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);
54 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
58 __xfrm4_selector_match(struct xfrm_selector *sel, struct flowi *fl)
60 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
61 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
62 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
63 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
64 (fl->proto == sel->proto || !sel->proto) &&
65 (fl->oif == sel->ifindex || !sel->ifindex);
69 __xfrm6_selector_match(struct xfrm_selector *sel, struct flowi *fl)
71 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
72 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
73 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
74 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
75 (fl->proto == sel->proto || !sel->proto) &&
76 (fl->oif == sel->ifindex || !sel->ifindex);
79 int xfrm_selector_match(struct xfrm_selector *sel, struct flowi *fl,
80 unsigned short family)
84 return __xfrm4_selector_match(sel, fl);
86 return __xfrm6_selector_match(sel, fl);
91 static inline struct dst_entry *__xfrm_dst_lookup(struct net *net, int tos,
92 xfrm_address_t *saddr,
93 xfrm_address_t *daddr,
96 struct xfrm_policy_afinfo *afinfo;
97 struct dst_entry *dst;
99 afinfo = xfrm_policy_get_afinfo(family);
100 if (unlikely(afinfo == NULL))
101 return ERR_PTR(-EAFNOSUPPORT);
103 dst = afinfo->dst_lookup(net, tos, saddr, daddr);
105 xfrm_policy_put_afinfo(afinfo);
110 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
111 xfrm_address_t *prev_saddr,
112 xfrm_address_t *prev_daddr,
115 struct net *net = xs_net(x);
116 xfrm_address_t *saddr = &x->props.saddr;
117 xfrm_address_t *daddr = &x->id.daddr;
118 struct dst_entry *dst;
120 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
124 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
129 dst = __xfrm_dst_lookup(net, tos, saddr, daddr, family);
132 if (prev_saddr != saddr)
133 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
134 if (prev_daddr != daddr)
135 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
141 static inline unsigned long make_jiffies(long secs)
143 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
144 return MAX_SCHEDULE_TIMEOUT-1;
149 static void xfrm_policy_timer(unsigned long data)
151 struct xfrm_policy *xp = (struct xfrm_policy*)data;
152 unsigned long now = get_seconds();
153 long next = LONG_MAX;
157 read_lock(&xp->lock);
159 if (unlikely(xp->walk.dead))
162 dir = xfrm_policy_id2dir(xp->index);
164 if (xp->lft.hard_add_expires_seconds) {
165 long tmo = xp->lft.hard_add_expires_seconds +
166 xp->curlft.add_time - now;
172 if (xp->lft.hard_use_expires_seconds) {
173 long tmo = xp->lft.hard_use_expires_seconds +
174 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
180 if (xp->lft.soft_add_expires_seconds) {
181 long tmo = xp->lft.soft_add_expires_seconds +
182 xp->curlft.add_time - now;
185 tmo = XFRM_KM_TIMEOUT;
190 if (xp->lft.soft_use_expires_seconds) {
191 long tmo = xp->lft.soft_use_expires_seconds +
192 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
195 tmo = XFRM_KM_TIMEOUT;
202 km_policy_expired(xp, dir, 0, 0);
203 if (next != LONG_MAX &&
204 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
208 read_unlock(&xp->lock);
213 read_unlock(&xp->lock);
214 if (!xfrm_policy_delete(xp, dir))
215 km_policy_expired(xp, dir, 1, 0);
219 static struct flow_cache_object *xfrm_policy_flo_get(struct flow_cache_object *flo)
221 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
223 if (unlikely(pol->walk.dead))
231 static int xfrm_policy_flo_check(struct flow_cache_object *flo)
233 struct xfrm_policy *pol = container_of(flo, struct xfrm_policy, flo);
235 return !pol->walk.dead;
238 static void xfrm_policy_flo_delete(struct flow_cache_object *flo)
240 xfrm_pol_put(container_of(flo, struct xfrm_policy, flo));
243 static const struct flow_cache_ops xfrm_policy_fc_ops = {
244 .get = xfrm_policy_flo_get,
245 .check = xfrm_policy_flo_check,
246 .delete = xfrm_policy_flo_delete,
249 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
253 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
255 struct xfrm_policy *policy;
257 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
260 write_pnet(&policy->xp_net, net);
261 INIT_LIST_HEAD(&policy->walk.all);
262 INIT_HLIST_NODE(&policy->bydst);
263 INIT_HLIST_NODE(&policy->byidx);
264 rwlock_init(&policy->lock);
265 atomic_set(&policy->refcnt, 1);
266 setup_timer(&policy->timer, xfrm_policy_timer,
267 (unsigned long)policy);
268 policy->flo.ops = &xfrm_policy_fc_ops;
272 EXPORT_SYMBOL(xfrm_policy_alloc);
274 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
276 void xfrm_policy_destroy(struct xfrm_policy *policy)
278 BUG_ON(!policy->walk.dead);
280 if (del_timer(&policy->timer))
283 security_xfrm_policy_free(policy->security);
286 EXPORT_SYMBOL(xfrm_policy_destroy);
288 /* Rule must be locked. Release descentant resources, announce
289 * entry dead. The rule must be unlinked from lists to the moment.
292 static void xfrm_policy_kill(struct xfrm_policy *policy)
294 policy->walk.dead = 1;
296 atomic_inc(&policy->genid);
298 if (del_timer(&policy->timer))
299 xfrm_pol_put(policy);
301 xfrm_pol_put(policy);
304 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
306 static inline unsigned int idx_hash(struct net *net, u32 index)
308 return __idx_hash(index, net->xfrm.policy_idx_hmask);
311 static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
313 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
314 unsigned int hash = __sel_hash(sel, family, hmask);
316 return (hash == hmask + 1 ?
