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(struct net
*net
, 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(net
, 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 struct net
*net
= xs_net(x
);
120 xfrm_address_t
*saddr
= &x
->props
.saddr
;
121 xfrm_address_t
*daddr
= &x
->id
.daddr
;
122 struct dst_entry
*dst
;
124 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
128 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
133 dst
= __xfrm_dst_lookup(net
, tos
, saddr
, daddr
, family
);
136 if (prev_saddr
!= saddr
)
137 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
138 if (prev_daddr
!= daddr
)
139 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
145 static inline unsigned long make_jiffies(long secs
)
147 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
148 return MAX_SCHEDULE_TIMEOUT
-1;
153 static void xfrm_policy_timer(unsigned long data
)
155 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
156 unsigned long now
= get_seconds();
157 long next
= LONG_MAX
;
161 read_lock(&xp
->lock
);
166 dir
= xfrm_policy_id2dir(xp
->index
);
168 if (xp
->lft
.hard_add_expires_seconds
) {
169 long tmo
= xp
->lft
.hard_add_expires_seconds
+
170 xp
->curlft
.add_time
- now
;
176 if (xp
->lft
.hard_use_expires_seconds
) {
177 long tmo
= xp
->lft
.hard_use_expires_seconds
+
178 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
184 if (xp
->lft
.soft_add_expires_seconds
) {
185 long tmo
= xp
->lft
.soft_add_expires_seconds
+
186 xp
->curlft
.add_time
- now
;
189 tmo
= XFRM_KM_TIMEOUT
;
194 if (xp
->lft
.soft_use_expires_seconds
) {
195 long tmo
= xp
->lft
.soft_use_expires_seconds
+
196 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
199 tmo
= XFRM_KM_TIMEOUT
;
206 km_policy_expired(xp
, dir
, 0, 0);
207 if (next
!= LONG_MAX
&&
208 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
212 read_unlock(&xp
->lock
);
217 read_unlock(&xp
->lock
);
218 if (!xfrm_policy_delete(xp
, dir
))
219 km_policy_expired(xp
, dir
, 1, 0);
224 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
228 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
230 struct xfrm_policy
*policy
;
232 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
235 write_pnet(&policy
->xp_net
, net
);
236 INIT_LIST_HEAD(&policy
->walk
.all
);
237 INIT_HLIST_NODE(&policy
->bydst
);
238 INIT_HLIST_NODE(&policy
->byidx
);
239 rwlock_init(&policy
->lock
);
240 atomic_set(&policy
->refcnt
, 1);
241 setup_timer(&policy
->timer
, xfrm_policy_timer
,
242 (unsigned long)policy
);
246 EXPORT_SYMBOL(xfrm_policy_alloc
);
248 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
250 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
252 BUG_ON(!policy
->walk
.dead
);
254 BUG_ON(policy
->bundles
);
256 if (del_timer(&policy
->timer
))
259 security_xfrm_policy_free(policy
->security
);
262 EXPORT_SYMBOL(xfrm_policy_destroy
);
264 static void xfrm_policy_gc_kill(struct xfrm_policy
*policy
)
266 struct dst_entry
*dst
;
268 while ((dst
= policy
->bundles
) != NULL
) {
269 policy
->bundles
= dst
->next
;
273 if (del_timer(&policy
->timer
))
274 atomic_dec(&policy
->refcnt
);
276 if (atomic_read(&policy
->refcnt
) > 1)
279 xfrm_pol_put(policy
);
282 static void xfrm_policy_gc_task(struct work_struct
*work
)
284 struct xfrm_policy
*policy
;
285 struct hlist_node
*entry
, *tmp
;
286 struct hlist_head gc_list
;
288 spin_lock_bh(&xfrm_policy_gc_lock
);
289 gc_list
.first
= xfrm_policy_gc_list
.first
;
290 INIT_HLIST_HEAD(&xfrm_policy_gc_list
);
291 spin_unlock_bh(&xfrm_policy_gc_lock
);
293 hlist_for_each_entry_safe(policy
, entry
, tmp
, &gc_list
, bydst
)
294 xfrm_policy_gc_kill(policy
);
296 static DECLARE_WORK(xfrm_policy_gc_work
, xfrm_policy_gc_task
);
298 /* Rule must be locked. Release descentant resources, announce
299 * entry dead. The rule must be unlinked from lists to the moment.
302 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
306 write_lock_bh(&policy
->lock
);
307 dead
= policy
->walk
.dead
;
308 policy
->walk
.dead
= 1;
309 write_unlock_bh(&policy
->lock
);
311 if (unlikely(dead
)) {
316 spin_lock_bh(&xfrm_policy_gc_lock
);
317 hlist_add_head(&policy
->bydst
, &xfrm_policy_gc_list
);
318 spin_unlock_bh(&xfrm_policy_gc_lock
);
320 schedule_work(&xfrm_policy_gc_work
);
323 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
325 static inline unsigned int idx_hash(struct net
*net
, u32 index
)
327 return __idx_hash(index
, net
->xfrm
.policy_idx_hmask
);
330 static struct hlist_head
*policy_hash_bysel(struct net
*net
, struct xfrm_selector
*sel
, unsigned short family
, int dir
)
332 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
333 unsigned int hash
= __sel_hash(sel
, family
, hmask
);
335 return (hash
== hmask
+ 1 ?
