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 int sysctl_xfrm_larval_drop __read_mostly = 1;
39 #ifdef CONFIG_XFRM_STATISTICS
40 DEFINE_SNMP_STAT(struct linux_xfrm_mib, xfrm_statistics) __read_mostly;
41 EXPORT_SYMBOL(xfrm_statistics);
44 DEFINE_MUTEX(xfrm_cfg_mutex);
45 EXPORT_SYMBOL(xfrm_cfg_mutex);
47 static DEFINE_RWLOCK(xfrm_policy_lock);
49 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
50 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
52 static struct kmem_cache *xfrm_dst_cache __read_mostly;
54 static HLIST_HEAD(xfrm_policy_gc_list);
55 static DEFINE_SPINLOCK(xfrm_policy_gc_lock);
57 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
58 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo);
59 static void xfrm_init_pmtu(struct dst_entry *dst);
62 __xfrm4_selector_match(struct xfrm_selector *sel, struct flowi *fl)
64 return addr_match(&fl->fl4_dst, &sel->daddr, sel->prefixlen_d) &&
65 addr_match(&fl->fl4_src, &sel->saddr, sel->prefixlen_s) &&
66 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
67 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
68 (fl->proto == sel->proto || !sel->proto) &&
69 (fl->oif == sel->ifindex || !sel->ifindex);
73 __xfrm6_selector_match(struct xfrm_selector *sel, struct flowi *fl)
75 return addr_match(&fl->fl6_dst, &sel->daddr, sel->prefixlen_d) &&
76 addr_match(&fl->fl6_src, &sel->saddr, sel->prefixlen_s) &&
77 !((xfrm_flowi_dport(fl) ^ sel->dport) & sel->dport_mask) &&
78 !((xfrm_flowi_sport(fl) ^ sel->sport) & sel->sport_mask) &&
79 (fl->proto == sel->proto || !sel->proto) &&
80 (fl->oif == sel->ifindex || !sel->ifindex);
83 int xfrm_selector_match(struct xfrm_selector *sel, struct flowi *fl,
84 unsigned short family)
88 return __xfrm4_selector_match(sel, fl);
90 return __xfrm6_selector_match(sel, fl);
95 static inline struct dst_entry *__xfrm_dst_lookup(int tos,
96 xfrm_address_t *saddr,
97 xfrm_address_t *daddr,
100 struct xfrm_policy_afinfo *afinfo;
101 struct dst_entry *dst;
103 afinfo = xfrm_policy_get_afinfo(family);
104 if (unlikely(afinfo == NULL))
105 return ERR_PTR(-EAFNOSUPPORT);
107 dst = afinfo->dst_lookup(tos, saddr, daddr);
109 xfrm_policy_put_afinfo(afinfo);
114 static inline struct dst_entry *xfrm_dst_lookup(struct xfrm_state *x, int tos,
115 xfrm_address_t *prev_saddr,
116 xfrm_address_t *prev_daddr,
119 xfrm_address_t *saddr = &x->props.saddr;
120 xfrm_address_t *daddr = &x->id.daddr;
121 struct dst_entry *dst;
123 if (x->type->flags & XFRM_TYPE_LOCAL_COADDR) {
127 if (x->type->flags & XFRM_TYPE_REMOTE_COADDR) {
132 dst = __xfrm_dst_lookup(tos, saddr, daddr, family);
135 if (prev_saddr != saddr)
136 memcpy(prev_saddr, saddr, sizeof(*prev_saddr));
137 if (prev_daddr != daddr)
138 memcpy(prev_daddr, daddr, sizeof(*prev_daddr));
144 static inline unsigned long make_jiffies(long secs)
146 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
147 return MAX_SCHEDULE_TIMEOUT-1;
152 static void xfrm_policy_timer(unsigned long data)
154 struct xfrm_policy *xp = (struct xfrm_policy*)data;
155 unsigned long now = get_seconds();
156 long next = LONG_MAX;
160 read_lock(&xp->lock);
165 dir = xfrm_policy_id2dir(xp->index);
167 if (xp->lft.hard_add_expires_seconds) {
168 long tmo = xp->lft.hard_add_expires_seconds +
169 xp->curlft.add_time - now;
175 if (xp->lft.hard_use_expires_seconds) {
176 long tmo = xp->lft.hard_use_expires_seconds +
177 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
183 if (xp->lft.soft_add_expires_seconds) {
184 long tmo = xp->lft.soft_add_expires_seconds +
185 xp->curlft.add_time - now;
188 tmo = XFRM_KM_TIMEOUT;
193 if (xp->lft.soft_use_expires_seconds) {
194 long tmo = xp->lft.soft_use_expires_seconds +
195 (xp->curlft.use_time ? : xp->curlft.add_time) - now;
198 tmo = XFRM_KM_TIMEOUT;
205 km_policy_expired(xp, dir, 0, 0);
206 if (next != LONG_MAX &&
207 !mod_timer(&xp->timer, jiffies + make_jiffies(next)))
211 read_unlock(&xp->lock);
216 read_unlock(&xp->lock);
217 if (!xfrm_policy_delete(xp, dir))
218 km_policy_expired(xp, dir, 1, 0);
223 /* Allocate xfrm_policy. Not used here, it is supposed to be used by pfkeyv2
227 struct xfrm_policy *xfrm_policy_alloc(struct net *net, gfp_t gfp)
229 struct xfrm_policy *policy;
231 policy = kzalloc(sizeof(struct xfrm_policy), gfp);
234 write_pnet(&policy->xp_net, net);
235 INIT_LIST_HEAD(&policy->walk.all);
236 INIT_HLIST_NODE(&policy->bydst);
237 INIT_HLIST_NODE(&policy->byidx);
238 rwlock_init(&policy->lock);
239 atomic_set(&policy->refcnt, 1);
240 setup_timer(&policy->timer, xfrm_policy_timer,
241 (unsigned long)policy);
245 EXPORT_SYMBOL(xfrm_policy_alloc);
247 /* Destroy xfrm_policy: descendant resources must be released to this moment. */
249 void xfrm_policy_destroy(struct xfrm_policy *policy)
251 BUG_ON(!policy->walk.dead);
253 BUG_ON(policy->bundles);
255 if (del_timer(&policy->timer))
258 security_xfrm_policy_free(policy->security);
261 EXPORT_SYMBOL(xfrm_policy_destroy);
263 static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
265 struct dst_entry *dst;
267 while ((dst = policy->bundles) != NULL) {
268 policy->bundles = dst->next;
272 if (del_timer(&policy->timer))
273 atomic_dec(&policy->refcnt);
275 if (atomic_read(&policy->refcnt) > 1)
278 xfrm_pol_put(policy);
281 static void xfrm_policy_gc_task(struct work_struct *work)
283 struct xfrm_policy *policy;
284 struct hlist_node *entry, *tmp;
285 struct hlist_head gc_list;
287 spin_lock_bh(&xfrm_policy_gc_lock);
288 gc_list.first = xfrm_policy_gc_list.first;
289 INIT_HLIST_HEAD(&xfrm_policy_gc_list);
290 spin_unlock_bh(&xfrm_policy_gc_lock);
292 hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
293 xfrm_policy_gc_kill(policy);
295 static DECLARE_WORK(xfrm_policy_gc_work, xfrm_policy_gc_task);
297 /* Rule must be locked. Release descentant resources, announce
298 * entry dead. The rule must be unlinked from lists to the moment.
301 static void xfrm_policy_kill(struct xfrm_policy *policy)
305 write_lock_bh(&policy->lock);
306 dead = policy->walk.dead;
307 policy->walk.dead = 1;
308 write_unlock_bh(&policy->lock);
310 if (unlikely(dead)) {
315 spin_lock_bh(&xfrm_policy_gc_lock);
316 hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
317 spin_unlock_bh(&xfrm_policy_gc_lock);
319 schedule_work(&xfrm_policy_gc_work);
322 static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
324 static inline unsigned int idx_hash(struct net *net, u32 index)
326 return __idx_hash(index, net->xfrm.policy_idx_hmask);
329 static struct hlist_head *policy_hash_bysel(struct xfrm_selector *sel, unsigned short family, int dir)
331 unsigned int hmask = init_net.xfrm.policy_bydst[dir].hmask;
332 unsigned int hash = __sel_hash(sel, family, hmask);
334 return (hash == hmask + 1 ?
