm_policy_all
, all
) {
889 pol
= container_of(x
, struct xfrm_policy
, walk
);
890 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
891 walk
->type
!= pol
->type
)
893 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
896 list_move_tail(&walk
->walk
.all
, &x
->all
);
901 if (walk
->seq
== 0) {
905 list_del_init(&walk
->walk
.all
);
907 write_unlock_bh(&xfrm_policy_lock
);
910 EXPORT_SYMBOL(xfrm_policy_walk
);
912 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
914 INIT_LIST_HEAD(&walk
->walk
.all
);
919 EXPORT_SYMBOL(xfrm_policy_walk_init
);
921 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
923 if (list_empty(&walk
->walk
.all
))
926 write_lock_bh(&xfrm_policy_lock
);
927 list_del(&walk
->walk
.all
);
928 write_unlock_bh(&xfrm_policy_lock
);
930 EXPORT_SYMBOL(xfrm_policy_walk_done
);
933 * Find policy to apply to this flow.
935 * Returns 0 if policy found, else an -errno.
937 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
938 u8 type
, u16 family
, int dir
)
940 struct xfrm_selector
*sel
= &pol
->selector
;
941 int match
, ret
= -ESRCH
;
943 if (pol
->family
!= family
||
947 match
= xfrm_selector_match(sel
, fl
, family
);
949 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
955 static struct xfrm_policy
*xfrm_policy_lookup_bytype(u8 type
, struct flowi
*fl
,
959 struct xfrm_policy
*pol
, *ret
;
960 xfrm_address_t
*daddr
, *saddr
;
961 struct hlist_node
*entry
;
962 struct hlist_head
*chain
;
965 daddr
= xfrm_flowi_daddr(fl
, family
);
966 saddr
= xfrm_flowi_saddr(fl
, family
);
967 if (unlikely(!daddr
|| !saddr
))
970 read_lock_bh(&xfrm_policy_lock
);
971 chain
= policy_hash_direct(daddr
, saddr
, family
, dir
);
973 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
974 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
984 priority
= ret
->priority
;
988 chain
= &xfrm_policy_inexact
[dir
];
989 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
990 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
998 } else if (pol
->priority
< priority
) {
1006 read_unlock_bh(&xfrm_policy_lock
);
1011 static int xfrm_policy_lookup(struct flowi
*fl
, u16 family
, u8 dir
,
1012 void **objp
, atomic_t
**obj_refp
)
1014 struct xfrm_policy
*pol
;
1017 #ifdef CONFIG_XFRM_SUB_POLICY
1018 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1026 pol
= xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1031 #ifdef CONFIG_XFRM_SUB_POLICY
1034 if ((*objp
= (void *) pol
) != NULL
)
1035 *obj_refp
= &pol
->refcnt
;
1039 static inline int policy_to_flow_dir(int dir
)
1041 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1042 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1043 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1047 case XFRM_POLICY_IN
:
1049 case XFRM_POLICY_OUT
:
1050 return FLOW_DIR_OUT
;
1051 case XFRM_POLICY_FWD
:
1052 return FLOW_DIR_FWD
;
1056 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1058 struct xfrm_policy
*pol
;
1060 read_lock_bh(&xfrm_policy_lock
);
1061 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1062 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1067 err
= security_xfrm_policy_lookup(pol
->security
,
1069 policy_to_flow_dir(dir
));
1072 else if (err
== -ESRCH
)
1079 read_unlock_bh(&xfrm_policy_lock
);
1083 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1085 struct hlist_head
*chain
= policy_hash_bysel(&pol
->selector
,
1088 list_add(&pol
->walk
.all
, &xfrm_policy_all
);
1089 hlist_add_head(&pol
->bydst
, chain
);
1090 hlist_add_head(&pol
->byidx
, xfrm_policy_byidx
+idx_hash(pol
->index
));
1091 xfrm_policy_count
[dir
]++;
1094 if (xfrm_bydst_should_resize(dir
, NULL
))
1095 schedule_work(&xfrm_hash_work
);
1098 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1101 if (hlist_unhashed(&pol
->bydst
))
1104 hlist_del(&pol
->bydst
);
1105 hlist_del(&pol
->byidx
);
1106 list_del(&pol
->walk
.all
);
1107 xfrm_policy_count
[dir
]--;
1112 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1114 write_lock_bh(&xfrm_policy_lock
);
1115 pol
= __xfrm_policy_unlink(pol
, dir
);
1116 write_unlock_bh(&xfrm_policy_lock
);
1118 if (dir
< XFRM_POLICY_MAX
)
1119 atomic_inc(&flow_cache_genid
);
1120 xfrm_policy_kill(pol
);
1125 EXPORT_SYMBOL(xfrm_policy_delete
);
1127 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1129 struct xfrm_policy
*old_pol
;
1131 #ifdef CONFIG_XFRM_SUB_POLICY
1132 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1136 write_lock_bh(&xfrm_policy_lock
);
1137 old_pol
= sk
->sk_policy
[dir
];
1138 sk
->sk_policy
[dir
] = pol
;
1140 pol
->curlft
.add_time
= get_seconds();
1141 pol
->index
= xfrm_gen_index(pol
->type
, XFRM_POLICY_MAX
+dir
);
