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 __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
48 static struct kmem_cache *xfrm_state_cache __ro_after_init;
50 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
51 static HLIST_HEAD(xfrm_state_gc_list);
53 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
55 return refcount_inc_not_zero(&x->refcnt);
58 static inline unsigned int xfrm_dst_hash(struct net *net,
59 const xfrm_address_t *daddr,
60 const xfrm_address_t *saddr,
62 unsigned short family)
64 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
67 static inline unsigned int xfrm_src_hash(struct net *net,
68 const xfrm_address_t *daddr,
69 const xfrm_address_t *saddr,
70 unsigned short family)
72 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
75 static inline unsigned int
76 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
77 __be32 spi, u8 proto, unsigned short family)
79 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
82 static void xfrm_hash_transfer(struct hlist_head *list,
83 struct hlist_head *ndsttable,
84 struct hlist_head *nsrctable,
85 struct hlist_head *nspitable,
86 unsigned int nhashmask)
88 struct hlist_node *tmp;
91 hlist_for_each_entry_safe(x, tmp, list, bydst) {
94 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
95 x->props.reqid, x->props.family,
97 hlist_add_head_rcu(&x->bydst, ndsttable + h);
99 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
102 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
105 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
106 x->id.proto, x->props.family,
108 hlist_add_head_rcu(&x->byspi, nspitable + h);
113 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
115 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
118 static void xfrm_hash_resize(struct work_struct *work)
120 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
121 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
122 unsigned long nsize, osize;
123 unsigned int nhashmask, ohashmask;
126 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
127 ndst = xfrm_hash_alloc(nsize);
130 nsrc = xfrm_hash_alloc(nsize);
132 xfrm_hash_free(ndst, nsize);
135 nspi = xfrm_hash_alloc(nsize);
137 xfrm_hash_free(ndst, nsize);
138 xfrm_hash_free(nsrc, nsize);
142 spin_lock_bh(&net->xfrm.xfrm_state_lock);
143 write_seqcount_begin(&xfrm_state_hash_generation);
145 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
146 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
147 for (i = net->xfrm.state_hmask; i >= 0; i--)
148 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
150 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
151 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
152 ohashmask = net->xfrm.state_hmask;
154 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
155 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
156 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
157 net->xfrm.state_hmask = nhashmask;
159 write_seqcount_end(&xfrm_state_hash_generation);
160 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
162 osize = (ohashmask + 1) * sizeof(struct hlist_head);
166 xfrm_hash_free(odst, osize);
167 xfrm_hash_free(osrc, osize);
168 xfrm_hash_free(ospi, osize);
171 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
172 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
174 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
176 int __xfrm_state_delete(struct xfrm_state *x);
178 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
179 static bool km_is_alive(const struct km_event *c);
180 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
182 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
184 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
188 return -EAFNOSUPPORT;
190 #define X(afi, T, name) do { \
191 WARN_ON((afi)->type_ ## name); \
192 (afi)->type_ ## name = (T); \
195 switch (type->proto) {
197 X(afinfo, type, comp);
203 X(afinfo, type, esp);
206 X(afinfo, type, ipip);
208 case IPPROTO_DSTOPTS:
209 X(afinfo, type, dstopts);
211 case IPPROTO_ROUTING:
212 X(afinfo, type, routing);
215 X(afinfo, type, ipip6);
219 err = -EPROTONOSUPPORT;
226 EXPORT_SYMBOL(xfrm_register_type);
228 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
230 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
232 if (unlikely(afinfo == NULL))
235 #define X(afi, T, name) do { \
236 WARN_ON((afi)->type_ ## name != (T)); \
237 (afi)->type_ ## name = NULL; \
240 switch (type->proto) {
242 X(afinfo, type, comp);
248 X(afinfo, type, esp);
251 X(afinfo, type, ipip);
253 case IPPROTO_DSTOPTS:
254 X(afinfo, type, dstopts);
256 case IPPROTO_ROUTING:
257 X(afinfo, type, routing);
260 X(afinfo, type, ipip6);
269 EXPORT_SYMBOL(xfrm_unregister_type);
271 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
273 const struct xfrm_type *type = NULL;
274 struct xfrm_state_afinfo *afinfo;
275 int modload_attempted = 0;
278 afinfo = xfrm_state_get_afinfo(family);
279 if (unlikely(afinfo == NULL))
284 type = afinfo->type_comp;
287 type = afinfo->type_ah;
290 type = afinfo->type_esp;
293 type = afinfo->type_ipip;
295 case IPPROTO_DSTOPTS:
296 type = afinfo->type_dstopts;
298 case IPPROTO_ROUTING:
299 type = afinfo->type_routing;
302 type = afinfo->type_ipip6;
308 if (unlikely(type && !try_module_get(type->owner)))
313 if (!type && !modload_attempted) {
314 request_module("xfrm-type-%d-%d", family, proto);
315 modload_attempted = 1;
322 static void xfrm_put_type(const struct xfrm_type *type)
324 module_put(type->owner);
327 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
328 unsigned short family)
330 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
333 if (unlikely(afinfo == NULL))
334 return -EAFNOSUPPORT;
336 switch (type->proto) {
338 WARN_ON(afinfo->type_offload_esp);
339 afinfo->type_offload_esp = type;
343 err = -EPROTONOSUPPORT;
350 EXPORT_SYMBOL(xfrm_register_type_offload);
352 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
353 unsigned short family)
355 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