335 &init_net.xfrm.policy_inexact[dir] :
336 init_net.xfrm.policy_bydst[dir].table + hash);
339 static struct hlist_head *policy_hash_direct(xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
341 unsigned int hmask = init_net.xfrm.policy_bydst[dir].hmask;
342 unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
344 return init_net.xfrm.policy_bydst[dir].table + hash;
347 static void xfrm_dst_hash_transfer(struct hlist_head *list,
348 struct hlist_head *ndsttable,
349 unsigned int nhashmask)
351 struct hlist_node *entry, *tmp, *entry0 = NULL;
352 struct xfrm_policy *pol;
356 hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
359 h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
360 pol->family, nhashmask);
363 hlist_add_head(&pol->bydst, ndsttable+h);
369 hlist_add_after(entry0, &pol->bydst);
373 if (!hlist_empty(list)) {
379 static void xfrm_idx_hash_transfer(struct hlist_head *list,
380 struct hlist_head *nidxtable,
381 unsigned int nhashmask)
383 struct hlist_node *entry, *tmp;
384 struct xfrm_policy *pol;
386 hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
389 h = __idx_hash(pol->index, nhashmask);
390 hlist_add_head(&pol->byidx, nidxtable+h);
394 static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
396 return ((old_hmask + 1) << 1) - 1;
399 static void xfrm_bydst_resize(struct net *net, int dir)
401 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
402 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
403 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
404 struct hlist_head *odst = net->xfrm.policy_bydst[dir].table;
405 struct hlist_head *ndst = xfrm_hash_alloc(nsize);
411 write_lock_bh(&xfrm_policy_lock);
413 for (i = hmask; i >= 0; i--)
414 xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
416 net->xfrm.policy_bydst[dir].table = ndst;
417 net->xfrm.policy_bydst[dir].hmask = nhashmask;
419 write_unlock_bh(&xfrm_policy_lock);
421 xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
424 static void xfrm_byidx_resize(struct net *net, int total)
426 unsigned int hmask = net->xfrm.policy_idx_hmask;
427 unsigned int nhashmask = xfrm_new_hash_mask(hmask);
428 unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
429 struct hlist_head *oidx = net->xfrm.policy_byidx;
430 struct hlist_head *nidx = xfrm_hash_alloc(nsize);
436 write_lock_bh(&xfrm_policy_lock);
438 for (i = hmask; i >= 0; i--)
439 xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
441 net->xfrm.policy_byidx = nidx;
442 net->xfrm.policy_idx_hmask = nhashmask;
444 write_unlock_bh(&xfrm_policy_lock);
446 xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
449 static inline int xfrm_bydst_should_resize(struct net *net, int dir, int *total)
451 unsigned int cnt = net->xfrm.policy_count[dir];
452 unsigned int hmask = net->xfrm.policy_bydst[dir].hmask;
457 if ((hmask + 1) < xfrm_policy_hashmax &&
464 static inline int xfrm_byidx_should_resize(struct net *net, int total)
466 unsigned int hmask = net->xfrm.policy_idx_hmask;
468 if ((hmask + 1) < xfrm_policy_hashmax &&
475 void xfrm_spd_getinfo(struct xfrmk_spdinfo *si)
477 read_lock_bh(&xfrm_policy_lock);
478 si->incnt = init_net.xfrm.policy_count[XFRM_POLICY_IN];
479 si->outcnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT];
480 si->fwdcnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD];
481 si->inscnt = init_net.xfrm.policy_count[XFRM_POLICY_IN+XFRM_POLICY_MAX];
482 si->outscnt = init_net.xfrm.policy_count[XFRM_POLICY_OUT+XFRM_POLICY_MAX];
483 si->fwdscnt = init_net.xfrm.policy_count[XFRM_POLICY_FWD+XFRM_POLICY_MAX];
484 si->spdhcnt = init_net.xfrm.policy_idx_hmask;
485 si->spdhmcnt = xfrm_policy_hashmax;
486 read_unlock_bh(&xfrm_policy_lock);
488 EXPORT_SYMBOL(xfrm_spd_getinfo);
490 static DEFINE_MUTEX(hash_resize_mutex);
491 static void xfrm_hash_resize(struct work_struct *work)
493 struct net *net = container_of(work, struct net, xfrm.policy_hash_work);
496 mutex_lock(&hash_resize_mutex);
499 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
500 if (xfrm_bydst_should_resize(net, dir, &total))
501 xfrm_bydst_resize(net, dir);
503 if (xfrm_byidx_should_resize(net, total))
504 xfrm_byidx_resize(net, total);
506 mutex_unlock(&hash_resize_mutex);
509 /* Generate new index... KAME seems to generate them ordered by cost
510 * of an absolute inpredictability of ordering of rules. This will not pass. */
511 static u32 xfrm_gen_index(int dir)
513 static u32 idx_generator;
516 struct hlist_node *entry;
517 struct hlist_head *list;
518 struct xfrm_policy *p;
522 idx = (idx_generator | dir);
526 list = init_net.xfrm.policy_byidx + idx_hash(&init_net, idx);
528 hlist_for_each_entry(p, entry, list, byidx) {
529 if (p->index == idx) {
539 static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
541 u32 *p1 = (u32 *) s1;
542 u32 *p2 = (u32 *) s2;
543 int len = sizeof(struct xfrm_selector) / sizeof(u32);
546 for (i = 0; i < len; i++) {
554 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
556 struct xfrm_policy *pol;
557 struct xfrm_policy *delpol;
558 struct hlist_head *chain;
559 struct hlist_node *entry, *newpos;
560 struct dst_entry *gc_list;
562 write_lock_bh(&xfrm_policy_lock);
563 chain = policy_hash_bysel(&policy->selector, policy->family, dir);
566 hlist_for_each_entry(pol, entry, chain, bydst) {
567 if (pol->type == policy->type &&
568 !selector_cmp(&pol->selector, &policy->selector) &&
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 init_net.xfrm.policy_count[dir]++;
591 atomic_inc(&flow_cache_genid);
593 hlist_del(&delpol->bydst);
594 hlist_del(&delpol->byidx);
595 list_del(&delpol->walk.all);
596 init_net.xfrm.policy_count[dir]--;
598 policy->index = delpol ? delpol->index : xfrm_gen_index(dir);
599 hlist_add_head(&policy->byidx, init_net.xfrm.policy_byidx+idx_hash(&init_net, policy->index));
600 policy->curlft.add_time = get_seconds();
601 policy->curlft.use_time = 0;
602 if (!mod_timer(&policy->timer, jiffies + HZ))
