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 atomic_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_mode_lock);
255 int xfrm_register_mode(struct xfrm_mode *mode, int family)
257 struct xfrm_state_afinfo *afinfo;
258 struct xfrm_mode **modemap;
261 if (unlikely(mode->encap >= XFRM_MODE_MAX))
264 afinfo = xfrm_state_get_afinfo(family);
265 if (unlikely(afinfo == NULL))
266 return -EAFNOSUPPORT;
269 modemap = afinfo->mode_map;
270 spin_lock_bh(&xfrm_mode_lock);
271 if (modemap[mode->encap])
275 if (!try_module_get(afinfo->owner))
278 mode->afinfo = afinfo;
279 modemap[mode->encap] = mode;
283 spin_unlock_bh(&xfrm_mode_lock);
287 EXPORT_SYMBOL(xfrm_register_mode);
289 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
291 struct xfrm_state_afinfo *afinfo;
292 struct xfrm_mode **modemap;
295 if (unlikely(mode->encap >= XFRM_MODE_MAX))
298 afinfo = xfrm_state_get_afinfo(family);
299 if (unlikely(afinfo == NULL))
300 return -EAFNOSUPPORT;
303 modemap = afinfo->mode_map;
304 spin_lock_bh(&xfrm_mode_lock);
305 if (likely(modemap[mode->encap] == mode)) {
306 modemap[mode->encap] = NULL;
307 module_put(mode->afinfo->owner);
311 spin_unlock_bh(&xfrm_mode_lock);
315 EXPORT_SYMBOL(xfrm_unregister_mode);
317 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
319 struct xfrm_state_afinfo *afinfo;
320 struct xfrm_mode *mode;
321 int modload_attempted = 0;
323 if (unlikely(encap >= XFRM_MODE_MAX))
327 afinfo = xfrm_state_get_afinfo(family);
328 if (unlikely(afinfo == NULL))
331 mode = READ_ONCE(afinfo->mode_map[encap]);
332 if (unlikely(mode && !try_module_get(mode->owner)))
336 if (!mode && !modload_attempted) {
337 request_module("xfrm-mode-%d-%d", family, encap);
338 modload_attempted = 1;
345 static void xfrm_put_mode(struct xfrm_mode *mode)
347 module_put(mode->owner);
350 static void xfrm_state_gc_destroy(struct xfrm_state *x)
352 tasklet_hrtimer_cancel(&x->mtimer);
353 del_timer_sync(&x->rtimer);
360 kfree(x->replay_esn);
361 kfree(x->preplay_esn);
363 xfrm_put_mode(x->inner_mode);
364 if (x->inner_mode_iaf)
365 xfrm_put_mode(x->inner_mode_iaf);
367 xfrm_put_mode(x->outer_mode);
369 x->type->destructor(x);
370 xfrm_put_type(x->type);
372 security_xfrm_state_free(x);
376 static void xfrm_state_gc_task(struct work_struct *work)
378 struct xfrm_state *x;
379 struct hlist_node *tmp;
380 struct hlist_head gc_list;
382 spin_lock_bh(&xfrm_state_gc_lock);
383 hlist_move_list(&xfrm_state_gc_list, &gc_list);
384 spin_unlock_bh(&xfrm_state_gc_lock);
388 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
389 xfrm_state_gc_destroy(x);
392 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
394 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
395 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
396 unsigned long now = get_seconds();
397 long next = LONG_MAX;
402 if (x->km.state == XFRM_STATE_DEAD)
404 if (x->km.state == XFRM_STATE_EXPIRED)
406 if (x->lft.hard_add_expires_seconds) {
407 long tmo = x->lft.hard_add_expires_seconds +
408 x->curlft.add_time - now;
410 if (x->xflags & XFRM_SOFT_EXPIRE) {
411 /* enter hard expire without soft expire first?!
412 * setting a new date could trigger this.
