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 inline unsigned int xfrm_dst_hash(struct net *net,
43 const xfrm_address_t *daddr,
44 const xfrm_address_t *saddr,
46 unsigned short family)
48 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
51 static inline unsigned int xfrm_src_hash(struct net *net,
52 const xfrm_address_t *daddr,
53 const xfrm_address_t *saddr,
54 unsigned short family)
56 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
59 static inline unsigned int
60 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
61 __be32 spi, u8 proto, unsigned short family)
63 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
66 static void xfrm_hash_transfer(struct hlist_head *list,
67 struct hlist_head *ndsttable,
68 struct hlist_head *nsrctable,
69 struct hlist_head *nspitable,
70 unsigned int nhashmask)
72 struct hlist_node *tmp;
75 hlist_for_each_entry_safe(x, tmp, list, bydst) {
78 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
79 x->props.reqid, x->props.family,
81 hlist_add_head(&x->bydst, ndsttable+h);
83 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
86 hlist_add_head(&x->bysrc, nsrctable+h);
89 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
90 x->id.proto, x->props.family,
92 hlist_add_head(&x->byspi, nspitable+h);
97 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
99 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
102 static DEFINE_MUTEX(hash_resize_mutex);
104 static void xfrm_hash_resize(struct work_struct *work)
106 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
107 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
108 unsigned long nsize, osize;
109 unsigned int nhashmask, ohashmask;
112 mutex_lock(&hash_resize_mutex);
114 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
115 ndst = xfrm_hash_alloc(nsize);
118 nsrc = xfrm_hash_alloc(nsize);
120 xfrm_hash_free(ndst, nsize);
123 nspi = xfrm_hash_alloc(nsize);
125 xfrm_hash_free(ndst, nsize);
126 xfrm_hash_free(nsrc, nsize);
130 spin_lock_bh(&xfrm_state_lock);
132 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
133 for (i = net->xfrm.state_hmask; i >= 0; i--)
134 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
137 odst = net->xfrm.state_bydst;
138 osrc = net->xfrm.state_bysrc;
139 ospi = net->xfrm.state_byspi;
140 ohashmask = net->xfrm.state_hmask;
142 net->xfrm.state_bydst = ndst;
143 net->xfrm.state_bysrc = nsrc;
144 net->xfrm.state_byspi = nspi;
145 net->xfrm.state_hmask = nhashmask;
147 spin_unlock_bh(&xfrm_state_lock);
149 osize = (ohashmask + 1) * sizeof(struct hlist_head);
150 xfrm_hash_free(odst, osize);
151 xfrm_hash_free(osrc, osize);
152 xfrm_hash_free(ospi, osize);
155 mutex_unlock(&hash_resize_mutex);
158 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
159 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
161 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
163 int __xfrm_state_delete(struct xfrm_state *x);
165 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
166 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
168 static DEFINE_SPINLOCK(xfrm_type_lock);
169 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
171 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
172 const struct xfrm_type **typemap;
175 if (unlikely(afinfo == NULL))
176 return -EAFNOSUPPORT;
177 typemap = afinfo->type_map;
178 spin_lock_bh(&xfrm_type_lock);
180 if (likely(typemap[type->proto] == NULL))
181 typemap[type->proto] = type;
184 spin_unlock_bh(&xfrm_type_lock);
185 xfrm_state_put_afinfo(afinfo);
188 EXPORT_SYMBOL(xfrm_register_type);
190 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
192 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
193 const struct xfrm_type **typemap;
196 if (unlikely(afinfo == NULL))
197 return -EAFNOSUPPORT;
198 typemap = afinfo->type_map;
199 spin_lock_bh(&xfrm_type_lock);
201 if (unlikely(typemap[type->proto] != type))
204 typemap[type->proto] = NULL;
205 spin_unlock_bh(&xfrm_type_lock);
206 xfrm_state_put_afinfo(afinfo);
209 EXPORT_SYMBOL(xfrm_unregister_type);
211 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
213 struct xfrm_state_afinfo *afinfo;
214 const struct xfrm_type **typemap;
215 const struct xfrm_type *type;
216 int modload_attempted = 0;
219 afinfo = xfrm_state_get_afinfo(family);
220 if (unlikely(afinfo == NULL))
222 typemap = afinfo->type_map;
224 type = typemap[proto];
225 if (unlikely(type && !try_module_get(type->owner)))
227 if (!type && !modload_attempted) {
228 xfrm_state_put_afinfo(afinfo);
229 request_module("xfrm-type-%d-%d", family, proto);
230 modload_attempted = 1;
234 xfrm_state_put_afinfo(afinfo);
238 static void xfrm_put_type(const struct xfrm_type *type)
240 module_put(type->owner);
243 static DEFINE_SPINLOCK(xfrm_mode_lock);
244 int xfrm_register_mode(struct xfrm_mode *mode, int family)
246 struct xfrm_state_afinfo *afinfo;
247 struct xfrm_mode **modemap;
250 if (unlikely(mode->encap >= XFRM_MODE_MAX))
253 afinfo = xfrm_state_get_afinfo(family);
254 if (unlikely(afinfo == NULL))
255 return -EAFNOSUPPORT;
258 modemap = afinfo->mode_map;
259 spin_lock_bh(&xfrm_mode_lock);
260 if (modemap[mode->encap])
264 if (!try_module_get(afinfo->owner))
267 mode->afinfo = afinfo;
268 modemap[mode->encap] = mode;
272 spin_unlock_bh(&xfrm_mode_lock);
273 xfrm_state_put_afinfo(afinfo);
276 EXPORT_SYMBOL(xfrm_register_mode);
278 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
280 struct xfrm_state_afinfo *afinfo;
281 struct xfrm_mode **modemap;
284 if (unlikely(mode->encap >= XFRM_MODE_MAX))
287 afinfo = xfrm_state_get_afinfo(family);
