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 <asm/uaccess.h>
24 #include <linux/ktime.h>
25 #include <linux/interrupt.h>
26 #include <linux/kernel.h>
28 #include "xfrm_hash.h"
30 /* Each xfrm_state may be linked to two tables:
32 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
33 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
34 destination/tunnel endpoint. (output)
37 static DEFINE_SPINLOCK(xfrm_state_lock);
39 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
40 static unsigned int xfrm_state_genid;
42 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
45 #ifdef CONFIG_AUDITSYSCALL
46 static void xfrm_audit_state_replay(struct xfrm_state *x,
47 struct sk_buff *skb, __be32 net_seq);
49 #define xfrm_audit_state_replay(x, s, sq) do { ; } while (0)
50 #endif /* CONFIG_AUDITSYSCALL */
52 static inline unsigned int xfrm_dst_hash(struct net *net,
53 xfrm_address_t *daddr,
54 xfrm_address_t *saddr,
56 unsigned short family)
58 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
61 static inline unsigned int xfrm_src_hash(struct net *net,
62 xfrm_address_t *daddr,
63 xfrm_address_t *saddr,
64 unsigned short family)
66 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
69 static inline unsigned int
70 xfrm_spi_hash(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
72 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
75 static void xfrm_hash_transfer(struct hlist_head *list,
76 struct hlist_head *ndsttable,
77 struct hlist_head *nsrctable,
78 struct hlist_head *nspitable,
79 unsigned int nhashmask)
81 struct hlist_node *entry, *tmp;
84 hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
87 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
88 x->props.reqid, x->props.family,
90 hlist_add_head(&x->bydst, ndsttable+h);
92 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
95 hlist_add_head(&x->bysrc, nsrctable+h);
98 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
99 x->id.proto, x->props.family,
101 hlist_add_head(&x->byspi, nspitable+h);
106 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
108 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
111 static DEFINE_MUTEX(hash_resize_mutex);
113 static void xfrm_hash_resize(struct work_struct *work)
115 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
116 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
117 unsigned long nsize, osize;
118 unsigned int nhashmask, ohashmask;
121 mutex_lock(&hash_resize_mutex);
123 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
124 ndst = xfrm_hash_alloc(nsize);
127 nsrc = xfrm_hash_alloc(nsize);
129 xfrm_hash_free(ndst, nsize);
132 nspi = xfrm_hash_alloc(nsize);
134 xfrm_hash_free(ndst, nsize);
135 xfrm_hash_free(nsrc, nsize);
139 spin_lock_bh(&xfrm_state_lock);
141 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
142 for (i = net->xfrm.state_hmask; i >= 0; i--)
143 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
146 odst = net->xfrm.state_bydst;
147 osrc = net->xfrm.state_bysrc;
148 ospi = net->xfrm.state_byspi;
149 ohashmask = net->xfrm.state_hmask;
151 net->xfrm.state_bydst = ndst;
152 net->xfrm.state_bysrc = nsrc;
153 net->xfrm.state_byspi = nspi;
154 net->xfrm.state_hmask = nhashmask;
156 spin_unlock_bh(&xfrm_state_lock);
158 osize = (ohashmask + 1) * sizeof(struct hlist_head);
159 xfrm_hash_free(odst, osize);
160 xfrm_hash_free(osrc, osize);
161 xfrm_hash_free(ospi, osize);
164 mutex_unlock(&hash_resize_mutex);
167 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
168 static struct xfrm_state_afinfo *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 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
177 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
179 struct xfrm_state_afinfo *afinfo;
180 if (unlikely(family >= NPROTO))
182 write_lock_bh(&xfrm_state_afinfo_lock);
183 afinfo = xfrm_state_afinfo[family];
184 if (unlikely(!afinfo))
185 write_unlock_bh(&xfrm_state_afinfo_lock);
189 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
190 __releases(xfrm_state_afinfo_lock)
192 write_unlock_bh(&xfrm_state_afinfo_lock);
195 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
197 struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
198 const struct xfrm_type **typemap;
201 if (unlikely(afinfo == NULL))
202 return -EAFNOSUPPORT;
203 typemap = afinfo->type_map;
205 if (likely(typemap[type->proto] == NULL))
206 typemap[type->proto] = type;
209 xfrm_state_unlock_afinfo(afinfo);
212 EXPORT_SYMBOL(xfrm_register_type);
214 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
216 struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
217 const struct xfrm_type **typemap;
220 if (unlikely(afinfo == NULL))
221 return -EAFNOSUPPORT;
222 typemap = afinfo->type_map;
224 if (unlikely(typemap[type->proto] != type))
227 typemap[type->proto] = NULL;
228 xfrm_state_unlock_afinfo(afinfo);
231 EXPORT_SYMBOL(xfrm_unregister_type);
233 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
235 struct xfrm_state_afinfo *afinfo;
236 const struct xfrm_type **typemap;
237 const struct xfrm_type *type;
238 int modload_attempted = 0;
241 afinfo = xfrm_state_get_afinfo(family);
242 if (unlikely(afinfo == NULL))
244 typemap = afinfo->type_map;
246 type = typemap[proto];
247 if (unlikely(type && !try_module_get(type->owner)))
249 if (!type && !modload_attempted) {
250 xfrm_state_put_afinfo(afinfo);
251 request_module("xfrm-type-%d-%d", family, proto);
252 modload_attempted = 1;
256 xfrm_state_put_afinfo(afinfo);
260 static void xfrm_put_type(const struct xfrm_type *type)