1142 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1145 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1146 write_unlock_bh(&xfrm_policy_lock
);
1149 xfrm_policy_kill(old_pol
);
1154 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1156 struct xfrm_policy
*newp
= xfrm_policy_alloc(GFP_ATOMIC
);
1159 newp
->selector
= old
->selector
;
1160 if (security_xfrm_policy_clone(old
->security
,
1163 return NULL
; /* ENOMEM */
1165 newp
->lft
= old
->lft
;
1166 newp
->curlft
= old
->curlft
;
1167 newp
->action
= old
->action
;
1168 newp
->flags
= old
->flags
;
1169 newp
->xfrm_nr
= old
->xfrm_nr
;
1170 newp
->index
= old
->index
;
1171 newp
->type
= old
->type
;
1172 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1173 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1174 write_lock_bh(&xfrm_policy_lock
);
1175 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1176 write_unlock_bh(&xfrm_policy_lock
);
1182 int __xfrm_sk_clone_policy(struct sock
*sk
)
1184 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1185 *p1
= sk
->sk_policy
[1];
1187 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1188 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1190 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1196 xfrm_get_saddr(xfrm_address_t
*local
, xfrm_address_t
*remote
,
1197 unsigned short family
)
1200 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1202 if (unlikely(afinfo
== NULL
))
1204 err
= afinfo
->get_saddr(local
, remote
);
1205 xfrm_policy_put_afinfo(afinfo
);
1209 /* Resolve list of templates for the flow, given policy. */
1212 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1213 struct xfrm_state
**xfrm
,
1214 unsigned short family
)
1218 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1219 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1222 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1223 struct xfrm_state
*x
;
1224 xfrm_address_t
*remote
= daddr
;
1225 xfrm_address_t
*local
= saddr
;
1226 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1228 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1229 tmpl
->mode
== XFRM_MODE_BEET
) {
1230 remote
= &tmpl
->id
.daddr
;
1231 local
= &tmpl
->saddr
;
1232 family
= tmpl
->encap_family
;
1233 if (xfrm_addr_any(local
, family
)) {
1234 error
= xfrm_get_saddr(&tmp
, remote
, family
);
1241 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1243 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1250 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1255 if (!tmpl
->optional
)
1261 for (nx
--; nx
>=0; nx
--)
1262 xfrm_state_put(xfrm
[nx
]);
1267 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1268 struct xfrm_state
**xfrm
,
1269 unsigned short family
)
1271 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1272 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1278 for (i
= 0; i
< npols
; i
++) {
1279 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1284 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1292 /* found states are sorted for outbound processing */
1294 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1299 for (cnx
--; cnx
>=0; cnx
--)
1300 xfrm_state_put(tpp
[cnx
]);
1305 /* Check that the bundle accepts the flow and its components are
1309 static struct dst_entry
*
1310 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1312 struct dst_entry
*x
;
1313 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1314 if (unlikely(afinfo
== NULL
))
1315 return ERR_PTR(-EINVAL
);
1316 x
= afinfo
->find_bundle(fl
, policy
);
1317 xfrm_policy_put_afinfo(afinfo
);
1321 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1323 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1329 tos
= afinfo
->get_tos(fl
);
1331 xfrm_policy_put_afinfo(afinfo
);
1336 static inline struct xfrm_dst
*xfrm_alloc_dst(int family
)
1338 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1339 struct xfrm_dst
*xdst
;
1342 return ERR_PTR(-EINVAL
);
1344 xdst
= dst_alloc(afinfo
->dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1346 xfrm_policy_put_afinfo(afinfo
);
1351 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1354 struct xfrm_policy_afinfo
*afinfo
=
1355 xfrm_policy_get_afinfo(dst
->ops
->family
);
1361 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1363 xfrm_policy_put_afinfo(afinfo
);
1368 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1370 struct xfrm_policy_afinfo
*afinfo
=
1371 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1377 err
= afinfo
->fill_dst(xdst
, dev
);
1379 xfrm_policy_put_afinfo(afinfo
);
1384 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1385 * all the metrics... Shortly, bundle a bundle.
