6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * Kazunori MIYAZAWA @USAGI
11 * Split up af-specific portion
12 * Derek Atkins <derek@ihtfp.com> Add the post_input processor
16 #include <linux/err.h>
17 #include <linux/slab.h>
18 #include <linux/kmod.h>
19 #include <linux/list.h>
20 #include <linux/spinlock.h>
21 #include <linux/workqueue.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/module.h>
26 #include <linux/cache.h>
27 #include <linux/audit.h>
31 #ifdef CONFIG_XFRM_STATISTICS
35 #include "xfrm_hash.h"
37 DEFINE_MUTEX(xfrm_cfg_mutex);
38 EXPORT_SYMBOL(xfrm_cfg_mutex);
40 static DEFINE_SPINLOCK(xfrm_policy_sk_bundle_lock);
41 static struct dst_entry *xfrm_policy_sk_bundles;
42 static DEFINE_RWLOCK(xfrm_policy_lock);
44 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
45 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
47 static struct kmem_cache *xfrm_dst_cache __read_mostly;
49 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
50 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
51 static void xfrm_init_pmtu(struct dst_entry *dst);
52 static int stale_bundle(struct dst_entry *dst);
53 static int xfrm_bundle_ok(struct xfrm_dst *xdst, int family);
56 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
60 __xfrm4_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
62 const struct flowi4 *fl4 = &fl->u.ip4;
64 return addr_match(&fl4->daddr, &sel->daddr, sel->prefixlen_d) &&
65 addr_match(&fl4->saddr, &sel->saddr, sel->prefixlen_s) &&
66 !((xfrm_flowi_dport(fl, &fl4->uli) ^ sel->dport) & sel->dport_mask) &&
67 !((xfrm_flowi_sport(fl, &fl4->uli) ^ sel->sport) & sel->sport_mask) &&
68 (fl4->flowi4_proto == sel->proto || !sel->proto) &&
69 (fl4->flowi4_oif == sel->ifindex || !sel->ifindex);
73 __xfrm6_selector_match(const struct xfrm_selector *sel, const struct flowi *fl)
75 const struct flowi6 *fl6 = &fl->u.ip6;
77 return addr_match(&fl6->daddr, &sel->daddr, sel->prefixlen_d) &&
78 addr_match(&fl6->saddr, &sel->saddr, sel->prefixlen_s) &&
79 !((xfrm_flowi_dport(fl, &fl6->uli) ^ sel->dport) & sel->dport_mask) &&
80 !((xfrm_flowi_sport(fl, &fl6->uli) ^ sel->sport) & sel->sport_mask) &&
81 (fl6->flowi6_proto == sel->proto || !sel->proto) &&
82 (fl6->flowi6_oif == sel->ifindex || !sel->ifindex);
85 int xfrm_selector_match(const struct xfrm_selector *sel, const struct flowi *fl,
86 unsigned short family)
90 return __xfrm4_selector_match(sel, fl);
92 return __xfrm6_selector_match(sel, fl);
97 static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
98 const xfrm_address_t *saddr,
99 const xfrm_address_t *daddr,
102 struct xfrm_policy_afinfo *afinfo;
103 struct dst_entry *dst;
105 afinfo = xfrm_policy_get_afinfo(family);
106 if (unlikely(afinfo == NULL))
107 return ERR_PTR(-EAFNOSUPPORT);
109 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
111 xfrm_policy_put_afinfo(afinfo);
116 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
117 xfrm_address_t *prev_saddr,
118 xfrm_address_t *prev_daddr,
121 struct net *net = xs_net(x);
122 xfrm_address_t *saddr = &x->props.saddr;
123 xfrm_address_t *daddr = &x->id.daddr;
124 struct dst_entry *dst;
126 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
130 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
135 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
138 if (prev_saddr != saddr)
139 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
140 if (prev_daddr != daddr)
141 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
147 static inline unsigned long make_jiffies(long secs)
149 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
150 return MAX_SCHEDULE_TIMEOUT-1;
155 static void xfrm_policy_timer(unsigned long data)
157 struct xfrm_policy *xp = (struct xfrm_policy*)data;
158 unsigned long now = get_seconds();
159 long next = LONG_MAX;
163 read_lock(&xp->lock);
165 if (unlikely(xp->walk.dead))
168 dir = xfrm_policy_id2dir(xp->index);
170 if (xp->lft.hard_add_expires_seconds) {
171 long tmo = xp->lft.hard_add_expires_seconds +
172 xp->curlft.add_time - now;
178 if (xp->lft.hard_use_expires_seconds) {
179 long tmo = xp->lft.hard_use_expires_seconds +
180 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
186 if (xp->lft.soft_add_expires_seconds) {
187 long tmo = xp->lft.soft_add_expires_seconds +
188 xp->curlft.add_time - now;
191 tmo = XFRM_KM_TIMEOUT;
196 if (xp->lft.soft_use_expires_seconds) {
197 long tmo = xp->lft.soft_use_expires_seconds +
198 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
201 tmo = XFRM_KM_TIMEOUT;
208 km_policy_expired(xp, dir, 0, 0);
209 if (next != LONG_MAX &&
210 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
214 read_unlock(&xp->lock);
219 read_unlock(&xp->lock);
220 if (!xfrm_policy_delete(xp, dir))
221 km_policy_expired(xp, dir, 1, 0);
225 static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
227 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
229 if (unlikely(pol->walk.dead))
237 static int xfrm_policy_flo_check(struct flow_cache_object *flo)
239 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
241 return !pol->walk.dead;
244 static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
246 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
249 static const struct flow_cache_ops xfrm_policy_fc_ops = {
250 .get = xfrm_policy_flo_get,
251 .check = xfrm_policy_flo_check,
252 .delete = xfrm_policy_flo_delete,
255 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
259 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
261 struct xfrm_policy *policy;
263 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
266 write_pnet(&policy->xp_net, net);
267 INIT_LIST_HEAD(&policy->walk.all);
268 INIT_HLIST_NODE(&policy->bydst);
269 INIT_HLIST_NODE(&policy->byidx);
270 rwlock_init(&policy->lock);
271 atomic_set(&policy->refcnt, 1);
272 setup_timer(&policy->timer, xfrm_policy_timer,
273 (unsigned long)policy);
274 policy->flo.ops = &xfrm_policy_fc_ops;
278 EXPORT_SYMBOL(xfrm_policy_alloc);
280 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
282 void xfrm_policy_destroy(struct xfrm_policy *policy)
284 BUG_ON(!policy->walk.dead);
286 if (del_timer(&policy->timer))
289 security_xfrm_policy_free(policy->security);
292 EXPORT_SYMBOL(xfrm_policy_destroy);
294 /* Rule must be locked. Release descentant resources, announce
295 * entry dead. The rule must be unlinked from lists to the moment.
298 static void xfrm_policy_kill(struct xfrm_policy *policy)
300 policy->walk.dead = 1;
302 atomic_inc(&policy->genid);
304 if (del_timer(&policy->timer))
305 xfrm_pol_put(policy);
307 xfrm_pol_put(policy);
310 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
312 static inline unsigned int idx_hash(struct net *net, u32 index)
314 return __idx_hash(index, net->xfrm.policy_idx_hmask);
317 static struct hlist_head *policy_hash_bysel(struct net *net,
318 const struct xfrm_selector *sel,
319 unsigned short family, int dir)
321 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
322 unsigned int hash = __sel_hash(sel, family, hmask);
324 return (hash == hmask + 1 ?