357 if (unlikely(afinfo == NULL))
360 switch (type->proto) {
362 WARN_ON(afinfo->type_offload_esp != type);
363 afinfo->type_offload_esp = NULL;
371 EXPORT_SYMBOL(xfrm_unregister_type_offload);
373 static const struct xfrm_type_offload *
374 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
376 const struct xfrm_type_offload *type = NULL;
377 struct xfrm_state_afinfo *afinfo;
380 afinfo = xfrm_state_get_afinfo(family);
381 if (unlikely(afinfo == NULL))
386 type = afinfo->type_offload_esp;
392 if ((type && !try_module_get(type->owner)))
397 if (!type && try_load) {
398 request_module("xfrm-offload-%d-%d", family, proto);
406 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
408 module_put(type->owner);
411 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
413 .encap = XFRM_MODE_BEET,
414 .flags = XFRM_MODE_FLAG_TUNNEL,
417 [XFRM_MODE_TRANSPORT] = {
418 .encap = XFRM_MODE_TRANSPORT,
421 [XFRM_MODE_TUNNEL] = {
422 .encap = XFRM_MODE_TUNNEL,
423 .flags = XFRM_MODE_FLAG_TUNNEL,
428 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
430 .encap = XFRM_MODE_BEET,
431 .flags = XFRM_MODE_FLAG_TUNNEL,
434 [XFRM_MODE_ROUTEOPTIMIZATION] = {
435 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
438 [XFRM_MODE_TRANSPORT] = {
439 .encap = XFRM_MODE_TRANSPORT,
442 [XFRM_MODE_TUNNEL] = {
443 .encap = XFRM_MODE_TUNNEL,
444 .flags = XFRM_MODE_FLAG_TUNNEL,
449 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
451 const struct xfrm_mode *mode;
453 if (unlikely(encap >= XFRM_MODE_MAX))
458 mode = &xfrm4_mode_map[encap];
459 if (mode->family == family)
463 mode = &xfrm6_mode_map[encap];
464 if (mode->family == family)
474 void xfrm_state_free(struct xfrm_state *x)
476 kmem_cache_free(xfrm_state_cache, x);
478 EXPORT_SYMBOL(xfrm_state_free);
480 static void ___xfrm_state_destroy(struct xfrm_state *x)
482 hrtimer_cancel(&x->mtimer);
483 del_timer_sync(&x->rtimer);
490 kfree(x->replay_esn);
491 kfree(x->preplay_esn);
493 xfrm_put_type_offload(x->type_offload);
495 x->type->destructor(x);
496 xfrm_put_type(x->type);
499 put_page(x->xfrag.page);
500 xfrm_dev_state_free(x);
501 security_xfrm_state_free(x);
505 static void xfrm_state_gc_task(struct work_struct *work)
507 struct xfrm_state *x;
508 struct hlist_node *tmp;
509 struct hlist_head gc_list;
511 spin_lock_bh(&xfrm_state_gc_lock);
512 hlist_move_list(&xfrm_state_gc_list, &gc_list);
513 spin_unlock_bh(&xfrm_state_gc_lock);
517 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
518 ___xfrm_state_destroy(x);
521 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
523 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
524 enum hrtimer_restart ret = HRTIMER_NORESTART;
525 time64_t now = ktime_get_real_seconds();
526 time64_t next = TIME64_MAX;
531 if (x->km.state == XFRM_STATE_DEAD)
533 if (x->km.state == XFRM_STATE_EXPIRED)
535 if (x->lft.hard_add_expires_seconds) {
536 long tmo = x->lft.hard_add_expires_seconds +
537 x->curlft.add_time - now;
539 if (x->xflags & XFRM_SOFT_EXPIRE) {
540 /* enter hard expire without soft expire first?!
541 * setting a new date could trigger this.
542 * workaround: fix x->curflt.add_time by below:
544 x->curlft.add_time = now - x->saved_tmo - 1;
545 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
552 if (x->lft.hard_use_expires_seconds) {
553 long tmo = x->lft.hard_use_expires_seconds +
554 (x->curlft.use_time ? : now) - now;
562 if (x->lft.soft_add_expires_seconds) {
563 long tmo = x->lft.soft_add_expires_seconds +
564 x->curlft.add_time - now;
567 x->xflags &= ~XFRM_SOFT_EXPIRE;
568 } else if (tmo < next) {
570 x->xflags |= XFRM_SOFT_EXPIRE;
574 if (x->lft.soft_use_expires_seconds) {
575 long tmo = x->lft.soft_use_expires_seconds +
576 (x->curlft.use_time ? : now) - now;
585 km_state_expired(x, 0, 0);
587 if (next != TIME64_MAX) {
588 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
589 ret = HRTIMER_RESTART;
595 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
596 x->km.state = XFRM_STATE_EXPIRED;
598 err = __xfrm_state_delete(x);
600 km_state_expired(x, 1, 0);
602 xfrm_audit_state_delete(x, err ? 0 : 1, true);
605 spin_unlock(&x->lock);
609 static void xfrm_replay_timer_handler(struct timer_list *t);
611 struct xfrm_state *xfrm_state_alloc(struct net *net)
613 struct xfrm_state *x;
615 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
618 write_pnet(&x->xs_net, net);
619 refcount_set(&x->refcnt, 1);
620 atomic_set(&x->tunnel_users, 0);
621 INIT_LIST_HEAD(&x->km.all);
622 INIT_HLIST_NODE(&x->bydst);
623 INIT_HLIST_NODE(&x->bysrc);
624 INIT_HLIST_NODE(&x->byspi);
625 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
626 x->mtimer.function = xfrm_timer_handler;
627 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
628 x->curlft.add_time = ktime_get_real_seconds();
629 x->lft.soft_byte_limit = XFRM_INF;
630 x->lft.soft_packet_limit = XFRM_INF;
631 x->lft.hard_byte_limit = XFRM_INF;
632 x->lft.hard_packet_limit = XFRM_INF;
633 x->replay_maxage = 0;
634 x->replay_maxdiff = 0;
635 spin_lock_init(&x->lock);
639 EXPORT_SYMBOL(xfrm_state_alloc);
641 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
643 WARN_ON(x->km.state != XFRM_STATE_DEAD);
647 ___xfrm_state_destroy(x);
649 spin_lock_bh(&xfrm_state_gc_lock);
650 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
651 spin_unlock_bh(&xfrm_state_gc_lock);
652 schedule_work(&xfrm_state_gc_work);
655 EXPORT_SYMBOL(__xfrm_state_destroy);
657 int __xfrm_state_delete(struct xfrm_state *x)
659 struct net *net = xs_net(x);
662 if (x->km.state != XFRM_STATE_DEAD) {
663 x->km.state = XFRM_STATE_DEAD;
664 spin_lock(&net->xfrm.xfrm_state_lock);
665 list_del(&x->km.all);
666 hlist_del_rcu(&x->bydst);
667 hlist_del_rcu(&x->bysrc);
669 hlist_del_rcu(&x->byspi);
670 net->xfrm.state_num--;
671 spin_unlock(&net->xfrm.xfrm_state_lock);
674 sock_put(rcu_dereference_raw(x->encap_sk));
676 xfrm_dev_state_delete(x);
678 /* All xfrm_state objects are created by xfrm_state_alloc.
679 * The xfrm_state_alloc call gives a reference, and that
680 * is what we are dropping here.