288 if (unlikely(afinfo == NULL))
289 return -EAFNOSUPPORT;
292 modemap = afinfo->mode_map;
293 spin_lock_bh(&xfrm_mode_lock);
294 if (likely(modemap[mode->encap] == mode)) {
295 modemap[mode->encap] = NULL;
296 module_put(mode->afinfo->owner);
300 spin_unlock_bh(&xfrm_mode_lock);
301 xfrm_state_put_afinfo(afinfo);
304 EXPORT_SYMBOL(xfrm_unregister_mode);
306 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
308 struct xfrm_state_afinfo *afinfo;
309 struct xfrm_mode *mode;
310 int modload_attempted = 0;
312 if (unlikely(encap >= XFRM_MODE_MAX))
316 afinfo = xfrm_state_get_afinfo(family);
317 if (unlikely(afinfo == NULL))
320 mode = afinfo->mode_map[encap];
321 if (unlikely(mode && !try_module_get(mode->owner)))
323 if (!mode && !modload_attempted) {
324 xfrm_state_put_afinfo(afinfo);
325 request_module("xfrm-mode-%d-%d", family, encap);
326 modload_attempted = 1;
330 xfrm_state_put_afinfo(afinfo);
334 static void xfrm_put_mode(struct xfrm_mode *mode)
336 module_put(mode->owner);
339 static void xfrm_state_gc_destroy(struct xfrm_state *x)
341 tasklet_hrtimer_cancel(&x->mtimer);
342 del_timer_sync(&x->rtimer);
348 kfree(x->replay_esn);
349 kfree(x->preplay_esn);
351 xfrm_put_mode(x->inner_mode);
352 if (x->inner_mode_iaf)
353 xfrm_put_mode(x->inner_mode_iaf);
355 xfrm_put_mode(x->outer_mode);
357 x->type->destructor(x);
358 xfrm_put_type(x->type);
360 security_xfrm_state_free(x);
364 static void xfrm_state_gc_task(struct work_struct *work)
366 struct net *net = container_of(work, struct net, xfrm.state_gc_work);
367 struct xfrm_state *x;
368 struct hlist_node *tmp;
369 struct hlist_head gc_list;
371 spin_lock_bh(&xfrm_state_gc_lock);
372 hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
373 spin_unlock_bh(&xfrm_state_gc_lock);
375 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
376 xfrm_state_gc_destroy(x);
378 wake_up(&net->xfrm.km_waitq);
381 static inline unsigned long make_jiffies(long secs)
383 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
384 return MAX_SCHEDULE_TIMEOUT-1;
389 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
391 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
392 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
393 struct net *net = xs_net(x);
394 unsigned long now = get_seconds();
395 long next = LONG_MAX;
400 if (x->km.state == XFRM_STATE_DEAD)
402 if (x->km.state == XFRM_STATE_EXPIRED)
404 if (x->lft.hard_add_expires_seconds) {
405 long tmo = x->lft.hard_add_expires_seconds +
406 x->curlft.add_time - now;
408 if (x->xflags & XFRM_SOFT_EXPIRE) {
409 /* enter hard expire without soft expire first?!
410 * setting a new date could trigger this.
411 * workarbound: fix x->curflt.add_time by below:
413 x->curlft.add_time = now - x->saved_tmo - 1;
414 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
421 if (x->lft.hard_use_expires_seconds) {
422 long tmo = x->lft.hard_use_expires_seconds +
423 (x->curlft.use_time ? : now) - now;
431 if (x->lft.soft_add_expires_seconds) {
432 long tmo = x->lft.soft_add_expires_seconds +
433 x->curlft.add_time - now;
436 x->xflags &= ~XFRM_SOFT_EXPIRE;
437 } else if (tmo < next) {
439 x->xflags |= XFRM_SOFT_EXPIRE;
443 if (x->lft.soft_use_expires_seconds) {
444 long tmo = x->lft.soft_use_expires_seconds +
445 (x->curlft.use_time ? : now) - now;
454 km_state_expired(x, 0, 0);
456 if (next != LONG_MAX){
457 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
463 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
464 x->km.state = XFRM_STATE_EXPIRED;
465 wake_up(&net->xfrm.km_waitq);
470 err = __xfrm_state_delete(x);
472 km_state_expired(x, 1, 0);
474 xfrm_audit_state_delete(x, err ? 0 : 1,
475 audit_get_loginuid(current),
476 audit_get_sessionid(current), 0);
479 spin_unlock(&x->lock);
480 return HRTIMER_NORESTART;
483 static void xfrm_replay_timer_handler(unsigned long data);
485 struct xfrm_state *xfrm_state_alloc(struct net *net)
487 struct xfrm_state *x;
489 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
492 write_pnet(&x->xs_net, net);
493 atomic_set(&x->refcnt, 1);
494 atomic_set(&x->tunnel_users, 0);
495 INIT_LIST_HEAD(&x->km.all);
496 INIT_HLIST_NODE(&x->bydst);
497 INIT_HLIST_NODE(&x->bysrc);
498 INIT_HLIST_NODE(&x->byspi);
499 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
500 CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
501 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
503 x->curlft.add_time = get_seconds();
504 x->lft.soft_byte_limit = XFRM_INF;
505 x->lft.soft_packet_limit = XFRM_INF;
506 x->lft.hard_byte_limit = XFRM_INF;
507 x->lft.hard_packet_limit = XFRM_INF;
508 x->replay_maxage = 0;
509 x->replay_maxdiff = 0;
510 x->inner_mode = NULL;
511 x->inner_mode_iaf = NULL;
512 spin_lock_init(&x->lock);
516 EXPORT_SYMBOL(xfrm_state_alloc);
518 void __xfrm_state_destroy(struct xfrm_state *x)
520 struct net *net = xs_net(x);
522 WARN_ON(x->km.state != XFRM_STATE_DEAD);
524 spin_lock_bh(&xfrm_state_gc_lock);
525 hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
526 spin_unlock_bh(&xfrm_state_gc_lock);
527 schedule_work(&net->xfrm.state_gc_work);
529 EXPORT_SYMBOL(__xfrm_state_destroy);
531 int __xfrm_state_delete(struct xfrm_state *x)
533 struct net *net = xs_net(x);
536 if (x->km.state != XFRM_STATE_DEAD) {
537 x->km.state = XFRM_STATE_DEAD;
538 spin_lock(&xfrm_state_lock);
539 list_del(&x->km.all);
540 hlist_del(&x->bydst);
541 hlist_del(&x->bysrc);
543 hlist_del(&x->byspi);
544 net->xfrm.state_num--;
545 spin_unlock(&xfrm_state_lock);
547 /* All xfrm_state objects are created by xfrm_state_alloc.
