6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * YOSHIFUJI Hideaki @USAGI
10 * Split up af-specific functions
11 * Derek Atkins <derek@ihtfp.com>
12 * Add UDP Encapsulation
16 #include <linux/workqueue.h>
18 #include <linux/pfkeyv2.h>
19 #include <linux/ipsec.h>
20 #include <linux/module.h>
21 #include <linux/cache.h>
22 #include <linux/audit.h>
23 #include <linux/uaccess.h>
24 #include <linux/ktime.h>
25 #include <linux/slab.h>
26 #include <linux/interrupt.h>
27 #include <linux/kernel.h>
29 #include "xfrm_hash.h"
31 #define xfrm_state_deref_prot(table, net) \
32 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
34 static void xfrm_state_gc_task(struct work_struct *work);
36 /* Each xfrm_state may be linked to two tables:
38 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
39 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
40 destination/tunnel endpoint. (output)
43 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
44 static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
46 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
47 static HLIST_HEAD(xfrm_state_gc_list);
49 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
51 return refcount_inc_not_zero(&x->refcnt);
54 static inline unsigned int xfrm_dst_hash(struct net *net,
55 const xfrm_address_t *daddr,
56 const xfrm_address_t *saddr,
58 unsigned short family)
60 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
63 static inline unsigned int xfrm_src_hash(struct net *net,
64 const xfrm_address_t *daddr,
65 const xfrm_address_t *saddr,
66 unsigned short family)
68 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
71 static inline unsigned int
72 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
73 __be32 spi, u8 proto, unsigned short family)
75 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
78 static void xfrm_hash_transfer(struct hlist_head *list,
79 struct hlist_head *ndsttable,
80 struct hlist_head *nsrctable,
81 struct hlist_head *nspitable,
82 unsigned int nhashmask)
84 struct hlist_node *tmp;
87 hlist_for_each_entry_safe(x, tmp, list, bydst) {
90 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
91 x->props.reqid, x->props.family,
93 hlist_add_head_rcu(&x->bydst, ndsttable + h);
95 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
98 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
101 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
102 x->id.proto, x->props.family,
104 hlist_add_head_rcu(&x->byspi, nspitable + h);
109 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
111 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
114 static void xfrm_hash_resize(struct work_struct *work)
116 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
117 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
118 unsigned long nsize, osize;
119 unsigned int nhashmask, ohashmask;
122 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
123 ndst = xfrm_hash_alloc(nsize);
126 nsrc = xfrm_hash_alloc(nsize);
128 xfrm_hash_free(ndst, nsize);
131 nspi = xfrm_hash_alloc(nsize);
133 xfrm_hash_free(ndst, nsize);
134 xfrm_hash_free(nsrc, nsize);
138 spin_lock_bh(&net->xfrm.xfrm_state_lock);
139 write_seqcount_begin(&xfrm_state_hash_generation);
141 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
142 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
143 for (i = net->xfrm.state_hmask; i >= 0; i--)
144 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
146 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
147 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
148 ohashmask = net->xfrm.state_hmask;
150 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
151 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
152 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
153 net->xfrm.state_hmask = nhashmask;
155 write_seqcount_end(&xfrm_state_hash_generation);
156 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
158 osize = (ohashmask + 1) * sizeof(struct hlist_head);
162 xfrm_hash_free(odst, osize);
163 xfrm_hash_free(osrc, osize);
164 xfrm_hash_free(ospi, osize);
167 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
168 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
170 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
172 int __xfrm_state_delete(struct xfrm_state *x);
174 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
175 bool km_is_alive(const struct km_event *c);
176 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
178 static DEFINE_SPINLOCK(xfrm_type_lock);
179 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
181 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
182 const struct xfrm_type **typemap;
185 if (unlikely(afinfo == NULL))
186 return -EAFNOSUPPORT;
187 typemap = afinfo->type_map;
188 spin_lock_bh(&xfrm_type_lock);
190 if (likely(typemap[type->proto] == NULL))
191 typemap[type->proto] = type;
194 spin_unlock_bh(&xfrm_type_lock);
198 EXPORT_SYMBOL(xfrm_register_type);
200 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
202 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
203 const struct xfrm_type **typemap;
206 if (unlikely(afinfo == NULL))
207 return -EAFNOSUPPORT;
208 typemap = afinfo->type_map;
209 spin_lock_bh(&xfrm_type_lock);
211 if (unlikely(typemap[type->proto] != type))
214 typemap[type->proto] = NULL;
215 spin_unlock_bh(&xfrm_type_lock);
219 EXPORT_SYMBOL(xfrm_unregister_type);
221 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