325 &net->xfrm.policy_inexact[dir] :
326 net->xfrm.policy_bydst[dir].table + hash);
329 static struct hlist_head *policy_hash_direct(struct net *net,
330 const xfrm_address_t *daddr,
331 const xfrm_address_t *saddr,
332 unsigned short family, int dir)
334 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
335 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
337 return net->xfrm.policy_bydst[dir].table + hash;
340 static void xfrm_dst_hash_transfer(struct hlist_head *list,
341 struct hlist_head *ndsttable,
342 unsigned int nhashmask)
344 struct hlist_node *entry, *tmp, *entry0 = NULL;
345 struct xfrm_policy *pol;
349 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
352 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
353 pol->family, nhashmask);
356 hlist_add_head(&pol->bydst, ndsttable+h);
362 hlist_add_after(entry0, &pol->bydst);
366 if (!hlist_empty(list)) {
372 static void xfrm_idx_hash_transfer(struct hlist_head *list,
373 struct hlist_head *nidxtable,
374 unsigned int nhashmask)
376 struct hlist_node *entry, *tmp;
377 struct xfrm_policy *pol;
379 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
382 h = __idx_hash(pol->index, nhashmask);
383 hlist_add_head(&pol->byidx, nidxtable+h);
387 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
389 return ((old_hmask + 1) << 1) - 1;
392 static void xfrm_bydst_resize(struct net *net, int dir)
394 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
395 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
396 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
397 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
398 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
404 write_lock_bh(&xfrm_policy_lock);
406 for (i = hmask; i >= 0; i--)
407 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
409 net->xfrm.policy_bydst[dir].table = ndst;
410 net->xfrm.policy_bydst[dir].hmask = nhashmask;
412 write_unlock_bh(&xfrm_policy_lock);
414 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
417 static void xfrm_byidx_resize(struct net *net, int total)
419 unsigned int hmask = net->xfrm.policy_idx_hmask;
420 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
421 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
422 struct hlist_head *oidx = net->xfrm.policy_byidx;
423 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
429 write_lock_bh(&xfrm_policy_lock);
431 for (i = hmask; i >= 0; i--)
432 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
434 net->xfrm.policy_byidx = nidx;
435 net->xfrm.policy_idx_hmask = nhashmask;
437 write_unlock_bh(&xfrm_policy_lock);
439 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
442 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
444 unsigned int cnt = net->xfrm.policy_count[dir];
445 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
450 if ((hmask + 1) < xfrm_policy_hashmax &&
457 static inline int xfrm_byidx_should_resize(struct net *net, int total)
459 unsigned int hmask = net->xfrm.policy_idx_hmask;
461 if ((hmask + 1) < xfrm_policy_hashmax &&
468 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
470 read_lock_bh(&xfrm_policy_lock);
471 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
472 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
473 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
474 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
475 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
476 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
477 si->spdhcnt = net->xfrm.policy_idx_hmask;
478 si->spdhmcnt = xfrm_policy_hashmax;
479 read_unlock_bh(&xfrm_policy_lock);
481 EXPORT_SYMBOL(xfrm_spd_getinfo);
483 static DEFINE_MUTEX(hash_resize_mutex);
484 static void xfrm_hash_resize(struct work_struct *work)
486 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
489 mutex_lock(&hash_resize_mutex);
492 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
493 if (xfrm_bydst_should_resize(net, dir, &total))
494 xfrm_bydst_resize(net, dir);
496 if (xfrm_byidx_should_resize(net, total))
497 xfrm_byidx_resize(net, total);
499 mutex_unlock(&hash_resize_mutex);
502 /* Generate new index... KAME seems to generate them ordered by cost
503 * of an absolute inpredictability of ordering of rules. This will not pass. */
504 static u32 xfrm_gen_index(struct net *net, int dir)
506 static u32 idx_generator;
509 struct hlist_node *entry;
510 struct hlist_head *list;
511 struct xfrm_policy *p;
515 idx = (idx_generator | dir);
519 list = net->xfrm.policy_byidx + idx_hash(net, idx);
521 hlist_for_each_entry(p, entry, list, byidx) {
522 if (p->index == idx) {
532 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
534 u32 *p1 = (u32 *) s1;
535 u32 *p2 = (u32 *) s2;
536 int len = sizeof(struct xfrm_selector) / sizeof(u32);
539 for (i = 0; i < len; i++) {
547 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
549 struct net *net = xp_net(policy);
550 struct xfrm_policy *pol;
551 struct xfrm_policy *delpol;
552 struct hlist_head *chain;
553 struct hlist_node *entry, *newpos;
554 u32 mark = policy->mark.v & policy->mark.m;
556 write_lock_bh(&xfrm_policy_lock);
557 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
560 hlist_for_each_entry(pol, entry, chain, bydst) {
561 if (pol->type == policy->type &&
562 !selector_cmp(&pol->selector, &policy->selector) &&
563 (mark & pol->mark.m) == pol->mark.v &&
564 xfrm_sec_ctx_match(pol->security, policy->security) &&
567 write_unlock_bh(&xfrm_policy_lock);
571 if (policy->priority > pol->priority)
573 } else if (policy->priority >= pol->priority) {
574 newpos = &pol->bydst;
581 hlist_add_after(newpos, &policy->bydst);
583 hlist_add_head(&policy->bydst, chain);
584 xfrm_pol_hold(policy);
585 net->xfrm.policy_count[dir]++;
586 atomic_inc(&flow_cache_genid);
588 __xfrm_policy_unlink(delpol, dir);
589 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
590 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
591 policy->curlft.add_time = get_seconds();
592 policy->curlft.use_time = 0;
593 if (!mod_timer(&policy->timer, jiffies + HZ))
594 xfrm_pol_hold(policy);
595 list_add(&policy->walk.all, &net->xfrm.policy_all);
596 write_unlock_bh(&xfrm_policy_lock);
599 xfrm_policy_kill(delpol);
600 else if (xfrm_bydst_should_resize(net, dir, NULL))
601 schedule_work(&net->xfrm.policy_hash_work);
605 EXPORT_SYMBOL(xfrm_policy_insert);
607 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
608 int dir, struct xfrm_selector *sel,
609 struct xfrm_sec_ctx *ctx, int delete,
612 struct xfrm_policy *pol, *ret;
613 struct hlist_head *chain;
614 struct hlist_node *entry;
617 write_lock_bh(&xfrm_policy_lock);
618 chain = policy_hash_bysel(net, sel, sel->family, dir);
620 hlist_for_each_entry(pol, entry, chain, bydst) {
621 if (pol->type == type &&
622 (mark & pol->mark.m) == pol->mark.v &&
623 !selector_cmp(sel, &pol->selector) &&
624 xfrm_sec_ctx_match(ctx, pol->security)) {
627 *err = security_xfrm_policy_delete(
630 write_unlock_bh(&xfrm_policy_lock);
633 __xfrm_policy_unlink(pol, dir);
639 write_unlock_bh(&xfrm_policy_lock);
642 xfrm_policy_kill(ret);
645 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
647 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
648 int dir, u32 id, int delete, int *err)
650 struct xfrm_policy *pol, *ret;
651 struct hlist_head *chain;
652 struct hlist_node *entry;
655 if (xfrm_policy_id2dir(id) != dir)
659 write_lock_bh(&xfrm_policy_lock);
660 chain = net->xfrm.policy_byidx + idx_hash(net, id);
662 hlist_for_each_entry(pol, entry, chain, byidx) {
663 if (pol->type == type && pol->index == id &&
664 (mark & pol->mark.m) == pol->mark.v) {
667 *err = security_xfrm_policy_delete(
670 write_unlock_bh(&xfrm_policy_lock);
673 __xfrm_policy_unlink(pol, dir);