317 &net->xfrm.policy_inexact[dir] :
318 net->xfrm.policy_bydst[dir].table + hash);
321 static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
323 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
324 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
326 return net->xfrm.policy_bydst[dir].table + hash;
329 static void xfrm_dst_hash_transfer(struct hlist_head *list,
330 struct hlist_head *ndsttable,
331 unsigned int nhashmask)
333 struct hlist_node *entry, *tmp, *entry0 = NULL;
334 struct xfrm_policy *pol;
338 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
341 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
342 pol->family, nhashmask);
345 hlist_add_head(&pol->bydst, ndsttable+h);
351 hlist_add_after(entry0, &pol->bydst);
355 if (!hlist_empty(list)) {
361 static void xfrm_idx_hash_transfer(struct hlist_head *list,
362 struct hlist_head *nidxtable,
363 unsigned int nhashmask)
365 struct hlist_node *entry, *tmp;
366 struct xfrm_policy *pol;
368 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
371 h = __idx_hash(pol->index, nhashmask);
372 hlist_add_head(&pol->byidx, nidxtable+h);
376 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
378 return ((old_hmask + 1) << 1) - 1;
381 static void xfrm_bydst_resize(struct net *net, int dir)
383 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
384 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
385 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
386 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
387 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
393 write_lock_bh(&xfrm_policy_lock);
395 for (i = hmask; i >= 0; i--)
396 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
398 net->xfrm.policy_bydst[dir].table = ndst;
399 net->xfrm.policy_bydst[dir].hmask = nhashmask;
401 write_unlock_bh(&xfrm_policy_lock);
403 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
406 static void xfrm_byidx_resize(struct net *net, int total)
408 unsigned int hmask = net->xfrm.policy_idx_hmask;
409 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
410 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
411 struct hlist_head *oidx = net->xfrm.policy_byidx;
412 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
418 write_lock_bh(&xfrm_policy_lock);
420 for (i = hmask; i >= 0; i--)
421 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
423 net->xfrm.policy_byidx = nidx;
424 net->xfrm.policy_idx_hmask = nhashmask;
426 write_unlock_bh(&xfrm_policy_lock);
428 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
431 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
433 unsigned int cnt = net->xfrm.policy_count[dir];
434 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
439 if ((hmask + 1) < xfrm_policy_hashmax &&
446 static inline int xfrm_byidx_should_resize(struct net *net, int total)
448 unsigned int hmask = net->xfrm.policy_idx_hmask;
450 if ((hmask + 1) < xfrm_policy_hashmax &&
457 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
459 read_lock_bh(&xfrm_policy_lock);
460 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
461 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
462 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
463 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
464 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
465 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
466 si->spdhcnt = net->xfrm.policy_idx_hmask;
467 si->spdhmcnt = xfrm_policy_hashmax;
468 read_unlock_bh(&xfrm_policy_lock);
470 EXPORT_SYMBOL(xfrm_spd_getinfo);
472 static DEFINE_MUTEX(hash_resize_mutex);
473 static void xfrm_hash_resize(struct work_struct *work)
475 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
478 mutex_lock(&hash_resize_mutex);
481 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
482 if (xfrm_bydst_should_resize(net, dir, &total))
483 xfrm_bydst_resize(net, dir);
485 if (xfrm_byidx_should_resize(net, total))
486 xfrm_byidx_resize(net, total);
488 mutex_unlock(&hash_resize_mutex);
491 /* Generate new index... KAME seems to generate them ordered by cost
492 * of an absolute inpredictability of ordering of rules. This will not pass. */
493 static u32 xfrm_gen_index(struct net *net, int dir)
495 static u32 idx_generator;
498 struct hlist_node *entry;
499 struct hlist_head *list;
500 struct xfrm_policy *p;
504 idx = (idx_generator | dir);
508 list = net->xfrm.policy_byidx + idx_hash(net, idx);
510 hlist_for_each_entry(p, entry, list, byidx) {
511 if (p->index == idx) {
521 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
523 u32 *p1 = (u32 *) s1;
524 u32 *p2 = (u32 *) s2;
525 int len = sizeof(struct xfrm_selector) / sizeof(u32);
528 for (i = 0; i < len; i++) {
536 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
538 struct net *net = xp_net(policy);
539 struct xfrm_policy *pol;
540 struct xfrm_policy *delpol;
541 struct hlist_head *chain;
542 struct hlist_node *entry, *newpos;
543 u32 mark = policy->mark.v & policy->mark.m;
545 write_lock_bh(&xfrm_policy_lock);
546 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
549 hlist_for_each_entry(pol, entry, chain, bydst) {
550 if (pol->type == policy->type &&
551 !selector_cmp(&pol->selector, &policy->selector) &&
552 (mark & pol->mark.m) == pol->mark.v &&
553 xfrm_sec_ctx_match(pol->security, policy->security) &&
556 write_unlock_bh(&xfrm_policy_lock);
560 if (policy->priority > pol->priority)
562 } else if (policy->priority >= pol->priority) {
563 newpos = &pol->bydst;
570 hlist_add_after(newpos, &policy->bydst);
572 hlist_add_head(&policy->bydst, chain);
573 xfrm_pol_hold(policy);
574 net->xfrm.policy_count[dir]++;
575 atomic_inc(&flow_cache_genid);
577 __xfrm_policy_unlink(delpol, dir);
578 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
579 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
580 policy->curlft.add_time = get_seconds();
581 policy->curlft.use_time = 0;
582 if (!mod_timer(&policy->timer, jiffies + HZ))
583 xfrm_pol_hold(policy);
584 list_add(&policy->walk.all, &net->xfrm.policy_all);
585 write_unlock_bh(&xfrm_policy_lock);
588 xfrm_policy_kill(delpol);
589 else if (xfrm_bydst_should_resize(net, dir, NULL))
590 schedule_work(&net->xfrm.policy_hash_work);
594 EXPORT_SYMBOL(xfrm_policy_insert);
596 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
597 int dir, struct xfrm_selector *sel,
598 struct xfrm_sec_ctx *ctx, int delete,
601 struct xfrm_policy *pol, *ret;
602 struct hlist_head *chain;
603 struct hlist_node *entry;
606 write_lock_bh(&xfrm_policy_lock);
607 chain = policy_hash_bysel(net, sel, sel->family, dir);
609 hlist_for_each_entry(pol, entry, chain, bydst) {
610 if (pol->type == type &&
611 (mark & pol->mark.m) == pol->mark.v &&
612 !selector_cmp(sel, &pol->selector) &&
613 xfrm_sec_ctx_match(ctx, pol->security)) {
616 *err = security_xfrm_policy_delete(
619 write_unlock_bh(&xfrm_policy_lock);
622 __xfrm_policy_unlink(pol, dir);
628 write_unlock_bh(&xfrm_policy_lock);
631 xfrm_policy_kill(ret);
634 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
636 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
637 int dir, u32 id, int delete, int *err)
639 struct xfrm_policy *pol, *ret;
640 struct hlist_head *chain;
641 struct hlist_node *entry;
644 if (xfrm_policy_id2dir(id) != dir)
648 write_lock_bh(&xfrm_policy_lock);
649 chain = net->xfrm.policy_byidx + idx_hash(net, id);
651 hlist_for_each_entry(pol, entry, chain, byidx) {
652 if (pol->type == type && pol->index == id &&
653 (mark & pol->mark.m) == pol->mark.v) {
656 *err = security_xfrm_policy_delete(
659 write_unlock_bh(&xfrm_policy_lock);
662 __xfrm_policy_unlink(pol, dir);