688 EXPORT_SYMBOL(__xfrm_state_delete);
690 int xfrm_state_delete(struct xfrm_state *x)
694 spin_lock_bh(&x->lock);
695 err = __xfrm_state_delete(x);
696 spin_unlock_bh(&x->lock);
700 EXPORT_SYMBOL(xfrm_state_delete);
702 #ifdef CONFIG_SECURITY_NETWORK_XFRM
704 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
708 for (i = 0; i <= net->xfrm.state_hmask; i++) {
709 struct xfrm_state *x;
711 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
712 if (xfrm_id_proto_match(x->id.proto, proto) &&
713 (err = security_xfrm_state_delete(x)) != 0) {
714 xfrm_audit_state_delete(x, 0, task_valid);
724 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
728 for (i = 0; i <= net->xfrm.state_hmask; i++) {
729 struct xfrm_state *x;
730 struct xfrm_state_offload *xso;
732 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
735 if (xso->dev == dev &&
736 (err = security_xfrm_state_delete(x)) != 0) {
737 xfrm_audit_state_delete(x, 0, task_valid);
747 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
753 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
759 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
761 int i, err = 0, cnt = 0;
763 spin_lock_bh(&net->xfrm.xfrm_state_lock);
764 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
769 for (i = 0; i <= net->xfrm.state_hmask; i++) {
770 struct xfrm_state *x;
772 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
773 if (!xfrm_state_kern(x) &&
774 xfrm_id_proto_match(x->id.proto, proto)) {
776 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
778 err = xfrm_state_delete(x);
779 xfrm_audit_state_delete(x, err ? 0 : 1,
782 xfrm_state_put_sync(x);
788 spin_lock_bh(&net->xfrm.xfrm_state_lock);
794 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
800 EXPORT_SYMBOL(xfrm_state_flush);
802 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
804 int i, err = 0, cnt = 0;
806 spin_lock_bh(&net->xfrm.xfrm_state_lock);
807 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
812 for (i = 0; i <= net->xfrm.state_hmask; i++) {
813 struct xfrm_state *x;
814 struct xfrm_state_offload *xso;
816 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
819 if (!xfrm_state_kern(x) && xso->dev == dev) {
821 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
823 err = xfrm_state_delete(x);
824 xfrm_audit_state_delete(x, err ? 0 : 1,
830 spin_lock_bh(&net->xfrm.xfrm_state_lock);
839 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
842 EXPORT_SYMBOL(xfrm_dev_state_flush);
844 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
846 spin_lock_bh(&net->xfrm.xfrm_state_lock);
847 si->sadcnt = net->xfrm.state_num;
848 si->sadhcnt = net->xfrm.state_hmask + 1;
849 si->sadhmcnt = xfrm_state_hashmax;
850 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
852 EXPORT_SYMBOL(xfrm_sad_getinfo);
855 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
857 const struct flowi4 *fl4 = &fl->u.ip4;
859 sel->daddr.a4 = fl4->daddr;
860 sel->saddr.a4 = fl4->saddr;
861 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
862 sel->dport_mask = htons(0xffff);
863 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
864 sel->sport_mask = htons(0xffff);
865 sel->family = AF_INET;
866 sel->prefixlen_d = 32;
867 sel->prefixlen_s = 32;
868 sel->proto = fl4->flowi4_proto;
869 sel->ifindex = fl4->flowi4_oif;
873 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
875 const struct flowi6 *fl6 = &fl->u.ip6;
877 /* Initialize temporary selector matching only to current session. */
878 *(struct in6_addr *)&sel->daddr = fl6->daddr;
879 *(struct in6_addr *)&sel->saddr = fl6->saddr;
880 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
881 sel->dport_mask = htons(0xffff);
882 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
883 sel->sport_mask = htons(0xffff);
884 sel->family = AF_INET6;
885 sel->prefixlen_d = 128;
886 sel->prefixlen_s = 128;
887 sel->proto = fl6->flowi6_proto;
888 sel->ifindex = fl6->flowi6_oif;
892 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
893 const struct xfrm_tmpl *tmpl,
894 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
895 unsigned short family)
899 __xfrm4_init_tempsel(&x->sel, fl);
902 __xfrm6_init_tempsel(&x->sel, fl);
908 switch (tmpl->encap_family) {
910 if (x->id.daddr.a4 == 0)
911 x->id.daddr.a4 = daddr->a4;
912 x->props.saddr = tmpl->saddr;
913 if (x->props.saddr.a4 == 0)
914 x->props.saddr.a4 = saddr->a4;
917 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
918 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
919 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
920 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
921 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
925 x->props.mode = tmpl->mode;
926 x->props.reqid = tmpl->reqid;
927 x->props.family = tmpl->encap_family;
930 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
931 const xfrm_address_t *daddr,
932 __be32 spi, u8 proto,
933 unsigned short family)
935 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
936 struct xfrm_state *x;
938 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
939 if (x->props.family != family ||
941 x->id.proto != proto ||
942 !xfrm_addr_equal(&x->id.daddr, daddr, family))
945 if ((mark & x->mark.m) != x->mark.v)
947 if (!xfrm_state_hold_rcu(x))
955 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
956 const xfrm_address_t *daddr,
957 const xfrm_address_t *saddr,
958 u8 proto, unsigned short family)
960 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
961 struct xfrm_state *x;
963 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
964 if (x->props.family != family ||
965 x->id.proto != proto ||
966 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
967 !xfrm_addr_equal(&x->props.saddr, saddr, family))
970 if ((mark & x->mark.m) != x->mark.v)
972 if (!xfrm_state_hold_rcu(x))
980 static inline struct xfrm_state *
981 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
983 struct net *net = xs_net(x);
984 u32 mark = x->mark.v & x->mark.m;
987 return __xfrm_state_lookup(net, mark, &x->id.daddr,
988 x->id.spi, x->id.proto, family);
990 return __xfrm_state_lookup_byaddr(net, mark,
993 x->id.proto, family);
996 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
998 if (have_hash_collision &&
999 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1000 net->xfrm.state_num > net->xfrm.state_hmask)
1001 schedule_work(&net->xfrm.state_hash_work);
1004 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1005 const struct flowi *fl, unsigned short family,
1006 struct xfrm_state **best, int *acq_in_progress,
1009 /* Resolution logic:
1010 * 1. There is a valid state with matching selector. Done.
1011 * 2. Valid state with inappropriate selector. Skip.
1013 * Entering area of "sysdeps".