548 * The xfrm_state_alloc call gives a reference, and that
549 * is what we are dropping here.
557 EXPORT_SYMBOL(__xfrm_state_delete);
559 int xfrm_state_delete(struct xfrm_state *x)
563 spin_lock_bh(&x->lock);
564 err = __xfrm_state_delete(x);
565 spin_unlock_bh(&x->lock);
569 EXPORT_SYMBOL(xfrm_state_delete);
571 #ifdef CONFIG_SECURITY_NETWORK_XFRM
573 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
577 for (i = 0; i <= net->xfrm.state_hmask; i++) {
578 struct xfrm_state *x;
580 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
581 if (xfrm_id_proto_match(x->id.proto, proto) &&
582 (err = security_xfrm_state_delete(x)) != 0) {
583 xfrm_audit_state_delete(x, 0,
584 audit_info->loginuid,
585 audit_info->sessionid,
596 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
602 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
604 int i, err = 0, cnt = 0;
606 spin_lock_bh(&xfrm_state_lock);
607 err = xfrm_state_flush_secctx_check(net, proto, audit_info);
612 for (i = 0; i <= net->xfrm.state_hmask; i++) {
613 struct xfrm_state *x;
615 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
616 if (!xfrm_state_kern(x) &&
617 xfrm_id_proto_match(x->id.proto, proto)) {
619 spin_unlock_bh(&xfrm_state_lock);
621 err = xfrm_state_delete(x);
622 xfrm_audit_state_delete(x, err ? 0 : 1,
623 audit_info->loginuid,
624 audit_info->sessionid,
630 spin_lock_bh(&xfrm_state_lock);
639 spin_unlock_bh(&xfrm_state_lock);
640 wake_up(&net->xfrm.km_waitq);
643 EXPORT_SYMBOL(xfrm_state_flush);
645 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
647 spin_lock_bh(&xfrm_state_lock);
648 si->sadcnt = net->xfrm.state_num;
649 si->sadhcnt = net->xfrm.state_hmask;
650 si->sadhmcnt = xfrm_state_hashmax;
651 spin_unlock_bh(&xfrm_state_lock);
653 EXPORT_SYMBOL(xfrm_sad_getinfo);
656 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
657 const struct xfrm_tmpl *tmpl,
658 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
659 unsigned short family)
661 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
664 afinfo->init_tempsel(&x->sel, fl);
666 if (family != tmpl->encap_family) {
667 xfrm_state_put_afinfo(afinfo);
668 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
672 afinfo->init_temprop(x, tmpl, daddr, saddr);
673 xfrm_state_put_afinfo(afinfo);
677 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
678 const xfrm_address_t *daddr,
679 __be32 spi, u8 proto,
680 unsigned short family)
682 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
683 struct xfrm_state *x;
685 hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
686 if (x->props.family != family ||
688 x->id.proto != proto ||
689 !xfrm_addr_equal(&x->id.daddr, daddr, family))
692 if ((mark & x->mark.m) != x->mark.v)
701 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
702 const xfrm_address_t *daddr,
703 const xfrm_address_t *saddr,
704 u8 proto, unsigned short family)
706 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
707 struct xfrm_state *x;
709 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
710 if (x->props.family != family ||
711 x->id.proto != proto ||
712 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
713 !xfrm_addr_equal(&x->props.saddr, saddr, family))
716 if ((mark & x->mark.m) != x->mark.v)
725 static inline struct xfrm_state *
726 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
728 struct net *net = xs_net(x);
729 u32 mark = x->mark.v & x->mark.m;
732 return __xfrm_state_lookup(net, mark, &x->id.daddr,
733 x->id.spi, x->id.proto, family);
735 return __xfrm_state_lookup_byaddr(net, mark,
738 x->id.proto, family);
741 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
743 if (have_hash_collision &&
744 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
745 net->xfrm.state_num > net->xfrm.state_hmask)
746 schedule_work(&net->xfrm.state_hash_work);
749 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
750 const struct flowi *fl, unsigned short family,
751 struct xfrm_state **best, int *acq_in_progress,
755 * 1. There is a valid state with matching selector. Done.
756 * 2. Valid state with inappropriate selector. Skip.
758 * Entering area of "sysdeps".
