1 // SPDX-License-Identifier: GPL-2.0-only
7 * Kazunori MIYAZAWA @USAGI
8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
10 * YOSHIFUJI Hideaki @USAGI
11 * Split up af-specific functions
12 * Derek Atkins <derek@ihtfp.com>
13 * Add UDP Encapsulation
17 #include <linux/workqueue.h>
19 #include <linux/pfkeyv2.h>
20 #include <linux/ipsec.h>
21 #include <linux/module.h>
22 #include <linux/cache.h>
23 #include <linux/audit.h>
24 #include <linux/uaccess.h>
25 #include <linux/ktime.h>
26 #include <linux/slab.h>
27 #include <linux/interrupt.h>
28 #include <linux/kernel.h>
30 #include <crypto/aead.h>
32 #include "xfrm_hash.h"
34 #define xfrm_state_deref_prot(table, net) \
35 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
37 static void xfrm_state_gc_task(struct work_struct *work);
39 /* Each xfrm_state may be linked to two tables:
41 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
42 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
43 destination/tunnel endpoint. (output)
46 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
47 static struct kmem_cache *xfrm_state_cache __ro_after_init;
49 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
50 static HLIST_HEAD(xfrm_state_gc_list);
52 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54 return refcount_inc_not_zero(&x->refcnt);
57 static inline unsigned int xfrm_dst_hash(struct net *net,
58 const xfrm_address_t *daddr,
59 const xfrm_address_t *saddr,
61 unsigned short family)
63 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
66 static inline unsigned int xfrm_src_hash(struct net *net,
67 const xfrm_address_t *daddr,
68 const xfrm_address_t *saddr,
69 unsigned short family)
71 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
74 static inline unsigned int
75 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
76 __be32 spi, u8 proto, unsigned short family)
78 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
81 static void xfrm_hash_transfer(struct hlist_head *list,
82 struct hlist_head *ndsttable,
83 struct hlist_head *nsrctable,
84 struct hlist_head *nspitable,
85 unsigned int nhashmask)
87 struct hlist_node *tmp;
90 hlist_for_each_entry_safe(x, tmp, list, bydst) {
93 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
94 x->props.reqid, x->props.family,
96 hlist_add_head_rcu(&x->bydst, ndsttable + h);
98 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
101 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
104 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
105 x->id.proto, x->props.family,
107 hlist_add_head_rcu(&x->byspi, nspitable + h);
112 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
114 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
117 static void xfrm_hash_resize(struct work_struct *work)
119 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
120 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
121 unsigned long nsize, osize;
122 unsigned int nhashmask, ohashmask;
125 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
126 ndst = xfrm_hash_alloc(nsize);
129 nsrc = xfrm_hash_alloc(nsize);
131 xfrm_hash_free(ndst, nsize);
134 nspi = xfrm_hash_alloc(nsize);
136 xfrm_hash_free(ndst, nsize);
137 xfrm_hash_free(nsrc, nsize);
141 spin_lock_bh(&net->xfrm.xfrm_state_lock);
142 write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
144 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
145 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
146 for (i = net->xfrm.state_hmask; i >= 0; i--)
147 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
149 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
150 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
151 ohashmask = net->xfrm.state_hmask;
153 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
154 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
155 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
156 net->xfrm.state_hmask = nhashmask;
158 write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
159 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
161 osize = (ohashmask + 1) * sizeof(struct hlist_head);
165 xfrm_hash_free(odst, osize);
166 xfrm_hash_free(osrc, osize);
167 xfrm_hash_free(ospi, osize);
170 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
171 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
173 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
175 int __xfrm_state_delete(struct xfrm_state *x);
177 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
178 static bool km_is_alive(const struct km_event *c);
179 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
181 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
183 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
187 return -EAFNOSUPPORT;
189 #define X(afi, T, name) do { \
190 WARN_ON((afi)->type_ ## name); \
191 (afi)->type_ ## name = (T); \
194 switch (type->proto) {
196 X(afinfo, type, comp);
202 X(afinfo, type, esp);
205 X(afinfo, type, ipip);
207 case IPPROTO_DSTOPTS:
208 X(afinfo, type, dstopts);
210 case IPPROTO_ROUTING:
211 X(afinfo, type, routing);
214 X(afinfo, type, ipip6);
218 err = -EPROTONOSUPPORT;
225 EXPORT_SYMBOL(xfrm_register_type);
227 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
229 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
231 if (unlikely(afinfo == NULL))
234 #define X(afi, T, name) do { \
235 WARN_ON((afi)->type_ ## name != (T)); \
236 (afi)->type_ ## name = NULL; \
239 switch (type->proto) {
241 X(afinfo, type, comp);
247 X(afinfo, type, esp);
250 X(afinfo, type, ipip);
252 case IPPROTO_DSTOPTS:
253 X(afinfo, type, dstopts);
255 case IPPROTO_ROUTING:
256 X(afinfo, type, routing);
259 X(afinfo, type, ipip6);
268 EXPORT_SYMBOL(xfrm_unregister_type);
270 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
272 const struct xfrm_type *type = NULL;
273 struct xfrm_state_afinfo *afinfo;
274 int modload_attempted = 0;
277 afinfo = xfrm_state_get_afinfo(family);
278 if (unlikely(afinfo == NULL))
283 type = afinfo->type_comp;
286 type = afinfo->type_ah;
289 type = afinfo->type_esp;
292 type = afinfo->type_ipip;
294 case IPPROTO_DSTOPTS:
295 type = afinfo->type_dstopts;
297 case IPPROTO_ROUTING:
298 type = afinfo->type_routing;
301 type = afinfo->type_ipip6;
307 if (unlikely(type && !try_module_get(type->owner)))
312 if (!type && !modload_attempted) {
313 request_module("xfrm-type-%d-%d", family, proto);
314 modload_attempted = 1;
321 static void xfrm_put_type(const struct xfrm_type *type)
323 module_put(type->owner);
326 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