223 struct xfrm_state_afinfo *afinfo;
224 const struct xfrm_type **typemap;
225 const struct xfrm_type *type;
226 int modload_attempted = 0;
229 afinfo = xfrm_state_get_afinfo(family);
230 if (unlikely(afinfo == NULL))
232 typemap = afinfo->type_map;
234 type = READ_ONCE(typemap[proto]);
235 if (unlikely(type && !try_module_get(type->owner)))
240 if (!type && !modload_attempted) {
241 request_module("xfrm-type-%d-%d", family, proto);
242 modload_attempted = 1;
249 static void xfrm_put_type(const struct xfrm_type *type)
251 module_put(type->owner);
254 static DEFINE_SPINLOCK(xfrm_type_offload_lock);
255 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
256 unsigned short family)
258 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
259 const struct xfrm_type_offload **typemap;
262 if (unlikely(afinfo == NULL))
263 return -EAFNOSUPPORT;
264 typemap = afinfo->type_offload_map;
265 spin_lock_bh(&xfrm_type_offload_lock);
267 if (likely(typemap[type->proto] == NULL))
268 typemap[type->proto] = type;
271 spin_unlock_bh(&xfrm_type_offload_lock);
275 EXPORT_SYMBOL(xfrm_register_type_offload);
277 int xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
278 unsigned short family)
280 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
281 const struct xfrm_type_offload **typemap;
284 if (unlikely(afinfo == NULL))
285 return -EAFNOSUPPORT;
286 typemap = afinfo->type_offload_map;
287 spin_lock_bh(&xfrm_type_offload_lock);
289 if (unlikely(typemap[type->proto] != type))
292 typemap[type->proto] = NULL;
293 spin_unlock_bh(&xfrm_type_offload_lock);
297 EXPORT_SYMBOL(xfrm_unregister_type_offload);
299 static const struct xfrm_type_offload *
300 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
302 struct xfrm_state_afinfo *afinfo;
303 const struct xfrm_type_offload **typemap;
304 const struct xfrm_type_offload *type;
307 afinfo = xfrm_state_get_afinfo(family);
308 if (unlikely(afinfo == NULL))
310 typemap = afinfo->type_offload_map;
312 type = typemap[proto];
313 if ((type && !try_module_get(type->owner)))
318 if (!type && try_load) {
319 request_module("xfrm-offload-%d-%d", family, proto);
327 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
329 module_put(type->owner);
332 static DEFINE_SPINLOCK(xfrm_mode_lock);
333 int xfrm_register_mode(struct xfrm_mode *mode, int family)
335 struct xfrm_state_afinfo *afinfo;
336 struct xfrm_mode **modemap;
339 if (unlikely(mode->encap >= XFRM_MODE_MAX))
342 afinfo = xfrm_state_get_afinfo(family);
343 if (unlikely(afinfo == NULL))
344 return -EAFNOSUPPORT;
347 modemap = afinfo->mode_map;
348 spin_lock_bh(&xfrm_mode_lock);
349 if (modemap[mode->encap])
353 if (!try_module_get(afinfo->owner))
356 mode->afinfo = afinfo;
357 modemap[mode->encap] = mode;
361 spin_unlock_bh(&xfrm_mode_lock);
365 EXPORT_SYMBOL(xfrm_register_mode);
367 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
369 struct xfrm_state_afinfo *afinfo;
370 struct xfrm_mode **modemap;
373 if (unlikely(mode->encap >= XFRM_MODE_MAX))
376 afinfo = xfrm_state_get_afinfo(family);
377 if (unlikely(afinfo == NULL))
378 return -EAFNOSUPPORT;
381 modemap = afinfo->mode_map;
382 spin_lock_bh(&xfrm_mode_lock);
383 if (likely(modemap[mode->encap] == mode)) {
384 modemap[mode->encap] = NULL;
385 module_put(mode->afinfo->owner);
389 spin_unlock_bh(&xfrm_mode_lock);
393 EXPORT_SYMBOL(xfrm_unregister_mode);
395 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
397 struct xfrm_state_afinfo *afinfo;
398 struct xfrm_mode *mode;
399 int modload_attempted = 0;
401 if (unlikely(encap >= XFRM_MODE_MAX))
405 afinfo = xfrm_state_get_afinfo(family);
406 if (unlikely(afinfo == NULL))
409 mode = READ_ONCE(afinfo->mode_map[encap]);
410 if (unlikely(mode && !try_module_get(mode->owner)))
414 if (!mode && !modload_attempted) {
415 request_module("xfrm-mode-%d-%d", family, encap);
416 modload_attempted = 1;
423 static void xfrm_put_mode(struct xfrm_mode *mode)
425 module_put(mode->owner);
428 static void xfrm_state_gc_destroy(struct xfrm_state *x)
430 tasklet_hrtimer_cancel(&x->mtimer);
431 del_timer_sync(&x->rtimer);
438 kfree(x->replay_esn);
439 kfree(x->preplay_esn);
441 xfrm_put_mode(x->inner_mode);
442 if (x->inner_mode_iaf)
443 xfrm_put_mode(x->inner_mode_iaf);
445 xfrm_put_mode(x->outer_mode);
447 xfrm_put_type_offload(x->type_offload);
449 x->type->destructor(x);
450 xfrm_put_type(x->type);
453 put_page(x->xfrag.page);
454 xfrm_dev_state_free(x);
455 security_xfrm_state_free(x);
459 static void xfrm_state_gc_task(struct work_struct *work)
461 struct xfrm_state *x;
462 struct hlist_node *tmp;
463 struct hlist_head gc_list;
465 spin_lock_bh(&xfrm_state_gc_lock);
466 hlist_move_list(&xfrm_state_gc_list, &gc_list);
467 spin_unlock_bh(&xfrm_state_gc_lock);
471 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
472 xfrm_state_gc_destroy(x);
475 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
477 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
478 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
479 unsigned long now = get_seconds();
480 long next = LONG_MAX;
485 if (x->km.state == XFRM_STATE_DEAD)
487 if (x->km.state == XFRM_STATE_EXPIRED)
489 if (x->lft.hard_add_expires_seconds) {
490 long tmo = x->lft.hard_add_expires_seconds +
491 x->curlft.add_time - now;
493 if (x->xflags & XFRM_SOFT_EXPIRE) {
494 /* enter hard expire without soft expire first?!
495 * setting a new date could trigger this.
