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);
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 addr4_match(fl4->daddr, sel->daddr.a4, sel->prefixlen_d) &&
65 addr4_match(fl4->saddr, sel->saddr.a4, 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 bool 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;
883 if (pol->family != family ||
884 (fl->flowi_mark & pol->mark.m) != pol->mark.v ||
888 match = xfrm_selector_match(sel, fl, family);
890 ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
896 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
897 const struct flowi *fl,
901 struct xfrm_policy *pol, *ret;
902 const xfrm_address_t *daddr, *saddr;
903 struct hlist_node *entry;
904 struct hlist_head *chain;
907 daddr = xfrm_flowi_daddr(fl, family);
908 saddr = xfrm_flowi_saddr(fl, family);
909 if (unlikely(!daddr || !saddr))
912 read_lock_bh(&xfrm_policy_lock);
913 chain = policy_hash_direct(net, daddr, saddr, family, dir);
915 hlist_for_each_entry(pol, entry, chain, bydst) {
916 err = xfrm_policy_match(pol, fl, type, family, dir);
926 priority = ret->priority;
930 chain = &net->xfrm.policy_inexact[dir];
931 hlist_for_each_entry(pol, entry, chain, bydst) {
932 err = xfrm_policy_match(pol, fl, type, family, dir);
940 } else if (pol->priority < priority) {
948 read_unlock_bh(&xfrm_policy_lock);
953 static struct xfrm_policy *
954 __xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir)
956 #ifdef CONFIG_XFRM_SUB_POLICY
957 struct xfrm_policy *pol;
959 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
963 return xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
966 static struct flow_cache_object *
967 xfrm_policy_lookup(struct net *net, const struct flowi *fl, u16 family,
968 u8 dir, struct flow_cache_object *old_obj, void *ctx)
970 struct xfrm_policy *pol;
973 xfrm_pol_put(container_of(old_obj, struct xfrm_policy, flo));
975 pol = __xfrm_policy_lookup(net, fl, family, dir);
976 if (IS_ERR_OR_NULL(pol))
977 return ERR_CAST(pol);
979 /* Resolver returns two references:
980 * one for cache and one for caller of flow_cache_lookup() */
986 static inline int policy_to_flow_dir(int dir)
988 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
989 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
990 XFRM_POLICY_FWD == FLOW_DIR_FWD)
996 case XFRM_POLICY_OUT:
998 case XFRM_POLICY_FWD:
1003 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir,
1004 const struct flowi *fl)
1006 struct xfrm_policy *pol;
1008 read_lock_bh(&xfrm_policy_lock);
1009 if ((pol = sk->sk_policy[dir]) != NULL) {
1010 bool match = xfrm_selector_match(&pol->selector, fl,
1015 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1019 err = security_xfrm_policy_lookup(pol->security,
1021 policy_to_flow_dir(dir));
1024 else if (err == -ESRCH)
1032 read_unlock_bh(&xfrm_policy_lock);
1036 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1038 struct net *net = xp_net(pol);
1039 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1042 list_add(&pol->walk.all, &net->xfrm.policy_all);
1043 hlist_add_head(&pol->bydst, chain);
1044 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1045 net->xfrm.policy_count[dir]++;
1048 if (xfrm_bydst_should_resize(net, dir, NULL))
1049 schedule_work(&net->xfrm.policy_hash_work);
1052 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1055 struct net *net = xp_net(pol);
1057 if (hlist_unhashed(&pol->bydst))
1060 hlist_del(&pol->bydst);
1061 hlist_del(&pol->byidx);
1062 list_del(&pol->walk.all);
1063 net->xfrm.policy_count[dir]--;
1068 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1070 write_lock_bh(&xfrm_policy_lock);
1071 pol = __xfrm_policy_unlink(pol, dir);
1072 write_unlock_bh(&xfrm_policy_lock);
1074 xfrm_policy_kill(pol);
1079 EXPORT_SYMBOL(xfrm_policy_delete);
1081 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1083 struct net *net = xp_net(pol);
1084 struct xfrm_policy *old_pol;
1086 #ifdef CONFIG_XFRM_SUB_POLICY
1087 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1091 write_lock_bh(&xfrm_policy_lock);
1092 old_pol = sk->sk_policy[dir];
1093 sk->sk_policy[dir] = pol;
1095 pol->curlft.add_time = get_seconds();
1096 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1097 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1100 /* Unlinking succeeds always. This is the only function
1101 * allowed to delete or replace socket policy.
1103 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1104 write_unlock_bh(&xfrm_policy_lock);
1107 xfrm_policy_kill(old_pol);
1112 static struct xfrm_policy *clone_policy(const struct xfrm_policy *old, int dir)
1114 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1117 newp->selector = old->selector;
1118 if (security_xfrm_policy_clone(old->security,
1121 return NULL; /* ENOMEM */
1123 newp->lft = old->lft;
1124 newp->curlft = old->curlft;
1125 newp->mark = old->mark;
1126 newp->action = old->action;
1127 newp->flags = old->flags;
1128 newp->xfrm_nr = old->xfrm_nr;
1129 newp->index = old->index;
1130 newp->type = old->type;
1131 memcpy(newp->xfrm_vec, old->xfrm_vec,
1132 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1133 write_lock_bh(&xfrm_policy_lock);
1134 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1135 write_unlock_bh(&xfrm_policy_lock);
1141 int __xfrm_sk_clone_policy(struct sock *sk)
1143 struct xfrm_policy *p0 = sk->sk_policy[0],
1144 *p1 = sk->sk_policy[1];
1146 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1147 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1149 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1155 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1156 unsigned short family)
1159 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1161 if (unlikely(afinfo == NULL))
1163 err = afinfo->get_saddr(net, local, remote);
1164 xfrm_policy_put_afinfo(afinfo);
1168 /* Resolve list of templates for the flow, given policy. */
1171 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, const struct flowi *fl,
1172 struct xfrm_state **xfrm, unsigned short family)
1174 struct net *net = xp_net(policy);
1177 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1178 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1181 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1182 struct xfrm_state *x;
1183 xfrm_address_t *remote = daddr;
1184 xfrm_address_t *local = saddr;
1185 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1187 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1188 tmpl->mode == XFRM_MODE_BEET) {
1189 remote = &tmpl->id.daddr;
1190 local = &tmpl->saddr;
1191 if (xfrm_addr_any(local, tmpl->encap_family)) {
1192 error = xfrm_get_saddr(net, &tmp, remote, tmpl->encap_family);
1199 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1201 if (x && x->km.state == XFRM_STATE_VALID) {
1208 error = (x->km.state == XFRM_STATE_ERROR ?