327 unsigned short family)
329 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
332 if (unlikely(afinfo == NULL))
333 return -EAFNOSUPPORT;
335 switch (type->proto) {
337 WARN_ON(afinfo->type_offload_esp);
338 afinfo->type_offload_esp = type;
342 err = -EPROTONOSUPPORT;
349 EXPORT_SYMBOL(xfrm_register_type_offload);
351 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
352 unsigned short family)
354 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
356 if (unlikely(afinfo == NULL))
359 switch (type->proto) {
361 WARN_ON(afinfo->type_offload_esp != type);
362 afinfo->type_offload_esp = NULL;
370 EXPORT_SYMBOL(xfrm_unregister_type_offload);
372 static const struct xfrm_type_offload *
373 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
375 const struct xfrm_type_offload *type = NULL;
376 struct xfrm_state_afinfo *afinfo;
379 afinfo = xfrm_state_get_afinfo(family);
380 if (unlikely(afinfo == NULL))
385 type = afinfo->type_offload_esp;
391 if ((type && !try_module_get(type->owner)))
396 if (!type && try_load) {
397 request_module("xfrm-offload-%d-%d", family, proto);
405 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
407 module_put(type->owner);
410 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
412 .encap = XFRM_MODE_BEET,
413 .flags = XFRM_MODE_FLAG_TUNNEL,
416 [XFRM_MODE_TRANSPORT] = {
417 .encap = XFRM_MODE_TRANSPORT,
420 [XFRM_MODE_TUNNEL] = {
421 .encap = XFRM_MODE_TUNNEL,
422 .flags = XFRM_MODE_FLAG_TUNNEL,
427 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
429 .encap = XFRM_MODE_BEET,
430 .flags = XFRM_MODE_FLAG_TUNNEL,
433 [XFRM_MODE_ROUTEOPTIMIZATION] = {
434 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
437 [XFRM_MODE_TRANSPORT] = {
438 .encap = XFRM_MODE_TRANSPORT,
441 [XFRM_MODE_TUNNEL] = {
442 .encap = XFRM_MODE_TUNNEL,
443 .flags = XFRM_MODE_FLAG_TUNNEL,
448 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
450 const struct xfrm_mode *mode;
452 if (unlikely(encap >= XFRM_MODE_MAX))
457 mode = &xfrm4_mode_map[encap];
458 if (mode->family == family)
462 mode = &xfrm6_mode_map[encap];
463 if (mode->family == family)
473 void xfrm_state_free(struct xfrm_state *x)
475 kmem_cache_free(xfrm_state_cache, x);
477 EXPORT_SYMBOL(xfrm_state_free);
479 static void ___xfrm_state_destroy(struct xfrm_state *x)
481 hrtimer_cancel(&x->mtimer);
482 del_timer_sync(&x->rtimer);
489 kfree(x->replay_esn);
490 kfree(x->preplay_esn);
492 xfrm_put_type_offload(x->type_offload);
494 x->type->destructor(x);
495 xfrm_put_type(x->type);
498 put_page(x->xfrag.page);
499 xfrm_dev_state_free(x);
500 security_xfrm_state_free(x);
504 static void xfrm_state_gc_task(struct work_struct *work)
506 struct xfrm_state *x;
507 struct hlist_node *tmp;
508 struct hlist_head gc_list;
510 spin_lock_bh(&xfrm_state_gc_lock);
511 hlist_move_list(&xfrm_state_gc_list, &gc_list);
512 spin_unlock_bh(&xfrm_state_gc_lock);
516 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
517 ___xfrm_state_destroy(x);
520 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
522 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
523 enum hrtimer_restart ret = HRTIMER_NORESTART;
524 time64_t now = ktime_get_real_seconds();
525 time64_t next = TIME64_MAX;
530 if (x->km.state == XFRM_STATE_DEAD)
532 if (x->km.state == XFRM_STATE_EXPIRED)
534 if (x->lft.hard_add_expires_seconds) {
535 long tmo = x->lft.hard_add_expires_seconds +
536 x->curlft.add_time - now;
538 if (x->xflags & XFRM_SOFT_EXPIRE) {
539 /* enter hard expire without soft expire first?!
540 * setting a new date could trigger this.
541 * workaround: fix x->curflt.add_time by below:
543 x->curlft.add_time = now - x->saved_tmo - 1;
544 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
551 if (x->lft.hard_use_expires_seconds) {
552 long tmo = x->lft.hard_use_expires_seconds +
553 (x->curlft.use_time ? : now) - now;
561 if (x->lft.soft_add_expires_seconds) {
562 long tmo = x->lft.soft_add_expires_seconds +
563 x->curlft.add_time - now;
566 x->xflags &= ~XFRM_SOFT_EXPIRE;
567 } else if (tmo < next) {
569 x->xflags |= XFRM_SOFT_EXPIRE;
573 if (x->lft.soft_use_expires_seconds) {
574 long tmo = x->lft.soft_use_expires_seconds +
575 (x->curlft.use_time ? : now) - now;
584 km_state_expired(x, 0, 0);
586 if (next != TIME64_MAX) {
587 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
588 ret = HRTIMER_RESTART;
594 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
595 x->km.state = XFRM_STATE_EXPIRED;
597 err = __xfrm_state_delete(x);
599 km_state_expired(x, 1, 0);
601 xfrm_audit_state_delete(x, err ? 0 : 1, true);
604 spin_unlock(&x->lock);
608 static void xfrm_replay_timer_handler(struct timer_list *t);
610 struct xfrm_state *xfrm_state_alloc(struct net *net)
612 struct xfrm_state *x;
614 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
617 write_pnet(&x->xs_net, net);
618 refcount_set(&x->refcnt, 1);
619 atomic_set(&x->tunnel_users, 0);
620 INIT_LIST_HEAD(&x->km.all);
621 INIT_HLIST_NODE(&x->bydst);
622 INIT_HLIST_NODE(&x->bysrc);
623 INIT_HLIST_NODE(&x->byspi);
624 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
625 x->mtimer.function = xfrm_timer_handler;
626 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
627 x->curlft.add_time = ktime_get_real_seconds();
628 x->lft.soft_byte_limit = XFRM_INF;
629 x->lft.soft_packet_limit = XFRM_INF;
630 x->lft.hard_byte_limit = XFRM_INF;
631 x->lft.hard_packet_limit = XFRM_INF;
632 x->replay_maxage = 0;
633 x->replay_maxdiff = 0;
634 spin_lock_init(&x->lock);
638 EXPORT_SYMBOL(xfrm_state_alloc);
640 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
642 WARN_ON(x->km.state != XFRM_STATE_DEAD);
646 ___xfrm_state_destroy(x);
648 spin_lock_bh(&xfrm_state_gc_lock);
649 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
650 spin_unlock_bh(&xfrm_state_gc_lock);
651 schedule_work(&xfrm_state_gc_work);
654 EXPORT_SYMBOL(__xfrm_state_destroy);
656 int __xfrm_state_delete(struct xfrm_state *x)
658 struct net *net = xs_net(x);
661 if (x->km.state != XFRM_STATE_DEAD) {
662 x->km.state = XFRM_STATE_DEAD;
663 spin_lock(&net->xfrm.xfrm_state_lock);
664 list_del(&x->km.all);
665 hlist_del_rcu(&x->bydst);
666 hlist_del_rcu(&x->bysrc);
668 hlist_del_rcu(&x->byspi);
669 net->xfrm.state_num--;
670 spin_unlock(&net->xfrm.xfrm_state_lock);
673 sock_put(rcu_dereference_raw(x->encap_sk));
675 xfrm_dev_state_delete(x);
677 /* All xfrm_state objects are created by xfrm_state_alloc.
678 * The xfrm_state_alloc call gives a reference, and that
679 * is what we are dropping here.
