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>
32 #ifdef CONFIG_XFRM_STATISTICS
36 #include "xfrm_hash.h"
38 #define XFRM_QUEUE_TMO_MIN ((unsigned)(HZ/10))
39 #define XFRM_QUEUE_TMO_MAX ((unsigned)(60*HZ))
40 #define XFRM_MAX_QUEUE_LEN 100
43 struct dst_entry
*dst_orig
;
47 static DEFINE_SPINLOCK(xfrm_policy_afinfo_lock
);
48 static struct xfrm_policy_afinfo __rcu
*xfrm_policy_afinfo
[NPROTO
]
51 static struct kmem_cache
*xfrm_dst_cache __read_mostly
;
53 static void xfrm_init_pmtu(struct dst_entry
*dst
);
54 static int stale_bundle(struct dst_entry
*dst
);
55 static int xfrm_bundle_ok(struct xfrm_dst
*xdst
);
56 static void xfrm_policy_queue_process(unsigned long arg
);
58 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
);
59 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
63 __xfrm4_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
)
65 const struct flowi4
*fl4
= &fl
->u
.ip4
;
67 return addr4_match(fl4
->daddr
, sel
->daddr
.a4
, sel
->prefixlen_d
) &&
68 addr4_match(fl4
->saddr
, sel
->saddr
.a4
, sel
->prefixlen_s
) &&
69 !((xfrm_flowi_dport(fl
, &fl4
->uli
) ^ sel
->dport
) & sel
->dport_mask
) &&
70 !((xfrm_flowi_sport(fl
, &fl4
->uli
) ^ sel
->sport
) & sel
->sport_mask
) &&
71 (fl4
->flowi4_proto
== sel
->proto
|| !sel
->proto
) &&
72 (fl4
->flowi4_oif
== sel
->ifindex
|| !sel
->ifindex
);
76 __xfrm6_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
)
78 const struct flowi6
*fl6
= &fl
->u
.ip6
;
80 return addr_match(&fl6
->daddr
, &sel
->daddr
, sel
->prefixlen_d
) &&
81 addr_match(&fl6
->saddr
, &sel
->saddr
, sel
->prefixlen_s
) &&
82 !((xfrm_flowi_dport(fl
, &fl6
->uli
) ^ sel
->dport
) & sel
->dport_mask
) &&
83 !((xfrm_flowi_sport(fl
, &fl6
->uli
) ^ sel
->sport
) & sel
->sport_mask
) &&
84 (fl6
->flowi6_proto
== sel
->proto
|| !sel
->proto
) &&
85 (fl6
->flowi6_oif
== sel
->ifindex
|| !sel
->ifindex
);
88 bool xfrm_selector_match(const struct xfrm_selector
*sel
, const struct flowi
*fl
,
89 unsigned short family
)
93 return __xfrm4_selector_match(sel
, fl
);
95 return __xfrm6_selector_match(sel
, fl
);
100 static struct xfrm_policy_afinfo
*xfrm_policy_get_afinfo(unsigned short family
)
102 struct xfrm_policy_afinfo
*afinfo
;
104 if (unlikely(family
>= NPROTO
))
107 afinfo
= rcu_dereference(xfrm_policy_afinfo
[family
]);
108 if (unlikely(!afinfo
))
113 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo
*afinfo
)
118 static inline struct dst_entry
*__xfrm_dst_lookup(struct net
*net
,
120 const xfrm_address_t
*saddr
,
121 const xfrm_address_t
*daddr
,
124 struct xfrm_policy_afinfo
*afinfo
;
125 struct dst_entry
*dst
;
127 afinfo
= xfrm_policy_get_afinfo(family
);
128 if (unlikely(afinfo
== NULL
))
129 return ERR_PTR(-EAFNOSUPPORT
);
131 dst
= afinfo
->dst_lookup(net
, tos
, oif
, saddr
, daddr
);
133 xfrm_policy_put_afinfo(afinfo
);
138 static inline struct dst_entry
*xfrm_dst_lookup(struct xfrm_state
*x
,
140 xfrm_address_t
*prev_saddr
,
141 xfrm_address_t
*prev_daddr
,
144 struct net
*net
= xs_net(x
);
145 xfrm_address_t
*saddr
= &x
->props
.saddr
;
146 xfrm_address_t
*daddr
= &x
->id
.daddr
;
147 struct dst_entry
*dst
;
149 if (x
->type
->flags
& XFRM_TYPE_LOCAL_COADDR
) {
153 if (x
->type
->flags
& XFRM_TYPE_REMOTE_COADDR
) {
158 dst
= __xfrm_dst_lookup(net
, tos
, oif
, saddr
, daddr
, family
);
161 if (prev_saddr
!= saddr
)
162 memcpy(prev_saddr
, saddr
, sizeof(*prev_saddr
));
163 if (prev_daddr
!= daddr
)
164 memcpy(prev_daddr
, daddr
, sizeof(*prev_daddr
));
170 static inline unsigned long make_jiffies(long secs
)
172 if (secs
>= (MAX_SCHEDULE_TIMEOUT
-1)/HZ
)
173 return MAX_SCHEDULE_TIMEOUT
-1;
178 static void xfrm_policy_timer(unsigned long data
)
180 struct xfrm_policy
*xp
= (struct xfrm_policy
*)data
;
181 unsigned long now
= get_seconds();
182 long next
= LONG_MAX
;
186 read_lock(&xp
->lock
);
188 if (unlikely(xp
->walk
.dead
))
191 dir
= xfrm_policy_id2dir(xp
->index
);
193 if (xp
->lft
.hard_add_expires_seconds
) {
194 long tmo
= xp
->lft
.hard_add_expires_seconds
+
195 xp
->curlft
.add_time
- now
;
201 if (xp
->lft
.hard_use_expires_seconds
) {
202 long tmo
= xp
->lft
.hard_use_expires_seconds
+
203 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
209 if (xp
->lft
.soft_add_expires_seconds
) {
210 long tmo
= xp
->lft
.soft_add_expires_seconds
+
211 xp
->curlft
.add_time
- now
;
214 tmo
= XFRM_KM_TIMEOUT
;
219 if (xp
->lft
.soft_use_expires_seconds
) {
220 long tmo
= xp
->lft
.soft_use_expires_seconds
+
221 (xp
->curlft
.use_time
? : xp
->curlft
.add_time
) - now
;
224 tmo
= XFRM_KM_TIMEOUT
;
231 km_policy_expired(xp
, dir
, 0, 0);
232 if (next
!= LONG_MAX
&&
233 !mod_timer(&xp
->timer
, jiffies
+ make_jiffies(next
)))
237 read_unlock(&xp
->lock
);
242 read_unlock(&xp
->lock
);
243 if (!xfrm_policy_delete(xp
, dir
))
244 km_policy_expired(xp
, dir
, 1, 0);
248 static struct flow_cache_object
*xfrm_policy_flo_get(struct flow_cache_object
*flo
)
250 struct xfrm_policy
*pol
= container_of(flo
, struct xfrm_policy
, flo
);
252 if (unlikely(pol
->walk
.dead
))
260 static int xfrm_policy_flo_check(struct flow_cache_object
*flo
)
262 struct xfrm_policy
*pol
= container_of(flo
, struct xfrm_policy
, flo
);
264 return !pol
->walk
.dead
;
267 static void xfrm_policy_flo_delete(struct flow_cache_object
*flo
)
269 xfrm_pol_put(container_of(flo
, struct xfrm_policy
, flo
));
272 static const struct flow_cache_ops xfrm_policy_fc_ops
= {
273 .get
= xfrm_policy_flo_get
,
274 .check
= xfrm_policy_flo_check
,
275 .delete = xfrm_policy_flo_delete
,
278 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
282 struct xfrm_policy
*xfrm_policy_alloc(struct net
*net
, gfp_t gfp
)
284 struct xfrm_policy
*policy
;
286 policy
= kzalloc(sizeof(struct xfrm_policy
), gfp
);
289 write_pnet(&policy
->xp_net
, net
);
290 INIT_LIST_HEAD(&policy
->walk
.all
);
291 INIT_HLIST_NODE(&policy
->bydst
);
292 INIT_HLIST_NODE(&policy
->byidx
);
293 rwlock_init(&policy
->lock
);
294 atomic_set(&policy
->refcnt
, 1);
295 skb_queue_head_init(&policy
->polq
.hold_queue
);
296 setup_timer(&policy
->timer
, xfrm_policy_timer
,
297 (unsigned long)policy
);
298 setup_timer(&policy
->polq
.hold_timer
, xfrm_policy_queue_process
,
299 (unsigned long)policy
);
300 policy
->flo
.ops
= &xfrm_policy_fc_ops
;
304 EXPORT_SYMBOL(xfrm_policy_alloc
);
306 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
308 void xfrm_policy_destroy(struct xfrm_policy
*policy
)
310 BUG_ON(!policy
->walk
.dead
);
312 if (del_timer(&policy
->timer
) || del_timer(&policy
->polq
.hold_timer
))
315 security_xfrm_policy_free(policy
->security
);
318 EXPORT_SYMBOL(xfrm_policy_destroy
);
320 /* Rule must be locked. Release descentant resources, announce
321 * entry dead. The rule must be unlinked from lists to the moment.
324 static void xfrm_policy_kill(struct xfrm_policy
*policy
)
326 policy
->walk
.dead
= 1;
328 atomic_inc(&policy
->genid
);
330 if (del_timer(&policy
->polq
.hold_timer
))
331 xfrm_pol_put(policy
);
332 skb_queue_purge(&policy
->polq
.hold_queue
);
334 if (del_timer(&policy
->timer
))
335 xfrm_pol_put(policy
);
337 xfrm_pol_put(policy
);
340 static unsigned int xfrm_policy_hashmax __read_mostly
= 1 * 1024 * 1024;
342 static inline unsigned int idx_hash(struct net
*net
, u32 index
)
344 return __idx_hash(index
, net
->xfrm
.policy_idx_hmask
);
347 /* calculate policy hash thresholds */
348 static void __get_hash_thresh(struct net
*net
,
349 unsigned short family
, int dir
,
350 u8
*dbits
, u8
*sbits
)
354 *dbits
= net
->xfrm
.policy_bydst
[dir
].dbits4
;
355 *sbits
= net
->xfrm
.policy_bydst
[dir
].sbits4
;
359 *dbits
= net
->xfrm
.policy_bydst
[dir
].dbits6
;
360 *sbits
= net
->xfrm
.policy_bydst
[dir
].sbits6
;
369 static struct hlist_head
*policy_hash_bysel(struct net
*net
,
370 const struct xfrm_selector
*sel
,
371 unsigned short family
, int dir
)
373 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
378 __get_hash_thresh(net
, family
, dir
, &dbits
, &sbits
);
379 hash
= __sel_hash(sel
, family
, hmask
, dbits
, sbits
);
381 return (hash
== hmask
+ 1 ?