603 xfrm_pol_hold(policy);
604 list_add(&policy->walk.all, &init_net.xfrm.policy_all);
605 write_unlock_bh(&xfrm_policy_lock);
608 xfrm_policy_kill(delpol);
609 else if (xfrm_bydst_should_resize(&init_net, dir, NULL))
610 schedule_work(&init_net.xfrm.policy_hash_work);
612 read_lock_bh(&xfrm_policy_lock);
614 entry = &policy->bydst;
615 hlist_for_each_entry_continue(policy, entry, bydst) {
616 struct dst_entry *dst;
618 write_lock(&policy->lock);
619 dst = policy->bundles;
621 struct dst_entry *tail = dst;
624 tail->next = gc_list;
627 policy->bundles = NULL;
629 write_unlock(&policy->lock);
631 read_unlock_bh(&xfrm_policy_lock);
634 struct dst_entry *dst = gc_list;
642 EXPORT_SYMBOL(xfrm_policy_insert);
644 struct xfrm_policy *xfrm_policy_bysel_ctx(u8 type, int dir,
645 struct xfrm_selector *sel,
646 struct xfrm_sec_ctx *ctx, int delete,
649 struct xfrm_policy *pol, *ret;
650 struct hlist_head *chain;
651 struct hlist_node *entry;
654 write_lock_bh(&xfrm_policy_lock);
655 chain = policy_hash_bysel(sel, sel->family, dir);
657 hlist_for_each_entry(pol, entry, chain, bydst) {
658 if (pol->type == type &&
659 !selector_cmp(sel, &pol->selector) &&
660 xfrm_sec_ctx_match(ctx, pol->security)) {
663 *err = security_xfrm_policy_delete(
666 write_unlock_bh(&xfrm_policy_lock);
669 hlist_del(&pol->bydst);
670 hlist_del(&pol->byidx);
671 list_del(&pol->walk.all);
672 init_net.xfrm.policy_count[dir]--;
678 write_unlock_bh(&xfrm_policy_lock);
681 atomic_inc(&flow_cache_genid);
682 xfrm_policy_kill(ret);
686 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
688 struct xfrm_policy *xfrm_policy_byid(u8 type, int dir, u32 id, int delete,
691 struct xfrm_policy *pol, *ret;
692 struct hlist_head *chain;
693 struct hlist_node *entry;
696 if (xfrm_policy_id2dir(id) != dir)
700 write_lock_bh(&xfrm_policy_lock);
701 chain = init_net.xfrm.policy_byidx + idx_hash(&init_net, id);
703 hlist_for_each_entry(pol, entry, chain, byidx) {
704 if (pol->type == type && pol->index == id) {
707 *err = security_xfrm_policy_delete(
710 write_unlock_bh(&xfrm_policy_lock);
713 hlist_del(&pol->bydst);
714 hlist_del(&pol->byidx);
715 list_del(&pol->walk.all);
716 init_net.xfrm.policy_count[dir]--;
722 write_unlock_bh(&xfrm_policy_lock);
725 atomic_inc(&flow_cache_genid);
726 xfrm_policy_kill(ret);
730 EXPORT_SYMBOL(xfrm_policy_byid);
732 #ifdef CONFIG_SECURITY_NETWORK_XFRM
734 xfrm_policy_flush_secctx_check(u8 type, struct xfrm_audit *audit_info)
738 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
739 struct xfrm_policy *pol;
740 struct hlist_node *entry;
743 hlist_for_each_entry(pol, entry,
744 &init_net.xfrm.policy_inexact[dir], bydst) {
745 if (pol->type != type)
747 err = security_xfrm_policy_delete(pol->security);
749 xfrm_audit_policy_delete(pol, 0,
750 audit_info->loginuid,
751 audit_info->sessionid,
756 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
757 hlist_for_each_entry(pol, entry,
758 init_net.xfrm.policy_bydst[dir].table + i,
760 if (pol->type != type)
762 err = security_xfrm_policy_delete(
765 xfrm_audit_policy_delete(pol, 0,
766 audit_info->loginuid,
767 audit_info->sessionid,
778 xfrm_policy_flush_secctx_check(u8 type, struct xfrm_audit *audit_info)
784 int xfrm_policy_flush(u8 type, struct xfrm_audit *audit_info)
788 write_lock_bh(&xfrm_policy_lock);
790 err = xfrm_policy_flush_secctx_check(type, audit_info);
794 for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
795 struct xfrm_policy *pol;
796 struct hlist_node *entry;
801 hlist_for_each_entry(pol, entry,
802 &init_net.xfrm.policy_inexact[dir], bydst) {
803 if (pol->type != type)
805 hlist_del(&pol->bydst);
806 hlist_del(&pol->byidx);
807 write_unlock_bh(&xfrm_policy_lock);
809 xfrm_audit_policy_delete(pol, 1, audit_info->loginuid,
810 audit_info->sessionid,
813 xfrm_policy_kill(pol);
816 write_lock_bh(&xfrm_policy_lock);
820 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
822 hlist_for_each_entry(pol, entry,
823 init_net.xfrm.policy_bydst[dir].table + i,
825 if (pol->type != type)
827 hlist_del(&pol->bydst);
828 hlist_del(&pol->byidx);
829 list_del(&pol->walk.all);
830 write_unlock_bh(&xfrm_policy_lock);
832 xfrm_audit_policy_delete(pol, 1,
833 audit_info->loginuid,
834 audit_info->sessionid,
836 xfrm_policy_kill(pol);
839 write_lock_bh(&xfrm_policy_lock);
844 init_net.xfrm.policy_count[dir] -= killed;
846 atomic_inc(&flow_cache_genid);
848 write_unlock_bh(&xfrm_policy_lock);
851 EXPORT_SYMBOL(xfrm_policy_flush);
853 int xfrm_policy_walk(struct xfrm_policy_walk *walk,
854 int (*func)(struct xfrm_policy *, int, int, void*),
857 struct xfrm_policy *pol;
858 struct xfrm_policy_walk_entry *x;
861 if (walk->type >= XFRM_POLICY_TYPE_MAX &&
862 walk->type != XFRM_POLICY_TYPE_ANY)
865 if (list_empty(&walk->walk.all) && walk->seq != 0)
868 write_lock_bh(&xfrm_policy_lock);
869 if (list_empty(&walk->walk.all))
870 x = list_first_entry(&init_net.xfrm.policy_all, struct xfrm_policy_walk_entry, all);
872 x = list_entry(&walk->walk.all, struct xfrm_policy_walk_entry, all);
873 list_for_each_entry_from(x, &init_net.xfrm.policy_all, all) {
876 pol = container_of(x, struct xfrm_policy, walk);
877 if (walk->type != XFRM_POLICY_TYPE_ANY &&
878 walk->type != pol->type)
880 error = func(pol, xfrm_policy_id2dir(pol->index),
883 list_move_tail(&walk->walk.all, &x->all);
888 if (walk->seq == 0) {
892 list_del_init(&walk->walk.all);
894 write_unlock_bh(&xfrm_policy_lock);
897 EXPORT_SYMBOL(xfrm_policy_walk);
899 void xfrm_policy_walk_init(struct xfrm_policy_walk *walk, u8 type)
901 INIT_LIST_HEAD(&walk->walk.all);
906 EXPORT_SYMBOL(xfrm_policy_walk_init);
908 void xfrm_policy_walk_done(struct xfrm_policy_walk *walk)
910 if (list_empty(&walk->walk.all))
913 write_lock_bh(&xfrm_policy_lock);
914 list_del(&walk->walk.all);
915 write_unlock_bh(&xfrm_policy_lock);
917 EXPORT_SYMBOL(xfrm_policy_walk_done);
920 * Find policy to apply to this flow.
922 * Returns 0 if policy found, else an -errno.