687 EXPORT_SYMBOL(__xfrm_state_delete);
689 int xfrm_state_delete(struct xfrm_state *x)
693 spin_lock_bh(&x->lock);
694 err = __xfrm_state_delete(x);
695 spin_unlock_bh(&x->lock);
699 EXPORT_SYMBOL(xfrm_state_delete);
701 #ifdef CONFIG_SECURITY_NETWORK_XFRM
703 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
707 for (i = 0; i <= net->xfrm.state_hmask; i++) {
708 struct xfrm_state *x;
710 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
711 if (xfrm_id_proto_match(x->id.proto, proto) &&
712 (err = security_xfrm_state_delete(x)) != 0) {
713 xfrm_audit_state_delete(x, 0, task_valid);
723 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
727 for (i = 0; i <= net->xfrm.state_hmask; i++) {
728 struct xfrm_state *x;
729 struct xfrm_state_offload *xso;
731 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
734 if (xso->dev == dev &&
735 (err = security_xfrm_state_delete(x)) != 0) {
736 xfrm_audit_state_delete(x, 0, task_valid);
746 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
752 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
758 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
760 int i, err = 0, cnt = 0;
762 spin_lock_bh(&net->xfrm.xfrm_state_lock);
763 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
768 for (i = 0; i <= net->xfrm.state_hmask; i++) {
769 struct xfrm_state *x;
771 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
772 if (!xfrm_state_kern(x) &&
773 xfrm_id_proto_match(x->id.proto, proto)) {
775 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
777 err = xfrm_state_delete(x);
778 xfrm_audit_state_delete(x, err ? 0 : 1,
781 xfrm_state_put_sync(x);
787 spin_lock_bh(&net->xfrm.xfrm_state_lock);
793 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
799 EXPORT_SYMBOL(xfrm_state_flush);
801 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
803 int i, err = 0, cnt = 0;
805 spin_lock_bh(&net->xfrm.xfrm_state_lock);
806 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
811 for (i = 0; i <= net->xfrm.state_hmask; i++) {
812 struct xfrm_state *x;
813 struct xfrm_state_offload *xso;
815 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
818 if (!xfrm_state_kern(x) && xso->dev == dev) {
820 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
822 err = xfrm_state_delete(x);
823 xfrm_audit_state_delete(x, err ? 0 : 1,
829 spin_lock_bh(&net->xfrm.xfrm_state_lock);
838 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
841 EXPORT_SYMBOL(xfrm_dev_state_flush);
843 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
845 spin_lock_bh(&net->xfrm.xfrm_state_lock);
846 si->sadcnt = net->xfrm.state_num;
847 si->sadhcnt = net->xfrm.state_hmask + 1;
848 si->sadhmcnt = xfrm_state_hashmax;
849 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
851 EXPORT_SYMBOL(xfrm_sad_getinfo);
854 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
856 const struct flowi4 *fl4 = &fl->u.ip4;
858 sel->daddr.a4 = fl4->daddr;
859 sel->saddr.a4 = fl4->saddr;
860 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
861 sel->dport_mask = htons(0xffff);
862 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
863 sel->sport_mask = htons(0xffff);
864 sel->family = AF_INET;
865 sel->prefixlen_d = 32;
866 sel->prefixlen_s = 32;
867 sel->proto = fl4->flowi4_proto;
868 sel->ifindex = fl4->flowi4_oif;
872 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
874 const struct flowi6 *fl6 = &fl->u.ip6;
876 /* Initialize temporary selector matching only to current session. */
877 *(struct in6_addr *)&sel->daddr = fl6->daddr;
878 *(struct in6_addr *)&sel->saddr = fl6->saddr;
879 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
880 sel->dport_mask = htons(0xffff);
881 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
882 sel->sport_mask = htons(0xffff);
883 sel->family = AF_INET6;
884 sel->prefixlen_d = 128;
885 sel->prefixlen_s = 128;
886 sel->proto = fl6->flowi6_proto;
887 sel->ifindex = fl6->flowi6_oif;
891 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
892 const struct xfrm_tmpl *tmpl,
893 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
894 unsigned short family)
898 __xfrm4_init_tempsel(&x->sel, fl);
901 __xfrm6_init_tempsel(&x->sel, fl);
907 switch (tmpl->encap_family) {
909 if (x->id.daddr.a4 == 0)
910 x->id.daddr.a4 = daddr->a4;
911 x->props.saddr = tmpl->saddr;
912 if (x->props.saddr.a4 == 0)
913 x->props.saddr.a4 = saddr->a4;
916 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
917 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
918 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
919 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
920 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
924 x->props.mode = tmpl->mode;
925 x->props.reqid = tmpl->reqid;
926 x->props.family = tmpl->encap_family;
929 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
930 const xfrm_address_t *daddr,
931 __be32 spi, u8 proto,
932 unsigned short family)
934 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
935 struct xfrm_state *x;
937 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
938 if (x->props.family != family ||
940 x->id.proto != proto ||
941 !xfrm_addr_equal(&x->id.daddr, daddr, family))
944 if ((mark & x->mark.m) != x->mark.v)
946 if (!xfrm_state_hold_rcu(x))
954 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
955 const xfrm_address_t *daddr,
956 const xfrm_address_t *saddr,
957 u8 proto, unsigned short family)
959 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
960 struct xfrm_state *x;
962 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
963 if (x->props.family != family ||
964 x->id.proto != proto ||
965 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
966 !xfrm_addr_equal(&x->props.saddr, saddr, family))
969 if ((mark & x->mark.m) != x->mark.v)
971 if (!xfrm_state_hold_rcu(x))
979 static inline struct xfrm_state *
980 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
982 struct net *net = xs_net(x);
983 u32 mark = x->mark.v & x->mark.m;
986 return __xfrm_state_lookup(net, mark, &x->id.daddr,
987 x->id.spi, x->id.proto, family);
989 return __xfrm_state_lookup_byaddr(net, mark,
992 x->id.proto, family);
995 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
997 if (have_hash_collision &&
998 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
999 net->xfrm.state_num > net->xfrm.state_hmask)
1000 schedule_work(&net->xfrm.state_hash_work);
1003 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1004 const struct flowi *fl, unsigned short family,
1005 struct xfrm_state **best, int *acq_in_progress,
1008 /* Resolution logic:
1009 * 1. There is a valid state with matching selector. Done.
1010 * 2. Valid state with inappropriate selector. Skip.
1012 * Entering area of "sysdeps".