382 &net
->xfrm
.policy_inexact
[dir
] :
383 net
->xfrm
.policy_bydst
[dir
].table
+ hash
);
386 static struct hlist_head
*policy_hash_direct(struct net
*net
,
387 const xfrm_address_t
*daddr
,
388 const xfrm_address_t
*saddr
,
389 unsigned short family
, int dir
)
391 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
396 __get_hash_thresh(net
, family
, dir
, &dbits
, &sbits
);
397 hash
= __addr_hash(daddr
, saddr
, family
, hmask
, dbits
, sbits
);
399 return net
->xfrm
.policy_bydst
[dir
].table
+ hash
;
402 static void xfrm_dst_hash_transfer(struct net
*net
,
403 struct hlist_head
*list
,
404 struct hlist_head
*ndsttable
,
405 unsigned int nhashmask
,
408 struct hlist_node
*tmp
, *entry0
= NULL
;
409 struct xfrm_policy
*pol
;
415 hlist_for_each_entry_safe(pol
, tmp
, list
, bydst
) {
418 __get_hash_thresh(net
, pol
->family
, dir
, &dbits
, &sbits
);
419 h
= __addr_hash(&pol
->selector
.daddr
, &pol
->selector
.saddr
,
420 pol
->family
, nhashmask
, dbits
, sbits
);
422 hlist_del(&pol
->bydst
);
423 hlist_add_head(&pol
->bydst
, ndsttable
+h
);
428 hlist_del(&pol
->bydst
);
429 hlist_add_behind(&pol
->bydst
, entry0
);
431 entry0
= &pol
->bydst
;
433 if (!hlist_empty(list
)) {
439 static void xfrm_idx_hash_transfer(struct hlist_head
*list
,
440 struct hlist_head
*nidxtable
,
441 unsigned int nhashmask
)
443 struct hlist_node
*tmp
;
444 struct xfrm_policy
*pol
;
446 hlist_for_each_entry_safe(pol
, tmp
, list
, byidx
) {
449 h
= __idx_hash(pol
->index
, nhashmask
);
450 hlist_add_head(&pol
->byidx
, nidxtable
+h
);
454 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask
)
456 return ((old_hmask
+ 1) << 1) - 1;
459 static void xfrm_bydst_resize(struct net
*net
, int dir
)
461 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
462 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
463 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
464 struct hlist_head
*odst
= net
->xfrm
.policy_bydst
[dir
].table
;
465 struct hlist_head
*ndst
= xfrm_hash_alloc(nsize
);
471 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
473 for (i
= hmask
; i
>= 0; i
--)
474 xfrm_dst_hash_transfer(net
, odst
+ i
, ndst
, nhashmask
, dir
);
476 net
->xfrm
.policy_bydst
[dir
].table
= ndst
;
477 net
->xfrm
.policy_bydst
[dir
].hmask
= nhashmask
;
479 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
481 xfrm_hash_free(odst
, (hmask
+ 1) * sizeof(struct hlist_head
));
484 static void xfrm_byidx_resize(struct net
*net
, int total
)
486 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
487 unsigned int nhashmask
= xfrm_new_hash_mask(hmask
);
488 unsigned int nsize
= (nhashmask
+ 1) * sizeof(struct hlist_head
);
489 struct hlist_head
*oidx
= net
->xfrm
.policy_byidx
;
490 struct hlist_head
*nidx
= xfrm_hash_alloc(nsize
);
496 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
498 for (i
= hmask
; i
>= 0; i
--)
499 xfrm_idx_hash_transfer(oidx
+ i
, nidx
, nhashmask
);
501 net
->xfrm
.policy_byidx
= nidx
;
502 net
->xfrm
.policy_idx_hmask
= nhashmask
;
504 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
506 xfrm_hash_free(oidx
, (hmask
+ 1) * sizeof(struct hlist_head
));
509 static inline int xfrm_bydst_should_resize(struct net
*net
, int dir
, int *total
)
511 unsigned int cnt
= net
->xfrm
.policy_count
[dir
];
512 unsigned int hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
517 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
524 static inline int xfrm_byidx_should_resize(struct net
*net
, int total
)
526 unsigned int hmask
= net
->xfrm
.policy_idx_hmask
;
528 if ((hmask
+ 1) < xfrm_policy_hashmax
&&
535 void xfrm_spd_getinfo(struct net
*net
, struct xfrmk_spdinfo
*si
)
537 read_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
538 si
->incnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
];
539 si
->outcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
];
540 si
->fwdcnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
];
541 si
->inscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_IN
+XFRM_POLICY_MAX
];
542 si
->outscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_OUT
+XFRM_POLICY_MAX
];
543 si
->fwdscnt
= net
->xfrm
.policy_count
[XFRM_POLICY_FWD
+XFRM_POLICY_MAX
];
544 si
->spdhcnt
= net
->xfrm
.policy_idx_hmask
;
545 si
->spdhmcnt
= xfrm_policy_hashmax
;
546 read_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
548 EXPORT_SYMBOL(xfrm_spd_getinfo
);
550 static DEFINE_MUTEX(hash_resize_mutex
);
551 static void xfrm_hash_resize(struct work_struct
*work
)
553 struct net
*net
= container_of(work
, struct net
, xfrm
.policy_hash_work
);
556 mutex_lock(&hash_resize_mutex
);
559 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
560 if (xfrm_bydst_should_resize(net
, dir
, &total
))
561 xfrm_bydst_resize(net
, dir
);
563 if (xfrm_byidx_should_resize(net
, total
))
564 xfrm_byidx_resize(net
, total
);
566 mutex_unlock(&hash_resize_mutex
);
569 static void xfrm_hash_rebuild(struct work_struct
*work
)
571 struct net
*net
= container_of(work
, struct net
,
572 xfrm
.policy_hthresh
.work
);
574 struct xfrm_policy
*pol
;
575 struct xfrm_policy
*policy
;
576 struct hlist_head
*chain
;
577 struct hlist_head
*odst
;
578 struct hlist_node
*newpos
;
582 u8 lbits4
, rbits4
, lbits6
, rbits6
;
584 mutex_lock(&hash_resize_mutex
);
586 /* read selector prefixlen thresholds */
588 seq
= read_seqbegin(&net
->xfrm
.policy_hthresh
.lock
);
590 lbits4
= net
->xfrm
.policy_hthresh
.lbits4
;
591 rbits4
= net
->xfrm
.policy_hthresh
.rbits4
;
592 lbits6
= net
->xfrm
.policy_hthresh
.lbits6
;
593 rbits6
= net
->xfrm
.policy_hthresh
.rbits6
;
594 } while (read_seqretry(&net
->xfrm
.policy_hthresh
.lock
, seq
));
596 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
598 /* reset the bydst and inexact table in all directions */
599 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
600 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
601 hmask
= net
->xfrm
.policy_bydst
[dir
].hmask
;
602 odst
= net
->xfrm
.policy_bydst
[dir
].table
;
603 for (i
= hmask
; i
>= 0; i
--)
604 INIT_HLIST_HEAD(odst
+ i
);
605 if ((dir
& XFRM_POLICY_MASK
) == XFRM_POLICY_OUT
) {
606 /* dir out => dst = remote, src = local */
607 net
->xfrm
.policy_bydst
[dir
].dbits4
= rbits4
;
608 net
->xfrm
.policy_bydst
[dir
].sbits4
= lbits4
;
609 net
->xfrm
.policy_bydst
[dir
].dbits6
= rbits6
;
610 net
->xfrm
.policy_bydst
[dir
].sbits6
= lbits6
;
612 /* dir in/fwd => dst = local, src = remote */
613 net
->xfrm
.policy_bydst
[dir
].dbits4
= lbits4
;
614 net
->xfrm
.policy_bydst
[dir
].sbits4
= rbits4
;
615 net
->xfrm
.policy_bydst
[dir
].dbits6
= lbits6
;
616 net
->xfrm
.policy_bydst
[dir
].sbits6
= rbits6
;
620 /* re-insert all policies by order of creation */
621 list_for_each_entry_reverse(policy
, &net
->xfrm
.policy_all
, walk
.all
) {
623 chain
= policy_hash_bysel(net
, &policy
->selector
,
625 xfrm_policy_id2dir(policy
->index
));
626 hlist_for_each_entry(pol
, chain
, bydst
) {
627 if (policy
->priority
>= pol
->priority
)
628 newpos
= &pol
->bydst
;
633 hlist_add_behind(&policy
->bydst
, newpos
);
635 hlist_add_head(&policy
->bydst
, chain
);
638 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
640 mutex_unlock(&hash_resize_mutex
);
643 void xfrm_policy_hash_rebuild(struct net
*net
)
645 schedule_work(&net
->xfrm
.policy_hthresh
.work
);
647 EXPORT_SYMBOL(xfrm_policy_hash_rebuild
);
649 /* Generate new index... KAME seems to generate them ordered by cost
650 * of an absolute inpredictability of ordering of rules. This will not pass. */
651 static u32
xfrm_gen_index(struct net
*net
, int dir
, u32 index
)
653 static u32 idx_generator
;
656 struct hlist_head
*list
;
657 struct xfrm_policy
*p
;
662 idx
= (idx_generator
| dir
);
671 list
= net
->xfrm
.policy_byidx
+ idx_hash(net
, idx
);
673 hlist_for_each_entry(p
, list
, byidx
) {
674 if (p
->index
== idx
) {
684 static inline int selector_cmp(struct xfrm_selector
*s1
, struct xfrm_selector
*s2
)
686 u32
*p1
= (u32
*) s1
;
687 u32
*p2
= (u32
*) s2
;
688 int len
= sizeof(struct xfrm_selector
) / sizeof(u32
);
691 for (i
= 0; i
< len
; i
++) {
699 static void xfrm_policy_requeue(struct xfrm_policy
*old
,
700 struct xfrm_policy
*new)
702 struct xfrm_policy_queue
*pq
= &old
->polq
;
703 struct sk_buff_head list
;
705 if (skb_queue_empty(&pq
->hold_queue
))
708 __skb_queue_head_init(&list
);
710 spin_lock_bh(&pq
->hold_queue
.lock
);
711 skb_queue_splice_init(&pq
->hold_queue
, &list
);
712 if (del_timer(&pq
->hold_timer
))
714 spin_unlock_bh(&pq
->hold_queue
.lock
);
718 spin_lock_bh(&pq
->hold_queue
.lock
);
719 skb_queue_splice(&list
, &pq
->hold_queue
);
720 pq
->timeout
= XFRM_QUEUE_TMO_MIN
;
721 if (!mod_timer(&pq
->hold_timer
, jiffies
))
723 spin_unlock_bh(&pq
->hold_queue
.lock
);
726 static bool xfrm_policy_mark_match(struct xfrm_policy
*policy
,
727 struct xfrm_policy
*pol
)
729 u32 mark
= policy
->mark
.v
& policy
->mark
.m
;
731 if (policy
->mark
.v
== pol
->mark
.v
&& policy
->mark
.m
== pol
->mark
.m
)
734 if ((mark
& pol
->mark
.m
) == pol
->mark
.v
&&
735 policy
->priority
== pol
->priority
)
741 int xfrm_policy_insert(int dir
, struct xfrm_policy
*policy
, int excl
)
743 struct net
*net
= xp_net(policy
);
744 struct xfrm_policy
*pol
;
745 struct xfrm_policy
*delpol
;
746 struct hlist_head
*chain
;
747 struct hlist_node
*newpos
;
749 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
750 chain
= policy_hash_bysel(net
, &policy
->selector
, policy
->family
, dir
);
753 hlist_for_each_entry(pol
, chain
, bydst
) {
754 if (pol
->type
== policy
->type
&&
755 !selector_cmp(&pol
->selector
, &policy
->selector
) &&
756 xfrm_policy_mark_match(policy
, pol
) &&
757 xfrm_sec_ctx_match(pol
->security
, policy
->security
) &&
760 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
764 if (policy
->priority
> pol
->priority
)
766 } else if (policy
->priority
>= pol
->priority
) {
767 newpos
= &pol
->bydst
;
774 hlist_add_behind(&policy
->bydst
, newpos
);
776 hlist_add_head(&policy
->bydst
, chain
);
777 __xfrm_policy_link(policy
, dir
);
778 atomic_inc(&net
->xfrm
.flow_cache_genid
);
780 /* After previous checking, family can either be AF_INET or AF_INET6 */
781 if (policy
->family
== AF_INET
)
782 rt_genid_bump_ipv4(net
);
784 rt_genid_bump_ipv6(net
);
787 xfrm_policy_requeue(delpol
, policy
);
788 __xfrm_policy_unlink(delpol
, dir
);
790 policy
->index
= delpol
? delpol
->index
: xfrm_gen_index(net
, dir
, policy
->index
);
791 hlist_add_head(&policy
->byidx
, net
->xfrm
.policy_byidx
+idx_hash(net
, policy
->index
));
792 policy
->curlft
.add_time
= get_seconds();
793 policy
->curlft
.use_time
= 0;
794 if (!mod_timer(&policy
->timer
, jiffies
+ HZ
))
795 xfrm_pol_hold(policy
);
796 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