496 * workaround: fix x->curflt.add_time by below:
498 x->curlft.add_time = now - x->saved_tmo - 1;
499 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
506 if (x->lft.hard_use_expires_seconds) {
507 long tmo = x->lft.hard_use_expires_seconds +
508 (x->curlft.use_time ? : now) - now;
516 if (x->lft.soft_add_expires_seconds) {
517 long tmo = x->lft.soft_add_expires_seconds +
518 x->curlft.add_time - now;
521 x->xflags &= ~XFRM_SOFT_EXPIRE;
522 } else if (tmo < next) {
524 x->xflags |= XFRM_SOFT_EXPIRE;
528 if (x->lft.soft_use_expires_seconds) {
529 long tmo = x->lft.soft_use_expires_seconds +
530 (x->curlft.use_time ? : now) - now;
539 km_state_expired(x, 0, 0);
541 if (next != LONG_MAX) {
542 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
548 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
549 x->km.state = XFRM_STATE_EXPIRED;
551 err = __xfrm_state_delete(x);
553 km_state_expired(x, 1, 0);
555 xfrm_audit_state_delete(x, err ? 0 : 1, true);
558 spin_unlock(&x->lock);
559 return HRTIMER_NORESTART;
562 static void xfrm_replay_timer_handler(unsigned long data);
564 struct xfrm_state *xfrm_state_alloc(struct net *net)
566 struct xfrm_state *x;
568 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
571 write_pnet(&x->xs_net, net);
572 refcount_set(&x->refcnt, 1);
573 atomic_set(&x->tunnel_users, 0);
574 INIT_LIST_HEAD(&x->km.all);
575 INIT_HLIST_NODE(&x->bydst);
576 INIT_HLIST_NODE(&x->bysrc);
577 INIT_HLIST_NODE(&x->byspi);
578 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
579 CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
580 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
582 x->curlft.add_time = get_seconds();
583 x->lft.soft_byte_limit = XFRM_INF;
584 x->lft.soft_packet_limit = XFRM_INF;
585 x->lft.hard_byte_limit = XFRM_INF;
586 x->lft.hard_packet_limit = XFRM_INF;
587 x->replay_maxage = 0;
588 x->replay_maxdiff = 0;
589 x->inner_mode = NULL;
590 x->inner_mode_iaf = NULL;
591 spin_lock_init(&x->lock);
595 EXPORT_SYMBOL(xfrm_state_alloc);
597 void __xfrm_state_destroy(struct xfrm_state *x)
599 WARN_ON(x->km.state != XFRM_STATE_DEAD);
601 spin_lock_bh(&xfrm_state_gc_lock);
602 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
603 spin_unlock_bh(&xfrm_state_gc_lock);
604 schedule_work(&xfrm_state_gc_work);
606 EXPORT_SYMBOL(__xfrm_state_destroy);
608 int __xfrm_state_delete(struct xfrm_state *x)
610 struct net *net = xs_net(x);
613 if (x->km.state != XFRM_STATE_DEAD) {
614 x->km.state = XFRM_STATE_DEAD;
615 spin_lock(&net->xfrm.xfrm_state_lock);
616 list_del(&x->km.all);
617 hlist_del_rcu(&x->bydst);
618 hlist_del_rcu(&x->bysrc);
620 hlist_del_rcu(&x->byspi);
621 net->xfrm.state_num--;
622 spin_unlock(&net->xfrm.xfrm_state_lock);
624 xfrm_dev_state_delete(x);
626 /* All xfrm_state objects are created by xfrm_state_alloc.
627 * The xfrm_state_alloc call gives a reference, and that
628 * is what we are dropping here.
636 EXPORT_SYMBOL(__xfrm_state_delete);
638 int xfrm_state_delete(struct xfrm_state *x)
642 spin_lock_bh(&x->lock);
643 err = __xfrm_state_delete(x);
644 spin_unlock_bh(&x->lock);
648 EXPORT_SYMBOL(xfrm_state_delete);
650 #ifdef CONFIG_SECURITY_NETWORK_XFRM
652 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
656 for (i = 0; i <= net->xfrm.state_hmask; i++) {
657 struct xfrm_state *x;
659 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
660 if (xfrm_id_proto_match(x->id.proto, proto) &&
661 (err = security_xfrm_state_delete(x)) != 0) {
662 xfrm_audit_state_delete(x, 0, task_valid);
672 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
676 for (i = 0; i <= net->xfrm.state_hmask; i++) {
677 struct xfrm_state *x;
678 struct xfrm_state_offload *xso;
680 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
683 if (xso->dev == dev &&
684 (err = security_xfrm_state_delete(x)) != 0) {
685 xfrm_audit_state_delete(x, 0, task_valid);
695 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
701 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
707 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
709 int i, err = 0, cnt = 0;
711 spin_lock_bh(&net->xfrm.xfrm_state_lock);
712 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
717 for (i = 0; i <= net->xfrm.state_hmask; i++) {
718 struct xfrm_state *x;
720 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
721 if (!xfrm_state_kern(x) &&
722 xfrm_id_proto_match(x->id.proto, proto)) {
724 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
726 err = xfrm_state_delete(x);
727 xfrm_audit_state_delete(x, err ? 0 : 1,
733 spin_lock_bh(&net->xfrm.xfrm_state_lock);
739 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
745 EXPORT_SYMBOL(xfrm_state_flush);
747 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
749 int i, err = 0, cnt = 0;
751 spin_lock_bh(&net->xfrm.xfrm_state_lock);
752 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
757 for (i = 0; i <= net->xfrm.state_hmask; i++) {
758 struct xfrm_state *x;
759 struct xfrm_state_offload *xso;
761 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
764 if (!xfrm_state_kern(x) && xso->dev == dev) {
766 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
768 err = xfrm_state_delete(x);
769 xfrm_audit_state_delete(x, err ? 0 : 1,
775 spin_lock_bh(&net->xfrm.xfrm_state_lock);
784 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
787 EXPORT_SYMBOL(xfrm_dev_state_flush);
789 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
791 spin_lock_bh(&net->xfrm.xfrm_state_lock);
792 si->sadcnt = net->xfrm.state_num;
793 si->sadhcnt = net->xfrm.state_hmask + 1;
794 si->sadhmcnt = xfrm_state_hashmax;
795 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
797 EXPORT_SYMBOL(xfrm_sad_getinfo);
800 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
801 const struct xfrm_tmpl *tmpl,
802 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
803 unsigned short family)
805 struct xfrm_state_afinfo *afinfo = xfrm_state_afinfo_get_rcu(family);
810 afinfo->init_tempsel(&x->sel, fl);
812 if (family != tmpl->encap_family) {
813 afinfo = xfrm_state_afinfo_get_rcu(tmpl->encap_family);
817 afinfo->init_temprop(x, tmpl, daddr, saddr);
820 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
821 const xfrm_address_t *daddr,
822 __be32 spi, u8 proto,
823 unsigned short family)
825 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
826 struct xfrm_state *x;
828 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
829 if (x->props.family != family ||
831 x->id.proto != proto ||
832 !xfrm_addr_equal(&x->id.daddr, daddr, family))
835 if ((mark & x->mark.m) != x->mark.v)
837 if (!xfrm_state_hold_rcu(x))
845 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
846 const xfrm_address_t *daddr,
847 const xfrm_address_t *saddr,
848 u8 proto, unsigned short family)
850 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
851 struct xfrm_state *x;
853 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
854 if (x->props.family != family ||
855 x->id.proto != proto ||
856 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
857 !xfrm_addr_equal(&x->props.saddr, saddr, family))
860 if ((mark & x->mark.m) != x->mark.v)
862 if (!xfrm_state_hold_rcu(x))
870 static inline struct xfrm_state *
871 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
873 struct net *net = xs_net(x);
874 u32 mark = x->mark.v & x->mark.m;
877 return __xfrm_state_lookup(net, mark, &x->id.daddr,
878 x->id.spi, x->id.proto, family);
880 return __xfrm_state_lookup_byaddr(net, mark,
883 x->id.proto, family);
886 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
888 if (have_hash_collision &&
889 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
890 net->xfrm.state_num > net->xfrm.state_hmask)
891 schedule_work(&net->xfrm.state_hash_work);
894 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
895 const struct flowi *fl, unsigned short family,
896 struct xfrm_state **best, int *acq_in_progress,
900 * 1. There is a valid state with matching selector. Done.