1014 * 3. If state is not valid, selector is temporary, it selects
1015 * only session which triggered previous resolution. Key
1016 * manager will do something to install a state with proper
1019 if (x->km.state == XFRM_STATE_VALID) {
1020 if ((x->sel.family &&
1021 (x->sel.family != family ||
1022 !xfrm_selector_match(&x->sel, fl, family))) ||
1023 !security_xfrm_state_pol_flow_match(x, pol,
1024 &fl->u.__fl_common))
1028 (*best)->km.dying > x->km.dying ||
1029 ((*best)->km.dying == x->km.dying &&
1030 (*best)->curlft.add_time < x->curlft.add_time))
1032 } else if (x->km.state == XFRM_STATE_ACQ) {
1033 *acq_in_progress = 1;
1034 } else if (x->km.state == XFRM_STATE_ERROR ||
1035 x->km.state == XFRM_STATE_EXPIRED) {
1036 if ((!x->sel.family ||
1037 (x->sel.family == family &&
1038 xfrm_selector_match(&x->sel, fl, family))) &&
1039 security_xfrm_state_pol_flow_match(x, pol,
1040 &fl->u.__fl_common))
1046 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1047 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1048 struct xfrm_policy *pol, int *err,
1049 unsigned short family, u32 if_id)
1051 static xfrm_address_t saddr_wildcard = { };
1052 struct net *net = xp_net(pol);
1053 unsigned int h, h_wildcard;
1054 struct xfrm_state *x, *x0, *to_put;
1055 int acquire_in_progress = 0;
1057 struct xfrm_state *best = NULL;
1058 u32 mark = pol->mark.v & pol->mark.m;
1059 unsigned short encap_family = tmpl->encap_family;
1060 unsigned int sequence;
1065 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1068 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1069 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1070 if (x->props.family == encap_family &&
1071 x->props.reqid == tmpl->reqid &&
1072 (mark & x->mark.m) == x->mark.v &&
1073 x->if_id == if_id &&
1074 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1075 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1076 tmpl->mode == x->props.mode &&
1077 tmpl->id.proto == x->id.proto &&
1078 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1079 xfrm_state_look_at(pol, x, fl, family,
1080 &best, &acquire_in_progress, &error);
1082 if (best || acquire_in_progress)
1085 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1086 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1087 if (x->props.family == encap_family &&
1088 x->props.reqid == tmpl->reqid &&
1089 (mark & x->mark.m) == x->mark.v &&
1090 x->if_id == if_id &&
1091 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1092 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1093 tmpl->mode == x->props.mode &&
1094 tmpl->id.proto == x->id.proto &&
1095 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1096 xfrm_state_look_at(pol, x, fl, family,
1097 &best, &acquire_in_progress, &error);
1102 if (!x && !error && !acquire_in_progress) {
1104 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1105 tmpl->id.proto, encap_family)) != NULL) {
1112 /* If the KMs have no listeners (yet...), avoid allocating an SA
1113 * for each and every packet - garbage collection might not
1116 if (!km_is_alive(&c)) {
1121 x = xfrm_state_alloc(net);
1126 /* Initialize temporary state matching only
1127 * to current session. */
1128 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1129 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1132 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1134 x->km.state = XFRM_STATE_DEAD;
1140 if (km_query(x, tmpl, pol) == 0) {
1141 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1142 x->km.state = XFRM_STATE_ACQ;
1143 list_add(&x->km.all, &net->xfrm.state_all);
1144 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1145 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1146 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1148 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1149 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1151 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1152 hrtimer_start(&x->mtimer,
1153 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1154 HRTIMER_MODE_REL_SOFT);
1155 net->xfrm.state_num++;
1156 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1157 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1159 x->km.state = XFRM_STATE_DEAD;
1167 if (!xfrm_state_hold_rcu(x)) {
1172 *err = acquire_in_progress ? -EAGAIN : error;
1176 xfrm_state_put(to_put);
1178 if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1190 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1191 xfrm_address_t *daddr, xfrm_address_t *saddr,
1192 unsigned short family, u8 mode, u8 proto, u32 reqid)
1195 struct xfrm_state *rx = NULL, *x = NULL;
1197 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1198 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1199 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1200 if (x->props.family == family &&
1201 x->props.reqid == reqid &&
1202 (mark & x->mark.m) == x->mark.v &&
1203 x->if_id == if_id &&
1204 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1205 xfrm_state_addr_check(x, daddr, saddr, family) &&
1206 mode == x->props.mode &&
1207 proto == x->id.proto &&
1208 x->km.state == XFRM_STATE_VALID) {
1215 xfrm_state_hold(rx);
1216 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1221 EXPORT_SYMBOL(xfrm_stateonly_find);
1223 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1224 unsigned short family)
1226 struct xfrm_state *x;
1227 struct xfrm_state_walk *w;
1229 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1230 list_for_each_entry(w, &net->xfrm.state_all, all) {
1231 x = container_of(w, struct xfrm_state, km);
1232 if (x->props.family != family ||
1237 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1240 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1243 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1245 static void __xfrm_state_insert(struct xfrm_state *x)
1247 struct net *net = xs_net(x);
1250 list_add(&x->km.all, &net->xfrm.state_all);
1252 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1253 x->props.reqid, x->props.family);
1254 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1256 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1257 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1260 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1263 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1266 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1267 if (x->replay_maxage)
1268 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1270 net->xfrm.state_num++;
1272 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1275 /* net->xfrm.xfrm_state_lock is held */
1276 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1278 struct net *net = xs_net(xnew);
1279 unsigned short family = xnew->props.family;
1280 u32 reqid = xnew->props.reqid;
1281 struct xfrm_state *x;
1283 u32 mark = xnew->mark.v & xnew->mark.m;
1284 u32 if_id = xnew->if_id;
1286 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1287 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1288 if (x->props.family == family &&
1289 x->props.reqid == reqid &&
1290 x->if_id == if_id &&
1291 (mark & x->mark.m) == x->mark.v &&
1292 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1293 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1298 void xfrm_state_insert(struct xfrm_state *x)
1300 struct net *net = xs_net(x);
1302 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1303 __xfrm_state_bump_genids(x);
1304 __xfrm_state_insert(x);
1305 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1307 EXPORT_SYMBOL(xfrm_state_insert);
1309 /* net->xfrm.xfrm_state_lock is held */
1310 static struct xfrm_state *__find_acq_core(struct net *net,
1311 const struct xfrm_mark *m,