679 write_unlock_bh(&xfrm_policy_lock);
682 xfrm_policy_kill(ret);
685 EXPORT_SYMBOL(xfrm_policy_byid);
687 #ifdef CONFIG_SECURITY_NETWORK_XFRM
689 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
693 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
694 struct xfrm_policy *pol;
695 struct hlist_node *entry;
698 hlist_for_each_entry(pol, entry,
699 &net->xfrm.policy_inexact[dir], bydst) {
700 if (pol->type != type)
702 err = security_xfrm_policy_delete(pol->security);
704 xfrm_audit_policy_delete(pol, 0,
705 audit_info->loginuid,
706 audit_info->sessionid,
711 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
712 hlist_for_each_entry(pol, entry,
713 net->xfrm.policy_bydst[dir].table + i,
715 if (pol->type != type)
717 err = security_xfrm_policy_delete(
720 xfrm_audit_policy_delete(pol, 0,
721 audit_info->loginuid,
722 audit_info->sessionid,
733 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
739 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
741 int dir, err = 0, cnt = 0;
743 write_lock_bh(&xfrm_policy_lock);
745 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
749 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
750 struct xfrm_policy *pol;
751 struct hlist_node *entry;
755 hlist_for_each_entry(pol, entry,
756 &net->xfrm.policy_inexact[dir], bydst) {
757 if (pol->type != type)
759 __xfrm_policy_unlink(pol, dir);
760 write_unlock_bh(&xfrm_policy_lock);
763 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
764 audit_info->sessionid,
767 xfrm_policy_kill(pol);
769 write_lock_bh(&xfrm_policy_lock);
773 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
775 hlist_for_each_entry(pol, entry,
776 net->xfrm.policy_bydst[dir].table + i,
778 if (pol->type != type)
780 __xfrm_policy_unlink(pol, dir);
781 write_unlock_bh(&xfrm_policy_lock);
784 xfrm_audit_policy_delete(pol, 1,
785 audit_info->loginuid,
786 audit_info->sessionid,
788 xfrm_policy_kill(pol);
790 write_lock_bh(&xfrm_policy_lock);
799 write_unlock_bh(&xfrm_policy_lock);
802 EXPORT_SYMBOL(xfrm_policy_flush);
804 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
805 int (*func)(struct xfrm_policy *, int, int, void*),
808 struct xfrm_policy *pol;
809 struct xfrm_policy_walk_entry *x;
812 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
813 walk->type != XFRM_POLICY_TYPE_ANY)
816 if (list_empty(&walk->walk.all) && walk->seq != 0)
819 write_lock_bh(&xfrm_policy_lock);
820 if (list_empty(&walk->walk.all))
821 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
823 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
824 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
827 pol = container_of(x, struct xfrm_policy, walk);
828 if (walk->type != XFRM_POLICY_TYPE_ANY &&
829 walk->type != pol->type)
831 error = func(pol, xfrm_policy_id2dir(pol->index),
834 list_move_tail(&walk->walk.all, &x->all);
839 if (walk->seq == 0) {
843 list_del_init(&walk->walk.all);
845 write_unlock_bh(&xfrm_policy_lock);
848 EXPORT_SYMBOL(xfrm_policy_walk);
850 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
852 INIT_LIST_HEAD(&walk->walk.all);
857 EXPORT_SYMBOL(xfrm_policy_walk_init);
859 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
861 if (list_empty(&walk->walk.all))
864 write_lock_bh(&xfrm_policy_lock);
865 list_del(&walk->walk.all);
866 write_unlock_bh(&xfrm_policy_lock);
868 EXPORT_SYMBOL(xfrm_policy_walk_done);
871 * Find policy to apply to this flow.
873 * Returns 0 if policy found, else an -errno.
875 static int xfrm_policy_match(const struct xfrm_policy *pol,
876 const struct flowi *fl,
877 u8 type, u16 family, int dir)
879 const struct xfrm_selector *sel = &pol->selector;
880 int match, ret = -ESRCH;
882 if (pol->family != family ||
883 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
887 match = xfrm_selector_match(sel, fl, family);
889 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
895 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
896 const struct flowi *fl,
900 struct xfrm_policy *pol, *ret;
901 const xfrm_address_t *daddr, *saddr;
902 struct hlist_node *entry;
903 struct hlist_head *chain;
906 daddr = xfrm_flowi_daddr(fl, family);
907 saddr = xfrm_flowi_saddr(fl, family);
908 if (unlikely(!daddr || !saddr))
911 read_lock_bh(&xfrm_policy_lock);
912 chain = policy_hash_direct(net, daddr, saddr, family, dir);
914 hlist_for_each_entry(pol, entry, chain, bydst) {
915 err = xfrm_policy_match(pol, fl, type, family, dir);
925 priority = ret->priority;
929 chain = &net->xfrm.policy_inexact[dir];
930 hlist_for_each_entry(pol, entry, chain, bydst) {
931 err = xfrm_policy_match(pol, fl, type, family, dir);
939 } else if (pol->priority < priority) {
947 read_unlock_bh(&xfrm_policy_lock);
952 static struct xfrm_policy *
953 __xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
955 #ifdef CONFIG_XFRM_SUB_POLICY
956 struct xfrm_policy *pol;
958 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
962 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
965 static struct flow_cache_object *
966 xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
967 u8 dir, struct flow_cache_object *old_obj, void *ctx)
969 struct xfrm_policy *pol;
972 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
974 pol = __xfrm_policy_lookup(net, fl, family, dir);
975 if (IS_ERR_OR_NULL(pol))
976 return ERR_CAST(pol);
978 /* Resolver returns two references:
979 * one for cache and one for caller of flow_cache_lookup() */
985 static inline int policy_to_flow_dir(int dir)
987 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
988 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
989 XFRM_POLICY_FWD == FLOW_DIR_FWD)
995 case XFRM_POLICY_OUT:
997 case XFRM_POLICY_FWD:
1002 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1003 const struct flowi *fl)
1005 struct xfrm_policy *pol;
1007 read_lock_bh(&xfrm_policy_lock);
1008 if ((pol = sk->sk_policy[dir]) != NULL) {
1009 int match = xfrm_selector_match(&pol->selector, fl,
1014 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1018 err = security_xfrm_policy_lookup(pol->security,
1020 policy_to_flow_dir(dir));
1023 else if (err == -ESRCH)
1031 read_unlock_bh(&xfrm_policy_lock);
1035 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1037 struct net *net = xp_net(pol);
1038 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1041 list_add(&pol->walk.all, &net->xfrm.policy_all);
1042 hlist_add_head(&pol->bydst, chain);
1043 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1044 net->xfrm.policy_count[dir]++;
1047 if (xfrm_bydst_should_resize(net, dir, NULL))
1048 schedule_work(&net->xfrm.policy_hash_work);
1051 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1054 struct net *net = xp_net(pol);
1056 if (hlist_unhashed(&pol->bydst))
1059 hlist_del(&pol->bydst);
1060 hlist_del(&pol->byidx);
1061 list_del(&pol->walk.all);
1062 net->xfrm.policy_count[dir]--;
1067 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1069 write_lock_bh(&xfrm_policy_lock);
1070 pol = __xfrm_policy_unlink(pol, dir);
1071 write_unlock_bh(&xfrm_policy_lock);
1073 xfrm_policy_kill(pol);
1078 EXPORT_SYMBOL(xfrm_policy_delete);
1080 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1082 struct net *net = xp_net(pol);
1083 struct xfrm_policy *old_pol;
1085 #ifdef CONFIG_XFRM_SUB_POLICY
1086 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1090 write_lock_bh(&xfrm_policy_lock);
1091 old_pol = sk->sk_policy[dir];
1092 sk->sk_policy[dir] = pol;
1094 pol->curlft.add_time = get_seconds();
1095 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1096 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1099 /* Unlinking succeeds always. This is the only function
1100 * allowed to delete or replace socket policy.