668 write_unlock_bh(&xfrm_policy_lock);
671 xfrm_policy_kill(ret);
674 EXPORT_SYMBOL(xfrm_policy_byid);
676 #ifdef CONFIG_SECURITY_NETWORK_XFRM
678 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
682 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
683 struct xfrm_policy *pol;
684 struct hlist_node *entry;
687 hlist_for_each_entry(pol, entry,
688 &net->xfrm.policy_inexact[dir], bydst) {
689 if (pol->type != type)
691 err = security_xfrm_policy_delete(pol->security);
693 xfrm_audit_policy_delete(pol, 0,
694 audit_info->loginuid,
695 audit_info->sessionid,
700 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
701 hlist_for_each_entry(pol, entry,
702 net->xfrm.policy_bydst[dir].table + i,
704 if (pol->type != type)
706 err = security_xfrm_policy_delete(
709 xfrm_audit_policy_delete(pol, 0,
710 audit_info->loginuid,
711 audit_info->sessionid,
722 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
728 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
730 int dir, err = 0, cnt = 0;
732 write_lock_bh(&xfrm_policy_lock);
734 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
738 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
739 struct xfrm_policy *pol;
740 struct hlist_node *entry;
744 hlist_for_each_entry(pol, entry,
745 &net->xfrm.policy_inexact[dir], bydst) {
746 if (pol->type != type)
748 __xfrm_policy_unlink(pol, dir);
749 write_unlock_bh(&xfrm_policy_lock);
752 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
753 audit_info->sessionid,
756 xfrm_policy_kill(pol);
758 write_lock_bh(&xfrm_policy_lock);
762 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
764 hlist_for_each_entry(pol, entry,
765 net->xfrm.policy_bydst[dir].table + i,
767 if (pol->type != type)
769 __xfrm_policy_unlink(pol, dir);
770 write_unlock_bh(&xfrm_policy_lock);
773 xfrm_audit_policy_delete(pol, 1,
774 audit_info->loginuid,
775 audit_info->sessionid,
777 xfrm_policy_kill(pol);
779 write_lock_bh(&xfrm_policy_lock);
788 write_unlock_bh(&xfrm_policy_lock);
791 EXPORT_SYMBOL(xfrm_policy_flush);
793 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
794 int (*func)(struct xfrm_policy *, int, int, void*),
797 struct xfrm_policy *pol;
798 struct xfrm_policy_walk_entry *x;
801 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
802 walk->type != XFRM_POLICY_TYPE_ANY)
805 if (list_empty(&walk->walk.all) && walk->seq != 0)
808 write_lock_bh(&xfrm_policy_lock);
809 if (list_empty(&walk->walk.all))
810 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
812 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
813 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
816 pol = container_of(x, struct xfrm_policy, walk);
817 if (walk->type != XFRM_POLICY_TYPE_ANY &&
818 walk->type != pol->type)
820 error = func(pol, xfrm_policy_id2dir(pol->index),
823 list_move_tail(&walk->walk.all, &x->all);
828 if (walk->seq == 0) {
832 list_del_init(&walk->walk.all);
834 write_unlock_bh(&xfrm_policy_lock);
837 EXPORT_SYMBOL(xfrm_policy_walk);
839 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
841 INIT_LIST_HEAD(&walk->walk.all);
846 EXPORT_SYMBOL(xfrm_policy_walk_init);
848 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
850 if (list_empty(&walk->walk.all))
853 write_lock_bh(&xfrm_policy_lock);
854 list_del(&walk->walk.all);
855 write_unlock_bh(&xfrm_policy_lock);
857 EXPORT_SYMBOL(xfrm_policy_walk_done);
860 * Find policy to apply to this flow.
862 * Returns 0 if policy found, else an -errno.
864 static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
865 u8 type, u16 family, int dir)
867 struct xfrm_selector *sel = &pol->selector;
868 int match, ret = -ESRCH;
870 if (pol->family != family ||
871 (fl->mark & pol->mark.m) != pol->mark.v ||
875 match = xfrm_selector_match(sel, fl, family);
877 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
883 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
888 struct xfrm_policy *pol, *ret;
889 xfrm_address_t *daddr, *saddr;
890 struct hlist_node *entry;
891 struct hlist_head *chain;
894 daddr = xfrm_flowi_daddr(fl, family);
895 saddr = xfrm_flowi_saddr(fl, family);
896 if (unlikely(!daddr || !saddr))
899 read_lock_bh(&xfrm_policy_lock);
900 chain = policy_hash_direct(net, daddr, saddr, family, dir);
902 hlist_for_each_entry(pol, entry, chain, bydst) {
903 err = xfrm_policy_match(pol, fl, type, family, dir);
913 priority = ret->priority;
917 chain = &net->xfrm.policy_inexact[dir];
918 hlist_for_each_entry(pol, entry, chain, bydst) {
919 err = xfrm_policy_match(pol, fl, type, family, dir);
927 } else if (pol->priority < priority) {
935 read_unlock_bh(&xfrm_policy_lock);
940 static struct xfrm_policy *
941 __xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family, u8 dir)
943 #ifdef CONFIG_XFRM_SUB_POLICY
944 struct xfrm_policy *pol;
946 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
950 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
953 static struct flow_cache_object *
954 xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
955 u8 dir, struct flow_cache_object *old_obj, void *ctx)
957 struct xfrm_policy *pol;
960 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
962 pol = __xfrm_policy_lookup(net, fl, family, dir);
963 if (IS_ERR_OR_NULL(pol))
964 return ERR_CAST(pol);
966 /* Resolver returns two references:
967 * one for cache and one for caller of flow_cache_lookup() */
973 static inline int policy_to_flow_dir(int dir)
975 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
976 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
977 XFRM_POLICY_FWD == FLOW_DIR_FWD)
983 case XFRM_POLICY_OUT:
985 case XFRM_POLICY_FWD:
990 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
992 struct xfrm_policy *pol;
994 read_lock_bh(&xfrm_policy_lock);
995 if ((pol = sk->sk_policy[dir]) != NULL) {
996 int match = xfrm_selector_match(&pol->selector, fl,
1001 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1005 err = security_xfrm_policy_lookup(pol->security,
1007 policy_to_flow_dir(dir));
1010 else if (err == -ESRCH)
1018 read_unlock_bh(&xfrm_policy_lock);
1022 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1024 struct net *net = xp_net(pol);
1025 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1028 list_add(&pol->walk.all, &net->xfrm.policy_all);
1029 hlist_add_head(&pol->bydst, chain);
1030 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1031 net->xfrm.policy_count[dir]++;
1034 if (xfrm_bydst_should_resize(net, dir, NULL))
1035 schedule_work(&net->xfrm.policy_hash_work);
1038 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1041 struct net *net = xp_net(pol);
1043 if (hlist_unhashed(&pol->bydst))
1046 hlist_del(&pol->bydst);
1047 hlist_del(&pol->byidx);
1048 list_del(&pol->walk.all);
1049 net->xfrm.policy_count[dir]--;
1054 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1056 write_lock_bh(&xfrm_policy_lock);
1057 pol = __xfrm_policy_unlink(pol, dir);
1058 write_unlock_bh(&xfrm_policy_lock);
1060 xfrm_policy_kill(pol);
1065 EXPORT_SYMBOL(xfrm_policy_delete);
1067 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1069 struct net *net = xp_net(pol);
1070 struct xfrm_policy *old_pol;
1072 #ifdef CONFIG_XFRM_SUB_POLICY
1073 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1077 write_lock_bh(&xfrm_policy_lock);
1078 old_pol = sk->sk_policy[dir];
1079 sk->sk_policy[dir] = pol;
1081 pol->curlft.add_time = get_seconds();
1082 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1083 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1086 /* Unlinking succeeds always. This is the only function
1087 * allowed to delete or replace socket policy.
