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/slab.h>
26 #include <linux/interrupt.h>
27 #include <linux/kernel.h>
29 #include "xfrm_hash.h"
31 /* Each xfrm_state may be linked to two tables:
33 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
34 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
35 destination/tunnel endpoint. (output)
38 static DEFINE_SPINLOCK(xfrm_state_lock);
40 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
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, u32 mark, 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))
685 if ((mark & x->mark.m) != x->mark.v)
694 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
696 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
697 struct xfrm_state *x;
698 struct hlist_node *entry;
700 hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
701 if (x->props.family != family ||
702 x->id.proto != proto ||
703 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
704 xfrm_addr_cmp(&x->props.saddr, saddr, family))
707 if ((mark & x->mark.m) != x->mark.v)
716 static inline struct xfrm_state *
717 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
719 struct net *net = xs_net(x);
720 u32 mark = x->mark.v & x->mark.m;
723 return __xfrm_state_lookup(net, mark, &x->id.daddr,
724 x->id.spi, x->id.proto, family);
726 return __xfrm_state_lookup_byaddr(net, mark,
729 x->id.proto, family);
732 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
734 if (have_hash_collision &&
735 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
736 net->xfrm.state_num > net->xfrm.state_hmask)
737 schedule_work(&net->xfrm.state_hash_work);
740 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
741 struct flowi *fl, unsigned short family,
742 xfrm_address_t *daddr, xfrm_address_t *saddr,
743 struct xfrm_state **best, int *acq_in_progress,
747 * 1. There is a valid state with matching selector. Done.
748 * 2. Valid state with inappropriate selector. Skip.
750 * Entering area of "sysdeps".
752 * 3. If state is not valid, selector is temporary, it selects
753 * only session which triggered previous resolution. Key
754 * manager will do something to install a state with proper
757 if (x->km.state == XFRM_STATE_VALID) {
758 if ((x->sel.family &&
759 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
760 !security_xfrm_state_pol_flow_match(x, pol, fl))
764 (*best)->km.dying > x->km.dying ||
765 ((*best)->km.dying == x->km.dying &&
766 (*best)->curlft.add_time < x->curlft.add_time))
768 } else if (x->km.state == XFRM_STATE_ACQ) {
769 *acq_in_progress = 1;
770 } else if (x->km.state == XFRM_STATE_ERROR ||
771 x->km.state == XFRM_STATE_EXPIRED) {
772 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
773 security_xfrm_state_pol_flow_match(x, pol, fl))
779 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
780 struct flowi *fl, struct xfrm_tmpl *tmpl,
781 struct xfrm_policy *pol, int *err,
782 unsigned short family)
784 static xfrm_address_t saddr_wildcard = { };
785 struct net *net = xp_net(pol);
786 unsigned int h, h_wildcard;
787 struct hlist_node *entry;
788 struct xfrm_state *x, *x0, *to_put;
789 int acquire_in_progress = 0;
791 struct xfrm_state *best = NULL;
792 u32 mark = pol->mark.v & pol->mark.m;
796 spin_lock_bh(&xfrm_state_lock);
797 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
798 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
799 if (x->props.family == family &&
800 x->props.reqid == tmpl->reqid &&
801 (mark & x->mark.m) == x->mark.v &&
802 !(x->props.flags & XFRM_STATE_WILDRECV) &&
803 xfrm_state_addr_check(x, daddr, saddr, family) &&
804 tmpl->mode == x->props.mode &&
805 tmpl->id.proto == x->id.proto &&
806 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
807 xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
808 &best, &acquire_in_progress, &error);
813 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
814 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
815 if (x->props.family == family &&
816 x->props.reqid == tmpl->reqid &&
817 (mark & x->mark.m) == x->mark.v &&
818 !(x->props.flags & XFRM_STATE_WILDRECV) &&
819 xfrm_state_addr_check(x, daddr, saddr, family) &&
820 tmpl->mode == x->props.mode &&
821 tmpl->id.proto == x->id.proto &&
822 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