1102 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1103 write_unlock_bh(&xfrm_policy_lock);
1106 xfrm_policy_kill(old_pol);
1111 static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1113 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1116 newp->selector = old->selector;
1117 if (security_xfrm_policy_clone(old->security,
1120 return NULL; /* ENOMEM */
1122 newp->lft = old->lft;
1123 newp->curlft = old->curlft;
1124 newp->mark = old->mark;
1125 newp->action = old->action;
1126 newp->flags = old->flags;
1127 newp->xfrm_nr = old->xfrm_nr;
1128 newp->index = old->index;
1129 newp->type = old->type;
1130 memcpy(newp->xfrm_vec, old->xfrm_vec,
1131 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1132 write_lock_bh(&xfrm_policy_lock);
1133 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1134 write_unlock_bh(&xfrm_policy_lock);
1140 int __xfrm_sk_clone_policy(struct sock *sk)
1142 struct xfrm_policy *p0 = sk->sk_policy[0],
1143 *p1 = sk->sk_policy[1];
1145 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1146 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1148 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1154 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1155 unsigned short family)
1158 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1160 if (unlikely(afinfo == NULL))
1162 err = afinfo->get_saddr(net, local, remote);
1163 xfrm_policy_put_afinfo(afinfo);
1167 /* Resolve list of templates for the flow, given policy. */
1170 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1171 struct xfrm_state **xfrm, unsigned short family)
1173 struct net *net = xp_net(policy);
1176 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1177 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1180 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1181 struct xfrm_state *x;
1182 xfrm_address_t *remote = daddr;
1183 xfrm_address_t *local = saddr;
1184 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1186 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1187 tmpl->mode == XFRM_MODE_BEET) {
1188 remote = &tmpl->id.daddr;
1189 local = &tmpl->saddr;
1190 if (xfrm_addr_any(local, tmpl->encap_family)) {
1191 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
1198 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1200 if (x && x->km.state == XFRM_STATE_VALID) {
1207 error = (x->km.state == XFRM_STATE_ERROR ?
1211 else if (error == -ESRCH)
1214 if (!tmpl->optional)
1220 for (nx--; nx>=0; nx--)
1221 xfrm_state_put(xfrm[nx]);
1226 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1227 struct xfrm_state **xfrm, unsigned short family)
1229 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1230 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1236 for (i = 0; i < npols; i++) {
1237 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1242 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1250 /* found states are sorted for outbound processing */
1252 xfrm_state_sort(xfrm, tpp, cnx, family);
1257 for (cnx--; cnx>=0; cnx--)
1258 xfrm_state_put(tpp[cnx]);
1263 /* Check that the bundle accepts the flow and its components are
1267 static inline int xfrm_get_tos(const struct flowi *fl, int family)
1269 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1275 tos = afinfo->get_tos(fl);
1277 xfrm_policy_put_afinfo(afinfo);
1282 static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1284 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1285 struct dst_entry *dst = &xdst->u.dst;
1287 if (xdst->route == NULL) {
1288 /* Dummy bundle - if it has xfrms we were not
1289 * able to build bundle as template resolution failed.
1290 * It means we need to try again resolving. */
1291 if (xdst->num_xfrms > 0)
1295 if (stale_bundle(dst))
1303 static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1305 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1306 struct dst_entry *dst = &xdst->u.dst;
1310 if (stale_bundle(dst))
1316 static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1318 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1319 struct dst_entry *dst = &xdst->u.dst;
1324 static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1325 .get = xfrm_bundle_flo_get,
1326 .check = xfrm_bundle_flo_check,
1327 .delete = xfrm_bundle_flo_delete,
1330 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1332 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1333 struct dst_ops *dst_ops;
1334 struct xfrm_dst *xdst;
1337 return ERR_PTR(-EINVAL);
1341 dst_ops = &net->xfrm.xfrm4_dst_ops;
1343 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1345 dst_ops = &net->xfrm.xfrm6_dst_ops;
1351 xdst = dst_alloc(dst_ops, 0);
1352 xfrm_policy_put_afinfo(afinfo);
1355 xdst->flo.ops = &xfrm_bundle_fc_ops;
1357 xdst = ERR_PTR(-ENOBUFS);
1362 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1365 struct xfrm_policy_afinfo *afinfo =
1366 xfrm_policy_get_afinfo(dst->ops->family);
1372 err = afinfo->init_path(path, dst, nfheader_len);
1374 xfrm_policy_put_afinfo(afinfo);
1379 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1380 const struct flowi *fl)
1382 struct xfrm_policy_afinfo *afinfo =
1383 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1389 err = afinfo->fill_dst(xdst, dev, fl);
1391 xfrm_policy_put_afinfo(afinfo);
1397 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1398 * all the metrics... Shortly, bundle a bundle.