1089 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1090 write_unlock_bh(&xfrm_policy_lock);
1093 xfrm_policy_kill(old_pol);
1098 static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1100 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1103 newp->selector = old->selector;
1104 if (security_xfrm_policy_clone(old->security,
1107 return NULL; /* ENOMEM */
1109 newp->lft = old->lft;
1110 newp->curlft = old->curlft;
1111 newp->mark = old->mark;
1112 newp->action = old->action;
1113 newp->flags = old->flags;
1114 newp->xfrm_nr = old->xfrm_nr;
1115 newp->index = old->index;
1116 newp->type = old->type;
1117 memcpy(newp->xfrm_vec, old->xfrm_vec,
1118 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1119 write_lock_bh(&xfrm_policy_lock);
1120 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1121 write_unlock_bh(&xfrm_policy_lock);
1127 int __xfrm_sk_clone_policy(struct sock *sk)
1129 struct xfrm_policy *p0 = sk->sk_policy[0],
1130 *p1 = sk->sk_policy[1];
1132 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1133 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1135 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1141 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1142 unsigned short family)
1145 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1147 if (unlikely(afinfo == NULL))
1149 err = afinfo->get_saddr(net, local, remote);
1150 xfrm_policy_put_afinfo(afinfo);
1154 /* Resolve list of templates for the flow, given policy. */
1157 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1158 struct xfrm_state **xfrm,
1159 unsigned short family)
1161 struct net *net = xp_net(policy);
1164 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1165 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1168 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1169 struct xfrm_state *x;
1170 xfrm_address_t *remote = daddr;
1171 xfrm_address_t *local = saddr;
1172 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1174 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1175 tmpl->mode == XFRM_MODE_BEET) {
1176 remote = &tmpl->id.daddr;
1177 local = &tmpl->saddr;
1178 if (xfrm_addr_any(local, tmpl->encap_family)) {
1179 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
1186 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1188 if (x && x->km.state == XFRM_STATE_VALID) {
1195 error = (x->km.state == XFRM_STATE_ERROR ?
1199 else if (error == -ESRCH)
1202 if (!tmpl->optional)
1208 for (nx--; nx>=0; nx--)
1209 xfrm_state_put(xfrm[nx]);
1214 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1215 struct xfrm_state **xfrm,
1216 unsigned short family)
1218 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1219 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1225 for (i = 0; i < npols; i++) {
1226 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1231 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1239 /* found states are sorted for outbound processing */
1241 xfrm_state_sort(xfrm, tpp, cnx, family);
1246 for (cnx--; cnx>=0; cnx--)
1247 xfrm_state_put(tpp[cnx]);
1252 /* Check that the bundle accepts the flow and its components are
1256 static inline int xfrm_get_tos(struct flowi *fl, int family)
1258 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1264 tos = afinfo->get_tos(fl);
1266 xfrm_policy_put_afinfo(afinfo);
1271 static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1273 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1274 struct dst_entry *dst = &xdst->u.dst;
1276 if (xdst->route == NULL) {
1277 /* Dummy bundle - if it has xfrms we were not
1278 * able to build bundle as template resolution failed.
1279 * It means we need to try again resolving. */
1280 if (xdst->num_xfrms > 0)
1284 if (stale_bundle(dst))
1292 static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1294 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1295 struct dst_entry *dst = &xdst->u.dst;
1299 if (stale_bundle(dst))
1305 static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1307 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1308 struct dst_entry *dst = &xdst->u.dst;
1313 static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1314 .get = xfrm_bundle_flo_get,
1315 .check = xfrm_bundle_flo_check,
1316 .delete = xfrm_bundle_flo_delete,
1319 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1321 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1322 struct dst_ops *dst_ops;
1323 struct xfrm_dst *xdst;
1326 return ERR_PTR(-EINVAL);
1330 dst_ops = &net->xfrm.xfrm4_dst_ops;
1332 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1334 dst_ops = &net->xfrm.xfrm6_dst_ops;
1340 xdst = dst_alloc(dst_ops) ?: ERR_PTR(-ENOBUFS);
1341 xfrm_policy_put_afinfo(afinfo);
1343 xdst->flo.ops = &xfrm_bundle_fc_ops;
1348 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1351 struct xfrm_policy_afinfo *afinfo =
1352 xfrm_policy_get_afinfo(dst->ops->family);
1358 err = afinfo->init_path(path, dst, nfheader_len);
1360 xfrm_policy_put_afinfo(afinfo);
1365 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1368 struct xfrm_policy_afinfo *afinfo =
1369 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1375 err = afinfo->fill_dst(xdst, dev, fl);
1377 xfrm_policy_put_afinfo(afinfo);
1383 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1384 * all the metrics... Shortly, bundle a bundle.
1387 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1388 struct xfrm_state **xfrm, int nx,
1390 struct dst_entry *dst)
1392 struct net *net = xp_net(policy);
1393 unsigned long now = jiffies;
1394 struct net_device *dev;
1395 struct dst_entry *dst_prev = NULL;
1396 struct dst_entry *dst0 = NULL;
1400 int nfheader_len = 0;
1401 int trailer_len = 0;
1403 int family = policy->selector.family;
1404 xfrm_address_t saddr, daddr;
1406 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1408 tos = xfrm_get_tos(fl, family);
1415 for (; i < nx; i++) {
1416 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1417 struct dst_entry *dst1 = &xdst->u.dst;
1419 err = PTR_ERR(xdst);
1428 dst_prev->child = dst_clone(dst1);
1429 dst1->flags |= DST_NOHASH;
1433 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1435 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1436 family = xfrm[i]->props.family;
1437 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1445 dst1->xfrm = xfrm[i];
1446 xdst->xfrm_genid = xfrm[i]->genid;
1448 dst1->obsolete = -1;
1449 dst1->flags |= DST_HOST;
1450 dst1->lastuse = now;
1452 dst1->input = dst_discard;
1453 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1455 dst1->next = dst_prev;
1458 header_len += xfrm[i]->props.header_len;
1459 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1460 nfheader_len += xfrm[i]->props.header_len;
1461 trailer_len += xfrm[i]->props.trailer_len;
1464 dst_prev->child = dst;
1472 /* Copy neighbour for reachability confirmation */
1473 dst0->neighbour = neigh_clone(dst->neighbour);
1475 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1476 xfrm_init_pmtu(dst_prev);
1478 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1479 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1481 err = xfrm_fill_dst(xdst, dev, fl);
1485 dst_prev->header_len = header_len;
1486 dst_prev->trailer_len = trailer_len;
1487 header_len -= xdst->u.dst.xfrm->props.header_len;
1488 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1496 xfrm_state_put(xfrm[i]);
1500 dst0 = ERR_PTR(err);
1505 xfrm_dst_alloc_copy(void **target, void *src, int size)
1508 *target = kmalloc(size, GFP_ATOMIC);
1512 memcpy(*target, src, size);
1517 xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1519 #ifdef CONFIG_XFRM_SUB_POLICY
1520 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1521 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1529 xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1531 #ifdef CONFIG_XFRM_SUB_POLICY
1532 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1533 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1539 static int xfrm_expand_policies(struct flowi *fl, u16 family,