799 xfrm_policy_kill(delpol
);
800 else if (xfrm_bydst_should_resize(net
, dir
, NULL
))
801 schedule_work(&net
->xfrm
.policy_hash_work
);
805 EXPORT_SYMBOL(xfrm_policy_insert
);
807 struct xfrm_policy
*xfrm_policy_bysel_ctx(struct net
*net
, u32 mark
, u8 type
,
808 int dir
, struct xfrm_selector
*sel
,
809 struct xfrm_sec_ctx
*ctx
, int delete,
812 struct xfrm_policy
*pol
, *ret
;
813 struct hlist_head
*chain
;
816 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
817 chain
= policy_hash_bysel(net
, sel
, sel
->family
, dir
);
819 hlist_for_each_entry(pol
, chain
, bydst
) {
820 if (pol
->type
== type
&&
821 (mark
& pol
->mark
.m
) == pol
->mark
.v
&&
822 !selector_cmp(sel
, &pol
->selector
) &&
823 xfrm_sec_ctx_match(ctx
, pol
->security
)) {
826 *err
= security_xfrm_policy_delete(
829 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
832 __xfrm_policy_unlink(pol
, dir
);
838 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
841 xfrm_policy_kill(ret
);
844 EXPORT_SYMBOL(xfrm_policy_bysel_ctx
);
846 struct xfrm_policy
*xfrm_policy_byid(struct net
*net
, u32 mark
, u8 type
,
847 int dir
, u32 id
, int delete, int *err
)
849 struct xfrm_policy
*pol
, *ret
;
850 struct hlist_head
*chain
;
853 if (xfrm_policy_id2dir(id
) != dir
)
857 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
858 chain
= net
->xfrm
.policy_byidx
+ idx_hash(net
, id
);
860 hlist_for_each_entry(pol
, chain
, byidx
) {
861 if (pol
->type
== type
&& pol
->index
== id
&&
862 (mark
& pol
->mark
.m
) == pol
->mark
.v
) {
865 *err
= security_xfrm_policy_delete(
868 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
871 __xfrm_policy_unlink(pol
, dir
);
877 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
880 xfrm_policy_kill(ret
);
883 EXPORT_SYMBOL(xfrm_policy_byid
);
885 #ifdef CONFIG_SECURITY_NETWORK_XFRM
887 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, bool task_valid
)
891 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
892 struct xfrm_policy
*pol
;
895 hlist_for_each_entry(pol
,
896 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
897 if (pol
->type
!= type
)
899 err
= security_xfrm_policy_delete(pol
->security
);
901 xfrm_audit_policy_delete(pol
, 0, task_valid
);
905 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
906 hlist_for_each_entry(pol
,
907 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
909 if (pol
->type
!= type
)
911 err
= security_xfrm_policy_delete(
914 xfrm_audit_policy_delete(pol
, 0,
925 xfrm_policy_flush_secctx_check(struct net
*net
, u8 type
, bool task_valid
)
931 int xfrm_policy_flush(struct net
*net
, u8 type
, bool task_valid
)
933 int dir
, err
= 0, cnt
= 0;
935 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
937 err
= xfrm_policy_flush_secctx_check(net
, type
, task_valid
);
941 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
942 struct xfrm_policy
*pol
;
946 hlist_for_each_entry(pol
,
947 &net
->xfrm
.policy_inexact
[dir
], bydst
) {
948 if (pol
->type
!= type
)
950 __xfrm_policy_unlink(pol
, dir
);
951 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
954 xfrm_audit_policy_delete(pol
, 1, task_valid
);
956 xfrm_policy_kill(pol
);
958 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
962 for (i
= net
->xfrm
.policy_bydst
[dir
].hmask
; i
>= 0; i
--) {
964 hlist_for_each_entry(pol
,
965 net
->xfrm
.policy_bydst
[dir
].table
+ i
,
967 if (pol
->type
!= type
)
969 __xfrm_policy_unlink(pol
, dir
);
970 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
973 xfrm_audit_policy_delete(pol
, 1, task_valid
);
974 xfrm_policy_kill(pol
);
976 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
985 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
988 EXPORT_SYMBOL(xfrm_policy_flush
);
990 int xfrm_policy_walk(struct net
*net
, struct xfrm_policy_walk
*walk
,
991 int (*func
)(struct xfrm_policy
*, int, int, void*),
994 struct xfrm_policy
*pol
;
995 struct xfrm_policy_walk_entry
*x
;
998 if (walk
->type
>= XFRM_POLICY_TYPE_MAX
&&
999 walk
->type
!= XFRM_POLICY_TYPE_ANY
)
1002 if (list_empty(&walk
->walk
.all
) && walk
->seq
!= 0)
1005 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1006 if (list_empty(&walk
->walk
.all
))
1007 x
= list_first_entry(&net
->xfrm
.policy_all
, struct xfrm_policy_walk_entry
, all
);
1009 x
= list_first_entry(&walk
->walk
.all
,
1010 struct xfrm_policy_walk_entry
, all
);
1012 list_for_each_entry_from(x
, &net
->xfrm
.policy_all
, all
) {
1015 pol
= container_of(x
, struct xfrm_policy
, walk
);
1016 if (walk
->type
!= XFRM_POLICY_TYPE_ANY
&&
1017 walk
->type
!= pol
->type
)
1019 error
= func(pol
, xfrm_policy_id2dir(pol
->index
),
1022 list_move_tail(&walk
->walk
.all
, &x
->all
);
1027 if (walk
->seq
== 0) {
1031 list_del_init(&walk
->walk
.all
);
1033 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1036 EXPORT_SYMBOL(xfrm_policy_walk
);
1038 void xfrm_policy_walk_init(struct xfrm_policy_walk
*walk
, u8 type
)
1040 INIT_LIST_HEAD(&walk
->walk
.all
);
1041 walk
->walk
.dead
= 1;
1045 EXPORT_SYMBOL(xfrm_policy_walk_init
);
1047 void xfrm_policy_walk_done(struct xfrm_policy_walk
*walk
, struct net
*net
)
1049 if (list_empty(&walk
->walk
.all
))
1052 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
); /*FIXME where is net? */
1053 list_del(&walk
->walk
.all
);
1054 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1056 EXPORT_SYMBOL(xfrm_policy_walk_done
);
1059 * Find policy to apply to this flow.
1061 * Returns 0 if policy found, else an -errno.
1063 static int xfrm_policy_match(const struct xfrm_policy
*pol
,
1064 const struct flowi
*fl
,
1065 u8 type
, u16 family
, int dir
)
1067 const struct xfrm_selector
*sel
= &pol
->selector
;
1071 if (pol
->family
!= family
||
1072 (fl
->flowi_mark
& pol
->mark
.m
) != pol
->mark
.v
||
1076 match
= xfrm_selector_match(sel
, fl
, family
);
1078 ret
= security_xfrm_policy_lookup(pol
->security
, fl
->flowi_secid
,
1084 static struct xfrm_policy
*xfrm_policy_lookup_bytype(struct net
*net
, u8 type
,
1085 const struct flowi
*fl
,
1089 struct xfrm_policy
*pol
, *ret
;
1090 const xfrm_address_t
*daddr
, *saddr
;
1091 struct hlist_head
*chain
;
1094 daddr
= xfrm_flowi_daddr(fl
, family
);
1095 saddr
= xfrm_flowi_saddr(fl
, family
);
1096 if (unlikely(!daddr
|| !saddr
))
1099 read_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1100 chain
= policy_hash_direct(net
, daddr
, saddr
, family
, dir
);
1102 hlist_for_each_entry(pol
, chain
, bydst
) {
1103 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
1113 priority
= ret
->priority
;
1117 chain
= &net
->xfrm
.policy_inexact
[dir
];
1118 hlist_for_each_entry(pol
, chain
, bydst
) {
1119 if ((pol
->priority
>= priority
) && ret
)
1122 err
= xfrm_policy_match(pol
, fl
, type
, family
, dir
);
1138 read_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1143 static struct xfrm_policy
*
1144 __xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
)
1146 #ifdef CONFIG_XFRM_SUB_POLICY
1147 struct xfrm_policy
*pol
;
1149 pol
= xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_SUB
, fl
, family
, dir
);
1153 return xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
, fl
, family
, dir
);
1156 static int flow_to_policy_dir(int dir
)
1158 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1159 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1160 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1166 return XFRM_POLICY_IN
;
1168 return XFRM_POLICY_OUT
;
1170 return XFRM_POLICY_FWD
;
1174 static struct flow_cache_object
*
1175 xfrm_policy_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
,
1176 u8 dir
, struct flow_cache_object
*old_obj
, void *ctx
)
1178 struct xfrm_policy
*pol
;
1181 xfrm_pol_put(container_of(old_obj
, struct xfrm_policy
, flo
));
1183 pol
= __xfrm_policy_lookup(net
, fl
, family
, flow_to_policy_dir(dir
));
1184 if (IS_ERR_OR_NULL(pol
))
1185 return ERR_CAST(pol
);
1187 /* Resolver returns two references:
1188 * one for cache and one for caller of flow_cache_lookup() */
1194 static inline int policy_to_flow_dir(int dir
)
1196 if (XFRM_POLICY_IN
== FLOW_DIR_IN
&&
1197 XFRM_POLICY_OUT
== FLOW_DIR_OUT
&&
1198 XFRM_POLICY_FWD
== FLOW_DIR_FWD
)
1202 case XFRM_POLICY_IN
:
1204 case XFRM_POLICY_OUT
:
1205 return FLOW_DIR_OUT
;
1206 case XFRM_POLICY_FWD
:
1207 return FLOW_DIR_FWD
;
1211 static struct xfrm_policy
*xfrm_sk_policy_lookup(struct sock
*sk
, int dir
,
1212 const struct flowi
*fl
)
1214 struct xfrm_policy
*pol
;
1215 struct net
*net
= sock_net(sk
);
1217 read_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1218 if ((pol
= sk
->sk_policy
[dir
]) != NULL
) {
1219 bool match
= xfrm_selector_match(&pol
->selector
, fl
,
1224 if ((sk
->sk_mark
& pol
->mark
.m
) != pol
->mark
.v
) {
1228 err
= security_xfrm_policy_lookup(pol
->security
,
1230 policy_to_flow_dir(dir
));
1233 else if (err
== -ESRCH
)
1241 read_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1245 static void __xfrm_policy_link(struct xfrm_policy
*pol
, int dir
)
1247 struct net
*net
= xp_net(pol
);
1249 list_add(&pol
->walk
.all
, &net
->xfrm
.policy_all
);
1250 net
->xfrm
.policy_count
[dir
]++;
1254 static struct xfrm_policy
*__xfrm_policy_unlink(struct xfrm_policy
*pol
,
1257 struct net
*net
= xp_net(pol
);
1259 if (list_empty(&pol
->walk
.all
))
1262 /* Socket policies are not hashed. */
1263 if (!hlist_unhashed(&pol
->bydst
)) {
1264 hlist_del(&pol
->bydst
);
1265 hlist_del(&pol
->byidx
);
1268 list_del_init(&pol
->walk
.all
);
1269 net
->xfrm
.policy_count
[dir
]--;
1274 static void xfrm_sk_policy_link(struct xfrm_policy
*pol
, int dir
)
1276 __xfrm_policy_link(pol
, XFRM_POLICY_MAX
+ dir
);
1279 static void xfrm_sk_policy_unlink(struct xfrm_policy
*pol
, int dir
)
1281 __xfrm_policy_unlink(pol
, XFRM_POLICY_MAX
+ dir
);
1284 int xfrm_policy_delete(struct xfrm_policy
*pol
, int dir
)
1286 struct net
*net
= xp_net(pol
);
1288 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1289 pol
= __xfrm_policy_unlink(pol
, dir
);
1290 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1292 xfrm_policy_kill(pol
);
1297 EXPORT_SYMBOL(xfrm_policy_delete
);
1299 int xfrm_sk_policy_insert(struct sock
*sk
, int dir
, struct xfrm_policy
*pol
)
1301 struct net
*net
= xp_net(pol
);
1302 struct xfrm_policy
*old_pol
;
1304 #ifdef CONFIG_XFRM_SUB_POLICY
1305 if (pol
&& pol
->type
!= XFRM_POLICY_TYPE_MAIN
)
1309 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1310 old_pol
= sk
->sk_policy
[dir
];
1311 sk
->sk_policy
[dir
] = pol
;
1313 pol
->curlft
.add_time
= get_seconds();
1314 pol
->index
= xfrm_gen_index(net
, XFRM_POLICY_MAX
+dir
, 0);
1315 xfrm_sk_policy_link(pol
, dir
);
1319 xfrm_policy_requeue(old_pol
, pol
);
1321 /* Unlinking succeeds always. This is the only function
1322 * allowed to delete or replace socket policy.