1401 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1402 struct xfrm_state **xfrm, int nx,
1403 const struct flowi *fl,
1404 struct dst_entry *dst)
1406 struct net *net = xp_net(policy);
1407 unsigned long now = jiffies;
1408 struct net_device *dev;
1409 struct dst_entry *dst_prev = NULL;
1410 struct dst_entry *dst0 = NULL;
1414 int nfheader_len = 0;
1415 int trailer_len = 0;
1417 int family = policy->selector.family;
1418 xfrm_address_t saddr, daddr;
1420 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1422 tos = xfrm_get_tos(fl, family);
1429 for (; i < nx; i++) {
1430 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1431 struct dst_entry *dst1 = &xdst->u.dst;
1433 err = PTR_ERR(xdst);
1442 dst_prev->child = dst_clone(dst1);
1443 dst1->flags |= DST_NOHASH;
1447 dst_copy_metrics(dst1, dst);
1449 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1450 family = xfrm[i]->props.family;
1451 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1459 dst1->xfrm = xfrm[i];
1460 xdst->xfrm_genid = xfrm[i]->genid;
1462 dst1->obsolete = -1;
1463 dst1->flags |= DST_HOST;
1464 dst1->lastuse = now;
1466 dst1->input = dst_discard;
1467 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1469 dst1->next = dst_prev;
1472 header_len += xfrm[i]->props.header_len;
1473 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1474 nfheader_len += xfrm[i]->props.header_len;
1475 trailer_len += xfrm[i]->props.trailer_len;
1478 dst_prev->child = dst;
1486 /* Copy neighbour for reachability confirmation */
1487 dst0->neighbour = neigh_clone(dst->neighbour);
1489 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1490 xfrm_init_pmtu(dst_prev);
1492 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1493 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1495 err = xfrm_fill_dst(xdst, dev, fl);
1499 dst_prev->header_len = header_len;
1500 dst_prev->trailer_len = trailer_len;
1501 header_len -= xdst->u.dst.xfrm->props.header_len;
1502 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1510 xfrm_state_put(xfrm[i]);
1514 dst0 = ERR_PTR(err);
1519 xfrm_dst_alloc_copy(void **target, const void *src, int size)
1522 *target = kmalloc(size, GFP_ATOMIC);
1526 memcpy(*target, src, size);
1531 xfrm_dst_update_parent(struct dst_entry *dst, const struct xfrm_selector *sel)
1533 #ifdef CONFIG_XFRM_SUB_POLICY
1534 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1535 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1543 xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
1545 #ifdef CONFIG_XFRM_SUB_POLICY
1546 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1547 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1553 static int xfrm_expand_policies(const struct flowi *fl, u16 family,
1554 struct xfrm_policy **pols,
1555 int *num_pols, int *num_xfrms)
1559 if (*num_pols == 0 || !pols[0]) {
1564 if (IS_ERR(pols[0]))
1565 return PTR_ERR(pols[0]);
1567 *num_xfrms = pols[0]->xfrm_nr;
1569 #ifdef CONFIG_XFRM_SUB_POLICY
1570 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1571 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1572 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1573 XFRM_POLICY_TYPE_MAIN,
1577 if (IS_ERR(pols[1])) {
1578 xfrm_pols_put(pols, *num_pols);
1579 return PTR_ERR(pols[1]);
1582 (*num_xfrms) += pols[1]->xfrm_nr;
1586 for (i = 0; i < *num_pols; i++) {
1587 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1597 static struct xfrm_dst *
1598 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1599 const struct flowi *fl, u16 family,
1600 struct dst_entry *dst_orig)
1602 struct net *net = xp_net(pols[0]);
1603 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1604 struct dst_entry *dst;
1605 struct xfrm_dst *xdst;
1608 /* Try to instantiate a bundle */
1609 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1611 if (err != 0 && err != -EAGAIN)
1612 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1613 return ERR_PTR(err);
1616 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1618 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1619 return ERR_CAST(dst);
1622 xdst = (struct xfrm_dst *)dst;
1623 xdst->num_xfrms = err;
1625 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1627 err = xfrm_dst_update_origin(dst, fl);
1628 if (unlikely(err)) {
1630 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1631 return ERR_PTR(err);
1634 xdst->num_pols = num_pols;
1635 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1636 xdst->policy_genid = atomic_read(&pols[0]->genid);
1641 static struct flow_cache_object *
1642 xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
1643 struct flow_cache_object *oldflo, void *ctx)
1645 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1646 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1647 struct xfrm_dst *xdst, *new_xdst;
1648 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1650 /* Check if the policies from old bundle are usable */
1653 xdst = container_of(oldflo, struct xfrm_dst, flo);
1654 num_pols = xdst->num_pols;
1655 num_xfrms = xdst->num_xfrms;
1657 for (i = 0; i < num_pols; i++) {
1658 pols[i] = xdst->pols[i];
1659 pol_dead |= pols[i]->walk.dead;
1662 dst_free(&xdst->u.dst);
1670 /* Resolve policies to use if we couldn't get them from
1671 * previous cache entry */
1674 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1675 err = xfrm_expand_policies(fl, family, pols,
1676 &num_pols, &num_xfrms);
1682 goto make_dummy_bundle;
1685 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1686 if (IS_ERR(new_xdst)) {
1687 err = PTR_ERR(new_xdst);
1691 goto make_dummy_bundle;
1692 dst_hold(&xdst->u.dst);
1694 } else if (new_xdst == NULL) {
1697 goto make_dummy_bundle;
1698 xdst->num_xfrms = 0;
1699 dst_hold(&xdst->u.dst);
1703 /* Kill the previous bundle */
1705 /* The policies were stolen for newly generated bundle */
1707 dst_free(&xdst->u.dst);
1710 /* Flow cache does not have reference, it dst_free()'s,
1711 * but we do need to return one reference for original caller */
1712 dst_hold(&new_xdst->u.dst);
1713 return &new_xdst->flo;
1716 /* We found policies, but there's no bundles to instantiate:
1717 * either because the policy blocks, has no transformations or
1718 * we could not build template (no xfrm_states).*/
1719 xdst = xfrm_alloc_dst(net, family);
1721 xfrm_pols_put(pols, num_pols);
1722 return ERR_CAST(xdst);
1724 xdst->num_pols = num_pols;
1725 xdst->num_xfrms = num_xfrms;
1726 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1728 dst_hold(&xdst->u.dst);
1732 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1735 dst_free(&xdst->u.dst);
1737 xfrm_pols_put(pols, num_pols);
1738 return ERR_PTR(err);
1741 static struct dst_entry *make_blackhole(struct net *net, u16 family,
1742 struct dst_entry *dst_orig)
1744 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1745 struct dst_entry *ret;
1748 dst_release(dst_orig);
1749 ret = ERR_PTR(-EINVAL);
1751 ret = afinfo->blackhole_route(net, dst_orig);
1753 xfrm_policy_put_afinfo(afinfo);
1758 /* Main function: finds/creates a bundle for given flow.
1760 * At the moment we eat a raw IP route. Mostly to speed up lookups
1761 * on interfaces with disabled IPsec.
1763 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
1764 const struct flowi *fl,
1765 struct sock *sk, int flags)
1767 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1768 struct flow_cache_object *flo;
1769 struct xfrm_dst *xdst;
1770 struct dst_entry *dst, *route;
1771 u16 family = dst_orig->ops->family;
1772 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1773 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1780 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1782 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1783 err = xfrm_expand_policies(fl, family, pols,
1784 &num_pols, &num_xfrms);
1789 if (num_xfrms <= 0) {
1790 drop_pols = num_pols;
1794 xdst = xfrm_resolve_and_create_bundle(
1798 xfrm_pols_put(pols, num_pols);
1799 err = PTR_ERR(xdst);
1801 } else if (xdst == NULL) {
1803 drop_pols = num_pols;
1807 dst_hold(&xdst->u.dst);
1809 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1810 xdst->u.dst.next = xfrm_policy_sk_bundles;
1811 xfrm_policy_sk_bundles = &xdst->u.dst;
1812 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1814 route = xdst->route;
1819 /* To accelerate a bit... */
1820 if ((dst_orig->flags & DST_NOXFRM) ||
1821 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1824 flo = flow_cache_lookup(net, fl, family, dir,
1825 xfrm_bundle_lookup, dst_orig);
1832 xdst = container_of(flo, struct xfrm_dst, flo);
1834 num_pols = xdst->num_pols;
1835 num_xfrms = xdst->num_xfrms;
1836 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1837 route = xdst->route;
1841 if (route == NULL && num_xfrms > 0) {
1842 /* The only case when xfrm_bundle_lookup() returns a
1843 * bundle with null route, is when the template could
1844 * not be resolved. It means policies are there, but
1845 * bundle could not be created, since we don't yet
1846 * have the xfrm_state's. We need to wait for KM to
1847 * negotiate new SA's or bail out with error.*/
1848 if (net->xfrm.sysctl_larval_drop) {
1849 /* EREMOTE tells the caller to generate
1850 * a one-shot blackhole route. */
1852 xfrm_pols_put(pols, drop_pols);
1853 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1855 return make_blackhole(net, family, dst_orig);
1857 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
1858 DECLARE_WAITQUEUE(wait, current);
1860 add_wait_queue(&net->xfrm.km_waitq, &wait);
1861 set_current_state(TASK_INTERRUPTIBLE);
1863 set_current_state(TASK_RUNNING);
1864 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1866 if (!signal_pending(current)) {
1875 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1883 if ((flags & XFRM_LOOKUP_ICMP) &&
1884 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1889 for (i = 0; i < num_pols; i++)
1890 pols[i]->curlft.use_time = get_seconds();
1892 if (num_xfrms < 0) {
1893 /* Prohibit the flow */
1894 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1897 } else if (num_xfrms > 0) {
1898 /* Flow transformed */
1899 dst_release(dst_orig);
1901 /* Flow passes untransformed */
1906 xfrm_pols_put(pols, drop_pols);
1910 if (!(flags & XFRM_LOOKUP_ICMP)) {
1918 dst_release(dst_orig);
1919 xfrm_pols_put(pols, drop_pols);
1920 return ERR_PTR(err);
1922 EXPORT_SYMBOL(xfrm_lookup);
1925 xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
1927 struct xfrm_state *x;
1929 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1931 x = skb->sp->xvec[idx];
1932 if (!x->type->reject)
1934 return x->type->reject(x, skb, fl);
1937 /* When skb is transformed back to its "native" form, we have to
1938 * check policy restrictions. At the moment we make this in maximally
1939 * stupid way. Shame on me. :-) Of course, connected sockets must
1940 * have policy cached at them.