901 * 2. Valid state with inappropriate selector. Skip.
903 * Entering area of "sysdeps".
905 * 3. If state is not valid, selector is temporary, it selects
906 * only session which triggered previous resolution. Key
907 * manager will do something to install a state with proper
910 if (x->km.state == XFRM_STATE_VALID) {
911 if ((x->sel.family &&
912 (x->sel.family != family ||
913 !xfrm_selector_match(&x->sel, fl, family))) ||
914 !security_xfrm_state_pol_flow_match(x, pol, fl))
918 (*best)->km.dying > x->km.dying ||
919 ((*best)->km.dying == x->km.dying &&
920 (*best)->curlft.add_time < x->curlft.add_time))
922 } else if (x->km.state == XFRM_STATE_ACQ) {
923 *acq_in_progress = 1;
924 } else if (x->km.state == XFRM_STATE_ERROR ||
925 x->km.state == XFRM_STATE_EXPIRED) {
926 if ((!x->sel.family ||
927 (x->sel.family == family &&
928 xfrm_selector_match(&x->sel, fl, family))) &&
929 security_xfrm_state_pol_flow_match(x, pol, fl))
935 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
936 const struct flowi *fl, struct xfrm_tmpl *tmpl,
937 struct xfrm_policy *pol, int *err,
938 unsigned short family)
940 static xfrm_address_t saddr_wildcard = { };
941 struct net *net = xp_net(pol);
942 unsigned int h, h_wildcard;
943 struct xfrm_state *x, *x0, *to_put;
944 int acquire_in_progress = 0;
946 struct xfrm_state *best = NULL;
947 u32 mark = pol->mark.v & pol->mark.m;
948 unsigned short encap_family = tmpl->encap_family;
949 unsigned int sequence;
954 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
957 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
958 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
959 if (x->props.family == encap_family &&
960 x->props.reqid == tmpl->reqid &&
961 (mark & x->mark.m) == x->mark.v &&
962 !(x->props.flags & XFRM_STATE_WILDRECV) &&
963 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
964 tmpl->mode == x->props.mode &&
965 tmpl->id.proto == x->id.proto &&
966 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
967 xfrm_state_look_at(pol, x, fl, family,
968 &best, &acquire_in_progress, &error);
970 if (best || acquire_in_progress)
973 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
974 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
975 if (x->props.family == encap_family &&
976 x->props.reqid == tmpl->reqid &&
977 (mark & x->mark.m) == x->mark.v &&
978 !(x->props.flags & XFRM_STATE_WILDRECV) &&
979 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
980 tmpl->mode == x->props.mode &&
981 tmpl->id.proto == x->id.proto &&
982 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
983 xfrm_state_look_at(pol, x, fl, family,
984 &best, &acquire_in_progress, &error);
989 if (!x && !error && !acquire_in_progress) {
991 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
992 tmpl->id.proto, encap_family)) != NULL) {
999 /* If the KMs have no listeners (yet...), avoid allocating an SA
1000 * for each and every packet - garbage collection might not
1003 if (!km_is_alive(&c)) {
1008 x = xfrm_state_alloc(net);
1013 /* Initialize temporary state matching only
1014 * to current session. */
1015 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1016 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1018 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1020 x->km.state = XFRM_STATE_DEAD;
1026 if (km_query(x, tmpl, pol) == 0) {
1027 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1028 x->km.state = XFRM_STATE_ACQ;
1029 list_add(&x->km.all, &net->xfrm.state_all);
1030 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1031 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1032 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1034 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1035 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1037 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1038 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1039 net->xfrm.state_num++;
1040 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1041 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1043 x->km.state = XFRM_STATE_DEAD;
1051 if (!xfrm_state_hold_rcu(x)) {
1056 *err = acquire_in_progress ? -EAGAIN : error;
1060 xfrm_state_put(to_put);
1062 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1074 xfrm_stateonly_find(struct net *net, u32 mark,
1075 xfrm_address_t *daddr, xfrm_address_t *saddr,
1076 unsigned short family, u8 mode, u8 proto, u32 reqid)
1079 struct xfrm_state *rx = NULL, *x = NULL;
1081 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1082 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1083 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1084 if (x->props.family == family &&
1085 x->props.reqid == reqid &&
1086 (mark & x->mark.m) == x->mark.v &&
1087 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1088 xfrm_state_addr_check(x, daddr, saddr, family) &&
1089 mode == x->props.mode &&
1090 proto == x->id.proto &&
1091 x->km.state == XFRM_STATE_VALID) {
1098 xfrm_state_hold(rx);
1099 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1104 EXPORT_SYMBOL(xfrm_stateonly_find);
1106 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1107 unsigned short family)
1109 struct xfrm_state *x;
1110 struct xfrm_state_walk *w;
1112 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1113 list_for_each_entry(w, &net->xfrm.state_all, all) {
1114 x = container_of(w, struct xfrm_state, km);
1115 if (x->props.family != family ||
1120 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1123 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1126 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1128 static void __xfrm_state_insert(struct xfrm_state *x)
1130 struct net *net = xs_net(x);
1133 list_add(&x->km.all, &net->xfrm.state_all);
1135 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1136 x->props.reqid, x->props.family);
1137 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1139 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1140 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1143 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1146 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1149 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1150 if (x->replay_maxage)
1151 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1153 net->xfrm.state_num++;
1155 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1158 /* net->xfrm.xfrm_state_lock is held */
1159 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1161 struct net *net = xs_net(xnew);
1162 unsigned short family = xnew->props.family;
1163 u32 reqid = xnew->props.reqid;
1164 struct xfrm_state *x;
1166 u32 mark = xnew->mark.v & xnew->mark.m;
1168 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1169 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1170 if (x->props.family == family &&