1212 else if (error == -ESRCH)
1215 if (!tmpl->optional)
1221 for (nx--; nx>=0; nx--)
1222 xfrm_state_put(xfrm[nx]);
1227 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, const struct flowi *fl,
1228 struct xfrm_state **xfrm, unsigned short family)
1230 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1231 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1237 for (i = 0; i < npols; i++) {
1238 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1243 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1251 /* found states are sorted for outbound processing */
1253 xfrm_state_sort(xfrm, tpp, cnx, family);
1258 for (cnx--; cnx>=0; cnx--)
1259 xfrm_state_put(tpp[cnx]);
1264 /* Check that the bundle accepts the flow and its components are
1268 static inline int xfrm_get_tos(const struct flowi *fl, int family)
1270 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1276 tos = afinfo->get_tos(fl);
1278 xfrm_policy_put_afinfo(afinfo);
1283 static struct flow_cache_object *xfrm_bundle_flo_get(struct flow_cache_object *flo)
1285 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1286 struct dst_entry *dst = &xdst->u.dst;
1288 if (xdst->route == NULL) {
1289 /* Dummy bundle - if it has xfrms we were not
1290 * able to build bundle as template resolution failed.
1291 * It means we need to try again resolving. */
1292 if (xdst->num_xfrms > 0)
1296 if (stale_bundle(dst))
1304 static int xfrm_bundle_flo_check(struct flow_cache_object *flo)
1306 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1307 struct dst_entry *dst = &xdst->u.dst;
1311 if (stale_bundle(dst))
1317 static void xfrm_bundle_flo_delete(struct flow_cache_object *flo)
1319 struct xfrm_dst *xdst = container_of(flo, struct xfrm_dst, flo);
1320 struct dst_entry *dst = &xdst->u.dst;
1325 static const struct flow_cache_ops xfrm_bundle_fc_ops = {
1326 .get = xfrm_bundle_flo_get,
1327 .check = xfrm_bundle_flo_check,
1328 .delete = xfrm_bundle_flo_delete,
1331 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1333 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1334 struct dst_ops *dst_ops;
1335 struct xfrm_dst *xdst;
1338 return ERR_PTR(-EINVAL);
1342 dst_ops = &net->xfrm.xfrm4_dst_ops;
1344 #if IS_ENABLED(CONFIG_IPV6)
1346 dst_ops = &net->xfrm.xfrm6_dst_ops;
1352 xdst = dst_alloc(dst_ops, NULL, 0, 0, 0);
1355 memset(&xdst->u.rt6.rt6i_table, 0,
1356 sizeof(*xdst) - sizeof(struct dst_entry));
1357 xdst->flo.ops = &xfrm_bundle_fc_ops;
1359 xdst = ERR_PTR(-ENOBUFS);
1361 xfrm_policy_put_afinfo(afinfo);
1366 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1369 struct xfrm_policy_afinfo *afinfo =
1370 xfrm_policy_get_afinfo(dst->ops->family);
1376 err = afinfo->init_path(path, dst, nfheader_len);
1378 xfrm_policy_put_afinfo(afinfo);
1383 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
1384 const struct flowi *fl)
1386 struct xfrm_policy_afinfo *afinfo =
1387 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1393 err = afinfo->fill_dst(xdst, dev, fl);
1395 xfrm_policy_put_afinfo(afinfo);
1401 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1402 * all the metrics... Shortly, bundle a bundle.