1540 struct xfrm_policy **pols,
1541 int *num_pols, int *num_xfrms)
1545 if (*num_pols == 0 || !pols[0]) {
1550 if (IS_ERR(pols[0]))
1551 return PTR_ERR(pols[0]);
1553 *num_xfrms = pols[0]->xfrm_nr;
1555 #ifdef CONFIG_XFRM_SUB_POLICY
1556 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1557 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1558 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1559 XFRM_POLICY_TYPE_MAIN,
1563 if (IS_ERR(pols[1])) {
1564 xfrm_pols_put(pols, *num_pols);
1565 return PTR_ERR(pols[1]);
1568 (*num_xfrms) += pols[1]->xfrm_nr;
1572 for (i = 0; i < *num_pols; i++) {
1573 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1583 static struct xfrm_dst *
1584 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1585 struct flowi *fl, u16 family,
1586 struct dst_entry *dst_orig)
1588 struct net *net = xp_net(pols[0]);
1589 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1590 struct dst_entry *dst;
1591 struct xfrm_dst *xdst;
1594 /* Try to instantiate a bundle */
1595 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1597 if (err != 0 && err != -EAGAIN)
1598 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1599 return ERR_PTR(err);
1602 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1604 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1605 return ERR_CAST(dst);
1608 xdst = (struct xfrm_dst *)dst;
1609 xdst->num_xfrms = err;
1611 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1613 err = xfrm_dst_update_origin(dst, fl);
1614 if (unlikely(err)) {
1616 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1617 return ERR_PTR(err);
1620 xdst->num_pols = num_pols;
1621 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1622 xdst->policy_genid = atomic_read(&pols[0]->genid);
1627 static struct flow_cache_object *
1628 xfrm_bundle_lookup(struct net *net, struct flowi *fl, u16 family, u8 dir,
1629 struct flow_cache_object *oldflo, void *ctx)
1631 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1632 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1633 struct xfrm_dst *xdst, *new_xdst;
1634 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1636 /* Check if the policies from old bundle are usable */
1639 xdst = container_of(oldflo, struct xfrm_dst, flo);
1640 num_pols = xdst->num_pols;
1641 num_xfrms = xdst->num_xfrms;
1643 for (i = 0; i < num_pols; i++) {
1644 pols[i] = xdst->pols[i];
1645 pol_dead |= pols[i]->walk.dead;
1648 dst_free(&xdst->u.dst);
1656 /* Resolve policies to use if we couldn't get them from
1657 * previous cache entry */
1660 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1661 err = xfrm_expand_policies(fl, family, pols,
1662 &num_pols, &num_xfrms);
1668 goto make_dummy_bundle;
1671 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1672 if (IS_ERR(new_xdst)) {
1673 err = PTR_ERR(new_xdst);
1677 goto make_dummy_bundle;
1678 dst_hold(&xdst->u.dst);
1680 } else if (new_xdst == NULL) {
1683 goto make_dummy_bundle;
1684 xdst->num_xfrms = 0;
1685 dst_hold(&xdst->u.dst);
1689 /* Kill the previous bundle */
1691 /* The policies were stolen for newly generated bundle */
1693 dst_free(&xdst->u.dst);
1696 /* Flow cache does not have reference, it dst_free()'s,
1697 * but we do need to return one reference for original caller */
1698 dst_hold(&new_xdst->u.dst);
1699 return &new_xdst->flo;
1702 /* We found policies, but there's no bundles to instantiate:
1703 * either because the policy blocks, has no transformations or
1704 * we could not build template (no xfrm_states).*/
1705 xdst = xfrm_alloc_dst(net, family);
1707 xfrm_pols_put(pols, num_pols);
1708 return ERR_CAST(xdst);
1710 xdst->num_pols = num_pols;
1711 xdst->num_xfrms = num_xfrms;
1712 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1714 dst_hold(&xdst->u.dst);
1718 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1721 dst_free(&xdst->u.dst);
1723 xfrm_pols_put(pols, num_pols);
1724 return ERR_PTR(err);
1727 /* Main function: finds/creates a bundle for given flow.
1729 * At the moment we eat a raw IP route. Mostly to speed up lookups
1730 * on interfaces with disabled IPsec.
1732 int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1733 struct sock *sk, int flags)
1735 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1736 struct flow_cache_object *flo;
1737 struct xfrm_dst *xdst;
1738 struct dst_entry *dst, *dst_orig = *dst_p, *route;
1739 u16 family = dst_orig->ops->family;
1740 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1741 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1748 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1750 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1751 err = xfrm_expand_policies(fl, family, pols,
1752 &num_pols, &num_xfrms);
1757 if (num_xfrms <= 0) {
1758 drop_pols = num_pols;
1762 xdst = xfrm_resolve_and_create_bundle(
1766 xfrm_pols_put(pols, num_pols);
1767 err = PTR_ERR(xdst);
1769 } else if (xdst == NULL) {
1771 drop_pols = num_pols;
1775 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1776 xdst->u.dst.next = xfrm_policy_sk_bundles;
1777 xfrm_policy_sk_bundles = &xdst->u.dst;
1778 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1780 route = xdst->route;
1785 /* To accelerate a bit... */
1786 if ((dst_orig->flags & DST_NOXFRM) ||
1787 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1790 flo = flow_cache_lookup(net, fl, family, dir,
1791 xfrm_bundle_lookup, dst_orig);
1798 xdst = container_of(flo, struct xfrm_dst, flo);
1800 num_pols = xdst->num_pols;
1801 num_xfrms = xdst->num_xfrms;
1802 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1803 route = xdst->route;
1807 if (route == NULL && num_xfrms > 0) {
1808 /* The only case when xfrm_bundle_lookup() returns a
1809 * bundle with null route, is when the template could
1810 * not be resolved. It means policies are there, but
1811 * bundle could not be created, since we don't yet
1812 * have the xfrm_state's. We need to wait for KM to
1813 * negotiate new SA's or bail out with error.*/
1814 if (net->xfrm.sysctl_larval_drop) {
1815 /* EREMOTE tells the caller to generate
1816 * a one-shot blackhole route. */
1818 xfrm_pols_put(pols, drop_pols);
1819 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1822 if (flags & XFRM_LOOKUP_WAIT) {
1823 DECLARE_WAITQUEUE(wait, current);
1825 add_wait_queue(&net->xfrm.km_waitq, &wait);
1826 set_current_state(TASK_INTERRUPTIBLE);
1828 set_current_state(TASK_RUNNING);
1829 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1831 if (!signal_pending(current)) {
1840 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1848 if ((flags & XFRM_LOOKUP_ICMP) &&
1849 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1854 for (i = 0; i < num_pols; i++)
1855 pols[i]->curlft.use_time = get_seconds();
1857 if (num_xfrms < 0) {
1858 /* Prohibit the flow */
1859 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1862 } else if (num_xfrms > 0) {
1863 /* Flow transformed */
1865 dst_release(dst_orig);
1867 /* Flow passes untransformed */
1871 xfrm_pols_put(pols, drop_pols);
1875 if (!(flags & XFRM_LOOKUP_ICMP))
1881 dst_release(dst_orig);
1883 xfrm_pols_put(pols, drop_pols);
1886 EXPORT_SYMBOL(__xfrm_lookup);
1888 int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1889 struct sock *sk, int flags)
1891 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
1893 if (err == -EREMOTE) {
1894 dst_release(*dst_p);
1901 EXPORT_SYMBOL(xfrm_lookup);
1904 xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1906 struct xfrm_state *x;
1908 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1910 x = skb->sp->xvec[idx];
1911 if (!x->type->reject)
1913 return x->type->reject(x, skb, fl);
1916 /* When skb is transformed back to its "native" form, we have to
1917 * check policy restrictions. At the moment we make this in maximally
1918 * stupid way. Shame on me. :-) Of course, connected sockets must
1919 * have policy cached at them.