1015 * 3. If state is not valid, selector is temporary, it selects
1016 * only session which triggered previous resolution. Key
1017 * manager will do something to install a state with proper
1020 if (x->km.state == XFRM_STATE_VALID) {
1021 if ((x->sel.family &&
1022 (x->sel.family != family ||
1023 !xfrm_selector_match(&x->sel, fl, family))) ||
1024 !security_xfrm_state_pol_flow_match(x, pol,
1025 &fl->u.__fl_common))
1029 (*best)->km.dying > x->km.dying ||
1030 ((*best)->km.dying == x->km.dying &&
1031 (*best)->curlft.add_time < x->curlft.add_time))
1033 } else if (x->km.state == XFRM_STATE_ACQ) {
1034 *acq_in_progress = 1;
1035 } else if (x->km.state == XFRM_STATE_ERROR ||
1036 x->km.state == XFRM_STATE_EXPIRED) {
1037 if ((!x->sel.family ||
1038 (x->sel.family == family &&
1039 xfrm_selector_match(&x->sel, fl, family))) &&
1040 security_xfrm_state_pol_flow_match(x, pol,
1041 &fl->u.__fl_common))
1047 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1048 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1049 struct xfrm_policy *pol, int *err,
1050 unsigned short family, u32 if_id)
1052 static xfrm_address_t saddr_wildcard = { };
1053 struct net *net = xp_net(pol);
1054 unsigned int h, h_wildcard;
1055 struct xfrm_state *x, *x0, *to_put;
1056 int acquire_in_progress = 0;
1058 struct xfrm_state *best = NULL;
1059 u32 mark = pol->mark.v & pol->mark.m;
1060 unsigned short encap_family = tmpl->encap_family;
1061 unsigned int sequence;
1066 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
1069 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1070 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1071 if (x->props.family == encap_family &&
1072 x->props.reqid == tmpl->reqid &&
1073 (mark & x->mark.m) == x->mark.v &&
1074 x->if_id == if_id &&
1075 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1076 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1077 tmpl->mode == x->props.mode &&
1078 tmpl->id.proto == x->id.proto &&
1079 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1080 xfrm_state_look_at(pol, x, fl, family,
1081 &best, &acquire_in_progress, &error);
1083 if (best || acquire_in_progress)
1086 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1087 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1088 if (x->props.family == encap_family &&
1089 x->props.reqid == tmpl->reqid &&
1090 (mark & x->mark.m) == x->mark.v &&
1091 x->if_id == if_id &&
1092 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1093 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1094 tmpl->mode == x->props.mode &&
1095 tmpl->id.proto == x->id.proto &&
1096 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1097 xfrm_state_look_at(pol, x, fl, family,
1098 &best, &acquire_in_progress, &error);
1103 if (!x && !error && !acquire_in_progress) {
1105 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1106 tmpl->id.proto, encap_family)) != NULL) {
1113 /* If the KMs have no listeners (yet...), avoid allocating an SA
1114 * for each and every packet - garbage collection might not
1117 if (!km_is_alive(&c)) {
1122 x = xfrm_state_alloc(net);
1127 /* Initialize temporary state matching only
1128 * to current session. */
1129 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1130 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1133 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1135 x->km.state = XFRM_STATE_DEAD;
1141 if (km_query(x, tmpl, pol) == 0) {
1142 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1143 x->km.state = XFRM_STATE_ACQ;
1144 list_add(&x->km.all, &net->xfrm.state_all);
1145 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1146 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1147 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1149 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1150 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1152 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1153 hrtimer_start(&x->mtimer,
1154 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1155 HRTIMER_MODE_REL_SOFT);
1156 net->xfrm.state_num++;
1157 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1158 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1160 x->km.state = XFRM_STATE_DEAD;
1168 if (!xfrm_state_hold_rcu(x)) {
1173 *err = acquire_in_progress ? -EAGAIN : error;
1177 xfrm_state_put(to_put);
1179 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1191 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1192 xfrm_address_t *daddr, xfrm_address_t *saddr,
1193 unsigned short family, u8 mode, u8 proto, u32 reqid)
1196 struct xfrm_state *rx = NULL, *x = NULL;
1198 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1199 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1200 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1201 if (x->props.family == family &&
1202 x->props.reqid == reqid &&
1203 (mark & x->mark.m) == x->mark.v &&
1204 x->if_id == if_id &&
1205 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1206 xfrm_state_addr_check(x, daddr, saddr, family) &&
1207 mode == x->props.mode &&
1208 proto == x->id.proto &&
1209 x->km.state == XFRM_STATE_VALID) {
1216 xfrm_state_hold(rx);
1217 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1222 EXPORT_SYMBOL(xfrm_stateonly_find);
1224 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1225 unsigned short family)
1227 struct xfrm_state *x;
1228 struct xfrm_state_walk *w;
1230 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1231 list_for_each_entry(w, &net->xfrm.state_all, all) {
1232 x = container_of(w, struct xfrm_state, km);
1233 if (x->props.family != family ||
1238 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1241 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1244 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1246 static void __xfrm_state_insert(struct xfrm_state *x)
1248 struct net *net = xs_net(x);
1251 list_add(&x->km.all, &net->xfrm.state_all);
1253 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1254 x->props.reqid, x->props.family);
1255 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1257 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1258 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1261 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1264 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1267 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1268 if (x->replay_maxage)
1269 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1271 net->xfrm.state_num++;
1273 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1276 /* net->xfrm.xfrm_state_lock is held */
1277 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1279 struct net *net = xs_net(xnew);
1280 unsigned short family = xnew->props.family;
1281 u32 reqid = xnew->props.reqid;
1282 struct xfrm_state *x;
1284 u32 mark = xnew->mark.v & xnew->mark.m;
1285 u32 if_id = xnew->if_id;
1287 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1288 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1289 if (x->props.family == family &&
1290 x->props.reqid == reqid &&
1291 x->if_id == if_id &&
1292 (mark & x->mark.m) == x->mark.v &&
1293 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1294 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1299 void xfrm_state_insert(struct xfrm_state *x)
1301 struct net *net = xs_net(x);
1303 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1304 __xfrm_state_bump_genids(x);
1305 __xfrm_state_insert(x);
1306 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1308 EXPORT_SYMBOL(xfrm_state_insert);
1310 /* net->xfrm.xfrm_state_lock is held */
1311 static struct xfrm_state *__find_acq_core(struct net *net,