924 static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
925 u8 type, u16 family, int dir)
927 struct xfrm_selector *sel = &pol->selector;
928 int match, ret = -ESRCH;
930 if (pol->family != family ||
934 match = xfrm_selector_match(sel, fl, family);
936 ret = security_xfrm_policy_lookup(pol->security, fl->secid,
942 static struct xfrm_policy *xfrm_policy_lookup_bytype(u8 type, struct flowi *fl,
946 struct xfrm_policy *pol, *ret;
947 xfrm_address_t *daddr, *saddr;
948 struct hlist_node *entry;
949 struct hlist_head *chain;
952 daddr = xfrm_flowi_daddr(fl, family);
953 saddr = xfrm_flowi_saddr(fl, family);
954 if (unlikely(!daddr || !saddr))
957 read_lock_bh(&xfrm_policy_lock);
958 chain = policy_hash_direct(daddr, saddr, family, dir);
960 hlist_for_each_entry(pol, entry, chain, bydst) {
961 err = xfrm_policy_match(pol, fl, type, family, dir);
971 priority = ret->priority;
975 chain = &init_net.xfrm.policy_inexact[dir];
976 hlist_for_each_entry(pol, entry, chain, bydst) {
977 err = xfrm_policy_match(pol, fl, type, family, dir);
985 } else if (pol->priority < priority) {
993 read_unlock_bh(&xfrm_policy_lock);
998 static int xfrm_policy_lookup(struct flowi *fl, u16 family, u8 dir,
999 void **objp, atomic_t **obj_refp)
1001 struct xfrm_policy *pol;
1004 #ifdef CONFIG_XFRM_SUB_POLICY
1005 pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB, fl, family, dir);
1013 pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, fl, family, dir);
1018 #ifdef CONFIG_XFRM_SUB_POLICY
1021 if ((*objp = (void *) pol) != NULL)
1022 *obj_refp = &pol->refcnt;
1026 static inline int policy_to_flow_dir(int dir)
1028 if (XFRM_POLICY_IN == FLOW_DIR_IN &&
1029 XFRM_POLICY_OUT == FLOW_DIR_OUT &&
1030 XFRM_POLICY_FWD == FLOW_DIR_FWD)
1034 case XFRM_POLICY_IN:
1036 case XFRM_POLICY_OUT:
1037 return FLOW_DIR_OUT;
1038 case XFRM_POLICY_FWD:
1039 return FLOW_DIR_FWD;
1043 static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
1045 struct xfrm_policy *pol;
1047 read_lock_bh(&xfrm_policy_lock);
1048 if ((pol = sk->sk_policy[dir]) != NULL) {
1049 int match = xfrm_selector_match(&pol->selector, fl,
1054 err = security_xfrm_policy_lookup(pol->security,
1056 policy_to_flow_dir(dir));
1059 else if (err == -ESRCH)
1066 read_unlock_bh(&xfrm_policy_lock);
1070 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
1072 struct net *net = xp_net(pol);
1073 struct hlist_head *chain = policy_hash_bysel(&pol->selector,
1076 list_add(&pol->walk.all, &net->xfrm.policy_all);
1077 hlist_add_head(&pol->bydst, chain);
1078 hlist_add_head(&pol->byidx, net->xfrm.policy_byidx+idx_hash(net, pol->index));
1079 net->xfrm.policy_count[dir]++;
1082 if (xfrm_bydst_should_resize(net, dir, NULL))
1083 schedule_work(&net->xfrm.policy_hash_work);
1086 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
1089 struct net *net = xp_net(pol);
1091 if (hlist_unhashed(&pol->bydst))
1094 hlist_del(&pol->bydst);
1095 hlist_del(&pol->byidx);
1096 list_del(&pol->walk.all);
1097 net->xfrm.policy_count[dir]--;
1102 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
1104 write_lock_bh(&xfrm_policy_lock);
1105 pol = __xfrm_policy_unlink(pol, dir);
1106 write_unlock_bh(&xfrm_policy_lock);
1108 if (dir < XFRM_POLICY_MAX)
1109 atomic_inc(&flow_cache_genid);
1110 xfrm_policy_kill(pol);
1115 EXPORT_SYMBOL(xfrm_policy_delete);
1117 int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *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(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(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(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)
1208 xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
1209 xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
1212 for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
1213 struct xfrm_state *x;
1214 xfrm_address_t *remote = daddr;
1215 xfrm_address_t *local = saddr;
1216 struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
1218 if (tmpl->mode == XFRM_MODE_TUNNEL ||
1219 tmpl->mode == XFRM_MODE_BEET) {
1220 remote = &tmpl->id.daddr;
1221 local = &tmpl->saddr;
1222 family = tmpl->encap_family;
1223 if (xfrm_addr_any(local, family)) {
1224 error = xfrm_get_saddr(&tmp, remote, family);
1231 x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
1233 if (x && x->km.state == XFRM_STATE_VALID) {
1240 error = (x->km.state == XFRM_STATE_ERROR ?
1244 else if (error == -ESRCH)
1247 if (!tmpl->optional)
1253 for (nx--; nx>=0; nx--)
1254 xfrm_state_put(xfrm[nx]);
1259 xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
1260 struct xfrm_state **xfrm,
1261 unsigned short family)
1263 struct xfrm_state *tp[XFRM_MAX_DEPTH];
1264 struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
1270 for (i = 0; i < npols; i++) {
1271 if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
1276 ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
1284 /* found states are sorted for outbound processing */
1286 xfrm_state_sort(xfrm, tpp, cnx, family);
1291 for (cnx--; cnx>=0; cnx--)
1292 xfrm_state_put(tpp[cnx]);
1297 /* Check that the bundle accepts the flow and its components are
1301 static struct dst_entry *
1302 xfrm_find_bundle(struct flowi *fl, struct xfrm_policy *policy, unsigned short family)
1304 struct dst_entry *x;
1305 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1306 if (unlikely(afinfo == NULL))
1307 return ERR_PTR(-EINVAL);
1308 x = afinfo->find_bundle(fl, policy);
1309 xfrm_policy_put_afinfo(afinfo);
1313 static inline int xfrm_get_tos(struct flowi *fl, int family)
1315 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1321 tos = afinfo->get_tos(fl);
1323 xfrm_policy_put_afinfo(afinfo);
1328 static inline struct xfrm_dst *xfrm_alloc_dst(int family)
1330 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1331 struct xfrm_dst *xdst;
1334 return ERR_PTR(-EINVAL);
1336 xdst = dst_alloc(afinfo->dst_ops) ?: ERR_PTR(-ENOBUFS);
1338 xfrm_policy_put_afinfo(afinfo);
1343 static inline int xfrm_init_path(struct xfrm_dst *path, struct dst_entry *dst,
1346 struct xfrm_policy_afinfo *afinfo =
1347 xfrm_policy_get_afinfo(dst->ops->family);
1353 err = afinfo->init_path(path, dst, nfheader_len);
1355 xfrm_policy_put_afinfo(afinfo);
1360 static inline int xfrm_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
1362 struct xfrm_policy_afinfo *afinfo =
1363 xfrm_policy_get_afinfo(xdst->u.dst.ops->family);
1369 err = afinfo->fill_dst(xdst, dev);
1371 xfrm_policy_put_afinfo(afinfo);
1376 /* Allocate chain of dst_entry's, attach known xfrm's, calculate
1377 * all the metrics... Shortly, bundle a bundle.
1380 static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,
1381 struct xfrm_state **xfrm, int nx,
1383 struct dst_entry *dst)
1385 unsigned long now = jiffies;
1386 struct net_device *dev;
1387 struct dst_entry *dst_prev = NULL;
1388 struct dst_entry *dst0 = NULL;
1392 int nfheader_len = 0;
1393 int trailer_len = 0;
1395 int family = policy->selector.family;
1396 xfrm_address_t saddr, daddr;
1398 xfrm_flowi_addr_get(fl, &saddr, &daddr, family);
1400 tos = xfrm_get_tos(fl, family);
1407 for (; i < nx; i++) {
1408 struct xfrm_dst *xdst = xfrm_alloc_dst(family);
1409 struct dst_entry *dst1 = &xdst->u.dst;
1411 err = PTR_ERR(xdst);
1420 dst_prev->child = dst_clone(dst1);
1421 dst1->flags |= DST_NOHASH;
1425 memcpy(&dst1->metrics, &dst->metrics, sizeof(dst->metrics));
1427 if (xfrm[i]->props.mode != XFRM_MODE_TRANSPORT) {
1428 family = xfrm[i]->props.family;
1429 dst = xfrm_dst_lookup(xfrm[i], tos, &saddr, &daddr,
1437 dst1->xfrm = xfrm[i];
1438 xdst->genid = xfrm[i]->genid;
1440 dst1->obsolete = -1;
1441 dst1->flags |= DST_HOST;
1442 dst1->lastuse = now;
1444 dst1->input = dst_discard;
1445 dst1->output = xfrm[i]->outer_mode->afinfo->output;
1447 dst1->next = dst_prev;
1450 header_len += xfrm[i]->props.header_len;
1451 if (xfrm[i]->type->flags & XFRM_TYPE_NON_FRAGMENT)
1452 nfheader_len += xfrm[i]->props.header_len;
1453 trailer_len += xfrm[i]->props.trailer_len;
1456 dst_prev->child = dst;
1464 /* Copy neighbout for reachability confirmation */
1465 dst0->neighbour = neigh_clone(dst->neighbour);
1467 xfrm_init_path((struct xfrm_dst *)dst0, dst, nfheader_len);
1468 xfrm_init_pmtu(dst_prev);
1470 for (dst_prev = dst0; dst_prev != dst; dst_prev = dst_prev->child) {
1471 struct xfrm_dst *xdst = (struct xfrm_dst *)dst_prev;
1473 err = xfrm_fill_dst(xdst, dev);
1477 dst_prev->header_len = header_len;
1478 dst_prev->trailer_len = trailer_len;
1479 header_len -= xdst->u.dst.xfrm->props.header_len;
1480 trailer_len -= xdst->u.dst.xfrm->props.trailer_len;
1488 xfrm_state_put(xfrm[i]);
1492 dst0 = ERR_PTR(err);
1497 xfrm_dst_alloc_copy(void **target, void *src, int size)
1500 *target = kmalloc(size, GFP_ATOMIC);
1504 memcpy(*target, src, size);
1509 xfrm_dst_update_parent(struct dst_entry *dst, struct xfrm_selector *sel)
1511 #ifdef CONFIG_XFRM_SUB_POLICY
1512 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1513 return xfrm_dst_alloc_copy((void **)&(xdst->partner),
1521 xfrm_dst_update_origin(struct dst_entry *dst, struct flowi *fl)
1523 #ifdef CONFIG_XFRM_SUB_POLICY
1524 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
1525 return xfrm_dst_alloc_copy((void **)&(xdst->origin), fl, sizeof(*fl));
1531 static int stale_bundle(struct dst_entry *dst);
1533 /* Main function: finds/creates a bundle for given flow.