413 * workaround: fix x->curflt.add_time by below:
415 x->curlft.add_time = now - x->saved_tmo - 1;
416 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
423 if (x->lft.hard_use_expires_seconds) {
424 long tmo = x->lft.hard_use_expires_seconds +
425 (x->curlft.use_time ? : now) - now;
433 if (x->lft.soft_add_expires_seconds) {
434 long tmo = x->lft.soft_add_expires_seconds +
435 x->curlft.add_time - now;
438 x->xflags &= ~XFRM_SOFT_EXPIRE;
439 } else if (tmo < next) {
441 x->xflags |= XFRM_SOFT_EXPIRE;
445 if (x->lft.soft_use_expires_seconds) {
446 long tmo = x->lft.soft_use_expires_seconds +
447 (x->curlft.use_time ? : now) - now;
456 km_state_expired(x, 0, 0);
458 if (next != LONG_MAX) {
459 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
465 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
466 x->km.state = XFRM_STATE_EXPIRED;
468 err = __xfrm_state_delete(x);
470 km_state_expired(x, 1, 0);
472 xfrm_audit_state_delete(x, err ? 0 : 1, true);
475 spin_unlock(&x->lock);
476 return HRTIMER_NORESTART;
479 static void xfrm_replay_timer_handler(unsigned long data);
481 struct xfrm_state *xfrm_state_alloc(struct net *net)
483 struct xfrm_state *x;
485 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
488 write_pnet(&x->xs_net, net);
489 atomic_set(&x->refcnt, 1);
490 atomic_set(&x->tunnel_users, 0);
491 INIT_LIST_HEAD(&x->km.all);
492 INIT_HLIST_NODE(&x->bydst);
493 INIT_HLIST_NODE(&x->bysrc);
494 INIT_HLIST_NODE(&x->byspi);
495 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
496 CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
497 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
499 x->curlft.add_time = get_seconds();
500 x->lft.soft_byte_limit = XFRM_INF;
501 x->lft.soft_packet_limit = XFRM_INF;
502 x->lft.hard_byte_limit = XFRM_INF;
503 x->lft.hard_packet_limit = XFRM_INF;
504 x->replay_maxage = 0;
505 x->replay_maxdiff = 0;
506 x->inner_mode = NULL;
507 x->inner_mode_iaf = NULL;
508 spin_lock_init(&x->lock);
512 EXPORT_SYMBOL(xfrm_state_alloc);
514 void __xfrm_state_destroy(struct xfrm_state *x)
516 WARN_ON(x->km.state != XFRM_STATE_DEAD);
518 spin_lock_bh(&xfrm_state_gc_lock);
519 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
520 spin_unlock_bh(&xfrm_state_gc_lock);
521 schedule_work(&xfrm_state_gc_work);
523 EXPORT_SYMBOL(__xfrm_state_destroy);
525 int __xfrm_state_delete(struct xfrm_state *x)
527 struct net *net = xs_net(x);
530 if (x->km.state != XFRM_STATE_DEAD) {
531 x->km.state = XFRM_STATE_DEAD;
532 spin_lock(&net->xfrm.xfrm_state_lock);
533 list_del(&x->km.all);
534 hlist_del_rcu(&x->bydst);
535 hlist_del_rcu(&x->bysrc);
537 hlist_del_rcu(&x->byspi);
538 net->xfrm.state_num--;
539 spin_unlock(&net->xfrm.xfrm_state_lock);
541 /* All xfrm_state objects are created by xfrm_state_alloc.
542 * The xfrm_state_alloc call gives a reference, and that
543 * is what we are dropping here.
551 EXPORT_SYMBOL(__xfrm_state_delete);
553 int xfrm_state_delete(struct xfrm_state *x)
557 spin_lock_bh(&x->lock);
558 err = __xfrm_state_delete(x);
559 spin_unlock_bh(&x->lock);
563 EXPORT_SYMBOL(xfrm_state_delete);
565 #ifdef CONFIG_SECURITY_NETWORK_XFRM
567 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
571 for (i = 0; i <= net->xfrm.state_hmask; i++) {
572 struct xfrm_state *x;
574 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
575 if (xfrm_id_proto_match(x->id.proto, proto) &&
576 (err = security_xfrm_state_delete(x)) != 0) {
577 xfrm_audit_state_delete(x, 0, task_valid);
587 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
593 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
595 int i, err = 0, cnt = 0;
597 spin_lock_bh(&net->xfrm.xfrm_state_lock);
598 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
603 for (i = 0; i <= net->xfrm.state_hmask; i++) {
604 struct xfrm_state *x;
606 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
607 if (!xfrm_state_kern(x) &&
608 xfrm_id_proto_match(x->id.proto, proto)) {
610 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
612 err = xfrm_state_delete(x);
613 xfrm_audit_state_delete(x, err ? 0 : 1,
619 spin_lock_bh(&net->xfrm.xfrm_state_lock);
628 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
631 EXPORT_SYMBOL(xfrm_state_flush);
633 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
635 spin_lock_bh(&net->xfrm.xfrm_state_lock);
636 si->sadcnt = net->xfrm.state_num;
637 si->sadhcnt = net->xfrm.state_hmask;
638 si->sadhmcnt = xfrm_state_hashmax;
639 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
641 EXPORT_SYMBOL(xfrm_sad_getinfo);
644 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
645 const struct xfrm_tmpl *tmpl,
646 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
647 unsigned short family)
649 struct xfrm_state_afinfo *afinfo = xfrm_state_afinfo_get_rcu(family);
654 afinfo->init_tempsel(&x->sel, fl);
656 if (family != tmpl->encap_family) {
657 afinfo = xfrm_state_afinfo_get_rcu(tmpl->encap_family);
661 afinfo->init_temprop(x, tmpl, daddr, saddr);
664 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
665 const xfrm_address_t *daddr,
666 __be32 spi, u8 proto,
667 unsigned short family)
669 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
670 struct xfrm_state *x;
672 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
673 if (x->props.family != family ||
675 x->id.proto != proto ||
676 !xfrm_addr_equal(&x->id.daddr, daddr, family))
679 if ((mark & x->mark.m) != x->mark.v)
681 if (!xfrm_state_hold_rcu(x))
689 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
690 const xfrm_address_t *daddr,
691 const xfrm_address_t *saddr,
692 u8 proto, unsigned short family)
694 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
695 struct xfrm_state *x;
697 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
698 if (x->props.family != family ||
699 x->id.proto != proto ||
700 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
701 !xfrm_addr_equal(&x->props.saddr, saddr, family))
704 if ((mark & x->mark.m) != x->mark.v)
706 if (!xfrm_state_hold_rcu(x))
714 static inline struct xfrm_state *
715 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
717 struct net *net = xs_net(x);
718 u32 mark = x->mark.v & x->mark.m;
721 return __xfrm_state_lookup(net, mark, &x->id.daddr,
722 x->id.spi, x->id.proto, family);
724 return __xfrm_state_lookup_byaddr(net, mark,
727 x->id.proto, family);
730 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
732 if (have_hash_collision &&
733 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
734 net->xfrm.state_num > net->xfrm.state_hmask)
735 schedule_work(&net->xfrm.state_hash_work);
738 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
739 const struct flowi *fl, unsigned short family,
740 struct xfrm_state **best, int *acq_in_progress,
744 * 1. There is a valid state with matching selector. Done.