1388 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1389 struct xfrm_state
**xfrm
, int nx
,
1391 struct dst_entry
*dst
)
1393 unsigned long now
= jiffies
;
1394 struct net_device
*dev
;
1395 struct dst_entry
*dst_prev
= NULL
;
1396 struct dst_entry
*dst0
= NULL
;
1400 int nfheader_len
= 0;
1401 int trailer_len
= 0;
1403 int family
= policy
->selector
.family
;
1404 xfrm_address_t saddr
, daddr
;
1406 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1408 tos
= xfrm_get_tos(fl
, family
);
1415 for (; i
< nx
; i
++) {
1416 struct xfrm_dst
*xdst
= xfrm_alloc_dst(family
);
1417 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1419 err
= PTR_ERR(xdst
);
1428 dst_prev
->child
= dst_clone(dst1
);
1429 dst1
->flags
|= DST_NOHASH
;
1433 memcpy(&dst1
->metrics
, &dst
->metrics
, sizeof(dst
->metrics
));
1435 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1436 family
= xfrm
[i
]->props
.family
;
1437 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1445 dst1
->xfrm
= xfrm
[i
];
1446 xdst
->genid
= xfrm
[i
]->genid
;
1448 dst1
->obsolete
= -1;
1449 dst1
->flags
|= DST_HOST
;
1450 dst1
->lastuse
= now
;
1452 dst1
->input
= dst_discard
;
1453 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1455 dst1
->next
= dst_prev
;
1458 header_len
+= xfrm
[i
]->props
.header_len
;
1459 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1460 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1461 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1464 dst_prev
->child
= dst
;
1472 /* Copy neighbout for reachability confirmation */
1473 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1475 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1476 xfrm_init_pmtu(dst_prev
);
1478 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1479 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1481 err
= xfrm_fill_dst(xdst
, dev
);
1485 dst_prev
->header_len
= header_len
;
1486 dst_prev
->trailer_len
= trailer_len
;
1487 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1488 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1496 xfrm_state_put(xfrm
[i
]);
1500 dst0
= ERR_PTR(err
);
1505 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1508 *target
= kmalloc(size
, GFP_ATOMIC
);
1512 memcpy(*target
, src
, size
);
1517 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1519 #ifdef CONFIG_XFRM_SUB_POLICY
1520 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1521 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1529 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1531 #ifdef CONFIG_XFRM_SUB_POLICY
1532 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1533 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1539 static int stale_bundle(struct dst_entry
*dst
);
1541 /* Main function: finds/creates a bundle for given flow.
1543 * At the moment we eat a raw IP route. Mostly to speed up lookups
1544 * on interfaces with disabled IPsec.
1546 int __xfrm_lookup(struct dst_entry
**dst_p
, struct flowi
*fl
,
1547 struct sock
*sk
, int flags
)
1549 struct xfrm_policy
*policy
;
1550 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1555 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1556 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1561 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1564 genid
= atomic_read(&flow_cache_genid
);
1566 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1572 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1573 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1574 err
= PTR_ERR(policy
);
1575 if (IS_ERR(policy
)) {
1576 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1582 /* To accelerate a bit... */
1583 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1584 !xfrm_policy_count
[XFRM_POLICY_OUT
])
1587 policy
= flow_cache_lookup(fl
, dst_orig
->ops
->family
,
1588 dir
, xfrm_policy_lookup
);
1589 err
= PTR_ERR(policy
);
1590 if (IS_ERR(policy
)) {
1591 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1599 family
= dst_orig
->ops
->family
;
1602 xfrm_nr
+= pols
[0]->xfrm_nr
;
1605 if ((flags
& XFRM_LOOKUP_ICMP
) && !(policy
->flags
& XFRM_POLICY_ICMP
))
1608 policy
->curlft
.use_time
= get_seconds();
1610 switch (policy
->action
) {
1612 case XFRM_POLICY_BLOCK
:
1613 /* Prohibit the flow */
1614 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1618 case XFRM_POLICY_ALLOW
:
1619 #ifndef CONFIG_XFRM_SUB_POLICY
1620 if (policy
->xfrm_nr
== 0) {
1621 /* Flow passes not transformed. */
1622 xfrm_pol_put(policy
);
1627 /* Try to find matching bundle.
1629 * LATER: help from flow cache. It is optional, this
1630 * is required only for output policy.
