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_RWLOCK(xfrm_policy_lock);
42 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
43 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
45 static struct kmem_cache *xfrm_dst_cache __read_mostly;
47 static HLIST_HEAD(xfrm_policy_gc_list);
48 static DEFINE_SPINLOCK(xfrm_policy_gc_lock);
50 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
51 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
52 static void xfrm_init_pmtu(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);
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);
220 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
224 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
226 struct xfrm_policy *policy;
228 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
231 write_pnet(&policy->xp_net, net);
232 INIT_LIST_HEAD(&policy->walk.all);
233 INIT_HLIST_NODE(&policy->bydst);
234 INIT_HLIST_NODE(&policy->byidx);
235 rwlock_init(&policy->lock);
236 atomic_set(&policy->refcnt, 1);
237 setup_timer(&policy->timer, xfrm_policy_timer,
238 (unsigned long)policy);
242 EXPORT_SYMBOL(xfrm_policy_alloc);
244 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
246 void xfrm_policy_destroy(struct xfrm_policy *policy)
248 BUG_ON(!policy->walk.dead);
250 BUG_ON(policy->bundles);
252 if (del_timer(&policy->timer))
255 security_xfrm_policy_free(policy->security);
258 EXPORT_SYMBOL(xfrm_policy_destroy);
260 static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
262 struct dst_entry *dst;
264 while ((dst = policy->bundles) != NULL) {
265 policy->bundles = dst->next;
269 if (del_timer(&policy->timer))
270 atomic_dec(&policy->refcnt);
272 if (atomic_read(&policy->refcnt) > 1)
275 xfrm_pol_put(policy);
278 static void xfrm_policy_gc_task(struct work_struct *work)
280 struct xfrm_policy *policy;
281 struct hlist_node *entry, *tmp;
282 struct hlist_head gc_list;
284 spin_lock_bh(&xfrm_policy_gc_lock);
285 gc_list.first = xfrm_policy_gc_list.first;
286 INIT_HLIST_HEAD(&xfrm_policy_gc_list);
287 spin_unlock_bh(&xfrm_policy_gc_lock);
289 hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
290 xfrm_policy_gc_kill(policy);
292 static DECLARE_WORK(xfrm_policy_gc_work, xfrm_policy_gc_task);
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)
302 write_lock_bh(&policy->lock);
303 dead = policy->walk.dead;
304 policy->walk.dead = 1;
305 write_unlock_bh(&policy->lock);
307 if (unlikely(dead)) {
312 spin_lock_bh(&xfrm_policy_gc_lock);
313 hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
314 spin_unlock_bh(&xfrm_policy_gc_lock);
316 schedule_work(&xfrm_policy_gc_work);
319 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
321 static inline unsigned int idx_hash(struct net *net, u32 index)
323 return __idx_hash(index, net->xfrm.policy_idx_hmask);
326 static struct hlist_head *policy_hash_bysel(struct net *net, struct xfrm_selector *sel, unsigned short family, int dir)
328 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
329 unsigned int hash = __sel_hash(sel, family, hmask);
331 return (hash == hmask + 1 ?
332 &net->xfrm.policy_inexact[dir] :
333 net->xfrm.policy_bydst[dir].table + hash);
336 static struct hlist_head *policy_hash_direct(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
338 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
339 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
341 return net->xfrm.policy_bydst[dir].table + hash;
344 static void xfrm_dst_hash_transfer(struct hlist_head *list,
345 struct hlist_head *ndsttable,
346 unsigned int nhashmask)
348 struct hlist_node *entry, *tmp, *entry0 = NULL;
349 struct xfrm_policy *pol;
353 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
356 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
357 pol->family, nhashmask);
360 hlist_add_head(&pol->bydst, ndsttable+h);
366 hlist_add_after(entry0, &pol->bydst);
370 if (!hlist_empty(list)) {
376 static void xfrm_idx_hash_transfer(struct hlist_head *list,
377 struct hlist_head *nidxtable,
378 unsigned int nhashmask)
380 struct hlist_node *entry, *tmp;
381 struct xfrm_policy *pol;
383 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
386 h = __idx_hash(pol->index, nhashmask);
387 hlist_add_head(&pol->byidx, nidxtable+h);
391 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
393 return ((old_hmask + 1) << 1) - 1;
396 static void xfrm_bydst_resize(struct net *net, int dir)
398 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
399 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
400 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
401 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
402 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
408 write_lock_bh(&xfrm_policy_lock);
410 for (i = hmask; i >= 0; i--)
411 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
413 net->xfrm.policy_bydst[dir].table = ndst;
414 net->xfrm.policy_bydst[dir].hmask = nhashmask;
416 write_unlock_bh(&xfrm_policy_lock);
418 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
421 static void xfrm_byidx_resize(struct net *net, int total)
423 unsigned int hmask = net->xfrm.policy_idx_hmask;
424 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
425 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
426 struct hlist_head *oidx = net->xfrm.policy_byidx;
427 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
433 write_lock_bh(&xfrm_policy_lock);
435 for (i = hmask; i >= 0; i--)
436 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
438 net->xfrm.policy_byidx = nidx;
439 net->xfrm.policy_idx_hmask = nhashmask;
441 write_unlock_bh(&xfrm_policy_lock);
443 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
446 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
448 unsigned int cnt = net->xfrm.policy_count[dir];
449 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
454 if ((hmask + 1) < xfrm_policy_hashmax &&
461 static inline int xfrm_byidx_should_resize(struct net *net, int total)
463 unsigned int hmask = net->xfrm.policy_idx_hmask;
465 if ((hmask + 1) < xfrm_policy_hashmax &&
472 void xfrm_spd_getinfo(struct net *net, struct xfrmk_spdinfo *si)
474 read_lock_bh(&xfrm_policy_lock);
475 si->incnt = net->xfrm.policy_count[XFRM_POLICY_IN];
476 si->outcnt = net->xfrm.policy_count[XFRM_POLICY_OUT];
477 si->fwdcnt = net->xfrm.policy_count[XFRM_POLICY_FWD];
478 si->inscnt = net->xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
479 si->outscnt = net->xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
480 si->fwdscnt = net->xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
481 si->spdhcnt = net->xfrm.policy_idx_hmask;
482 si->spdhmcnt = xfrm_policy_hashmax;
483 read_unlock_bh(&xfrm_policy_lock);
485 EXPORT_SYMBOL(xfrm_spd_getinfo);
487 static DEFINE_MUTEX(hash_resize_mutex);
488 static void xfrm_hash_resize(struct work_struct *work)
490 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
493 mutex_lock(&hash_resize_mutex);
496 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
497 if (xfrm_bydst_should_resize(net, dir, &total))
498 xfrm_bydst_resize(net, dir);
500 if (xfrm_byidx_should_resize(net, total))
501 xfrm_byidx_resize(net, total);
503 mutex_unlock(&hash_resize_mutex);
506 /* Generate new index... KAME seems to generate them ordered by cost
507 * of an absolute inpredictability of ordering of rules. This will not pass. */
508 static u32 xfrm_gen_index(struct net *net, int dir)
510 static u32 idx_generator;
513 struct hlist_node *entry;
514 struct hlist_head *list;