760 * 3. If state is not valid, selector is temporary, it selects
761 * only session which triggered previous resolution. Key
762 * manager will do something to install a state with proper
765 if (x->km.state == XFRM_STATE_VALID) {
766 if ((x->sel.family &&
767 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
768 !security_xfrm_state_pol_flow_match(x, pol, fl))
772 (*best)->km.dying > x->km.dying ||
773 ((*best)->km.dying == x->km.dying &&
774 (*best)->curlft.add_time < x->curlft.add_time))
776 } else if (x->km.state == XFRM_STATE_ACQ) {
777 *acq_in_progress = 1;
778 } else if (x->km.state == XFRM_STATE_ERROR ||
779 x->km.state == XFRM_STATE_EXPIRED) {
780 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
781 security_xfrm_state_pol_flow_match(x, pol, fl))
787 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
788 const struct flowi *fl, struct xfrm_tmpl *tmpl,
789 struct xfrm_policy *pol, int *err,
790 unsigned short family)
792 static xfrm_address_t saddr_wildcard = { };
793 struct net *net = xp_net(pol);
794 unsigned int h, h_wildcard;
795 struct xfrm_state *x, *x0, *to_put;
796 int acquire_in_progress = 0;
798 struct xfrm_state *best = NULL;
799 u32 mark = pol->mark.v & pol->mark.m;
800 unsigned short encap_family = tmpl->encap_family;
804 spin_lock_bh(&xfrm_state_lock);
805 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
806 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
807 if (x->props.family == encap_family &&
808 x->props.reqid == tmpl->reqid &&
809 (mark & x->mark.m) == x->mark.v &&
810 !(x->props.flags & XFRM_STATE_WILDRECV) &&
811 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
812 tmpl->mode == x->props.mode &&
813 tmpl->id.proto == x->id.proto &&
814 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
815 xfrm_state_look_at(pol, x, fl, encap_family,
816 &best, &acquire_in_progress, &error);
818 if (best || acquire_in_progress)
821 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
822 hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
823 if (x->props.family == encap_family &&
824 x->props.reqid == tmpl->reqid &&
825 (mark & x->mark.m) == x->mark.v &&
826 !(x->props.flags & XFRM_STATE_WILDRECV) &&
827 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
828 tmpl->mode == x->props.mode &&
829 tmpl->id.proto == x->id.proto &&
830 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
831 xfrm_state_look_at(pol, x, fl, encap_family,
832 &best, &acquire_in_progress, &error);
837 if (!x && !error && !acquire_in_progress) {
839 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
840 tmpl->id.proto, encap_family)) != NULL) {
845 x = xfrm_state_alloc(net);
850 /* Initialize temporary state matching only
851 * to current session. */
852 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
853 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
855 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
857 x->km.state = XFRM_STATE_DEAD;
863 if (km_query(x, tmpl, pol) == 0) {
864 x->km.state = XFRM_STATE_ACQ;
865 list_add(&x->km.all, &net->xfrm.state_all);
866 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
867 h = xfrm_src_hash(net, daddr, saddr, encap_family);
868 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
870 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
871 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
873 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
874 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
875 net->xfrm.state_num++;
876 xfrm_hash_grow_check(net, x->bydst.next != NULL);
878 x->km.state = XFRM_STATE_DEAD;
888 *err = acquire_in_progress ? -EAGAIN : error;
889 spin_unlock_bh(&xfrm_state_lock);
891 xfrm_state_put(to_put);
896 xfrm_stateonly_find(struct net *net, u32 mark,
897 xfrm_address_t *daddr, xfrm_address_t *saddr,
898 unsigned short family, u8 mode, u8 proto, u32 reqid)
901 struct xfrm_state *rx = NULL, *x = NULL;
903 spin_lock(&xfrm_state_lock);
904 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
905 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
906 if (x->props.family == family &&
907 x->props.reqid == reqid &&
908 (mark & x->mark.m) == x->mark.v &&
909 !(x->props.flags & XFRM_STATE_WILDRECV) &&
910 xfrm_state_addr_check(x, daddr, saddr, family) &&
911 mode == x->props.mode &&
912 proto == x->id.proto &&
913 x->km.state == XFRM_STATE_VALID) {
921 spin_unlock(&xfrm_state_lock);
926 EXPORT_SYMBOL(xfrm_stateonly_find);
928 static void __xfrm_state_insert(struct xfrm_state *x)
930 struct net *net = xs_net(x);
933 list_add(&x->km.all, &net->xfrm.state_all);
935 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
936 x->props.reqid, x->props.family);
937 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
939 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
940 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
943 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
946 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
949 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
950 if (x->replay_maxage)
951 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
953 wake_up(&net->xfrm.km_waitq);
955 net->xfrm.state_num++;
957 xfrm_hash_grow_check(net, x->bydst.next != NULL);
960 /* xfrm_state_lock is held */
961 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
963 struct net *net = xs_net(xnew);
964 unsigned short family = xnew->props.family;
965 u32 reqid = xnew->props.reqid;
966 struct xfrm_state *x;
968 u32 mark = xnew->mark.v & xnew->mark.m;
970 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
971 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
972 if (x->props.family == family &&
973 x->props.reqid == reqid &&
974 (mark & x->mark.m) == x->mark.v &&
975 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
976 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
981 void xfrm_state_insert(struct xfrm_state *x)
983 spin_lock_bh(&xfrm_state_lock);
984 __xfrm_state_bump_genids(x);
985 __xfrm_state_insert(x);
986 spin_unlock_bh(&xfrm_state_lock);
988 EXPORT_SYMBOL(xfrm_state_insert);
990 /* xfrm_state_lock is held */
991 static struct xfrm_state *__find_acq_core(struct net *net,