336 &net
->xfrm
.policy_inexact
[dir
] :
337 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
340 static struct hlist_head
*policy_hash_direct(struct net
*net
, xfrm_address_t
*daddr
, xfrm_address_t
*saddr
, unsigned short family
, int dir
)
342 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
343 unsigned int hash
= __addr_hash(daddr
, saddr
, family
, hmask
);
345 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
348 static void xfrm_dst_hash_transfer(struct hlist_head
*list
,
349 struct hlist_head
*ndsttable
,
350 unsigned int nhashmask
)
352 struct hlist_node
*entry
, *tmp
, *entry0
= NULL
;
353 struct xfrm_policy
*pol
;
357 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, bydst
) {
360 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
361 pol
->family
, nhashmask
);
364 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
370 hlist_add_after(entry0
, &pol
->bydst
);
374 if (!hlist_empty(list
)) {
380 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
381 struct hlist_head
*nidxtable
,
382 unsigned int nhashmask
)
384 struct hlist_node
*entry
, *tmp
;
385 struct xfrm_policy
*pol
;
387 hlist_for_each_entry_safe(pol
, entry
, tmp
, list
, byidx
) {
390 h
= __idx_hash(pol
->index
, nhashmask
);
391 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
395 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
397 return ((old_hmask
+ 1) << 1) - 1;
400 static void xfrm_bydst_resize(struct net
*net
, int dir
)
402 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
403 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
404 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
405 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
406 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
412 write_lock_bh(&xfrm_policy_lock
);
414 for (i
= hmask
; i
>= 0; i
--)
415 xfrm_dst_hash_transfer(odst
+ i
, ndst
, nhashmask
);
417 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
418 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
420 write_unlock_bh(&xfrm_policy_lock
);
422 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
425 static void xfrm_byidx_resize(struct net
*net
, int total
)
427 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
428 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
429 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
430 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
431 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
437 write_lock_bh(&xfrm_policy_lock
);
439 for (i
= hmask
; i
>= 0; i
--)
440 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
442 net
->xfrm
.policy_byidx
= nidx
;
443 net
->xfrm
.policy_idx_hmask
= nhashmask
;
445 write_unlock_bh(&xfrm_policy_lock
);
447 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
450 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
452 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
453 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
458 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
465 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
467 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
469 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
476 void xfrm_spd_getinfo(struct xfrmk_spdinfo
*si
)
478 read_lock_bh(&xfrm_policy_lock
);
479 si
->incnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
];
480 si
->outcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
];
481 si
->fwdcnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
];
482 si
->inscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
483 si
->outscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
484 si
->fwdscnt
= init_net
.xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
485 si
->spdhcnt
= init_net
.xfrm
.policy_idx_hmask
;
486 si
->spdhmcnt
= xfrm_policy_hashmax
;
487 read_unlock_bh(&xfrm_policy_lock
);
489 EXPORT_SYMBOL(xfrm_spd_getinfo
);
491 static DEFINE_MUTEX(hash_resize_mutex
);
492 static void xfrm_hash_resize(struct work_struct
*work
)
494 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
497 mutex_lock(&hash_resize_mutex
);
500 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
501 if (xfrm_bydst_should_resize(net
, dir
, &total
))
502 xfrm_bydst_resize(net
, dir
);
504 if (xfrm_byidx_should_resize(net
, total
))
505 xfrm_byidx_resize(net
, total
);
507 mutex_unlock(&hash_resize_mutex
);
510 /* Generate new index... KAME seems to generate them ordered by cost
511 * of an absolute inpredictability of ordering of rules. This will not pass. */
512 static u32
xfrm_gen_index(struct net
*net
, int dir
)
514 static u32 idx_generator
;
517 struct hlist_node
*entry
;
518 struct hlist_head
*list
;
519 struct xfrm_policy
*p
;
523 idx
= (idx_generator
| dir
);
527 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
529 hlist_for_each_entry(p
, entry
, list
, byidx
) {
530 if (p
->index
== idx
) {
540 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
542 u32
*p1
= (u32
*) s1
;
543 u32
*p2
= (u32
*) s2
;
544 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
547 for (i
= 0; i
< len
; i
++) {
555 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
557 struct net
*net
= xp_net(policy
);
558 struct xfrm_policy
*pol
;
559 struct xfrm_policy
*delpol
;
560 struct hlist_head
*chain
;
561 struct hlist_node
*entry
, *newpos
;
562 struct dst_entry
*gc_list
;
564 write_lock_bh(&xfrm_policy_lock
);
565 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
568 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
569 if (pol
->type
== policy
->type
&&
570 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
571 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
574 write_unlock_bh(&xfrm_policy_lock
);
578 if (policy
->priority
> pol
->priority
)
580 } else if (policy
->priority
>= pol
->priority
) {
581 newpos
= &pol
->bydst
;
588 hlist_add_after(newpos
, &policy
->bydst
);
590 hlist_add_head(&policy
->bydst
, chain
);
591 xfrm_pol_hold(policy
);
592 net
->xfrm
.policy_count
[dir
]++;
593 atomic_inc(&flow_cache_genid
);
595 hlist_del(&delpol
->bydst
);
596 hlist_del(&delpol
->byidx
);
597 list_del(&delpol
->walk
.all
);
598 net
->xfrm
.policy_count
[dir
]--;
600 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
);
601 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
602 policy
->curlft
.add_time
= get_seconds();