1632 dst
= xfrm_find_bundle(fl
, policy
, family
);
1634 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1642 #ifdef CONFIG_XFRM_SUB_POLICY
1643 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1644 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1648 if (IS_ERR(pols
[1])) {
1649 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1650 err
= PTR_ERR(pols
[1]);
1653 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1654 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1659 xfrm_nr
+= pols
[1]->xfrm_nr
;
1664 * Because neither flowi nor bundle information knows about
1665 * transformation template size. On more than one policy usage
1666 * we can realize whether all of them is bypass or not after
1667 * they are searched. See above not-transformed bypass
1668 * is surrounded by non-sub policy configuration, too.
1671 /* Flow passes not transformed. */
1672 xfrm_pols_put(pols
, npols
);
1677 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1679 if (unlikely(nx
<0)) {
1681 if (err
== -EAGAIN
&& sysctl_xfrm_larval_drop
) {
1682 /* EREMOTE tells the caller to generate
1683 * a one-shot blackhole route.
1685 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1686 xfrm_pol_put(policy
);
1689 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1690 DECLARE_WAITQUEUE(wait
, current
);
1692 add_wait_queue(&km_waitq
, &wait
);
1693 set_current_state(TASK_INTERRUPTIBLE
);
1695 set_current_state(TASK_RUNNING
);
1696 remove_wait_queue(&km_waitq
, &wait
);
1698 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1700 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1701 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1705 if (nx
== -EAGAIN
||
1706 genid
!= atomic_read(&flow_cache_genid
)) {
1707 xfrm_pols_put(pols
, npols
);
1713 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1718 /* Flow passes not transformed. */
1719 xfrm_pols_put(pols
, npols
);
1723 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1726 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1730 for (pi
= 0; pi
< npols
; pi
++) {
1731 read_lock_bh(&pols
[pi
]->lock
);
1732 pol_dead
|= pols
[pi
]->walk
.dead
;
1733 read_unlock_bh(&pols
[pi
]->lock
);
1736 write_lock_bh(&policy
->lock
);
1737 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1738 /* Wow! While we worked on resolving, this
1739 * policy has gone. Retry. It is not paranoia,
1740 * we just cannot enlist new bundle to dead object.
1741 * We can't enlist stable bundles either.
1743 write_unlock_bh(&policy
->lock
);
1747 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD
);
1749 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1750 err
= -EHOSTUNREACH
;
1755 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1757 err
= xfrm_dst_update_origin(dst
, fl
);
1758 if (unlikely(err
)) {
1759 write_unlock_bh(&policy
->lock
);
1761 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1765 dst
->next
= policy
->bundles
;
1766 policy
->bundles
= dst
;
1768 write_unlock_bh(&policy
->lock
);
1771 dst_release(dst_orig
);
1772 xfrm_pols_put(pols
, npols
);
1776 xfrm_pols_put(pols
, npols
);
1778 dst_release(dst_orig
);
1784 if (flags
& XFRM_LOOKUP_ICMP
)
1788 EXPORT_SYMBOL(__xfrm_lookup
);
1790 int xfrm_lookup(struct dst_entry
**dst_p
, struct flowi
*fl
,
1791 struct sock
*sk
, int flags
)
1793 int err
= __xfrm_lookup(dst_p
, fl
, sk
, flags
);
1795 if (err
== -EREMOTE
) {
1796 dst_release(*dst_p
);
1803 EXPORT_SYMBOL(xfrm_lookup
);
1806 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1808 struct xfrm_state
*x
;
1810 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1812 x
= skb
->sp
->xvec
[idx
];
1813 if (!x
->type
->reject
)
1815 return x
->type
->reject(x
, skb
, fl
);
1818 /* When skb is transformed back to its "native" form, we have to
1819 * check policy restrictions. At the moment we make this in maximally
1820 * stupid way. Shame on me. :-) Of course, connected sockets must
1821 * have policy cached at them.
1825 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1826 unsigned short family
)
1828 if (xfrm_state_kern(x
))
1829 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1830 return x
->id
.proto
== tmpl
->id
.proto
&&
1831 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1832 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1833 x
->props
.mode
== tmpl
->mode
&&
1834 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1835 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1836 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1837 xfrm_state_addr_cmp(tmpl
, x
, family
));
1841 * 0 or more than 0 is returned when validation is succeeded (either bypass
1842 * because of optional transport mode, or next index of the mathced secpath
1843 * state with the template.
1844 * -1 is returned when no matching template is found.
1845 * Otherwise "-2 - errored_index" is returned.