992 const struct xfrm_mark *m,
993 unsigned short family, u8 mode,
995 const xfrm_address_t *daddr,
996 const xfrm_address_t *saddr,
999 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1000 struct xfrm_state *x;
1001 u32 mark = m->v & m->m;
1003 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1004 if (x->props.reqid != reqid ||
1005 x->props.mode != mode ||
1006 x->props.family != family ||
1007 x->km.state != XFRM_STATE_ACQ ||
1009 x->id.proto != proto ||
1010 (mark & x->mark.m) != x->mark.v ||
1011 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1012 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1022 x = xfrm_state_alloc(net);
1026 x->sel.daddr.a4 = daddr->a4;
1027 x->sel.saddr.a4 = saddr->a4;
1028 x->sel.prefixlen_d = 32;
1029 x->sel.prefixlen_s = 32;
1030 x->props.saddr.a4 = saddr->a4;
1031 x->id.daddr.a4 = daddr->a4;
1035 *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1036 *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1037 x->sel.prefixlen_d = 128;
1038 x->sel.prefixlen_s = 128;
1039 *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1040 *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1044 x->km.state = XFRM_STATE_ACQ;
1045 x->id.proto = proto;
1046 x->props.family = family;
1047 x->props.mode = mode;
1048 x->props.reqid = reqid;
1051 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1053 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1054 list_add(&x->km.all, &net->xfrm.state_all);
1055 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1056 h = xfrm_src_hash(net, daddr, saddr, family);
1057 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1059 net->xfrm.state_num++;
1061 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1067 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1069 int xfrm_state_add(struct xfrm_state *x)
1071 struct net *net = xs_net(x);
1072 struct xfrm_state *x1, *to_put;
1075 u32 mark = x->mark.v & x->mark.m;
1076 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1078 family = x->props.family;
1082 spin_lock_bh(&xfrm_state_lock);
1084 x1 = __xfrm_state_locate(x, use_spi, family);
1092 if (use_spi && x->km.seq) {
1093 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1094 if (x1 && ((x1->id.proto != x->id.proto) ||
1095 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1102 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1103 x->props.reqid, x->id.proto,
1104 &x->id.daddr, &x->props.saddr, 0);
1106 __xfrm_state_bump_genids(x);
1107 __xfrm_state_insert(x);
1111 spin_unlock_bh(&xfrm_state_lock);
1114 xfrm_state_delete(x1);
1119 xfrm_state_put(to_put);
1123 EXPORT_SYMBOL(xfrm_state_add);
1125 #ifdef CONFIG_XFRM_MIGRATE
1126 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1128 struct net *net = xs_net(orig);
1130 struct xfrm_state *x = xfrm_state_alloc(net);
1134 memcpy(&x->id, &orig->id, sizeof(x->id));
1135 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1136 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1137 x->props.mode = orig->props.mode;
1138 x->props.replay_window = orig->props.replay_window;
1139 x->props.reqid = orig->props.reqid;
1140 x->props.family = orig->props.family;
1141 x->props.saddr = orig->props.saddr;
1144 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1148 x->props.aalgo = orig->props.aalgo;
1151 x->ealg = xfrm_algo_clone(orig->ealg);
1155 x->props.ealgo = orig->props.ealgo;
1158 x->calg = xfrm_algo_clone(orig->calg);
1162 x->props.calgo = orig->props.calgo;
1165 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1171 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1177 if (orig->replay_esn) {
1178 err = xfrm_replay_clone(x, orig);
1183 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1185 err = xfrm_init_state(x);
1189 x->props.flags = orig->props.flags;
1190 x->props.extra_flags = orig->props.extra_flags;
1192 x->curlft.add_time = orig->curlft.add_time;
1193 x->km.state = orig->km.state;
1194 x->km.seq = orig->km.seq;
1206 /* xfrm_state_lock is held */
1207 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1210 struct xfrm_state *x;
1213 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1214 m->reqid, m->old_family);
1215 hlist_for_each_entry(x, init_net.xfrm.state_bydst+h, bydst) {
1216 if (x->props.mode != m->mode ||
1217 x->id.proto != m->proto)
1219 if (m->reqid && x->props.reqid != m->reqid)
1221 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1223 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1230 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1232 hlist_for_each_entry(x, init_net.xfrm.state_bysrc+h, bysrc) {
1233 if (x->props.mode != m->mode ||
1234 x->id.proto != m->proto)
1236 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1238 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1248 EXPORT_SYMBOL(xfrm_migrate_state_find);
1250 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1251 struct xfrm_migrate *m)
1253 struct xfrm_state *xc;
1256 xc = xfrm_state_clone(x, &err);
1260 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1261 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1264 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1265 /* a care is needed when the destination address of the
1266 state is to be updated as it is a part of triplet */
1267 xfrm_state_insert(xc);
1269 if ((err = xfrm_state_add(xc)) < 0)
1278 EXPORT_SYMBOL(xfrm_state_migrate);
1281 int xfrm_state_update(struct xfrm_state *x)
1283 struct xfrm_state *x1, *to_put;
1285 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1289 spin_lock_bh(&xfrm_state_lock);
1290 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1296 if (xfrm_state_kern(x1)) {
1302 if (x1->km.state == XFRM_STATE_ACQ) {
1303 __xfrm_state_insert(x);
1309 spin_unlock_bh(&xfrm_state_lock);
1312 xfrm_state_put(to_put);
1318 xfrm_state_delete(x1);
1324 spin_lock_bh(&x1->lock);
1325 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1326 if (x->encap && x1->encap)
1327 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1328 if (x->coaddr && x1->coaddr) {