1312 unsigned short family, u8 mode,
1313 u32 reqid, u32 if_id, u8 proto,
1314 const xfrm_address_t *daddr,
1315 const xfrm_address_t *saddr,
1318 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1319 struct xfrm_state *x;
1320 u32 mark = m->v & m->m;
1322 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1323 if (x->props.reqid != reqid ||
1324 x->props.mode != mode ||
1325 x->props.family != family ||
1326 x->km.state != XFRM_STATE_ACQ ||
1328 x->id.proto != proto ||
1329 (mark & x->mark.m) != x->mark.v ||
1330 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1331 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1341 x = xfrm_state_alloc(net);
1345 x->sel.daddr.a4 = daddr->a4;
1346 x->sel.saddr.a4 = saddr->a4;
1347 x->sel.prefixlen_d = 32;
1348 x->sel.prefixlen_s = 32;
1349 x->props.saddr.a4 = saddr->a4;
1350 x->id.daddr.a4 = daddr->a4;
1354 x->sel.daddr.in6 = daddr->in6;
1355 x->sel.saddr.in6 = saddr->in6;
1356 x->sel.prefixlen_d = 128;
1357 x->sel.prefixlen_s = 128;
1358 x->props.saddr.in6 = saddr->in6;
1359 x->id.daddr.in6 = daddr->in6;
1363 x->km.state = XFRM_STATE_ACQ;
1364 x->id.proto = proto;
1365 x->props.family = family;
1366 x->props.mode = mode;
1367 x->props.reqid = reqid;
1371 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1373 hrtimer_start(&x->mtimer,
1374 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1375 HRTIMER_MODE_REL_SOFT);
1376 list_add(&x->km.all, &net->xfrm.state_all);
1377 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1378 h = xfrm_src_hash(net, daddr, saddr, family);
1379 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1381 net->xfrm.state_num++;
1383 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1389 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1391 int xfrm_state_add(struct xfrm_state *x)
1393 struct net *net = xs_net(x);
1394 struct xfrm_state *x1, *to_put;
1397 u32 mark = x->mark.v & x->mark.m;
1398 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1400 family = x->props.family;
1404 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1406 x1 = __xfrm_state_locate(x, use_spi, family);
1414 if (use_spi && x->km.seq) {
1415 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1416 if (x1 && ((x1->id.proto != x->id.proto) ||
1417 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1424 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1425 x->props.reqid, x->if_id, x->id.proto,
1426 &x->id.daddr, &x->props.saddr, 0);
1428 __xfrm_state_bump_genids(x);
1429 __xfrm_state_insert(x);
1433 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1436 xfrm_state_delete(x1);
1441 xfrm_state_put(to_put);
1445 EXPORT_SYMBOL(xfrm_state_add);
1447 #ifdef CONFIG_XFRM_MIGRATE
1448 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1450 struct xfrm_user_sec_ctx *uctx;
1451 int size = sizeof(*uctx) + security->ctx_len;
1454 uctx = kmalloc(size, GFP_KERNEL);
1458 uctx->exttype = XFRMA_SEC_CTX;
1460 uctx->ctx_doi = security->ctx_doi;
1461 uctx->ctx_alg = security->ctx_alg;
1462 uctx->ctx_len = security->ctx_len;
1463 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1464 err = security_xfrm_state_alloc(x, uctx);
1472 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1473 struct xfrm_encap_tmpl *encap)
1475 struct net *net = xs_net(orig);
1476 struct xfrm_state *x = xfrm_state_alloc(net);
1480 memcpy(&x->id, &orig->id, sizeof(x->id));
1481 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1482 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1483 x->props.mode = orig->props.mode;
1484 x->props.replay_window = orig->props.replay_window;
1485 x->props.reqid = orig->props.reqid;
1486 x->props.family = orig->props.family;
1487 x->props.saddr = orig->props.saddr;
1490 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1494 x->props.aalgo = orig->props.aalgo;
1497 x->aead = xfrm_algo_aead_clone(orig->aead);
1498 x->geniv = orig->geniv;
1503 x->ealg = xfrm_algo_clone(orig->ealg);
1507 x->props.ealgo = orig->props.ealgo;
1510 x->calg = xfrm_algo_clone(orig->calg);
1514 x->props.calgo = orig->props.calgo;
1516 if (encap || orig->encap) {
1518 x->encap = kmemdup(encap, sizeof(*x->encap),
1521 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1529 if (clone_security(x, orig->security))
1533 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1539 if (orig->replay_esn) {
1540 if (xfrm_replay_clone(x, orig))
1544 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1545 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1547 if (xfrm_init_state(x) < 0)
1550 x->props.flags = orig->props.flags;
1551 x->props.extra_flags = orig->props.extra_flags;
1553 x->if_id = orig->if_id;
1554 x->tfcpad = orig->tfcpad;
1555 x->replay_maxdiff = orig->replay_maxdiff;
1556 x->replay_maxage = orig->replay_maxage;
1557 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1558 x->km.state = orig->km.state;
1559 x->km.seq = orig->km.seq;
1560 x->replay = orig->replay;
1561 x->preplay = orig->preplay;
1571 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1574 struct xfrm_state *x = NULL;
1576 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1579 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1580 m->reqid, m->old_family);
1581 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1582 if (x->props.mode != m->mode ||
1583 x->id.proto != m->proto)
1585 if (m->reqid && x->props.reqid != m->reqid)
1587 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1589 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1596 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1598 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1599 if (x->props.mode != m->mode ||
1600 x->id.proto != m->proto)
1602 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1604 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1612 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1616 EXPORT_SYMBOL(xfrm_migrate_state_find);
1618 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1619 struct xfrm_migrate *m,
1620 struct xfrm_encap_tmpl *encap)
1622 struct xfrm_state *xc;
1624 xc = xfrm_state_clone(x, encap);
1628 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1629 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1632 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1633 /* a care is needed when the destination address of the
1634 state is to be updated as it is a part of triplet */
1635 xfrm_state_insert(xc);
1637 if (xfrm_state_add(xc) < 0)
1646 EXPORT_SYMBOL(xfrm_state_migrate);
1649 int xfrm_state_update(struct xfrm_state *x)
1651 struct xfrm_state *x1, *to_put;
1653 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1654 struct net *net = xs_net(x);
1658 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1659 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1665 if (xfrm_state_kern(x1)) {
1671 if (x1->km.state == XFRM_STATE_ACQ) {
1672 __xfrm_state_insert(x);
1678 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1681 xfrm_state_put(to_put);
1687 xfrm_state_delete(x1);
1693 spin_lock_bh(&x1->lock);
1694 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1695 if (x->encap && x1->encap &&
1696 x->encap->encap_type == x1->encap->encap_type)
1697 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1698 else if (x->encap || x1->encap)
1701 if (x->coaddr && x1->coaddr) {
1702 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1704 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1705 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1706 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1709 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1710 HRTIMER_MODE_REL_SOFT);
1711 if (x1->curlft.use_time)
1712 xfrm_state_check_expire(x1);