262 module_put(type->owner);
265 int xfrm_register_mode(struct xfrm_mode *mode, int family)
267 struct xfrm_state_afinfo *afinfo;
268 struct xfrm_mode **modemap;
271 if (unlikely(mode->encap >= XFRM_MODE_MAX))
274 afinfo = xfrm_state_lock_afinfo(family);
275 if (unlikely(afinfo == NULL))
276 return -EAFNOSUPPORT;
279 modemap = afinfo->mode_map;
280 if (modemap[mode->encap])
284 if (!try_module_get(afinfo->owner))
287 mode->afinfo = afinfo;
288 modemap[mode->encap] = mode;
292 xfrm_state_unlock_afinfo(afinfo);
295 EXPORT_SYMBOL(xfrm_register_mode);
297 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
299 struct xfrm_state_afinfo *afinfo;
300 struct xfrm_mode **modemap;
303 if (unlikely(mode->encap >= XFRM_MODE_MAX))
306 afinfo = xfrm_state_lock_afinfo(family);
307 if (unlikely(afinfo == NULL))
308 return -EAFNOSUPPORT;
311 modemap = afinfo->mode_map;
312 if (likely(modemap[mode->encap] == mode)) {
313 modemap[mode->encap] = NULL;
314 module_put(mode->afinfo->owner);
318 xfrm_state_unlock_afinfo(afinfo);
321 EXPORT_SYMBOL(xfrm_unregister_mode);
323 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
325 struct xfrm_state_afinfo *afinfo;
326 struct xfrm_mode *mode;
327 int modload_attempted = 0;
329 if (unlikely(encap >= XFRM_MODE_MAX))
333 afinfo = xfrm_state_get_afinfo(family);
334 if (unlikely(afinfo == NULL))
337 mode = afinfo->mode_map[encap];
338 if (unlikely(mode && !try_module_get(mode->owner)))
340 if (!mode && !modload_attempted) {
341 xfrm_state_put_afinfo(afinfo);
342 request_module("xfrm-mode-%d-%d", family, encap);
343 modload_attempted = 1;
347 xfrm_state_put_afinfo(afinfo);
351 static void xfrm_put_mode(struct xfrm_mode *mode)
353 module_put(mode->owner);
356 static void xfrm_state_gc_destroy(struct xfrm_state *x)
358 tasklet_hrtimer_cancel(&x->mtimer);
359 del_timer_sync(&x->rtimer);
366 xfrm_put_mode(x->inner_mode);
367 if (x->inner_mode_iaf)
368 xfrm_put_mode(x->inner_mode_iaf);
370 xfrm_put_mode(x->outer_mode);
372 x->type->destructor(x);
373 xfrm_put_type(x->type);
375 security_xfrm_state_free(x);
379 static void xfrm_state_gc_task(struct work_struct *work)
381 struct net *net = container_of(work, struct net, xfrm.state_gc_work);
382 struct xfrm_state *x;
383 struct hlist_node *entry, *tmp;
384 struct hlist_head gc_list;
386 spin_lock_bh(&xfrm_state_gc_lock);
387 hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
388 spin_unlock_bh(&xfrm_state_gc_lock);
390 hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
391 xfrm_state_gc_destroy(x);
393 wake_up(&net->xfrm.km_waitq);
396 static inline unsigned long make_jiffies(long secs)
398 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
399 return MAX_SCHEDULE_TIMEOUT-1;
404 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
406 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
407 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
408 struct net *net = xs_net(x);
409 unsigned long now = get_seconds();
410 long next = LONG_MAX;
415 if (x->km.state == XFRM_STATE_DEAD)
417 if (x->km.state == XFRM_STATE_EXPIRED)
419 if (x->lft.hard_add_expires_seconds) {
420 long tmo = x->lft.hard_add_expires_seconds +
421 x->curlft.add_time - now;
427 if (x->lft.hard_use_expires_seconds) {
428 long tmo = x->lft.hard_use_expires_seconds +
429 (x->curlft.use_time ? : now) - now;
437 if (x->lft.soft_add_expires_seconds) {
438 long tmo = x->lft.soft_add_expires_seconds +
439 x->curlft.add_time - now;
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;
467 wake_up(&net->xfrm.km_waitq);
472 err = __xfrm_state_delete(x);
473 if (!err && x->id.spi)
474 km_state_expired(x, 1, 0);
476 xfrm_audit_state_delete(x, err ? 0 : 1,
477 audit_get_loginuid(current),
478 audit_get_sessionid(current), 0);
481 spin_unlock(&x->lock);
482 return HRTIMER_NORESTART;
485 static void xfrm_replay_timer_handler(unsigned long data);
487 struct xfrm_state *xfrm_state_alloc(struct net *net)
489 struct xfrm_state *x;
491 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
494 write_pnet(&x->xs_net, net);
495 atomic_set(&x->refcnt, 1);
496 atomic_set(&x->tunnel_users, 0);
497 INIT_LIST_HEAD(&x->km.all);
498 INIT_HLIST_NODE(&x->bydst);
499 INIT_HLIST_NODE(&x->bysrc);
500 INIT_HLIST_NODE(&x->byspi);
501 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
502 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
504 x->curlft.add_time = get_seconds();
505 x->lft.soft_byte_limit = XFRM_INF;
506 x->lft.soft_packet_limit = XFRM_INF;
507 x->lft.hard_byte_limit = XFRM_INF;
508 x->lft.hard_packet_limit = XFRM_INF;
509 x->replay_maxage = 0;
510 x->replay_maxdiff = 0;
511 x->inner_mode = NULL;
512 x->inner_mode_iaf = NULL;
513 spin_lock_init(&x->lock);
517 EXPORT_SYMBOL(xfrm_state_alloc);
519 void __xfrm_state_destroy(struct xfrm_state *x)
521 struct net *net = xs_net(x);
523 WARN_ON(x->km.state != XFRM_STATE_DEAD);
525 spin_lock_bh(&xfrm_state_gc_lock);
526 hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
527 spin_unlock_bh(&xfrm_state_gc_lock);
528 schedule_work(&net->xfrm.state_gc_work);
530 EXPORT_SYMBOL(__xfrm_state_destroy);
532 int __xfrm_state_delete(struct xfrm_state *x)
534 struct net *net = xs_net(x);
537 if (x->km.state != XFRM_STATE_DEAD) {
538 x->km.state = XFRM_STATE_DEAD;
539 spin_lock(&xfrm_state_lock);
540 list_del(&x->km.all);
541 hlist_del(&x->bydst);
542 hlist_del(&x->bysrc);
544 hlist_del(&x->byspi);
545 net->xfrm.state_num--;
546 spin_unlock(&xfrm_state_lock);
548 /* All xfrm_state objects are created by xfrm_state_alloc.