1329 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1331 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1332 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1333 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1336 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1337 if (x1->curlft.use_time)
1338 xfrm_state_check_expire(x1);
1341 x->km.state = XFRM_STATE_DEAD;
1342 __xfrm_state_put(x);
1344 spin_unlock_bh(&x1->lock);
1350 EXPORT_SYMBOL(xfrm_state_update);
1352 int xfrm_state_check_expire(struct xfrm_state *x)
1354 if (!x->curlft.use_time)
1355 x->curlft.use_time = get_seconds();
1357 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1358 x->curlft.packets >= x->lft.hard_packet_limit) {
1359 x->km.state = XFRM_STATE_EXPIRED;
1360 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1365 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1366 x->curlft.packets >= x->lft.soft_packet_limit)) {
1368 km_state_expired(x, 0, 0);
1372 EXPORT_SYMBOL(xfrm_state_check_expire);
1375 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1376 u8 proto, unsigned short family)
1378 struct xfrm_state *x;
1380 spin_lock_bh(&xfrm_state_lock);
1381 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1382 spin_unlock_bh(&xfrm_state_lock);
1385 EXPORT_SYMBOL(xfrm_state_lookup);
1388 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1389 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1390 u8 proto, unsigned short family)
1392 struct xfrm_state *x;
1394 spin_lock_bh(&xfrm_state_lock);
1395 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1396 spin_unlock_bh(&xfrm_state_lock);
1399 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1402 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1403 u8 proto, const xfrm_address_t *daddr,
1404 const xfrm_address_t *saddr, int create, unsigned short family)
1406 struct xfrm_state *x;
1408 spin_lock_bh(&xfrm_state_lock);
1409 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1410 spin_unlock_bh(&xfrm_state_lock);
1414 EXPORT_SYMBOL(xfrm_find_acq);
1416 #ifdef CONFIG_XFRM_SUB_POLICY
1418 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1419 unsigned short family)
1422 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1424 return -EAFNOSUPPORT;
1426 spin_lock_bh(&xfrm_state_lock);
1427 if (afinfo->tmpl_sort)
1428 err = afinfo->tmpl_sort(dst, src, n);
1429 spin_unlock_bh(&xfrm_state_lock);
1430 xfrm_state_put_afinfo(afinfo);
1433 EXPORT_SYMBOL(xfrm_tmpl_sort);
1436 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1437 unsigned short family)
1440 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1442 return -EAFNOSUPPORT;
1444 spin_lock_bh(&xfrm_state_lock);
1445 if (afinfo->state_sort)
1446 err = afinfo->state_sort(dst, src, n);
1447 spin_unlock_bh(&xfrm_state_lock);
1448 xfrm_state_put_afinfo(afinfo);
1451 EXPORT_SYMBOL(xfrm_state_sort);
1454 /* Silly enough, but I'm lazy to build resolution list */
1456 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1460 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1461 struct xfrm_state *x;
1463 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1464 if (x->km.seq == seq &&
1465 (mark & x->mark.m) == x->mark.v &&
1466 x->km.state == XFRM_STATE_ACQ) {
1475 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1477 struct xfrm_state *x;
1479 spin_lock_bh(&xfrm_state_lock);
1480 x = __xfrm_find_acq_byseq(net, mark, seq);
1481 spin_unlock_bh(&xfrm_state_lock);
1484 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1486 u32 xfrm_get_acqseq(void)
1489 static atomic_t acqseq;
1492 res = atomic_inc_return(&acqseq);
1497 EXPORT_SYMBOL(xfrm_get_acqseq);
1499 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1501 struct net *net = xs_net(x);
1503 struct xfrm_state *x0;
1505 __be32 minspi = htonl(low);
1506 __be32 maxspi = htonl(high);
1507 u32 mark = x->mark.v & x->mark.m;
1509 spin_lock_bh(&x->lock);
1510 if (x->km.state == XFRM_STATE_DEAD)
1519 if (minspi == maxspi) {
1520 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1528 for (h=0; h<high-low+1; h++) {
1529 spi = low + net_random()%(high-low+1);
1530 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1532 x->id.spi = htonl(spi);
1539 spin_lock_bh(&xfrm_state_lock);
1540 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1541 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1542 spin_unlock_bh(&xfrm_state_lock);
1548 spin_unlock_bh(&x->lock);
1552 EXPORT_SYMBOL(xfrm_alloc_spi);
1554 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1555 int (*func)(struct xfrm_state *, int, void*),
1558 struct xfrm_state *state;
1559 struct xfrm_state_walk *x;
1562 if (walk->seq != 0 && list_empty(&walk->all))
1565 spin_lock_bh(&xfrm_state_lock);
1566 if (list_empty(&walk->all))
1567 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1569 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1570 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1571 if (x->state == XFRM_STATE_DEAD)
1573 state = container_of(x, struct xfrm_state, km);
1574 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1576 err = func(state, walk->seq, data);
1578 list_move_tail(&walk->all, &x->all);
1583 if (walk->seq == 0) {
1587 list_del_init(&walk->all);
1589 spin_unlock_bh(&xfrm_state_lock);
1592 EXPORT_SYMBOL(xfrm_state_walk);
1594 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1596 INIT_LIST_HEAD(&walk->all);
1597 walk->proto = proto;
1598 walk->state = XFRM_STATE_DEAD;
1601 EXPORT_SYMBOL(xfrm_state_walk_init);
1603 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1605 if (list_empty(&walk->all))
1608 spin_lock_bh(&xfrm_state_lock);
1609 list_del(&walk->all);
1610 spin_unlock_bh(&xfrm_state_lock);
1612 EXPORT_SYMBOL(xfrm_state_walk_done);
1614 static void xfrm_replay_timer_handler(unsigned long data)
1616 struct xfrm_state *x = (struct xfrm_state*)data;
1618 spin_lock(&x->lock);
1620 if (x->km.state == XFRM_STATE_VALID) {
1621 if (xfrm_aevent_is_on(xs_net(x)))
1622 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1624 x->xflags |= XFRM_TIME_DEFER;
1627 spin_unlock(&x->lock);
1630 static LIST_HEAD(xfrm_km_list);
1632 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1634 struct xfrm_mgr *km;