603 policy
->curlft
.use_time
= 0;
604 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
605 xfrm_pol_hold(policy
);
606 list_add(&policy
->walk
.all
, &net
->xfrm
.policy_all
);
607 write_unlock_bh(&xfrm_policy_lock
);
610 xfrm_policy_kill(delpol
);
611 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
612 schedule_work(&net
->xfrm
.policy_hash_work
);
614 read_lock_bh(&xfrm_policy_lock
);
616 entry
= &policy
->bydst
;
617 hlist_for_each_entry_continue(policy
, entry
, bydst
) {
618 struct dst_entry
*dst
;
620 write_lock(&policy
->lock
);
621 dst
= policy
->bundles
;
623 struct dst_entry
*tail
= dst
;
626 tail
->next
= gc_list
;
629 policy
->bundles
= NULL
;
631 write_unlock(&policy
->lock
);
633 read_unlock_bh(&xfrm_policy_lock
);
636 struct dst_entry
*dst
= gc_list
;
644 EXPORT_SYMBOL(xfrm_policy_insert
);
646 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u8 type
, int dir
,
647 struct xfrm_selector
*sel
,
648 struct xfrm_sec_ctx
*ctx
, int delete,
651 struct xfrm_policy
*pol
, *ret
;
652 struct hlist_head
*chain
;
653 struct hlist_node
*entry
;
656 write_lock_bh(&xfrm_policy_lock
);
657 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
659 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
660 if (pol
->type
== type
&&
661 !selector_cmp(sel
, &pol
->selector
) &&
662 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
665 *err
= security_xfrm_policy_delete(
668 write_unlock_bh(&xfrm_policy_lock
);
671 hlist_del(&pol
->bydst
);
672 hlist_del(&pol
->byidx
);
673 list_del(&pol
->walk
.all
);
674 net
->xfrm
.policy_count
[dir
]--;
680 write_unlock_bh(&xfrm_policy_lock
);
683 atomic_inc(&flow_cache_genid
);
684 xfrm_policy_kill(ret
);
688 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
690 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u8 type
, int dir
, u32 id
,
691 int delete, int *err
)
693 struct xfrm_policy
*pol
, *ret
;
694 struct hlist_head
*chain
;
695 struct hlist_node
*entry
;
698 if (xfrm_policy_id2dir(id
) != dir
)
702 write_lock_bh(&xfrm_policy_lock
);
703 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
705 hlist_for_each_entry(pol
, entry
, chain
, byidx
) {
706 if (pol
->type
== type
&& pol
->index
== id
) {
709 *err
= security_xfrm_policy_delete(
712 write_unlock_bh(&xfrm_policy_lock
);
715 hlist_del(&pol
->bydst
);
716 hlist_del(&pol
->byidx
);
717 list_del(&pol
->walk
.all
);
718 net
->xfrm
.policy_count
[dir
]--;
724 write_unlock_bh(&xfrm_policy_lock
);
727 atomic_inc(&flow_cache_genid
);
728 xfrm_policy_kill(ret
);
732 EXPORT_SYMBOL(xfrm_policy_byid
);
734 #ifdef CONFIG_SECURITY_NETWORK_XFRM
736 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
740 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
741 struct xfrm_policy
*pol
;
742 struct hlist_node
*entry
;
745 hlist_for_each_entry(pol
, entry
,
746 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
747 if (pol
->type
!= type
)
749 err
= security_xfrm_policy_delete(pol
->security
);
751 xfrm_audit_policy_delete(pol
, 0,
752 audit_info
->loginuid
,
753 audit_info
->sessionid
,
758 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
759 hlist_for_each_entry(pol
, entry
,
760 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
762 if (pol
->type
!= type
)
764 err
= security_xfrm_policy_delete(
767 xfrm_audit_policy_delete(pol
, 0,
768 audit_info
->loginuid
,
769 audit_info
->sessionid
,
780 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
786 int xfrm_policy_flush(struct net
*net
, u8 type
, struct xfrm_audit
*audit_info
)
790 write_lock_bh(&xfrm_policy_lock
);
792 err
= xfrm_policy_flush_secctx_check(net
, type
, audit_info
);
796 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
797 struct xfrm_policy
*pol
;
798 struct hlist_node
*entry
;
803 hlist_for_each_entry(pol
, entry
,
804 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
805 if (pol
->type
!= type
)
807 hlist_del(&pol
->bydst
);
808 hlist_del(&pol
->byidx
);
809 write_unlock_bh(&xfrm_policy_lock
);
811 xfrm_audit_policy_delete(pol
, 1, audit_info
->loginuid
,
812 audit_info
->sessionid
,
815 xfrm_policy_kill(pol
);
818 write_lock_bh(&xfrm_policy_lock
);
822 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
824 hlist_for_each_entry(pol
, entry
,
825 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
827 if (pol
->type
!= type
)
829 hlist_del(&pol
->bydst
);
830 hlist_del(&pol
->byidx
);
831 list_del(&pol
->walk
.all
);
832 write_unlock_bh(&xfrm_policy_lock
);
834 xfrm_audit_policy_delete(pol
, 1,
835 audit_info
->loginuid
,
836 audit_info
->sessionid
,
838 xfrm_policy_kill(pol
);
841 write_lock_bh(&xfrm_policy_lock
);
846 net
->xfrm
.policy_count
[dir
] -= killed
;
848 atomic_inc(&flow_cache_genid
);
850 write_unlock_bh(&xfrm_policy_lock
);
853 EXPORT_SYMBOL(xfrm_policy_flush
);
855 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
856 int (*func
)(struct xfrm_policy
*, int, int, void*),
859 struct xfrm_policy
*pol
;
860 struct xfrm_policy_walk_entry
*x
;
863 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
864 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
867 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
870 write_lock_bh(&xfrm_policy_lock
);
871 if (list_empty(&walk
->walk
.all
))
872 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
874 x
= list_entry(&walk
->walk
.all
, struct xfrm_policy_walk_entry
, all
);
875 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
878 pol
= container_of(x
, struct xfrm_policy
, walk
);
879 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
880 walk
->type
!= pol
->type
)
882 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
885 list_move_tail(&walk
->walk
.all
, &x
->all
);
890 if (walk
->seq
== 0) {
894 list_del_init(&walk
->walk
.all
);
896 write_unlock_bh(&xfrm_policy_lock
);
899 EXPORT_SYMBOL(xfrm_policy_walk
);
901 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
903 INIT_LIST_HEAD(&walk
->walk
.all
);
908 EXPORT_SYMBOL(xfrm_policy_walk_init
);
910 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
)
912 if (list_empty(&walk
->walk
.all
))
915 write_lock_bh(&xfrm_policy_lock
);
916 list_del(&walk
->walk
.all
);
917 write_unlock_bh(&xfrm_policy_lock
);
919 EXPORT_SYMBOL(xfrm_policy_walk_done
);
922 * Find policy to apply to this flow.
924 * Returns 0 if policy found, else an -errno.