1312 const struct xfrm_mark *m,
1313 unsigned short family, u8 mode,
1314 u32 reqid, u32 if_id, u8 proto,
1315 const xfrm_address_t *daddr,
1316 const xfrm_address_t *saddr,
1319 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1320 struct xfrm_state *x;
1321 u32 mark = m->v & m->m;
1323 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1324 if (x->props.reqid != reqid ||
1325 x->props.mode != mode ||
1326 x->props.family != family ||
1327 x->km.state != XFRM_STATE_ACQ ||
1329 x->id.proto != proto ||
1330 (mark & x->mark.m) != x->mark.v ||
1331 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1332 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1342 x = xfrm_state_alloc(net);
1346 x->sel.daddr.a4 = daddr->a4;
1347 x->sel.saddr.a4 = saddr->a4;
1348 x->sel.prefixlen_d = 32;
1349 x->sel.prefixlen_s = 32;
1350 x->props.saddr.a4 = saddr->a4;
1351 x->id.daddr.a4 = daddr->a4;
1355 x->sel.daddr.in6 = daddr->in6;
1356 x->sel.saddr.in6 = saddr->in6;
1357 x->sel.prefixlen_d = 128;
1358 x->sel.prefixlen_s = 128;
1359 x->props.saddr.in6 = saddr->in6;
1360 x->id.daddr.in6 = daddr->in6;
1364 x->km.state = XFRM_STATE_ACQ;
1365 x->id.proto = proto;
1366 x->props.family = family;
1367 x->props.mode = mode;
1368 x->props.reqid = reqid;
1372 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1374 hrtimer_start(&x->mtimer,
1375 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1376 HRTIMER_MODE_REL_SOFT);
1377 list_add(&x->km.all, &net->xfrm.state_all);
1378 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1379 h = xfrm_src_hash(net, daddr, saddr, family);
1380 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1382 net->xfrm.state_num++;
1384 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1390 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1392 int xfrm_state_add(struct xfrm_state *x)
1394 struct net *net = xs_net(x);
1395 struct xfrm_state *x1, *to_put;
1398 u32 mark = x->mark.v & x->mark.m;
1399 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1401 family = x->props.family;
1405 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1407 x1 = __xfrm_state_locate(x, use_spi, family);
1415 if (use_spi && x->km.seq) {
1416 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1417 if (x1 && ((x1->id.proto != x->id.proto) ||
1418 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1425 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1426 x->props.reqid, x->if_id, x->id.proto,
1427 &x->id.daddr, &x->props.saddr, 0);
1429 __xfrm_state_bump_genids(x);
1430 __xfrm_state_insert(x);
1434 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1437 xfrm_state_delete(x1);
1442 xfrm_state_put(to_put);
1446 EXPORT_SYMBOL(xfrm_state_add);
1448 #ifdef CONFIG_XFRM_MIGRATE
1449 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1451 struct xfrm_user_sec_ctx *uctx;
1452 int size = sizeof(*uctx) + security->ctx_len;
1455 uctx = kmalloc(size, GFP_KERNEL);
1459 uctx->exttype = XFRMA_SEC_CTX;
1461 uctx->ctx_doi = security->ctx_doi;
1462 uctx->ctx_alg = security->ctx_alg;
1463 uctx->ctx_len = security->ctx_len;
1464 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1465 err = security_xfrm_state_alloc(x, uctx);
1473 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1474 struct xfrm_encap_tmpl *encap)
1476 struct net *net = xs_net(orig);
1477 struct xfrm_state *x = xfrm_state_alloc(net);
1481 memcpy(&x->id, &orig->id, sizeof(x->id));
1482 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1483 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1484 x->props.mode = orig->props.mode;
1485 x->props.replay_window = orig->props.replay_window;
1486 x->props.reqid = orig->props.reqid;
1487 x->props.family = orig->props.family;
1488 x->props.saddr = orig->props.saddr;
1491 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1495 x->props.aalgo = orig->props.aalgo;
1498 x->aead = xfrm_algo_aead_clone(orig->aead);
1499 x->geniv = orig->geniv;
1504 x->ealg = xfrm_algo_clone(orig->ealg);
1508 x->props.ealgo = orig->props.ealgo;
1511 x->calg = xfrm_algo_clone(orig->calg);
1515 x->props.calgo = orig->props.calgo;
1517 if (encap || orig->encap) {
1519 x->encap = kmemdup(encap, sizeof(*x->encap),
1522 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1530 if (clone_security(x, orig->security))
1534 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1540 if (orig->replay_esn) {
1541 if (xfrm_replay_clone(x, orig))
1545 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1546 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1548 if (xfrm_init_state(x) < 0)
1551 x->props.flags = orig->props.flags;
1552 x->props.extra_flags = orig->props.extra_flags;
1554 x->if_id = orig->if_id;
1555 x->tfcpad = orig->tfcpad;
1556 x->replay_maxdiff = orig->replay_maxdiff;
1557 x->replay_maxage = orig->replay_maxage;
1558 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1559 x->km.state = orig->km.state;
1560 x->km.seq = orig->km.seq;
1561 x->replay = orig->replay;
1562 x->preplay = orig->preplay;
1572 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1575 struct xfrm_state *x = NULL;
1577 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1580 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1581 m->reqid, m->old_family);
1582 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1583 if (x->props.mode != m->mode ||
1584 x->id.proto != m->proto)
1586 if (m->reqid && x->props.reqid != m->reqid)
1588 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1590 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1597 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1599 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1600 if (x->props.mode != m->mode ||
1601 x->id.proto != m->proto)
1603 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1605 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1613 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1617 EXPORT_SYMBOL(xfrm_migrate_state_find);
1619 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1620 struct xfrm_migrate *m,
1621 struct xfrm_encap_tmpl *encap)
1623 struct xfrm_state *xc;
1625 xc = xfrm_state_clone(x, encap);
1629 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1630 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1633 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1634 /* a care is needed when the destination address of the
1635 state is to be updated as it is a part of triplet */
1636 xfrm_state_insert(xc);
1638 if (xfrm_state_add(xc) < 0)
1647 EXPORT_SYMBOL(xfrm_state_migrate);
1650 int xfrm_state_update(struct xfrm_state *x)
1652 struct xfrm_state *x1, *to_put;
1654 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1655 struct net *net = xs_net(x);
1659 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1660 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1666 if (xfrm_state_kern(x1)) {
1672 if (x1->km.state == XFRM_STATE_ACQ) {
1673 __xfrm_state_insert(x);
1679 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1682 xfrm_state_put(to_put);
1688 xfrm_state_delete(x1);
1694 spin_lock_bh(&x1->lock);
1695 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1696 if (x->encap && x1->encap &&
1697 x->encap->encap_type == x1->encap->encap_type)
1698 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1699 else if (x->encap || x1->encap)
1702 if (x->coaddr && x1->coaddr) {
1703 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1705 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1706 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1707 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1710 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1711 HRTIMER_MODE_REL_SOFT);