515 struct xfrm_policy *p;
519 idx = (idx_generator | dir);
523 list = net->xfrm.policy_byidx + idx_hash(net, idx);
525 hlist_for_each_entry(p, entry, list, byidx) {
526 if (p->index == idx) {
536 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
538 u32 *p1 = (u32 *) s1;
539 u32 *p2 = (u32 *) s2;
540 int len = sizeof(struct xfrm_selector) / sizeof(u32);
543 for (i = 0; i < len; i++) {
551 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
553 struct net *net = xp_net(policy);
554 struct xfrm_policy *pol;
555 struct xfrm_policy *delpol;
556 struct hlist_head *chain;
557 struct hlist_node *entry, *newpos;
558 struct dst_entry *gc_list;
559 u32 mark = policy->mark.v & policy->mark.m;
561 write_lock_bh(&xfrm_policy_lock);
562 chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
565 hlist_for_each_entry(pol, entry, chain, bydst) {
566 if (pol->type == policy->type &&
567 !selector_cmp(&pol->selector, &policy->selector) &&
568 (mark & pol->mark.m) == pol->mark.v &&
569 xfrm_sec_ctx_match(pol->security, policy->security) &&
572 write_unlock_bh(&xfrm_policy_lock);
576 if (policy->priority > pol->priority)
578 } else if (policy->priority >= pol->priority) {
579 newpos = &pol->bydst;
586 hlist_add_after(newpos, &policy->bydst);
588 hlist_add_head(&policy->bydst, chain);
589 xfrm_pol_hold(policy);
590 net->xfrm.policy_count[dir]++;
591 atomic_inc(&flow_cache_genid);
593 __xfrm_policy_unlink(delpol, dir);
594 policy->index = delpol ? delpol->index : xfrm_gen_index(net, dir);
595 hlist_add_head(&policy->byidx, net->xfrm.policy_byidx+idx_hash(net, policy->index));
596 policy->curlft.add_time = get_seconds();
597 policy->curlft.use_time = 0;
598 if (!mod_timer(&policy->timer, jiffies + HZ))
599 xfrm_pol_hold(policy);
600 list_add(&policy->walk.all, &net->xfrm.policy_all);
601 write_unlock_bh(&xfrm_policy_lock);
604 xfrm_policy_kill(delpol);
605 else if (xfrm_bydst_should_resize(net, dir, NULL))
606 schedule_work(&net->xfrm.policy_hash_work);
608 read_lock_bh(&xfrm_policy_lock);
610 entry = &policy->bydst;
611 hlist_for_each_entry_continue(policy, entry, bydst) {
612 struct dst_entry *dst;
614 write_lock(&policy->lock);
615 dst = policy->bundles;
617 struct dst_entry *tail = dst;
620 tail->next = gc_list;
623 policy->bundles = NULL;
625 write_unlock(&policy->lock);
627 read_unlock_bh(&xfrm_policy_lock);
630 struct dst_entry *dst = gc_list;
638 EXPORT_SYMBOL(xfrm_policy_insert);
640 struct xfrm_policy *xfrm_policy_bysel_ctx(struct net *net, u32 mark, u8 type,
641 int dir, struct xfrm_selector *sel,
642 struct xfrm_sec_ctx *ctx, int delete,
645 struct xfrm_policy *pol, *ret;
646 struct hlist_head *chain;
647 struct hlist_node *entry;
650 write_lock_bh(&xfrm_policy_lock);
651 chain = policy_hash_bysel(net, sel, sel->family, dir);
653 hlist_for_each_entry(pol, entry, chain, bydst) {
654 if (pol->type == type &&
655 (mark & pol->mark.m) == pol->mark.v &&
656 !selector_cmp(sel, &pol->selector) &&
657 xfrm_sec_ctx_match(ctx, pol->security)) {
660 *err = security_xfrm_policy_delete(
663 write_unlock_bh(&xfrm_policy_lock);
666 __xfrm_policy_unlink(pol, dir);
672 write_unlock_bh(&xfrm_policy_lock);
675 atomic_inc(&flow_cache_genid);
676 xfrm_policy_kill(ret);
680 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
682 struct xfrm_policy *xfrm_policy_byid(struct net *net, u32 mark, u8 type,
683 int dir, u32 id, int delete, int *err)
685 struct xfrm_policy *pol, *ret;
686 struct hlist_head *chain;
687 struct hlist_node *entry;
690 if (xfrm_policy_id2dir(id) != dir)
694 write_lock_bh(&xfrm_policy_lock);
695 chain = net->xfrm.policy_byidx + idx_hash(net, id);
697 hlist_for_each_entry(pol, entry, chain, byidx) {
698 if (pol->type == type && pol->index == id &&
699 (mark & pol->mark.m) == pol->mark.v) {
702 *err = security_xfrm_policy_delete(
705 write_unlock_bh(&xfrm_policy_lock);
708 __xfrm_policy_unlink(pol, dir);
714 write_unlock_bh(&xfrm_policy_lock);
717 atomic_inc(&flow_cache_genid);
718 xfrm_policy_kill(ret);
722 EXPORT_SYMBOL(xfrm_policy_byid);
724 #ifdef CONFIG_SECURITY_NETWORK_XFRM
726 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
730 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
731 struct xfrm_policy *pol;
732 struct hlist_node *entry;
735 hlist_for_each_entry(pol, entry,
736 &net->xfrm.policy_inexact[dir], bydst) {
737 if (pol->type != type)
739 err = security_xfrm_policy_delete(pol->security);
741 xfrm_audit_policy_delete(pol, 0,
742 audit_info->loginuid,
743 audit_info->sessionid,
748 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
749 hlist_for_each_entry(pol, entry,
750 net->xfrm.policy_bydst[dir].table + i,
752 if (pol->type != type)
754 err = security_xfrm_policy_delete(
757 xfrm_audit_policy_delete(pol, 0,
758 audit_info->loginuid,
759 audit_info->sessionid,
770 xfrm_policy_flush_secctx_check(struct net *net, u8 type, struct xfrm_audit *audit_info)
776 int xfrm_policy_flush(struct net *net, u8 type, struct xfrm_audit *audit_info)
778 int dir, err = 0, cnt = 0;
779 struct xfrm_policy *dp;
781 write_lock_bh(&xfrm_policy_lock);
783 err = xfrm_policy_flush_secctx_check(net, type, audit_info);
787 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
788 struct xfrm_policy *pol;
789 struct hlist_node *entry;
793 hlist_for_each_entry(pol, entry,
794 &net->xfrm.policy_inexact[dir], bydst) {
795 if (pol->type != type)
797 dp = __xfrm_policy_unlink(pol, dir);
798 write_unlock_bh(&xfrm_policy_lock);
802 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
803 audit_info->sessionid,
806 xfrm_policy_kill(pol);
808 write_lock_bh(&xfrm_policy_lock);
812 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
814 hlist_for_each_entry(pol, entry,
815 net->xfrm.policy_bydst[dir].table + i,
817 if (pol->type != type)
819 dp = __xfrm_policy_unlink(pol, dir);
820 write_unlock_bh(&xfrm_policy_lock);
824 xfrm_audit_policy_delete(pol, 1,
825 audit_info->loginuid,
826 audit_info->sessionid,
828 xfrm_policy_kill(pol);
830 write_lock_bh(&xfrm_policy_lock);
838 atomic_inc(&flow_cache_genid);
840 write_unlock_bh(&xfrm_policy_lock);
843 EXPORT_SYMBOL(xfrm_policy_flush);
845 int xfrm_policy_walk(struct net *net, struct xfrm_policy_walk *walk,
846 int (*func)(struct xfrm_policy *, int, int, void*),
849 struct xfrm_policy *pol;
850 struct xfrm_policy_walk_entry *x;
853 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
854 walk->type != XFRM_POLICY_TYPE_ANY)
857 if (list_empty(&walk->walk.all) && walk->seq != 0)
860 write_lock_bh(&xfrm_policy_lock);
861 if (list_empty(&walk->walk.all))
862 x = list_first_entry(&net->xfrm.policy_all, struct xfrm_policy_walk_entry, all);
864 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
865 list_for_each_entry_from(x, &net->xfrm.policy_all, all) {
868 pol = container_of(x, struct xfrm_policy, walk);
869 if (walk->type != XFRM_POLICY_TYPE_ANY &&
870 walk->type != pol->type)
872 error = func(pol, xfrm_policy_id2dir(pol->index),
875 list_move_tail(&walk->walk.all, &x->all);
880 if (walk->seq == 0) {
884 list_del_init(&walk->walk.all);
886 write_unlock_bh(&xfrm_policy_lock);
889 EXPORT_SYMBOL(xfrm_policy_walk);
891 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
893 INIT_LIST_HEAD(&walk->walk.all);
898 EXPORT_SYMBOL(xfrm_policy_walk_init);
900 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
902 if (list_empty(&walk->walk.all))
905 write_lock_bh(&xfrm_policy_lock);
906 list_del(&walk->walk.all);
907 write_unlock_bh(&xfrm_policy_lock);
909 EXPORT_SYMBOL(xfrm_policy_walk_done);
912 * Find policy to apply to this flow.