549 * The xfrm_state_alloc call gives a reference, and that
550 * is what we are dropping here.
558 EXPORT_SYMBOL(__xfrm_state_delete);
560 int xfrm_state_delete(struct xfrm_state *x)
564 spin_lock_bh(&x->lock);
565 err = __xfrm_state_delete(x);
566 spin_unlock_bh(&x->lock);
570 EXPORT_SYMBOL(xfrm_state_delete);
572 #ifdef CONFIG_SECURITY_NETWORK_XFRM
574 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
578 for (i = 0; i <= net->xfrm.state_hmask; i++) {
579 struct hlist_node *entry;
580 struct xfrm_state *x;
582 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
583 if (xfrm_id_proto_match(x->id.proto, proto) &&
584 (err = security_xfrm_state_delete(x)) != 0) {
585 xfrm_audit_state_delete(x, 0,
586 audit_info->loginuid,
587 audit_info->sessionid,
598 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
604 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
606 int i, err = 0, cnt = 0;
608 spin_lock_bh(&xfrm_state_lock);
609 err = xfrm_state_flush_secctx_check(net, proto, audit_info);
614 for (i = 0; i <= net->xfrm.state_hmask; i++) {
615 struct hlist_node *entry;
616 struct xfrm_state *x;
618 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
619 if (!xfrm_state_kern(x) &&
620 xfrm_id_proto_match(x->id.proto, proto)) {
622 spin_unlock_bh(&xfrm_state_lock);
624 err = xfrm_state_delete(x);
625 xfrm_audit_state_delete(x, err ? 0 : 1,
626 audit_info->loginuid,
627 audit_info->sessionid,
633 spin_lock_bh(&xfrm_state_lock);
642 spin_unlock_bh(&xfrm_state_lock);
643 wake_up(&net->xfrm.km_waitq);
646 EXPORT_SYMBOL(xfrm_state_flush);
648 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
650 spin_lock_bh(&xfrm_state_lock);
651 si->sadcnt = net->xfrm.state_num;
652 si->sadhcnt = net->xfrm.state_hmask;
653 si->sadhmcnt = xfrm_state_hashmax;
654 spin_unlock_bh(&xfrm_state_lock);
656 EXPORT_SYMBOL(xfrm_sad_getinfo);
659 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
660 struct xfrm_tmpl *tmpl,
661 xfrm_address_t *daddr, xfrm_address_t *saddr,
662 unsigned short family)
664 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
667 afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
668 xfrm_state_put_afinfo(afinfo);
672 static struct xfrm_state *__xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
674 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
675 struct xfrm_state *x;
676 struct hlist_node *entry;
678 hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
679 if (x->props.family != family ||
681 x->id.proto != proto ||
682 xfrm_addr_cmp(&x->id.daddr, daddr, family))
692 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
694 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
695 struct xfrm_state *x;
696 struct hlist_node *entry;
698 hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
699 if (x->props.family != family ||
700 x->id.proto != proto ||
701 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
702 xfrm_addr_cmp(&x->props.saddr, saddr, family))
712 static inline struct xfrm_state *
713 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
715 struct net *net = xs_net(x);
718 return __xfrm_state_lookup(net, &x->id.daddr, x->id.spi,
719 x->id.proto, family);
721 return __xfrm_state_lookup_byaddr(net, &x->id.daddr,
723 x->id.proto, family);
726 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
728 if (have_hash_collision &&
729 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
730 net->xfrm.state_num > net->xfrm.state_hmask)
731 schedule_work(&net->xfrm.state_hash_work);
734 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
735 struct flowi *fl, unsigned short family,
736 xfrm_address_t *daddr, xfrm_address_t *saddr,
737 struct xfrm_state **best, int *acq_in_progress,
741 * 1. There is a valid state with matching selector. Done.
742 * 2. Valid state with inappropriate selector. Skip.
744 * Entering area of "sysdeps".
746 * 3. If state is not valid, selector is temporary, it selects
747 * only session which triggered previous resolution. Key
748 * manager will do something to install a state with proper
751 if (x->km.state == XFRM_STATE_VALID) {
752 if ((x->sel.family &&
753 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
754 !security_xfrm_state_pol_flow_match(x, pol, fl))
758 (*best)->km.dying > x->km.dying ||
759 ((*best)->km.dying == x->km.dying &&
760 (*best)->curlft.add_time < x->curlft.add_time))
762 } else if (x->km.state == XFRM_STATE_ACQ) {
763 *acq_in_progress = 1;
764 } else if (x->km.state == XFRM_STATE_ERROR ||
765 x->km.state == XFRM_STATE_EXPIRED) {
766 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
767 security_xfrm_state_pol_flow_match(x, pol, fl))
773 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
774 struct flowi *fl, struct xfrm_tmpl *tmpl,
775 struct xfrm_policy *pol, int *err,
776 unsigned short family)
778 static xfrm_address_t saddr_wildcard = { };
779 struct net *net = xp_net(pol);
780 unsigned int h, h_wildcard;
781 struct hlist_node *entry;
782 struct xfrm_state *x, *x0, *to_put;
783 int acquire_in_progress = 0;
785 struct xfrm_state *best = NULL;
789 spin_lock_bh(&xfrm_state_lock);
790 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
791 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
792 if (x->props.family == family &&
793 x->props.reqid == tmpl->reqid &&
794 !(x->props.flags & XFRM_STATE_WILDRECV) &&
795 xfrm_state_addr_check(x, daddr, saddr, family) &&
796 tmpl->mode == x->props.mode &&
797 tmpl->id.proto == x->id.proto &&
798 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
799 xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
800 &best, &acquire_in_progress, &error);
805 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
806 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
807 if (x->props.family == family &&
808 x->props.reqid == tmpl->reqid &&
809 !(x->props.flags & XFRM_STATE_WILDRECV) &&
810 xfrm_state_addr_check(x, daddr, saddr, family) &&
811 tmpl->mode == x->props.mode &&
812 tmpl->id.proto == x->id.proto &&
813 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