1712 if (x1->curlft.use_time)
1713 xfrm_state_check_expire(x1);
1715 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1716 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1718 if (x->props.smark.m || x->props.smark.v)
1719 x1->props.smark = x->props.smark;
1722 x1->if_id = x->if_id;
1724 __xfrm_state_bump_genids(x1);
1725 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1729 x->km.state = XFRM_STATE_DEAD;
1730 __xfrm_state_put(x);
1734 spin_unlock_bh(&x1->lock);
1740 EXPORT_SYMBOL(xfrm_state_update);
1742 int xfrm_state_check_expire(struct xfrm_state *x)
1744 if (!x->curlft.use_time)
1745 x->curlft.use_time = ktime_get_real_seconds();
1747 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1748 x->curlft.packets >= x->lft.hard_packet_limit) {
1749 x->km.state = XFRM_STATE_EXPIRED;
1750 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1755 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1756 x->curlft.packets >= x->lft.soft_packet_limit)) {
1758 km_state_expired(x, 0, 0);
1762 EXPORT_SYMBOL(xfrm_state_check_expire);
1765 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1766 u8 proto, unsigned short family)
1768 struct xfrm_state *x;
1771 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1775 EXPORT_SYMBOL(xfrm_state_lookup);
1778 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1779 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1780 u8 proto, unsigned short family)
1782 struct xfrm_state *x;
1784 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1785 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1786 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1789 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1792 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1793 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1794 const xfrm_address_t *saddr, int create, unsigned short family)
1796 struct xfrm_state *x;
1798 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1799 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1800 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1804 EXPORT_SYMBOL(xfrm_find_acq);
1806 #ifdef CONFIG_XFRM_SUB_POLICY
1807 #if IS_ENABLED(CONFIG_IPV6)
1808 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1810 __xfrm6_sort(void **dst, void **src, int n,
1811 int (*cmp)(const void *p), int maxclass)
1813 int count[XFRM_MAX_DEPTH] = { };
1814 int class[XFRM_MAX_DEPTH];
1817 for (i = 0; i < n; i++) {
1818 int c = cmp(src[i]);
1824 for (i = 2; i < maxclass; i++)
1825 count[i] += count[i - 1];
1827 for (i = 0; i < n; i++) {
1828 dst[count[class[i] - 1]++] = src[i];
1833 /* Rule for xfrm_state:
1835 * rule 1: select IPsec transport except AH
1836 * rule 2: select MIPv6 RO or inbound trigger
1837 * rule 3: select IPsec transport AH
1838 * rule 4: select IPsec tunnel
1841 static int __xfrm6_state_sort_cmp(const void *p)
1843 const struct xfrm_state *v = p;
1845 switch (v->props.mode) {
1846 case XFRM_MODE_TRANSPORT:
1847 if (v->id.proto != IPPROTO_AH)
1851 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1852 case XFRM_MODE_ROUTEOPTIMIZATION:
1853 case XFRM_MODE_IN_TRIGGER:
1856 case XFRM_MODE_TUNNEL:
1857 case XFRM_MODE_BEET:
1863 /* Rule for xfrm_tmpl:
1865 * rule 1: select IPsec transport
1866 * rule 2: select MIPv6 RO or inbound trigger
1867 * rule 3: select IPsec tunnel
1870 static int __xfrm6_tmpl_sort_cmp(const void *p)
1872 const struct xfrm_tmpl *v = p;
1875 case XFRM_MODE_TRANSPORT:
1877 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1878 case XFRM_MODE_ROUTEOPTIMIZATION:
1879 case XFRM_MODE_IN_TRIGGER:
1882 case XFRM_MODE_TUNNEL:
1883 case XFRM_MODE_BEET:
1889 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1890 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1893 __xfrm6_sort(void **dst, void **src, int n,
1894 int (*cmp)(const void *p), int maxclass)
1898 for (i = 0; i < n; i++)
1901 #endif /* CONFIG_IPV6 */
1904 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1905 unsigned short family)
1909 if (family == AF_INET6)
1910 __xfrm6_sort((void **)dst, (void **)src, n,
1911 __xfrm6_tmpl_sort_cmp, 5);
1913 for (i = 0; i < n; i++)
1918 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1919 unsigned short family)
1923 if (family == AF_INET6)
1924 __xfrm6_sort((void **)dst, (void **)src, n,
1925 __xfrm6_state_sort_cmp, 6);
1927 for (i = 0; i < n; i++)
1932 /* Silly enough, but I'm lazy to build resolution list */
1934 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1938 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1939 struct xfrm_state *x;
1941 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1942 if (x->km.seq == seq &&
1943 (mark & x->mark.m) == x->mark.v &&
1944 x->km.state == XFRM_STATE_ACQ) {
1953 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1955 struct xfrm_state *x;
1957 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1958 x = __xfrm_find_acq_byseq(net, mark, seq);
1959 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1962 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1964 u32 xfrm_get_acqseq(void)
1967 static atomic_t acqseq;
1970 res = atomic_inc_return(&acqseq);
1975 EXPORT_SYMBOL(xfrm_get_acqseq);
1977 int verify_spi_info(u8 proto, u32 min, u32 max)
1985 /* IPCOMP spi is 16-bits. */
1999 EXPORT_SYMBOL(verify_spi_info);
2001 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2003 struct net *net = xs_net(x);
2005 struct xfrm_state *x0;
2007 __be32 minspi = htonl(low);
2008 __be32 maxspi = htonl(high);
2010 u32 mark = x->mark.v & x->mark.m;
2012 spin_lock_bh(&x->lock);
2013 if (x->km.state == XFRM_STATE_DEAD)
2022 if (minspi == maxspi) {
2023 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2031 for (h = 0; h < high-low+1; h++) {
2032 spi = low + prandom_u32()%(high-low+1);
2033 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2035 newspi = htonl(spi);
2042 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2044 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2045 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2046 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2052 spin_unlock_bh(&x->lock);
2056 EXPORT_SYMBOL(xfrm_alloc_spi);
2058 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2059 struct xfrm_address_filter *filter)
2062 if ((filter->family == AF_INET ||
2063 filter->family == AF_INET6) &&
2064 x->props.family != filter->family)
2067 return addr_match(&x->props.saddr, &filter->saddr,
2069 addr_match(&x->id.daddr, &filter->daddr,
2075 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2076 int (*func)(struct xfrm_state *, int, void*),
2079 struct xfrm_state *state;
2080 struct xfrm_state_walk *x;
2083 if (walk->seq != 0 && list_empty(&walk->all))
2086 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2087 if (list_empty(&walk->all))
2088 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2090 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2091 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2092 if (x->state == XFRM_STATE_DEAD)
2094 state = container_of(x, struct xfrm_state, km);
2095 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2097 if (!__xfrm_state_filter_match(state, walk->filter))
2099 err = func(state, walk->seq, data);
2101 list_move_tail(&walk->all, &x->all);
2106 if (walk->seq == 0) {
2110 list_del_init(&walk->all);
2112 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2115 EXPORT_SYMBOL(xfrm_state_walk);
2117 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2118 struct xfrm_address_filter *filter)
2120 INIT_LIST_HEAD(&walk->all);
2121 walk->proto = proto;
2122 walk->state = XFRM_STATE_DEAD;