1848 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1849 unsigned short family
)
1853 if (tmpl
->optional
) {
1854 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1858 for (; idx
< sp
->len
; idx
++) {
1859 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1861 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1870 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1871 unsigned int family
, int reverse
)
1873 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1876 if (unlikely(afinfo
== NULL
))
1877 return -EAFNOSUPPORT
;
1879 afinfo
->decode_session(skb
, fl
, reverse
);
1880 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1881 xfrm_policy_put_afinfo(afinfo
);
1884 EXPORT_SYMBOL(__xfrm_decode_session
);
1886 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1888 for (; k
< sp
->len
; k
++) {
1889 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1898 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1899 unsigned short family
)
1901 struct xfrm_policy
*pol
;
1902 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1911 reverse
= dir
& ~XFRM_POLICY_MASK
;
1912 dir
&= XFRM_POLICY_MASK
;
1913 fl_dir
= policy_to_flow_dir(dir
);
1915 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1916 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
1920 nf_nat_decode_session(skb
, &fl
, family
);
1922 /* First, check used SA against their selectors. */
1926 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1927 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1928 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1929 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH
);
1936 if (sk
&& sk
->sk_policy
[dir
]) {
1937 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1939 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1945 pol
= flow_cache_lookup(&fl
, family
, fl_dir
,
1946 xfrm_policy_lookup
);
1949 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1954 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1955 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1956 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS
);
1962 pol
->curlft
.use_time
= get_seconds();
1966 #ifdef CONFIG_XFRM_SUB_POLICY
1967 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1968 pols
[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN
,
1972 if (IS_ERR(pols
[1])) {
1973 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1976 pols
[1]->curlft
.use_time
= get_seconds();
1982 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1983 struct sec_path
*sp
;
1984 static struct sec_path dummy
;
1985 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1986 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1987 struct xfrm_tmpl
**tpp
= tp
;
1991 if ((sp
= skb
->sp
) == NULL
)
1994 for (pi
= 0; pi
< npols
; pi
++) {
1995 if (pols
[pi
] != pol
&&
1996 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
1997 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
2000 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2001 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR
);
2004 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2005 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2009 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2013 /* For each tunnel xfrm, find the first matching tmpl.
2014 * For each tmpl before that, find corresponding xfrm.
2015 * Order is _important_. Later we will implement
2016 * some barriers, but at the moment barriers
2017 * are implied between each two transformations.
2019 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2020 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2023 /* "-2 - errored_index" returned */
2025 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2030 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2031 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2035 xfrm_pols_put(pols
, npols
);
2038 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
2041 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2043 xfrm_pols_put(pols
, npols
);
2046 EXPORT_SYMBOL(__xfrm_policy_check
);
2048 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2052 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2053 /* XXX: we should have something like FWDHDRERROR here. */
2054 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
2058 return xfrm_lookup(&skb
->dst
, &fl
, NULL
, 0) == 0;
2060 EXPORT_SYMBOL(__xfrm_route_forward
);
2062 /* Optimize later using cookies and generation ids. */
2064 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2066 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2067 * to "-1" to force all XFRM destinations to get validated by
2068 * dst_ops->check on every use. We do this because when a
2069 * normal route referenced by an XFRM dst is obsoleted we do
2070 * not go looking around for all parent referencing XFRM dsts
2071 * so that we can invalidate them. It is just too much work.
2072 * Instead we make the checks here on every use. For example:
2074 * XFRM dst A --> IPv4 dst X
2076 * X is the "xdst->route" of A (X is also the "dst->path" of A
2077 * in this example). If X is marked obsolete, "A" will not
2078 * notice. That's what we are validating here via the
2079 * stale_bundle() check.
2081 * When a policy's bundle is pruned, we dst_free() the XFRM
2082 * dst which causes it's ->obsolete field to be set to a
2083 * positive non-zero integer. If an XFRM dst has been pruned
2084 * like this, we want to force a new route lookup.