745 * 2. Valid state with inappropriate selector. Skip.
747 * Entering area of "sysdeps".
749 * 3. If state is not valid, selector is temporary, it selects
750 * only session which triggered previous resolution. Key
751 * manager will do something to install a state with proper
754 if (x->km.state == XFRM_STATE_VALID) {
755 if ((x->sel.family &&
756 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
757 !security_xfrm_state_pol_flow_match(x, pol, fl))
761 (*best)->km.dying > x->km.dying ||
762 ((*best)->km.dying == x->km.dying &&
763 (*best)->curlft.add_time < x->curlft.add_time))
765 } else if (x->km.state == XFRM_STATE_ACQ) {
766 *acq_in_progress = 1;
767 } else if (x->km.state == XFRM_STATE_ERROR ||
768 x->km.state == XFRM_STATE_EXPIRED) {
769 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
770 security_xfrm_state_pol_flow_match(x, pol, fl))
776 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
777 const struct flowi *fl, struct xfrm_tmpl *tmpl,
778 struct xfrm_policy *pol, int *err,
779 unsigned short family)
781 static xfrm_address_t saddr_wildcard = { };
782 struct net *net = xp_net(pol);
783 unsigned int h, h_wildcard;
784 struct xfrm_state *x, *x0, *to_put;
785 int acquire_in_progress = 0;
787 struct xfrm_state *best = NULL;
788 u32 mark = pol->mark.v & pol->mark.m;
789 unsigned short encap_family = tmpl->encap_family;
790 unsigned int sequence;
795 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
798 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
799 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
800 if (x->props.family == encap_family &&
801 x->props.reqid == tmpl->reqid &&
802 (mark & x->mark.m) == x->mark.v &&
803 !(x->props.flags & XFRM_STATE_WILDRECV) &&
804 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
805 tmpl->mode == x->props.mode &&
806 tmpl->id.proto == x->id.proto &&
807 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
808 xfrm_state_look_at(pol, x, fl, encap_family,
809 &best, &acquire_in_progress, &error);
811 if (best || acquire_in_progress)
814 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
815 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
816 if (x->props.family == encap_family &&
817 x->props.reqid == tmpl->reqid &&
818 (mark & x->mark.m) == x->mark.v &&
819 !(x->props.flags & XFRM_STATE_WILDRECV) &&
820 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
821 tmpl->mode == x->props.mode &&
822 tmpl->id.proto == x->id.proto &&
823 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
824 xfrm_state_look_at(pol, x, fl, encap_family,
825 &best, &acquire_in_progress, &error);
830 if (!x && !error && !acquire_in_progress) {
832 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
833 tmpl->id.proto, encap_family)) != NULL) {
840 /* If the KMs have no listeners (yet...), avoid allocating an SA
841 * for each and every packet - garbage collection might not
844 if (!km_is_alive(&c)) {
849 x = xfrm_state_alloc(net);
854 /* Initialize temporary state matching only
855 * to current session. */
856 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
857 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
859 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
861 x->km.state = XFRM_STATE_DEAD;
867 if (km_query(x, tmpl, pol) == 0) {
868 spin_lock_bh(&net->xfrm.xfrm_state_lock);
869 x->km.state = XFRM_STATE_ACQ;
870 list_add(&x->km.all, &net->xfrm.state_all);
871 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
872 h = xfrm_src_hash(net, daddr, saddr, encap_family);
873 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
875 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
876 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
878 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
879 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
880 net->xfrm.state_num++;
881 xfrm_hash_grow_check(net, x->bydst.next != NULL);
882 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
884 x->km.state = XFRM_STATE_DEAD;
892 if (!xfrm_state_hold_rcu(x)) {
897 *err = acquire_in_progress ? -EAGAIN : error;
901 xfrm_state_put(to_put);
903 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
915 xfrm_stateonly_find(struct net *net, u32 mark,
916 xfrm_address_t *daddr, xfrm_address_t *saddr,
917 unsigned short family, u8 mode, u8 proto, u32 reqid)
920 struct xfrm_state *rx = NULL, *x = NULL;
922 spin_lock_bh(&net->xfrm.xfrm_state_lock);
923 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
924 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
925 if (x->props.family == family &&
926 x->props.reqid == reqid &&
927 (mark & x->mark.m) == x->mark.v &&
928 !(x->props.flags & XFRM_STATE_WILDRECV) &&
929 xfrm_state_addr_check(x, daddr, saddr, family) &&
930 mode == x->props.mode &&
931 proto == x->id.proto &&
932 x->km.state == XFRM_STATE_VALID) {
940 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
945 EXPORT_SYMBOL(xfrm_stateonly_find);
947 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
948 unsigned short family)
950 struct xfrm_state *x;
951 struct xfrm_state_walk *w;
953 spin_lock_bh(&net->xfrm.xfrm_state_lock);
954 list_for_each_entry(w, &net->xfrm.state_all, all) {
955 x = container_of(w, struct xfrm_state, km);
956 if (x->props.family != family ||
961 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
964 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
967 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