1535 * At the moment we eat a raw IP route. Mostly to speed up lookups
1536 * on interfaces with disabled IPsec.
1538 int __xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
1539 struct sock *sk, int flags)
1541 struct xfrm_policy *policy;
1542 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1547 struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
1548 struct dst_entry *dst, *dst_orig = *dst_p;
1553 u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
1556 genid = atomic_read(&flow_cache_genid);
1558 for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
1564 if (sk && sk->sk_policy[XFRM_POLICY_OUT]) {
1565 policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
1566 err = PTR_ERR(policy);
1567 if (IS_ERR(policy)) {
1568 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
1574 /* To accelerate a bit... */
1575 if ((dst_orig->flags & DST_NOXFRM) ||
1576 !init_net.xfrm.policy_count[XFRM_POLICY_OUT])
1579 policy = flow_cache_lookup(fl, dst_orig->ops->family,
1580 dir, xfrm_policy_lookup);
1581 err = PTR_ERR(policy);
1582 if (IS_ERR(policy)) {
1583 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
1591 family = dst_orig->ops->family;
1594 xfrm_nr += pols[0]->xfrm_nr;
1597 if ((flags & XFRM_LOOKUP_ICMP) && !(policy->flags & XFRM_POLICY_ICMP))
1600 policy->curlft.use_time = get_seconds();
1602 switch (policy->action) {
1604 case XFRM_POLICY_BLOCK:
1605 /* Prohibit the flow */
1606 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
1610 case XFRM_POLICY_ALLOW:
1611 #ifndef CONFIG_XFRM_SUB_POLICY
1612 if (policy->xfrm_nr == 0) {
1613 /* Flow passes not transformed. */
1614 xfrm_pol_put(policy);
1619 /* Try to find matching bundle.
1621 * LATER: help from flow cache. It is optional, this
1622 * is required only for output policy.
1624 dst = xfrm_find_bundle(fl, policy, family);
1626 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1634 #ifdef CONFIG_XFRM_SUB_POLICY
1635 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1636 pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
1640 if (IS_ERR(pols[1])) {
1641 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLERROR);
1642 err = PTR_ERR(pols[1]);
1645 if (pols[1]->action == XFRM_POLICY_BLOCK) {
1646 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLBLOCK);
1651 xfrm_nr += pols[1]->xfrm_nr;
1656 * Because neither flowi nor bundle information knows about
1657 * transformation template size. On more than one policy usage
1658 * we can realize whether all of them is bypass or not after
1659 * they are searched. See above not-transformed bypass
1660 * is surrounded by non-sub policy configuration, too.
1663 /* Flow passes not transformed. */
1664 xfrm_pols_put(pols, npols);
1669 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1671 if (unlikely(nx<0)) {
1673 if (err == -EAGAIN && sysctl_xfrm_larval_drop) {
1674 /* EREMOTE tells the caller to generate
1675 * a one-shot blackhole route.
1677 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1678 xfrm_pol_put(policy);
1681 if (err == -EAGAIN && (flags & XFRM_LOOKUP_WAIT)) {
1682 DECLARE_WAITQUEUE(wait, current);
1684 add_wait_queue(&init_net.xfrm.km_waitq, &wait);
1685 set_current_state(TASK_INTERRUPTIBLE);
1687 set_current_state(TASK_RUNNING);
1688 remove_wait_queue(&init_net.xfrm.km_waitq, &wait);
1690 nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
1692 if (nx == -EAGAIN && signal_pending(current)) {
1693 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1697 if (nx == -EAGAIN ||
1698 genid != atomic_read(&flow_cache_genid)) {
1699 xfrm_pols_put(pols, npols);
1705 XFRM_INC_STATS(LINUX_MIB_XFRMOUTNOSTATES);
1710 /* Flow passes not transformed. */
1711 xfrm_pols_put(pols, npols);
1715 dst = xfrm_bundle_create(policy, xfrm, nx, fl, dst_orig);
1718 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLEGENERROR);
1722 for (pi = 0; pi < npols; pi++) {
1723 read_lock_bh(&pols[pi]->lock);
1724 pol_dead |= pols[pi]->walk.dead;
1725 read_unlock_bh(&pols[pi]->lock);
1728 write_lock_bh(&policy->lock);
1729 if (unlikely(pol_dead || stale_bundle(dst))) {
1730 /* Wow! While we worked on resolving, this
1731 * policy has gone. Retry. It is not paranoia,
1732 * we just cannot enlist new bundle to dead object.
1733 * We can't enlist stable bundles either.
1735 write_unlock_bh(&policy->lock);
1739 XFRM_INC_STATS(LINUX_MIB_XFRMOUTPOLDEAD);
1741 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1742 err = -EHOSTUNREACH;
1747 err = xfrm_dst_update_parent(dst, &pols[1]->selector);
1749 err = xfrm_dst_update_origin(dst, fl);
1750 if (unlikely(err)) {
1751 write_unlock_bh(&policy->lock);
1753 XFRM_INC_STATS(LINUX_MIB_XFRMOUTBUNDLECHECKERROR);
1757 dst->next = policy->bundles;
1758 policy->bundles = dst;
1760 write_unlock_bh(&policy->lock);
1763 dst_release(dst_orig);
1764 xfrm_pols_put(pols, npols);
1768 xfrm_pols_put(pols, npols);
1770 dst_release(dst_orig);
1776 if (flags & XFRM_LOOKUP_ICMP)
1780 EXPORT_SYMBOL(__xfrm_lookup);
1782 int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
1783 struct sock *sk, int flags)
1785 int err = __xfrm_lookup(dst_p, fl, sk, flags);
1787 if (err == -EREMOTE) {
1788 dst_release(*dst_p);
1795 EXPORT_SYMBOL(xfrm_lookup);
1798 xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
1800 struct xfrm_state *x;
1802 if (!skb->sp || idx < 0 || idx >= skb->sp->len)
1804 x = skb->sp->xvec[idx];
1805 if (!x->type->reject)
1807 return x->type->reject(x, skb, fl);
1810 /* When skb is transformed back to its "native" form, we have to
1811 * check policy restrictions. At the moment we make this in maximally
1812 * stupid way. Shame on me. :-) Of course, connected sockets must
1813 * have policy cached at them.
1817 xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
1818 unsigned short family)
1820 if (xfrm_state_kern(x))
1821 return tmpl->optional && !xfrm_state_addr_cmp(tmpl, x, tmpl->encap_family);
1822 return x->id.proto == tmpl->id.proto &&
1823 (x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
1824 (x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
1825 x->props.mode == tmpl->mode &&
1826 (tmpl->allalgs || (tmpl->aalgos & (1<<x->props.aalgo)) ||
1827 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
1828 !(x->props.mode != XFRM_MODE_TRANSPORT &&
1829 xfrm_state_addr_cmp(tmpl, x, family));
1833 * 0 or more than 0 is returned when validation is succeeded (either bypass
1834 * because of optional transport mode, or next index of the mathced secpath
1835 * state with the template.
1836 * -1 is returned when no matching template is found.
1837 * Otherwise "-2 - errored_index" is returned.