926 static int xfrm_policy_match(struct xfrm_policy
*pol
, struct flowi
*fl
,
927 u8 type
, u16 family
, int dir
)
929 struct xfrm_selector
*sel
= &pol
->selector
;
930 int match
, ret
= -ESRCH
;
932 if (pol
->family
!= family
||
936 match
= xfrm_selector_match(sel
, fl
, family
);
938 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->secid
,
944 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
949 struct xfrm_policy
*pol
, *ret
;
950 xfrm_address_t
*daddr
, *saddr
;
951 struct hlist_node
*entry
;
952 struct hlist_head
*chain
;
955 daddr
= xfrm_flowi_daddr(fl
, family
);
956 saddr
= xfrm_flowi_saddr(fl
, family
);
957 if (unlikely(!daddr
|| !saddr
))
960 read_lock_bh(&xfrm_policy_lock
);
961 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
963 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
964 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
974 priority
= ret
->priority
;
978 chain
= &net
->xfrm
.policy_inexact
[dir
];
979 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
980 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
988 } else if (pol
->priority
< priority
) {
996 read_unlock_bh(&xfrm_policy_lock
);
1001 static int xfrm_policy_lookup(struct net
*net
, struct flowi
*fl
, u16 family
,
1002 u8 dir
, void **objp
, atomic_t
**obj_refp
)
1004 struct xfrm_policy
*pol
;
1007 #ifdef CONFIG_XFRM_SUB_POLICY
1008 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1016 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1021 #ifdef CONFIG_XFRM_SUB_POLICY
1024 if ((*objp
= (void *) pol
) != NULL
)
1025 *obj_refp
= &pol
->refcnt
;
1029 static inline int policy_to_flow_dir(int dir
)
1031 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1032 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1033 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1037 case XFRM_POLICY_IN
:
1039 case XFRM_POLICY_OUT
:
1040 return FLOW_DIR_OUT
;
1041 case XFRM_POLICY_FWD
:
1042 return FLOW_DIR_FWD
;
1046 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
, struct flowi
*fl
)
1048 struct xfrm_policy
*pol
;
1050 read_lock_bh(&xfrm_policy_lock
);
1051 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1052 int match
= xfrm_selector_match(&pol
->selector
, fl
,
1057 err
= security_xfrm_policy_lookup(pol
->security
,
1059 policy_to_flow_dir(dir
));
1062 else if (err
== -ESRCH
)
1069 read_unlock_bh(&xfrm_policy_lock
);
1073 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1075 struct net
*net
= xp_net(pol
);
1076 struct hlist_head
*chain
= policy_hash_bysel(net
, &pol
->selector
,
1079 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1080 hlist_add_head(&pol
->bydst
, chain
);
1081 hlist_add_head(&pol
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, pol
->index
));
1082 net
->xfrm
.policy_count
[dir
]++;
1085 if (xfrm_bydst_should_resize(net
, dir
, NULL
))
1086 schedule_work(&net
->xfrm
.policy_hash_work
);
1089 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1092 struct net
*net
= xp_net(pol
);
1094 if (hlist_unhashed(&pol
->bydst
))
1097 hlist_del(&pol
->bydst
);
1098 hlist_del(&pol
->byidx
);
1099 list_del(&pol
->walk
.all
);
1100 net
->xfrm
.policy_count
[dir
]--;
1105 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1107 write_lock_bh(&xfrm_policy_lock
);
1108 pol
= __xfrm_policy_unlink(pol
, dir
);
1109 write_unlock_bh(&xfrm_policy_lock
);
1111 if (dir
< XFRM_POLICY_MAX
)
1112 atomic_inc(&flow_cache_genid
);
1113 xfrm_policy_kill(pol
);
1118 EXPORT_SYMBOL(xfrm_policy_delete
);
1120 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1122 struct net
*net
= xp_net(pol
);
1123 struct xfrm_policy
*old_pol
;
1125 #ifdef CONFIG_XFRM_SUB_POLICY
1126 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1130 write_lock_bh(&xfrm_policy_lock
);
1131 old_pol
= sk
->sk_policy
[dir
];
1132 sk
->sk_policy
[dir
] = pol
;
1134 pol
->curlft
.add_time
= get_seconds();
1135 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
);
1136 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+dir
);
1139 __xfrm_policy_unlink(old_pol
, XFRM_POLICY_MAX
+dir
);
1140 write_unlock_bh(&xfrm_policy_lock
);
1143 xfrm_policy_kill(old_pol
);
1148 static struct xfrm_policy
*clone_policy(struct xfrm_policy
*old
, int dir
)
1150 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1153 newp
->selector
= old
->selector
;
1154 if (security_xfrm_policy_clone(old
->security
,
1157 return NULL
; /* ENOMEM */
1159 newp
->lft
= old
->lft
;
1160 newp
->curlft
= old
->curlft
;
1161 newp
->action
= old
->action
;
1162 newp
->flags
= old
->flags
;
1163 newp
->xfrm_nr
= old
->xfrm_nr
;
1164 newp
->index
= old
->index
;
1165 newp
->type
= old
->type
;
1166 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1167 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1168 write_lock_bh(&xfrm_policy_lock
);
1169 __xfrm_policy_link(newp
, XFRM_POLICY_MAX
+dir
);
1170 write_unlock_bh(&xfrm_policy_lock
);
1176 int __xfrm_sk_clone_policy(struct sock
*sk
)
1178 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1179 *p1
= sk
->sk_policy
[1];
1181 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1182 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1184 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1190 xfrm_get_saddr(struct net
*net
, xfrm_address_t
*local
, xfrm_address_t
*remote
,
1191 unsigned short family
)
1194 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1196 if (unlikely(afinfo
== NULL
))
1198 err
= afinfo
->get_saddr(net
, local
, remote
);
1199 xfrm_policy_put_afinfo(afinfo
);
1203 /* Resolve list of templates for the flow, given policy. */
1206 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, struct flowi
*fl
,
1207 struct xfrm_state
**xfrm
,
1208 unsigned short family
)
1210 struct net
*net
= xp_net(policy
);
1213 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1214 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1217 for (nx
=0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1218 struct xfrm_state
*x
;
1219 xfrm_address_t
*remote
= daddr
;
1220 xfrm_address_t
*local
= saddr
;
1221 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1223 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1224 tmpl
->mode
== XFRM_MODE_BEET
) {
1225 remote
= &tmpl
->id
.daddr
;
1226 local
= &tmpl
->saddr
;
1227 family
= tmpl
->encap_family
;
1228 if (xfrm_addr_any(local
, family
)) {
1229 error
= xfrm_get_saddr(net
, &tmp
, remote
, family
);
1236 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1238 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1245 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1249 else if (error
== -ESRCH
)
1252 if (!tmpl
->optional
)
1258 for (nx
--; nx
>=0; nx
--)
1259 xfrm_state_put(xfrm
[nx
]);
1264 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, struct flowi
*fl
,
1265 struct xfrm_state
**xfrm
,
1266 unsigned short family
)
1268 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1269 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1275 for (i
= 0; i
< npols
; i
++) {
1276 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1281 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1289 /* found states are sorted for outbound processing */
1291 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1296 for (cnx
--; cnx
>=0; cnx
--)
1297 xfrm_state_put(tpp
[cnx
]);
1302 /* Check that the bundle accepts the flow and its components are
1306 static struct dst_entry
*
1307 xfrm_find_bundle(struct flowi
*fl
, struct xfrm_policy
*policy
, unsigned short family
)
1309 struct dst_entry
*x
;
1310 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1311 if (unlikely(afinfo
== NULL
))
1312 return ERR_PTR(-EINVAL
);
1313 x
= afinfo
->find_bundle(fl
, policy
);
1314 xfrm_policy_put_afinfo(afinfo
);
1318 static inline int xfrm_get_tos(struct flowi
*fl
, int family
)
1320 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1326 tos
= afinfo
->get_tos(fl
);
1328 xfrm_policy_put_afinfo(afinfo
);
1333 static inline struct xfrm_dst
*xfrm_alloc_dst(int family
)
1335 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1336 struct xfrm_dst
*xdst
;
1339 return ERR_PTR(-EINVAL
);
1341 xdst
= dst_alloc(afinfo
->dst_ops
) ?: ERR_PTR(-ENOBUFS
);
1343 xfrm_policy_put_afinfo(afinfo
);
1348 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1351 struct xfrm_policy_afinfo
*afinfo
=
1352 xfrm_policy_get_afinfo(dst
->ops
->family
);
1358 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1360 xfrm_policy_put_afinfo(afinfo
);
1365 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
)
1367 struct xfrm_policy_afinfo
*afinfo
=
1368 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1374 err
= afinfo
->fill_dst(xdst
, dev
);
1376 xfrm_policy_put_afinfo(afinfo
);
1381 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1382 * all the metrics... Shortly, bundle a bundle.