914 * Returns 0 if policy found, else an -errno.
916 static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
917 u8 type, u16 family, int dir)
919 struct xfrm_selector *sel = &pol->selector;
920 int match, ret = -ESRCH;
922 if (pol->family != family ||
923 (fl->mark & pol->mark.m) != pol->mark.v ||
927 match = xfrm_selector_match(sel, fl, family);
929 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
935 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
940 struct xfrm_policy *pol, *ret;
941 xfrm_address_t *daddr, *saddr;
942 struct hlist_node *entry;
943 struct hlist_head *chain;
946 daddr = xfrm_flowi_daddr(fl, family);
947 saddr = xfrm_flowi_saddr(fl, family);
948 if (unlikely(!daddr || !saddr))
951 read_lock_bh(&xfrm_policy_lock);
952 chain = policy_hash_direct(net, daddr, saddr, family, dir);
954 hlist_for_each_entry(pol, entry, chain, bydst) {
955 err = xfrm_policy_match(pol, fl, type, family, dir);
965 priority = ret->priority;
969 chain = &net->xfrm.policy_inexact[dir];
970 hlist_for_each_entry(pol, entry, chain, bydst) {
971 err = xfrm_policy_match(pol, fl, type, family, dir);
979 } else if (pol->priority < priority) {
987 read_unlock_bh(&xfrm_policy_lock);
992 static int xfrm_policy_lookup(struct net *net, struct flowi *fl, u16 family,
993 u8 dir, void **objp, atomic_t **obj_refp)
995 struct xfrm_policy *pol;
998 #ifdef CONFIG_XFRM_SUB_POLICY
999 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_SUB, fl, family, dir);
1007 pol = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1012 #ifdef CONFIG_XFRM_SUB_POLICY
1015 if ((*objp = (void *) pol) != NULL)
1016 *obj_refp = &pol->refcnt;
1020 static inline int policy_to_flow_dir(int dir)
1022 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1023 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1024 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1028 case XFRM_POLICY_IN:
1030 case XFRM_POLICY_OUT:
1031 return FLOW_DIR_OUT;
1032 case XFRM_POLICY_FWD:
1033 return FLOW_DIR_FWD;
1037 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
1039 struct xfrm_policy *pol;
1041 read_lock_bh(&xfrm_policy_lock);
1042 if ((pol = sk->sk_policy[dir]) != NULL) {
1043 int match = xfrm_selector_match(&pol->selector, fl,
1048 if ((sk->sk_mark & pol->mark.m) != pol->mark.v) {
1052 err = security_xfrm_policy_lookup(pol->security,
1054 policy_to_flow_dir(dir));
1057 else if (err == -ESRCH)
1065 read_unlock_bh(&xfrm_policy_lock);
1069 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1071 struct net *net = xp_net(pol);
1072 struct hlist_head *chain = policy_hash_bysel(net, &pol->selector,
1075 list_add(&pol->walk.all, &net->xfrm.policy_all);
1076 hlist_add_head(&pol->bydst, chain);
1077 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1078 net->xfrm.policy_count[dir]++;
1081 if (xfrm_bydst_should_resize(net, dir, NULL))
1082 schedule_work(&net->xfrm.policy_hash_work);
1085 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1088 struct net *net = xp_net(pol);
1090 if (hlist_unhashed(&pol->bydst))
1093 hlist_del(&pol->bydst);
1094 hlist_del(&pol->byidx);
1095 list_del(&pol->walk.all);
1096 net->xfrm.policy_count[dir]--;
1101 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1103 write_lock_bh(&xfrm_policy_lock);
1104 pol = __xfrm_policy_unlink(pol, dir);
1105 write_unlock_bh(&xfrm_policy_lock);
1107 if (dir < XFRM_POLICY_MAX)
1108 atomic_inc(&flow_cache_genid);
1109 xfrm_policy_kill(pol);
1114 EXPORT_SYMBOL(xfrm_policy_delete);
1116 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
1118 struct net *net = xp_net(pol);
1119 struct xfrm_policy *old_pol;
1121 #ifdef CONFIG_XFRM_SUB_POLICY
1122 if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
1126 write_lock_bh(&xfrm_policy_lock);
1127 old_pol = sk->sk_policy[dir];
1128 sk->sk_policy[dir] = pol;
1130 pol->curlft.add_time = get_seconds();
1131 pol->index = xfrm_gen_index(net, XFRM_POLICY_MAX+dir);
1132 __xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
1135 __xfrm_policy_unlink(old_pol, XFRM_POLICY_MAX+dir);
1136 write_unlock_bh(&xfrm_policy_lock);
1139 xfrm_policy_kill(old_pol);
1144 static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
1146 struct xfrm_policy *newp = xfrm_policy_alloc(xp_net(old), GFP_ATOMIC);
1149 newp->selector = old->selector;
1150 if (security_xfrm_policy_clone(old->security,
1153 return NULL; /* ENOMEM */
1155 newp->lft = old->lft;
1156 newp->curlft = old->curlft;
1157 newp->action = old->action;
1158 newp->flags = old->flags;
1159 newp->xfrm_nr = old->xfrm_nr;
1160 newp->index = old->index;
1161 newp->type = old->type;
1162 memcpy(newp->xfrm_vec, old->xfrm_vec,
1163 newp->xfrm_nr*sizeof(struct xfrm_tmpl));
1164 write_lock_bh(&xfrm_policy_lock);
1165 __xfrm_policy_link(newp, XFRM_POLICY_MAX+dir);
1166 write_unlock_bh(&xfrm_policy_lock);
1172 int __xfrm_sk_clone_policy(struct sock *sk)
1174 struct xfrm_policy *p0 = sk->sk_policy[0],
1175 *p1 = sk->sk_policy[1];
1177 sk->sk_policy[0] = sk->sk_policy[1] = NULL;
1178 if (p0 && (sk->sk_policy[0] = clone_policy(p0, 0)) == NULL)
1180 if (p1 && (sk->sk_policy[1] = clone_policy(p1, 1)) == NULL)
1186 xfrm_get_saddr(struct net *net, xfrm_address_t *local, xfrm_address_t *remote,
1187 unsigned short family)
1190 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1192 if (unlikely(afinfo == NULL))
1194 err = afinfo->get_saddr(net, local, remote);
1195 xfrm_policy_put_afinfo(afinfo);
1199 /* Resolve list of templates for the flow, given policy. */
1202 xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
1203 struct xfrm_state **xfrm,
1204 unsigned short family)
1206 struct net *net = xp_net(policy);
1209 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1210 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1213 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1214 struct xfrm_state *x;
1215 xfrm_address_t *remote = daddr;
1216 xfrm_address_t *local = saddr;
1217 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1219 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1220 tmpl->mode == XFRM_MODE_BEET) {
1221 remote = &tmpl->id.daddr;
1222 local = &tmpl->saddr;
1223 family = tmpl->encap_family;
1224 if (xfrm_addr_any(local, family)) {
1225 error = xfrm_get_saddr(net, &tmp, remote, family);
1232 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1234 if (x && x->km.state == XFRM_STATE_VALID) {
1241 error = (x->km.state == XFRM_STATE_ERROR ?