1171 x->props.reqid == reqid &&
1172 (mark & x->mark.m) == x->mark.v &&
1173 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1174 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1179 void xfrm_state_insert(struct xfrm_state *x)
1181 struct net *net = xs_net(x);
1183 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1184 __xfrm_state_bump_genids(x);
1185 __xfrm_state_insert(x);
1186 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1188 EXPORT_SYMBOL(xfrm_state_insert);
1190 /* net->xfrm.xfrm_state_lock is held */
1191 static struct xfrm_state *__find_acq_core(struct net *net,
1192 const struct xfrm_mark *m,
1193 unsigned short family, u8 mode,
1194 u32 reqid, u8 proto,
1195 const xfrm_address_t *daddr,
1196 const xfrm_address_t *saddr,
1199 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1200 struct xfrm_state *x;
1201 u32 mark = m->v & m->m;
1203 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1204 if (x->props.reqid != reqid ||
1205 x->props.mode != mode ||
1206 x->props.family != family ||
1207 x->km.state != XFRM_STATE_ACQ ||
1209 x->id.proto != proto ||
1210 (mark & x->mark.m) != x->mark.v ||
1211 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1212 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1222 x = xfrm_state_alloc(net);
1226 x->sel.daddr.a4 = daddr->a4;
1227 x->sel.saddr.a4 = saddr->a4;
1228 x->sel.prefixlen_d = 32;
1229 x->sel.prefixlen_s = 32;
1230 x->props.saddr.a4 = saddr->a4;
1231 x->id.daddr.a4 = daddr->a4;
1235 x->sel.daddr.in6 = daddr->in6;
1236 x->sel.saddr.in6 = saddr->in6;
1237 x->sel.prefixlen_d = 128;
1238 x->sel.prefixlen_s = 128;
1239 x->props.saddr.in6 = saddr->in6;
1240 x->id.daddr.in6 = daddr->in6;
1244 x->km.state = XFRM_STATE_ACQ;
1245 x->id.proto = proto;
1246 x->props.family = family;
1247 x->props.mode = mode;
1248 x->props.reqid = reqid;
1251 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1253 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1254 list_add(&x->km.all, &net->xfrm.state_all);
1255 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1256 h = xfrm_src_hash(net, daddr, saddr, family);
1257 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1259 net->xfrm.state_num++;
1261 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1267 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1269 int xfrm_state_add(struct xfrm_state *x)
1271 struct net *net = xs_net(x);
1272 struct xfrm_state *x1, *to_put;
1275 u32 mark = x->mark.v & x->mark.m;
1276 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1278 family = x->props.family;
1282 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1284 x1 = __xfrm_state_locate(x, use_spi, family);
1292 if (use_spi && x->km.seq) {
1293 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1294 if (x1 && ((x1->id.proto != x->id.proto) ||
1295 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1302 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1303 x->props.reqid, x->id.proto,
1304 &x->id.daddr, &x->props.saddr, 0);
1306 __xfrm_state_bump_genids(x);
1307 __xfrm_state_insert(x);
1311 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1314 xfrm_state_delete(x1);
1319 xfrm_state_put(to_put);
1323 EXPORT_SYMBOL(xfrm_state_add);
1325 #ifdef CONFIG_XFRM_MIGRATE
1326 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1328 struct xfrm_user_sec_ctx *uctx;
1329 int size = sizeof(*uctx) + security->ctx_len;
1332 uctx = kmalloc(size, GFP_KERNEL);
1336 uctx->exttype = XFRMA_SEC_CTX;
1338 uctx->ctx_doi = security->ctx_doi;
1339 uctx->ctx_alg = security->ctx_alg;
1340 uctx->ctx_len = security->ctx_len;
1341 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1342 err = security_xfrm_state_alloc(x, uctx);
1350 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1351 struct xfrm_encap_tmpl *encap)
1353 struct net *net = xs_net(orig);
1354 struct xfrm_state *x = xfrm_state_alloc(net);
1358 memcpy(&x->id, &orig->id, sizeof(x->id));
1359 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1360 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1361 x->props.mode = orig->props.mode;
1362 x->props.replay_window = orig->props.replay_window;
1363 x->props.reqid = orig->props.reqid;
1364 x->props.family = orig->props.family;
1365 x->props.saddr = orig->props.saddr;
1368 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1372 x->props.aalgo = orig->props.aalgo;
1375 x->aead = xfrm_algo_aead_clone(orig->aead);
1376 x->geniv = orig->geniv;
1381 x->ealg = xfrm_algo_clone(orig->ealg);
1385 x->props.ealgo = orig->props.ealgo;
1388 x->calg = xfrm_algo_clone(orig->calg);
1392 x->props.calgo = orig->props.calgo;
1394 if (encap || orig->encap) {
1396 x->encap = kmemdup(encap, sizeof(*x->encap),
1399 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1407 if (clone_security(x, orig->security))
1411 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1417 if (orig->replay_esn) {
1418 if (xfrm_replay_clone(x, orig))
1422 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1424 x->props.flags = orig->props.flags;
1425 x->props.extra_flags = orig->props.extra_flags;
1427 x->tfcpad = orig->tfcpad;
1428 x->replay_maxdiff = orig->replay_maxdiff;
1429 x->replay_maxage = orig->replay_maxage;
1430 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1431 x->km.state = orig->km.state;
1432 x->km.seq = orig->km.seq;
1433 x->replay = orig->replay;
1434 x->preplay = orig->preplay;
1444 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1447 struct xfrm_state *x = NULL;
1449 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1452 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1453 m->reqid, m->old_family);
1454 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1455 if (x->props.mode != m->mode ||
1456 x->id.proto != m->proto)
1458 if (m->reqid && x->props.reqid != m->reqid)
1460 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1462 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1469 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1471 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1472 if (x->props.mode != m->mode ||
1473 x->id.proto != m->proto)
1475 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1477 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1485 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1489 EXPORT_SYMBOL(xfrm_migrate_state_find);
1491 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1492 struct xfrm_migrate *m,
1493 struct xfrm_encap_tmpl *encap)
1495 struct xfrm_state *xc;
1497 xc = xfrm_state_clone(x, encap);
1501 xc->props.family = m->new_family;
1503 if (xfrm_init_state(xc) < 0)
1506 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1507 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1510 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1511 /* a care is needed when the destination address of the