823 xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
824 &best, &acquire_in_progress, &error);
829 if (!x && !error && !acquire_in_progress) {
831 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
832 tmpl->id.proto, family)) != NULL) {
837 x = xfrm_state_alloc(net);
842 /* Initialize temporary selector matching only
843 * to current session. */
844 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
845 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
847 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
849 x->km.state = XFRM_STATE_DEAD;
855 if (km_query(x, tmpl, pol) == 0) {
856 x->km.state = XFRM_STATE_ACQ;
857 list_add(&x->km.all, &net->xfrm.state_all);
858 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
859 h = xfrm_src_hash(net, daddr, saddr, family);
860 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
862 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
863 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
865 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
866 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
867 net->xfrm.state_num++;
868 xfrm_hash_grow_check(net, x->bydst.next != NULL);
870 x->km.state = XFRM_STATE_DEAD;
880 *err = acquire_in_progress ? -EAGAIN : error;
881 spin_unlock_bh(&xfrm_state_lock);
883 xfrm_state_put(to_put);
888 xfrm_stateonly_find(struct net *net, u32 mark,
889 xfrm_address_t *daddr, xfrm_address_t *saddr,
890 unsigned short family, u8 mode, u8 proto, u32 reqid)
893 struct xfrm_state *rx = NULL, *x = NULL;
894 struct hlist_node *entry;
896 spin_lock(&xfrm_state_lock);
897 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
898 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
899 if (x->props.family == family &&
900 x->props.reqid == reqid &&
901 (mark & x->mark.m) == x->mark.v &&
902 !(x->props.flags & XFRM_STATE_WILDRECV) &&
903 xfrm_state_addr_check(x, daddr, saddr, family) &&
904 mode == x->props.mode &&
905 proto == x->id.proto &&
906 x->km.state == XFRM_STATE_VALID) {
914 spin_unlock(&xfrm_state_lock);
919 EXPORT_SYMBOL(xfrm_stateonly_find);
921 static void __xfrm_state_insert(struct xfrm_state *x)
923 struct net *net = xs_net(x);
926 list_add(&x->km.all, &net->xfrm.state_all);
928 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
929 x->props.reqid, x->props.family);
930 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
932 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
933 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
936 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
939 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
942 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
943 if (x->replay_maxage)
944 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
946 wake_up(&net->xfrm.km_waitq);
948 net->xfrm.state_num++;
950 xfrm_hash_grow_check(net, x->bydst.next != NULL);
953 /* xfrm_state_lock is held */
954 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
956 struct net *net = xs_net(xnew);
957 unsigned short family = xnew->props.family;
958 u32 reqid = xnew->props.reqid;
959 struct xfrm_state *x;
960 struct hlist_node *entry;
962 u32 mark = xnew->mark.v & xnew->mark.m;
964 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
965 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
966 if (x->props.family == family &&
967 x->props.reqid == reqid &&
968 (mark & x->mark.m) == x->mark.v &&
969 !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
970 !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
975 void xfrm_state_insert(struct xfrm_state *x)
977 spin_lock_bh(&xfrm_state_lock);
978 __xfrm_state_bump_genids(x);
979 __xfrm_state_insert(x);
980 spin_unlock_bh(&xfrm_state_lock);
982 EXPORT_SYMBOL(xfrm_state_insert);
984 /* xfrm_state_lock is held */
985 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
987 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
988 struct hlist_node *entry;
989 struct xfrm_state *x;
990 u32 mark = m->v & m->m;
992 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
993 if (x->props.reqid != reqid ||
994 x->props.mode != mode ||
995 x->props.family != family ||
996 x->km.state != XFRM_STATE_ACQ ||
998 x->id.proto != proto ||
999 (mark & x->mark.m) != x->mark.v ||
1000 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1001 xfrm_addr_cmp(&x->props.saddr, saddr, family))