814 xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
815 &best, &acquire_in_progress, &error);
820 if (!x && !error && !acquire_in_progress) {
822 (x0 = __xfrm_state_lookup(net, daddr, tmpl->id.spi,
823 tmpl->id.proto, family)) != NULL) {
828 x = xfrm_state_alloc(net);
833 /* Initialize temporary selector matching only
834 * to current session. */
835 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
837 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
839 x->km.state = XFRM_STATE_DEAD;
845 if (km_query(x, tmpl, pol) == 0) {
846 x->km.state = XFRM_STATE_ACQ;
847 list_add(&x->km.all, &net->xfrm.state_all);
848 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
849 h = xfrm_src_hash(net, daddr, saddr, family);
850 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
852 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
853 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
855 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
856 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
857 net->xfrm.state_num++;
858 xfrm_hash_grow_check(net, x->bydst.next != NULL);
860 x->km.state = XFRM_STATE_DEAD;
870 *err = acquire_in_progress ? -EAGAIN : error;
871 spin_unlock_bh(&xfrm_state_lock);
873 xfrm_state_put(to_put);
878 xfrm_stateonly_find(struct net *net,
879 xfrm_address_t *daddr, xfrm_address_t *saddr,
880 unsigned short family, u8 mode, u8 proto, u32 reqid)
883 struct xfrm_state *rx = NULL, *x = NULL;
884 struct hlist_node *entry;
886 spin_lock(&xfrm_state_lock);
887 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
888 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
889 if (x->props.family == family &&
890 x->props.reqid == reqid &&
891 !(x->props.flags & XFRM_STATE_WILDRECV) &&
892 xfrm_state_addr_check(x, daddr, saddr, family) &&
893 mode == x->props.mode &&
894 proto == x->id.proto &&
895 x->km.state == XFRM_STATE_VALID) {
903 spin_unlock(&xfrm_state_lock);
908 EXPORT_SYMBOL(xfrm_stateonly_find);
910 static void __xfrm_state_insert(struct xfrm_state *x)
912 struct net *net = xs_net(x);
915 x->genid = ++xfrm_state_genid;
917 list_add(&x->km.all, &net->xfrm.state_all);
919 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
920 x->props.reqid, x->props.family);
921 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
923 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
924 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
927 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
930 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
933 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
934 if (x->replay_maxage)
935 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
937 wake_up(&net->xfrm.km_waitq);
939 net->xfrm.state_num++;
941 xfrm_hash_grow_check(net, x->bydst.next != NULL);
944 /* xfrm_state_lock is held */
945 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
947 struct net *net = xs_net(xnew);
948 unsigned short family = xnew->props.family;
949 u32 reqid = xnew->props.reqid;
950 struct xfrm_state *x;
951 struct hlist_node *entry;
954 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
955 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
956 if (x->props.family == family &&
957 x->props.reqid == reqid &&
958 !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
959 !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
960 x->genid = xfrm_state_genid;
964 void xfrm_state_insert(struct xfrm_state *x)
966 spin_lock_bh(&xfrm_state_lock);
967 __xfrm_state_bump_genids(x);
968 __xfrm_state_insert(x);
969 spin_unlock_bh(&xfrm_state_lock);
971 EXPORT_SYMBOL(xfrm_state_insert);
973 /* xfrm_state_lock is held */
974 static struct xfrm_state *__find_acq_core(struct net *net, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
976 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
977 struct hlist_node *entry;
978 struct xfrm_state *x;
980 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
981 if (x->props.reqid != reqid ||
982 x->props.mode != mode ||
983 x->props.family != family ||
984 x->km.state != XFRM_STATE_ACQ ||
986 x->id.proto != proto ||
987 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
988 xfrm_addr_cmp(&x->props.saddr, saddr, family))
998 x = xfrm_state_alloc(net);
1002 x->sel.daddr.a4 = daddr->a4;
1003 x->sel.saddr.a4 = saddr->a4;
1004 x->sel.prefixlen_d = 32;
1005 x->sel.prefixlen_s = 32;
1006 x->props.saddr.a4 = saddr->a4;
1007 x->id.daddr.a4 = daddr->a4;
1011 ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1012 (struct in6_addr *)daddr);
1013 ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1014 (struct in6_addr *)saddr);
1015 x->sel.prefixlen_d = 128;
1016 x->sel.prefixlen_s = 128;
1017 ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1018 (struct in6_addr *)saddr);
1019 ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1020 (struct in6_addr *)daddr);
1024 x->km.state = XFRM_STATE_ACQ;
1025 x->id.proto = proto;
1026 x->props.family = family;
1027 x->props.mode = mode;
1028 x->props.reqid = reqid;
1029 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1031 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1032 list_add(&x->km.all, &net->xfrm.state_all);
1033 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1034 h = xfrm_src_hash(net, daddr, saddr, family);
1035 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1037 net->xfrm.state_num++;
1039 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1045 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq);
1047 int xfrm_state_add(struct xfrm_state *x)
1049 struct net *net = xs_net(x);
1050 struct xfrm_state *x1, *to_put;
1053 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1055 family = x->props.family;
1059 spin_lock_bh(&xfrm_state_lock);
1061 x1 = __xfrm_state_locate(x, use_spi, family);
1069 if (use_spi && x->km.seq) {
1070 x1 = __xfrm_find_acq_byseq(net, x->km.seq);
1071 if (x1 && ((x1->id.proto != x->id.proto) ||
1072 xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1079 x1 = __find_acq_core(net, family, x->props.mode, x->props.reqid,