1405 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1406 struct xfrm_state **xfrm, int nx,
1407 const struct flowi *fl,
1408 struct dst_entry *dst)
1410 struct net *net = xp_net(policy);
1411 unsigned long now = jiffies;
1412 struct net_device *dev;
1413 struct xfrm_mode *inner_mode;
1414 struct dst_entry *dst_prev = NULL;
1415 struct dst_entry *dst0 = NULL;
1419 int nfheader_len = 0;
1420 int trailer_len = 0;
1422 int family = policy->selector.family;
1423 xfrm_address_t saddr, daddr;
1425 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1427 tos = xfrm_get_tos(fl, family);
1434 for (; i < nx; i++) {
1435 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1436 struct dst_entry *dst1 = &xdst->u.dst;
1438 err = PTR_ERR(xdst);
1444 if (xfrm[i]->sel.family == AF_UNSPEC) {
1445 inner_mode = xfrm_ip2inner_mode(xfrm[i],
1446 xfrm_af2proto(family));
1448 err = -EAFNOSUPPORT;
1453 inner_mode = xfrm[i]->inner_mode;
1458 dst_prev->child = dst_clone(dst1);
1459 dst1->flags |= DST_NOHASH;
1463 dst_copy_metrics(dst1, dst);
1465 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1466 family = xfrm[i]->props.family;
1467 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1475 dst1->xfrm = xfrm[i];
1476 xdst->xfrm_genid = xfrm[i]->genid;
1478 dst1->obsolete = -1;
1479 dst1->flags |= DST_HOST;
1480 dst1->lastuse = now;
1482 dst1->input = dst_discard;
1483 dst1->output = inner_mode->afinfo->output;
1485 dst1->next = dst_prev;
1488 header_len += xfrm[i]->props.header_len;
1489 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1490 nfheader_len += xfrm[i]->props.header_len;
1491 trailer_len += xfrm[i]->props.trailer_len;
1494 dst_prev->child = dst;
1502 /* Copy neighbour for reachability confirmation */
1503 dst_set_neighbour(dst0, neigh_clone(dst_get_neighbour_noref(dst)));
1505 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1506 xfrm_init_pmtu(dst_prev);
1508 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1509 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1511 err = xfrm_fill_dst(xdst, dev, fl);
1515 dst_prev->header_len = header_len;
1516 dst_prev->trailer_len = trailer_len;
1517 header_len -= xdst->u.dst.xfrm->props.header_len;
1518 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1526 xfrm_state_put(xfrm[i]);
1530 dst0 = ERR_PTR(err);
1535 xfrm_dst_alloc_copy(void **target, const void *src, int size)
1538 *target = kmalloc(size, GFP_ATOMIC);
1542 memcpy(*target, src, size);
1547 xfrm_dst_update_parent(struct dst_entry *dst, const struct xfrm_selector *sel)
1549 #ifdef CONFIG_XFRM_SUB_POLICY
1550 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1551 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1559 xfrm_dst_update_origin(struct dst_entry *dst, const struct flowi *fl)
1561 #ifdef CONFIG_XFRM_SUB_POLICY
1562 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1563 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1569 static int xfrm_expand_policies(const struct flowi *fl, u16 family,
1570 struct xfrm_policy **pols,
1571 int *num_pols, int *num_xfrms)
1575 if (*num_pols == 0 || !pols[0]) {
1580 if (IS_ERR(pols[0]))
1581 return PTR_ERR(pols[0]);
1583 *num_xfrms = pols[0]->xfrm_nr;
1585 #ifdef CONFIG_XFRM_SUB_POLICY
1586 if (pols[0] && pols[0]->action == XFRM_POLICY_ALLOW &&
1587 pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1588 pols[1] = xfrm_policy_lookup_bytype(xp_net(pols[0]),
1589 XFRM_POLICY_TYPE_MAIN,
1593 if (IS_ERR(pols[1])) {
1594 xfrm_pols_put(pols, *num_pols);
1595 return PTR_ERR(pols[1]);
1598 (*num_xfrms) += pols[1]->xfrm_nr;
1602 for (i = 0; i < *num_pols; i++) {
1603 if (pols[i]->action != XFRM_POLICY_ALLOW) {
1613 static struct xfrm_dst *
1614 xfrm_resolve_and_create_bundle(struct xfrm_policy **pols, int num_pols,
1615 const struct flowi *fl, u16 family,
1616 struct dst_entry *dst_orig)
1618 struct net *net = xp_net(pols[0]);
1619 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1620 struct dst_entry *dst;
1621 struct xfrm_dst *xdst;
1624 /* Try to instantiate a bundle */
1625 err = xfrm_tmpl_resolve(pols, num_pols, fl, xfrm, family);
1627 if (err != 0 && err != -EAGAIN)
1628 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1629 return ERR_PTR(err);
1632 dst = xfrm_bundle_create(pols[0], xfrm, err, fl, dst_orig);
1634 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1635 return ERR_CAST(dst);
1638 xdst = (struct xfrm_dst *)dst;
1639 xdst->num_xfrms = err;
1641 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1643 err = xfrm_dst_update_origin(dst, fl);
1644 if (unlikely(err)) {
1646 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1647 return ERR_PTR(err);
1650 xdst->num_pols = num_pols;
1651 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1652 xdst->policy_genid = atomic_read(&pols[0]->genid);
1657 static struct flow_cache_object *
1658 xfrm_bundle_lookup(struct net *net, const struct flowi *fl, u16 family, u8 dir,
1659 struct flow_cache_object *oldflo, void *ctx)
1661 struct dst_entry *dst_orig = (struct dst_entry *)ctx;
1662 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1663 struct xfrm_dst *xdst, *new_xdst;
1664 int num_pols = 0, num_xfrms = 0, i, err, pol_dead;
1666 /* Check if the policies from old bundle are usable */
1669 xdst = container_of(oldflo, struct xfrm_dst, flo);
1670 num_pols = xdst->num_pols;
1671 num_xfrms = xdst->num_xfrms;
1673 for (i = 0; i < num_pols; i++) {
1674 pols[i] = xdst->pols[i];
1675 pol_dead |= pols[i]->walk.dead;
1678 dst_free(&xdst->u.dst);
1686 /* Resolve policies to use if we couldn't get them from
1687 * previous cache entry */
1690 pols[0] = __xfrm_policy_lookup(net, fl, family, dir);
1691 err = xfrm_expand_policies(fl, family, pols,
1692 &num_pols, &num_xfrms);
1698 goto make_dummy_bundle;
1701 new_xdst = xfrm_resolve_and_create_bundle(pols, num_pols, fl, family, dst_orig);
1702 if (IS_ERR(new_xdst)) {
1703 err = PTR_ERR(new_xdst);
1707 goto make_dummy_bundle;
1708 dst_hold(&xdst->u.dst);
1710 } else if (new_xdst == NULL) {
1713 goto make_dummy_bundle;
1714 xdst->num_xfrms = 0;
1715 dst_hold(&xdst->u.dst);
1719 /* Kill the previous bundle */
1721 /* The policies were stolen for newly generated bundle */
1723 dst_free(&xdst->u.dst);
1726 /* Flow cache does not have reference, it dst_free()'s,
1727 * but we do need to return one reference for original caller */
1728 dst_hold(&new_xdst->u.dst);
1729 return &new_xdst->flo;
1732 /* We found policies, but there's no bundles to instantiate:
1733 * either because the policy blocks, has no transformations or
1734 * we could not build template (no xfrm_states).*/
1735 xdst = xfrm_alloc_dst(net, family);
1737 xfrm_pols_put(pols, num_pols);
1738 return ERR_CAST(xdst);
1740 xdst->num_pols = num_pols;
1741 xdst->num_xfrms = num_xfrms;
1742 memcpy(xdst->pols, pols, sizeof(struct xfrm_policy*) * num_pols);
1744 dst_hold(&xdst->u.dst);
1748 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1751 dst_free(&xdst->u.dst);
1753 xfrm_pols_put(pols, num_pols);
1754 return ERR_PTR(err);
1757 static struct dst_entry *make_blackhole(struct net *net, u16 family,
1758 struct dst_entry *dst_orig)
1760 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1761 struct dst_entry *ret;
1764 dst_release(dst_orig);
1765 ret = ERR_PTR(-EINVAL);
1767 ret = afinfo->blackhole_route(net, dst_orig);
1769 xfrm_policy_put_afinfo(afinfo);
1774 /* Main function: finds/creates a bundle for given flow.