1011 x = xfrm_state_alloc(net);
1015 x->sel.daddr.a4 = daddr->a4;
1016 x->sel.saddr.a4 = saddr->a4;
1017 x->sel.prefixlen_d = 32;
1018 x->sel.prefixlen_s = 32;
1019 x->props.saddr.a4 = saddr->a4;
1020 x->id.daddr.a4 = daddr->a4;
1024 ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1025 (struct in6_addr *)daddr);
1026 ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1027 (struct in6_addr *)saddr);
1028 x->sel.prefixlen_d = 128;
1029 x->sel.prefixlen_s = 128;
1030 ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1031 (struct in6_addr *)saddr);
1032 ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1033 (struct in6_addr *)daddr);
1037 x->km.state = XFRM_STATE_ACQ;
1038 x->id.proto = proto;
1039 x->props.family = family;
1040 x->props.mode = mode;
1041 x->props.reqid = reqid;
1044 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1046 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1047 list_add(&x->km.all, &net->xfrm.state_all);
1048 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1049 h = xfrm_src_hash(net, daddr, saddr, family);
1050 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1052 net->xfrm.state_num++;
1054 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1060 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1062 int xfrm_state_add(struct xfrm_state *x)
1064 struct net *net = xs_net(x);
1065 struct xfrm_state *x1, *to_put;
1068 u32 mark = x->mark.v & x->mark.m;
1069 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1071 family = x->props.family;
1075 spin_lock_bh(&xfrm_state_lock);
1077 x1 = __xfrm_state_locate(x, use_spi, family);
1085 if (use_spi && x->km.seq) {
1086 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1087 if (x1 && ((x1->id.proto != x->id.proto) ||
1088 xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1095 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1096 x->props.reqid, x->id.proto,
1097 &x->id.daddr, &x->props.saddr, 0);
1099 __xfrm_state_bump_genids(x);
1100 __xfrm_state_insert(x);
1104 spin_unlock_bh(&xfrm_state_lock);
1107 xfrm_state_delete(x1);
1112 xfrm_state_put(to_put);
1116 EXPORT_SYMBOL(xfrm_state_add);
1118 #ifdef CONFIG_XFRM_MIGRATE
1119 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1121 struct net *net = xs_net(orig);
1123 struct xfrm_state *x = xfrm_state_alloc(net);
1127 memcpy(&x->id, &orig->id, sizeof(x->id));
1128 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1129 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1130 x->props.mode = orig->props.mode;
1131 x->props.replay_window = orig->props.replay_window;
1132 x->props.reqid = orig->props.reqid;
1133 x->props.family = orig->props.family;
1134 x->props.saddr = orig->props.saddr;
1137 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1141 x->props.aalgo = orig->props.aalgo;
1144 x->ealg = xfrm_algo_clone(orig->ealg);
1148 x->props.ealgo = orig->props.ealgo;
1151 x->calg = xfrm_algo_clone(orig->calg);
1155 x->props.calgo = orig->props.calgo;
1158 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1164 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1170 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1172 err = xfrm_init_state(x);
1176 x->props.flags = orig->props.flags;
1178 x->curlft.add_time = orig->curlft.add_time;
1179 x->km.state = orig->km.state;
1180 x->km.seq = orig->km.seq;
1192 /* xfrm_state_lock is held */
1193 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1196 struct xfrm_state *x;
1197 struct hlist_node *entry;
1200 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1201 m->reqid, m->old_family);
1202 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1203 if (x->props.mode != m->mode ||
1204 x->id.proto != m->proto)
1206 if (m->reqid && x->props.reqid != m->reqid)
1208 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1210 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1217 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1219 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1220 if (x->props.mode != m->mode ||
1221 x->id.proto != m->proto)
1223 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1225 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1235 EXPORT_SYMBOL(xfrm_migrate_state_find);
1237 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1238 struct xfrm_migrate *m)
1240 struct xfrm_state *xc;
1243 xc = xfrm_state_clone(x, &err);