2086 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2092 static int stale_bundle(struct dst_entry
*dst
)
2094 return !xfrm_bundle_ok(NULL
, (struct xfrm_dst
*)dst
, NULL
, AF_UNSPEC
, 0);
2097 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2099 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2100 dst
->dev
= dev_net(dev
)->loopback_dev
;
2105 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2107 static void xfrm_link_failure(struct sk_buff
*skb
)
2109 /* Impossible. Such dst must be popped before reaches point of failure. */
2113 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2116 if (dst
->obsolete
) {
2124 static void prune_one_bundle(struct xfrm_policy
*pol
, int (*func
)(struct dst_entry
*), struct dst_entry
**gc_list_p
)
2126 struct dst_entry
*dst
, **dstp
;
2128 write_lock(&pol
->lock
);
2129 dstp
= &pol
->bundles
;
2130 while ((dst
=*dstp
) != NULL
) {
2133 dst
->next
= *gc_list_p
;
2139 write_unlock(&pol
->lock
);
2142 static void xfrm_prune_bundles(int (*func
)(struct dst_entry
*))
2144 struct dst_entry
*gc_list
= NULL
;
2147 read_lock_bh(&xfrm_policy_lock
);
2148 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2149 struct xfrm_policy
*pol
;
2150 struct hlist_node
*entry
;
2151 struct hlist_head
*table
;
2154 hlist_for_each_entry(pol
, entry
,
2155 &xfrm_policy_inexact
[dir
], bydst
)
2156 prune_one_bundle(pol
, func
, &gc_list
);
2158 table
= xfrm_policy_bydst
[dir
].table
;
2159 for (i
= xfrm_policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
2160 hlist_for_each_entry(pol
, entry
, table
+ i
, bydst
)
2161 prune_one_bundle(pol
, func
, &gc_list
);
2164 read_unlock_bh(&xfrm_policy_lock
);
2167 struct dst_entry
*dst
= gc_list
;
2168 gc_list
= dst
->next
;
2173 static int unused_bundle(struct dst_entry
*dst
)
2175 return !atomic_read(&dst
->__refcnt
);
2178 static void __xfrm_garbage_collect(void)
2180 xfrm_prune_bundles(unused_bundle
);
2183 static int xfrm_flush_bundles(void)
2185 xfrm_prune_bundles(stale_bundle
);
2189 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2192 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2193 u32 pmtu
, route_mtu_cached
;
2195 pmtu
= dst_mtu(dst
->child
);
2196 xdst
->child_mtu_cached
= pmtu
;
2198 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2200 route_mtu_cached
= dst_mtu(xdst
->route
);
2201 xdst
->route_mtu_cached
= route_mtu_cached
;
2203 if (pmtu
> route_mtu_cached
)
2204 pmtu
= route_mtu_cached
;
2206 dst
->metrics
[RTAX_MTU
-1] = pmtu
;
2207 } while ((dst
= dst
->next
));
2210 /* Check that the bundle accepts the flow and its components are
2214 int xfrm_bundle_ok(struct xfrm_policy
*pol
, struct xfrm_dst
*first
,
2215 struct flowi
*fl
, int family
, int strict
)
2217 struct dst_entry
*dst
= &first
->u
.dst
;
2218 struct xfrm_dst
*last
;
2221 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2222 (dst
->dev
&& !netif_running(dst
->dev
)))
2224 #ifdef CONFIG_XFRM_SUB_POLICY
2226 if (first
->origin
&& !flow_cache_uli_match(first
->origin
, fl
))
2228 if (first
->partner
&&
2229 !xfrm_selector_match(first
->partner
, fl
, family
))
2237 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2239 if (fl
&& !xfrm_selector_match(&dst
->xfrm
->sel
, fl
, family
))
2242 !security_xfrm_state_pol_flow_match(dst
->xfrm
, pol
, fl
))
2244 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2246 if (xdst
->genid
!= dst
->xfrm
->genid
)
2250 !(dst
->xfrm
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) &&
2251 !xfrm_state_addr_flow_check(dst
->xfrm
, fl
, family
))
2254 mtu
= dst_mtu(dst
->child
);
2255 if (xdst
->child_mtu_cached
!= mtu
) {
2257 xdst
->child_mtu_cached
= mtu
;
2260 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2262 mtu
= dst_mtu(xdst
->route
);
2263 if (xdst
->route_mtu_cached
!= mtu
) {
2265 xdst
->route_mtu_cached
= mtu
;
2269 } while (dst
->xfrm
);
2274 mtu
= last
->child_mtu_cached
;
2278 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2279 if (mtu
> last
->route_mtu_cached
)
2280 mtu
= last
->route_mtu_cached
;
2281 dst
->metrics
[RTAX_MTU
-1] = mtu
;
2286 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2287 last
->child_mtu_cached
= mtu
;
2293 EXPORT_SYMBOL(xfrm_bundle_ok
);
2295 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2298 if (unlikely(afinfo
== NULL
))
2300 if (unlikely(afinfo
->family
>= NPROTO
))
2301 return -EAFNOSUPPORT
;
2302 write_lock_bh(&xfrm_policy_afinfo_lock
);
2303 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2306 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2307 if (likely(dst_ops
->kmem_cachep
== NULL
))
2308 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2309 if (likely(dst_ops
->check
== NULL
))
2310 dst_ops
->check
= xfrm_dst_check
;
2311 if (likely(dst_ops
->negative_advice
== NULL
))
2312 dst_ops
->negative_advice
= xfrm_negative_advice
;
2313 if (likely(dst_ops
->link_failure
== NULL
))
2314 dst_ops
->link_failure
= xfrm_link_failure
;
2315 if (likely(afinfo
->garbage_collect