1714 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1715 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1717 if (x->props.smark.m || x->props.smark.v)
1718 x1->props.smark = x->props.smark;
1721 x1->if_id = x->if_id;
1723 __xfrm_state_bump_genids(x1);
1724 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1728 x->km.state = XFRM_STATE_DEAD;
1729 __xfrm_state_put(x);
1733 spin_unlock_bh(&x1->lock);
1739 EXPORT_SYMBOL(xfrm_state_update);
1741 int xfrm_state_check_expire(struct xfrm_state *x)
1743 if (!x->curlft.use_time)
1744 x->curlft.use_time = ktime_get_real_seconds();
1746 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1747 x->curlft.packets >= x->lft.hard_packet_limit) {
1748 x->km.state = XFRM_STATE_EXPIRED;
1749 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1754 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1755 x->curlft.packets >= x->lft.soft_packet_limit)) {
1757 km_state_expired(x, 0, 0);
1761 EXPORT_SYMBOL(xfrm_state_check_expire);
1764 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1765 u8 proto, unsigned short family)
1767 struct xfrm_state *x;
1770 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1774 EXPORT_SYMBOL(xfrm_state_lookup);
1777 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1778 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1779 u8 proto, unsigned short family)
1781 struct xfrm_state *x;
1783 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1784 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1785 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1788 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1791 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1792 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1793 const xfrm_address_t *saddr, int create, unsigned short family)
1795 struct xfrm_state *x;
1797 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1798 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1799 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1803 EXPORT_SYMBOL(xfrm_find_acq);
1805 #ifdef CONFIG_XFRM_SUB_POLICY
1806 #if IS_ENABLED(CONFIG_IPV6)
1807 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1809 __xfrm6_sort(void **dst, void **src, int n,
1810 int (*cmp)(const void *p), int maxclass)
1812 int count[XFRM_MAX_DEPTH] = { };
1813 int class[XFRM_MAX_DEPTH];
1816 for (i = 0; i < n; i++) {
1817 int c = cmp(src[i]);
1823 for (i = 2; i < maxclass; i++)
1824 count[i] += count[i - 1];
1826 for (i = 0; i < n; i++) {
1827 dst[count[class[i] - 1]++] = src[i];
1832 /* Rule for xfrm_state:
1834 * rule 1: select IPsec transport except AH
1835 * rule 2: select MIPv6 RO or inbound trigger
1836 * rule 3: select IPsec transport AH
1837 * rule 4: select IPsec tunnel
1840 static int __xfrm6_state_sort_cmp(const void *p)
1842 const struct xfrm_state *v = p;
1844 switch (v->props.mode) {
1845 case XFRM_MODE_TRANSPORT:
1846 if (v->id.proto != IPPROTO_AH)
1850 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1851 case XFRM_MODE_ROUTEOPTIMIZATION:
1852 case XFRM_MODE_IN_TRIGGER:
1855 case XFRM_MODE_TUNNEL:
1856 case XFRM_MODE_BEET:
1862 /* Rule for xfrm_tmpl:
1864 * rule 1: select IPsec transport
1865 * rule 2: select MIPv6 RO or inbound trigger
1866 * rule 3: select IPsec tunnel
1869 static int __xfrm6_tmpl_sort_cmp(const void *p)
1871 const struct xfrm_tmpl *v = p;
1874 case XFRM_MODE_TRANSPORT:
1876 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1877 case XFRM_MODE_ROUTEOPTIMIZATION:
1878 case XFRM_MODE_IN_TRIGGER:
1881 case XFRM_MODE_TUNNEL:
1882 case XFRM_MODE_BEET:
1888 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1889 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1892 __xfrm6_sort(void **dst, void **src, int n,
1893 int (*cmp)(const void *p), int maxclass)
1897 for (i = 0; i < n; i++)
1900 #endif /* CONFIG_IPV6 */
1903 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1904 unsigned short family)
1908 if (family == AF_INET6)
1909 __xfrm6_sort((void **)dst, (void **)src, n,
1910 __xfrm6_tmpl_sort_cmp, 5);
1912 for (i = 0; i < n; i++)
1917 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1918 unsigned short family)
1922 if (family == AF_INET6)
1923 __xfrm6_sort((void **)dst, (void **)src, n,
1924 __xfrm6_state_sort_cmp, 6);
1926 for (i = 0; i < n; i++)
1931 /* Silly enough, but I'm lazy to build resolution list */
1933 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1937 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1938 struct xfrm_state *x;
1940 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1941 if (x->km.seq == seq &&
1942 (mark & x->mark.m) == x->mark.v &&
1943 x->km.state == XFRM_STATE_ACQ) {
1952 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1954 struct xfrm_state *x;
1956 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1957 x = __xfrm_find_acq_byseq(net, mark, seq);
1958 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1961 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1963 u32 xfrm_get_acqseq(void)
1966 static atomic_t acqseq;
1969 res = atomic_inc_return(&acqseq);
1974 EXPORT_SYMBOL(xfrm_get_acqseq);
1976 int verify_spi_info(u8 proto, u32 min, u32 max)
1984 /* IPCOMP spi is 16-bits. */
1998 EXPORT_SYMBOL(verify_spi_info);
2000 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2002 struct net *net = xs_net(x);
2004 struct xfrm_state *x0;
2006 __be32 minspi = htonl(low);
2007 __be32 maxspi = htonl(high);
2009 u32 mark = x->mark.v & x->mark.m;
2011 spin_lock_bh(&x->lock);
2012 if (x->km.state == XFRM_STATE_DEAD)
2021 if (minspi == maxspi) {
2022 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2030 for (h = 0; h < high-low+1; h++) {
2031 spi = low + prandom_u32()%(high-low+1);
2032 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2034 newspi = htonl(spi);
2041 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2043 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2044 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2045 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2051 spin_unlock_bh(&x->lock);
2055 EXPORT_SYMBOL(xfrm_alloc_spi);
2057 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2058 struct xfrm_address_filter *filter)
2061 if ((filter->family == AF_INET ||
2062 filter->family == AF_INET6) &&
2063 x->props.family != filter->family)
2066 return addr_match(&x->props.saddr, &filter->saddr,
2068 addr_match(&x->id.daddr, &filter->daddr,
2074 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2075 int (*func)(struct xfrm_state *, int, void*),
2078 struct xfrm_state *state;
2079 struct xfrm_state_walk *x;
2082 if (walk->seq != 0 && list_empty(&walk->all))
2085 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2086 if (list_empty(&walk->all))
2087 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2089 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2090 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2091 if (x->state == XFRM_STATE_DEAD)
2093 state = container_of(x, struct xfrm_state, km);
2094 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2096 if (!__xfrm_state_filter_match(state, walk->filter))
2098 err = func(state, walk->seq, data);
2100 list_move_tail(&walk->all, &x->all);
2105 if (walk->seq == 0) {
2109 list_del_init(&walk->all);
2111 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2114 EXPORT_SYMBOL(xfrm_state_walk);
2116 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2117 struct xfrm_address_filter *filter)
2119 INIT_LIST_HEAD(&walk->all);
2120 walk->proto = proto;
2121 walk->state = XFRM_STATE_DEAD;
2123 walk->filter = filter;
2125 EXPORT_SYMBOL(xfrm_state_walk_init);
2127 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2129 kfree(walk->filter);