1324 xfrm_sk_policy_unlink(old_pol
, dir
);
1326 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1329 xfrm_policy_kill(old_pol
);
1334 static struct xfrm_policy
*clone_policy(const struct xfrm_policy
*old
, int dir
)
1336 struct xfrm_policy
*newp
= xfrm_policy_alloc(xp_net(old
), GFP_ATOMIC
);
1337 struct net
*net
= xp_net(old
);
1340 newp
->selector
= old
->selector
;
1341 if (security_xfrm_policy_clone(old
->security
,
1344 return NULL
; /* ENOMEM */
1346 newp
->lft
= old
->lft
;
1347 newp
->curlft
= old
->curlft
;
1348 newp
->mark
= old
->mark
;
1349 newp
->action
= old
->action
;
1350 newp
->flags
= old
->flags
;
1351 newp
->xfrm_nr
= old
->xfrm_nr
;
1352 newp
->index
= old
->index
;
1353 newp
->type
= old
->type
;
1354 memcpy(newp
->xfrm_vec
, old
->xfrm_vec
,
1355 newp
->xfrm_nr
*sizeof(struct xfrm_tmpl
));
1356 write_lock_bh(&net
->xfrm
.xfrm_policy_lock
);
1357 xfrm_sk_policy_link(newp
, dir
);
1358 write_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
1364 int __xfrm_sk_clone_policy(struct sock
*sk
)
1366 struct xfrm_policy
*p0
= sk
->sk_policy
[0],
1367 *p1
= sk
->sk_policy
[1];
1369 sk
->sk_policy
[0] = sk
->sk_policy
[1] = NULL
;
1370 if (p0
&& (sk
->sk_policy
[0] = clone_policy(p0
, 0)) == NULL
)
1372 if (p1
&& (sk
->sk_policy
[1] = clone_policy(p1
, 1)) == NULL
)
1378 xfrm_get_saddr(struct net
*net
, int oif
, xfrm_address_t
*local
,
1379 xfrm_address_t
*remote
, unsigned short family
)
1382 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1384 if (unlikely(afinfo
== NULL
))
1386 err
= afinfo
->get_saddr(net
, oif
, local
, remote
);
1387 xfrm_policy_put_afinfo(afinfo
);
1391 /* Resolve list of templates for the flow, given policy. */
1394 xfrm_tmpl_resolve_one(struct xfrm_policy
*policy
, const struct flowi
*fl
,
1395 struct xfrm_state
**xfrm
, unsigned short family
)
1397 struct net
*net
= xp_net(policy
);
1400 xfrm_address_t
*daddr
= xfrm_flowi_daddr(fl
, family
);
1401 xfrm_address_t
*saddr
= xfrm_flowi_saddr(fl
, family
);
1404 for (nx
= 0, i
= 0; i
< policy
->xfrm_nr
; i
++) {
1405 struct xfrm_state
*x
;
1406 xfrm_address_t
*remote
= daddr
;
1407 xfrm_address_t
*local
= saddr
;
1408 struct xfrm_tmpl
*tmpl
= &policy
->xfrm_vec
[i
];
1410 if (tmpl
->mode
== XFRM_MODE_TUNNEL
||
1411 tmpl
->mode
== XFRM_MODE_BEET
) {
1412 remote
= &tmpl
->id
.daddr
;
1413 local
= &tmpl
->saddr
;
1414 if (xfrm_addr_any(local
, tmpl
->encap_family
)) {
1415 error
= xfrm_get_saddr(net
, fl
->flowi_oif
,
1417 tmpl
->encap_family
);
1424 x
= xfrm_state_find(remote
, local
, fl
, tmpl
, policy
, &error
, family
);
1426 if (x
&& x
->km
.state
== XFRM_STATE_VALID
) {
1433 error
= (x
->km
.state
== XFRM_STATE_ERROR
?
1436 } else if (error
== -ESRCH
) {
1440 if (!tmpl
->optional
)
1446 for (nx
--; nx
>= 0; nx
--)
1447 xfrm_state_put(xfrm
[nx
]);
1452 xfrm_tmpl_resolve(struct xfrm_policy
**pols
, int npols
, const struct flowi
*fl
,
1453 struct xfrm_state
**xfrm
, unsigned short family
)
1455 struct xfrm_state
*tp
[XFRM_MAX_DEPTH
];
1456 struct xfrm_state
**tpp
= (npols
> 1) ? tp
: xfrm
;
1462 for (i
= 0; i
< npols
; i
++) {
1463 if (cnx
+ pols
[i
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
1468 ret
= xfrm_tmpl_resolve_one(pols
[i
], fl
, &tpp
[cnx
], family
);
1476 /* found states are sorted for outbound processing */
1478 xfrm_state_sort(xfrm
, tpp
, cnx
, family
);
1483 for (cnx
--; cnx
>= 0; cnx
--)
1484 xfrm_state_put(tpp
[cnx
]);
1489 /* Check that the bundle accepts the flow and its components are
1493 static inline int xfrm_get_tos(const struct flowi
*fl
, int family
)
1495 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1501 tos
= afinfo
->get_tos(fl
);
1503 xfrm_policy_put_afinfo(afinfo
);
1508 static struct flow_cache_object
*xfrm_bundle_flo_get(struct flow_cache_object
*flo
)
1510 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1511 struct dst_entry
*dst
= &xdst
->u
.dst
;
1513 if (xdst
->route
== NULL
) {
1514 /* Dummy bundle - if it has xfrms we were not
1515 * able to build bundle as template resolution failed.
1516 * It means we need to try again resolving. */
1517 if (xdst
->num_xfrms
> 0)
1519 } else if (dst
->flags
& DST_XFRM_QUEUE
) {
1523 if (stale_bundle(dst
))
1531 static int xfrm_bundle_flo_check(struct flow_cache_object
*flo
)
1533 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1534 struct dst_entry
*dst
= &xdst
->u
.dst
;
1538 if (stale_bundle(dst
))
1544 static void xfrm_bundle_flo_delete(struct flow_cache_object
*flo
)
1546 struct xfrm_dst
*xdst
= container_of(flo
, struct xfrm_dst
, flo
);
1547 struct dst_entry
*dst
= &xdst
->u
.dst
;
1552 static const struct flow_cache_ops xfrm_bundle_fc_ops
= {
1553 .get
= xfrm_bundle_flo_get
,
1554 .check
= xfrm_bundle_flo_check
,
1555 .delete = xfrm_bundle_flo_delete
,
1558 static inline struct xfrm_dst
*xfrm_alloc_dst(struct net
*net
, int family
)
1560 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
1561 struct dst_ops
*dst_ops
;
1562 struct xfrm_dst
*xdst
;
1565 return ERR_PTR(-EINVAL
);
1569 dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
1571 #if IS_ENABLED(CONFIG_IPV6)
1573 dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
1579 xdst
= dst_alloc(dst_ops
, NULL
, 0, DST_OBSOLETE_NONE
, 0);
1582 struct dst_entry
*dst
= &xdst
->u
.dst
;
1584 memset(dst
+ 1, 0, sizeof(*xdst
) - sizeof(*dst
));
1585 xdst
->flo
.ops
= &xfrm_bundle_fc_ops
;
1586 if (afinfo
->init_dst
)
1587 afinfo
->init_dst(net
, xdst
);
1589 xdst
= ERR_PTR(-ENOBUFS
);
1591 xfrm_policy_put_afinfo(afinfo
);
1596 static inline int xfrm_init_path(struct xfrm_dst
*path
, struct dst_entry
*dst
,
1599 struct xfrm_policy_afinfo
*afinfo
=
1600 xfrm_policy_get_afinfo(dst
->ops
->family
);
1606 err
= afinfo
->init_path(path
, dst
, nfheader_len
);
1608 xfrm_policy_put_afinfo(afinfo
);
1613 static inline int xfrm_fill_dst(struct xfrm_dst
*xdst
, struct net_device
*dev
,
1614 const struct flowi
*fl
)
1616 struct xfrm_policy_afinfo
*afinfo
=
1617 xfrm_policy_get_afinfo(xdst
->u
.dst
.ops
->family
);
1623 err
= afinfo
->fill_dst(xdst
, dev
, fl
);
1625 xfrm_policy_put_afinfo(afinfo
);
1631 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1632 * all the metrics... Shortly, bundle a bundle.
1635 static struct dst_entry
*xfrm_bundle_create(struct xfrm_policy
*policy
,
1636 struct xfrm_state
**xfrm
, int nx
,
1637 const struct flowi
*fl
,
1638 struct dst_entry
*dst
)
1640 struct net
*net
= xp_net(policy
);
1641 unsigned long now
= jiffies
;
1642 struct net_device
*dev
;
1643 struct xfrm_mode
*inner_mode
;
1644 struct dst_entry
*dst_prev
= NULL
;
1645 struct dst_entry
*dst0
= NULL
;
1649 int nfheader_len
= 0;
1650 int trailer_len
= 0;
1652 int family
= policy
->selector
.family
;
1653 xfrm_address_t saddr
, daddr
;
1655 xfrm_flowi_addr_get(fl
, &saddr
, &daddr
, family
);
1657 tos
= xfrm_get_tos(fl
, family
);
1664 for (; i
< nx
; i
++) {
1665 struct xfrm_dst
*xdst
= xfrm_alloc_dst(net
, family
);
1666 struct dst_entry
*dst1
= &xdst
->u
.dst
;
1668 err
= PTR_ERR(xdst
);
1674 if (xfrm
[i
]->sel
.family
== AF_UNSPEC
) {
1675 inner_mode
= xfrm_ip2inner_mode(xfrm
[i
],
1676 xfrm_af2proto(family
));
1678 err
= -EAFNOSUPPORT
;
1683 inner_mode
= xfrm
[i
]->inner_mode
;
1688 dst_prev
->child
= dst_clone(dst1
);
1689 dst1
->flags
|= DST_NOHASH
;
1693 dst_copy_metrics(dst1
, dst
);
1695 if (xfrm
[i
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
1696 family
= xfrm
[i
]->props
.family
;
1697 dst
= xfrm_dst_lookup(xfrm
[i
], tos
, fl
->flowi_oif
,
1698 &saddr
, &daddr
, family
);
1705 dst1
->xfrm
= xfrm
[i
];
1706 xdst
->xfrm_genid
= xfrm
[i
]->genid
;
1708 dst1
->obsolete
= DST_OBSOLETE_FORCE_CHK
;
1709 dst1
->flags
|= DST_HOST
;
1710 dst1
->lastuse
= now
;
1712 dst1
->input
= dst_discard
;
1713 dst1
->output
= inner_mode
->afinfo
->output
;
1715 dst1
->next
= dst_prev
;
1718 header_len
+= xfrm
[i
]->props
.header_len
;
1719 if (xfrm
[i
]->type
->flags
& XFRM_TYPE_NON_FRAGMENT
)
1720 nfheader_len
+= xfrm
[i
]->props
.header_len
;
1721 trailer_len
+= xfrm
[i
]->props
.trailer_len
;
1724 dst_prev
->child
= dst
;
1732 xfrm_init_path((struct xfrm_dst
*)dst0
, dst
, nfheader_len
);
1733 xfrm_init_pmtu(dst_prev
);
1735 for (dst_prev
= dst0
; dst_prev
!= dst
; dst_prev
= dst_prev
->child
) {
1736 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst_prev
;
1738 err
= xfrm_fill_dst(xdst
, dev
, fl
);
1742 dst_prev
->header_len
= header_len
;
1743 dst_prev
->trailer_len
= trailer_len
;
1744 header_len
-= xdst
->u
.dst
.xfrm
->props
.header_len
;
1745 trailer_len
-= xdst
->u
.dst
.xfrm
->props