969 static void __xfrm_state_insert(struct xfrm_state *x)
971 struct net *net = xs_net(x);
974 list_add(&x->km.all, &net->xfrm.state_all);
976 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
977 x->props.reqid, x->props.family);
978 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
980 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
981 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
984 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
987 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
990 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
991 if (x->replay_maxage)
992 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
994 net->xfrm.state_num++;
996 xfrm_hash_grow_check(net, x->bydst.next != NULL);
999 /* net->xfrm.xfrm_state_lock is held */
1000 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1002 struct net *net = xs_net(xnew);
1003 unsigned short family = xnew->props.family;
1004 u32 reqid = xnew->props.reqid;
1005 struct xfrm_state *x;
1007 u32 mark = xnew->mark.v & xnew->mark.m;
1009 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1010 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1011 if (x->props.family == family &&
1012 x->props.reqid == reqid &&
1013 (mark & x->mark.m) == x->mark.v &&
1014 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1015 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1020 void xfrm_state_insert(struct xfrm_state *x)
1022 struct net *net = xs_net(x);
1024 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1025 __xfrm_state_bump_genids(x);
1026 __xfrm_state_insert(x);
1027 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1029 EXPORT_SYMBOL(xfrm_state_insert);
1031 /* net->xfrm.xfrm_state_lock is held */
1032 static struct xfrm_state *__find_acq_core(struct net *net,
1033 const struct xfrm_mark *m,
1034 unsigned short family, u8 mode,
1035 u32 reqid, u8 proto,
1036 const xfrm_address_t *daddr,
1037 const xfrm_address_t *saddr,
1040 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1041 struct xfrm_state *x;
1042 u32 mark = m->v & m->m;
1044 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1045 if (x->props.reqid != reqid ||
1046 x->props.mode != mode ||
1047 x->props.family != family ||
1048 x->km.state != XFRM_STATE_ACQ ||
1050 x->id.proto != proto ||
1051 (mark & x->mark.m) != x->mark.v ||
1052 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1053 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1063 x = xfrm_state_alloc(net);
1067 x->sel.daddr.a4 = daddr->a4;
1068 x->sel.saddr.a4 = saddr->a4;
1069 x->sel.prefixlen_d = 32;
1070 x->sel.prefixlen_s = 32;
1071 x->props.saddr.a4 = saddr->a4;
1072 x->id.daddr.a4 = daddr->a4;
1076 x->sel.daddr.in6 = daddr->in6;
1077 x->sel.saddr.in6 = saddr->in6;
1078 x->sel.prefixlen_d = 128;
1079 x->sel.prefixlen_s = 128;
1080 x->props.saddr.in6 = saddr->in6;
1081 x->id.daddr.in6 = daddr->in6;
1085 x->km.state = XFRM_STATE_ACQ;
1086 x->id.proto = proto;
1087 x->props.family = family;
1088 x->props.mode = mode;
1089 x->props.reqid = reqid;
1092 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1094 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1095 list_add(&x->km.all, &net->xfrm.state_all);
1096 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1097 h = xfrm_src_hash(net, daddr, saddr, family);
1098 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1100 net->xfrm.state_num++;
1102 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1108 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1110 int xfrm_state_add(struct xfrm_state *x)
1112 struct net *net = xs_net(x);
1113 struct xfrm_state *x1, *to_put;
1116 u32 mark = x->mark.v & x->mark.m;
1117 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1119 family = x->props.family;
1123 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1125 x1 = __xfrm_state_locate(x, use_spi, family);
1133 if (use_spi && x->km.seq) {
1134 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1135 if (x1 && ((x1->id.proto != x->id.proto) ||
1136 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1143 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1144 x->props.reqid, x->id.proto,
1145 &x->id.daddr, &x->props.saddr, 0);
1147 __xfrm_state_bump_genids(x);
1148 __xfrm_state_insert(x);
1152 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1155 xfrm_state_delete(x1);
1160 xfrm_state_put(to_put);
1164 EXPORT_SYMBOL(xfrm_state_add);
1166 #ifdef CONFIG_XFRM_MIGRATE
1167 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1169 struct net *net = xs_net(orig);
1170 struct xfrm_state *x = xfrm_state_alloc(net);
1174 memcpy(&x->id, &orig->id, sizeof(x->id));
1175 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1176 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1177 x->props.mode = orig->props.mode;
1178 x->props.replay_window = orig->props.replay_window;
1179 x->props.reqid = orig->props.reqid;
1180 x->props.family = orig->props.family;
1181 x->props.saddr = orig->props.saddr;
1184 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1188 x->props.aalgo = orig->props.aalgo;
1191 x->aead = xfrm_algo_aead_clone(orig->aead);
1196 x->ealg = xfrm_algo_clone(orig->ealg);
1200 x->props.ealgo = orig->props.ealgo;
1203 x->calg = xfrm_algo_clone(orig->calg);
1207 x->props.calgo = orig->props.calgo;
1210 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1216 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1222 if (orig->replay_esn) {