1840 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
1841 unsigned short family)
1845 if (tmpl->optional) {
1846 if (tmpl->mode == XFRM_MODE_TRANSPORT)
1850 for (; idx < sp->len; idx++) {
1851 if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
1853 if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
1862 int __xfrm_decode_session(struct sk_buff *skb, struct flowi *fl,
1863 unsigned int family, int reverse)
1865 struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
1868 if (unlikely(afinfo == NULL))
1869 return -EAFNOSUPPORT;
1871 afinfo->decode_session(skb, fl, reverse);
1872 err = security_xfrm_decode_session(skb, &fl->secid);
1873 xfrm_policy_put_afinfo(afinfo);
1876 EXPORT_SYMBOL(__xfrm_decode_session);
1878 static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
1880 for (; k < sp->len; k++) {
1881 if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
1890 int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
1891 unsigned short family)
1893 struct xfrm_policy *pol;
1894 struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
1903 reverse = dir & ~XFRM_POLICY_MASK;
1904 dir &= XFRM_POLICY_MASK;
1905 fl_dir = policy_to_flow_dir(dir);
1907 if (__xfrm_decode_session(skb, &fl, family, reverse) < 0) {
1908 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
1912 nf_nat_decode_session(skb, &fl, family);
1914 /* First, check used SA against their selectors. */
1918 for (i=skb->sp->len-1; i>=0; i--) {
1919 struct xfrm_state *x = skb->sp->xvec[i];
1920 if (!xfrm_selector_match(&x->sel, &fl, family)) {
1921 XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMISMATCH);
1928 if (sk && sk->sk_policy[dir]) {
1929 pol = xfrm_sk_policy_lookup(sk, dir, &fl);
1931 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
1937 pol = flow_cache_lookup(&fl, family, fl_dir,
1938 xfrm_policy_lookup);
1941 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
1946 if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
1947 xfrm_secpath_reject(xerr_idx, skb, &fl);
1948 XFRM_INC_STATS(LINUX_MIB_XFRMINNOPOLS);
1954 pol->curlft.use_time = get_seconds();
1958 #ifdef CONFIG_XFRM_SUB_POLICY
1959 if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
1960 pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
1964 if (IS_ERR(pols[1])) {
1965 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLERROR);
1968 pols[1]->curlft.use_time = get_seconds();
1974 if (pol->action == XFRM_POLICY_ALLOW) {
1975 struct sec_path *sp;
1976 static struct sec_path dummy;
1977 struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
1978 struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
1979 struct xfrm_tmpl **tpp = tp;
1983 if ((sp = skb->sp) == NULL)
1986 for (pi = 0; pi < npols; pi++) {
1987 if (pols[pi] != pol &&
1988 pols[pi]->action != XFRM_POLICY_ALLOW) {
1989 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
1992 if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH) {
1993 XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR);
1996 for (i = 0; i < pols[pi]->xfrm_nr; i++)
1997 tpp[ti++] = &pols[pi]->xfrm_vec[i];
2001 xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
2005 /* For each tunnel xfrm, find the first matching tmpl.
2006 * For each tmpl before that, find corresponding xfrm.
2007 * Order is _important_. Later we will implement
2008 * some barriers, but at the moment barriers
2009 * are implied between each two transformations.
2011 for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
2012 k = xfrm_policy_ok(tpp[i], sp, k, family);
2015 /* "-2 - errored_index" returned */
2017 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
2022 if (secpath_has_nontransport(sp, k, &xerr_idx)) {
2023 XFRM_INC_STATS(LINUX_MIB_XFRMINTMPLMISMATCH);
2027 xfrm_pols_put(pols, npols);
2030 XFRM_INC_STATS(LINUX_MIB_XFRMINPOLBLOCK);
2033 xfrm_secpath_reject(xerr_idx, skb, &fl);
2035 xfrm_pols_put(pols, npols);
2038 EXPORT_SYMBOL(__xfrm_policy_check);
2040 int __xfrm_route_forward(struct sk_buff *skb, unsigned short family)
2044 if (xfrm_decode_session(skb, &fl, family) < 0) {
2045 /* XXX: we should have something like FWDHDRERROR here. */
2046 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
2050 return xfrm_lookup(&skb->dst, &fl, NULL, 0) == 0;
2052 EXPORT_SYMBOL(__xfrm_route_forward);
2054 /* Optimize later using cookies and generation ids. */
2056 static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
2058 /* Code (such as __xfrm4_bundle_create()) sets dst->obsolete
2059 * to "-1" to force all XFRM destinations to get validated by
2060 * dst_ops->check on every use. We do this because when a
2061 * normal route referenced by an XFRM dst is obsoleted we do
2062 * not go looking around for all parent referencing XFRM dsts
2063 * so that we can invalidate them. It is just too much work.
2064 * Instead we make the checks here on every use. For example:
2066 * XFRM dst A --> IPv4 dst X
2068 * X is the "xdst->route" of A (X is also the "dst->path" of A
2069 * in this example). If X is marked obsolete, "A" will not
2070 * notice. That's what we are validating here via the
2071 * stale_bundle() check.
2073 * When a policy's bundle is pruned, we dst_free() the XFRM
2074 * dst which causes it's ->obsolete field to be set to a
2075 * positive non-zero integer. If an XFRM dst has been pruned
2076 * like this, we want to force a new route lookup.
2078 if (dst->obsolete < 0 && !stale_bundle(dst))
2084 static int stale_bundle(struct dst_entry *dst)
2086 return !xfrm_bundle_ok(NULL, (struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
2089 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
2091 while ((dst = dst->child) && dst->xfrm && dst->dev == dev) {
2092 dst->dev = dev_net(dev)->loopback_dev;
2097 EXPORT_SYMBOL(xfrm_dst_ifdown);
2099 static void xfrm_link_failure(struct sk_buff *skb)
2101 /* Impossible. Such dst must be popped before reaches point of failure. */
2105 static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
2108 if (dst->obsolete) {
2116 static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
2118 struct dst_entry *dst, **dstp;
2120 write_lock(&pol->lock);
2121 dstp = &pol->bundles;
2122 while ((dst=*dstp) != NULL) {
2125 dst->next = *gc_list_p;
2131 write_unlock(&pol->lock);
2134 static void xfrm_prune_bundles(int (*func)(struct dst_entry *))
2136 struct dst_entry *gc_list = NULL;
2139 read_lock_bh(&xfrm_policy_lock);
2140 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2141 struct xfrm_policy *pol;
2142 struct hlist_node *entry;
2143 struct hlist_head *table;
2146 hlist_for_each_entry(pol, entry,
2147 &init_net.xfrm.policy_inexact[dir], bydst)
2148 prune_one_bundle(pol, func, &gc_list);
2150 table = init_net.xfrm.policy_bydst[dir].table;
2151 for (i = init_net.xfrm.policy_bydst[dir].hmask; i >= 0; i--) {
2152 hlist_for_each_entry(pol, entry, table + i, bydst)
2153 prune_one_bundle(pol, func, &gc_list);
2156 read_unlock_bh(&xfrm_policy_lock);
2159 struct dst_entry *dst = gc_list;