1944 xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1945 unsigned short family)
1947 if (xfrm_state_kern(x))
1948 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1949 return x->id.proto == tmpl->id.proto &&
1950 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1951 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1952 x->props.mode == tmpl->mode &&
1953 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1954 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1955 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1956 xfrm_state_addr_cmp(tmpl, x, family));
1960 * 0 or more than 0 is returned when validation is succeeded (either bypass
1961 * because of optional transport mode, or next index of the mathced secpath
1962 * state with the template.
1963 * -1 is returned when no matching template is found.
1964 * Otherwise "-2 - errored_index" is returned.
1967 xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1968 unsigned short family)
1972 if (tmpl->optional) {
1973 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1977 for (; idx < sp->len; idx++) {
1978 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1980 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1989 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1990 unsigned int family, int reverse)
1992 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1995 if (unlikely(afinfo == NULL))
1996 return -EAFNOSUPPORT;
1998 afinfo->decode_session(skb, fl, reverse);
1999 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
2000 xfrm_policy_put_afinfo(afinfo);
2003 EXPORT_SYMBOL(__xfrm_decode_session);
2005 static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
2007 for (; k < sp->len; k++) {
2008 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
2017 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
2018 unsigned short family)
2020 struct net *net = dev_net(skb->dev);
2021 struct xfrm_policy *pol;
2022 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2031 reverse = dir & ~XFRM_POLICY_MASK;
2032 dir &= XFRM_POLICY_MASK;
2033 fl_dir = policy_to_flow_dir(dir);
2035 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2036 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2040 nf_nat_decode_session(skb, &fl, family);
2042 /* First, check used SA against their selectors. */
2046 for (i=skb->sp->len-1; i>=0; i--) {
2047 struct xfrm_state *x = skb->sp->xvec[i];
2048 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2049 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2056 if (sk && sk->sk_policy[dir]) {
2057 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2059 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2065 struct flow_cache_object *flo;
2067 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2068 xfrm_policy_lookup, NULL);
2069 if (IS_ERR_OR_NULL(flo))
2070 pol = ERR_CAST(flo);
2072 pol = container_of(flo, struct xfrm_policy, flo);
2076 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2081 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2082 xfrm_secpath_reject(xerr_idx, skb, &fl);
2083 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2089 pol->curlft.use_time = get_seconds();
2093 #ifdef CONFIG_XFRM_SUB_POLICY
2094 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2095 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2099 if (IS_ERR(pols[1])) {
2100 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2103 pols[1]->curlft.use_time = get_seconds();
2109 if (pol->action == XFRM_POLICY_ALLOW) {
2110 struct sec_path *sp;
2111 static struct sec_path dummy;
2112 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2113 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2114 struct xfrm_tmpl **tpp = tp;
2118 if ((sp = skb->sp) == NULL)
2121 for (pi = 0; pi < npols; pi++) {
2122 if (pols[pi] != pol &&
2123 pols[pi]->action != XFRM_POLICY_ALLOW) {
2124 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2127 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2128 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2131 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2132 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2136 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2140 /* For each tunnel xfrm, find the first matching tmpl.
2141 * For each tmpl before that, find corresponding xfrm.
2142 * Order is _important_. Later we will implement
2143 * some barriers, but at the moment barriers
2144 * are implied between each two transformations.
2146 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2147 k = xfrm_policy_ok(tpp[i], sp, k, family);
2150 /* "-2 - errored_index" returned */
2152 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2157 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2158 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2162 xfrm_pols_put(pols, npols);
2165 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2168 xfrm_secpath_reject(xerr_idx, skb, &fl);
2170 xfrm_pols_put(pols, npols);
2173 EXPORT_SYMBOL(__xfrm_policy_check);
2175 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2177 struct net *net = dev_net(skb->dev);
2179 struct dst_entry *dst;
2182 if (xfrm_decode_session(skb, &fl, family) < 0) {
2183 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2189 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2194 skb_dst_set(skb, dst);
2197 EXPORT_SYMBOL(__xfrm_route_forward);
2199 /* Optimize later using cookies and generation ids. */
2201 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2203 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2204 * to "-1" to force all XFRM destinations to get validated by
2205 * dst_ops->check on every use. We do this because when a
2206 * normal route referenced by an XFRM dst is obsoleted we do
2207 * not go looking around for all parent referencing XFRM dsts
2208 * so that we can invalidate them. It is just too much work.
2209 * Instead we make the checks here on every use. For example:
2211 * XFRM dst A --> IPv4 dst X
2213 * X is the "xdst->route" of A (X is also the "dst->path" of A
2214 * in this example). If X is marked obsolete, "A" will not
2215 * notice. That's what we are validating here via the
2216 * stale_bundle() check.
2218 * When a policy's bundle is pruned, we dst_free() the XFRM
2219 * dst which causes it's ->obsolete field to be set to a
2220 * positive non-zero integer. If an XFRM dst has been pruned
2221 * like this, we want to force a new route lookup.