1512 state is to be updated as it is a part of triplet */
1513 xfrm_state_insert(xc);
1515 if (xfrm_state_add(xc) < 0)
1524 EXPORT_SYMBOL(xfrm_state_migrate);
1527 int xfrm_state_update(struct xfrm_state *x)
1529 struct xfrm_state *x1, *to_put;
1531 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1532 struct net *net = xs_net(x);
1536 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1537 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1543 if (xfrm_state_kern(x1)) {
1549 if (x1->km.state == XFRM_STATE_ACQ) {
1550 __xfrm_state_insert(x);
1556 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1559 xfrm_state_put(to_put);
1565 xfrm_state_delete(x1);
1571 spin_lock_bh(&x1->lock);
1572 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1573 if (x->encap && x1->encap)
1574 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1575 if (x->coaddr && x1->coaddr) {
1576 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1578 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1579 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1580 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1583 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1584 if (x1->curlft.use_time)
1585 xfrm_state_check_expire(x1);
1588 x->km.state = XFRM_STATE_DEAD;
1589 __xfrm_state_put(x);
1591 spin_unlock_bh(&x1->lock);
1597 EXPORT_SYMBOL(xfrm_state_update);
1599 int xfrm_state_check_expire(struct xfrm_state *x)
1601 if (!x->curlft.use_time)
1602 x->curlft.use_time = get_seconds();
1604 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1605 x->curlft.packets >= x->lft.hard_packet_limit) {
1606 x->km.state = XFRM_STATE_EXPIRED;
1607 tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1612 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1613 x->curlft.packets >= x->lft.soft_packet_limit)) {
1615 km_state_expired(x, 0, 0);
1619 EXPORT_SYMBOL(xfrm_state_check_expire);
1622 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1623 u8 proto, unsigned short family)
1625 struct xfrm_state *x;
1628 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1632 EXPORT_SYMBOL(xfrm_state_lookup);
1635 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1636 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1637 u8 proto, unsigned short family)
1639 struct xfrm_state *x;
1641 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1642 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1643 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1646 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1649 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1650 u8 proto, const xfrm_address_t *daddr,
1651 const xfrm_address_t *saddr, int create, unsigned short family)
1653 struct xfrm_state *x;
1655 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1656 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1657 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1661 EXPORT_SYMBOL(xfrm_find_acq);
1663 #ifdef CONFIG_XFRM_SUB_POLICY
1665 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1666 unsigned short family, struct net *net)
1670 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1672 return -EAFNOSUPPORT;
1674 spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1675 if (afinfo->tmpl_sort)
1676 err = afinfo->tmpl_sort(dst, src, n);
1678 for (i = 0; i < n; i++)
1680 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1684 EXPORT_SYMBOL(xfrm_tmpl_sort);
1687 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1688 unsigned short family)
1692 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1693 struct net *net = xs_net(*src);
1696 return -EAFNOSUPPORT;
1698 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1699 if (afinfo->state_sort)
1700 err = afinfo->state_sort(dst, src, n);
1702 for (i = 0; i < n; i++)
1704 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1708 EXPORT_SYMBOL(xfrm_state_sort);
1711 /* Silly enough, but I'm lazy to build resolution list */
1713 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1717 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1718 struct xfrm_state *x;
1720 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1721 if (x->km.seq == seq &&
1722 (mark & x->mark.m) == x->mark.v &&
1723 x->km.state == XFRM_STATE_ACQ) {
1732 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1734 struct xfrm_state *x;
1736 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1737 x = __xfrm_find_acq_byseq(net, mark, seq);
1738 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1741 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1743 u32 xfrm_get_acqseq(void)
1746 static atomic_t acqseq;
1749 res = atomic_inc_return(&acqseq);
1754 EXPORT_SYMBOL(xfrm_get_acqseq);
1756 int verify_spi_info(u8 proto, u32 min, u32 max)
1764 /* IPCOMP spi is 16-bits. */
1778 EXPORT_SYMBOL(verify_spi_info);
1780 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1782 struct net *net = xs_net(x);
1784 struct xfrm_state *x0;
1786 __be32 minspi = htonl(low);
1787 __be32 maxspi = htonl(high);
1789 u32 mark = x->mark.v & x->mark.m;
1791 spin_lock_bh(&x->lock);
1792 if (x->km.state == XFRM_STATE_DEAD)
1801 if (minspi == maxspi) {
1802 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1810 for (h = 0; h < high-low+1; h++) {
1811 spi = low + prandom_u32()%(high-low+1);
1812 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1814 newspi = htonl(spi);
1821 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1823 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1824 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1825 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1831 spin_unlock_bh(&x->lock);
1835 EXPORT_SYMBOL(xfrm_alloc_spi);
1837 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1838 struct xfrm_address_filter *filter)
1841 if ((filter->family == AF_INET ||
1842 filter->family == AF_INET6) &&
1843 x->props.family != filter->family)
1846 return addr_match(&x->props.saddr, &filter->saddr,
1848 addr_match(&x->id.daddr, &filter->daddr,
1854 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1855 int (*func)(struct xfrm_state *, int, void*),
1858 struct xfrm_state *state;
1859 struct xfrm_state_walk *x;
1862 if (walk->seq != 0 && list_empty(&walk->all))
1865 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1866 if (list_empty(&walk->all))
1867 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1869 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1870 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1871 if (x->state == XFRM_STATE_DEAD)
1873 state = container_of(x, struct xfrm_state, km);
1874 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1876 if (!__xfrm_state_filter_match(state, walk->filter))