1081 &x->id.daddr, &x->props.saddr, 0);
1083 __xfrm_state_bump_genids(x);
1084 __xfrm_state_insert(x);
1088 spin_unlock_bh(&xfrm_state_lock);
1091 xfrm_state_delete(x1);
1096 xfrm_state_put(to_put);
1100 EXPORT_SYMBOL(xfrm_state_add);
1102 #ifdef CONFIG_XFRM_MIGRATE
1103 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1105 struct net *net = xs_net(orig);
1107 struct xfrm_state *x = xfrm_state_alloc(net);
1111 memcpy(&x->id, &orig->id, sizeof(x->id));
1112 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1113 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1114 x->props.mode = orig->props.mode;
1115 x->props.replay_window = orig->props.replay_window;
1116 x->props.reqid = orig->props.reqid;
1117 x->props.family = orig->props.family;
1118 x->props.saddr = orig->props.saddr;
1121 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1125 x->props.aalgo = orig->props.aalgo;
1128 x->ealg = xfrm_algo_clone(orig->ealg);
1132 x->props.ealgo = orig->props.ealgo;
1135 x->calg = xfrm_algo_clone(orig->calg);
1139 x->props.calgo = orig->props.calgo;
1142 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1148 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1154 err = xfrm_init_state(x);
1158 x->props.flags = orig->props.flags;
1160 x->curlft.add_time = orig->curlft.add_time;
1161 x->km.state = orig->km.state;
1162 x->km.seq = orig->km.seq;
1174 /* xfrm_state_lock is held */
1175 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1178 struct xfrm_state *x;
1179 struct hlist_node *entry;
1182 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1183 m->reqid, m->old_family);
1184 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1185 if (x->props.mode != m->mode ||
1186 x->id.proto != m->proto)
1188 if (m->reqid && x->props.reqid != m->reqid)
1190 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1192 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1199 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1201 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1202 if (x->props.mode != m->mode ||
1203 x->id.proto != m->proto)
1205 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1207 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1217 EXPORT_SYMBOL(xfrm_migrate_state_find);
1219 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1220 struct xfrm_migrate *m)
1222 struct xfrm_state *xc;
1225 xc = xfrm_state_clone(x, &err);
1229 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1230 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1233 if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1234 /* a care is needed when the destination address of the
1235 state is to be updated as it is a part of triplet */
1236 xfrm_state_insert(xc);
1238 if ((err = xfrm_state_add(xc)) < 0)
1247 EXPORT_SYMBOL(xfrm_state_migrate);
1250 int xfrm_state_update(struct xfrm_state *x)
1252 struct xfrm_state *x1, *to_put;
1254 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1258 spin_lock_bh(&xfrm_state_lock);
1259 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1265 if (xfrm_state_kern(x1)) {
1271 if (x1->km.state == XFRM_STATE_ACQ) {
1272 __xfrm_state_insert(x);
1278 spin_unlock_bh(&xfrm_state_lock);
1281 xfrm_state_put(to_put);
1287 xfrm_state_delete(x1);
1293 spin_lock_bh(&x1->lock);
1294 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1295 if (x->encap && x1->encap)
1296 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1297 if (x->coaddr && x1->coaddr) {
1298 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1300 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1301 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1302 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1305 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1306 if (x1->curlft.use_time)
1307 xfrm_state_check_expire(x1);
1311 spin_unlock_bh(&x1->lock);
1317 EXPORT_SYMBOL(xfrm_state_update);
1319 int xfrm_state_check_expire(struct xfrm_state *x)
1321 if (!x->curlft.use_time)
1322 x->curlft.use_time = get_seconds();
1324 if (x->km.state != XFRM_STATE_VALID)
1327 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1328 x->curlft.packets >= x->lft.hard_packet_limit) {
1329 x->km.state = XFRM_STATE_EXPIRED;
1330 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1335 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1336 x->curlft.packets >= x->lft.soft_packet_limit)) {
1338 km_state_expired(x, 0, 0);
1342 EXPORT_SYMBOL(xfrm_state_check_expire);
1345 xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto,
1346 unsigned short family)
1348 struct xfrm_state *x;
1350 spin_lock_bh(&xfrm_state_lock);
1351 x = __xfrm_state_lookup(net, daddr, spi, proto, family);
1352 spin_unlock_bh(&xfrm_state_lock);
1355 EXPORT_SYMBOL(xfrm_state_lookup);
1358 xfrm_state_lookup_byaddr(struct net *net,
1359 xfrm_address_t *daddr, xfrm_address_t *saddr,
1360 u8 proto, unsigned short family)
1362 struct xfrm_state *x;
1364 spin_lock_bh(&xfrm_state_lock);
1365 x = __xfrm_state_lookup_byaddr(net, daddr, saddr, proto, family);
1366 spin_unlock_bh(&xfrm_state_lock);
1369 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1372 xfrm_find_acq(struct net *net, u8 mode, u32 reqid, u8 proto,
1373 xfrm_address_t *daddr, xfrm_address_t *saddr,
1374 int create, unsigned short family)
1376 struct xfrm_state *x;
1378 spin_lock_bh(&xfrm_state_lock);
1379 x = __find_acq_core(net, family, mode, reqid, proto, daddr, saddr, create);
1380 spin_unlock_bh(&xfrm_state_lock);
1384 EXPORT_SYMBOL(xfrm_find_acq);
1386 #ifdef CONFIG_XFRM_SUB_POLICY
1388 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1389 unsigned short family)
1392 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1394 return -EAFNOSUPPORT;
1396 spin_lock_bh(&xfrm_state_lock);
1397 if (afinfo->tmpl_sort)
1398 err = afinfo->tmpl_sort(dst, src, n);
1399 spin_unlock_bh(&xfrm_state_lock);
1400 xfrm_state_put_afinfo(afinfo);
1403 EXPORT_SYMBOL(xfrm_tmpl_sort);
1406 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1407 unsigned short family)
1410 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1412 return -EAFNOSUPPORT;