1245 else if (error == -ESRCH)
1248 if (!tmpl->optional)
1254 for (nx--; nx>=0; nx--)
1255 xfrm_state_put(xfrm[nx]);
1260 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1261 struct xfrm_state **xfrm,
1262 unsigned short family)
1264 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1265 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1271 for (i = 0; i < npols; i++) {
1272 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1277 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1285 /* found states are sorted for outbound processing */
1287 xfrm_state_sort(xfrm, tpp, cnx, family);
1292 for (cnx--; cnx>=0; cnx--)
1293 xfrm_state_put(tpp[cnx]);
1298 /* Check that the bundle accepts the flow and its components are
1302 static struct dst_entry *
1303 xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family)
1305 struct dst_entry *x;
1306 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1307 if (unlikely(afinfo == NULL))
1308 return ERR_PTR(-EINVAL);
1309 x = afinfo->find_bundle(fl, policy);
1310 xfrm_policy_put_afinfo(afinfo);
1314 static inline int xfrm_get_tos(struct flowi *fl, int family)
1316 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1322 tos = afinfo->get_tos(fl);
1324 xfrm_policy_put_afinfo(afinfo);
1329 static inline struct xfrm_dst *xfrm_alloc_dst(struct net *net, int family)
1331 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1332 struct dst_ops *dst_ops;
1333 struct xfrm_dst *xdst;
1336 return ERR_PTR(-EINVAL);
1340 dst_ops = &net->xfrm.xfrm4_dst_ops;
1342 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1344 dst_ops = &net->xfrm.xfrm6_dst_ops;
1350 xdst = dst_alloc(dst_ops) ?: ERR_PTR(-ENOBUFS);
1352 xfrm_policy_put_afinfo(afinfo);
1357 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1360 struct xfrm_policy_afinfo *afinfo =
1361 xfrm_policy_get_afinfo(dst->ops->family);
1367 err = afinfo->init_path(path, dst, nfheader_len);
1369 xfrm_policy_put_afinfo(afinfo);
1374 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
1376 struct xfrm_policy_afinfo *afinfo =
1377 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1383 err = afinfo->fill_dst(xdst, dev);
1385 xfrm_policy_put_afinfo(afinfo);
1390 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1391 * all the metrics... Shortly, bundle a bundle.
1394 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1395 struct xfrm_state **xfrm, int nx,
1397 struct dst_entry *dst)
1399 struct net *net = xp_net(policy);
1400 unsigned long now = jiffies;
1401 struct net_device *dev;
1402 struct dst_entry *dst_prev = NULL;
1403 struct dst_entry *dst0 = NULL;
1407 int nfheader_len = 0;
1408 int trailer_len = 0;
1410 int family = policy->selector.family;
1411 xfrm_address_t saddr, daddr;
1413 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1415 tos = xfrm_get_tos(fl, family);
1422 for (; i < nx; i++) {
1423 struct xfrm_dst *xdst = xfrm_alloc_dst(net, family);
1424 struct dst_entry *dst1 = &xdst->u.dst;
1426 err = PTR_ERR(xdst);
1435 dst_prev->child = dst_clone(dst1);
1436 dst1->flags |= DST_NOHASH;
1440 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1442 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1443 family = xfrm[i]->props.family;
1444 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1452 dst1->xfrm = xfrm[i];
1453 xdst->genid = xfrm[i]->genid;
1455 dst1->obsolete = -1;
1456 dst1->flags |= DST_HOST;
1457 dst1->lastuse = now;
1459 dst1->input = dst_discard;
1460 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1462 dst1->next = dst_prev;
1465 header_len += xfrm[i]->props.header_len;
1466 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1467 nfheader_len += xfrm[i]->props.header_len;
1468 trailer_len += xfrm[i]->props.trailer_len;
1471 dst_prev->child = dst;
1479 /* Copy neighbour for reachability confirmation */
1480 dst0->neighbour = neigh_clone(dst->neighbour);
1482 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1483 xfrm_init_pmtu(dst_prev);
1485 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1486 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1488 err = xfrm_fill_dst(xdst, dev);
1492 dst_prev->header_len = header_len;
1493 dst_prev->trailer_len = trailer_len;
1494 header_len -= xdst->u.dst.xfrm->props.header_len;
1495 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1503 xfrm_state_put(xfrm[i]);
1507 dst0 = ERR_PTR(err);
1512 xfrm_dst_alloc_copy(void **target, void *src, int size)
1515 *target = kmalloc(size, GFP_ATOMIC);
1519 memcpy(*target, src, size);
1524 xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1526 #ifdef CONFIG_XFRM_SUB_POLICY
1527 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1528 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1536 xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1538 #ifdef CONFIG_XFRM_SUB_POLICY
1539 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1540 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1546 static int stale_bundle(struct dst_entry *dst);
1548 /* Main function: finds/creates a bundle for given flow.
1550 * At the moment we eat a raw IP route. Mostly to speed up lookups
1551 * on interfaces with disabled IPsec.