1878 err = func(state, walk->seq, data);
1880 list_move_tail(&walk->all, &x->all);
1885 if (walk->seq == 0) {
1889 list_del_init(&walk->all);
1891 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1894 EXPORT_SYMBOL(xfrm_state_walk);
1896 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1897 struct xfrm_address_filter *filter)
1899 INIT_LIST_HEAD(&walk->all);
1900 walk->proto = proto;
1901 walk->state = XFRM_STATE_DEAD;
1903 walk->filter = filter;
1905 EXPORT_SYMBOL(xfrm_state_walk_init);
1907 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1909 kfree(walk->filter);
1911 if (list_empty(&walk->all))
1914 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1915 list_del(&walk->all);
1916 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1918 EXPORT_SYMBOL(xfrm_state_walk_done);
1920 static void xfrm_replay_timer_handler(unsigned long data)
1922 struct xfrm_state *x = (struct xfrm_state *)data;
1924 spin_lock(&x->lock);
1926 if (x->km.state == XFRM_STATE_VALID) {
1927 if (xfrm_aevent_is_on(xs_net(x)))
1928 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1930 x->xflags |= XFRM_TIME_DEFER;
1933 spin_unlock(&x->lock);
1936 static LIST_HEAD(xfrm_km_list);
1938 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1940 struct xfrm_mgr *km;
1943 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1944 if (km->notify_policy)
1945 km->notify_policy(xp, dir, c);
1949 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1951 struct xfrm_mgr *km;
1953 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1959 EXPORT_SYMBOL(km_policy_notify);
1960 EXPORT_SYMBOL(km_state_notify);
1962 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1968 c.event = XFRM_MSG_EXPIRE;
1969 km_state_notify(x, &c);
1972 EXPORT_SYMBOL(km_state_expired);
1974 * We send to all registered managers regardless of failure
1975 * We are happy with one success
1977 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1979 int err = -EINVAL, acqret;
1980 struct xfrm_mgr *km;
1983 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1984 acqret = km->acquire(x, t, pol);
1991 EXPORT_SYMBOL(km_query);
1993 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1996 struct xfrm_mgr *km;
1999 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2000 if (km->new_mapping)
2001 err = km->new_mapping(x, ipaddr, sport);
2008 EXPORT_SYMBOL(km_new_mapping);
2010 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2016 c.event = XFRM_MSG_POLEXPIRE;
2017 km_policy_notify(pol, dir, &c);
2019 EXPORT_SYMBOL(km_policy_expired);
2021 #ifdef CONFIG_XFRM_MIGRATE
2022 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2023 const struct xfrm_migrate *m, int num_migrate,
2024 const struct xfrm_kmaddress *k,
2025 const struct xfrm_encap_tmpl *encap)
2029 struct xfrm_mgr *km;
2032 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2034 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2043 EXPORT_SYMBOL(km_migrate);
2046 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2050 struct xfrm_mgr *km;
2053 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2055 ret = km->report(net, proto, sel, addr);
2063 EXPORT_SYMBOL(km_report);
2065 bool km_is_alive(const struct km_event *c)
2067 struct xfrm_mgr *km;
2068 bool is_alive = false;
2071 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2072 if (km->is_alive && km->is_alive(c)) {
2081 EXPORT_SYMBOL(km_is_alive);
2083 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2087 struct xfrm_mgr *km;
2088 struct xfrm_policy *pol = NULL;
2090 #ifdef CONFIG_COMPAT
2091 if (in_compat_syscall())
2095 if (!optval && !optlen) {
2096 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2097 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2102 if (optlen <= 0 || optlen > PAGE_SIZE)
2105 data = memdup_user(optval, optlen);
2107 return PTR_ERR(data);
2111 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2112 pol = km->compile_policy(sk, optname, data,
2120 xfrm_sk_policy_insert(sk, err, pol);
2129 EXPORT_SYMBOL(xfrm_user_policy);
2131 static DEFINE_SPINLOCK(xfrm_km_lock);
2133 int xfrm_register_km(struct xfrm_mgr *km)
2135 spin_lock_bh(&xfrm_km_lock);
2136 list_add_tail_rcu(&km->list, &xfrm_km_list);
2137 spin_unlock_bh(&xfrm_km_lock);
2140 EXPORT_SYMBOL(xfrm_register_km);
2142 int xfrm_unregister_km(struct xfrm_mgr *km)
2144 spin_lock_bh(&xfrm_km_lock);
2145 list_del_rcu(&km->list);
2146 spin_unlock_bh(&xfrm_km_lock);
2150 EXPORT_SYMBOL(xfrm_unregister_km);
2152 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2156 if (WARN_ON(afinfo->family >= NPROTO))
2157 return -EAFNOSUPPORT;
2159 spin_lock_bh(&xfrm_state_afinfo_lock);
2160 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2163 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2164 spin_unlock_bh(&xfrm_state_afinfo_lock);
2167 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2169 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2171 int err = 0, family = afinfo->family;
2173 if (WARN_ON(family >= NPROTO))
2174 return -EAFNOSUPPORT;
2176 spin_lock_bh(&xfrm_state_afinfo_lock);
2177 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2178 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2181 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2183 spin_unlock_bh(&xfrm_state_afinfo_lock);
2187 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2189 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2191 if (unlikely(family >= NPROTO))
2194 return rcu_dereference(xfrm_state_afinfo[family]);
2197 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2199 struct xfrm_state_afinfo *afinfo;
2200 if (unlikely(family >= NPROTO))
2203 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2204 if (unlikely(!afinfo))
2209 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2210 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2213 struct xfrm_state *t = x->tunnel;
2215 if (atomic_read(&t->tunnel_users) == 2)
2216 xfrm_state_delete(t);
2217 atomic_dec(&t->tunnel_users);
2222 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2224 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2226 const struct xfrm_type *type = READ_ONCE(x->type);
2228 if (x->km.state == XFRM_STATE_VALID &&
2229 type && type->get_mtu)
2230 return type->get_mtu(x, mtu);
2232 return mtu - x->props.header_len;
2235 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2237 struct xfrm_state_afinfo *afinfo;
2238 struct xfrm_mode *inner_mode;
2239 int family = x->props.family;
2242 err = -EAFNOSUPPORT;
2243 afinfo = xfrm_state_get_afinfo(family);
2248 if (afinfo->init_flags)
2249 err = afinfo->init_flags(x);
2256 err = -EPROTONOSUPPORT;
2258 if (x->sel.family != AF_UNSPEC) {
2259 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2260 if (inner_mode == NULL)
2263 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2264 family != x->sel.family) {