2160 gc_list = dst->next;
2165 static int unused_bundle(struct dst_entry *dst)
2167 return !atomic_read(&dst->__refcnt);
2170 static void __xfrm_garbage_collect(void)
2172 xfrm_prune_bundles(unused_bundle);
2175 static int xfrm_flush_bundles(void)
2177 xfrm_prune_bundles(stale_bundle);
2181 static void xfrm_init_pmtu(struct dst_entry *dst)
2184 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2185 u32 pmtu, route_mtu_cached;
2187 pmtu = dst_mtu(dst->child);
2188 xdst->child_mtu_cached = pmtu;
2190 pmtu = xfrm_state_mtu(dst->xfrm, pmtu);
2192 route_mtu_cached = dst_mtu(xdst->route);
2193 xdst->route_mtu_cached = route_mtu_cached;
2195 if (pmtu > route_mtu_cached)
2196 pmtu = route_mtu_cached;
2198 dst->metrics[RTAX_MTU-1] = pmtu;
2199 } while ((dst = dst->next));
2202 /* Check that the bundle accepts the flow and its components are
2206 int xfrm_bundle_ok(struct xfrm_policy *pol, struct xfrm_dst *first,
2207 struct flowi *fl, int family, int strict)
2209 struct dst_entry *dst = &first->u.dst;
2210 struct xfrm_dst *last;
2213 if (!dst_check(dst->path, ((struct xfrm_dst *)dst)->path_cookie) ||
2214 (dst->dev && !netif_running(dst->dev)))
2216 #ifdef CONFIG_XFRM_SUB_POLICY
2218 if (first->origin && !flow_cache_uli_match(first->origin, fl))
2220 if (first->partner &&
2221 !xfrm_selector_match(first->partner, fl, family))
2229 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
2231 if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
2234 !security_xfrm_state_pol_flow_match(dst->xfrm, pol, fl))
2236 if (dst->xfrm->km.state != XFRM_STATE_VALID)
2238 if (xdst->genid != dst->xfrm->genid)
2242 !(dst->xfrm->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2243 !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
2246 mtu = dst_mtu(dst->child);
2247 if (xdst->child_mtu_cached != mtu) {
2249 xdst->child_mtu_cached = mtu;
2252 if (!dst_check(xdst->route, xdst->route_cookie))
2254 mtu = dst_mtu(xdst->route);
2255 if (xdst->route_mtu_cached != mtu) {
2257 xdst->route_mtu_cached = mtu;
2261 } while (dst->xfrm);
2266 mtu = last->child_mtu_cached;
2270 mtu = xfrm_state_mtu(dst->xfrm, mtu);
2271 if (mtu > last->route_mtu_cached)
2272 mtu = last->route_mtu_cached;
2273 dst->metrics[RTAX_MTU-1] = mtu;
2278 last = (struct xfrm_dst *)last->u.dst.next;
2279 last->child_mtu_cached = mtu;
2285 EXPORT_SYMBOL(xfrm_bundle_ok);
2287 int xfrm_policy_register_afinfo(struct xfrm_policy_afinfo *afinfo)
2290 if (unlikely(afinfo == NULL))
2292 if (unlikely(afinfo->family >= NPROTO))
2293 return -EAFNOSUPPORT;
2294 write_lock_bh(&xfrm_policy_afinfo_lock);
2295 if (unlikely(xfrm_policy_afinfo[afinfo->family] != NULL))
2298 struct dst_ops *dst_ops = afinfo->dst_ops;
2299 if (likely(dst_ops->kmem_cachep == NULL))
2300 dst_ops->kmem_cachep = xfrm_dst_cache;
2301 if (likely(dst_ops->check == NULL))
2302 dst_ops->check = xfrm_dst_check;
2303 if (likely(dst_ops->negative_advice == NULL))
2304 dst_ops->negative_advice = xfrm_negative_advice;
2305 if (likely(dst_ops->link_failure == NULL))
2306 dst_ops->link_failure = xfrm_link_failure;
2307 if (likely(afinfo->garbage_collect == NULL))
2308 afinfo->garbage_collect = __xfrm_garbage_collect;
2309 xfrm_policy_afinfo[afinfo->family] = afinfo;
2311 write_unlock_bh(&xfrm_policy_afinfo_lock);
2314 EXPORT_SYMBOL(xfrm_policy_register_afinfo);
2316 int xfrm_policy_unregister_afinfo(struct xfrm_policy_afinfo *afinfo)
2319 if (unlikely(afinfo == NULL))
2321 if (unlikely(afinfo->family >= NPROTO))
2322 return -EAFNOSUPPORT;
2323 write_lock_bh(&xfrm_policy_afinfo_lock);
2324 if (likely(xfrm_policy_afinfo[afinfo->family] != NULL)) {
2325 if (unlikely(xfrm_policy_afinfo[afinfo->family] != afinfo))
2328 struct dst_ops *dst_ops = afinfo->dst_ops;
2329 xfrm_policy_afinfo[afinfo->family] = NULL;
2330 dst_ops->kmem_cachep = NULL;
2331 dst_ops->check = NULL;
2332 dst_ops->negative_advice = NULL;
2333 dst_ops->link_failure = NULL;
2334 afinfo->garbage_collect = NULL;
2337 write_unlock_bh(&xfrm_policy_afinfo_lock);
2340 EXPORT_SYMBOL(xfrm_policy_unregister_afinfo);
2342 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family)
2344 struct xfrm_policy_afinfo *afinfo;
2345 if (unlikely(family >= NPROTO))
2347 read_lock(&xfrm_policy_afinfo_lock);
2348 afinfo = xfrm_policy_afinfo[family];
2349 if (unlikely(!afinfo))
2350 read_unlock(&xfrm_policy_afinfo_lock);
2354 static void xfrm_policy_put_afinfo(struct xfrm_policy_afinfo *afinfo)
2356 read_unlock(&xfrm_policy_afinfo_lock);
2359 static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
2361 struct net_device *dev = ptr;
2363 if (!net_eq(dev_net(dev), &init_net))
2368 xfrm_flush_bundles();
2373 static struct notifier_block xfrm_dev_notifier = {
2374 .notifier_call = xfrm_dev_event,
2377 #ifdef CONFIG_XFRM_STATISTICS
2378 static int __init xfrm_statistics_init(void)
2380 if (snmp_mib_init((void **)xfrm_statistics,
2381 sizeof(struct linux_xfrm_mib)) < 0)
2387 static int __net_init xfrm_policy_init(struct net *net)
2389 unsigned int hmask, sz;
2392 if (net_eq(net, &init_net))
2393 xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
2394 sizeof(struct xfrm_dst),
2395 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
2399 sz = (hmask+1) * sizeof(struct hlist_head);
2401 net->xfrm.policy_byidx = xfrm_hash_alloc(sz);
2402 if (!net->xfrm.policy_byidx)
2404 net->xfrm.policy_idx_hmask = hmask;
2406 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2407 struct xfrm_policy_hash *htab;
2409 net->xfrm.policy_count[dir] = 0;
2410 INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
2412 htab = &net->xfrm.policy_bydst[dir];
2413 htab->table = xfrm_hash_alloc(sz);
2416 htab->hmask = hmask;
2419 INIT_LIST_HEAD(&net->xfrm.policy_all);
2420 INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
2421 if (net_eq(net, &init_net))
2422 register_netdevice_notifier(&xfrm_dev_notifier);
2426 for (dir--; dir >= 0; dir--) {
2427 struct xfrm_policy_hash *htab;
2429 htab = &net->xfrm.policy_bydst[dir];
2430 xfrm_hash_free(htab->table, sz);
2432 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2437 static void xfrm_policy_fini(struct net *net)
2442 WARN_ON(!list_empty(&net->xfrm.policy_all));
2444 for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
2445 struct xfrm_policy_hash *htab;
2447 WARN_ON(!hlist_empty(&net->xfrm.policy_inexact[dir]));
2449 htab = &net->xfrm.policy_bydst[dir];
2450 sz = (htab->hmask + 1);
2451 WARN_ON(!hlist_empty(htab->table));
2452 xfrm_hash_free(htab->table, sz);
2455 sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
2456 WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
2457 xfrm_hash_free(net->xfrm.policy_byidx, sz);
2460 static int __net_init xfrm_net_init(struct net *net)
2464 rv = xfrm_state_init(net);
2467 rv = xfrm_policy_init(net);
2473 xfrm_state_fini(net);
2478 static void __net_exit xfrm_net_exit(struct net *net)
2480 xfrm_policy_fini(net);
2481 xfrm_state_fini(net);
2484 static struct pernet_operations __net_initdata xfrm_net_ops = {