1923 xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1924 unsigned short family)
1926 if (xfrm_state_kern(x))
1927 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1928 return x->id.proto == tmpl->id.proto &&
1929 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1930 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1931 x->props.mode == tmpl->mode &&
1932 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1933 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1934 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1935 xfrm_state_addr_cmp(tmpl, x, family));
1939 * 0 or more than 0 is returned when validation is succeeded (either bypass
1940 * because of optional transport mode, or next index of the mathced secpath
1941 * state with the template.
1942 * -1 is returned when no matching template is found.
1943 * Otherwise "-2 - errored_index" is returned.
1946 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1947 unsigned short family)
1951 if (tmpl->optional) {
1952 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1956 for (; idx < sp->len; idx++) {
1957 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1959 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1968 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1969 unsigned int family, int reverse)
1971 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1974 if (unlikely(afinfo == NULL))
1975 return -EAFNOSUPPORT;
1977 afinfo->decode_session(skb, fl, reverse);
1978 err = security_xfrm_decode_session(skb, &fl->secid);
1979 xfrm_policy_put_afinfo(afinfo);
1982 EXPORT_SYMBOL(__xfrm_decode_session);
1984 static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1986 for (; k < sp->len; k++) {
1987 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
1996 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1997 unsigned short family)
1999 struct net *net = dev_net(skb->dev);
2000 struct xfrm_policy *pol;
2001 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2010 reverse = dir & ~XFRM_POLICY_MASK;
2011 dir &= XFRM_POLICY_MASK;
2012 fl_dir = policy_to_flow_dir(dir);
2014 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2015 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2019 nf_nat_decode_session(skb, &fl, family);
2021 /* First, check used SA against their selectors. */
2025 for (i=skb->sp->len-1; i>=0; i--) {
2026 struct xfrm_state *x = skb->sp->xvec[i];
2027 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2028 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2035 if (sk && sk->sk_policy[dir]) {
2036 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2038 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2044 struct flow_cache_object *flo;
2046 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2047 xfrm_policy_lookup, NULL);
2048 if (IS_ERR_OR_NULL(flo))
2049 pol = ERR_CAST(flo);
2051 pol = container_of(flo, struct xfrm_policy, flo);
2055 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2060 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2061 xfrm_secpath_reject(xerr_idx, skb, &fl);
2062 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2068 pol->curlft.use_time = get_seconds();
2072 #ifdef CONFIG_XFRM_SUB_POLICY
2073 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2074 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2078 if (IS_ERR(pols[1])) {
2079 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2082 pols[1]->curlft.use_time = get_seconds();
2088 if (pol->action == XFRM_POLICY_ALLOW) {
2089 struct sec_path *sp;
2090 static struct sec_path dummy;
2091 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2092 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2093 struct xfrm_tmpl **tpp = tp;
2097 if ((sp = skb->sp) == NULL)
2100 for (pi = 0; pi < npols; pi++) {
2101 if (pols[pi] != pol &&
2102 pols[pi]->action != XFRM_POLICY_ALLOW) {
2103 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2106 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2107 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2110 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2111 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2115 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2119 /* For each tunnel xfrm, find the first matching tmpl.
2120 * For each tmpl before that, find corresponding xfrm.
2121 * Order is _important_. Later we will implement
2122 * some barriers, but at the moment barriers
2123 * are implied between each two transformations.
2125 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2126 k = xfrm_policy_ok(tpp[i], sp, k, family);
2129 /* "-2 - errored_index" returned */
2131 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2136 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2137 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2141 xfrm_pols_put(pols, npols);
2144 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2147 xfrm_secpath_reject(xerr_idx, skb, &fl);
2149 xfrm_pols_put(pols, npols);
2152 EXPORT_SYMBOL(__xfrm_policy_check);
2154 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2156 struct net *net = dev_net(skb->dev);
2158 struct dst_entry *dst;
2161 if (xfrm_decode_session(skb, &fl, family) < 0) {
2162 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2169 res = xfrm_lookup(net, &dst, &fl, NULL, 0) == 0;
2170 skb_dst_set(skb, dst);
2173 EXPORT_SYMBOL(__xfrm_route_forward);
2175 /* Optimize later using cookies and generation ids. */
2177 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2179 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2180 * to "-1" to force all XFRM destinations to get validated by
2181 * dst_ops->check on every use. We do this because when a
2182 * normal route referenced by an XFRM dst is obsoleted we do
2183 * not go looking around for all parent referencing XFRM dsts
2184 * so that we can invalidate them. It is just too much work.
2185 * Instead we make the checks here on every use. For example:
2187 * XFRM dst A --> IPv4 dst X
2189 * X is the "xdst->route" of A (X is also the "dst->path" of A
2190 * in this example). If X is marked obsolete, "A" will not
2191 * notice. That's what we are validating here via the
2192 * stale_bundle() check.
2194 * When a policy's bundle is pruned, we dst_free() the XFRM
2195 * dst which causes it's ->obsolete field to be set to a
2196 * positive non-zero integer. If an XFRM dst has been pruned
2197 * like this, we want to force a new route lookup.