1553 int __xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1554 struct sock *sk, int flags)
1556 struct xfrm_policy *policy;
1557 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1562 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1563 struct dst_entry *dst, *dst_orig = *dst_p;
1568 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1571 genid = atomic_read(&flow_cache_genid);
1573 for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
1579 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1580 policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1581 err = PTR_ERR(policy);
1582 if (IS_ERR(policy)) {
1583 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1589 /* To accelerate a bit... */
1590 if ((dst_orig->flags & DST_NOXFRM) ||
1591 !net->xfrm.policy_count[XFRM_POLICY_OUT])
1594 policy = flow_cache_lookup(net, fl, dst_orig->ops->family,
1595 dir, xfrm_policy_lookup);
1596 err = PTR_ERR(policy);
1597 if (IS_ERR(policy)) {
1598 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1606 family = dst_orig->ops->family;
1609 xfrm_nr += pols[0]->xfrm_nr;
1612 if ((flags & XFRM_LOOKUP_ICMP) && !(policy->flags & XFRM_POLICY_ICMP))
1615 policy->curlft.use_time = get_seconds();
1617 switch (policy->action) {
1619 case XFRM_POLICY_BLOCK:
1620 /* Prohibit the flow */
1621 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1625 case XFRM_POLICY_ALLOW:
1626 #ifndef CONFIG_XFRM_SUB_POLICY
1627 if (policy->xfrm_nr == 0) {
1628 /* Flow passes not transformed. */
1629 xfrm_pol_put(policy);
1634 /* Try to find matching bundle.
1636 * LATER: help from flow cache. It is optional, this
1637 * is required only for output policy.
1639 dst = xfrm_find_bundle(fl, policy, family);
1641 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1649 #ifdef CONFIG_XFRM_SUB_POLICY
1650 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1651 pols[1] = xfrm_policy_lookup_bytype(net,
1652 XFRM_POLICY_TYPE_MAIN,
1656 if (IS_ERR(pols[1])) {
1657 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLERROR);
1658 err = PTR_ERR(pols[1]);
1661 if (pols[1]->action == XFRM_POLICY_BLOCK) {
1662 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLBLOCK);
1667 xfrm_nr += pols[1]->xfrm_nr;
1672 * Because neither flowi nor bundle information knows about
1673 * transformation template size. On more than one policy usage
1674 * we can realize whether all of them is bypass or not after
1675 * they are searched. See above not-transformed bypass
1676 * is surrounded by non-sub policy configuration, too.
1679 /* Flow passes not transformed. */
1680 xfrm_pols_put(pols, npols);
1685 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1687 if (unlikely(nx<0)) {
1689 if (err == -EAGAIN && net->xfrm.sysctl_larval_drop) {
1690 /* EREMOTE tells the caller to generate
1691 * a one-shot blackhole route.
1693 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1694 xfrm_pol_put(policy);
1697 if (err == -EAGAIN && (flags & XFRM_LOOKUP_WAIT)) {
1698 DECLARE_WAITQUEUE(wait, current);
1700 add_wait_queue(&net->xfrm.km_waitq, &wait);
1701 set_current_state(TASK_INTERRUPTIBLE);
1703 set_current_state(TASK_RUNNING);
1704 remove_wait_queue(&net->xfrm.km_waitq, &wait);
1706 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1708 if (nx == -EAGAIN && signal_pending(current)) {
1709 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1713 if (nx == -EAGAIN ||
1714 genid != atomic_read(&flow_cache_genid)) {
1715 xfrm_pols_put(pols, npols);
1721 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTNOSTATES);
1726 /* Flow passes not transformed. */
1727 xfrm_pols_put(pols, npols);
1731 dst = xfrm_bundle_create(policy, xfrm, nx, fl, dst_orig);
1734 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1738 for (pi = 0; pi < npols; pi++) {
1739 read_lock_bh(&pols[pi]->lock);
1740 pol_dead |= pols[pi]->walk.dead;
1741 read_unlock_bh(&pols[pi]->lock);
1744 write_lock_bh(&policy->lock);
1745 if (unlikely(pol_dead || stale_bundle(dst))) {
1746 /* Wow! While we worked on resolving, this
1747 * policy has gone. Retry. It is not paranoia,
1748 * we just cannot enlist new bundle to dead object.
1749 * We can't enlist stable bundles either.
1751 write_unlock_bh(&policy->lock);
1755 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTPOLDEAD);
1757 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1758 err = -EHOSTUNREACH;
1763 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1765 err = xfrm_dst_update_origin(dst, fl);
1766 if (unlikely(err)) {
1767 write_unlock_bh(&policy->lock);
1769 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1773 dst->next = policy->bundles;
1774 policy->bundles = dst;
1776 write_unlock_bh(&policy->lock);
1779 dst_release(dst_orig);
1780 xfrm_pols_put(pols, npols);
1784 xfrm_pols_put(pols, npols);
1786 dst_release(dst_orig);
1792 if (flags & XFRM_LOOKUP_ICMP)
1796 EXPORT_SYMBOL(__xfrm_lookup);
1798 int xfrm_lookup(struct net *net, struct dst_entry **dst_p, struct flowi *fl,
1799 struct sock *sk, int flags)
1801 int err = __xfrm_lookup(net, dst_p, fl, sk, flags);
1803 if (err == -EREMOTE) {
1804 dst_release(*dst_p);
1811 EXPORT_SYMBOL(xfrm_lookup);
1814 xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1816 struct xfrm_state *x;
1818 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1820 x = skb->sp->xvec[idx];
1821 if (!x->type->reject)
1823 return x->type->reject(x, skb, fl);
1826 /* When skb is transformed back to its "native" form, we have to
1827 * check policy restrictions. At the moment we make this in maximally
1828 * stupid way. Shame on me. :-) Of course, connected sockets must
1829 * have policy cached at them.
1833 xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1834 unsigned short family)
1836 if (xfrm_state_kern(x))
1837 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1838 return x->id.proto == tmpl->id.proto &&
1839 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1840 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1841 x->props.mode == tmpl->mode &&
1842 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1843 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1844 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1845 xfrm_state_addr_cmp(tmpl, x, family));
1849 * 0 or more than 0 is returned when validation is succeeded (either bypass
1850 * because of optional transport mode, or next index of the mathced secpath
1851 * state with the template.
1852 * -1 is returned when no matching template is found.
1853 * Otherwise "-2 - errored_index" is returned.