1385 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1386 struct xfrm_state
**xfrm
, int nx
,
1388 struct dst_entry
*dst
)
1390 unsigned long now
= jiffies
;
1391 struct net_device
*dev
;
1392 struct dst_entry
*dst_prev
= NULL
;
1393 struct dst_entry
*dst0
= NULL
;
1397 int nfheader_len
= 0;
1398 int trailer_len
= 0;
1400 int family
= policy
->selector
.family
;
1401 xfrm_address_t saddr
, daddr
;
1403 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1405 tos
= xfrm_get_tos(fl
, family
);
1412 for (; i
< nx
; i
++) {
1413 struct xfrm_dst
*xdst
= xfrm_alloc_dst(family
);
1414 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1416 err
= PTR_ERR(xdst
);
1425 dst_prev
->child
= dst_clone(dst1
);
1426 dst1
->flags
|= DST_NOHASH
;
1430 memcpy(&dst1
->metrics
, &dst
->metrics
, sizeof(dst
->metrics
));
1432 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1433 family
= xfrm
[i
]->props
.family
;
1434 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, &saddr
, &daddr
,
1442 dst1
->xfrm
= xfrm
[i
];
1443 xdst
->genid
= xfrm
[i
]->genid
;
1445 dst1
->obsolete
= -1;
1446 dst1
->flags
|= DST_HOST
;
1447 dst1
->lastuse
= now
;
1449 dst1
->input
= dst_discard
;
1450 dst1
->output
= xfrm
[i
]->outer_mode
->afinfo
->output
;
1452 dst1
->next
= dst_prev
;
1455 header_len
+= xfrm
[i
]->props
.header_len
;
1456 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1457 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1458 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1461 dst_prev
->child
= dst
;
1469 /* Copy neighbout for reachability confirmation */
1470 dst0
->neighbour
= neigh_clone(dst
->neighbour
);
1472 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1473 xfrm_init_pmtu(dst_prev
);
1475 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1476 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1478 err
= xfrm_fill_dst(xdst
, dev
);
1482 dst_prev
->header_len
= header_len
;
1483 dst_prev
->trailer_len
= trailer_len
;
1484 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1485 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1493 xfrm_state_put(xfrm
[i
]);
1497 dst0
= ERR_PTR(err
);
1502 xfrm_dst_alloc_copy(void **target
, void *src
, int size
)
1505 *target
= kmalloc(size
, GFP_ATOMIC
);
1509 memcpy(*target
, src
, size
);
1514 xfrm_dst_update_parent(struct dst_entry
*dst
, struct xfrm_selector
*sel
)
1516 #ifdef CONFIG_XFRM_SUB_POLICY
1517 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1518 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1526 xfrm_dst_update_origin(struct dst_entry
*dst
, struct flowi
*fl
)
1528 #ifdef CONFIG_XFRM_SUB_POLICY
1529 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1530 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1536 static int stale_bundle(struct dst_entry
*dst
);
1538 /* Main function: finds/creates a bundle for given flow.
1540 * At the moment we eat a raw IP route. Mostly to speed up lookups
1541 * on interfaces with disabled IPsec.
1543 int __xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1544 struct sock
*sk
, int flags
)
1546 struct xfrm_policy
*policy
;
1547 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1552 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1553 struct dst_entry
*dst
, *dst_orig
= *dst_p
;
1558 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
1561 genid
= atomic_read(&flow_cache_genid
);
1563 for (pi
= 0; pi
< ARRAY_SIZE(pols
); pi
++)
1569 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
1570 policy
= xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
1571 err
= PTR_ERR(policy
);
1572 if (IS_ERR(policy
)) {
1573 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1579 /* To accelerate a bit... */
1580 if ((dst_orig
->flags
& DST_NOXFRM
) ||
1581 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
1584 policy
= flow_cache_lookup(net
, fl
, dst_orig
->ops
->family
,
1585 dir
, xfrm_policy_lookup
);
1586 err
= PTR_ERR(policy
);
1587 if (IS_ERR(policy
)) {
1588 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1596 family
= dst_orig
->ops
->family
;
1599 xfrm_nr
+= pols
[0]->xfrm_nr
;
1602 if ((flags
& XFRM_LOOKUP_ICMP
) && !(policy
->flags
& XFRM_POLICY_ICMP
))
1605 policy
->curlft
.use_time
= get_seconds();
1607 switch (policy
->action
) {
1609 case XFRM_POLICY_BLOCK
:
1610 /* Prohibit the flow */
1611 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1615 case XFRM_POLICY_ALLOW
:
1616 #ifndef CONFIG_XFRM_SUB_POLICY
1617 if (policy
->xfrm_nr
== 0) {
1618 /* Flow passes not transformed. */
1619 xfrm_pol_put(policy
);
1624 /* Try to find matching bundle.
1626 * LATER: help from flow cache. It is optional, this
1627 * is required only for output policy.