1776 * At the moment we eat a raw IP route. Mostly to speed up lookups
1777 * on interfaces with disabled IPsec.
1779 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
1780 const struct flowi *fl,
1781 struct sock *sk, int flags)
1783 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1784 struct flow_cache_object *flo;
1785 struct xfrm_dst *xdst;
1786 struct dst_entry *dst, *route;
1787 u16 family = dst_orig->ops->family;
1788 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1789 int i, err, num_pols, num_xfrms = 0, drop_pols = 0;
1796 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1798 pols[0] = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1799 err = xfrm_expand_policies(fl, family, pols,
1800 &num_pols, &num_xfrms);
1805 if (num_xfrms <= 0) {
1806 drop_pols = num_pols;
1810 xdst = xfrm_resolve_and_create_bundle(
1814 xfrm_pols_put(pols, num_pols);
1815 err = PTR_ERR(xdst);
1817 } else if (xdst == NULL) {
1819 drop_pols = num_pols;
1823 dst_hold(&xdst->u.dst);
1825 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
1826 xdst->u.dst.next = xfrm_policy_sk_bundles;
1827 xfrm_policy_sk_bundles = &xdst->u.dst;
1828 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
1830 route = xdst->route;
1835 /* To accelerate a bit... */
1836 if ((dst_orig->flags & DST_NOXFRM) ||
1837 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1840 flo = flow_cache_lookup(net, fl, family, dir,
1841 xfrm_bundle_lookup, dst_orig);
1848 xdst = container_of(flo, struct xfrm_dst, flo);
1850 num_pols = xdst->num_pols;
1851 num_xfrms = xdst->num_xfrms;
1852 memcpy(pols, xdst->pols, sizeof(struct xfrm_policy*) * num_pols);
1853 route = xdst->route;
1857 if (route == NULL && num_xfrms > 0) {
1858 /* The only case when xfrm_bundle_lookup() returns a
1859 * bundle with null route, is when the template could
1860 * not be resolved. It means policies are there, but
1861 * bundle could not be created, since we don't yet
1862 * have the xfrm_state's. We need to wait for KM to
1863 * negotiate new SA's or bail out with error.*/
1864 if (net->xfrm.sysctl_larval_drop) {
1865 /* EREMOTE tells the caller to generate
1866 * a one-shot blackhole route. */
1868 xfrm_pols_put(pols, drop_pols);
1869 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1871 return make_blackhole(net, family, dst_orig);
1873 if (fl->flowi_flags & FLOWI_FLAG_CAN_SLEEP) {
1874 DECLARE_WAITQUEUE(wait, current);
1876 add_wait_queue(&net->xfrm.km_waitq, &wait);
1877 set_current_state(TASK_INTERRUPTIBLE);
1879 set_current_state(TASK_RUNNING);
1880 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1882 if (!signal_pending(current)) {
1891 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1899 if ((flags & XFRM_LOOKUP_ICMP) &&
1900 !(pols[0]->flags & XFRM_POLICY_ICMP)) {
1905 for (i = 0; i < num_pols; i++)
1906 pols[i]->curlft.use_time = get_seconds();
1908 if (num_xfrms < 0) {
1909 /* Prohibit the flow */
1910 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1913 } else if (num_xfrms > 0) {
1914 /* Flow transformed */
1915 dst_release(dst_orig);
1917 /* Flow passes untransformed */
1922 xfrm_pols_put(pols, drop_pols);
1926 if (!(flags & XFRM_LOOKUP_ICMP)) {
1934 dst_release(dst_orig);
1935 xfrm_pols_put(pols, drop_pols);
1936 return ERR_PTR(err);
1938 EXPORT_SYMBOL(xfrm_lookup);
1941 xfrm_secpath_reject(int idx, struct sk_buff *skb, const struct flowi *fl)
1943 struct xfrm_state *x;
1945 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1947 x = skb->sp->xvec[idx];
1948 if (!x->type->reject)
1950 return x->type->reject(x, skb, fl);
1953 /* When skb is transformed back to its "native" form, we have to
1954 * check policy restrictions. At the moment we make this in maximally
1955 * stupid way. Shame on me. :-) Of course, connected sockets must
1956 * have policy cached at them.