2131 if (list_empty(&walk->all))
2134 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2135 list_del(&walk->all);
2136 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2138 EXPORT_SYMBOL(xfrm_state_walk_done);
2140 static void xfrm_replay_timer_handler(struct timer_list *t)
2142 struct xfrm_state *x = from_timer(x, t, rtimer);
2144 spin_lock(&x->lock);
2146 if (x->km.state == XFRM_STATE_VALID) {
2147 if (xfrm_aevent_is_on(xs_net(x)))
2148 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2150 x->xflags |= XFRM_TIME_DEFER;
2153 spin_unlock(&x->lock);
2156 static LIST_HEAD(xfrm_km_list);
2158 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2160 struct xfrm_mgr *km;
2163 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2164 if (km->notify_policy)
2165 km->notify_policy(xp, dir, c);
2169 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2171 struct xfrm_mgr *km;
2173 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2179 EXPORT_SYMBOL(km_policy_notify);
2180 EXPORT_SYMBOL(km_state_notify);
2182 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2188 c.event = XFRM_MSG_EXPIRE;
2189 km_state_notify(x, &c);
2192 EXPORT_SYMBOL(km_state_expired);
2194 * We send to all registered managers regardless of failure
2195 * We are happy with one success
2197 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2199 int err = -EINVAL, acqret;
2200 struct xfrm_mgr *km;
2203 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2204 acqret = km->acquire(x, t, pol);
2211 EXPORT_SYMBOL(km_query);
2213 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2216 struct xfrm_mgr *km;
2219 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2220 if (km->new_mapping)
2221 err = km->new_mapping(x, ipaddr, sport);
2228 EXPORT_SYMBOL(km_new_mapping);
2230 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2236 c.event = XFRM_MSG_POLEXPIRE;
2237 km_policy_notify(pol, dir, &c);
2239 EXPORT_SYMBOL(km_policy_expired);
2241 #ifdef CONFIG_XFRM_MIGRATE
2242 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2243 const struct xfrm_migrate *m, int num_migrate,
2244 const struct xfrm_kmaddress *k,
2245 const struct xfrm_encap_tmpl *encap)
2249 struct xfrm_mgr *km;
2252 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2254 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2263 EXPORT_SYMBOL(km_migrate);
2266 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2270 struct xfrm_mgr *km;
2273 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2275 ret = km->report(net, proto, sel, addr);
2283 EXPORT_SYMBOL(km_report);
2285 static bool km_is_alive(const struct km_event *c)
2287 struct xfrm_mgr *km;
2288 bool is_alive = false;
2291 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2292 if (km->is_alive && km->is_alive(c)) {
2302 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2303 static DEFINE_SPINLOCK(xfrm_translator_lock);
2304 static struct xfrm_translator __rcu *xfrm_translator;
2306 struct xfrm_translator *xfrm_get_translator(void)
2308 struct xfrm_translator *xtr;
2311 xtr = rcu_dereference(xfrm_translator);
2314 if (!try_module_get(xtr->owner))
2320 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2322 void xfrm_put_translator(struct xfrm_translator *xtr)
2324 module_put(xtr->owner);
2326 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2328 int xfrm_register_translator(struct xfrm_translator *xtr)
2332 spin_lock_bh(&xfrm_translator_lock);
2333 if (unlikely(xfrm_translator != NULL))
2336 rcu_assign_pointer(xfrm_translator, xtr);
2337 spin_unlock_bh(&xfrm_translator_lock);
2341 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2343 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2347 spin_lock_bh(&xfrm_translator_lock);
2348 if (likely(xfrm_translator != NULL)) {
2349 if (rcu_access_pointer(xfrm_translator) != xtr)
2352 RCU_INIT_POINTER(xfrm_translator, NULL);
2354 spin_unlock_bh(&xfrm_translator_lock);
2359 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2362 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2366 struct xfrm_mgr *km;
2367 struct xfrm_policy *pol = NULL;
2369 if (sockptr_is_null(optval) && !optlen) {
2370 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2371 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2376 if (optlen <= 0 || optlen > PAGE_SIZE)
2379 data = memdup_sockptr(optval, optlen);
2381 return PTR_ERR(data);
2383 if (in_compat_syscall()) {
2384 struct xfrm_translator *xtr = xfrm_get_translator();
2391 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2392 xfrm_put_translator(xtr);
2401 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2402 pol = km->compile_policy(sk, optname, data,
2410 xfrm_sk_policy_insert(sk, err, pol);
2419 EXPORT_SYMBOL(xfrm_user_policy);
2421 static DEFINE_SPINLOCK(xfrm_km_lock);
2423 int xfrm_register_km(struct xfrm_mgr *km)
2425 spin_lock_bh(&xfrm_km_lock);
2426 list_add_tail_rcu(&km->list, &xfrm_km_list);
2427 spin_unlock_bh(&xfrm_km_lock);
2430 EXPORT_SYMBOL(xfrm_register_km);
2432 int xfrm_unregister_km(struct xfrm_mgr *km)
2434 spin_lock_bh(&xfrm_km_lock);
2435 list_del_rcu(&km->list);
2436 spin_unlock_bh(&xfrm_km_lock);
2440 EXPORT_SYMBOL(xfrm_unregister_km);
2442 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2446 if (WARN_ON(afinfo->family >= NPROTO))
2447 return -EAFNOSUPPORT;
2449 spin_lock_bh(&xfrm_state_afinfo_lock);
2450 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2453 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2454 spin_unlock_bh(&xfrm_state_afinfo_lock);
2457 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2459 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2461 int err = 0, family = afinfo->family;
2463 if (WARN_ON(family >= NPROTO))
2464 return -EAFNOSUPPORT;
2466 spin_lock_bh(&xfrm_state_afinfo_lock);
2467 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2468 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2471 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2473 spin_unlock_bh(&xfrm_state_afinfo_lock);
2477 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2479 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2481 if (unlikely(family >= NPROTO))
2484 return rcu_dereference(xfrm_state_afinfo[family]);
2486 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2488 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2490 struct xfrm_state_afinfo *afinfo;
2491 if (unlikely(family >= NPROTO))
2494 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2495 if (unlikely(!afinfo))
2500 void xfrm_flush_gc(void)
2502 flush_work(&xfrm_state_gc_work);
2504 EXPORT_SYMBOL(xfrm_flush_gc);
2506 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2507 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2510 struct xfrm_state *t = x->tunnel;
2512 if (atomic_read(&t->tunnel_users) == 2)
2513 xfrm_state_delete(t);
2514 atomic_dec(&t->tunnel_users);
2515 xfrm_state_put_sync(t);
2519 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2521 u32 __xfrm_state_mtu(struct xfrm_state *x, int mtu)
2523 const struct xfrm_type *type = READ_ONCE(x->type);
2524 struct crypto_aead *aead;
2525 u32 blksize, net_adj = 0;
2527 if (x->km.state != XFRM_STATE_VALID ||
2528 !type || type->proto != IPPROTO_ESP)
2529 return mtu - x->props.header_len;
2532 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2534 switch (x->props.mode) {
2535 case XFRM_MODE_TRANSPORT:
2536 case XFRM_MODE_BEET:
2537 if (x->props.family == AF_INET)
2538 net_adj = sizeof(struct iphdr);
2539 else if (x->props.family == AF_INET6)
2540 net_adj = sizeof(struct ipv6hdr);
2542 case XFRM_MODE_TUNNEL:
2549 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2550 net_adj) & ~(blksize - 1)) + net_adj - 2;