1629 dst
= xfrm_find_bundle(fl
, policy
, family
);
1631 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1639 #ifdef CONFIG_XFRM_SUB_POLICY
1640 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1641 pols
[1] = xfrm_policy_lookup_bytype(net
,
1642 XFRM_POLICY_TYPE_MAIN
,
1646 if (IS_ERR(pols
[1])) {
1647 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR
);
1648 err
= PTR_ERR(pols
[1]);
1651 if (pols
[1]->action
== XFRM_POLICY_BLOCK
) {
1652 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK
);
1657 xfrm_nr
+= pols
[1]->xfrm_nr
;
1662 * Because neither flowi nor bundle information knows about
1663 * transformation template size. On more than one policy usage
1664 * we can realize whether all of them is bypass or not after
1665 * they are searched. See above not-transformed bypass
1666 * is surrounded by non-sub policy configuration, too.
1669 /* Flow passes not transformed. */
1670 xfrm_pols_put(pols
, npols
);
1675 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1677 if (unlikely(nx
<0)) {
1679 if (err
== -EAGAIN
&& sysctl_xfrm_larval_drop
) {
1680 /* EREMOTE tells the caller to generate
1681 * a one-shot blackhole route.
1683 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1684 xfrm_pol_put(policy
);
1687 if (err
== -EAGAIN
&& (flags
& XFRM_LOOKUP_WAIT
)) {
1688 DECLARE_WAITQUEUE(wait
, current
);
1690 add_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1691 set_current_state(TASK_INTERRUPTIBLE
);
1693 set_current_state(TASK_RUNNING
);
1694 remove_wait_queue(&net
->xfrm
.km_waitq
, &wait
);
1696 nx
= xfrm_tmpl_resolve(pols
, npols
, fl
, xfrm
, family
);
1698 if (nx
== -EAGAIN
&& signal_pending(current
)) {
1699 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1703 if (nx
== -EAGAIN
||
1704 genid
!= atomic_read(&flow_cache_genid
)) {
1705 xfrm_pols_put(pols
, npols
);
1711 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES
);
1716 /* Flow passes not transformed. */
1717 xfrm_pols_put(pols
, npols
);
1721 dst
= xfrm_bundle_create(policy
, xfrm
, nx
, fl
, dst_orig
);
1724 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1728 for (pi
= 0; pi
< npols
; pi
++) {
1729 read_lock_bh(&pols
[pi
]->lock
);
1730 pol_dead
|= pols
[pi
]->walk
.dead
;
1731 read_unlock_bh(&pols
[pi
]->lock
);
1734 write_lock_bh(&policy
->lock
);
1735 if (unlikely(pol_dead
|| stale_bundle(dst
))) {
1736 /* Wow! While we worked on resolving, this
1737 * policy has gone. Retry. It is not paranoia,
1738 * we just cannot enlist new bundle to dead object.
1739 * We can't enlist stable bundles either.
1741 write_unlock_bh(&policy
->lock
);
1745 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD
);
1747 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1748 err
= -EHOSTUNREACH
;
1753 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1755 err
= xfrm_dst_update_origin(dst
, fl
);
1756 if (unlikely(err
)) {
1757 write_unlock_bh(&policy
->lock
);
1759 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1763 dst
->next
= policy
->bundles
;
1764 policy
->bundles
= dst
;
1766 write_unlock_bh(&policy
->lock
);
1769 dst_release(dst_orig
);
1770 xfrm_pols_put(pols
, npols
);
1774 xfrm_pols_put(pols
, npols
);
1776 dst_release(dst_orig
);
1782 if (flags
& XFRM_LOOKUP_ICMP
)
1786 EXPORT_SYMBOL(__xfrm_lookup
);
1788 int xfrm_lookup(struct net
*net
, struct dst_entry
**dst_p
, struct flowi
*fl
,
1789 struct sock
*sk
, int flags
)
1791 int err
= __xfrm_lookup(net
, dst_p
, fl
, sk
, flags
);
1793 if (err
== -EREMOTE
) {
1794 dst_release(*dst_p
);
1801 EXPORT_SYMBOL(xfrm_lookup
);
1804 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, struct flowi
*fl
)
1806 struct xfrm_state
*x
;
1808 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
1810 x
= skb
->sp
->xvec
[idx
];
1811 if (!x
->type
->reject
)
1813 return x
->type
->reject(x
, skb
, fl
);
1816 /* When skb is transformed back to its "native" form, we have to
1817 * check policy restrictions. At the moment we make this in maximally
1818 * stupid way. Shame on me. :-) Of course, connected sockets must
1819 * have policy cached at them.
1823 xfrm_state_ok(struct xfrm_tmpl
*tmpl
, struct xfrm_state
*x
,
1824 unsigned short family
)
1826 if (xfrm_state_kern(x
))
1827 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
1828 return x
->id
.proto
== tmpl
->id
.proto
&&
1829 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
1830 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
1831 x
->props
.mode
== tmpl
->mode
&&
1832 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
1833 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
1834 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
1835 xfrm_state_addr_cmp(tmpl
, x
, family
));
1839 * 0 or more than 0 is returned when validation is succeeded (either bypass
1840 * because of optional transport mode, or next index of the mathced secpath
1841 * state with the template.
1842 * -1 is returned when no matching template is found.
1843 * Otherwise "-2 - errored_index" is returned.
1846 xfrm_policy_ok(struct xfrm_tmpl
*tmpl
, struct sec_path
*sp
, int start
,
1847 unsigned short family
)
1851 if (tmpl
->optional
) {
1852 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
1856 for (; idx
< sp
->len
; idx
++) {
1857 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
1859 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1868 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
1869 unsigned int family
, int reverse
)
1871 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1874 if (unlikely(afinfo
== NULL
))
1875 return -EAFNOSUPPORT
;
1877 afinfo
->decode_session(skb
, fl
, reverse
);
1878 err
= security_xfrm_decode_session(skb
, &fl
->secid
);
1879 xfrm_policy_put_afinfo(afinfo
);
1882 EXPORT_SYMBOL(__xfrm_decode_session
);
1884 static inline int secpath_has_nontransport(struct sec_path
*sp
, int k
, int *idxp
)
1886 for (; k
< sp
->len
; k
++) {
1887 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1896 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
1897 unsigned short family
)
1899 struct net
*net
= dev_net(skb
->dev
);
1900 struct xfrm_policy
*pol
;
1901 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
1910 reverse
= dir
& ~XFRM_POLICY_MASK
;
1911 dir
&= XFRM_POLICY_MASK
;
1912 fl_dir
= policy_to_flow_dir(dir
);
1914 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
1915 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR
);
1919 nf_nat_decode_session(skb
, &fl
, family
);
1921 /* First, check used SA against their selectors. */
1925 for (i
=skb
->sp
->len
-1; i
>=0; i
--) {
1926 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
1927 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
1928 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH
);
1935 if (sk
&& sk
->sk_policy
[dir
]) {
1936 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
1938 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1944 pol
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
1945 xfrm_policy_lookup
);
1948 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1953 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
1954 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
1955 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS
);
1961 pol
->curlft
.use_time
= get_seconds();
1965 #ifdef CONFIG_XFRM_SUB_POLICY
1966 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1967 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
1971 if (IS_ERR(pols
[1])) {
1972 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR
);
1975 pols
[1]->curlft
.use_time
= get_seconds();
1981 if (pol
->action
== XFRM_POLICY_ALLOW
) {
1982 struct sec_path
*sp
;
1983 static struct sec_path dummy
;
1984 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
1985 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
1986 struct xfrm_tmpl
**tpp
= tp
;
1990 if ((sp
= skb
->sp
) == NULL
)
1993 for (pi
= 0; pi
< npols
; pi
++) {
1994 if (pols
[pi
] != pol
&&
1995 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
1996 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
1999 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2000 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR
);
2003 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2004 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2008 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
);
2012 /* For each tunnel xfrm, find the first matching tmpl.