1960 xfrm_state_ok(const struct xfrm_tmpl *tmpl, const struct xfrm_state *x,
1961 unsigned short family)
1963 if (xfrm_state_kern(x))
1964 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1965 return x->id.proto == tmpl->id.proto &&
1966 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1967 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1968 x->props.mode == tmpl->mode &&
1969 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1970 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1971 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1972 xfrm_state_addr_cmp(tmpl, x, family));
1976 * 0 or more than 0 is returned when validation is succeeded (either bypass
1977 * because of optional transport mode, or next index of the mathced secpath
1978 * state with the template.
1979 * -1 is returned when no matching template is found.
1980 * Otherwise "-2 - errored_index" is returned.
1983 xfrm_policy_ok(const struct xfrm_tmpl *tmpl, const struct sec_path *sp, int start,
1984 unsigned short family)
1988 if (tmpl->optional) {
1989 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1993 for (; idx < sp->len; idx++) {
1994 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1996 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
2005 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
2006 unsigned int family, int reverse)
2008 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
2011 if (unlikely(afinfo == NULL))
2012 return -EAFNOSUPPORT;
2014 afinfo->decode_session(skb, fl, reverse);
2015 err = security_xfrm_decode_session(skb, &fl->flowi_secid);
2016 xfrm_policy_put_afinfo(afinfo);
2019 EXPORT_SYMBOL(__xfrm_decode_session);
2021 static inline int secpath_has_nontransport(const struct sec_path *sp, int k, int *idxp)
2023 for (; k < sp->len; k++) {
2024 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
2033 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
2034 unsigned short family)
2036 struct net *net = dev_net(skb->dev);
2037 struct xfrm_policy *pol;
2038 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
2047 reverse = dir & ~XFRM_POLICY_MASK;
2048 dir &= XFRM_POLICY_MASK;
2049 fl_dir = policy_to_flow_dir(dir);
2051 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
2052 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
2056 nf_nat_decode_session(skb, &fl, family);
2058 /* First, check used SA against their selectors. */
2062 for (i=skb->sp->len-1; i>=0; i--) {
2063 struct xfrm_state *x = skb->sp->xvec[i];
2064 if (!xfrm_selector_match(&x->sel, &fl, family)) {
2065 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
2072 if (sk && sk->sk_policy[dir]) {
2073 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
2075 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2081 struct flow_cache_object *flo;
2083 flo = flow_cache_lookup(net, &fl, family, fl_dir,
2084 xfrm_policy_lookup, NULL);
2085 if (IS_ERR_OR_NULL(flo))
2086 pol = ERR_CAST(flo);
2088 pol = container_of(flo, struct xfrm_policy, flo);
2092 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2097 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
2098 xfrm_secpath_reject(xerr_idx, skb, &fl);
2099 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
2105 pol->curlft.use_time = get_seconds();
2109 #ifdef CONFIG_XFRM_SUB_POLICY
2110 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
2111 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
2115 if (IS_ERR(pols[1])) {
2116 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
2119 pols[1]->curlft.use_time = get_seconds();
2125 if (pol->action == XFRM_POLICY_ALLOW) {
2126 struct sec_path *sp;
2127 static struct sec_path dummy;
2128 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
2129 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
2130 struct xfrm_tmpl **tpp = tp;
2134 if ((sp = skb->sp) == NULL)
2137 for (pi = 0; pi < npols; pi++) {
2138 if (pols[pi] != pol &&
2139 pols[pi]->action != XFRM_POLICY_ALLOW) {
2140 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2143 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2144 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2147 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2148 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2152 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2156 /* For each tunnel xfrm, find the first matching tmpl.
2157 * For each tmpl before that, find corresponding xfrm.
2158 * Order is _important_. Later we will implement
2159 * some barriers, but at the moment barriers
2160 * are implied between each two transformations.
2162 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2163 k = xfrm_policy_ok(tpp[i], sp, k, family);
2166 /* "-2 - errored_index" returned */
2168 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2173 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2174 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2178 xfrm_pols_put(pols, npols);
2181 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2184 xfrm_secpath_reject(xerr_idx, skb, &fl);
2186 xfrm_pols_put(pols, npols);
2189 EXPORT_SYMBOL(__xfrm_policy_check);
2191 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2193 struct net *net = dev_net(skb->dev);
2195 struct dst_entry *dst;
2198 if (xfrm_decode_session(skb, &fl, family) < 0) {
2199 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2205 dst = xfrm_lookup(net, skb_dst(skb), &fl, NULL, 0);
2210 skb_dst_set(skb, dst);
2213 EXPORT_SYMBOL(__xfrm_route_forward);
2215 /* Optimize later using cookies and generation ids. */
2217 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2219 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2220 * to "-1" to force all XFRM destinations to get validated by
2221 * dst_ops->check on every use. We do this because when a
2222 * normal route referenced by an XFRM dst is obsoleted we do
2223 * not go looking around for all parent referencing XFRM dsts
2224 * so that we can invalidate them. It is just too much work.
2225 * Instead we make the checks here on every use. For example:
2227 * XFRM dst A --> IPv4 dst X
2229 * X is the "xdst->route" of A (X is also the "dst->path" of A
2230 * in this example). If X is marked obsolete, "A" will not
2231 * notice. That's what we are validating here via the
2232 * stale_bundle() check.
2234 * When a policy's bundle is pruned, we dst_free() the XFRM
2235 * dst which causes it's ->obsolete field to be set to a
2236 * positive non-zero integer. If an XFRM dst has been pruned
2237 * like this, we want to force a new route lookup.