2552 EXPORT_SYMBOL_GPL(__xfrm_state_mtu);
2554 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2556 mtu = __xfrm_state_mtu(x, mtu);
2558 if (x->props.family == AF_INET6 && mtu < IPV6_MIN_MTU)
2559 return IPV6_MIN_MTU;
2564 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2566 const struct xfrm_mode *inner_mode;
2567 const struct xfrm_mode *outer_mode;
2568 int family = x->props.family;
2571 if (family == AF_INET &&
2572 xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2573 x->props.flags |= XFRM_STATE_NOPMTUDISC;
2575 err = -EPROTONOSUPPORT;
2577 if (x->sel.family != AF_UNSPEC) {
2578 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2579 if (inner_mode == NULL)
2582 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2583 family != x->sel.family)
2586 x->inner_mode = *inner_mode;
2588 const struct xfrm_mode *inner_mode_iaf;
2589 int iafamily = AF_INET;
2591 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2592 if (inner_mode == NULL)
2595 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2598 x->inner_mode = *inner_mode;
2600 if (x->props.family == AF_INET)
2601 iafamily = AF_INET6;
2603 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2604 if (inner_mode_iaf) {
2605 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2606 x->inner_mode_iaf = *inner_mode_iaf;
2610 x->type = xfrm_get_type(x->id.proto, family);
2611 if (x->type == NULL)
2614 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2616 err = x->type->init_state(x);
2620 outer_mode = xfrm_get_mode(x->props.mode, family);
2622 err = -EPROTONOSUPPORT;
2626 x->outer_mode = *outer_mode;
2628 err = xfrm_init_replay(x);
2637 EXPORT_SYMBOL(__xfrm_init_state);
2639 int xfrm_init_state(struct xfrm_state *x)
2643 err = __xfrm_init_state(x, true, false);
2645 x->km.state = XFRM_STATE_VALID;
2650 EXPORT_SYMBOL(xfrm_init_state);
2652 int __net_init xfrm_state_init(struct net *net)
2656 if (net_eq(net, &init_net))
2657 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2658 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2660 INIT_LIST_HEAD(&net->xfrm.state_all);
2662 sz = sizeof(struct hlist_head) * 8;
2664 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2665 if (!net->xfrm.state_bydst)
2667 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2668 if (!net->xfrm.state_bysrc)
2670 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2671 if (!net->xfrm.state_byspi)
2673 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2675 net->xfrm.state_num = 0;
2676 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2677 spin_lock_init(&net->xfrm.xfrm_state_lock);
2678 seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2679 &net->xfrm.xfrm_state_lock);
2683 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2685 xfrm_hash_free(net->xfrm.state_bydst, sz);
2690 void xfrm_state_fini(struct net *net)
2694 flush_work(&net->xfrm.state_hash_work);
2695 flush_work(&xfrm_state_gc_work);
2696 xfrm_state_flush(net, 0, false, true);
2698 WARN_ON(!list_empty(&net->xfrm.state_all));
2700 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2701 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2702 xfrm_hash_free(net->xfrm.state_byspi, sz);
2703 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2704 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2705 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2706 xfrm_hash_free(net->xfrm.state_bydst, sz);
2709 #ifdef CONFIG_AUDITSYSCALL
2710 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2711 struct audit_buffer *audit_buf)
2713 struct xfrm_sec_ctx *ctx = x->security;
2714 u32 spi = ntohl(x->id.spi);
2717 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2718 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2720 switch (x->props.family) {
2722 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2723 &x->props.saddr.a4, &x->id.daddr.a4);
2726 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2727 x->props.saddr.a6, x->id.daddr.a6);
2731 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2734 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2735 struct audit_buffer *audit_buf)
2737 const struct iphdr *iph4;
2738 const struct ipv6hdr *iph6;
2743 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2744 &iph4->saddr, &iph4->daddr);
2747 iph6 = ipv6_hdr(skb);
2748 audit_log_format(audit_buf,
2749 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2750 &iph6->saddr, &iph6->daddr,
2751 iph6->flow_lbl[0] & 0x0f,
2758 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2760 struct audit_buffer *audit_buf;
2762 audit_buf = xfrm_audit_start("SAD-add");
2763 if (audit_buf == NULL)
2765 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2766 xfrm_audit_helper_sainfo(x, audit_buf);
2767 audit_log_format(audit_buf, " res=%u", result);
2768 audit_log_end(audit_buf);
2770 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2772 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2774 struct audit_buffer *audit_buf;
2776 audit_buf = xfrm_audit_start("SAD-delete");
2777 if (audit_buf == NULL)
2779 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2780 xfrm_audit_helper_sainfo(x, audit_buf);
2781 audit_log_format(audit_buf, " res=%u", result);
2782 audit_log_end(audit_buf);
2784 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2786 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2787 struct sk_buff *skb)
2789 struct audit_buffer *audit_buf;
2792 audit_buf = xfrm_audit_start("SA-replay-overflow");
2793 if (audit_buf == NULL)
2795 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2796 /* don't record the sequence number because it's inherent in this kind
2797 * of audit message */
2798 spi = ntohl(x->id.spi);
2799 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2800 audit_log_end(audit_buf);
2802 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2804 void xfrm_audit_state_replay(struct xfrm_state *x,
2805 struct sk_buff *skb, __be32 net_seq)
2807 struct audit_buffer *audit_buf;
2810 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2811 if (audit_buf == NULL)
2813 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2814 spi = ntohl(x->id.spi);
2815 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2816 spi, spi, ntohl(net_seq));
2817 audit_log_end(audit_buf);
2819 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2821 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2823 struct audit_buffer *audit_buf;
2825 audit_buf = xfrm_audit_start("SA-notfound");
2826 if (audit_buf == NULL)
2828 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2829 audit_log_end(audit_buf);
2831 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2833 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2834 __be32 net_spi, __be32 net_seq)
2836 struct audit_buffer *audit_buf;
2839 audit_buf = xfrm_audit_start("SA-notfound");
2840 if (audit_buf == NULL)
2842 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2843 spi = ntohl(net_spi);
2844 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2845 spi, spi, ntohl(net_seq));
2846 audit_log_end(audit_buf);
2848 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2850 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2851 struct sk_buff *skb, u8 proto)
2853 struct audit_buffer *audit_buf;
2857 audit_buf = xfrm_audit_start("SA-icv-failure");
2858 if (audit_buf == NULL)
2860 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2861 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2862 u32 spi = ntohl(net_spi);
2863 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2864 spi, spi, ntohl(net_seq));
2866 audit_log_end(audit_buf);
2868 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2869 #endif /* CONFIG_AUDITSYSCALL */