1223 if (xfrm_replay_clone(x, orig))
1227 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1229 if (xfrm_init_state(x) < 0)
1232 x->props.flags = orig->props.flags;
1233 x->props.extra_flags = orig->props.extra_flags;
1235 x->tfcpad = orig->tfcpad;
1236 x->replay_maxdiff = orig->replay_maxdiff;
1237 x->replay_maxage = orig->replay_maxage;
1238 x->curlft.add_time = orig->curlft.add_time;
1239 x->km.state = orig->km.state;
1240 x->km.seq = orig->km.seq;
1250 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1253 struct xfrm_state *x = NULL;
1255 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1258 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1259 m->reqid, m->old_family);
1260 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1261 if (x->props.mode != m->mode ||
1262 x->id.proto != m->proto)
1264 if (m->reqid && x->props.reqid != m->reqid)
1266 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1268 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1275 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1277 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1278 if (x->props.mode != m->mode ||
1279 x->id.proto != m->proto)
1281 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1283 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1291 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1295 EXPORT_SYMBOL(xfrm_migrate_state_find);
1297 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1298 struct xfrm_migrate *m)
1300 struct xfrm_state *xc;
1302 xc = xfrm_state_clone(x);
1306 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1307 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1310 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1311 /* a care is needed when the destination address of the
1312 state is to be updated as it is a part of triplet */
1313 xfrm_state_insert(xc);
1315 if (xfrm_state_add(xc) < 0)
1324 EXPORT_SYMBOL(xfrm_state_migrate);
1327 int xfrm_state_update(struct xfrm_state *x)
1329 struct xfrm_state *x1, *to_put;
1331 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1332 struct net *net = xs_net(x);
1336 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1337 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1343 if (xfrm_state_kern(x1)) {
1349 if (x1->km.state == XFRM_STATE_ACQ) {
1350 __xfrm_state_insert(x);
1356 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1359 xfrm_state_put(to_put);
1365 xfrm_state_delete(x1);
1371 spin_lock_bh(&x1->lock);
1372 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1373 if (x->encap && x1->encap)
1374 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1375 if (x->coaddr && x1->coaddr) {
1376 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1378 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1379 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1380 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1383 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1384 if (x1->curlft.use_time)
1385 xfrm_state_check_expire(x1);
1388 x->km.state = XFRM_STATE_DEAD;
1389 __xfrm_state_put(x);
1391 spin_unlock_bh(&x1->lock);
1397 EXPORT_SYMBOL(xfrm_state_update);
1399 int xfrm_state_check_expire(struct xfrm_state *x)
1401 if (!x->curlft.use_time)
1402 x->curlft.use_time = get_seconds();
1404 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1405 x->curlft.packets >= x->lft.hard_packet_limit) {
1406 x->km.state = XFRM_STATE_EXPIRED;
1407 tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1412 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1413 x->curlft.packets >= x->lft.soft_packet_limit)) {
1415 km_state_expired(x, 0, 0);
1419 EXPORT_SYMBOL(xfrm_state_check_expire);
1422 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1423 u8 proto, unsigned short family)
1425 struct xfrm_state *x;
1428 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1432 EXPORT_SYMBOL(xfrm_state_lookup);
1435 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1436 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1437 u8 proto, unsigned short family)
1439 struct xfrm_state *x;
1441 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1442 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1443 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1446 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1449 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1450 u8 proto, const xfrm_address_t *daddr,
1451 const xfrm_address_t *saddr, int create, unsigned short family)
1453 struct xfrm_state *x;
1455 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1456 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1457 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1461 EXPORT_SYMBOL(xfrm_find_acq);
1463 #ifdef CONFIG_XFRM_SUB_POLICY
1465 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1466 unsigned short family, struct net *net)
1469 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1471 return -EAFNOSUPPORT;
1473 spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1474 if (afinfo->tmpl_sort)
1475 err = afinfo->tmpl_sort(dst, src, n);
1476 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1480 EXPORT_SYMBOL(xfrm_tmpl_sort);
1483 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1484 unsigned short family)
1487 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1488 struct net *net = xs_net(*src);
1491 return -EAFNOSUPPORT;
1493 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1494 if (afinfo->state_sort)