2239 if (dst->obsolete < 0 && !stale_bundle(dst))
2245 static int stale_bundle(struct dst_entry *dst)
2247 return !xfrm_bundle_ok((struct xfrm_dst *)dst);
2250 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2252 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2253 dst->dev = dev_net(dev)->loopback_dev;
2258 EXPORT_SYMBOL(xfrm_dst_ifdown);
2260 static void xfrm_link_failure(struct sk_buff *skb)
2262 /* Impossible. Such dst must be popped before reaches point of failure. */
2265 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2268 if (dst->obsolete) {
2276 static void __xfrm_garbage_collect(struct net *net)
2278 struct dst_entry *head, *next;
2280 spin_lock_bh(&xfrm_policy_sk_bundle_lock);
2281 head = xfrm_policy_sk_bundles;
2282 xfrm_policy_sk_bundles = NULL;
2283 spin_unlock_bh(&xfrm_policy_sk_bundle_lock);
2292 static void xfrm_garbage_collect(struct net *net)
2295 __xfrm_garbage_collect(net);
2298 static void xfrm_garbage_collect_deferred(struct net *net)
2300 flow_cache_flush_deferred();
2301 __xfrm_garbage_collect(net);
2304 static void xfrm_init_pmtu(struct dst_entry *dst)
2307 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2308 u32 pmtu, route_mtu_cached;
2310 pmtu = dst_mtu(dst->child);
2311 xdst->child_mtu_cached = pmtu;
2313 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2315 route_mtu_cached = dst_mtu(xdst->route);
2316 xdst->route_mtu_cached = route_mtu_cached;
2318 if (pmtu > route_mtu_cached)
2319 pmtu = route_mtu_cached;
2321 dst_metric_set(dst, RTAX_MTU, pmtu);
2322 } while ((dst = dst->next));
2325 /* Check that the bundle accepts the flow and its components are
2329 static int xfrm_bundle_ok(struct xfrm_dst *first)
2331 struct dst_entry *dst = &first->u.dst;
2332 struct xfrm_dst *last;
2335 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2336 (dst->dev && !netif_running(dst->dev)))
2342 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2344 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2346 if (xdst->xfrm_genid != dst->xfrm->genid)
2348 if (xdst->num_pols > 0 &&
2349 xdst->policy_genid != atomic_read(&xdst->pols[0]->genid))
2352 mtu = dst_mtu(dst->child);
2353 if (xdst->child_mtu_cached != mtu) {
2355 xdst->child_mtu_cached = mtu;
2358 if (!dst_check(xdst->route, xdst->route_cookie))
2360 mtu = dst_mtu(xdst->route);
2361 if (xdst->route_mtu_cached != mtu) {
2363 xdst->route_mtu_cached = mtu;
2367 } while (dst->xfrm);
2372 mtu = last->child_mtu_cached;
2376 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2377 if (mtu > last->route_mtu_cached)
2378 mtu = last->route_mtu_cached;
2379 dst_metric_set(dst, RTAX_MTU, mtu);
2384 last = (struct xfrm_dst *)last->u.dst.next;
2385 last->child_mtu_cached = mtu;
2391 static unsigned int xfrm_default_advmss(const struct dst_entry *dst)
2393 return dst_metric_advmss(dst->path);
2396 static unsigned int xfrm_mtu(const struct dst_entry *dst)
2398 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
2400 return mtu ? : dst_mtu(dst->path);
2403 static struct neighbour *xfrm_neigh_lookup(const struct dst_entry *dst, const void *daddr)
2405 return dst_neigh_lookup(dst->path, daddr);
2408 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2412 if (unlikely(afinfo == NULL))
2414 if (unlikely(afinfo->family >= NPROTO))
2415 return -EAFNOSUPPORT;
2416 write_lock_bh(&xfrm_policy_afinfo_lock);
2417 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2420 struct dst_ops *dst_ops = afinfo->dst_ops;
2421 if (likely(dst_ops->kmem_cachep == NULL))
2422 dst_ops->kmem_cachep = xfrm_dst_cache;
2423 if (likely(dst_ops->check == NULL))
2424 dst_ops->check = xfrm_dst_check;
2425 if (likely(dst_ops->default_advmss == NULL))
2426 dst_ops->default_advmss = xfrm_default_advmss;
2427 if (likely(dst_ops->mtu == NULL))
2428 dst_ops->mtu = xfrm_mtu;
2429 if (likely(dst_ops->negative_advice == NULL))
2430 dst_ops->negative_advice = xfrm_negative_advice;
2431 if (likely(dst_ops->link_failure == NULL))
2432 dst_ops->link_failure = xfrm_link_failure;
2433 if (likely(dst_ops->neigh_lookup == NULL))
2434 dst_ops->neigh_lookup = xfrm_neigh_lookup;
2435 if (likely(afinfo->garbage_collect == NULL))
2436 afinfo->garbage_collect = xfrm_garbage_collect_deferred;
2437 xfrm_policy_afinfo[afinfo->family] = afinfo;
2439 write_unlock_bh(&xfrm_policy_afinfo_lock);
2443 struct dst_ops *xfrm_dst_ops;
2445 switch (afinfo->family) {
2447 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2449 #if IS_ENABLED(CONFIG_IPV6)
2451 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2457 *xfrm_dst_ops = *afinfo->dst_ops;
2463 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2465 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2468 if (unlikely(afinfo == NULL))
2470 if (unlikely(afinfo->family >= NPROTO))
2471 return -EAFNOSUPPORT;
2472 write_lock_bh(&xfrm_policy_afinfo_lock);
2473 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2474 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2477 struct dst_ops *dst_ops = afinfo->dst_ops;
2478 xfrm_policy_afinfo[afinfo->family] = NULL;
2479 dst_ops->kmem_cachep = NULL;
2480 dst_ops->check = NULL;
2481 dst_ops->negative_advice = NULL;
2482 dst_ops->link_failure = NULL;
2483 afinfo->garbage_collect = NULL;
2486 write_unlock_bh(&xfrm_policy_afinfo_lock);
2489 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2491 static void __net_init xfrm_dst_ops_init(struct net *net)
2493 struct xfrm_policy_afinfo *afinfo;
2495 read_lock_bh(&xfrm_policy_afinfo_lock);
2496 afinfo = xfrm_policy_afinfo[AF_INET];