.trailer_len
;
1753 xfrm_state_put(xfrm
[i
]);
1757 dst0
= ERR_PTR(err
);
1761 #ifdef CONFIG_XFRM_SUB_POLICY
1762 static int xfrm_dst_alloc_copy(void **target
, const void *src
, int size
)
1765 *target
= kmalloc(size
, GFP_ATOMIC
);
1770 memcpy(*target
, src
, size
);
1775 static int xfrm_dst_update_parent(struct dst_entry
*dst
,
1776 const struct xfrm_selector
*sel
)
1778 #ifdef CONFIG_XFRM_SUB_POLICY
1779 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1780 return xfrm_dst_alloc_copy((void **)&(xdst
->partner
),
1787 static int xfrm_dst_update_origin(struct dst_entry
*dst
,
1788 const struct flowi
*fl
)
1790 #ifdef CONFIG_XFRM_SUB_POLICY
1791 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
1792 return xfrm_dst_alloc_copy((void **)&(xdst
->origin
), fl
, sizeof(*fl
));
1798 static int xfrm_expand_policies(const struct flowi
*fl
, u16 family
,
1799 struct xfrm_policy
**pols
,
1800 int *num_pols
, int *num_xfrms
)
1804 if (*num_pols
== 0 || !pols
[0]) {
1809 if (IS_ERR(pols
[0]))
1810 return PTR_ERR(pols
[0]);
1812 *num_xfrms
= pols
[0]->xfrm_nr
;
1814 #ifdef CONFIG_XFRM_SUB_POLICY
1815 if (pols
[0] && pols
[0]->action
== XFRM_POLICY_ALLOW
&&
1816 pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
1817 pols
[1] = xfrm_policy_lookup_bytype(xp_net(pols
[0]),
1818 XFRM_POLICY_TYPE_MAIN
,
1822 if (IS_ERR(pols
[1])) {
1823 xfrm_pols_put(pols
, *num_pols
);
1824 return PTR_ERR(pols
[1]);
1827 (*num_xfrms
) += pols
[1]->xfrm_nr
;
1831 for (i
= 0; i
< *num_pols
; i
++) {
1832 if (pols
[i
]->action
!= XFRM_POLICY_ALLOW
) {
1842 static struct xfrm_dst
*
1843 xfrm_resolve_and_create_bundle(struct xfrm_policy
**pols
, int num_pols
,
1844 const struct flowi
*fl
, u16 family
,
1845 struct dst_entry
*dst_orig
)
1847 struct net
*net
= xp_net(pols
[0]);
1848 struct xfrm_state
*xfrm
[XFRM_MAX_DEPTH
];
1849 struct dst_entry
*dst
;
1850 struct xfrm_dst
*xdst
;
1853 /* Try to instantiate a bundle */
1854 err
= xfrm_tmpl_resolve(pols
, num_pols
, fl
, xfrm
, family
);
1856 if (err
!= 0 && err
!= -EAGAIN
)
1857 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
1858 return ERR_PTR(err
);
1861 dst
= xfrm_bundle_create(pols
[0], xfrm
, err
, fl
, dst_orig
);
1863 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLEGENERROR
);
1864 return ERR_CAST(dst
);
1867 xdst
= (struct xfrm_dst
*)dst
;
1868 xdst
->num_xfrms
= err
;
1870 err
= xfrm_dst_update_parent(dst
, &pols
[1]->selector
);
1872 err
= xfrm_dst_update_origin(dst
, fl
);
1873 if (unlikely(err
)) {
1875 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTBUNDLECHECKERROR
);
1876 return ERR_PTR(err
);
1879 xdst
->num_pols
= num_pols
;
1880 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
1881 xdst
->policy_genid
= atomic_read(&pols
[0]->genid
);
1886 static void xfrm_policy_queue_process(unsigned long arg
)
1888 struct sk_buff
*skb
;
1890 struct dst_entry
*dst
;
1891 struct xfrm_policy
*pol
= (struct xfrm_policy
*)arg
;
1892 struct xfrm_policy_queue
*pq
= &pol
->polq
;
1894 struct sk_buff_head list
;
1896 spin_lock(&pq
->hold_queue
.lock
);
1897 skb
= skb_peek(&pq
->hold_queue
);
1899 spin_unlock(&pq
->hold_queue
.lock
);
1904 xfrm_decode_session(skb
, &fl
, dst
->ops
->family
);
1905 spin_unlock(&pq
->hold_queue
.lock
);
1907 dst_hold(dst
->path
);
1908 dst
= xfrm_lookup(xp_net(pol
), dst
->path
, &fl
,
1913 if (dst
->flags
& DST_XFRM_QUEUE
) {
1916 if (pq
->timeout
>= XFRM_QUEUE_TMO_MAX
)
1919 pq
->timeout
= pq
->timeout
<< 1;
1920 if (!mod_timer(&pq
->hold_timer
, jiffies
+ pq
->timeout
))
1927 __skb_queue_head_init(&list
);
1929 spin_lock(&pq
->hold_queue
.lock
);
1931 skb_queue_splice_init(&pq
->hold_queue
, &list
);
1932 spin_unlock(&pq
->hold_queue
.lock
);
1934 while (!skb_queue_empty(&list
)) {
1935 skb
= __skb_dequeue(&list
);
1937 xfrm_decode_session(skb
, &fl
, skb_dst(skb
)->ops
->family
);
1938 dst_hold(skb_dst(skb
)->path
);
1939 dst
= xfrm_lookup(xp_net(pol
), skb_dst(skb
)->path
,
1948 skb_dst_set(skb
, dst
);
1959 skb_queue_purge(&pq
->hold_queue
);
1963 static int xdst_queue_output(struct sock
*sk
, struct sk_buff
*skb
)
1965 unsigned long sched_next
;
1966 struct dst_entry
*dst
= skb_dst(skb
);
1967 struct xfrm_dst
*xdst
= (struct xfrm_dst
*) dst
;
1968 struct xfrm_policy
*pol
= xdst
->pols
[0];
1969 struct xfrm_policy_queue
*pq
= &pol
->polq
;
1971 if (unlikely(skb_fclone_busy(sk
, skb
))) {
1976 if (pq
->hold_queue
.qlen
> XFRM_MAX_QUEUE_LEN
) {
1983 spin_lock_bh(&pq
->hold_queue
.lock
);
1986 pq
->timeout
= XFRM_QUEUE_TMO_MIN
;
1988 sched_next
= jiffies
+ pq
->timeout
;
1990 if (del_timer(&pq
->hold_timer
)) {
1991 if (time_before(pq
->hold_timer
.expires
, sched_next
))
1992 sched_next
= pq
->hold_timer
.expires
;
1996 __skb_queue_tail(&pq
->hold_queue
, skb
);
1997 if (!mod_timer(&pq
->hold_timer
, sched_next
))
2000 spin_unlock_bh(&pq
->hold_queue
.lock
);
2005 static struct xfrm_dst
*xfrm_create_dummy_bundle(struct net
*net
,
2006 struct xfrm_flo
*xflo
,
2007 const struct flowi
*fl
,
2012 struct net_device
*dev
;
2013 struct dst_entry
*dst
;
2014 struct dst_entry
*dst1
;
2015 struct xfrm_dst
*xdst
;
2017 xdst
= xfrm_alloc_dst(net
, family
);
2021 if (!(xflo
->flags
& XFRM_LOOKUP_QUEUE
) ||
2022 net
->xfrm
.sysctl_larval_drop
||
2026 dst
= xflo
->dst_orig
;
2027 dst1
= &xdst
->u
.dst
;
2031 dst_copy_metrics(dst1
, dst
);
2033 dst1
->obsolete
= DST_OBSOLETE_FORCE_CHK
;
2034 dst1
->flags
|= DST_HOST
| DST_XFRM_QUEUE
;
2035 dst1
->lastuse
= jiffies
;
2037 dst1
->input
= dst_discard
;
2038 dst1
->output
= xdst_queue_output
;
2044 xfrm_init_path((struct xfrm_dst
*)dst1
, dst
, 0);
2051 err
= xfrm_fill_dst(xdst
, dev
, fl
);
2060 xdst
= ERR_PTR(err
);
2064 static struct flow_cache_object
*
2065 xfrm_bundle_lookup(struct net
*net
, const struct flowi
*fl
, u16 family
, u8 dir
,
2066 struct flow_cache_object
*oldflo
, void *ctx
)
2068 struct xfrm_flo
*xflo
= (struct xfrm_flo
*)ctx
;
2069 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2070 struct xfrm_dst
*xdst
, *new_xdst
;
2071 int num_pols
= 0, num_xfrms
= 0, i
, err
, pol_dead
;
2073 /* Check if the policies from old bundle are usable */
2076 xdst
= container_of(oldflo
, struct xfrm_dst
, flo
);
2077 num_pols
= xdst
->num_pols
;
2078 num_xfrms
= xdst
->num_xfrms
;
2080 for (i
= 0; i
< num_pols
; i
++) {
2081 pols
[i
] = xdst
->pols
[i
];
2082 pol_dead
|= pols
[i
]->walk
.dead
;
2085 dst_free(&xdst
->u
.dst
);
2093 /* Resolve policies to use if we couldn't get them from
2094 * previous cache entry */
2097 pols
[0] = __xfrm_policy_lookup(net
, fl
, family
,
2098 flow_to_policy_dir(dir
));
2099 err
= xfrm_expand_policies(fl
, family
, pols
,
2100 &num_pols
, &num_xfrms
);
2106 goto make_dummy_bundle
;
2109 new_xdst
= xfrm_resolve_and_create_bundle(pols
, num_pols
, fl
, family
,
2111 if (IS_ERR(new_xdst
)) {
2112 err
= PTR_ERR(new_xdst
);
2116 goto make_dummy_bundle
;
2117 dst_hold(&xdst
->u
.dst
);
2119 } else if (new_xdst
== NULL
) {
2122 goto make_dummy_bundle
;
2123 xdst
->num_xfrms
= 0;
2124 dst_hold(&xdst
->u
.dst
);
2128 /* Kill the previous bundle */
2130 /* The policies were stolen for newly generated bundle */
2132 dst_free(&xdst
->u
.dst
);
2135 /* Flow cache does not have reference, it dst_free()'s,
2136 * but we do need to return one reference for original caller */
2137 dst_hold(&new_xdst
->u
.dst
);
2138 return &new_xdst
->flo
;
2141 /* We found policies, but there's no bundles to instantiate:
2142 * either because the policy blocks, has no transformations or
2143 * we could not build template (no xfrm_states).*/
2144 xdst
= xfrm_create_dummy_bundle(net
, xflo
, fl
, num_xfrms
, family
);
2146 xfrm_pols_put(pols
, num_pols
);
2147 return ERR_CAST(xdst
);
2149 xdst
->num_pols
= num_pols
;
2150 xdst
->num_xfrms
= num_xfrms
;
2151 memcpy(xdst
->pols
, pols
, sizeof(struct xfrm_policy
*) * num_pols
);
2153 dst_hold(&xdst
->u
.dst
);
2157 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLERROR
);
2160 dst_free(&xdst
->u
.dst
);
2162 xfrm_pols_put(pols
, num_pols
);
2163 return ERR_PTR(err
);
2166 static struct dst_entry
*make_blackhole(struct net
*net
, u16 family
,
2167 struct dst_entry
*dst_orig
)
2169 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
2170 struct dst_entry
*ret
;
2173 dst_release(dst_orig
);
2174 return ERR_PTR(-EINVAL
);
2176 ret
= afinfo
->blackhole_route(net
, dst_orig
);
2178 xfrm_policy_put_afinfo(afinfo
);
2183 /* Main function: finds/creates a bundle for given flow.
2185 * At the moment we eat a raw IP route. Mostly to speed up lookups
2186 * on interfaces with disabled IPsec.