1856 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1857 unsigned short family)
1861 if (tmpl->optional) {
1862 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1866 for (; idx < sp->len; idx++) {
1867 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1869 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1878 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1879 unsigned int family, int reverse)
1881 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1884 if (unlikely(afinfo == NULL))
1885 return -EAFNOSUPPORT;
1887 afinfo->decode_session(skb, fl, reverse);
1888 err = security_xfrm_decode_session(skb, &fl->secid);
1889 xfrm_policy_put_afinfo(afinfo);
1892 EXPORT_SYMBOL(__xfrm_decode_session);
1894 static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1896 for (; k < sp->len; k++) {
1897 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
1906 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1907 unsigned short family)
1909 struct net *net = dev_net(skb->dev);
1910 struct xfrm_policy *pol;
1911 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1920 reverse = dir & ~XFRM_POLICY_MASK;
1921 dir &= XFRM_POLICY_MASK;
1922 fl_dir = policy_to_flow_dir(dir);
1924 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
1925 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
1929 nf_nat_decode_session(skb, &fl, family);
1931 /* First, check used SA against their selectors. */
1935 for (i=skb->sp->len-1; i>=0; i--) {
1936 struct xfrm_state *x = skb->sp->xvec[i];
1937 if (!xfrm_selector_match(&x->sel, &fl, family)) {
1938 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
1945 if (sk && sk->sk_policy[dir]) {
1946 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
1948 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
1954 pol = flow_cache_lookup(net, &fl, family, fl_dir,
1955 xfrm_policy_lookup);
1958 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
1963 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
1964 xfrm_secpath_reject(xerr_idx, skb, &fl);
1965 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOPOLS);
1971 pol->curlft.use_time = get_seconds();
1975 #ifdef CONFIG_XFRM_SUB_POLICY
1976 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1977 pols[1] = xfrm_policy_lookup_bytype(net, XFRM_POLICY_TYPE_MAIN,
1981 if (IS_ERR(pols[1])) {
1982 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLERROR);
1985 pols[1]->curlft.use_time = get_seconds();
1991 if (pol->action == XFRM_POLICY_ALLOW) {
1992 struct sec_path *sp;
1993 static struct sec_path dummy;
1994 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
1995 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
1996 struct xfrm_tmpl **tpp = tp;
2000 if ((sp = skb->sp) == NULL)
2003 for (pi = 0; pi < npols; pi++) {
2004 if (pols[pi] != pol &&
2005 pols[pi]->action != XFRM_POLICY_ALLOW) {
2006 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2009 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
2010 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
2013 for (i = 0; i < pols[pi]->xfrm_nr; i++)
2014 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2018 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2022 /* For each tunnel xfrm, find the first matching tmpl.
2023 * For each tmpl before that, find corresponding xfrm.
2024 * Order is _important_. Later we will implement
2025 * some barriers, but at the moment barriers
2026 * are implied between each two transformations.
2028 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2029 k = xfrm_policy_ok(tpp[i], sp, k, family);
2032 /* "-2 - errored_index" returned */
2034 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2039 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2040 XFRM_INC_STATS(net, LINUX_MIB_XFRMINTMPLMISMATCH);
2044 xfrm_pols_put(pols, npols);
2047 XFRM_INC_STATS(net, LINUX_MIB_XFRMINPOLBLOCK);
2050 xfrm_secpath_reject(xerr_idx, skb, &fl);
2052 xfrm_pols_put(pols, npols);
2055 EXPORT_SYMBOL(__xfrm_policy_check);
2057 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2059 struct net *net = dev_net(skb->dev);
2061 struct dst_entry *dst;
2064 if (xfrm_decode_session(skb, &fl, family) < 0) {
2065 XFRM_INC_STATS(net, LINUX_MIB_XFRMFWDHDRERROR);
2071 res = xfrm_lookup(net, &dst, &fl, NULL, 0) == 0;
2072 skb_dst_set(skb, dst);
2075 EXPORT_SYMBOL(__xfrm_route_forward);
2077 /* Optimize later using cookies and generation ids. */
2079 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2081 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2082 * to "-1" to force all XFRM destinations to get validated by
2083 * dst_ops->check on every use. We do this because when a
2084 * normal route referenced by an XFRM dst is obsoleted we do
2085 * not go looking around for all parent referencing XFRM dsts
2086 * so that we can invalidate them. It is just too much work.
2087 * Instead we make the checks here on every use. For example:
2089 * XFRM dst A --> IPv4 dst X
2091 * X is the "xdst->route" of A (X is also the "dst->path" of A
2092 * in this example). If X is marked obsolete, "A" will not
2093 * notice. That's what we are validating here via the
2094 * stale_bundle() check.
2096 * When a policy's bundle is pruned, we dst_free() the XFRM
2097 * dst which causes it's ->obsolete field to be set to a
2098 * positive non-zero integer. If an XFRM dst has been pruned
2099 * like this, we want to force a new route lookup.
2101 if (dst->obsolete < 0 && !stale_bundle(dst))
2107 static int stale_bundle(struct dst_entry *dst)
2109 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
2112 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2114 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2115 dst->dev = dev_net(dev)->loopback_dev;
2120 EXPORT_SYMBOL(xfrm_dst_ifdown);
2122 static void xfrm_link_failure(struct sk_buff *skb)
2124 /* Impossible. Such dst must be popped before reaches point of failure. */
2128 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2131 if (dst->obsolete) {
2139 static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
2141 struct dst_entry *dst, **dstp;
2143 write_lock(&pol->lock);
2144 dstp = &pol->bundles;
2145 while ((dst=*dstp) != NULL) {
2148 dst->next = *gc_list_p;
2154 write_unlock(&pol->lock);
2157 static void xfrm_prune_bundles(struct net *net, int (*func)(struct dst_entry *))
2159 struct dst_entry *gc_list = NULL;
2162 read_lock_bh(&xfrm_policy_lock);
2163 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2164 struct xfrm_policy *pol;
2165 struct hlist_node *entry;
2166 struct hlist_head *table;
2169 hlist_for_each_entry(pol, entry,
2170 &net->xfrm.policy_inexact[dir], bydst)
2171 prune_one_bundle(pol, func, &gc_list);
2173 table = net->xfrm.policy_bydst[dir].table;
2174 for (i = net->xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2175 hlist_for_each_entry(pol, entry, table + i, bydst)
2176 prune_one_bundle(pol, func, &gc_list);
2179 read_unlock_bh(&xfrm_policy_lock);
2182 struct dst_entry *dst = gc_list;
2183 gc_list = dst->next;
2188 static int unused_bundle(struct dst_entry *dst)
2190 return !atomic_read(&dst->__refcnt);
2193 static void __xfrm_garbage_collect(struct net *net)
2195 xfrm_prune_bundles(net, unused_bundle);
2198 static int xfrm_flush_bundles(struct net *net)
2200 xfrm_prune_bundles(net, stale_bundle);
2204 static void xfrm_init_pmtu(struct dst_entry *dst)
2207 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2208 u32 pmtu, route_mtu_cached;
2210 pmtu = dst_mtu(dst->child);
2211 xdst->child_mtu_cached = pmtu;