2124 walk->filter = filter;
2126 EXPORT_SYMBOL(xfrm_state_walk_init);
2128 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2130 kfree(walk->filter);
2132 if (list_empty(&walk->all))
2135 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2136 list_del(&walk->all);
2137 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2139 EXPORT_SYMBOL(xfrm_state_walk_done);
2141 static void xfrm_replay_timer_handler(struct timer_list *t)
2143 struct xfrm_state *x = from_timer(x, t, rtimer);
2145 spin_lock(&x->lock);
2147 if (x->km.state == XFRM_STATE_VALID) {
2148 if (xfrm_aevent_is_on(xs_net(x)))
2149 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2151 x->xflags |= XFRM_TIME_DEFER;
2154 spin_unlock(&x->lock);
2157 static LIST_HEAD(xfrm_km_list);
2159 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2161 struct xfrm_mgr *km;
2164 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2165 if (km->notify_policy)
2166 km->notify_policy(xp, dir, c);
2170 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2172 struct xfrm_mgr *km;
2174 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2180 EXPORT_SYMBOL(km_policy_notify);
2181 EXPORT_SYMBOL(km_state_notify);
2183 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2189 c.event = XFRM_MSG_EXPIRE;
2190 km_state_notify(x, &c);
2193 EXPORT_SYMBOL(km_state_expired);
2195 * We send to all registered managers regardless of failure
2196 * We are happy with one success
2198 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2200 int err = -EINVAL, acqret;
2201 struct xfrm_mgr *km;
2204 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2205 acqret = km->acquire(x, t, pol);
2212 EXPORT_SYMBOL(km_query);
2214 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2217 struct xfrm_mgr *km;
2220 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2221 if (km->new_mapping)
2222 err = km->new_mapping(x, ipaddr, sport);
2229 EXPORT_SYMBOL(km_new_mapping);
2231 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2237 c.event = XFRM_MSG_POLEXPIRE;
2238 km_policy_notify(pol, dir, &c);
2240 EXPORT_SYMBOL(km_policy_expired);
2242 #ifdef CONFIG_XFRM_MIGRATE
2243 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2244 const struct xfrm_migrate *m, int num_migrate,
2245 const struct xfrm_kmaddress *k,
2246 const struct xfrm_encap_tmpl *encap)
2250 struct xfrm_mgr *km;
2253 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2255 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2264 EXPORT_SYMBOL(km_migrate);
2267 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2271 struct xfrm_mgr *km;
2274 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2276 ret = km->report(net, proto, sel, addr);
2284 EXPORT_SYMBOL(km_report);
2286 static bool km_is_alive(const struct km_event *c)
2288 struct xfrm_mgr *km;
2289 bool is_alive = false;
2292 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2293 if (km->is_alive && km->is_alive(c)) {
2303 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2304 static DEFINE_SPINLOCK(xfrm_translator_lock);
2305 static struct xfrm_translator __rcu *xfrm_translator;
2307 struct xfrm_translator *xfrm_get_translator(void)
2309 struct xfrm_translator *xtr;
2312 xtr = rcu_dereference(xfrm_translator);
2315 if (!try_module_get(xtr->owner))
2321 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2323 void xfrm_put_translator(struct xfrm_translator *xtr)
2325 module_put(xtr->owner);
2327 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2329 int xfrm_register_translator(struct xfrm_translator *xtr)
2333 spin_lock_bh(&xfrm_translator_lock);
2334 if (unlikely(xfrm_translator != NULL))
2337 rcu_assign_pointer(xfrm_translator, xtr);
2338 spin_unlock_bh(&xfrm_translator_lock);
2342 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2344 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2348 spin_lock_bh(&xfrm_translator_lock);
2349 if (likely(xfrm_translator != NULL)) {
2350 if (rcu_access_pointer(xfrm_translator) != xtr)
2353 RCU_INIT_POINTER(xfrm_translator, NULL);
2355 spin_unlock_bh(&xfrm_translator_lock);
2360 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2363 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2367 struct xfrm_mgr *km;
2368 struct xfrm_policy *pol = NULL;
2370 if (sockptr_is_null(optval) && !optlen) {
2371 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2372 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2377 if (optlen <= 0 || optlen > PAGE_SIZE)
2380 data = memdup_sockptr(optval, optlen);
2382 return PTR_ERR(data);
2384 if (in_compat_syscall()) {
2385 struct xfrm_translator *xtr = xfrm_get_translator();
2392 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2393 xfrm_put_translator(xtr);
2402 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2403 pol = km->compile_policy(sk, optname, data,
2411 xfrm_sk_policy_insert(sk, err, pol);
2420 EXPORT_SYMBOL(xfrm_user_policy);
2422 static DEFINE_SPINLOCK(xfrm_km_lock);
2424 int xfrm_register_km(struct xfrm_mgr *km)
2426 spin_lock_bh(&xfrm_km_lock);
2427 list_add_tail_rcu(&km->list, &xfrm_km_list);
2428 spin_unlock_bh(&xfrm_km_lock);
2431 EXPORT_SYMBOL(xfrm_register_km);
2433 int xfrm_unregister_km(struct xfrm_mgr *km)
2435 spin_lock_bh(&xfrm_km_lock);
2436 list_del_rcu(&km->list);
2437 spin_unlock_bh(&xfrm_km_lock);
2441 EXPORT_SYMBOL(xfrm_unregister_km);
2443 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2447 if (WARN_ON(afinfo->family >= NPROTO))
2448 return -EAFNOSUPPORT;
2450 spin_lock_bh(&xfrm_state_afinfo_lock);
2451 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2454 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2455 spin_unlock_bh(&xfrm_state_afinfo_lock);
2458 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2460 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2462 int err = 0, family = afinfo->family;
2464 if (WARN_ON(family >= NPROTO))
2465 return -EAFNOSUPPORT;
2467 spin_lock_bh(&xfrm_state_afinfo_lock);
2468 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2469 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2472 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2474 spin_unlock_bh(&xfrm_state_afinfo_lock);
2478 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2480 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2482 if (unlikely(family >= NPROTO))
2485 return rcu_dereference(xfrm_state_afinfo[family]);
2487 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2489 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2491 struct xfrm_state_afinfo *afinfo;
2492 if (unlikely(family >= NPROTO))
2495 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2496 if (unlikely(!afinfo))
2501 void xfrm_flush_gc(void)
2503 flush_work(&xfrm_state_gc_work);
2505 EXPORT_SYMBOL(xfrm_flush_gc);
2507 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2508 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2511 struct xfrm_state *t = x->tunnel;
2513 if (atomic_read(&t->tunnel_users) == 2)
2514 xfrm_state_delete(t);
2515 atomic_dec(&t->tunnel_users);
2516 xfrm_state_put_sync(t);
2520 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2522 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2524 const struct xfrm_type *type = READ_ONCE(x->type);
2525 struct crypto_aead *aead;
2526 u32 blksize, net_adj = 0;
2528 if (x->km.state != XFRM_STATE_VALID ||
2529 !type || type->proto != IPPROTO_ESP)
2530 return mtu - x->props.header_len;
2533 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2535 switch (x->props.mode) {
2536 case XFRM_MODE_TRANSPORT:
2537 case XFRM_MODE_BEET:
2538 if (x->props.family == AF_INET)
2539 net_adj = sizeof(struct iphdr);