1247 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1248 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1251 if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1252 /* a care is needed when the destination address of the
1253 state is to be updated as it is a part of triplet */
1254 xfrm_state_insert(xc);
1256 if ((err = xfrm_state_add(xc)) < 0)
1265 EXPORT_SYMBOL(xfrm_state_migrate);
1268 int xfrm_state_update(struct xfrm_state *x)
1270 struct xfrm_state *x1, *to_put;
1272 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1276 spin_lock_bh(&xfrm_state_lock);
1277 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1283 if (xfrm_state_kern(x1)) {
1289 if (x1->km.state == XFRM_STATE_ACQ) {
1290 __xfrm_state_insert(x);
1296 spin_unlock_bh(&xfrm_state_lock);
1299 xfrm_state_put(to_put);
1305 xfrm_state_delete(x1);
1311 spin_lock_bh(&x1->lock);
1312 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1313 if (x->encap && x1->encap)
1314 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1315 if (x->coaddr && x1->coaddr) {
1316 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1318 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1319 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1320 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1323 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1324 if (x1->curlft.use_time)
1325 xfrm_state_check_expire(x1);
1329 spin_unlock_bh(&x1->lock);
1335 EXPORT_SYMBOL(xfrm_state_update);
1337 int xfrm_state_check_expire(struct xfrm_state *x)
1339 if (!x->curlft.use_time)
1340 x->curlft.use_time = get_seconds();
1342 if (x->km.state != XFRM_STATE_VALID)
1345 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1346 x->curlft.packets >= x->lft.hard_packet_limit) {
1347 x->km.state = XFRM_STATE_EXPIRED;
1348 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1353 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1354 x->curlft.packets >= x->lft.soft_packet_limit)) {
1356 km_state_expired(x, 0, 0);
1360 EXPORT_SYMBOL(xfrm_state_check_expire);
1363 xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi,
1364 u8 proto, unsigned short family)
1366 struct xfrm_state *x;
1368 spin_lock_bh(&xfrm_state_lock);
1369 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1370 spin_unlock_bh(&xfrm_state_lock);
1373 EXPORT_SYMBOL(xfrm_state_lookup);
1376 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1377 xfrm_address_t *daddr, xfrm_address_t *saddr,
1378 u8 proto, unsigned short family)
1380 struct xfrm_state *x;
1382 spin_lock_bh(&xfrm_state_lock);
1383 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1384 spin_unlock_bh(&xfrm_state_lock);
1387 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1390 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1391 xfrm_address_t *daddr, xfrm_address_t *saddr,
1392 int create, unsigned short family)
1394 struct xfrm_state *x;
1396 spin_lock_bh(&xfrm_state_lock);
1397 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1398 spin_unlock_bh(&xfrm_state_lock);
1402 EXPORT_SYMBOL(xfrm_find_acq);
1404 #ifdef CONFIG_XFRM_SUB_POLICY
1406 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **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->tmpl_sort)
1416 err = afinfo->tmpl_sort(dst, src, n);
1417 spin_unlock_bh(&xfrm_state_lock);
1418 xfrm_state_put_afinfo(afinfo);
1421 EXPORT_SYMBOL(xfrm_tmpl_sort);
1424 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1425 unsigned short family)
1428 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1430 return -EAFNOSUPPORT;
1432 spin_lock_bh(&xfrm_state_lock);
1433 if (afinfo->state_sort)
1434 err = afinfo->state_sort(dst, src, n);
1435 spin_unlock_bh(&xfrm_state_lock);
1436 xfrm_state_put_afinfo(afinfo);
1439 EXPORT_SYMBOL(xfrm_state_sort);
1442 /* Silly enough, but I'm lazy to build resolution list */
1444 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1448 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1449 struct hlist_node *entry;
1450 struct xfrm_state *x;
1452 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1453 if (x->km.seq == seq &&
1454 (mark & x->mark.m) == x->mark.v &&
1455 x->km.state == XFRM_STATE_ACQ) {
1464 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1466 struct xfrm_state *x;
1468 spin_lock_bh(&xfrm_state_lock);