1414 spin_lock_bh(&xfrm_state_lock);
1415 if (afinfo->state_sort)
1416 err = afinfo->state_sort(dst, src, n);
1417 spin_unlock_bh(&xfrm_state_lock);
1418 xfrm_state_put_afinfo(afinfo);
1421 EXPORT_SYMBOL(xfrm_state_sort);
1424 /* Silly enough, but I'm lazy to build resolution list */
1426 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq)
1430 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1431 struct hlist_node *entry;
1432 struct xfrm_state *x;
1434 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1435 if (x->km.seq == seq &&
1436 x->km.state == XFRM_STATE_ACQ) {
1445 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 seq)
1447 struct xfrm_state *x;
1449 spin_lock_bh(&xfrm_state_lock);
1450 x = __xfrm_find_acq_byseq(net, seq);
1451 spin_unlock_bh(&xfrm_state_lock);
1454 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1456 u32 xfrm_get_acqseq(void)
1459 static atomic_t acqseq;
1462 res = atomic_inc_return(&acqseq);
1467 EXPORT_SYMBOL(xfrm_get_acqseq);
1469 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1471 struct net *net = xs_net(x);
1473 struct xfrm_state *x0;
1475 __be32 minspi = htonl(low);
1476 __be32 maxspi = htonl(high);
1478 spin_lock_bh(&x->lock);
1479 if (x->km.state == XFRM_STATE_DEAD)
1488 if (minspi == maxspi) {
1489 x0 = xfrm_state_lookup(net, &x->id.daddr, minspi, x->id.proto, x->props.family);
1497 for (h=0; h<high-low+1; h++) {
1498 spi = low + net_random()%(high-low+1);
1499 x0 = xfrm_state_lookup(net, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1501 x->id.spi = htonl(spi);
1508 spin_lock_bh(&xfrm_state_lock);
1509 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1510 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1511 spin_unlock_bh(&xfrm_state_lock);
1517 spin_unlock_bh(&x->lock);
1521 EXPORT_SYMBOL(xfrm_alloc_spi);
1523 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1524 int (*func)(struct xfrm_state *, int, void*),
1527 struct xfrm_state *state;
1528 struct xfrm_state_walk *x;
1531 if (walk->seq != 0 && list_empty(&walk->all))
1534 spin_lock_bh(&xfrm_state_lock);
1535 if (list_empty(&walk->all))
1536 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1538 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1539 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1540 if (x->state == XFRM_STATE_DEAD)
1542 state = container_of(x, struct xfrm_state, km);
1543 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1545 err = func(state, walk->seq, data);
1547 list_move_tail(&walk->all, &x->all);
1552 if (walk->seq == 0) {
1556 list_del_init(&walk->all);
1558 spin_unlock_bh(&xfrm_state_lock);
1561 EXPORT_SYMBOL(xfrm_state_walk);
1563 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1565 INIT_LIST_HEAD(&walk->all);
1566 walk->proto = proto;
1567 walk->state = XFRM_STATE_DEAD;
1570 EXPORT_SYMBOL(xfrm_state_walk_init);
1572 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1574 if (list_empty(&walk->all))
1577 spin_lock_bh(&xfrm_state_lock);
1578 list_del(&walk->all);
1579 spin_unlock_bh(&xfrm_state_lock);
1581 EXPORT_SYMBOL(xfrm_state_walk_done);
1584 void xfrm_replay_notify(struct xfrm_state *x, int event)
1587 /* we send notify messages in case
1588 * 1. we updated on of the sequence numbers, and the seqno difference
1589 * is at least x->replay_maxdiff, in this case we also update the
1590 * timeout of our timer function
1591 * 2. if x->replay_maxage has elapsed since last update,
1592 * and there were changes
1594 * The state structure must be locked!
1598 case XFRM_REPLAY_UPDATE:
1599 if (x->replay_maxdiff &&
1600 (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1601 (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1602 if (x->xflags & XFRM_TIME_DEFER)
1603 event = XFRM_REPLAY_TIMEOUT;
1610 case XFRM_REPLAY_TIMEOUT:
1611 if ((x->replay.seq == x->preplay.seq) &&
1612 (x->replay.bitmap == x->preplay.bitmap) &&
1613 (x->replay.oseq == x->preplay.oseq)) {
1614 x->xflags |= XFRM_TIME_DEFER;
1621 memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1622 c.event = XFRM_MSG_NEWAE;
1623 c.data.aevent = event;
1624 km_state_notify(x, &c);
1626 if (x->replay_maxage &&
1627 !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1628 x->xflags &= ~XFRM_TIME_DEFER;
1631 static void xfrm_replay_timer_handler(unsigned long data)
1633 struct xfrm_state *x = (struct xfrm_state*)data;
1635 spin_lock(&x->lock);
1637 if (x->km.state == XFRM_STATE_VALID) {
1638 if (xfrm_aevent_is_on(xs_net(x)))
1639 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1641 x->xflags |= XFRM_TIME_DEFER;
1644 spin_unlock(&x->lock);
1647 int xfrm_replay_check(struct xfrm_state *x,
1648 struct sk_buff *skb, __be32 net_seq)
1651 u32 seq = ntohl(net_seq);
1653 if (unlikely(seq == 0))
1656 if (likely(seq > x->replay.seq))
1659 diff = x->replay.seq - seq;
1660 if (diff >= min_t(unsigned int, x->props.replay_window,
1661 sizeof(x->replay.bitmap) * 8)) {
1662 x->stats.replay_window++;
1666 if (x->replay.bitmap & (1U << diff)) {
1673 xfrm_audit_state_replay(x, skb, net_seq);
1677 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1680 u32 seq = ntohl(net_seq);
1682 if (seq > x->replay.seq) {
1683 diff = seq - x->replay.seq;
1684 if (diff < x->props.replay_window)
1685 x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1687 x->replay.bitmap = 1;
1688 x->replay.seq = seq;
1690 diff = x->replay.seq - seq;
1691 x->replay.bitmap |= (1U << diff);
1694 if (xfrm_aevent_is_on(xs_net(x)))
1695 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1698 static LIST_HEAD(xfrm_km_list);
1699 static DEFINE_RWLOCK(xfrm_km_lock);
1701 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1703 struct xfrm_mgr *km;
1705 read_lock(&xfrm_km_lock);
1706 list_for_each_entry(km, &xfrm_km_list, list)
1707 if (km->notify_policy)
1708 km->notify_policy(xp, dir, c);
1709 read_unlock(&xfrm_km_lock);
1712 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1714 struct xfrm_mgr *km;
1715 read_lock(&xfrm_km_lock);
1716 list_for_each_entry(km, &xfrm_km_list, list)
1719 read_unlock(&xfrm_km_lock);