2188 struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
2189 const struct flowi
*fl
,
2190 struct sock
*sk
, int flags
)
2192 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2193 struct flow_cache_object
*flo
;
2194 struct xfrm_dst
*xdst
;
2195 struct dst_entry
*dst
, *route
;
2196 u16 family
= dst_orig
->ops
->family
;
2197 u8 dir
= policy_to_flow_dir(XFRM_POLICY_OUT
);
2198 int i
, err
, num_pols
, num_xfrms
= 0, drop_pols
= 0;
2204 if (sk
&& sk
->sk_policy
[XFRM_POLICY_OUT
]) {
2206 pols
[0] = xfrm_sk_policy_lookup(sk
, XFRM_POLICY_OUT
, fl
);
2207 err
= xfrm_expand_policies(fl
, family
, pols
,
2208 &num_pols
, &num_xfrms
);
2213 if (num_xfrms
<= 0) {
2214 drop_pols
= num_pols
;
2218 xdst
= xfrm_resolve_and_create_bundle(
2222 xfrm_pols_put(pols
, num_pols
);
2223 err
= PTR_ERR(xdst
);
2225 } else if (xdst
== NULL
) {
2227 drop_pols
= num_pols
;
2231 dst_hold(&xdst
->u
.dst
);
2232 xdst
->u
.dst
.flags
|= DST_NOCACHE
;
2233 route
= xdst
->route
;
2238 struct xfrm_flo xflo
;
2240 xflo
.dst_orig
= dst_orig
;
2243 /* To accelerate a bit... */
2244 if ((dst_orig
->flags
& DST_NOXFRM
) ||
2245 !net
->xfrm
.policy_count
[XFRM_POLICY_OUT
])
2248 flo
= flow_cache_lookup(net
, fl
, family
, dir
,
2249 xfrm_bundle_lookup
, &xflo
);
2256 xdst
= container_of(flo
, struct xfrm_dst
, flo
);
2258 num_pols
= xdst
->num_pols
;
2259 num_xfrms
= xdst
->num_xfrms
;
2260 memcpy(pols
, xdst
->pols
, sizeof(struct xfrm_policy
*) * num_pols
);
2261 route
= xdst
->route
;
2265 if (route
== NULL
&& num_xfrms
> 0) {
2266 /* The only case when xfrm_bundle_lookup() returns a
2267 * bundle with null route, is when the template could
2268 * not be resolved. It means policies are there, but
2269 * bundle could not be created, since we don't yet
2270 * have the xfrm_state's. We need to wait for KM to
2271 * negotiate new SA's or bail out with error.*/
2272 if (net
->xfrm
.sysctl_larval_drop
) {
2273 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
2280 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTNOSTATES
);
2288 if ((flags
& XFRM_LOOKUP_ICMP
) &&
2289 !(pols
[0]->flags
& XFRM_POLICY_ICMP
)) {
2294 for (i
= 0; i
< num_pols
; i
++)
2295 pols
[i
]->curlft
.use_time
= get_seconds();
2297 if (num_xfrms
< 0) {
2298 /* Prohibit the flow */
2299 XFRM_INC_STATS(net
, LINUX_MIB_XFRMOUTPOLBLOCK
);
2302 } else if (num_xfrms
> 0) {
2303 /* Flow transformed */
2304 dst_release(dst_orig
);
2306 /* Flow passes untransformed */
2311 xfrm_pols_put(pols
, drop_pols
);
2312 if (dst
&& dst
->xfrm
&&
2313 dst
->xfrm
->props
.mode
== XFRM_MODE_TUNNEL
)
2314 dst
->flags
|= DST_XFRM_TUNNEL
;
2318 if (!(flags
& XFRM_LOOKUP_ICMP
)) {
2326 if (!(flags
& XFRM_LOOKUP_KEEP_DST_REF
))
2327 dst_release(dst_orig
);
2328 xfrm_pols_put(pols
, drop_pols
);
2329 return ERR_PTR(err
);
2331 EXPORT_SYMBOL(xfrm_lookup
);
2333 /* Callers of xfrm_lookup_route() must ensure a call to dst_output().
2334 * Otherwise we may send out blackholed packets.
2336 struct dst_entry
*xfrm_lookup_route(struct net
*net
, struct dst_entry
*dst_orig
,
2337 const struct flowi
*fl
,
2338 struct sock
*sk
, int flags
)
2340 struct dst_entry
*dst
= xfrm_lookup(net
, dst_orig
, fl
, sk
,
2341 flags
| XFRM_LOOKUP_QUEUE
|
2342 XFRM_LOOKUP_KEEP_DST_REF
);
2344 if (IS_ERR(dst
) && PTR_ERR(dst
) == -EREMOTE
)
2345 return make_blackhole(net
, dst_orig
->ops
->family
, dst_orig
);
2349 EXPORT_SYMBOL(xfrm_lookup_route
);
2352 xfrm_secpath_reject(int idx
, struct sk_buff
*skb
, const struct flowi
*fl
)
2354 struct xfrm_state
*x
;
2356 if (!skb
->sp
|| idx
< 0 || idx
>= skb
->sp
->len
)
2358 x
= skb
->sp
->xvec
[idx
];
2359 if (!x
->type
->reject
)
2361 return x
->type
->reject(x
, skb
, fl
);
2364 /* When skb is transformed back to its "native" form, we have to
2365 * check policy restrictions. At the moment we make this in maximally
2366 * stupid way. Shame on me. :-) Of course, connected sockets must
2367 * have policy cached at them.
2371 xfrm_state_ok(const struct xfrm_tmpl
*tmpl
, const struct xfrm_state
*x
,
2372 unsigned short family
)
2374 if (xfrm_state_kern(x
))
2375 return tmpl
->optional
&& !xfrm_state_addr_cmp(tmpl
, x
, tmpl
->encap_family
);
2376 return x
->id
.proto
== tmpl
->id
.proto
&&
2377 (x
->id
.spi
== tmpl
->id
.spi
|| !tmpl
->id
.spi
) &&
2378 (x
->props
.reqid
== tmpl
->reqid
|| !tmpl
->reqid
) &&
2379 x
->props
.mode
== tmpl
->mode
&&
2380 (tmpl
->allalgs
|| (tmpl
->aalgos
& (1<<x
->props
.aalgo
)) ||
2381 !(xfrm_id_proto_match(tmpl
->id
.proto
, IPSEC_PROTO_ANY
))) &&
2382 !(x
->props
.mode
!= XFRM_MODE_TRANSPORT
&&
2383 xfrm_state_addr_cmp(tmpl
, x
, family
));
2387 * 0 or more than 0 is returned when validation is succeeded (either bypass
2388 * because of optional transport mode, or next index of the mathced secpath
2389 * state with the template.
2390 * -1 is returned when no matching template is found.
2391 * Otherwise "-2 - errored_index" is returned.
2394 xfrm_policy_ok(const struct xfrm_tmpl
*tmpl
, const struct sec_path
*sp
, int start
,
2395 unsigned short family
)
2399 if (tmpl
->optional
) {
2400 if (tmpl
->mode
== XFRM_MODE_TRANSPORT
)
2404 for (; idx
< sp
->len
; idx
++) {
2405 if (xfrm_state_ok(tmpl
, sp
->xvec
[idx
], family
))
2407 if (sp
->xvec
[idx
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
2416 int __xfrm_decode_session(struct sk_buff
*skb
, struct flowi
*fl
,
2417 unsigned int family
, int reverse
)
2419 struct xfrm_policy_afinfo
*afinfo
= xfrm_policy_get_afinfo(family
);
2422 if (unlikely(afinfo
== NULL
))
2423 return -EAFNOSUPPORT
;
2425 afinfo
->decode_session(skb
, fl
, reverse
);
2426 err
= security_xfrm_decode_session(skb
, &fl
->flowi_secid
);
2427 xfrm_policy_put_afinfo(afinfo
);
2430 EXPORT_SYMBOL(__xfrm_decode_session
);
2432 static inline int secpath_has_nontransport(const struct sec_path
*sp
, int k
, int *idxp
)
2434 for (; k
< sp
->len
; k
++) {
2435 if (sp
->xvec
[k
]->props
.mode
!= XFRM_MODE_TRANSPORT
) {
2444 int __xfrm_policy_check(struct sock
*sk
, int dir
, struct sk_buff
*skb
,
2445 unsigned short family
)
2447 struct net
*net
= dev_net(skb
->dev
);
2448 struct xfrm_policy
*pol
;
2449 struct xfrm_policy
*pols
[XFRM_POLICY_TYPE_MAX
];
2458 reverse
= dir
& ~XFRM_POLICY_MASK
;
2459 dir
&= XFRM_POLICY_MASK
;
2460 fl_dir
= policy_to_flow_dir(dir
);
2462 if (__xfrm_decode_session(skb
, &fl
, family
, reverse
) < 0) {
2463 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
2467 nf_nat_decode_session(skb
, &fl
, family
);
2469 /* First, check used SA against their selectors. */
2473 for (i
= skb
->sp
->len
-1; i
>= 0; i
--) {
2474 struct xfrm_state
*x
= skb
->sp
->xvec
[i
];
2475 if (!xfrm_selector_match(&x
->sel
, &fl
, family
)) {
2476 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
2483 if (sk
&& sk
->sk_policy
[dir
]) {
2484 pol
= xfrm_sk_policy_lookup(sk
, dir
, &fl
);
2486 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2492 struct flow_cache_object
*flo
;
2494 flo
= flow_cache_lookup(net
, &fl
, family
, fl_dir
,
2495 xfrm_policy_lookup
, NULL
);
2496 if (IS_ERR_OR_NULL(flo
))
2497 pol
= ERR_CAST(flo
);
2499 pol
= container_of(flo
, struct xfrm_policy
, flo
);
2503 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2508 if (skb
->sp
&& secpath_has_nontransport(skb
->sp
, 0, &xerr_idx
)) {
2509 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2510 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOPOLS
);
2516 pol
->curlft
.use_time
= get_seconds();
2520 #ifdef CONFIG_XFRM_SUB_POLICY
2521 if (pols
[0]->type
!= XFRM_POLICY_TYPE_MAIN
) {
2522 pols
[1] = xfrm_policy_lookup_bytype(net
, XFRM_POLICY_TYPE_MAIN
,
2526 if (IS_ERR(pols
[1])) {
2527 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLERROR
);
2530 pols
[1]->curlft
.use_time
= get_seconds();
2536 if (pol
->action
== XFRM_POLICY_ALLOW
) {
2537 struct sec_path
*sp
;
2538 static struct sec_path dummy
;
2539 struct xfrm_tmpl
*tp
[XFRM_MAX_DEPTH
];
2540 struct xfrm_tmpl
*stp
[XFRM_MAX_DEPTH
];
2541 struct xfrm_tmpl
**tpp
= tp
;
2545 if ((sp
= skb
->sp
) == NULL
)
2548 for (pi
= 0; pi
< npols
; pi
++) {
2549 if (pols
[pi
] != pol
&&
2550 pols
[pi
]->action
!= XFRM_POLICY_ALLOW
) {
2551 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2554 if (ti
+ pols
[pi
]->xfrm_nr
>= XFRM_MAX_DEPTH
) {
2555 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
2558 for (i
= 0; i
< pols
[pi
]->xfrm_nr
; i
++)
2559 tpp
[ti
++] = &pols
[pi
]->xfrm_vec
[i
];
2563 xfrm_tmpl_sort(stp
, tpp
, xfrm_nr
, family
, net
);
2567 /* For each tunnel xfrm, find the first matching tmpl.
2568 * For each tmpl before that, find corresponding xfrm.
2569 * Order is _important_. Later we will implement
2570 * some barriers, but at the moment barriers
2571 * are implied between each two transformations.
2573 for (i
= xfrm_nr
-1, k
= 0; i
>= 0; i
--) {
2574 k
= xfrm_policy_ok(tpp
[i
], sp
, k
, family
);
2577 /* "-2 - errored_index" returned */
2579 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2584 if (secpath_has_nontransport(sp
, k
, &xerr_idx
)) {
2585 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINTMPLMISMATCH
);
2589 xfrm_pols_put(pols
, npols
);
2592 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINPOLBLOCK
);
2595 xfrm_secpath_reject(xerr_idx
, skb
, &fl
);
2597 xfrm_pols_put(pols
, npols
);
2600 EXPORT_SYMBOL(__xfrm_policy_check
);
2602 int __xfrm_route_forward(struct sk_buff
*skb
, unsigned short family
)
2604 struct net
*net
= dev_net(skb
->dev
);
2606 struct dst_entry
*dst
;
2609 if (xfrm_decode_session(skb
, &fl
, family
) < 0) {
2610 XFRM_INC_STATS(net
, LINUX_MIB_XFRMFWDHDRERROR
);
2616 dst
= xfrm_lookup(net
, skb_dst(skb
), &fl
, NULL
, XFRM_LOOKUP_QUEUE
);
2621 skb_dst_set(skb
, dst
);
2624 EXPORT_SYMBOL(__xfrm_route_forward
);
2626 /* Optimize later using cookies and generation ids. */
2628 static struct dst_entry
*xfrm_dst_check(struct dst_entry
*dst
, u32 cookie
)
2630 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2631 * to DST_OBSOLETE_FORCE_CHK to force all XFRM destinations to
2632 * get validated by dst_ops->check on every use. We do this
2633 * because when a normal route referenced by an XFRM dst is
2634 * obsoleted we do not go looking around for all parent
2635 * referencing XFRM dsts so that we can invalidate them. It
2636 * is just too much work. Instead we make the checks here on
2637 * every use. For example:
2639 * XFRM dst A --> IPv4 dst X
2641 * X is the "xdst->route" of A (X is also the "dst->path" of A
2642 * in this example). If X is marked obsolete, "A" will not
2643 * notice. That's what we are validating here via the
2644 * stale_bundle() check.