1637 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1638 if (km->notify_policy)
1639 km->notify_policy(xp, dir, c);
1643 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1645 struct xfrm_mgr *km;
1647 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1653 EXPORT_SYMBOL(km_policy_notify);
1654 EXPORT_SYMBOL(km_state_notify);
1656 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1658 struct net *net = xs_net(x);
1663 c.event = XFRM_MSG_EXPIRE;
1664 km_state_notify(x, &c);
1667 wake_up(&net->xfrm.km_waitq);
1670 EXPORT_SYMBOL(km_state_expired);
1672 * We send to all registered managers regardless of failure
1673 * We are happy with one success
1675 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1677 int err = -EINVAL, acqret;
1678 struct xfrm_mgr *km;
1681 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1682 acqret = km->acquire(x, t, pol);
1689 EXPORT_SYMBOL(km_query);
1691 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1694 struct xfrm_mgr *km;
1697 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1698 if (km->new_mapping)
1699 err = km->new_mapping(x, ipaddr, sport);
1706 EXPORT_SYMBOL(km_new_mapping);
1708 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1710 struct net *net = xp_net(pol);
1715 c.event = XFRM_MSG_POLEXPIRE;
1716 km_policy_notify(pol, dir, &c);
1719 wake_up(&net->xfrm.km_waitq);
1721 EXPORT_SYMBOL(km_policy_expired);
1723 #ifdef CONFIG_XFRM_MIGRATE
1724 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1725 const struct xfrm_migrate *m, int num_migrate,
1726 const struct xfrm_kmaddress *k)
1730 struct xfrm_mgr *km;
1733 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1735 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1743 EXPORT_SYMBOL(km_migrate);
1746 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1750 struct xfrm_mgr *km;
1753 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1755 ret = km->report(net, proto, sel, addr);
1763 EXPORT_SYMBOL(km_report);
1765 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1769 struct xfrm_mgr *km;
1770 struct xfrm_policy *pol = NULL;
1772 if (optlen <= 0 || optlen > PAGE_SIZE)
1775 data = kmalloc(optlen, GFP_KERNEL);
1780 if (copy_from_user(data, optval, optlen))
1785 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1786 pol = km->compile_policy(sk, optname, data,
1794 xfrm_sk_policy_insert(sk, err, pol);
1803 EXPORT_SYMBOL(xfrm_user_policy);
1805 static DEFINE_SPINLOCK(xfrm_km_lock);
1807 int xfrm_register_km(struct xfrm_mgr *km)
1809 spin_lock_bh(&xfrm_km_lock);
1810 list_add_tail_rcu(&km->list, &xfrm_km_list);
1811 spin_unlock_bh(&xfrm_km_lock);
1814 EXPORT_SYMBOL(xfrm_register_km);
1816 int xfrm_unregister_km(struct xfrm_mgr *km)
1818 spin_lock_bh(&xfrm_km_lock);
1819 list_del_rcu(&km->list);
1820 spin_unlock_bh(&xfrm_km_lock);
1824 EXPORT_SYMBOL(xfrm_unregister_km);
1826 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1829 if (unlikely(afinfo == NULL))
1831 if (unlikely(afinfo->family >= NPROTO))
1832 return -EAFNOSUPPORT;
1833 spin_lock_bh(&xfrm_state_afinfo_lock);
1834 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1837 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1838 spin_unlock_bh(&xfrm_state_afinfo_lock);
1841 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1843 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1846 if (unlikely(afinfo == NULL))
1848 if (unlikely(afinfo->family >= NPROTO))
1849 return -EAFNOSUPPORT;
1850 spin_lock_bh(&xfrm_state_afinfo_lock);
1851 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1852 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1855 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1857 spin_unlock_bh(&xfrm_state_afinfo_lock);
1861 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1863 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1865 struct xfrm_state_afinfo *afinfo;
1866 if (unlikely(family >= NPROTO))
1869 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1870 if (unlikely(!afinfo))
1875 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1880 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1881 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1884 struct xfrm_state *t = x->tunnel;
1886 if (atomic_read(&t->tunnel_users) == 2)
1887 xfrm_state_delete(t);
1888 atomic_dec(&t->tunnel_users);
1893 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1895 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1899 spin_lock_bh(&x->lock);
1900 if (x->km.state == XFRM_STATE_VALID &&
1901 x->type && x->type->get_mtu)
1902 res = x->type->get_mtu(x, mtu);
1904 res = mtu - x->props.header_len;
1905 spin_unlock_bh(&x->lock);
1909 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1911 struct xfrm_state_afinfo *afinfo;
1912 struct xfrm_mode *inner_mode;
1913 int family = x->props.family;
1916 err = -EAFNOSUPPORT;
1917 afinfo = xfrm_state_get_afinfo(family);
1922 if (afinfo->init_flags)
1923 err = afinfo->init_flags(x);
1925 xfrm_state_put_afinfo(afinfo);
1930 err = -EPROTONOSUPPORT;
1932 if (x->sel.family != AF_UNSPEC) {
1933 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1934 if (inner_mode == NULL)
1937 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1938 family != x->sel.family) {
1939 xfrm_put_mode(inner_mode);
1943 x->inner_mode = inner_mode;
1945 struct xfrm_mode *inner_mode_iaf;
1946 int iafamily = AF_INET;
1948 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1949 if (inner_mode == NULL)
1952 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1953 xfrm_put_mode(inner_mode);
1956 x->inner_mode = inner_mode;
1958 if (x->props.family == AF_INET)
1959 iafamily = AF_INET6;
1961 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1962 if (inner_mode_iaf) {
1963 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1964 x->inner_mode_iaf = inner_mode_iaf;
1966 xfrm_put_mode(inner_mode_iaf);
1970 x->type = xfrm_get_type(x->id.proto, family);
1971 if (x->type == NULL)
1974 err = x->type->init_state(x);
1978 x->outer_mode = xfrm_get_mode(x->props.mode, family);
1979 if (x->outer_mode == NULL) {
1980 err = -EPROTONOSUPPORT;
1985 err = xfrm_init_replay(x);
1990 x->km.state = XFRM_STATE_VALID;