1469 x = __xfrm_find_acq_byseq(net, mark, seq);
1470 spin_unlock_bh(&xfrm_state_lock);
1473 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1475 u32 xfrm_get_acqseq(void)
1478 static atomic_t acqseq;
1481 res = atomic_inc_return(&acqseq);
1486 EXPORT_SYMBOL(xfrm_get_acqseq);
1488 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1490 struct net *net = xs_net(x);
1492 struct xfrm_state *x0;
1494 __be32 minspi = htonl(low);
1495 __be32 maxspi = htonl(high);
1496 u32 mark = x->mark.v & x->mark.m;
1498 spin_lock_bh(&x->lock);
1499 if (x->km.state == XFRM_STATE_DEAD)
1508 if (minspi == maxspi) {
1509 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1517 for (h=0; h<high-low+1; h++) {
1518 spi = low + net_random()%(high-low+1);
1519 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1521 x->id.spi = htonl(spi);
1528 spin_lock_bh(&xfrm_state_lock);
1529 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1530 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1531 spin_unlock_bh(&xfrm_state_lock);
1537 spin_unlock_bh(&x->lock);
1541 EXPORT_SYMBOL(xfrm_alloc_spi);
1543 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1544 int (*func)(struct xfrm_state *, int, void*),
1547 struct xfrm_state *state;
1548 struct xfrm_state_walk *x;
1551 if (walk->seq != 0 && list_empty(&walk->all))
1554 spin_lock_bh(&xfrm_state_lock);
1555 if (list_empty(&walk->all))
1556 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1558 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1559 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1560 if (x->state == XFRM_STATE_DEAD)
1562 state = container_of(x, struct xfrm_state, km);
1563 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1565 err = func(state, walk->seq, data);
1567 list_move_tail(&walk->all, &x->all);
1572 if (walk->seq == 0) {
1576 list_del_init(&walk->all);
1578 spin_unlock_bh(&xfrm_state_lock);
1581 EXPORT_SYMBOL(xfrm_state_walk);
1583 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1585 INIT_LIST_HEAD(&walk->all);
1586 walk->proto = proto;
1587 walk->state = XFRM_STATE_DEAD;
1590 EXPORT_SYMBOL(xfrm_state_walk_init);
1592 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1594 if (list_empty(&walk->all))
1597 spin_lock_bh(&xfrm_state_lock);
1598 list_del(&walk->all);
1599 spin_unlock_bh(&xfrm_state_lock);
1601 EXPORT_SYMBOL(xfrm_state_walk_done);
1604 void xfrm_replay_notify(struct xfrm_state *x, int event)
1607 /* we send notify messages in case
1608 * 1. we updated on of the sequence numbers, and the seqno difference
1609 * is at least x->replay_maxdiff, in this case we also update the
1610 * timeout of our timer function
1611 * 2. if x->replay_maxage has elapsed since last update,
1612 * and there were changes
1614 * The state structure must be locked!
1618 case XFRM_REPLAY_UPDATE:
1619 if (x->replay_maxdiff &&
1620 (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1621 (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1622 if (x->xflags & XFRM_TIME_DEFER)
1623 event = XFRM_REPLAY_TIMEOUT;
1630 case XFRM_REPLAY_TIMEOUT:
1631 if ((x->replay.seq == x->preplay.seq) &&
1632 (x->replay.bitmap == x->preplay.bitmap) &&
1633 (x->replay.oseq == x->preplay.oseq)) {
1634 x->xflags |= XFRM_TIME_DEFER;
1641 memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1642 c.event = XFRM_MSG_NEWAE;
1643 c.data.aevent = event;
1644 km_state_notify(x, &c);
1646 if (x->replay_maxage &&
1647 !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1648 x->xflags &= ~XFRM_TIME_DEFER;
1651 static void xfrm_replay_timer_handler(unsigned long data)
1653 struct xfrm_state *x = (struct xfrm_state*)data;
1655 spin_lock(&x->lock);
1657 if (x->km.state == XFRM_STATE_VALID) {
1658 if (xfrm_aevent_is_on(xs_net(x)))
1659 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1661 x->xflags |= XFRM_TIME_DEFER;
1664 spin_unlock(&x->lock);
1667 int xfrm_replay_check(struct xfrm_state *x,
1668 struct sk_buff *skb, __be32 net_seq)
1671 u32 seq = ntohl(net_seq);
1673 if (unlikely(seq == 0))
1676 if (likely(seq > x->replay.seq))
1679 diff = x->replay.seq - seq;
1680 if (diff >= min_t(unsigned int, x->props.replay_window,
1681 sizeof(x->replay.bitmap) * 8)) {
1682 x->stats.replay_window++;
1686 if (x->replay.bitmap & (1U << diff)) {
1693 xfrm_audit_state_replay(x, skb, net_seq);
1697 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1700 u32 seq = ntohl(net_seq);