2213 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2215 route_mtu_cached = dst_mtu(xdst->route);
2216 xdst->route_mtu_cached = route_mtu_cached;
2218 if (pmtu > route_mtu_cached)
2219 pmtu = route_mtu_cached;
2221 dst->metrics[RTAX_MTU-1] = pmtu;
2222 } while ((dst = dst->next));
2225 /* Check that the bundle accepts the flow and its components are
2229 int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2230 struct flowi *fl, int family, int strict)
2232 struct dst_entry *dst = &first->u.dst;
2233 struct xfrm_dst *last;
2236 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2237 (dst->dev && !netif_running(dst->dev)))
2239 #ifdef CONFIG_XFRM_SUB_POLICY
2241 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2243 if (first->partner &&
2244 !xfrm_selector_match(first->partner, fl, family))
2252 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2254 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2257 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
2259 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2261 if (xdst->genid != dst->xfrm->genid)
2265 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2266 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2269 mtu = dst_mtu(dst->child);
2270 if (xdst->child_mtu_cached != mtu) {
2272 xdst->child_mtu_cached = mtu;
2275 if (!dst_check(xdst->route, xdst->route_cookie))
2277 mtu = dst_mtu(xdst->route);
2278 if (xdst->route_mtu_cached != mtu) {
2280 xdst->route_mtu_cached = mtu;
2284 } while (dst->xfrm);
2289 mtu = last->child_mtu_cached;
2293 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2294 if (mtu > last->route_mtu_cached)
2295 mtu = last->route_mtu_cached;
2296 dst->metrics[RTAX_MTU-1] = mtu;
2301 last = (struct xfrm_dst *)last->u.dst.next;
2302 last->child_mtu_cached = mtu;
2308 EXPORT_SYMBOL(xfrm_bundle_ok);
2310 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2314 if (unlikely(afinfo == NULL))
2316 if (unlikely(afinfo->family >= NPROTO))
2317 return -EAFNOSUPPORT;
2318 write_lock_bh(&xfrm_policy_afinfo_lock);
2319 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2322 struct dst_ops *dst_ops = afinfo->dst_ops;
2323 if (likely(dst_ops->kmem_cachep == NULL))
2324 dst_ops->kmem_cachep = xfrm_dst_cache;
2325 if (likely(dst_ops->check == NULL))
2326 dst_ops->check = xfrm_dst_check;
2327 if (likely(dst_ops->negative_advice == NULL))
2328 dst_ops->negative_advice = xfrm_negative_advice;
2329 if (likely(dst_ops->link_failure == NULL))
2330 dst_ops->link_failure = xfrm_link_failure;
2331 if (likely(afinfo->garbage_collect == NULL))
2332 afinfo->garbage_collect = __xfrm_garbage_collect;
2333 xfrm_policy_afinfo[afinfo->family] = afinfo;
2335 write_unlock_bh(&xfrm_policy_afinfo_lock);
2339 struct dst_ops *xfrm_dst_ops;
2341 switch (afinfo->family) {
2343 xfrm_dst_ops = &net->xfrm.xfrm4_dst_ops;
2345 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2347 xfrm_dst_ops = &net->xfrm.xfrm6_dst_ops;
2353 *xfrm_dst_ops = *afinfo->dst_ops;
2359 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2361 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2364 if (unlikely(afinfo == NULL))
2366 if (unlikely(afinfo->family >= NPROTO))
2367 return -EAFNOSUPPORT;
2368 write_lock_bh(&xfrm_policy_afinfo_lock);
2369 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2370 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2373 struct dst_ops *dst_ops = afinfo->dst_ops;
2374 xfrm_policy_afinfo[afinfo->family] = NULL;
2375 dst_ops->kmem_cachep = NULL;
2376 dst_ops->check = NULL;
2377 dst_ops->negative_advice = NULL;
2378 dst_ops->link_failure = NULL;
2379 afinfo->garbage_collect = NULL;
2382 write_unlock_bh(&xfrm_policy_afinfo_lock);
2385 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2387 static void __net_init xfrm_dst_ops_init(struct net *net)
2389 struct xfrm_policy_afinfo *afinfo;
2391 read_lock_bh(&xfrm_policy_afinfo_lock);
2392 afinfo = xfrm_policy_afinfo[AF_INET];
2394 net->xfrm.xfrm4_dst_ops = *afinfo->dst_ops;
2395 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
2396 afinfo = xfrm_policy_afinfo[AF_INET6];
2398 net->xfrm.xfrm6_dst_ops = *afinfo->dst_ops;
2400 read_unlock_bh(&xfrm_policy_afinfo_lock);
2403 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2405 struct xfrm_policy_afinfo *afinfo;
2406 if (unlikely(family >= NPROTO))
2408 read_lock(&xfrm_policy_afinfo_lock);
2409 afinfo = xfrm_policy_afinfo[family];
2410 if (unlikely(!afinfo))
2411 read_unlock(&xfrm_policy_afinfo_lock);
2415 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2417 read_unlock(&xfrm_policy_afinfo_lock);
2420 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2422 struct net_device *dev = ptr;
2426 xfrm_flush_bundles(dev_net(dev));
2431 static struct notifier_block xfrm_dev_notifier = {
2432 .notifier_call = xfrm_dev_event,
2435 #ifdef CONFIG_XFRM_STATISTICS
2436 static int __net_init xfrm_statistics_init(struct net *net)
2440 if (snmp_mib_init((void __percpu **)net->mib.xfrm_statistics,
2441 sizeof(struct linux_xfrm_mib)) < 0)
2443 rv = xfrm_proc_init(net);
2445 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2449 static void xfrm_statistics_fini(struct net *net)
2451 xfrm_proc_fini(net);
2452 snmp_mib_free((void __percpu **)net->mib.xfrm_statistics);
2455 static int __net_init xfrm_statistics_init(struct net *net)
2460 static void xfrm_statistics_fini(struct net *net)
2465 static int __net_init xfrm_policy_init(struct net *net)
2467 unsigned int hmask, sz;
2470 if (net_eq(net, &init_net))
2471 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2472 sizeof(struct xfrm_dst),
2473 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2477 sz = (hmask+1) * sizeof(struct hlist_head);
2479 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2480 if (!net->xfrm.policy_byidx)
2482 net->xfrm.policy_idx_hmask = hmask;
2484 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2485 struct xfrm_policy_hash *htab;
2487 net->xfrm.policy_count[dir] = 0;
2488 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2490 htab = &net->xfrm.policy_bydst[dir];
2491 htab->table = xfrm_hash_alloc(sz);
2494 htab->hmask = hmask;
2497 INIT_LIST_HEAD(&net->xfrm.policy_all);
2498 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2499 if (net_eq(net, &init_net))
2500 register_netdevice_notifier(&xfrm_dev_notifier);
2504 for (dir--; dir >= 0; dir--) {
2505 struct xfrm_policy_hash *htab;
2507 htab = &net->xfrm.policy_bydst[dir];
2508 xfrm_hash_free(htab->table, sz);
2510 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2515 static void xfrm_policy_fini(struct net *net)
2517 struct xfrm_audit audit_info;
2521 flush_work(&net->xfrm.policy_hash_work);
2522 #ifdef CONFIG_XFRM_SUB_POLICY
2523 audit_info.loginuid = -1;
2524 audit_info.sessionid = -1;
2525 audit_info.secid = 0;
2526 xfrm_policy_flush(net, XFRM_POLICY_TYPE_SUB, &audit_info);
2528 audit_info.loginuid = -1;
2529 audit_info.sessionid = -1;
2530 audit_info.secid = 0;
2531 xfrm_policy_flush(net, XFRM_POLICY_TYPE_MAIN, &audit_info);
2532 flush_work(&xfrm_policy_gc_work);
2534 WARN_ON(!list_empty(&net->xfrm.policy_all));
2536 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2537 struct xfrm_policy_hash *htab;
2539 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2541 htab = &net->xfrm.policy_bydst[dir];
2542 sz = (htab->hmask + 1);
2543 WARN_ON(!hlist_empty(htab->table));
2544 xfrm_hash_free(htab->table, sz);
2547 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2548 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2549 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2552 static int __net_init xfrm_net_init(struct net *net)