1996 EXPORT_SYMBOL(__xfrm_init_state);
1998 int xfrm_init_state(struct xfrm_state *x)
2000 return __xfrm_init_state(x, true);
2003 EXPORT_SYMBOL(xfrm_init_state);
2005 int __net_init xfrm_state_init(struct net *net)
2009 INIT_LIST_HEAD(&net->xfrm.state_all);
2011 sz = sizeof(struct hlist_head) * 8;
2013 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2014 if (!net->xfrm.state_bydst)
2016 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2017 if (!net->xfrm.state_bysrc)
2019 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2020 if (!net->xfrm.state_byspi)
2022 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2024 net->xfrm.state_num = 0;
2025 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2026 INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2027 INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2028 init_waitqueue_head(&net->xfrm.km_waitq);
2032 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2034 xfrm_hash_free(net->xfrm.state_bydst, sz);
2039 void xfrm_state_fini(struct net *net)
2041 struct xfrm_audit audit_info;
2044 flush_work(&net->xfrm.state_hash_work);
2045 audit_info.loginuid = INVALID_UID;
2046 audit_info.sessionid = -1;
2047 audit_info.secid = 0;
2048 xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2049 flush_work(&net->xfrm.state_gc_work);
2051 WARN_ON(!list_empty(&net->xfrm.state_all));
2053 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2054 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2055 xfrm_hash_free(net->xfrm.state_byspi, sz);
2056 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2057 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2058 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2059 xfrm_hash_free(net->xfrm.state_bydst, sz);
2062 #ifdef CONFIG_AUDITSYSCALL
2063 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2064 struct audit_buffer *audit_buf)
2066 struct xfrm_sec_ctx *ctx = x->security;
2067 u32 spi = ntohl(x->id.spi);
2070 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2071 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2073 switch(x->props.family) {
2075 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2076 &x->props.saddr.a4, &x->id.daddr.a4);
2079 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2080 x->props.saddr.a6, x->id.daddr.a6);
2084 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2087 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2088 struct audit_buffer *audit_buf)
2090 const struct iphdr *iph4;
2091 const struct ipv6hdr *iph6;
2096 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2097 &iph4->saddr, &iph4->daddr);
2100 iph6 = ipv6_hdr(skb);
2101 audit_log_format(audit_buf,
2102 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2103 &iph6->saddr,&iph6->daddr,
2104 iph6->flow_lbl[0] & 0x0f,
2111 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2112 kuid_t auid, u32 sessionid, u32 secid)
2114 struct audit_buffer *audit_buf;
2116 audit_buf = xfrm_audit_start("SAD-add");
2117 if (audit_buf == NULL)
2119 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2120 xfrm_audit_helper_sainfo(x, audit_buf);
2121 audit_log_format(audit_buf, " res=%u", result);
2122 audit_log_end(audit_buf);
2124 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2126 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2127 kuid_t auid, u32 sessionid, u32 secid)
2129 struct audit_buffer *audit_buf;
2131 audit_buf = xfrm_audit_start("SAD-delete");
2132 if (audit_buf == NULL)
2134 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2135 xfrm_audit_helper_sainfo(x, audit_buf);
2136 audit_log_format(audit_buf, " res=%u", result);
2137 audit_log_end(audit_buf);
2139 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2141 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2142 struct sk_buff *skb)
2144 struct audit_buffer *audit_buf;
2147 audit_buf = xfrm_audit_start("SA-replay-overflow");
2148 if (audit_buf == NULL)
2150 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2151 /* don't record the sequence number because it's inherent in this kind
2152 * of audit message */
2153 spi = ntohl(x->id.spi);
2154 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2155 audit_log_end(audit_buf);
2157 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2159 void xfrm_audit_state_replay(struct xfrm_state *x,
2160 struct sk_buff *skb, __be32 net_seq)
2162 struct audit_buffer *audit_buf;
2165 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2166 if (audit_buf == NULL)
2168 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2169 spi = ntohl(x->id.spi);
2170 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2171 spi, spi, ntohl(net_seq));
2172 audit_log_end(audit_buf);
2174 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2176 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2178 struct audit_buffer *audit_buf;
2180 audit_buf = xfrm_audit_start("SA-notfound");
2181 if (audit_buf == NULL)
2183 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2184 audit_log_end(audit_buf);
2186 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2188 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2189 __be32 net_spi, __be32 net_seq)
2191 struct audit_buffer *audit_buf;
2194 audit_buf = xfrm_audit_start("SA-notfound");
2195 if (audit_buf == NULL)
2197 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2198 spi = ntohl(net_spi);
2199 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2200 spi, spi, ntohl(net_seq));
2201 audit_log_end(audit_buf);
2203 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2205 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2206 struct sk_buff *skb, u8 proto)
2208 struct audit_buffer *audit_buf;
2212 audit_buf = xfrm_audit_start("SA-icv-failure");
2213 if (audit_buf == NULL)
2215 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2216 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2217 u32 spi = ntohl(net_spi);
2218 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2219 spi, spi, ntohl(net_seq));
2221 audit_log_end(audit_buf);
2223 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2224 #endif /* CONFIG_AUDITSYSCALL */