1495 err = afinfo->state_sort(dst, src, n);
1496 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1500 EXPORT_SYMBOL(xfrm_state_sort);
1503 /* Silly enough, but I'm lazy to build resolution list */
1505 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1509 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1510 struct xfrm_state *x;
1512 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1513 if (x->km.seq == seq &&
1514 (mark & x->mark.m) == x->mark.v &&
1515 x->km.state == XFRM_STATE_ACQ) {
1524 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1526 struct xfrm_state *x;
1528 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1529 x = __xfrm_find_acq_byseq(net, mark, seq);
1530 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1533 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1535 u32 xfrm_get_acqseq(void)
1538 static atomic_t acqseq;
1541 res = atomic_inc_return(&acqseq);
1546 EXPORT_SYMBOL(xfrm_get_acqseq);
1548 int verify_spi_info(u8 proto, u32 min, u32 max)
1556 /* IPCOMP spi is 16-bits. */
1570 EXPORT_SYMBOL(verify_spi_info);
1572 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1574 struct net *net = xs_net(x);
1576 struct xfrm_state *x0;
1578 __be32 minspi = htonl(low);
1579 __be32 maxspi = htonl(high);
1580 u32 mark = x->mark.v & x->mark.m;
1582 spin_lock_bh(&x->lock);
1583 if (x->km.state == XFRM_STATE_DEAD)
1592 if (minspi == maxspi) {
1593 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1601 for (h = 0; h < high-low+1; h++) {
1602 spi = low + prandom_u32()%(high-low+1);
1603 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1605 x->id.spi = htonl(spi);
1612 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1613 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1614 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1615 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1621 spin_unlock_bh(&x->lock);
1625 EXPORT_SYMBOL(xfrm_alloc_spi);
1627 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1628 struct xfrm_address_filter *filter)
1631 if ((filter->family == AF_INET ||
1632 filter->family == AF_INET6) &&
1633 x->props.family != filter->family)
1636 return addr_match(&x->props.saddr, &filter->saddr,
1638 addr_match(&x->id.daddr, &filter->daddr,
1644 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1645 int (*func)(struct xfrm_state *, int, void*),
1648 struct xfrm_state *state;
1649 struct xfrm_state_walk *x;
1652 if (walk->seq != 0 && list_empty(&walk->all))
1655 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1656 if (list_empty(&walk->all))
1657 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1659 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1660 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1661 if (x->state == XFRM_STATE_DEAD)
1663 state = container_of(x, struct xfrm_state, km);
1664 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1666 if (!__xfrm_state_filter_match(state, walk->filter))
1668 err = func(state, walk->seq, data);
1670 list_move_tail(&walk->all, &x->all);
1675 if (walk->seq == 0) {
1679 list_del_init(&walk->all);
1681 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1684 EXPORT_SYMBOL(xfrm_state_walk);
1686 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1687 struct xfrm_address_filter *filter)
1689 INIT_LIST_HEAD(&walk->all);
1690 walk->proto = proto;
1691 walk->state = XFRM_STATE_DEAD;
1693 walk->filter = filter;
1695 EXPORT_SYMBOL(xfrm_state_walk_init);
1697 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1699 kfree(walk->filter);
1701 if (list_empty(&walk->all))
1704 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1705 list_del(&walk->all);
1706 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1708 EXPORT_SYMBOL(xfrm_state_walk_done);
1710 static void xfrm_replay_timer_handler(unsigned long data)
1712 struct xfrm_state *x = (struct xfrm_state *)data;
1714 spin_lock(&x->lock);
1716 if (x->km.state == XFRM_STATE_VALID) {
1717 if (xfrm_aevent_is_on(xs_net(x)))
1718 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1720 x->xflags |= XFRM_TIME_DEFER;
1723 spin_unlock(&x->lock);
1726 static LIST_HEAD(xfrm_km_list);
1728 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1730 struct xfrm_mgr *km;
1733 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1734 if (km->notify_policy)
1735 km->notify_policy(xp, dir, c);
1739 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1741 struct xfrm_mgr *km;
1743 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1749 EXPORT_SYMBOL(km_policy_notify);
1750 EXPORT_SYMBOL(km_state_notify);
1752 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1758 c.event = XFRM_MSG_EXPIRE;
1759 km_state_notify(x, &c);
1762 EXPORT_SYMBOL(km_state_expired);
1764 * We send to all registered managers regardless of failure
1765 * We are happy with one success
1767 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1769 int err = -EINVAL, acqret;
1770 struct xfrm_mgr *km;
1773 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1774 acqret = km->acquire(x, t, pol);
1781 EXPORT_SYMBOL(km_query);
1783 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1786 struct xfrm_mgr *km;
1789 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1790 if (km->new_mapping)
1791 err = km->new_mapping(x, ipaddr, sport);
1798 EXPORT_SYMBOL(km_new_mapping);
1800 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1806 c.event = XFRM_MSG_POLEXPIRE;