2223 if (dst->obsolete < 0 && !stale_bundle(dst))
2229 static int stale_bundle(struct dst_entry *dst)
2231 return !xfrm_bundle_ok((struct xfrm_dst *)dst, AF_UNSPEC);
2234 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2236 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2237 dst->dev = dev_net(dev)->loopback_dev;
2242 EXPORT_SYMBOL(xfrm_dst_ifdown);
2244 static void xfrm_link_failure(struct sk_buff *skb)
2246 /* Impossible. Such dst must be popped before reaches point of failure. */
2249 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2252 if (dst->obsolete) {
2260 static void __xfrm_garbage_collect(struct net *net)
2262 struct dst_entry *head, *next;
2266 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2267 head = xfrm_policy_sk_bundles;
2268 xfrm_policy_sk_bundles = NULL;
2269 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2278 static void xfrm_init_pmtu(struct dst_entry *dst)
2281 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2282 u32 pmtu, route_mtu_cached;
2284 pmtu = dst_mtu(dst->child);
2285 xdst->child_mtu_cached = pmtu;
2287 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2289 route_mtu_cached = dst_mtu(xdst->route);
2290 xdst->route_mtu_cached = route_mtu_cached;
2292 if (pmtu > route_mtu_cached)
2293 pmtu = route_mtu_cached;
2295 dst_metric_set(dst, RTAX_MTU, pmtu);
2296 } while ((dst = dst->next));
2299 /* Check that the bundle accepts the flow and its components are
2303 static int xfrm_bundle_ok(struct xfrm_dst *first, int family)
2305 struct dst_entry *dst = &first->u.dst;
2306 struct xfrm_dst *last;
2309 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2310 (dst->dev && !netif_running(dst->dev)))
2316 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2318 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2320 if (xdst->xfrm_genid != dst->xfrm->genid)
2322 if (xdst->num_pols > 0 &&
2323 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2326 mtu = dst_mtu(dst->child);
2327 if (xdst->child_mtu_cached != mtu) {
2329 xdst->child_mtu_cached = mtu;
2332 if (!dst_check(xdst->route, xdst->route_cookie))
2334 mtu = dst_mtu(xdst->route);
2335 if (xdst->route_mtu_cached != mtu) {
2337 xdst->route_mtu_cached = mtu;
2341 } while (dst->xfrm);
2346 mtu = last->child_mtu_cached;
2350 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2351 if (mtu > last->route_mtu_cached)
2352 mtu = last->route_mtu_cached;
2353 dst_metric_set(dst, RTAX_MTU, mtu);
2358 last = (struct xfrm_dst *)last->u.dst.next;
2359 last->child_mtu_cached = mtu;
2365 static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2367 return dst_metric_advmss(dst->path);
2370 static unsigned int xfrm_default_mtu(const struct dst_entry *dst)
2372 return dst_mtu(dst->path);
2375 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2379 if (unlikely(afinfo == NULL))
2381 if (unlikely(afinfo->family >= NPROTO))
2382 return -EAFNOSUPPORT;
2383 write_lock_bh(&xfrm_policy_afinfo_lock);
2384 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2387 struct dst_ops *dst_ops = afinfo->dst_ops;
2388 if (likely(dst_ops->kmem_cachep == NULL))
2389 dst_ops->kmem_cachep = xfrm_dst_cache;
2390 if (likely(dst_ops->check == NULL))
2391 dst_ops->check = xfrm_dst_check;
2392 if (likely(dst_ops->default_advmss == NULL))
2393 dst_ops->default_advmss = xfrm_default_advmss;
2394 if (likely(dst_ops->default_mtu == NULL))
2395 dst_ops->default_mtu = xfrm_default_mtu;
2396 if (likely(dst_ops->negative_advice == NULL))
2397 dst_ops->negative_advice = xfrm_negative_advice;
2398 if (likely(dst_ops->link_failure == NULL))
2399 dst_ops->link_failure = xfrm_link_failure;
2400 if (likely(afinfo->garbage_collect == NULL))
2401 afinfo->garbage_collect = __xfrm_garbage_collect;
2402 xfrm_policy_afinfo[afinfo->family] = afinfo;
2404 write_unlock_bh(&xfrm_policy_afinfo_lock);
2408 struct dst_ops *xfrm_dst_ops;
2410 switch (afinfo->family) {
2412 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2414 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2416 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2422 *xfrm_dst_ops = *afinfo->dst_ops;
2428 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2430 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2433 if (unlikely(afinfo == NULL))
2435 if (unlikely(afinfo->family >= NPROTO))
2436 return -EAFNOSUPPORT;
2437 write_lock_bh(&xfrm_policy_afinfo_lock);
2438 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2439 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2442 struct dst_ops *dst_ops = afinfo->dst_ops;
2443 xfrm_policy_afinfo[afinfo->family] = NULL;
2444 dst_ops->kmem_cachep = NULL;
2445 dst_ops->check = NULL;
2446 dst_ops->negative_advice = NULL;
2447 dst_ops->link_failure = NULL;
2448 afinfo->garbage_collect = NULL;
2451 write_unlock_bh(&xfrm_policy_afinfo_lock);
2454 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2456 static void __net_init xfrm_dst_ops_init(struct net *net)
2458 struct xfrm_policy_afinfo *afinfo;
2460 read_lock_bh(&xfrm_policy_afinfo_lock);
2461 afinfo = xfrm_policy_afinfo[AF_INET];
2463 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2464 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2465 afinfo = xfrm_policy_afinfo[AF_INET6];
2467 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2469 read_unlock_bh(&xfrm_policy_afinfo_lock);
2472 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2474 struct xfrm_policy_afinfo *afinfo;
2475 if (unlikely(family >= NPROTO))
2477 read_lock(&xfrm_policy_afinfo_lock);
2478 afinfo = xfrm_policy_afinfo[family];
2479 if (unlikely(!afinfo))
2480 read_unlock(&xfrm_policy_afinfo_lock);
2484 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2486 read_unlock(&xfrm_policy_afinfo_lock);
2489 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2491 struct net_device *dev = ptr;
2495 __xfrm_garbage_collect(dev_net(dev));
2500 static struct notifier_block xfrm_dev_notifier = {
2501 .notifier_call = xfrm_dev_event,
2504 #ifdef CONFIG_XFRM_STATISTICS
2505 static int __net_init xfrm_statistics_init(struct net *net)
2509 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2510 sizeof(struct linux_xfrm_mib),
2511 __alignof__(struct linux_xfrm_mib)) < 0)
2513 rv = xfrm_proc_init(net);
2515 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2519 static void xfrm_statistics_fini(struct net *net)
2521 xfrm_proc_fini(net);
2522 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2525 static int __net_init xfrm_statistics_init(struct net *net)
2530 static void xfrm_statistics_fini(struct net *net)
2535 static int __net_init xfrm_policy_init(struct net *net)
2537 unsigned int hmask, sz;
2540 if (net_eq(net, &init_net))
2541 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2542 sizeof(struct xfrm_dst),
2543 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2547 sz = (hmask+1) * sizeof(struct hlist_head);
2549 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2550 if (!net->xfrm.policy_byidx)
2552 net->xfrm.policy_idx_hmask = hmask;
2554 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2555 struct xfrm_policy_hash *htab;
2557 net->xfrm.policy_count[dir] = 0;
2558 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2560 htab = &net->xfrm.policy_bydst[dir];
2561 htab->table = xfrm_hash_alloc(sz);
2564 htab->hmask = hmask;
2567 INIT_LIST_HEAD(&net->xfrm.policy_all);
2568 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2569 if (net_eq(net, &init_net))
2570 register_netdevice_notifier(&xfrm_dev_notifier);
2574 for (dir--; dir >= 0; dir--) {
2575 struct xfrm_policy_hash *htab;
2577 htab = &net->xfrm.policy_bydst[dir];
2578 xfrm_hash_free(htab->table, sz);
2580 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2585 static void xfrm_policy_fini(struct net *net)
2587 struct xfrm_audit audit_info;
2591 flush_work(&net->xfrm.policy_hash_work);
2592 #ifdef CONFIG_XFRM_SUB_POLICY
2593 audit_info.loginuid = -1;
2594 audit_info.sessionid = -1;
2595 audit_info.secid = 0;
2596 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2598 audit_info.loginuid = -1;