2498 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2499 #if IS_ENABLED(CONFIG_IPV6)
2500 afinfo = xfrm_policy_afinfo[AF_INET6];
2502 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2504 read_unlock_bh(&xfrm_policy_afinfo_lock);
2507 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2509 struct xfrm_policy_afinfo *afinfo;
2510 if (unlikely(family >= NPROTO))
2512 read_lock(&xfrm_policy_afinfo_lock);
2513 afinfo = xfrm_policy_afinfo[family];
2514 if (unlikely(!afinfo))
2515 read_unlock(&xfrm_policy_afinfo_lock);
2519 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2521 read_unlock(&xfrm_policy_afinfo_lock);
2524 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2526 struct net_device *dev = ptr;
2530 xfrm_garbage_collect(dev_net(dev));
2535 static struct notifier_block xfrm_dev_notifier = {
2536 .notifier_call = xfrm_dev_event,
2539 #ifdef CONFIG_XFRM_STATISTICS
2540 static int __net_init xfrm_statistics_init(struct net *net)
2544 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2545 sizeof(struct linux_xfrm_mib),
2546 __alignof__(struct linux_xfrm_mib)) < 0)
2548 rv = xfrm_proc_init(net);
2550 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2554 static void xfrm_statistics_fini(struct net *net)
2556 xfrm_proc_fini(net);
2557 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2560 static int __net_init xfrm_statistics_init(struct net *net)
2565 static void xfrm_statistics_fini(struct net *net)
2570 static int __net_init xfrm_policy_init(struct net *net)
2572 unsigned int hmask, sz;
2575 if (net_eq(net, &init_net))
2576 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2577 sizeof(struct xfrm_dst),
2578 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2582 sz = (hmask+1) * sizeof(struct hlist_head);
2584 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2585 if (!net->xfrm.policy_byidx)
2587 net->xfrm.policy_idx_hmask = hmask;
2589 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2590 struct xfrm_policy_hash *htab;
2592 net->xfrm.policy_count[dir] = 0;
2593 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2595 htab = &net->xfrm.policy_bydst[dir];
2596 htab->table = xfrm_hash_alloc(sz);
2599 htab->hmask = hmask;
2602 INIT_LIST_HEAD(&net->xfrm.policy_all);
2603 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2604 if (net_eq(net, &init_net))
2605 register_netdevice_notifier(&xfrm_dev_notifier);
2609 for (dir--; dir >= 0; dir--) {
2610 struct xfrm_policy_hash *htab;
2612 htab = &net->xfrm.policy_bydst[dir];
2613 xfrm_hash_free(htab->table, sz);
2615 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2620 static void xfrm_policy_fini(struct net *net)
2622 struct xfrm_audit audit_info;
2626 flush_work(&net->xfrm.policy_hash_work);
2627 #ifdef CONFIG_XFRM_SUB_POLICY
2628 audit_info.loginuid = -1;
2629 audit_info.sessionid = -1;
2630 audit_info.secid = 0;
2631 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2633 audit_info.loginuid = -1;
2634 audit_info.sessionid = -1;
2635 audit_info.secid = 0;
2636 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2638 WARN_ON(!list_empty(&net->xfrm.policy_all));
2640 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2641 struct xfrm_policy_hash *htab;
2643 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2645 htab = &net->xfrm.policy_bydst[dir];
2646 sz = (htab->hmask + 1);
2647 WARN_ON(!hlist_empty(htab->table));
2648 xfrm_hash_free(htab->table, sz);
2651 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2652 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2653 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2656 static int __net_init xfrm_net_init(struct net *net)
2660 rv = xfrm_statistics_init(net);
2662 goto out_statistics;
2663 rv = xfrm_state_init(net);
2666 rv = xfrm_policy_init(net);
2669 xfrm_dst_ops_init(net);
2670 rv = xfrm_sysctl_init(net);
2676 xfrm_policy_fini(net);
2678 xfrm_state_fini(net);
2680 xfrm_statistics_fini(net);
2685 static void __net_exit xfrm_net_exit(struct net *net)
2687 xfrm_sysctl_fini(net);
2688 xfrm_policy_fini(net);
2689 xfrm_state_fini(net);
2690 xfrm_statistics_fini(net);
2693 static struct pernet_operations __net_initdata xfrm_net_ops = {
2694 .init = xfrm_net_init,
2695 .exit = xfrm_net_exit,
2698 void __init xfrm_init(void)
2700 register_pernet_subsys(&xfrm_net_ops);
2704 #ifdef CONFIG_AUDITSYSCALL
2705 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2706 struct audit_buffer *audit_buf)
2708 struct xfrm_sec_ctx *ctx = xp->security;
2709 struct xfrm_selector *sel = &xp->selector;
2712 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2713 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2715 switch(sel->family) {
2717 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2718 if (sel->prefixlen_s != 32)
2719 audit_log_format(audit_buf, " src_prefixlen=%d",
2721 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2722 if (sel->prefixlen_d != 32)
2723 audit_log_format(audit_buf, " dst_prefixlen=%d",
2727 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2728 if (sel->prefixlen_s != 128)
2729 audit_log_format(audit_buf, " src_prefixlen=%d",
2731 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2732 if (sel->prefixlen_d != 128)
2733 audit_log_format(audit_buf, " dst_prefixlen=%d",
2739 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2740 uid_t auid, u32 sessionid, u32 secid)
2742 struct audit_buffer *audit_buf;
2744 audit_buf = xfrm_audit_start("SPD-add");
2745 if (audit_buf == NULL)
2747 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2748 audit_log_format(audit_buf, " res=%u", result);