2556 rv = xfrm_statistics_init(net);
2558 goto out_statistics;
2559 rv = xfrm_state_init(net);
2562 rv = xfrm_policy_init(net);
2565 xfrm_dst_ops_init(net);
2566 rv = xfrm_sysctl_init(net);
2572 xfrm_policy_fini(net);
2574 xfrm_state_fini(net);
2576 xfrm_statistics_fini(net);
2581 static void __net_exit xfrm_net_exit(struct net *net)
2583 xfrm_sysctl_fini(net);
2584 xfrm_policy_fini(net);
2585 xfrm_state_fini(net);
2586 xfrm_statistics_fini(net);
2589 static struct pernet_operations __net_initdata xfrm_net_ops = {
2590 .init = xfrm_net_init,
2591 .exit = xfrm_net_exit,
2594 void __init xfrm_init(void)
2596 register_pernet_subsys(&xfrm_net_ops);
2600 #ifdef CONFIG_AUDITSYSCALL
2601 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2602 struct audit_buffer *audit_buf)
2604 struct xfrm_sec_ctx *ctx = xp->security;
2605 struct xfrm_selector *sel = &xp->selector;
2608 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2609 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2611 switch(sel->family) {
2613 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2614 if (sel->prefixlen_s != 32)
2615 audit_log_format(audit_buf, " src_prefixlen=%d",
2617 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2618 if (sel->prefixlen_d != 32)
2619 audit_log_format(audit_buf, " dst_prefixlen=%d",
2623 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2624 if (sel->prefixlen_s != 128)
2625 audit_log_format(audit_buf, " src_prefixlen=%d",
2627 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2628 if (sel->prefixlen_d != 128)
2629 audit_log_format(audit_buf, " dst_prefixlen=%d",
2635 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2636 uid_t auid, u32 sessionid, u32 secid)
2638 struct audit_buffer *audit_buf;
2640 audit_buf = xfrm_audit_start("SPD-add");
2641 if (audit_buf == NULL)
2643 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2644 audit_log_format(audit_buf, " res=%u", result);
2645 xfrm_audit_common_policyinfo(xp, audit_buf);
2646 audit_log_end(audit_buf);
2648 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2650 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2651 uid_t auid, u32 sessionid, u32 secid)
2653 struct audit_buffer *audit_buf;
2655 audit_buf = xfrm_audit_start("SPD-delete");
2656 if (audit_buf == NULL)
2658 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2659 audit_log_format(audit_buf, " res=%u", result);
2660 xfrm_audit_common_policyinfo(xp, audit_buf);
2661 audit_log_end(audit_buf);
2663 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2666 #ifdef CONFIG_XFRM_MIGRATE
2667 static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2668 struct xfrm_selector *sel_tgt)
2670 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2671 if (sel_tgt->family == sel_cmp->family &&
2672 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2673 sel_cmp->family) == 0 &&
2674 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2675 sel_cmp->family) == 0 &&
2676 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2677 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2681 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2688 static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2691 struct xfrm_policy *pol, *ret = NULL;
2692 struct hlist_node *entry;
2693 struct hlist_head *chain;
2696 read_lock_bh(&xfrm_policy_lock);
2697 chain = policy_hash_direct(&init_net, &sel->daddr, &sel->saddr, sel->family, dir);
2698 hlist_for_each_entry(pol, entry, chain, bydst) {
2699 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2700 pol->type == type) {
2702 priority = ret->priority;
2706 chain = &init_net.xfrm.policy_inexact[dir];
2707 hlist_for_each_entry(pol, entry, chain, bydst) {
2708 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2709 pol->type == type &&
2710 pol->priority < priority) {
2719 read_unlock_bh(&xfrm_policy_lock);
2724 static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2728 if (t->mode == m->mode && t->id.proto == m->proto &&
2729 (m->reqid == 0 || t->reqid == m->reqid)) {
2731 case XFRM_MODE_TUNNEL:
2732 case XFRM_MODE_BEET:
2733 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2734 m->old_family) == 0 &&
2735 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2736 m->old_family) == 0) {
2740 case XFRM_MODE_TRANSPORT:
2741 /* in case of transport mode, template does not store
2742 any IP addresses, hence we just compare mode and
2753 /* update endpoint address(es) of template(s) */
2754 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2755 struct xfrm_migrate *m, int num_migrate)
2757 struct xfrm_migrate *mp;
2758 struct dst_entry *dst;
2761 write_lock_bh(&pol->lock);
2762 if (unlikely(pol->walk.dead)) {
2763 /* target policy has been deleted */
2764 write_unlock_bh(&pol->lock);
2768 for (i = 0; i < pol->xfrm_nr; i++) {
2769 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2770 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2773 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2774 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2776 /* update endpoints */
2777 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2778 sizeof(pol->xfrm_vec[i].id.daddr));
2779 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2780 sizeof(pol->xfrm_vec[i].saddr));
2781 pol->xfrm_vec[i].encap_family = mp->new_family;
2783 while ((dst = pol->bundles) != NULL) {
2784 pol->bundles = dst->next;
2790 write_unlock_bh(&pol->lock);
2798 static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2802 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2805 for (i = 0; i < num_migrate; i++) {
2806 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2807 m[i].old_family) == 0) &&
2808 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2809 m[i].old_family) == 0))
2811 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2812 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2815 /* check if there is any duplicated entry */
2816 for (j = i + 1; j < num_migrate; j++) {
2817 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2818 sizeof(m[i].old_daddr)) &&
2819 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2820 sizeof(m[i].old_saddr)) &&
2821 m[i].proto == m[j].proto &&
2822 m[i].mode == m[j].mode &&
2823 m[i].reqid == m[j].reqid &&
2824 m[i].old_family == m[j].old_family)
2832 int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
2833 struct xfrm_migrate *m, int num_migrate,
2834 struct xfrm_kmaddress *k)
2836 int i, err, nx_cur = 0, nx_new = 0;
2837 struct xfrm_policy *pol = NULL;
2838 struct xfrm_state *x, *xc;
2839 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2840 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2841 struct xfrm_migrate *mp;
2843 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2846 /* Stage 1 - find policy */
2847 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2852 /* Stage 2 - find and update state(s) */
2853 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2854 if ((x = xfrm_migrate_state_find(mp))) {
2857 if ((xc = xfrm_state_migrate(x, mp))) {
2867 /* Stage 3 - update policy */
2868 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2871 /* Stage 4 - delete old state(s) */
2873 xfrm_states_put(x_cur, nx_cur);
2874 xfrm_states_delete(x_cur, nx_cur);
2877 /* Stage 5 - announce */
2878 km_migrate(sel, dir, type, m, num_migrate, k);
2890 xfrm_states_put(x_cur, nx_cur);
2892 xfrm_states_delete(x_new, nx_new);
2896 EXPORT_SYMBOL(xfrm_migrate);