1722 EXPORT_SYMBOL(km_policy_notify);
1723 EXPORT_SYMBOL(km_state_notify);
1725 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1727 struct net *net = xs_net(x);
1732 c.event = XFRM_MSG_EXPIRE;
1733 km_state_notify(x, &c);
1736 wake_up(&net->xfrm.km_waitq);
1739 EXPORT_SYMBOL(km_state_expired);
1741 * We send to all registered managers regardless of failure
1742 * We are happy with one success
1744 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1746 int err = -EINVAL, acqret;
1747 struct xfrm_mgr *km;
1749 read_lock(&xfrm_km_lock);
1750 list_for_each_entry(km, &xfrm_km_list, list) {
1751 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1755 read_unlock(&xfrm_km_lock);
1758 EXPORT_SYMBOL(km_query);
1760 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1763 struct xfrm_mgr *km;
1765 read_lock(&xfrm_km_lock);
1766 list_for_each_entry(km, &xfrm_km_list, list) {
1767 if (km->new_mapping)
1768 err = km->new_mapping(x, ipaddr, sport);
1772 read_unlock(&xfrm_km_lock);
1775 EXPORT_SYMBOL(km_new_mapping);
1777 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1779 struct net *net = xp_net(pol);
1784 c.event = XFRM_MSG_POLEXPIRE;
1785 km_policy_notify(pol, dir, &c);
1788 wake_up(&net->xfrm.km_waitq);
1790 EXPORT_SYMBOL(km_policy_expired);
1792 #ifdef CONFIG_XFRM_MIGRATE
1793 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1794 struct xfrm_migrate *m, int num_migrate,
1795 struct xfrm_kmaddress *k)
1799 struct xfrm_mgr *km;
1801 read_lock(&xfrm_km_lock);
1802 list_for_each_entry(km, &xfrm_km_list, list) {
1804 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1809 read_unlock(&xfrm_km_lock);
1812 EXPORT_SYMBOL(km_migrate);
1815 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1819 struct xfrm_mgr *km;
1821 read_lock(&xfrm_km_lock);
1822 list_for_each_entry(km, &xfrm_km_list, list) {
1824 ret = km->report(net, proto, sel, addr);
1829 read_unlock(&xfrm_km_lock);
1832 EXPORT_SYMBOL(km_report);
1834 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1838 struct xfrm_mgr *km;
1839 struct xfrm_policy *pol = NULL;
1841 if (optlen <= 0 || optlen > PAGE_SIZE)
1844 data = kmalloc(optlen, GFP_KERNEL);
1849 if (copy_from_user(data, optval, optlen))
1853 read_lock(&xfrm_km_lock);
1854 list_for_each_entry(km, &xfrm_km_list, list) {
1855 pol = km->compile_policy(sk, optname, data,
1860 read_unlock(&xfrm_km_lock);
1863 xfrm_sk_policy_insert(sk, err, pol);
1872 EXPORT_SYMBOL(xfrm_user_policy);
1874 int xfrm_register_km(struct xfrm_mgr *km)
1876 write_lock_bh(&xfrm_km_lock);
1877 list_add_tail(&km->list, &xfrm_km_list);
1878 write_unlock_bh(&xfrm_km_lock);
1881 EXPORT_SYMBOL(xfrm_register_km);
1883 int xfrm_unregister_km(struct xfrm_mgr *km)
1885 write_lock_bh(&xfrm_km_lock);
1886 list_del(&km->list);
1887 write_unlock_bh(&xfrm_km_lock);
1890 EXPORT_SYMBOL(xfrm_unregister_km);
1892 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1895 if (unlikely(afinfo == NULL))
1897 if (unlikely(afinfo->family >= NPROTO))
1898 return -EAFNOSUPPORT;
1899 write_lock_bh(&xfrm_state_afinfo_lock);
1900 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1903 xfrm_state_afinfo[afinfo->family] = afinfo;
1904 write_unlock_bh(&xfrm_state_afinfo_lock);
1907 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1909 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1912 if (unlikely(afinfo == NULL))
1914 if (unlikely(afinfo->family >= NPROTO))
1915 return -EAFNOSUPPORT;
1916 write_lock_bh(&xfrm_state_afinfo_lock);
1917 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1918 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1921 xfrm_state_afinfo[afinfo->family] = NULL;
1923 write_unlock_bh(&xfrm_state_afinfo_lock);
1926 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1928 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1930 struct xfrm_state_afinfo *afinfo;
1931 if (unlikely(family >= NPROTO))
1933 read_lock(&xfrm_state_afinfo_lock);
1934 afinfo = xfrm_state_afinfo[family];
1935 if (unlikely(!afinfo))
1936 read_unlock(&xfrm_state_afinfo_lock);
1940 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1941 __releases(xfrm_state_afinfo_lock)
1943 read_unlock(&xfrm_state_afinfo_lock);
1946 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1947 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1950 struct xfrm_state *t = x->tunnel;
1952 if (atomic_read(&t->tunnel_users) == 2)
1953 xfrm_state_delete(t);
1954 atomic_dec(&t->tunnel_users);
1959 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1961 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1965 spin_lock_bh(&x->lock);
1966 if (x->km.state == XFRM_STATE_VALID &&
1967 x->type && x->type->get_mtu)
1968 res = x->type->get_mtu(x, mtu);
1970 res = mtu - x->props.header_len;
1971 spin_unlock_bh(&x->lock);
1975 int xfrm_init_state(struct xfrm_state *x)
1977 struct xfrm_state_afinfo *afinfo;
1978 struct xfrm_mode *inner_mode;
1979 int family = x->props.family;
1982 err = -EAFNOSUPPORT;
1983 afinfo = xfrm_state_get_afinfo(family);
1988 if (afinfo->init_flags)
1989 err = afinfo->init_flags(x);
1991 xfrm_state_put_afinfo(afinfo);
1996 err = -EPROTONOSUPPORT;
1998 if (x->sel.family != AF_UNSPEC) {
1999 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2000 if (inner_mode == NULL)
2003 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2004 family != x->sel.family) {
2005 xfrm_put_mode(inner_mode);
2009 x->inner_mode = inner_mode;
2011 struct xfrm_mode *inner_mode_iaf;
2012 int iafamily = AF_INET;
2014 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2015 if (inner_mode == NULL)
2018 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2019 xfrm_put_mode(inner_mode);
2022 x->inner_mode = inner_mode;
2024 if (x->props.family == AF_INET)
2025 iafamily = AF_INET6;
2027 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2028 if (inner_mode_iaf) {
2029 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2030 x->inner_mode_iaf = inner_mode_iaf;
2032 xfrm_put_mode(inner_mode_iaf);
2036 x->type = xfrm_get_type(x->id.proto, family);
2037 if (x->type == NULL)