2646 * When a policy's bundle is pruned, we dst_free() the XFRM
2647 * dst which causes it's ->obsolete field to be set to
2648 * DST_OBSOLETE_DEAD. If an XFRM dst has been pruned like
2649 * this, we want to force a new route lookup.
2651 if (dst
->obsolete
< 0 && !stale_bundle(dst
))
2657 static int stale_bundle(struct dst_entry
*dst
)
2659 return !xfrm_bundle_ok((struct xfrm_dst
*)dst
);
2662 void xfrm_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
)
2664 while ((dst
= dst
->child
) && dst
->xfrm
&& dst
->dev
== dev
) {
2665 dst
->dev
= dev_net(dev
)->loopback_dev
;
2670 EXPORT_SYMBOL(xfrm_dst_ifdown
);
2672 static void xfrm_link_failure(struct sk_buff
*skb
)
2674 /* Impossible. Such dst must be popped before reaches point of failure. */
2677 static struct dst_entry
*xfrm_negative_advice(struct dst_entry
*dst
)
2680 if (dst
->obsolete
) {
2688 void xfrm_garbage_collect(struct net
*net
)
2690 flow_cache_flush(net
);
2692 EXPORT_SYMBOL(xfrm_garbage_collect
);
2694 static void xfrm_garbage_collect_deferred(struct net
*net
)
2696 flow_cache_flush_deferred(net
);
2699 static void xfrm_init_pmtu(struct dst_entry
*dst
)
2702 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2703 u32 pmtu
, route_mtu_cached
;
2705 pmtu
= dst_mtu(dst
->child
);
2706 xdst
->child_mtu_cached
= pmtu
;
2708 pmtu
= xfrm_state_mtu(dst
->xfrm
, pmtu
);
2710 route_mtu_cached
= dst_mtu(xdst
->route
);
2711 xdst
->route_mtu_cached
= route_mtu_cached
;
2713 if (pmtu
> route_mtu_cached
)
2714 pmtu
= route_mtu_cached
;
2716 dst_metric_set(dst
, RTAX_MTU
, pmtu
);
2717 } while ((dst
= dst
->next
));
2720 /* Check that the bundle accepts the flow and its components are
2724 static int xfrm_bundle_ok(struct xfrm_dst
*first
)
2726 struct dst_entry
*dst
= &first
->u
.dst
;
2727 struct xfrm_dst
*last
;
2730 if (!dst_check(dst
->path
, ((struct xfrm_dst
*)dst
)->path_cookie
) ||
2731 (dst
->dev
&& !netif_running(dst
->dev
)))
2734 if (dst
->flags
& DST_XFRM_QUEUE
)
2740 struct xfrm_dst
*xdst
= (struct xfrm_dst
*)dst
;
2742 if (dst
->xfrm
->km
.state
!= XFRM_STATE_VALID
)
2744 if (xdst
->xfrm_genid
!= dst
->xfrm
->genid
)
2746 if (xdst
->num_pols
> 0 &&
2747 xdst
->policy_genid
!= atomic_read(&xdst
->pols
[0]->genid
))
2750 mtu
= dst_mtu(dst
->child
);
2751 if (xdst
->child_mtu_cached
!= mtu
) {
2753 xdst
->child_mtu_cached
= mtu
;
2756 if (!dst_check(xdst
->route
, xdst
->route_cookie
))
2758 mtu
= dst_mtu(xdst
->route
);
2759 if (xdst
->route_mtu_cached
!= mtu
) {
2761 xdst
->route_mtu_cached
= mtu
;
2765 } while (dst
->xfrm
);
2770 mtu
= last
->child_mtu_cached
;
2774 mtu
= xfrm_state_mtu(dst
->xfrm
, mtu
);
2775 if (mtu
> last
->route_mtu_cached
)
2776 mtu
= last
->route_mtu_cached
;
2777 dst_metric_set(dst
, RTAX_MTU
, mtu
);
2782 last
= (struct xfrm_dst
*)last
->u
.dst
.next
;
2783 last
->child_mtu_cached
= mtu
;
2789 static unsigned int xfrm_default_advmss(const struct dst_entry
*dst
)
2791 return dst_metric_advmss(dst
->path
);
2794 static unsigned int xfrm_mtu(const struct dst_entry
*dst
)
2796 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2798 return mtu
? : dst_mtu(dst
->path
);
2801 static struct neighbour
*xfrm_neigh_lookup(const struct dst_entry
*dst
,
2802 struct sk_buff
*skb
,
2805 return dst
->path
->ops
->neigh_lookup(dst
, skb
, daddr
);
2808 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2812 if (unlikely(afinfo
== NULL
))
2814 if (unlikely(afinfo
->family
>= NPROTO
))
2815 return -EAFNOSUPPORT
;
2816 spin_lock(&xfrm_policy_afinfo_lock
);
2817 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
))
2820 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2821 if (likely(dst_ops
->kmem_cachep
== NULL
))
2822 dst_ops
->kmem_cachep
= xfrm_dst_cache
;
2823 if (likely(dst_ops
->check
== NULL
))
2824 dst_ops
->check
= xfrm_dst_check
;
2825 if (likely(dst_ops
->default_advmss
== NULL
))
2826 dst_ops
->default_advmss
= xfrm_default_advmss
;
2827 if (likely(dst_ops
->mtu
== NULL
))
2828 dst_ops
->mtu
= xfrm_mtu
;
2829 if (likely(dst_ops
->negative_advice
== NULL
))
2830 dst_ops
->negative_advice
= xfrm_negative_advice
;
2831 if (likely(dst_ops
->link_failure
== NULL
))
2832 dst_ops
->link_failure
= xfrm_link_failure
;
2833 if (likely(dst_ops
->neigh_lookup
== NULL
))
2834 dst_ops
->neigh_lookup
= xfrm_neigh_lookup
;
2835 if (likely(afinfo
->garbage_collect
== NULL
))
2836 afinfo
->garbage_collect
= xfrm_garbage_collect_deferred
;
2837 rcu_assign_pointer(xfrm_policy_afinfo
[afinfo
->family
], afinfo
);
2839 spin_unlock(&xfrm_policy_afinfo_lock
);
2843 struct dst_ops
*xfrm_dst_ops
;
2845 switch (afinfo
->family
) {
2847 xfrm_dst_ops
= &net
->xfrm
.xfrm4_dst_ops
;
2849 #if IS_ENABLED(CONFIG_IPV6)
2851 xfrm_dst_ops
= &net
->xfrm
.xfrm6_dst_ops
;
2857 *xfrm_dst_ops
= *afinfo
->dst_ops
;
2863 EXPORT_SYMBOL(xfrm_policy_register_afinfo
);
2865 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo
*afinfo
)
2868 if (unlikely(afinfo
== NULL
))
2870 if (unlikely(afinfo
->family
>= NPROTO
))
2871 return -EAFNOSUPPORT
;
2872 spin_lock(&xfrm_policy_afinfo_lock
);
2873 if (likely(xfrm_policy_afinfo
[afinfo
->family
] != NULL
)) {
2874 if (unlikely(xfrm_policy_afinfo
[afinfo
->family
] != afinfo
))
2877 RCU_INIT_POINTER(xfrm_policy_afinfo
[afinfo
->family
],
2880 spin_unlock(&xfrm_policy_afinfo_lock
);
2882 struct dst_ops
*dst_ops
= afinfo
->dst_ops
;
2886 dst_ops
->kmem_cachep
= NULL
;
2887 dst_ops
->check
= NULL
;
2888 dst_ops
->negative_advice
= NULL
;
2889 dst_ops
->link_failure
= NULL
;
2890 afinfo
->garbage_collect
= NULL
;
2894 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo
);
2896 static void __net_init
xfrm_dst_ops_init(struct net
*net
)
2898 struct xfrm_policy_afinfo
*afinfo
;
2901 afinfo
= rcu_dereference(xfrm_policy_afinfo
[AF_INET
]);
2903 net
->xfrm
.xfrm4_dst_ops
= *afinfo
->dst_ops
;
2904 #if IS_ENABLED(CONFIG_IPV6)
2905 afinfo
= rcu_dereference(xfrm_policy_afinfo
[AF_INET6
]);
2907 net
->xfrm
.xfrm6_dst_ops
= *afinfo
->dst_ops
;
2912 static int xfrm_dev_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2914 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2918 xfrm_garbage_collect(dev_net(dev
));
2923 static struct notifier_block xfrm_dev_notifier
= {
2924 .notifier_call
= xfrm_dev_event
,
2927 #ifdef CONFIG_XFRM_STATISTICS
2928 static int __net_init
xfrm_statistics_init(struct net
*net
)
2931 net
->mib
.xfrm_statistics
= alloc_percpu(struct linux_xfrm_mib
);
2932 if (!net
->mib
.xfrm_statistics
)
2934 rv
= xfrm_proc_init(net
);
2936 free_percpu(net
->mib
.xfrm_statistics
);
2940 static void xfrm_statistics_fini(struct net
*net
)
2942 xfrm_proc_fini(net
);
2943 free_percpu(net
->mib
.xfrm_statistics
);
2946 static int __net_init
xfrm_statistics_init(struct net
*net
)
2951 static void xfrm_statistics_fini(struct net
*net
)
2956 static int __net_init
xfrm_policy_init(struct net
*net
)
2958 unsigned int hmask
, sz
;
2961 if (net_eq(net
, &init_net
))
2962 xfrm_dst_cache
= kmem_cache_create("xfrm_dst_cache",
2963 sizeof(struct xfrm_dst
),
2964 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2968 sz
= (hmask
+1) * sizeof(struct hlist_head
);
2970 net
->xfrm
.policy_byidx
= xfrm_hash_alloc(sz
);
2971 if (!net
->xfrm
.policy_byidx
)
2973 net
->xfrm
.policy_idx_hmask
= hmask
;
2975 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
2976 struct xfrm_policy_hash
*htab
;
2978 net
->xfrm
.policy_count
[dir
] = 0;
2979 net
->xfrm
.policy_count
[XFRM_POLICY_MAX
+ dir
] = 0;
2980 INIT_HLIST_HEAD(&net
->xfrm
.policy_inexact
[dir
]);
2982 htab
= &net
->xfrm
.policy_bydst
[dir
];
2983 htab
->table
= xfrm_hash_alloc(sz
);
2986 htab
->hmask
= hmask
;
2992 net
->xfrm
.policy_hthresh
.lbits4
= 32;
2993 net
->xfrm
.policy_hthresh
.rbits4
= 32;
2994 net
->xfrm
.policy_hthresh
.lbits6
= 128;
2995 net
->xfrm
.policy_hthresh
.rbits6
= 128;
2997 seqlock_init(&net
->xfrm
.policy_hthresh
.lock
);
2999 INIT_LIST_HEAD(&net
->xfrm
.policy_all
);
3000 INIT_WORK(&net
->xfrm
.policy_hash_work
, xfrm_hash_resize
);
3001 INIT_WORK(&net
->xfrm
.policy_hthresh
.work
, xfrm_hash_rebuild
);
3002 if (net_eq(net
, &init_net
))
3003 register_netdevice_notifier(&xfrm_dev_notifier
);
3007 for (dir
--; dir
>= 0; dir
--) {
3008 struct xfrm_policy_hash
*htab
;
3010 htab
= &net
->xfrm
.policy_bydst
[dir
];
3011 xfrm_hash_free(htab
->table
, sz
);
3013 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
3018 static void xfrm_policy_fini(struct net
*net
)
3023 flush_work(&net
->xfrm
.policy_hash_work
);
3024 #ifdef CONFIG_XFRM_SUB_POLICY
3025 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_SUB
, false);