== NULL
))
2316 afinfo
->garbage_collect
= __xfrm_garbage_collect
;
2317 xfrm_policy_afinfo
[afinfo
->family
] = afinfo
;
2319 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2322 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2324 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2327 if (unlikely(afinfo
== NULL
))
2329 if (unlikely(afinfo
->family
>= NPROTO
))
2330 return -EAFNOSUPPORT
;
2331 write_lock_bh(&xfrm_policy_afinfo_lock
);
2332 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2333 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2336 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2337 xfrm_policy_afinfo
[afinfo
->family
] = NULL
;
2338 dst_ops
->kmem_cachep
= NULL
;
2339 dst_ops
->check
= NULL
;
2340 dst_ops
->negative_advice
= NULL
;
2341 dst_ops
->link_failure
= NULL
;
2342 afinfo
->garbage_collect
= NULL
;
2345 write_unlock_bh(&xfrm_policy_afinfo_lock
);
2348 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2350 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
2352 struct xfrm_policy_afinfo
*afinfo
;
2353 if (unlikely(family
>= NPROTO
))
2355 read_lock(&xfrm_policy_afinfo_lock
);
2356 afinfo
= xfrm_policy_afinfo
[family
];
2357 if (unlikely(!afinfo
))
2358 read_unlock(&xfrm_policy_afinfo_lock
);
2362 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2364 read_unlock(&xfrm_policy_afinfo_lock
);
2367 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2369 struct net_device
*dev
= ptr
;
2371 if (!net_eq(dev_net(dev
), &init_net
))
2376 xfrm_flush_bundles();
2381 static struct notifier_block xfrm_dev_notifier
= {
2387 #ifdef CONFIG_XFRM_STATISTICS
2388 static int __init
xfrm_statistics_init(void)
2390 if (snmp_mib_init((void **)xfrm_statistics
,
2391 sizeof(struct linux_xfrm_mib
)) < 0)
2397 static void __init
xfrm_policy_init(void)
2399 unsigned int hmask
, sz
;
2402 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2403 sizeof(struct xfrm_dst
),
2404 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2408 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2410 xfrm_policy_byidx
= xfrm_hash_alloc(sz
);
2411 xfrm_idx_hmask
= hmask
;
2412 if (!xfrm_policy_byidx
)
2413 panic("XFRM: failed to allocate byidx hash\n");
2415 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2416 struct xfrm_policy_hash
*htab
;
2418 INIT_HLIST_HEAD(&xfrm_policy_inexact
[dir
]);
2420 htab
= &xfrm_policy_bydst
[dir
];
2421 htab
->table
= xfrm_hash_alloc(sz
);
2422 htab
->hmask
= hmask
;
2424 panic("XFRM: failed to allocate bydst hash\n");
2427 INIT_LIST_HEAD(&xfrm_policy_all
);
2428 INIT_WORK(&xfrm_policy_gc_work
, xfrm_policy_gc_task
);
2429 register_netdevice_notifier(&xfrm_dev_notifier
);
2432 void __init
xfrm_init(void)
2434 #ifdef CONFIG_XFRM_STATISTICS
2435 xfrm_statistics_init();
2440 #ifdef CONFIG_XFRM_STATISTICS
2445 #ifdef CONFIG_AUDITSYSCALL
2446 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2447 struct audit_buffer
*audit_buf
)
2449 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2450 struct xfrm_selector
*sel
= &xp
->selector
;
2453 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2454 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2456 switch(sel
->family
) {
2458 audit_log_format(audit_buf
, " src=" NIPQUAD_FMT
,
2459 NIPQUAD(sel
->saddr
.a4
));
2460 if (sel
->prefixlen_s
!= 32)
2461 audit_log_format(audit_buf
, " src_prefixlen=%d",
2463 audit_log_format(audit_buf
, " dst=" NIPQUAD_FMT
,
2464 NIPQUAD(sel
->daddr
.a4
));
2465 if (sel
->prefixlen_d
!= 32)
2466 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2470 audit_log_format(audit_buf
, " src=" NIP6_FMT
,
2471 NIP6(*(struct in6_addr
*)sel
->saddr
.a6
));
2472 if (sel
->prefixlen_s
!= 128)
2473 audit_log_format(audit_buf
, " src_prefixlen=%d",
2475 audit_log_format(audit_buf
, " dst=" NIP6_FMT
,
2476 NIP6(*(struct in6_addr
*)sel
->daddr
.a6
));
2477 if (sel
->prefixlen_d
!= 128)
2478 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2484 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2485 uid_t auid
, u32 sessionid
, u32 secid
)
2487 struct audit_buffer
*audit_buf
;
2489 audit_buf
= xfrm_audit_start("SPD-add");
2490 if (audit_buf
== NULL
)
2492 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2493 audit_log_format(audit_buf
, " res=%u", result
);
2494 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2495 audit_log_end(audit_buf
);
2497 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2499 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2500 uid_t auid
, u32 sessionid
, u32 secid
)
2502 struct audit_buffer
*audit_buf
;
2504 audit_buf
= xfrm_audit_start("SPD-delete");
2505 if (audit_buf
== NULL
)
2507 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2508 audit_log_format(audit_buf
, " res=%u", result
);
2509 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2510 audit_log_end(audit_buf
);
2512 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2515 #ifdef CONFIG_XFRM_MIGRATE