2013 * For each tmpl before that, find corresponding xfrm.
2014 * Order is _important_. Later we will implement
2015 * some barriers, but at the moment barriers
2016 * are implied between each two transformations.
2018 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2019 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2022 /* "-2 - errored_index" returned */
2024 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2029 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2030 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH
);
2034 xfrm_pols_put(pols
, npols
);
2037 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK
);
2040 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2042 xfrm_pols_put(pols
, npols
);
2045 EXPORT_SYMBOL(__xfrm_policy_check
);
2047 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2049 struct net
*net
= dev_net(skb
->dev
);
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(net
, &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(struct net
*net
, 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 &net
->xfrm
.policy_inexact
[dir
], bydst
)
2156 prune_one_bundle(pol
, func
, &gc_list
);
2158 table
= net
->xfrm
.policy_bydst
[dir
].table
;
2159 for (i
= net
->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(struct net
*net
)
2180 xfrm_prune_bundles(net
, unused_bundle
);
2183 static int xfrm_flush_bundles(struct net
*net
)
2185 xfrm_prune_bundles(net
, 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
;
2373 xfrm_flush_bundles(dev_net(dev
));
2378 static struct notifier_block xfrm_dev_notifier
= {
2379 .notifier_call
= xfrm_dev_event
,
2382 #ifdef CONFIG_XFRM_STATISTICS
2383 static int __init
xfrm_statistics_init(void)
2385 if (snmp_mib_init((void **)xfrm_statistics
,
2386 sizeof(struct linux_xfrm_mib
)) < 0)
2392 static int __net_init
xfrm_policy_init(struct net
*net
)
2394 unsigned int hmask
, sz
;
2397 if (net_eq(net
, &init_net
))
2398 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2399 sizeof(struct xfrm_dst
),
2400 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2404 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2406 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2407 if (!net
->xfrm
.policy_byidx
)
2409 net
->xfrm
.policy_idx_hmask
= hmask
;
2411 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2412 struct xfrm_policy_hash
*htab
;
2414 net
->xfrm
.policy_count
[dir
] = 0;
2415 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2417 htab
= &net
->xfrm
.policy_bydst
[dir
];
2418 htab
->table
= xfrm_hash_alloc(sz
);
2421 htab
->hmask
= hmask
;
2424 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
2425 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
2426 if (net_eq(net
, &init_net
))
2427 register_netdevice_notifier(&xfrm_dev_notifier
);
2431 for (dir
--; dir
>= 0; dir
--) {
2432 struct xfrm_policy_hash
*htab
;
2434 htab
= &net
->xfrm
.policy_bydst
[dir
];
2435 xfrm_hash_free(htab
->table
, sz
);
2437 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2442 static void xfrm_policy_fini(struct net
*net
)
2444 struct xfrm_audit audit_info
;
2448 flush_work(&net
->xfrm
.policy_hash_work
);
2449 #ifdef CONFIG_XFRM_SUB_POLICY
2450 audit_info
.loginuid
= -1;
2451 audit_info
.sessionid
= -1;
2452 audit_info
.secid
= 0;
2453 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, &audit_info
);
2455 audit_info
.loginuid
= -1;
2456 audit_info
.sessionid
= -1;
2457 audit_info
.secid
= 0;
2458 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, &audit_info
);
2459 flush_work(&xfrm_policy_gc_work
);
2461 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
2463 for (dir
= 0; dir
< XFRM_POLICY_MAX
* 2; dir
++) {
2464 struct xfrm_policy_hash
*htab
;
2466 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
2468 htab
= &net
->xfrm
.policy_bydst
[dir
];
2469 sz
= (htab
->hmask
+ 1);
2470 WARN_ON(!hlist_empty(htab
->table
));
2471 xfrm_hash_free(htab
->table
, sz
);
2474 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
2475 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
2476 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
2479 static int __net_init
xfrm_net_init(struct net
*net
)
2483 rv
= xfrm_state_init(net
);
2486 rv
= xfrm_policy_init(net
);
2492 xfrm_state_fini(net
);
2497 static void __net_exit
xfrm_net_exit(struct net
*net
)
2499 xfrm_policy_fini(net
);
2500 xfrm_state_fini(net
);
2503 static struct pernet_operations __net_initdata xfrm_net_ops
= {
2504 .init
= xfrm_net_init
,
2505 .exit
= xfrm_net_exit
,
2508 void __init
xfrm_init(void)
2510 register_pernet_subsys(&xfrm_net_ops
);
2511 #ifdef CONFIG_XFRM_STATISTICS
2512 xfrm_statistics_init();
2515 #ifdef CONFIG_XFRM_STATISTICS
2520 #ifdef CONFIG_AUDITSYSCALL
2521 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
2522 struct audit_buffer
*audit_buf
)
2524 struct xfrm_sec_ctx
*ctx
= xp
->security
;
2525 struct xfrm_selector
*sel
= &xp
->selector
;
2528 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
2529 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
2531 switch(sel
->family
) {
2533 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
2534 if (sel
->prefixlen_s
!= 32)
2535 audit_log_format(audit_buf
, " src_prefixlen=%d",
2537 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
2538 if (sel
->prefixlen_d
!= 32)
2539 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2543 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
2544 if (sel
->prefixlen_s
!= 128)
2545 audit_log_format(audit_buf
, " src_prefixlen=%d",
2547 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
2548 if (sel
->prefixlen_d
!= 128)
2549 audit_log_format(audit_buf
, " dst_prefixlen=%d",
2555 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
,
2556 uid_t auid
, u32 sessionid
, u32 secid
)
2558 struct audit_buffer
*audit_buf
;
2560 audit_buf
= xfrm_audit_start("SPD-add");
2561 if (audit_buf
== NULL
)
2563 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2564 audit_log_format(audit_buf