1702 if (seq > x->replay.seq) {
1703 diff = seq - x->replay.seq;
1704 if (diff < x->props.replay_window)
1705 x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1707 x->replay.bitmap = 1;
1708 x->replay.seq = seq;
1710 diff = x->replay.seq - seq;
1711 x->replay.bitmap |= (1U << diff);
1714 if (xfrm_aevent_is_on(xs_net(x)))
1715 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1718 static LIST_HEAD(xfrm_km_list);
1719 static DEFINE_RWLOCK(xfrm_km_lock);
1721 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1723 struct xfrm_mgr *km;
1725 read_lock(&xfrm_km_lock);
1726 list_for_each_entry(km, &xfrm_km_list, list)
1727 if (km->notify_policy)
1728 km->notify_policy(xp, dir, c);
1729 read_unlock(&xfrm_km_lock);
1732 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1734 struct xfrm_mgr *km;
1735 read_lock(&xfrm_km_lock);
1736 list_for_each_entry(km, &xfrm_km_list, list)
1739 read_unlock(&xfrm_km_lock);
1742 EXPORT_SYMBOL(km_policy_notify);
1743 EXPORT_SYMBOL(km_state_notify);
1745 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1747 struct net *net = xs_net(x);
1752 c.event = XFRM_MSG_EXPIRE;
1753 km_state_notify(x, &c);
1756 wake_up(&net->xfrm.km_waitq);
1759 EXPORT_SYMBOL(km_state_expired);
1761 * We send to all registered managers regardless of failure
1762 * We are happy with one success
1764 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1766 int err = -EINVAL, acqret;
1767 struct xfrm_mgr *km;
1769 read_lock(&xfrm_km_lock);
1770 list_for_each_entry(km, &xfrm_km_list, list) {
1771 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1775 read_unlock(&xfrm_km_lock);
1778 EXPORT_SYMBOL(km_query);
1780 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1783 struct xfrm_mgr *km;
1785 read_lock(&xfrm_km_lock);
1786 list_for_each_entry(km, &xfrm_km_list, list) {
1787 if (km->new_mapping)
1788 err = km->new_mapping(x, ipaddr, sport);
1792 read_unlock(&xfrm_km_lock);
1795 EXPORT_SYMBOL(km_new_mapping);
1797 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1799 struct net *net = xp_net(pol);
1804 c.event = XFRM_MSG_POLEXPIRE;
1805 km_policy_notify(pol, dir, &c);
1808 wake_up(&net->xfrm.km_waitq);
1810 EXPORT_SYMBOL(km_policy_expired);
1812 #ifdef CONFIG_XFRM_MIGRATE
1813 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1814 struct xfrm_migrate *m, int num_migrate,
1815 struct xfrm_kmaddress *k)
1819 struct xfrm_mgr *km;
1821 read_lock(&xfrm_km_lock);
1822 list_for_each_entry(km, &xfrm_km_list, list) {
1824 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1829 read_unlock(&xfrm_km_lock);
1832 EXPORT_SYMBOL(km_migrate);
1835 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1839 struct xfrm_mgr *km;
1841 read_lock(&xfrm_km_lock);
1842 list_for_each_entry(km, &xfrm_km_list, list) {
1844 ret = km->report(net, proto, sel, addr);
1849 read_unlock(&xfrm_km_lock);
1852 EXPORT_SYMBOL(km_report);
1854 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1858 struct xfrm_mgr *km;
1859 struct xfrm_policy *pol = NULL;
1861 if (optlen <= 0 || optlen > PAGE_SIZE)
1864 data = kmalloc(optlen, GFP_KERNEL);
1869 if (copy_from_user(data, optval, optlen))
1873 read_lock(&xfrm_km_lock);
1874 list_for_each_entry(km, &xfrm_km_list, list) {
1875 pol = km->compile_policy(sk, optname, data,
1880 read_unlock(&xfrm_km_lock);
1883 xfrm_sk_policy_insert(sk, err, pol);
1892 EXPORT_SYMBOL(xfrm_user_policy);
1894 int xfrm_register_km(struct xfrm_mgr *km)
1896 write_lock_bh(&xfrm_km_lock);
1897 list_add_tail(&km->list, &xfrm_km_list);
1898 write_unlock_bh(&xfrm_km_lock);
1901 EXPORT_SYMBOL(xfrm_register_km);
1903 int xfrm_unregister_km(struct xfrm_mgr *km)
1905 write_lock_bh(&xfrm_km_lock);
1906 list_del(&km->list);
1907 write_unlock_bh(&xfrm_km_lock);
1910 EXPORT_SYMBOL(xfrm_unregister_km);
1912 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1915 if (unlikely(afinfo == NULL))
1917 if (unlikely(afinfo->family >= NPROTO))
1918 return -EAFNOSUPPORT;
1919 write_lock_bh(&xfrm_state_afinfo_lock);
1920 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1923 xfrm_state_afinfo[afinfo->family] = afinfo;
1924 write_unlock_bh(&xfrm_state_afinfo_lock);
1927 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1929 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1932 if (unlikely(afinfo == NULL))