2540 else if (x->props.family == AF_INET6)
2541 net_adj = sizeof(struct ipv6hdr);
2543 case XFRM_MODE_TUNNEL:
2550 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2551 net_adj) & ~(blksize - 1)) + net_adj - 2;
2553 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2555 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2557 const struct xfrm_mode *inner_mode;
2558 const struct xfrm_mode *outer_mode;
2559 int family = x->props.family;
2562 if (family == AF_INET &&
2563 xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2564 x->props.flags |= XFRM_STATE_NOPMTUDISC;
2566 err = -EPROTONOSUPPORT;
2568 if (x->sel.family != AF_UNSPEC) {
2569 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2570 if (inner_mode == NULL)
2573 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2574 family != x->sel.family)
2577 x->inner_mode = *inner_mode;
2579 const struct xfrm_mode *inner_mode_iaf;
2580 int iafamily = AF_INET;
2582 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2583 if (inner_mode == NULL)
2586 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2589 x->inner_mode = *inner_mode;
2591 if (x->props.family == AF_INET)
2592 iafamily = AF_INET6;
2594 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2595 if (inner_mode_iaf) {
2596 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2597 x->inner_mode_iaf = *inner_mode_iaf;
2601 x->type = xfrm_get_type(x->id.proto, family);
2602 if (x->type == NULL)
2605 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2607 err = x->type->init_state(x);
2611 outer_mode = xfrm_get_mode(x->props.mode, family);
2613 err = -EPROTONOSUPPORT;
2617 x->outer_mode = *outer_mode;
2619 err = xfrm_init_replay(x);
2628 EXPORT_SYMBOL(__xfrm_init_state);
2630 int xfrm_init_state(struct xfrm_state *x)
2634 err = __xfrm_init_state(x, true, false);
2636 x->km.state = XFRM_STATE_VALID;
2641 EXPORT_SYMBOL(xfrm_init_state);
2643 int __net_init xfrm_state_init(struct net *net)
2647 if (net_eq(net, &init_net))
2648 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2649 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2651 INIT_LIST_HEAD(&net->xfrm.state_all);
2653 sz = sizeof(struct hlist_head) * 8;
2655 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2656 if (!net->xfrm.state_bydst)
2658 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2659 if (!net->xfrm.state_bysrc)
2661 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2662 if (!net->xfrm.state_byspi)
2664 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2666 net->xfrm.state_num = 0;
2667 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2668 spin_lock_init(&net->xfrm.xfrm_state_lock);
2672 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2674 xfrm_hash_free(net->xfrm.state_bydst, sz);
2679 void xfrm_state_fini(struct net *net)
2683 flush_work(&net->xfrm.state_hash_work);
2684 flush_work(&xfrm_state_gc_work);
2685 xfrm_state_flush(net, 0, false, true);
2687 WARN_ON(!list_empty(&net->xfrm.state_all));
2689 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2690 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2691 xfrm_hash_free(net->xfrm.state_byspi, sz);
2692 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2693 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2694 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2695 xfrm_hash_free(net->xfrm.state_bydst, sz);
2698 #ifdef CONFIG_AUDITSYSCALL
2699 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2700 struct audit_buffer *audit_buf)
2702 struct xfrm_sec_ctx *ctx = x->security;
2703 u32 spi = ntohl(x->id.spi);
2706 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2707 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2709 switch (x->props.family) {
2711 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2712 &x->props.saddr.a4, &x->id.daddr.a4);
2715 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2716 x->props.saddr.a6, x->id.daddr.a6);
2720 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2723 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2724 struct audit_buffer *audit_buf)
2726 const struct iphdr *iph4;
2727 const struct ipv6hdr *iph6;
2732 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2733 &iph4->saddr, &iph4->daddr);
2736 iph6 = ipv6_hdr(skb);
2737 audit_log_format(audit_buf,
2738 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2739 &iph6->saddr, &iph6->daddr,
2740 iph6->flow_lbl[0] & 0x0f,
2747 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2749 struct audit_buffer *audit_buf;
2751 audit_buf = xfrm_audit_start("SAD-add");
2752 if (audit_buf == NULL)
2754 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2755 xfrm_audit_helper_sainfo(x, audit_buf);
2756 audit_log_format(audit_buf, " res=%u", result);
2757 audit_log_end(audit_buf);
2759 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2761 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2763 struct audit_buffer *audit_buf;
2765 audit_buf = xfrm_audit_start("SAD-delete");
2766 if (audit_buf == NULL)
2768 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2769 xfrm_audit_helper_sainfo(x, audit_buf);
2770 audit_log_format(audit_buf, " res=%u", result);
2771 audit_log_end(audit_buf);
2773 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2775 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2776 struct sk_buff *skb)
2778 struct audit_buffer *audit_buf;
2781 audit_buf = xfrm_audit_start("SA-replay-overflow");
2782 if (audit_buf == NULL)
2784 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2785 /* don't record the sequence number because it's inherent in this kind
2786 * of audit message */
2787 spi = ntohl(x->id.spi);
2788 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2789 audit_log_end(audit_buf);
2791 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2793 void xfrm_audit_state_replay(struct xfrm_state *x,
2794 struct sk_buff *skb, __be32 net_seq)
2796 struct audit_buffer *audit_buf;
2799 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2800 if (audit_buf == NULL)
2802 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2803 spi = ntohl(x->id.spi);
2804 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2805 spi, spi, ntohl(net_seq));
2806 audit_log_end(audit_buf);
2808 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2810 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2812 struct audit_buffer *audit_buf;
2814 audit_buf = xfrm_audit_start("SA-notfound");
2815 if (audit_buf == NULL)
2817 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2818 audit_log_end(audit_buf);
2820 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2822 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2823 __be32 net_spi, __be32 net_seq)
2825 struct audit_buffer *audit_buf;
2828 audit_buf = xfrm_audit_start("SA-notfound");
2829 if (audit_buf == NULL)
2831 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2832 spi = ntohl(net_spi);
2833 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2834 spi, spi, ntohl(net_seq));
2835 audit_log_end(audit_buf);
2837 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2839 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2840 struct sk_buff *skb, u8 proto)
2842 struct audit_buffer *audit_buf;
2846 audit_buf = xfrm_audit_start("SA-icv-failure");
2847 if (audit_buf == NULL)
2849 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2850 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2851 u32 spi = ntohl(net_spi);
2852 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2853 spi, spi, ntohl(net_seq));
2855 audit_log_end(audit_buf);
2857 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2858 #endif /* CONFIG_AUDITSYSCALL */