, " res=%u", result
);
2565 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2566 audit_log_end(audit_buf
);
2568 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
2570 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
2571 uid_t auid
, u32 sessionid
, u32 secid
)
2573 struct audit_buffer
*audit_buf
;
2575 audit_buf
= xfrm_audit_start("SPD-delete");
2576 if (audit_buf
== NULL
)
2578 xfrm_audit_helper_usrinfo(auid
, sessionid
, secid
, audit_buf
);
2579 audit_log_format(audit_buf
, " res=%u", result
);
2580 xfrm_audit_common_policyinfo(xp
, audit_buf
);
2581 audit_log_end(audit_buf
);
2583 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
2586 #ifdef CONFIG_XFRM_MIGRATE
2587 static int xfrm_migrate_selector_match(struct xfrm_selector
*sel_cmp
,
2588 struct xfrm_selector
*sel_tgt
)
2590 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
2591 if (sel_tgt
->family
== sel_cmp
->family
&&
2592 xfrm_addr_cmp(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
2593 sel_cmp
->family
) == 0 &&
2594 xfrm_addr_cmp(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
2595 sel_cmp
->family
) == 0 &&
2596 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
2597 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
2601 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
2608 static struct xfrm_policy
* xfrm_migrate_policy_find(struct xfrm_selector
*sel
,
2611 struct xfrm_policy
*pol
, *ret
= NULL
;
2612 struct hlist_node
*entry
;
2613 struct hlist_head
*chain
;
2616 read_lock_bh(&xfrm_policy_lock
);
2617 chain
= policy_hash_direct(&init_net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
2618 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2619 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2620 pol
->type
== type
) {
2622 priority
= ret
->priority
;
2626 chain
= &init_net
.xfrm
.policy_inexact
[dir
];
2627 hlist_for_each_entry(pol
, entry
, chain
, bydst
) {
2628 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
2629 pol
->type
== type
&&
2630 pol
->priority
< priority
) {
2639 read_unlock_bh(&xfrm_policy_lock
);
2644 static int migrate_tmpl_match(struct xfrm_migrate
*m
, struct xfrm_tmpl
*t
)
2648 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
2649 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
2651 case XFRM_MODE_TUNNEL
:
2652 case XFRM_MODE_BEET
:
2653 if (xfrm_addr_cmp(&t
->id
.daddr
, &m
->old_daddr
,
2654 m
->old_family
) == 0 &&
2655 xfrm_addr_cmp(&t
->saddr
, &m
->old_saddr
,
2656 m
->old_family
) == 0) {
2660 case XFRM_MODE_TRANSPORT
:
2661 /* in case of transport mode, template does not store
2662 any IP addresses, hence we just compare mode and
2673 /* update endpoint address(es) of template(s) */
2674 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
2675 struct xfrm_migrate
*m
, int num_migrate
)
2677 struct xfrm_migrate
*mp
;
2678 struct dst_entry
*dst
;
2681 write_lock_bh(&pol
->lock
);
2682 if (unlikely(pol
->walk
.dead
)) {
2683 /* target policy has been deleted */
2684 write_unlock_bh(&pol
->lock
);
2688 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
2689 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
2690 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
2693 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
2694 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
2696 /* update endpoints */
2697 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
2698 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
2699 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
2700 sizeof(pol
->xfrm_vec
[i
].saddr
));
2701 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
2703 while ((dst
= pol
->bundles
) != NULL
) {
2704 pol
->bundles
= dst
->next
;
2710 write_unlock_bh(&pol
->lock
);
2718 static int xfrm_migrate_check(struct xfrm_migrate
*m
, int num_migrate
)
2722 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
2725 for (i
= 0; i
< num_migrate
; i
++) {
2726 if ((xfrm_addr_cmp(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
2727 m
[i
].old_family
) == 0) &&
2728 (xfrm_addr_cmp(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
2729 m
[i
].old_family
) == 0))
2731 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
2732 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
2735 /* check if there is any duplicated entry */
2736 for (j
= i
+ 1; j
< num_migrate
; j
++) {
2737 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
2738 sizeof(m
[i
].old_daddr
)) &&
2739 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
2740 sizeof(m
[i
].old_saddr
)) &&
2741 m
[i
].proto
== m
[j
].proto
&&
2742 m
[i
].mode
== m
[j
].mode
&&
2743 m
[i
].reqid
== m
[j
].reqid
&&
2744 m
[i
].old_family
== m
[j
].old_family
)
2752 int xfrm_migrate(struct xfrm_selector
*sel
, u8 dir
, u8 type
,
2753 struct xfrm_migrate
*m
, int num_migrate
,
2754 struct xfrm_kmaddress
*k
)
2756 int i
, err
, nx_cur
= 0, nx_new
= 0;
2757 struct xfrm_policy
*pol
= NULL
;
2758 struct xfrm_state
*x
, *xc
;
2759 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
2760 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
2761 struct xfrm_migrate
*mp
;
2763 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
2766 /* Stage 1 - find policy */
2767 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
)) == NULL
) {
2772 /* Stage 2 - find and update state(s) */
2773 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
2774 if ((x
= xfrm_migrate_state_find(mp
))) {
2777 if ((xc
= xfrm_state_migrate(x
, mp
))) {
2787 /* Stage 3 - update policy */
2788 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
2791 /* Stage 4 - delete old state(s) */
2793 xfrm_states_put(x_cur
, nx_cur
);
2794 xfrm_states_delete(x_cur
, nx_cur
);
2797 /* Stage 5 - announce */
2798 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
2810 xfrm_states_put(x_cur
, nx_cur
);
2812 xfrm_states_delete(x_new
, nx_new
);
2816 EXPORT_SYMBOL(xfrm_migrate
);