1807 km_policy_notify(pol, dir, &c);
1809 EXPORT_SYMBOL(km_policy_expired);
1811 #ifdef CONFIG_XFRM_MIGRATE
1812 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1813 const struct xfrm_migrate *m, int num_migrate,
1814 const struct xfrm_kmaddress *k)
1818 struct xfrm_mgr *km;
1821 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1823 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1831 EXPORT_SYMBOL(km_migrate);
1834 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1838 struct xfrm_mgr *km;
1841 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1843 ret = km->report(net, proto, sel, addr);
1851 EXPORT_SYMBOL(km_report);
1853 bool km_is_alive(const struct km_event *c)
1855 struct xfrm_mgr *km;
1856 bool is_alive = false;
1859 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1860 if (km->is_alive && km->is_alive(c)) {
1869 EXPORT_SYMBOL(km_is_alive);
1871 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1875 struct xfrm_mgr *km;
1876 struct xfrm_policy *pol = NULL;
1878 if (optlen <= 0 || optlen > PAGE_SIZE)
1881 data = kmalloc(optlen, GFP_KERNEL);
1886 if (copy_from_user(data, optval, optlen))
1891 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1892 pol = km->compile_policy(sk, optname, data,
1900 xfrm_sk_policy_insert(sk, err, pol);
1909 EXPORT_SYMBOL(xfrm_user_policy);
1911 static DEFINE_SPINLOCK(xfrm_km_lock);
1913 int xfrm_register_km(struct xfrm_mgr *km)
1915 spin_lock_bh(&xfrm_km_lock);
1916 list_add_tail_rcu(&km->list, &xfrm_km_list);
1917 spin_unlock_bh(&xfrm_km_lock);
1920 EXPORT_SYMBOL(xfrm_register_km);
1922 int xfrm_unregister_km(struct xfrm_mgr *km)
1924 spin_lock_bh(&xfrm_km_lock);
1925 list_del_rcu(&km->list);
1926 spin_unlock_bh(&xfrm_km_lock);
1930 EXPORT_SYMBOL(xfrm_unregister_km);
1932 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1936 if (WARN_ON(afinfo->family >= NPROTO))
1937 return -EAFNOSUPPORT;
1939 spin_lock_bh(&xfrm_state_afinfo_lock);
1940 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1943 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1944 spin_unlock_bh(&xfrm_state_afinfo_lock);
1947 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1949 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1951 int err = 0, family = afinfo->family;
1953 if (WARN_ON(family >= NPROTO))
1954 return -EAFNOSUPPORT;
1956 spin_lock_bh(&xfrm_state_afinfo_lock);
1957 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1958 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
1961 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1963 spin_unlock_bh(&xfrm_state_afinfo_lock);
1967 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1969 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
1971 if (unlikely(family >= NPROTO))
1974 return rcu_dereference(xfrm_state_afinfo[family]);
1977 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1979 struct xfrm_state_afinfo *afinfo;
1980 if (unlikely(family >= NPROTO))
1983 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1984 if (unlikely(!afinfo))
1989 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1990 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1993 struct xfrm_state *t = x->tunnel;
1995 if (atomic_read(&t->tunnel_users) == 2)
1996 xfrm_state_delete(t);
1997 atomic_dec(&t->tunnel_users);
2002 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2004 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2006 const struct xfrm_type *type = READ_ONCE(x->type);
2008 if (x->km.state == XFRM_STATE_VALID &&
2009 type && type->get_mtu)
2010 return type->get_mtu(x, mtu);
2012 return mtu - x->props.header_len;
2015 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
2017 struct xfrm_state_afinfo *afinfo;
2018 struct xfrm_mode *inner_mode;
2019 int family = x->props.family;
2022 err = -EAFNOSUPPORT;
2023 afinfo = xfrm_state_get_afinfo(family);
2028 if (afinfo->init_flags)
2029 err = afinfo->init_flags(x);
2036 err = -EPROTONOSUPPORT;
2038 if (x->sel.family != AF_UNSPEC) {
2039 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2040 if (inner_mode == NULL)
2043 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2044 family != x->sel.family) {
2045 xfrm_put_mode(inner_mode);
2049 x->inner_mode = inner_mode;
2051 struct xfrm_mode *inner_mode_iaf;
2052 int iafamily = AF_INET;
2054 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2055 if (inner_mode == NULL)
2058 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2059 xfrm_put_mode(inner_mode);
2062 x->inner_mode = inner_mode;
2064 if (x->props.family == AF_INET)
2065 iafamily = AF_INET6;
2067 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2068 if (inner_mode_iaf) {
2069 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2070 x->inner_mode_iaf = inner_mode_iaf;
2072 xfrm_put_mode(inner_mode_iaf);
2076 x->type = xfrm_get_type(x->id.proto, family);
2077 if (x->type == NULL)
2080 err = x->type->init_state(x);
2084 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2085 if (x->outer_mode == NULL) {
2086 err = -EPROTONOSUPPORT;
2091 err = xfrm_init_replay(x);
2096 x->km.state = XFRM_STATE_VALID;
2102 EXPORT_SYMBOL(__xfrm_init_state);
2104 int xfrm_init_state(struct xfrm_state *x)
2106 return __xfrm_init_state(x, true);
2109 EXPORT_SYMBOL(xfrm_init_state);
2111 int __net_init xfrm_state_init(struct net *net)
2115 INIT_LIST_HEAD(&net->xfrm.state_all);
2117 sz = sizeof(struct hlist_head) * 8;