2749 xfrm_audit_common_policyinfo(xp, audit_buf);
2750 audit_log_end(audit_buf);
2752 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2754 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2755 uid_t auid, u32 sessionid, u32 secid)
2757 struct audit_buffer *audit_buf;
2759 audit_buf = xfrm_audit_start("SPD-delete");
2760 if (audit_buf == NULL)
2762 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2763 audit_log_format(audit_buf, " res=%u", result);
2764 xfrm_audit_common_policyinfo(xp, audit_buf);
2765 audit_log_end(audit_buf);
2767 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2770 #ifdef CONFIG_XFRM_MIGRATE
2771 static bool xfrm_migrate_selector_match(const struct xfrm_selector *sel_cmp,
2772 const struct xfrm_selector *sel_tgt)
2774 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2775 if (sel_tgt->family == sel_cmp->family &&
2776 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2777 sel_cmp->family) == 0 &&
2778 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2779 sel_cmp->family) == 0 &&
2780 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2781 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2785 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2792 static struct xfrm_policy * xfrm_migrate_policy_find(const struct xfrm_selector *sel,
2795 struct xfrm_policy *pol, *ret = NULL;
2796 struct hlist_node *entry;
2797 struct hlist_head *chain;
2800 read_lock_bh(&xfrm_policy_lock);
2801 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2802 hlist_for_each_entry(pol, entry, chain, bydst) {
2803 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2804 pol->type == type) {
2806 priority = ret->priority;
2810 chain = &init_net.xfrm.policy_inexact[dir];
2811 hlist_for_each_entry(pol, entry, chain, bydst) {
2812 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2813 pol->type == type &&
2814 pol->priority < priority) {
2823 read_unlock_bh(&xfrm_policy_lock);
2828 static int migrate_tmpl_match(const struct xfrm_migrate *m, const struct xfrm_tmpl *t)
2832 if (t->mode == m->mode && t->id.proto == m->proto &&
2833 (m->reqid == 0 || t->reqid == m->reqid)) {
2835 case XFRM_MODE_TUNNEL:
2836 case XFRM_MODE_BEET:
2837 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2838 m->old_family) == 0 &&
2839 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2840 m->old_family) == 0) {
2844 case XFRM_MODE_TRANSPORT:
2845 /* in case of transport mode, template does not store
2846 any IP addresses, hence we just compare mode and
2857 /* update endpoint address(es) of template(s) */
2858 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2859 struct xfrm_migrate *m, int num_migrate)
2861 struct xfrm_migrate *mp;
2864 write_lock_bh(&pol->lock);
2865 if (unlikely(pol->walk.dead)) {
2866 /* target policy has been deleted */
2867 write_unlock_bh(&pol->lock);
2871 for (i = 0; i < pol->xfrm_nr; i++) {
2872 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2873 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2876 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2877 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2879 /* update endpoints */
2880 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2881 sizeof(pol->xfrm_vec[i].id.daddr));
2882 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2883 sizeof(pol->xfrm_vec[i].saddr));
2884 pol->xfrm_vec[i].encap_family = mp->new_family;
2886 atomic_inc(&pol->genid);
2890 write_unlock_bh(&pol->lock);
2898 static int xfrm_migrate_check(const struct xfrm_migrate *m, int num_migrate)
2902 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2905 for (i = 0; i < num_migrate; i++) {
2906 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2907 m[i].old_family) == 0) &&
2908 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2909 m[i].old_family) == 0))
2911 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2912 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2915 /* check if there is any duplicated entry */
2916 for (j = i + 1; j < num_migrate; j++) {
2917 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2918 sizeof(m[i].old_daddr)) &&
2919 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2920 sizeof(m[i].old_saddr)) &&
2921 m[i].proto == m[j].proto &&
2922 m[i].mode == m[j].mode &&
2923 m[i].reqid == m[j].reqid &&
2924 m[i].old_family == m[j].old_family)
2932 int xfrm_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2933 struct xfrm_migrate *m, int num_migrate,
2934 struct xfrm_kmaddress *k)
2936 int i, err, nx_cur = 0, nx_new = 0;
2937 struct xfrm_policy *pol = NULL;
2938 struct xfrm_state *x, *xc;
2939 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2940 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2941 struct xfrm_migrate *mp;
2943 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2946 /* Stage 1 - find policy */
2947 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2952 /* Stage 2 - find and update state(s) */
2953 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2954 if ((x = xfrm_migrate_state_find(mp))) {
2957 if ((xc = xfrm_state_migrate(x, mp))) {
2967 /* Stage 3 - update policy */
2968 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2971 /* Stage 4 - delete old state(s) */
2973 xfrm_states_put(x_cur, nx_cur);
2974 xfrm_states_delete(x_cur, nx_cur);
2977 /* Stage 5 - announce */
2978 km_migrate(sel, dir, type, m, num_migrate, k);
2990 xfrm_states_put(x_cur, nx_cur);
2992 xfrm_states_delete(x_new, nx_new);
2996 EXPORT_SYMBOL(xfrm_migrate);