2516 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2517 struct xfrm_selector
*sel_tgt
)
2519 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2520 if (sel_tgt
->family
== sel_cmp
->family
&&
2521 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2522 sel_cmp
->family
) == 0 &&
2523 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2524 sel_cmp
->family
) == 0 &&
2525 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2526 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2530 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2537 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2540 struct xfrm_policy
*pol
, *ret
= NULL
;
2541 struct hlist_node
*entry
;
2542 struct hlist_head
*chain
;
2545 read_lock_bh(&xfrm_policy_lock
);
2546 chain
= policy_hash_direct(&sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2547 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2548 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2549 pol
->type
== type
) {
2551 priority
= ret
->priority
;
2555 chain
= &xfrm_policy_inexact
[dir
];
2556 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2557 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2558 pol
->type
== type
&&
2559 pol
->priority
< priority
) {
2568 read_unlock_bh(&xfrm_policy_lock
);
2573 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2577 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2578 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2580 case XFRM_MODE_TUNNEL
:
2581 case XFRM_MODE_BEET
:
2582 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2583 m
->old_family
) == 0 &&
2584 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2585 m
->old_family
) == 0) {
2589 case XFRM_MODE_TRANSPORT
:
2590 /* in case of transport mode, template does not store
2591 any IP addresses, hence we just compare mode and
2602 /* update endpoint address(es) of template(s) */
2603 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2604 struct xfrm_migrate
*m
, int num_migrate
)
2606 struct xfrm_migrate
*mp
;
2607 struct dst_entry
*dst
;
2610 write_lock_bh(&pol
->lock
);
2611 if (unlikely(pol
->walk
.dead
)) {
2612 /* target policy has been deleted */
2613 write_unlock_bh(&pol
->lock
);
2617 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2618 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2619 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2622 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2623 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2625 /* update endpoints */
2626 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2627 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2628 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2629 sizeof(pol
->xfrm_vec
[i
].saddr
));
2630 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2632 while ((dst
= pol
->bundles
) != NULL
) {
2633 pol
->bundles
= dst
->next
;
2639 write_unlock_bh(&pol
->lock
);
2647 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2651 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2654 for (i
= 0; i
< num_migrate
; i
++) {
2655 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2656 m
[i
].old_family
) == 0) &&
2657 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2658 m
[i
].old_family
) == 0))
2660 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2661 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2664 /* check if there is any duplicated entry */
2665 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2666 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2667 sizeof(m
[i
].old_daddr
)) &&
2668 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2669 sizeof(m
[i
].old_saddr
)) &&
2670 m
[i
].proto
== m
[j
].proto
&&
2671 m
[i
].mode
== m
[j
].mode
&&
2672 m
[i
].reqid
== m
[j
].reqid
&&
2673 m
[i
].old_family
== m
[j
].old_family
)
2681 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2682 struct xfrm_migrate
*m
, int num_migrate
,
2683 struct xfrm_kmaddress
*k
)
2685 int i
, err
, nx_cur
= 0, nx_new
= 0;
2686 struct xfrm_policy
*pol
= NULL
;
2687 struct xfrm_state
*x
, *xc
;
2688 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2689 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2690 struct xfrm_migrate
*mp
;
2692 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2695 /* Stage 1 - find policy */
2696 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2701 /* Stage 2 - find and update state(s) */
2702 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2703 if ((x
= xfrm_migrate_state_find(mp
))) {
2706 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2716 /* Stage 3 - update policy */
2717 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2720 /* Stage 4 - delete old state(s) */
2722 xfrm_states_put(x_cur
, nx_cur
);
2723 xfrm_states_delete(x_cur
, nx_cur
);
2726 /* Stage 5 - announce */
2727 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2739 xfrm_states_put(x_cur
, nx_cur
);
2741 xfrm_states_delete(x_new
, nx_new
);
2745 EXPORT_SYMBOL(xfrm_migrate
);