2119 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2120 if (!net->xfrm.state_bydst)
2122 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2123 if (!net->xfrm.state_bysrc)
2125 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2126 if (!net->xfrm.state_byspi)
2128 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2130 net->xfrm.state_num = 0;
2131 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2132 spin_lock_init(&net->xfrm.xfrm_state_lock);
2136 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2138 xfrm_hash_free(net->xfrm.state_bydst, sz);
2143 void xfrm_state_fini(struct net *net)
2147 flush_work(&net->xfrm.state_hash_work);
2148 xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2149 flush_work(&xfrm_state_gc_work);
2151 WARN_ON(!list_empty(&net->xfrm.state_all));
2153 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2154 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2155 xfrm_hash_free(net->xfrm.state_byspi, sz);
2156 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2157 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2158 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2159 xfrm_hash_free(net->xfrm.state_bydst, sz);
2162 #ifdef CONFIG_AUDITSYSCALL
2163 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2164 struct audit_buffer *audit_buf)
2166 struct xfrm_sec_ctx *ctx = x->security;
2167 u32 spi = ntohl(x->id.spi);
2170 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2171 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2173 switch (x->props.family) {
2175 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2176 &x->props.saddr.a4, &x->id.daddr.a4);
2179 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2180 x->props.saddr.a6, x->id.daddr.a6);
2184 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2187 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2188 struct audit_buffer *audit_buf)
2190 const struct iphdr *iph4;
2191 const struct ipv6hdr *iph6;
2196 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2197 &iph4->saddr, &iph4->daddr);
2200 iph6 = ipv6_hdr(skb);
2201 audit_log_format(audit_buf,
2202 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2203 &iph6->saddr, &iph6->daddr,
2204 iph6->flow_lbl[0] & 0x0f,
2211 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2213 struct audit_buffer *audit_buf;
2215 audit_buf = xfrm_audit_start("SAD-add");
2216 if (audit_buf == NULL)
2218 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2219 xfrm_audit_helper_sainfo(x, audit_buf);
2220 audit_log_format(audit_buf, " res=%u", result);
2221 audit_log_end(audit_buf);
2223 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2225 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2227 struct audit_buffer *audit_buf;
2229 audit_buf = xfrm_audit_start("SAD-delete");
2230 if (audit_buf == NULL)
2232 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2233 xfrm_audit_helper_sainfo(x, audit_buf);
2234 audit_log_format(audit_buf, " res=%u", result);
2235 audit_log_end(audit_buf);
2237 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2239 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2240 struct sk_buff *skb)
2242 struct audit_buffer *audit_buf;
2245 audit_buf = xfrm_audit_start("SA-replay-overflow");
2246 if (audit_buf == NULL)
2248 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2249 /* don't record the sequence number because it's inherent in this kind
2250 * of audit message */
2251 spi = ntohl(x->id.spi);
2252 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2253 audit_log_end(audit_buf);
2255 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2257 void xfrm_audit_state_replay(struct xfrm_state *x,
2258 struct sk_buff *skb, __be32 net_seq)
2260 struct audit_buffer *audit_buf;
2263 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2264 if (audit_buf == NULL)
2266 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2267 spi = ntohl(x->id.spi);
2268 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2269 spi, spi, ntohl(net_seq));
2270 audit_log_end(audit_buf);
2272 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2274 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2276 struct audit_buffer *audit_buf;
2278 audit_buf = xfrm_audit_start("SA-notfound");
2279 if (audit_buf == NULL)
2281 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2282 audit_log_end(audit_buf);
2284 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2286 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2287 __be32 net_spi, __be32 net_seq)
2289 struct audit_buffer *audit_buf;
2292 audit_buf = xfrm_audit_start("SA-notfound");
2293 if (audit_buf == NULL)
2295 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2296 spi = ntohl(net_spi);
2297 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2298 spi, spi, ntohl(net_seq));
2299 audit_log_end(audit_buf);
2301 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2303 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2304 struct sk_buff *skb, u8 proto)
2306 struct audit_buffer *audit_buf;
2310 audit_buf = xfrm_audit_start("SA-icv-failure");
2311 if (audit_buf == NULL)
2313 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2314 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2315 u32 spi = ntohl(net_spi);
2316 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2317 spi, spi, ntohl(net_seq));
2319 audit_log_end(audit_buf);
2321 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2322 #endif /* CONFIG_AUDITSYSCALL */