2485 .init = xfrm_net_init,
2486 .exit = xfrm_net_exit,
2489 void __init xfrm_init(void)
2491 register_pernet_subsys(&xfrm_net_ops);
2492 #ifdef CONFIG_XFRM_STATISTICS
2493 xfrm_statistics_init();
2496 #ifdef CONFIG_XFRM_STATISTICS
2501 #ifdef CONFIG_AUDITSYSCALL
2502 static void xfrm_audit_common_policyinfo(struct xfrm_policy *xp,
2503 struct audit_buffer *audit_buf)
2505 struct xfrm_sec_ctx *ctx = xp->security;
2506 struct xfrm_selector *sel = &xp->selector;
2509 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2510 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2512 switch(sel->family) {
2514 audit_log_format(audit_buf, " src=%pI4", &sel->saddr.a4);
2515 if (sel->prefixlen_s != 32)
2516 audit_log_format(audit_buf, " src_prefixlen=%d",
2518 audit_log_format(audit_buf, " dst=%pI4", &sel->daddr.a4);
2519 if (sel->prefixlen_d != 32)
2520 audit_log_format(audit_buf, " dst_prefixlen=%d",
2524 audit_log_format(audit_buf, " src=%pI6", sel->saddr.a6);
2525 if (sel->prefixlen_s != 128)
2526 audit_log_format(audit_buf, " src_prefixlen=%d",
2528 audit_log_format(audit_buf, " dst=%pI6", sel->daddr.a6);
2529 if (sel->prefixlen_d != 128)
2530 audit_log_format(audit_buf, " dst_prefixlen=%d",
2536 void xfrm_audit_policy_add(struct xfrm_policy *xp, int result,
2537 uid_t auid, u32 sessionid, u32 secid)
2539 struct audit_buffer *audit_buf;
2541 audit_buf = xfrm_audit_start("SPD-add");
2542 if (audit_buf == NULL)
2544 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2545 audit_log_format(audit_buf, " res=%u", result);
2546 xfrm_audit_common_policyinfo(xp, audit_buf);
2547 audit_log_end(audit_buf);
2549 EXPORT_SYMBOL_GPL(xfrm_audit_policy_add);
2551 void xfrm_audit_policy_delete(struct xfrm_policy *xp, int result,
2552 uid_t auid, u32 sessionid, u32 secid)
2554 struct audit_buffer *audit_buf;
2556 audit_buf = xfrm_audit_start("SPD-delete");
2557 if (audit_buf == NULL)
2559 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2560 audit_log_format(audit_buf, " res=%u", result);
2561 xfrm_audit_common_policyinfo(xp, audit_buf);
2562 audit_log_end(audit_buf);
2564 EXPORT_SYMBOL_GPL(xfrm_audit_policy_delete);
2567 #ifdef CONFIG_XFRM_MIGRATE
2568 static int xfrm_migrate_selector_match(struct xfrm_selector *sel_cmp,
2569 struct xfrm_selector *sel_tgt)
2571 if (sel_cmp->proto == IPSEC_ULPROTO_ANY) {
2572 if (sel_tgt->family == sel_cmp->family &&
2573 xfrm_addr_cmp(&sel_tgt->daddr, &sel_cmp->daddr,
2574 sel_cmp->family) == 0 &&
2575 xfrm_addr_cmp(&sel_tgt->saddr, &sel_cmp->saddr,
2576 sel_cmp->family) == 0 &&
2577 sel_tgt->prefixlen_d == sel_cmp->prefixlen_d &&
2578 sel_tgt->prefixlen_s == sel_cmp->prefixlen_s) {
2582 if (memcmp(sel_tgt, sel_cmp, sizeof(*sel_tgt)) == 0) {
2589 static struct xfrm_policy * xfrm_migrate_policy_find(struct xfrm_selector *sel,
2592 struct xfrm_policy *pol, *ret = NULL;
2593 struct hlist_node *entry;
2594 struct hlist_head *chain;
2597 read_lock_bh(&xfrm_policy_lock);
2598 chain = policy_hash_direct(&sel->daddr, &sel->saddr, sel->family, dir);
2599 hlist_for_each_entry(pol, entry, chain, bydst) {
2600 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2601 pol->type == type) {
2603 priority = ret->priority;
2607 chain = &init_net.xfrm.policy_inexact[dir];
2608 hlist_for_each_entry(pol, entry, chain, bydst) {
2609 if (xfrm_migrate_selector_match(sel, &pol->selector) &&
2610 pol->type == type &&
2611 pol->priority < priority) {
2620 read_unlock_bh(&xfrm_policy_lock);
2625 static int migrate_tmpl_match(struct xfrm_migrate *m, struct xfrm_tmpl *t)
2629 if (t->mode == m->mode && t->id.proto == m->proto &&
2630 (m->reqid == 0 || t->reqid == m->reqid)) {
2632 case XFRM_MODE_TUNNEL:
2633 case XFRM_MODE_BEET:
2634 if (xfrm_addr_cmp(&t->id.daddr, &m->old_daddr,
2635 m->old_family) == 0 &&
2636 xfrm_addr_cmp(&t->saddr, &m->old_saddr,
2637 m->old_family) == 0) {
2641 case XFRM_MODE_TRANSPORT:
2642 /* in case of transport mode, template does not store
2643 any IP addresses, hence we just compare mode and
2654 /* update endpoint address(es) of template(s) */
2655 static int xfrm_policy_migrate(struct xfrm_policy *pol,
2656 struct xfrm_migrate *m, int num_migrate)
2658 struct xfrm_migrate *mp;
2659 struct dst_entry *dst;
2662 write_lock_bh(&pol->lock);
2663 if (unlikely(pol->walk.dead)) {
2664 /* target policy has been deleted */
2665 write_unlock_bh(&pol->lock);
2669 for (i = 0; i < pol->xfrm_nr; i++) {
2670 for (j = 0, mp = m; j < num_migrate; j++, mp++) {
2671 if (!migrate_tmpl_match(mp, &pol->xfrm_vec[i]))
2674 if (pol->xfrm_vec[i].mode != XFRM_MODE_TUNNEL &&
2675 pol->xfrm_vec[i].mode != XFRM_MODE_BEET)
2677 /* update endpoints */
2678 memcpy(&pol->xfrm_vec[i].id.daddr, &mp->new_daddr,
2679 sizeof(pol->xfrm_vec[i].id.daddr));
2680 memcpy(&pol->xfrm_vec[i].saddr, &mp->new_saddr,
2681 sizeof(pol->xfrm_vec[i].saddr));
2682 pol->xfrm_vec[i].encap_family = mp->new_family;
2684 while ((dst = pol->bundles) != NULL) {
2685 pol->bundles = dst->next;
2691 write_unlock_bh(&pol->lock);
2699 static int xfrm_migrate_check(struct xfrm_migrate *m, int num_migrate)
2703 if (num_migrate < 1 || num_migrate > XFRM_MAX_DEPTH)
2706 for (i = 0; i < num_migrate; i++) {
2707 if ((xfrm_addr_cmp(&m[i].old_daddr, &m[i].new_daddr,
2708 m[i].old_family) == 0) &&
2709 (xfrm_addr_cmp(&m[i].old_saddr, &m[i].new_saddr,
2710 m[i].old_family) == 0))
2712 if (xfrm_addr_any(&m[i].new_daddr, m[i].new_family) ||
2713 xfrm_addr_any(&m[i].new_saddr, m[i].new_family))
2716 /* check if there is any duplicated entry */
2717 for (j = i + 1; j < num_migrate; j++) {
2718 if (!memcmp(&m[i].old_daddr, &m[j].old_daddr,
2719 sizeof(m[i].old_daddr)) &&
2720 !memcmp(&m[i].old_saddr, &m[j].old_saddr,
2721 sizeof(m[i].old_saddr)) &&
2722 m[i].proto == m[j].proto &&
2723 m[i].mode == m[j].mode &&
2724 m[i].reqid == m[j].reqid &&
2725 m[i].old_family == m[j].old_family)
2733 int xfrm_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
2734 struct xfrm_migrate *m, int num_migrate,
2735 struct xfrm_kmaddress *k)
2737 int i, err, nx_cur = 0, nx_new = 0;
2738 struct xfrm_policy *pol = NULL;
2739 struct xfrm_state *x, *xc;
2740 struct xfrm_state *x_cur[XFRM_MAX_DEPTH];
2741 struct xfrm_state *x_new[XFRM_MAX_DEPTH];
2742 struct xfrm_migrate *mp;
2744 if ((err = xfrm_migrate_check(m, num_migrate)) < 0)
2747 /* Stage 1 - find policy */
2748 if ((pol = xfrm_migrate_policy_find(sel, dir, type)) == NULL) {
2753 /* Stage 2 - find and update state(s) */
2754 for (i = 0, mp = m; i < num_migrate; i++, mp++) {
2755 if ((x = xfrm_migrate_state_find(mp))) {
2758 if ((xc = xfrm_state_migrate(x, mp))) {
2768 /* Stage 3 - update policy */
2769 if ((err = xfrm_policy_migrate(pol, m, num_migrate)) < 0)
2772 /* Stage 4 - delete old state(s) */
2774 xfrm_states_put(x_cur, nx_cur);
2775 xfrm_states_delete(x_cur, nx_cur);
2778 /* Stage 5 - announce */
2779 km_migrate(sel, dir, type, m, num_migrate, k);
2791 xfrm_states_put(x_cur, nx_cur);
2793 xfrm_states_delete(x_new, nx_new);
2797 EXPORT_SYMBOL(xfrm_migrate);