2599 audit_info.sessionid = -1;
2600 audit_info.secid = 0;
2601 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2603 WARN_ON(!list_empty(&net->xfrm.policy_all));
2605 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2606 struct xfrm_policy_hash *htab;
2608 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2610 htab = &net->xfrm.policy_bydst[dir];
2611 sz = (htab->hmask + 1);
2612 WARN_ON(!hlist_empty(htab->table));
2613 xfrm_hash_free(htab->table, sz);
2616 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2617 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2618 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2621 static int __net_init xfrm_net_init(struct net *net)
2625 rv = xfrm_statistics_init(net);
2627 goto out_statistics;
2628 rv = xfrm_state_init(net);
2631 rv = xfrm_policy_init(net);
2634 xfrm_dst_ops_init(net);
2635 rv = xfrm_sysctl_init(net);
2641 xfrm_policy_fini(net);
2643 xfrm_state_fini(net);
2645 xfrm_statistics_fini(net);
2650 static void __net_exit xfrm_net_exit(struct net *net)
2652 xfrm_sysctl_fini(net);
2653 xfrm_policy_fini(net);
2654 xfrm_state_fini(net);
2655 xfrm_statistics_fini(net);
2658 static struct pernet_operations __net_initdata xfrm_net_ops = {
2659 .init = xfrm_net_init,
2660 .exit = xfrm_net_exit,
2663 void __init xfrm_init(void)
2665 register_pernet_subsys(&xfrm_net_ops);
2669 #ifdef CONFIG_AUDITSYSCALL
2670 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2671 struct audit_buffer *audit_buf)
2673 struct xfrm_sec_ctx *ctx = xp->security;
2674 struct xfrm_selector *sel = &xp->selector;
2677 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2678 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2680 switch(sel->family) {
2682 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2683 if (sel->prefixlen_s != 32)
2684 audit_log_format(audit_buf, " src_prefixlen=%d",
2686 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2687 if (sel->prefixlen_d != 32)
2688 audit_log_format(audit_buf, " dst_prefixlen=%d",
2692 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2693 if (sel->prefixlen_s != 128)
2694 audit_log_format(audit_buf, " src_prefixlen=%d",
2696 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2697 if (sel->prefixlen_d != 128)
2698 audit_log_format(audit_buf, " dst_prefixlen=%d",
2704 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2705 uid_t auid, u32 sessionid, u32 secid)
2707 struct audit_buffer *audit_buf;
2709 audit_buf = xfrm_audit_start("SPD-add");
2710 if (audit_buf == NULL)
2712 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2713 audit_log_format(audit_buf, " res=%u", result);
2714 xfrm_audit_common_policyinfo(xp, audit_buf);
2715 audit_log_end(audit_buf);
2717 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2719 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2720 uid_t auid, u32 sessionid, u32 secid)
2722 struct audit_buffer *audit_buf;
2724 audit_buf = xfrm_audit_start("SPD-delete");
2725 if (audit_buf == NULL)
2727 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2728 audit_log_format(audit_buf, " res=%u", result);
2729 xfrm_audit_common_policyinfo(xp, audit_buf);
2730 audit_log_end(audit_buf);
2732 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2735 #ifdef CONFIG_XFRM_MIGRATE
2736 static int xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
2737 const struct xfrm_selector *sel_tgt)
2739 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2740 if (sel_tgt->family == sel_cmp->family &&
2741 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2742 sel_cmp->family) == 0 &&
2743 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2744 sel_cmp->family) == 0 &&
2745 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2746 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2750 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2757 static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
2760 struct xfrm_policy *pol, *ret = NULL;
2761 struct hlist_node *entry;
2762 struct hlist_head *chain;
2765 read_lock_bh(&xfrm_policy_lock);
2766 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2767 hlist_for_each_entry(pol, entry, chain, bydst) {
2768 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2769 pol->type == type) {
2771 priority = ret->priority;
2775 chain = &init_net.xfrm.policy_inexact[dir];
2776 hlist_for_each_entry(pol, entry, chain, bydst) {
2777 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2778 pol->type == type &&
2779 pol->priority < priority) {
2788 read_unlock_bh(&xfrm_policy_lock);
2793 static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
2797 if (t->mode == m->mode && t->id.proto == m->proto &&
2798 (m->reqid == 0 || t->reqid == m->reqid)) {
2800 case XFRM_MODE_TUNNEL:
2801 case XFRM_MODE_BEET:
2802 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2803 m->old_family) == 0 &&
2804 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2805 m->old_family) == 0) {
2809 case XFRM_MODE_TRANSPORT:
2810 /* in case of transport mode, template does not store
2811 any IP addresses, hence we just compare mode and
2822 /* update endpoint address(es) of template(s) */
2823 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2824 struct xfrm_migrate *m, int num_migrate)
2826 struct xfrm_migrate *mp;
2829 write_lock_bh(&pol->lock);
2830 if (unlikely(pol->walk.dead)) {
2831 /* target policy has been deleted */
2832 write_unlock_bh(&pol->lock);
2836 for (i = 0; i < pol->xfrm_nr; i++) {
2837 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2838 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2841 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2842 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2844 /* update endpoints */
2845 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2846 sizeof(pol->xfrm_vec[i].id.daddr));
2847 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2848 sizeof(pol->xfrm_vec[i].saddr));
2849 pol->xfrm_vec[i].encap_family = mp->new_family;
2851 atomic_inc(&pol->genid);
2855 write_unlock_bh(&pol->lock);
2863 static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
2867 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2870 for (i = 0; i < num_migrate; i++) {
2871 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2872 m[i].old_family) == 0) &&
2873 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2874 m[i].old_family) == 0))
2876 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2877 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2880 /* check if there is any duplicated entry */
2881 for (j = i + 1; j < num_migrate; j++) {
2882 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2883 sizeof(m[i].old_daddr)) &&
2884 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2885 sizeof(m[i].old_saddr)) &&
2886 m[i].proto == m[j].proto &&
2887 m[i].mode == m[j].mode &&
2888 m[i].reqid == m[j].reqid &&
2889 m[i].old_family == m[j].old_family)
2897 int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2898 struct xfrm_migrate *m, int num_migrate,
2899 struct xfrm_kmaddress *k)
2901 int i, err, nx_cur = 0, nx_new = 0;
2902 struct xfrm_policy *pol = NULL;
2903 struct xfrm_state *x, *xc;
2904 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2905 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2906 struct xfrm_migrate *mp;
2908 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2911 /* Stage 1 - find policy */
2912 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2917 /* Stage 2 - find and update state(s) */
2918 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2919 if ((x = xfrm_migrate_state_find(mp))) {
2922 if ((xc = xfrm_state_migrate(x, mp))) {
2932 /* Stage 3 - update policy */
2933 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2936 /* Stage 4 - delete old state(s) */
2938 xfrm_states_put(x_cur, nx_cur);
2939 xfrm_states_delete(x_cur, nx_cur);
2942 /* Stage 5 - announce */
2943 km_migrate(sel, dir, type, m, num_migrate, k);
2955 xfrm_states_put(x_cur, nx_cur);
2957 xfrm_states_delete(x_new, nx_new);
2961 EXPORT_SYMBOL(xfrm_migrate);