2040 err = x->type->init_state(x);
2044 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2045 if (x->outer_mode == NULL)
2048 x->km.state = XFRM_STATE_VALID;
2054 EXPORT_SYMBOL(xfrm_init_state);
2056 int __net_init xfrm_state_init(struct net *net)
2060 INIT_LIST_HEAD(&net->xfrm.state_all);
2062 sz = sizeof(struct hlist_head) * 8;
2064 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2065 if (!net->xfrm.state_bydst)
2067 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2068 if (!net->xfrm.state_bysrc)
2070 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2071 if (!net->xfrm.state_byspi)
2073 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2075 net->xfrm.state_num = 0;
2076 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2077 INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2078 INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2079 init_waitqueue_head(&net->xfrm.km_waitq);
2083 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2085 xfrm_hash_free(net->xfrm.state_bydst, sz);
2090 void xfrm_state_fini(struct net *net)
2092 struct xfrm_audit audit_info;
2095 flush_work(&net->xfrm.state_hash_work);
2096 audit_info.loginuid = -1;
2097 audit_info.sessionid = -1;
2098 audit_info.secid = 0;
2099 xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2100 flush_work(&net->xfrm.state_gc_work);
2102 WARN_ON(!list_empty(&net->xfrm.state_all));
2104 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2105 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2106 xfrm_hash_free(net->xfrm.state_byspi, sz);
2107 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2108 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2109 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2110 xfrm_hash_free(net->xfrm.state_bydst, sz);
2113 #ifdef CONFIG_AUDITSYSCALL
2114 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2115 struct audit_buffer *audit_buf)
2117 struct xfrm_sec_ctx *ctx = x->security;
2118 u32 spi = ntohl(x->id.spi);
2121 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2122 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2124 switch(x->props.family) {
2126 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2127 &x->props.saddr.a4, &x->id.daddr.a4);
2130 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2131 x->props.saddr.a6, x->id.daddr.a6);
2135 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2138 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2139 struct audit_buffer *audit_buf)
2142 struct ipv6hdr *iph6;
2147 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2148 &iph4->saddr, &iph4->daddr);
2151 iph6 = ipv6_hdr(skb);
2152 audit_log_format(audit_buf,
2153 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2154 &iph6->saddr,&iph6->daddr,
2155 iph6->flow_lbl[0] & 0x0f,
2162 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2163 uid_t auid, u32 sessionid, u32 secid)
2165 struct audit_buffer *audit_buf;
2167 audit_buf = xfrm_audit_start("SAD-add");
2168 if (audit_buf == NULL)
2170 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2171 xfrm_audit_helper_sainfo(x, audit_buf);
2172 audit_log_format(audit_buf, " res=%u", result);
2173 audit_log_end(audit_buf);
2175 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2177 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2178 uid_t auid, u32 sessionid, u32 secid)
2180 struct audit_buffer *audit_buf;
2182 audit_buf = xfrm_audit_start("SAD-delete");
2183 if (audit_buf == NULL)
2185 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2186 xfrm_audit_helper_sainfo(x, audit_buf);
2187 audit_log_format(audit_buf, " res=%u", result);
2188 audit_log_end(audit_buf);
2190 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2192 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2193 struct sk_buff *skb)
2195 struct audit_buffer *audit_buf;
2198 audit_buf = xfrm_audit_start("SA-replay-overflow");
2199 if (audit_buf == NULL)
2201 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2202 /* don't record the sequence number because it's inherent in this kind
2203 * of audit message */
2204 spi = ntohl(x->id.spi);
2205 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2206 audit_log_end(audit_buf);
2208 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2210 static void xfrm_audit_state_replay(struct xfrm_state *x,
2211 struct sk_buff *skb, __be32 net_seq)
2213 struct audit_buffer *audit_buf;
2216 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2217 if (audit_buf == NULL)
2219 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2220 spi = ntohl(x->id.spi);
2221 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2222 spi, spi, ntohl(net_seq));
2223 audit_log_end(audit_buf);
2226 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2228 struct audit_buffer *audit_buf;
2230 audit_buf = xfrm_audit_start("SA-notfound");
2231 if (audit_buf == NULL)
2233 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2234 audit_log_end(audit_buf);
2236 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2238 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2239 __be32 net_spi, __be32 net_seq)
2241 struct audit_buffer *audit_buf;
2244 audit_buf = xfrm_audit_start("SA-notfound");
2245 if (audit_buf == NULL)
2247 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2248 spi = ntohl(net_spi);
2249 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2250 spi, spi, ntohl(net_seq));
2251 audit_log_end(audit_buf);
2253 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2255 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2256 struct sk_buff *skb, u8 proto)
2258 struct audit_buffer *audit_buf;
2262 audit_buf = xfrm_audit_start("SA-icv-failure");
2263 if (audit_buf == NULL)
2265 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2266 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2267 u32 spi = ntohl(net_spi);
2268 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2269 spi, spi, ntohl(net_seq));
2271 audit_log_end(audit_buf);
2273 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2274 #endif /* CONFIG_AUDITSYSCALL */