1934 if (unlikely(afinfo->family >= NPROTO))
1935 return -EAFNOSUPPORT;
1936 write_lock_bh(&xfrm_state_afinfo_lock);
1937 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1938 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1941 xfrm_state_afinfo[afinfo->family] = NULL;
1943 write_unlock_bh(&xfrm_state_afinfo_lock);
1946 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1948 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1950 struct xfrm_state_afinfo *afinfo;
1951 if (unlikely(family >= NPROTO))
1953 read_lock(&xfrm_state_afinfo_lock);
1954 afinfo = xfrm_state_afinfo[family];
1955 if (unlikely(!afinfo))
1956 read_unlock(&xfrm_state_afinfo_lock);
1960 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1961 __releases(xfrm_state_afinfo_lock)
1963 read_unlock(&xfrm_state_afinfo_lock);
1966 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1967 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1970 struct xfrm_state *t = x->tunnel;
1972 if (atomic_read(&t->tunnel_users) == 2)
1973 xfrm_state_delete(t);
1974 atomic_dec(&t->tunnel_users);
1979 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1981 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1985 spin_lock_bh(&x->lock);
1986 if (x->km.state == XFRM_STATE_VALID &&
1987 x->type && x->type->get_mtu)
1988 res = x->type->get_mtu(x, mtu);
1990 res = mtu - x->props.header_len;
1991 spin_unlock_bh(&x->lock);
1995 int xfrm_init_state(struct xfrm_state *x)
1997 struct xfrm_state_afinfo *afinfo;
1998 struct xfrm_mode *inner_mode;
1999 int family = x->props.family;
2002 err = -EAFNOSUPPORT;
2003 afinfo = xfrm_state_get_afinfo(family);
2008 if (afinfo->init_flags)
2009 err = afinfo->init_flags(x);
2011 xfrm_state_put_afinfo(afinfo);
2016 err = -EPROTONOSUPPORT;
2018 if (x->sel.family != AF_UNSPEC) {
2019 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2020 if (inner_mode == NULL)
2023 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2024 family != x->sel.family) {
2025 xfrm_put_mode(inner_mode);
2029 x->inner_mode = inner_mode;
2031 struct xfrm_mode *inner_mode_iaf;
2032 int iafamily = AF_INET;
2034 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2035 if (inner_mode == NULL)
2038 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2039 xfrm_put_mode(inner_mode);
2042 x->inner_mode = inner_mode;
2044 if (x->props.family == AF_INET)
2045 iafamily = AF_INET6;
2047 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2048 if (inner_mode_iaf) {
2049 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2050 x->inner_mode_iaf = inner_mode_iaf;
2052 xfrm_put_mode(inner_mode_iaf);
2056 x->type = xfrm_get_type(x->id.proto, family);
2057 if (x->type == NULL)
2060 err = x->type->init_state(x);
2064 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2065 if (x->outer_mode == NULL)
2068 x->km.state = XFRM_STATE_VALID;
2074 EXPORT_SYMBOL(xfrm_init_state);
2076 int __net_init xfrm_state_init(struct net *net)
2080 INIT_LIST_HEAD(&net->xfrm.state_all);
2082 sz = sizeof(struct hlist_head) * 8;
2084 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2085 if (!net->xfrm.state_bydst)
2087 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2088 if (!net->xfrm.state_bysrc)
2090 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2091 if (!net->xfrm.state_byspi)
2093 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2095 net->xfrm.state_num = 0;
2096 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2097 INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2098 INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2099 init_waitqueue_head(&net->xfrm.km_waitq);
2103 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2105 xfrm_hash_free(net->xfrm.state_bydst, sz);
2110 void xfrm_state_fini(struct net *net)
2112 struct xfrm_audit audit_info;
2115 flush_work(&net->xfrm.state_hash_work);
2116 audit_info.loginuid = -1;
2117 audit_info.sessionid = -1;
2118 audit_info.secid = 0;
2119 xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2120 flush_work(&net->xfrm.state_gc_work);
2122 WARN_ON(!list_empty(&net->xfrm.state_all));
2124 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2125 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2126 xfrm_hash_free(net->xfrm.state_byspi, sz);