2265 xfrm_put_mode(inner_mode);
2269 x->inner_mode = inner_mode;
2271 struct xfrm_mode *inner_mode_iaf;
2272 int iafamily = AF_INET;
2274 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2275 if (inner_mode == NULL)
2278 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2279 xfrm_put_mode(inner_mode);
2282 x->inner_mode = inner_mode;
2284 if (x->props.family == AF_INET)
2285 iafamily = AF_INET6;
2287 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2288 if (inner_mode_iaf) {
2289 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2290 x->inner_mode_iaf = inner_mode_iaf;
2292 xfrm_put_mode(inner_mode_iaf);
2296 x->type = xfrm_get_type(x->id.proto, family);
2297 if (x->type == NULL)
2300 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2302 err = x->type->init_state(x);
2306 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2307 if (x->outer_mode == NULL) {
2308 err = -EPROTONOSUPPORT;
2313 err = xfrm_init_replay(x);
2318 x->km.state = XFRM_STATE_VALID;
2324 EXPORT_SYMBOL(__xfrm_init_state);
2326 int xfrm_init_state(struct xfrm_state *x)
2328 return __xfrm_init_state(x, true, false);
2331 EXPORT_SYMBOL(xfrm_init_state);
2333 int __net_init xfrm_state_init(struct net *net)
2337 INIT_LIST_HEAD(&net->xfrm.state_all);
2339 sz = sizeof(struct hlist_head) * 8;
2341 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2342 if (!net->xfrm.state_bydst)
2344 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2345 if (!net->xfrm.state_bysrc)
2347 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2348 if (!net->xfrm.state_byspi)
2350 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2352 net->xfrm.state_num = 0;
2353 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2354 spin_lock_init(&net->xfrm.xfrm_state_lock);
2358 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2360 xfrm_hash_free(net->xfrm.state_bydst, sz);
2365 void xfrm_state_fini(struct net *net)
2369 flush_work(&net->xfrm.state_hash_work);
2370 xfrm_state_flush(net, 0, false);
2371 flush_work(&xfrm_state_gc_work);
2373 WARN_ON(!list_empty(&net->xfrm.state_all));
2375 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2376 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2377 xfrm_hash_free(net->xfrm.state_byspi, sz);
2378 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2379 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2380 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2381 xfrm_hash_free(net->xfrm.state_bydst, sz);
2384 #ifdef CONFIG_AUDITSYSCALL
2385 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2386 struct audit_buffer *audit_buf)
2388 struct xfrm_sec_ctx *ctx = x->security;
2389 u32 spi = ntohl(x->id.spi);
2392 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2393 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2395 switch (x->props.family) {
2397 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2398 &x->props.saddr.a4, &x->id.daddr.a4);
2401 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2402 x->props.saddr.a6, x->id.daddr.a6);
2406 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2409 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2410 struct audit_buffer *audit_buf)
2412 const struct iphdr *iph4;
2413 const struct ipv6hdr *iph6;
2418 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2419 &iph4->saddr, &iph4->daddr);
2422 iph6 = ipv6_hdr(skb);
2423 audit_log_format(audit_buf,
2424 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2425 &iph6->saddr, &iph6->daddr,
2426 iph6->flow_lbl[0] & 0x0f,
2433 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2435 struct audit_buffer *audit_buf;
2437 audit_buf = xfrm_audit_start("SAD-add");
2438 if (audit_buf == NULL)
2440 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2441 xfrm_audit_helper_sainfo(x, audit_buf);
2442 audit_log_format(audit_buf, " res=%u", result);
2443 audit_log_end(audit_buf);
2445 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2447 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2449 struct audit_buffer *audit_buf;
2451 audit_buf = xfrm_audit_start("SAD-delete");
2452 if (audit_buf == NULL)
2454 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2455 xfrm_audit_helper_sainfo(x, audit_buf);
2456 audit_log_format(audit_buf, " res=%u", result);
2457 audit_log_end(audit_buf);
2459 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2461 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2462 struct sk_buff *skb)
2464 struct audit_buffer *audit_buf;
2467 audit_buf = xfrm_audit_start("SA-replay-overflow");
2468 if (audit_buf == NULL)
2470 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2471 /* don't record the sequence number because it's inherent in this kind
2472 * of audit message */
2473 spi = ntohl(x->id.spi);
2474 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2475 audit_log_end(audit_buf);
2477 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2479 void xfrm_audit_state_replay(struct xfrm_state *x,
2480 struct sk_buff *skb, __be32 net_seq)
2482 struct audit_buffer *audit_buf;
2485 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2486 if (audit_buf == NULL)
2488 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2489 spi = ntohl(x->id.spi);
2490 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2491 spi, spi, ntohl(net_seq));
2492 audit_log_end(audit_buf);
2494 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2496 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2498 struct audit_buffer *audit_buf;
2500 audit_buf = xfrm_audit_start("SA-notfound");
2501 if (audit_buf == NULL)
2503 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2504 audit_log_end(audit_buf);
2506 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2508 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2509 __be32 net_spi, __be32 net_seq)
2511 struct audit_buffer *audit_buf;
2514 audit_buf = xfrm_audit_start("SA-notfound");
2515 if (audit_buf == NULL)
2517 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2518 spi = ntohl(net_spi);
2519 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2520 spi, spi, ntohl(net_seq));
2521 audit_log_end(audit_buf);
2523 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2525 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2526 struct sk_buff *skb, u8 proto)
2528 struct audit_buffer *audit_buf;
2532 audit_buf = xfrm_audit_start("SA-icv-failure");
2533 if (audit_buf == NULL)
2535 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2536 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2537 u32 spi = ntohl(net_spi);
2538 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2539 spi, spi, ntohl(net_seq));
2541 audit_log_end(audit_buf);
2543 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2544 #endif /* CONFIG_AUDITSYSCALL */