3027 xfrm_policy_flush(net
, XFRM_POLICY_TYPE_MAIN
, false);
3029 WARN_ON(!list_empty(&net
->xfrm
.policy_all
));
3031 for (dir
= 0; dir
< XFRM_POLICY_MAX
; dir
++) {
3032 struct xfrm_policy_hash
*htab
;
3034 WARN_ON(!hlist_empty(&net
->xfrm
.policy_inexact
[dir
]));
3036 htab
= &net
->xfrm
.policy_bydst
[dir
];
3037 sz
= (htab
->hmask
+ 1) * sizeof(struct hlist_head
);
3038 WARN_ON(!hlist_empty(htab
->table
));
3039 xfrm_hash_free(htab
->table
, sz
);
3042 sz
= (net
->xfrm
.policy_idx_hmask
+ 1) * sizeof(struct hlist_head
);
3043 WARN_ON(!hlist_empty(net
->xfrm
.policy_byidx
));
3044 xfrm_hash_free(net
->xfrm
.policy_byidx
, sz
);
3047 static int __net_init
xfrm_net_init(struct net
*net
)
3051 rv
= xfrm_statistics_init(net
);
3053 goto out_statistics
;
3054 rv
= xfrm_state_init(net
);
3057 rv
= xfrm_policy_init(net
);
3060 xfrm_dst_ops_init(net
);
3061 rv
= xfrm_sysctl_init(net
);
3064 rv
= flow_cache_init(net
);
3068 /* Initialize the per-net locks here */
3069 spin_lock_init(&net
->xfrm
.xfrm_state_lock
);
3070 rwlock_init(&net
->xfrm
.xfrm_policy_lock
);
3071 mutex_init(&net
->xfrm
.xfrm_cfg_mutex
);
3076 xfrm_sysctl_fini(net
);
3078 xfrm_policy_fini(net
);
3080 xfrm_state_fini(net
);
3082 xfrm_statistics_fini(net
);
3087 static void __net_exit
xfrm_net_exit(struct net
*net
)
3089 flow_cache_fini(net
);
3090 xfrm_sysctl_fini(net
);
3091 xfrm_policy_fini(net
);
3092 xfrm_state_fini(net
);
3093 xfrm_statistics_fini(net
);
3096 static struct pernet_operations __net_initdata xfrm_net_ops
= {
3097 .init
= xfrm_net_init
,
3098 .exit
= xfrm_net_exit
,
3101 void __init
xfrm_init(void)
3103 register_pernet_subsys(&xfrm_net_ops
);
3107 #ifdef CONFIG_AUDITSYSCALL
3108 static void xfrm_audit_common_policyinfo(struct xfrm_policy
*xp
,
3109 struct audit_buffer
*audit_buf
)
3111 struct xfrm_sec_ctx
*ctx
= xp
->security
;
3112 struct xfrm_selector
*sel
= &xp
->selector
;
3115 audit_log_format(audit_buf
, " sec_alg=%u sec_doi=%u sec_obj=%s",
3116 ctx
->ctx_alg
, ctx
->ctx_doi
, ctx
->ctx_str
);
3118 switch (sel
->family
) {
3120 audit_log_format(audit_buf
, " src=%pI4", &sel
->saddr
.a4
);
3121 if (sel
->prefixlen_s
!= 32)
3122 audit_log_format(audit_buf
, " src_prefixlen=%d",
3124 audit_log_format(audit_buf
, " dst=%pI4", &sel
->daddr
.a4
);
3125 if (sel
->prefixlen_d
!= 32)
3126 audit_log_format(audit_buf
, " dst_prefixlen=%d",
3130 audit_log_format(audit_buf
, " src=%pI6", sel
->saddr
.a6
);
3131 if (sel
->prefixlen_s
!= 128)
3132 audit_log_format(audit_buf
, " src_prefixlen=%d",
3134 audit_log_format(audit_buf
, " dst=%pI6", sel
->daddr
.a6
);
3135 if (sel
->prefixlen_d
!= 128)
3136 audit_log_format(audit_buf
, " dst_prefixlen=%d",
3142 void xfrm_audit_policy_add(struct xfrm_policy
*xp
, int result
, bool task_valid
)
3144 struct audit_buffer
*audit_buf
;
3146 audit_buf
= xfrm_audit_start("SPD-add");
3147 if (audit_buf
== NULL
)
3149 xfrm_audit_helper_usrinfo(task_valid
, audit_buf
);
3150 audit_log_format(audit_buf
, " res=%u", result
);
3151 xfrm_audit_common_policyinfo(xp
, audit_buf
);
3152 audit_log_end(audit_buf
);
3154 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add
);
3156 void xfrm_audit_policy_delete(struct xfrm_policy
*xp
, int result
,
3159 struct audit_buffer
*audit_buf
;
3161 audit_buf
= xfrm_audit_start("SPD-delete");
3162 if (audit_buf
== NULL
)
3164 xfrm_audit_helper_usrinfo(task_valid
, audit_buf
);
3165 audit_log_format(audit_buf
, " res=%u", result
);
3166 xfrm_audit_common_policyinfo(xp
, audit_buf
);
3167 audit_log_end(audit_buf
);
3169 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete
);
3172 #ifdef CONFIG_XFRM_MIGRATE
3173 static bool xfrm_migrate_selector_match(const struct xfrm_selector
*sel_cmp
,
3174 const struct xfrm_selector
*sel_tgt
)
3176 if (sel_cmp
->proto
== IPSEC_ULPROTO_ANY
) {
3177 if (sel_tgt
->family
== sel_cmp
->family
&&
3178 xfrm_addr_equal(&sel_tgt
->daddr
, &sel_cmp
->daddr
,
3180 xfrm_addr_equal(&sel_tgt
->saddr
, &sel_cmp
->saddr
,
3182 sel_tgt
->prefixlen_d
== sel_cmp
->prefixlen_d
&&
3183 sel_tgt
->prefixlen_s
== sel_cmp
->prefixlen_s
) {
3187 if (memcmp(sel_tgt
, sel_cmp
, sizeof(*sel_tgt
)) == 0) {
3194 static struct xfrm_policy
*xfrm_migrate_policy_find(const struct xfrm_selector
*sel
,
3195 u8 dir
, u8 type
, struct net
*net
)
3197 struct xfrm_policy
*pol
, *ret
= NULL
;
3198 struct hlist_head
*chain
;
3201 read_lock_bh(&net
->xfrm
.xfrm_policy_lock
); /*FIXME*/
3202 chain
= policy_hash_direct(net
, &sel
->daddr
, &sel
->saddr
, sel
->family
, dir
);
3203 hlist_for_each_entry(pol
, chain
, bydst
) {
3204 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
3205 pol
->type
== type
) {
3207 priority
= ret
->priority
;
3211 chain
= &net
->xfrm
.policy_inexact
[dir
];
3212 hlist_for_each_entry(pol
, chain
, bydst
) {
3213 if ((pol
->priority
>= priority
) && ret
)
3216 if (xfrm_migrate_selector_match(sel
, &pol
->selector
) &&
3217 pol
->type
== type
) {
3225 read_unlock_bh(&net
->xfrm
.xfrm_policy_lock
);
3230 static int migrate_tmpl_match(const struct xfrm_migrate
*m
, const struct xfrm_tmpl
*t
)
3234 if (t
->mode
== m
->mode
&& t
->id
.proto
== m
->proto
&&
3235 (m
->reqid
== 0 || t
->reqid
== m
->reqid
)) {
3237 case XFRM_MODE_TUNNEL
:
3238 case XFRM_MODE_BEET
:
3239 if (xfrm_addr_equal(&t
->id
.daddr
, &m
->old_daddr
,
3241 xfrm_addr_equal(&t
->saddr
, &m
->old_saddr
,
3246 case XFRM_MODE_TRANSPORT
:
3247 /* in case of transport mode, template does not store
3248 any IP addresses, hence we just compare mode and
3259 /* update endpoint address(es) of template(s) */
3260 static int xfrm_policy_migrate(struct xfrm_policy
*pol
,
3261 struct xfrm_migrate
*m
, int num_migrate
)
3263 struct xfrm_migrate
*mp
;
3266 write_lock_bh(&pol
->lock
);
3267 if (unlikely(pol
->walk
.dead
)) {
3268 /* target policy has been deleted */
3269 write_unlock_bh(&pol
->lock
);
3273 for (i
= 0; i
< pol
->xfrm_nr
; i
++) {
3274 for (j
= 0, mp
= m
; j
< num_migrate
; j
++, mp
++) {
3275 if (!migrate_tmpl_match(mp
, &pol
->xfrm_vec
[i
]))
3278 if (pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_TUNNEL
&&
3279 pol
->xfrm_vec
[i
].mode
!= XFRM_MODE_BEET
)
3281 /* update endpoints */
3282 memcpy(&pol
->xfrm_vec
[i
].id
.daddr
, &mp
->new_daddr
,
3283 sizeof(pol
->xfrm_vec
[i
].id
.daddr
));
3284 memcpy(&pol
->xfrm_vec
[i
].saddr
, &mp
->new_saddr
,
3285 sizeof(pol
->xfrm_vec
[i
].saddr
));
3286 pol
->xfrm_vec
[i
].encap_family
= mp
->new_family
;
3288 atomic_inc(&pol
->genid
);
3292 write_unlock_bh(&pol
->lock
);
3300 static int xfrm_migrate_check(const struct xfrm_migrate
*m
, int num_migrate
)
3304 if (num_migrate
< 1 || num_migrate
> XFRM_MAX_DEPTH
)
3307 for (i
= 0; i
< num_migrate
; i
++) {
3308 if (xfrm_addr_equal(&m
[i
].old_daddr
, &m
[i
].new_daddr
,
3310 xfrm_addr_equal(&m
[i
].old_saddr
, &m
[i
].new_saddr
,
3313 if (xfrm_addr_any(&m
[i
].new_daddr
, m
[i
].new_family
) ||
3314 xfrm_addr_any(&m
[i
].new_saddr
, m
[i
].new_family
))
3317 /* check if there is any duplicated entry */
3318 for (j
= i
+ 1; j
< num_migrate
; j
++) {
3319 if (!memcmp(&m
[i
].old_daddr
, &m
[j
].old_daddr
,
3320 sizeof(m
[i
].old_daddr
)) &&
3321 !memcmp(&m
[i
].old_saddr
, &m
[j
].old_saddr
,
3322 sizeof(m
[i
].old_saddr
)) &&
3323 m
[i
].proto
== m
[j
].proto
&&
3324 m
[i
].mode
== m
[j
].mode
&&
3325 m
[i
].reqid
== m
[j
].reqid
&&
3326 m
[i
].old_family
== m
[j
].old_family
)
3334 int xfrm_migrate(const struct xfrm_selector
*sel
, u8 dir
, u8 type
,
3335 struct xfrm_migrate
*m
, int num_migrate
,
3336 struct xfrm_kmaddress
*k
, struct net
*net
)
3338 int i
, err
, nx_cur
= 0, nx_new
= 0;
3339 struct xfrm_policy
*pol
= NULL
;
3340 struct xfrm_state
*x
, *xc
;
3341 struct xfrm_state
*x_cur
[XFRM_MAX_DEPTH
];
3342 struct xfrm_state
*x_new
[XFRM_MAX_DEPTH
];
3343 struct xfrm_migrate
*mp
;
3345 if ((err
= xfrm_migrate_check(m
, num_migrate
)) < 0)
3348 /* Stage 1 - find policy */
3349 if ((pol
= xfrm_migrate_policy_find(sel
, dir
, type
, net
)) == NULL
) {
3354 /* Stage 2 - find and update state(s) */
3355 for (i
= 0, mp
= m
; i
< num_migrate
; i
++, mp
++) {
3356 if ((x
= xfrm_migrate_state_find(mp
, net
))) {
3359 if ((xc
= xfrm_state_migrate(x
, mp
))) {
3369 /* Stage 3 - update policy */
3370 if ((err
= xfrm_policy_migrate(pol
, m
, num_migrate
)) < 0)
3373 /* Stage 4 - delete old state(s) */
3375 xfrm_states_put(x_cur
, nx_cur
);
3376 xfrm_states_delete(x_cur
, nx_cur
);
3379 /* Stage 5 - announce */
3380 km_migrate(sel
, dir
, type
, m
, num_migrate
, k
);
3392 xfrm_states_put(x_cur
, nx_cur
);
3394 xfrm_states_delete(x_new
, nx_new
);
3398 EXPORT_SYMBOL(xfrm_migrate
);