2127 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2128 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2129 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2130 xfrm_hash_free(net->xfrm.state_bydst, sz);
2133 #ifdef CONFIG_AUDITSYSCALL
2134 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2135 struct audit_buffer *audit_buf)
2137 struct xfrm_sec_ctx *ctx = x->security;
2138 u32 spi = ntohl(x->id.spi);
2141 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2142 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2144 switch(x->props.family) {
2146 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2147 &x->props.saddr.a4, &x->id.daddr.a4);
2150 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2151 x->props.saddr.a6, x->id.daddr.a6);
2155 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2159 struct audit_buffer *audit_buf)
2162 struct ipv6hdr *iph6;
2167 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2168 &iph4->saddr, &iph4->daddr);
2171 iph6 = ipv6_hdr(skb);
2172 audit_log_format(audit_buf,
2173 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2174 &iph6->saddr,&iph6->daddr,
2175 iph6->flow_lbl[0] & 0x0f,
2182 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2183 uid_t auid, u32 sessionid, u32 secid)
2185 struct audit_buffer *audit_buf;
2187 audit_buf = xfrm_audit_start("SAD-add");
2188 if (audit_buf == NULL)
2190 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2191 xfrm_audit_helper_sainfo(x, audit_buf);
2192 audit_log_format(audit_buf, " res=%u", result);
2193 audit_log_end(audit_buf);
2195 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2197 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2198 uid_t auid, u32 sessionid, u32 secid)
2200 struct audit_buffer *audit_buf;
2202 audit_buf = xfrm_audit_start("SAD-delete");
2203 if (audit_buf == NULL)
2205 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2206 xfrm_audit_helper_sainfo(x, audit_buf);
2207 audit_log_format(audit_buf, " res=%u", result);
2208 audit_log_end(audit_buf);
2210 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2212 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2213 struct sk_buff *skb)
2215 struct audit_buffer *audit_buf;
2218 audit_buf = xfrm_audit_start("SA-replay-overflow");
2219 if (audit_buf == NULL)
2221 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2222 /* don't record the sequence number because it's inherent in this kind
2223 * of audit message */
2224 spi = ntohl(x->id.spi);
2225 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2226 audit_log_end(audit_buf);
2228 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2230 static void xfrm_audit_state_replay(struct xfrm_state *x,
2231 struct sk_buff *skb, __be32 net_seq)
2233 struct audit_buffer *audit_buf;
2236 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2237 if (audit_buf == NULL)
2239 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2240 spi = ntohl(x->id.spi);
2241 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2242 spi, spi, ntohl(net_seq));
2243 audit_log_end(audit_buf);
2246 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2248 struct audit_buffer *audit_buf;
2250 audit_buf = xfrm_audit_start("SA-notfound");
2251 if (audit_buf == NULL)
2253 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2254 audit_log_end(audit_buf);
2256 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2258 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2259 __be32 net_spi, __be32 net_seq)
2261 struct audit_buffer *audit_buf;
2264 audit_buf = xfrm_audit_start("SA-notfound");
2265 if (audit_buf == NULL)
2267 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2268 spi = ntohl(net_spi);
2269 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2270 spi, spi, ntohl(net_seq));
2271 audit_log_end(audit_buf);
2273 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2275 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2276 struct sk_buff *skb, u8 proto)
2278 struct audit_buffer *audit_buf;
2282 audit_buf = xfrm_audit_start("SA-icv-failure");
2283 if (audit_buf == NULL)
2285 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2286 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2287 u32 spi = ntohl(net_spi);
2288 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2289 spi, spi, ntohl(net_seq));
2291 audit_log_end(audit_buf);
2293 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2294 #endif /* CONFIG_AUDITSYSCALL */