2199 if (dst->obsolete < 0 && !stale_bundle(dst))
2205 static int stale_bundle(struct dst_entry *dst)
2207 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
2210 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2212 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2213 dst->dev = dev_net(dev)->loopback_dev;
2218 EXPORT_SYMBOL(xfrm_dst_ifdown);
2220 static void xfrm_link_failure(struct sk_buff *skb)
2222 /* Impossible. Such dst must be popped before reaches point of failure. */
2225 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2228 if (dst->obsolete) {
2236 static void __xfrm_garbage_collect(struct net *net)
2238 struct dst_entry *head, *next;
2242 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2243 head = xfrm_policy_sk_bundles;
2244 xfrm_policy_sk_bundles = NULL;
2245 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2254 static void xfrm_init_pmtu(struct dst_entry *dst)
2257 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2258 u32 pmtu, route_mtu_cached;
2260 pmtu = dst_mtu(dst->child);
2261 xdst->child_mtu_cached = pmtu;
2263 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2265 route_mtu_cached = dst_mtu(xdst->route);
2266 xdst->route_mtu_cached = route_mtu_cached;
2268 if (pmtu > route_mtu_cached)
2269 pmtu = route_mtu_cached;
2271 dst->metrics[RTAX_MTU-1] = pmtu;
2272 } while ((dst = dst->next));
2275 /* Check that the bundle accepts the flow and its components are
2279 int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2280 struct flowi *fl, int family, int strict)
2282 struct dst_entry *dst = &first->u.dst;
2283 struct xfrm_dst *last;
2286 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2287 (dst->dev && !netif_running(dst->dev)))
2289 #ifdef CONFIG_XFRM_SUB_POLICY
2291 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2293 if (first->partner &&
2294 !xfrm_selector_match(first->partner, fl, family))
2302 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2304 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2307 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
2309 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2311 if (xdst->xfrm_genid != dst->xfrm->genid)
2313 if (xdst->num_pols > 0 &&
2314 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2318 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2319 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2322 mtu = dst_mtu(dst->child);
2323 if (xdst->child_mtu_cached != mtu) {
2325 xdst->child_mtu_cached = mtu;
2328 if (!dst_check(xdst->route, xdst->route_cookie))
2330 mtu = dst_mtu(xdst->route);
2331 if (xdst->route_mtu_cached != mtu) {
2333 xdst->route_mtu_cached = mtu;
2337 } while (dst->xfrm);
2342 mtu = last->child_mtu_cached;
2346 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2347 if (mtu > last->route_mtu_cached)
2348 mtu = last->route_mtu_cached;
2349 dst->metrics[RTAX_MTU-1] = mtu;
2354 last = (struct xfrm_dst *)last->u.dst.next;
2355 last->child_mtu_cached = mtu;
2361 EXPORT_SYMBOL(xfrm_bundle_ok);
2363 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2367 if (unlikely(afinfo == NULL))
2369 if (unlikely(afinfo->family >= NPROTO))
2370 return -EAFNOSUPPORT;
2371 write_lock_bh(&xfrm_policy_afinfo_lock);
2372 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2375 struct dst_ops *dst_ops = afinfo->dst_ops;
2376 if (likely(dst_ops->kmem_cachep == NULL))
2377 dst_ops->kmem_cachep = xfrm_dst_cache;
2378 if (likely(dst_ops->check == NULL))
2379 dst_ops->check = xfrm_dst_check;
2380 if (likely(dst_ops->negative_advice == NULL))
2381 dst_ops->negative_advice = xfrm_negative_advice;
2382 if (likely(dst_ops->link_failure == NULL))
2383 dst_ops->link_failure = xfrm_link_failure;
2384 if (likely(afinfo->garbage_collect == NULL))
2385 afinfo->garbage_collect = __xfrm_garbage_collect;
2386 xfrm_policy_afinfo[afinfo->family] = afinfo;
2388 write_unlock_bh(&xfrm_policy_afinfo_lock);
2392 struct dst_ops *xfrm_dst_ops;
2394 switch (afinfo->family) {
2396 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2398 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2400 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2406 *xfrm_dst_ops = *afinfo->dst_ops;
2412 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2414 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2417 if (unlikely(afinfo == NULL))
2419 if (unlikely(afinfo->family >= NPROTO))
2420 return -EAFNOSUPPORT;
2421 write_lock_bh(&xfrm_policy_afinfo_lock);
2422 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2423 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2426 struct dst_ops *dst_ops = afinfo->dst_ops;
2427 xfrm_policy_afinfo[afinfo->family] = NULL;
2428 dst_ops->kmem_cachep = NULL;
2429 dst_ops->check = NULL;
2430 dst_ops->negative_advice = NULL;
2431 dst_ops->link_failure = NULL;
2432 afinfo->garbage_collect = NULL;
2435 write_unlock_bh(&xfrm_policy_afinfo_lock);
2438 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2440 static void __net_init xfrm_dst_ops_init(struct net *net)
2442 struct xfrm_policy_afinfo *afinfo;
2444 read_lock_bh(&xfrm_policy_afinfo_lock);
2445 afinfo = xfrm_policy_afinfo[AF_INET];
2447 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2448 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2449 afinfo = xfrm_policy_afinfo[AF_INET6];
2451 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2453 read_unlock_bh(&xfrm_policy_afinfo_lock);
2456 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2458 struct xfrm_policy_afinfo *afinfo;
2459 if (unlikely(family >= NPROTO))
2461 read_lock(&xfrm_policy_afinfo_lock);
2462 afinfo = xfrm_policy_afinfo[family];
2463 if (unlikely(!afinfo))
2464 read_unlock(&xfrm_policy_afinfo_lock);
2468 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2470 read_unlock(&xfrm_policy_afinfo_lock);
2473 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2475 struct net_device *dev = ptr;
2479 __xfrm_garbage_collect(dev_net(dev));
2484 static struct notifier_block xfrm_dev_notifier = {
2485 .notifier_call = xfrm_dev_event,
2488 #ifdef CONFIG_XFRM_STATISTICS
2489 static int __net_init xfrm_statistics_init(struct net *net)
2493 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2494 sizeof(struct linux_xfrm_mib),
2495 __alignof__(struct linux_xfrm_mib)) < 0)
2497 rv = xfrm_proc_init(net);
2499 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2503 static void xfrm_statistics_fini(struct net *net)
2505 xfrm_proc_fini(net);
2506 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2509 static int __net_init xfrm_statistics_init(struct net *net)
2514 static void xfrm_statistics_fini(struct net *net)
2519 static int __net_init xfrm_policy_init(struct net *net)
2521 unsigned int hmask, sz;
2524 if (net_eq(net, &init_net))
2525 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2526 sizeof(struct xfrm_dst),
2527 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2531 sz = (hmask+1) * sizeof(struct hlist_head);
2533 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2534 if (!net->xfrm.policy_byidx)
2536 net->xfrm.policy_idx_hmask = hmask;
2538 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2539 struct xfrm_policy_hash *htab;
2541 net->xfrm.policy_count[dir] = 0;
2542 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2544 htab = &net->xfrm.policy_bydst[dir];
2545 htab->table = xfrm_hash_alloc(sz);
2548 htab->hmask = hmask;
2551 INIT_LIST_HEAD(&net->xfrm.policy_all);
2552 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2553 if (net_eq(net, &init_net))
2554 register_netdevice_notifier(&xfrm_dev_notifier);
2558 for (dir--; dir >= 0; dir--) {
2559 struct xfrm_policy_hash *htab;
2561 htab = &net->xfrm.policy_bydst[dir];
2562 xfrm_hash_free(htab->table, sz);
2564 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2569 static void xfrm_policy_fini(struct net *net)
2571 struct xfrm_audit audit_info;
2575 flush_work(&net->xfrm.policy_hash_work);
2576 #ifdef CONFIG_XFRM_SUB_POLICY
2577 audit_info.loginuid = -1;
2578 audit_info.sessionid = -1;
2579 audit_info.secid = 0;
2580 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2582 audit_info.loginuid = -1;
2583 audit_info.sessionid = -1;
2584 audit_info.secid = 0;
2585 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2587 WARN_ON(!list_empty(&net->xfrm.policy_all));
2589 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2590 struct xfrm_policy_hash *htab;
2592 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2594 htab = &net->xfrm.policy_bydst[dir];
2595 sz = (htab->hmask + 1);
2596 WARN_ON(!hlist_empty(htab->table));
2597 xfrm_hash_free(htab->table, sz);
2600 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2601 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2602 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2605 static int __net_init xfrm_net_init(struct net *net)
2609 rv = xfrm_statistics_init(net);
2611 goto out_statistics;
2612 rv = xfrm_state_init(net);
2615 rv = xfrm_policy_init(net);
2618 xfrm_dst_ops_init(net);
2619 rv = xfrm_sysctl_init(net);
2625 xfrm_policy_fini(net);
2627 xfrm_state_fini(net);
2629 xfrm_statistics_fini(net);
2634 static void __net_exit xfrm_net_exit(struct net *net)
2636 xfrm_sysctl_fini(net);
2637 xfrm_policy_fini(net);
2638 xfrm_state_fini(net);
2639 xfrm_statistics_fini(net);
2642 static struct pernet_operations __net_initdata xfrm_net_ops = {
2643 .init = xfrm_net_init,
2644 .exit = xfrm_net_exit,
2647 void __init xfrm_init(void)
2649 register_pernet_subsys(&xfrm_net_ops);
2653 #ifdef CONFIG_AUDITSYSCALL
2654 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2655 struct audit_buffer *audit_buf)
2657 struct xfrm_sec_ctx *ctx = xp->security;
2658 struct xfrm_selector *sel = &xp->selector;
2661 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2662 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2664 switch(sel->family) {
2666 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2667 if (sel->prefixlen_s != 32)
2668 audit_log_format(audit_buf, " src_prefixlen=%d",
2670 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2671 if (sel->prefixlen_d != 32)
2672 audit_log_format(audit_buf, " dst_prefixlen=%d",
2676 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2677 if (sel->prefixlen_s != 128)
2678 audit_log_format(audit_buf, " src_prefixlen=%d",
2680 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2681 if (sel->prefixlen_d != 128)
2682 audit_log_format(audit_buf, " dst_prefixlen=%d",
2688 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2689 uid_t auid, u32 sessionid, u32 secid)
2691 struct audit_buffer *audit_buf;
2693 audit_buf = xfrm_audit_start("SPD-add");
2694 if (audit_buf == NULL)
2696 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2697 audit_log_format(audit_buf, " res=%u", result);
2698 xfrm_audit_common_policyinfo(xp, audit_buf);
2699 audit_log_end(audit_buf);
2701 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2703 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2704 uid_t auid, u32 sessionid, u32 secid)
2706 struct audit_buffer *audit_buf;
2708 audit_buf = xfrm_audit_start("SPD-delete");
2709 if (audit_buf == NULL)
2711 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2712 audit_log_format(audit_buf, " res=%u", result);
2713 xfrm_audit_common_policyinfo(xp, audit_buf);
2714 audit_log_end(audit_buf);
2716 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2719 #ifdef CONFIG_XFRM_MIGRATE
2720 static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2721 struct xfrm_selector *sel_tgt)
2723 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2724 if (sel_tgt->family == sel_cmp->family &&
2725 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2726 sel_cmp->family) == 0 &&
2727 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2728 sel_cmp->family) == 0 &&
2729 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2730 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2734 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2741 static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2744 struct xfrm_policy *pol, *ret = NULL;
2745 struct hlist_node *entry;
2746 struct hlist_head *chain;
2749 read_lock_bh(&xfrm_policy_lock);
2750 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2751 hlist_for_each_entry(pol, entry, chain, bydst) {
2752 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2753 pol->type == type) {
2755 priority = ret->priority;
2759 chain = &init_net.xfrm.policy_inexact[dir];
2760 hlist_for_each_entry(pol, entry, chain, bydst) {
2761 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2762 pol->type == type &&
2763 pol->priority < priority) {
2772 read_unlock_bh(&xfrm_policy_lock);
2777 static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2781 if (t->mode == m->mode && t->id.proto == m->proto &&
2782 (m->reqid == 0 || t->reqid == m->reqid)) {
2784 case XFRM_MODE_TUNNEL:
2785 case XFRM_MODE_BEET:
2786 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2787 m->old_family) == 0 &&
2788 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2789 m->old_family) == 0) {
2793 case XFRM_MODE_TRANSPORT:
2794 /* in case of transport mode, template does not store
2795 any IP addresses, hence we just compare mode and
2806 /* update endpoint address(es) of template(s) */
2807 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2808 struct xfrm_migrate *m, int num_migrate)
2810 struct xfrm_migrate *mp;
2813 write_lock_bh(&pol->lock);
2814 if (unlikely(pol->walk.dead)) {
2815 /* target policy has been deleted */
2816 write_unlock_bh(&pol->lock);
2820 for (i = 0; i < pol->xfrm_nr; i++) {
2821 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2822 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2825 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2826 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2828 /* update endpoints */
2829 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2830 sizeof(pol->xfrm_vec[i].id.daddr));
2831 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2832 sizeof(pol->xfrm_vec[i].saddr));
2833 pol->xfrm_vec[i].encap_family = mp->new_family;
2835 atomic_inc(&pol->genid);
2839 write_unlock_bh(&pol->lock);
2847 static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2851 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2854 for (i = 0; i < num_migrate; i++) {
2855 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2856 m[i].old_family) == 0) &&
2857 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2858 m[i].old_family) == 0))
2860 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2861 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2864 /* check if there is any duplicated entry */
2865 for (j = i + 1; j < num_migrate; j++) {
2866 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2867 sizeof(m[i].old_daddr)) &&
2868 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2869 sizeof(m[i].old_saddr)) &&
2870 m[i].proto == m[j].proto &&
2871 m[i].mode == m[j].mode &&
2872 m[i].reqid == m[j].reqid &&
2873 m[i].old_family == m[j].old_family)
2881 int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
2882 struct xfrm_migrate *m, int num_migrate,
2883 struct xfrm_kmaddress *k)
2885 int i, err, nx_cur = 0, nx_new = 0;
2886 struct xfrm_policy *pol = NULL;
2887 struct xfrm_state *x, *xc;
2888 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2889 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2890 struct xfrm_migrate *mp;
2892 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2895 /* Stage 1 - find policy */
2896 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2901 /* Stage 2 - find and update state(s) */
2902 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2903 if ((x = xfrm_migrate_state_find(mp))) {
2906 if ((xc = xfrm_state_migrate(x, mp))) {
2916 /* Stage 3 - update policy */
2917 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2920 /* Stage 4 - delete old state(s) */
2922 xfrm_states_put(x_cur, nx_cur);
2923 xfrm_states_delete(x_cur, nx_cur);
2926 /* Stage 5 - announce */
2927 km_migrate(sel, dir, type, m, num_migrate, k);
2939 xfrm_states_put(x_cur, nx_cur);
2941 xfrm_states_delete(x_new, nx_new);
2945 EXPORT_SYMBOL(xfrm_migrate);