Loading...
1/*
2 * xfrm_state.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * YOSHIFUJI Hideaki @USAGI
10 * Split up af-specific functions
11 * Derek Atkins <derek@ihtfp.com>
12 * Add UDP Encapsulation
13 *
14 */
15
16#include <linux/workqueue.h>
17#include <net/xfrm.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>
28
29#include "xfrm_hash.h"
30
31/* Each xfrm_state may be linked to two tables:
32
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)
36 */
37
38static DEFINE_SPINLOCK(xfrm_state_lock);
39
40static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
41
42static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
43static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
44
45static inline unsigned int xfrm_dst_hash(struct net *net,
46 const xfrm_address_t *daddr,
47 const xfrm_address_t *saddr,
48 u32 reqid,
49 unsigned short family)
50{
51 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
52}
53
54static inline unsigned int xfrm_src_hash(struct net *net,
55 const xfrm_address_t *daddr,
56 const xfrm_address_t *saddr,
57 unsigned short family)
58{
59 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
60}
61
62static inline unsigned int
63xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
64 __be32 spi, u8 proto, unsigned short family)
65{
66 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
67}
68
69static void xfrm_hash_transfer(struct hlist_head *list,
70 struct hlist_head *ndsttable,
71 struct hlist_head *nsrctable,
72 struct hlist_head *nspitable,
73 unsigned int nhashmask)
74{
75 struct hlist_node *entry, *tmp;
76 struct xfrm_state *x;
77
78 hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
79 unsigned int h;
80
81 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
82 x->props.reqid, x->props.family,
83 nhashmask);
84 hlist_add_head(&x->bydst, ndsttable+h);
85
86 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
87 x->props.family,
88 nhashmask);
89 hlist_add_head(&x->bysrc, nsrctable+h);
90
91 if (x->id.spi) {
92 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
93 x->id.proto, x->props.family,
94 nhashmask);
95 hlist_add_head(&x->byspi, nspitable+h);
96 }
97 }
98}
99
100static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
101{
102 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
103}
104
105static DEFINE_MUTEX(hash_resize_mutex);
106
107static void xfrm_hash_resize(struct work_struct *work)
108{
109 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
110 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
111 unsigned long nsize, osize;
112 unsigned int nhashmask, ohashmask;
113 int i;
114
115 mutex_lock(&hash_resize_mutex);
116
117 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
118 ndst = xfrm_hash_alloc(nsize);
119 if (!ndst)
120 goto out_unlock;
121 nsrc = xfrm_hash_alloc(nsize);
122 if (!nsrc) {
123 xfrm_hash_free(ndst, nsize);
124 goto out_unlock;
125 }
126 nspi = xfrm_hash_alloc(nsize);
127 if (!nspi) {
128 xfrm_hash_free(ndst, nsize);
129 xfrm_hash_free(nsrc, nsize);
130 goto out_unlock;
131 }
132
133 spin_lock_bh(&xfrm_state_lock);
134
135 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
136 for (i = net->xfrm.state_hmask; i >= 0; i--)
137 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
138 nhashmask);
139
140 odst = net->xfrm.state_bydst;
141 osrc = net->xfrm.state_bysrc;
142 ospi = net->xfrm.state_byspi;
143 ohashmask = net->xfrm.state_hmask;
144
145 net->xfrm.state_bydst = ndst;
146 net->xfrm.state_bysrc = nsrc;
147 net->xfrm.state_byspi = nspi;
148 net->xfrm.state_hmask = nhashmask;
149
150 spin_unlock_bh(&xfrm_state_lock);
151
152 osize = (ohashmask + 1) * sizeof(struct hlist_head);
153 xfrm_hash_free(odst, osize);
154 xfrm_hash_free(osrc, osize);
155 xfrm_hash_free(ospi, osize);
156
157out_unlock:
158 mutex_unlock(&hash_resize_mutex);
159}
160
161static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
162static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
163
164static DEFINE_SPINLOCK(xfrm_state_gc_lock);
165
166int __xfrm_state_delete(struct xfrm_state *x);
167
168int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
169void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
170
171static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
172{
173 struct xfrm_state_afinfo *afinfo;
174 if (unlikely(family >= NPROTO))
175 return NULL;
176 write_lock_bh(&xfrm_state_afinfo_lock);
177 afinfo = xfrm_state_afinfo[family];
178 if (unlikely(!afinfo))
179 write_unlock_bh(&xfrm_state_afinfo_lock);
180 return afinfo;
181}
182
183static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
184 __releases(xfrm_state_afinfo_lock)
185{
186 write_unlock_bh(&xfrm_state_afinfo_lock);
187}
188
189int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
190{
191 struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
192 const struct xfrm_type **typemap;
193 int err = 0;
194
195 if (unlikely(afinfo == NULL))
196 return -EAFNOSUPPORT;
197 typemap = afinfo->type_map;
198
199 if (likely(typemap[type->proto] == NULL))
200 typemap[type->proto] = type;
201 else
202 err = -EEXIST;
203 xfrm_state_unlock_afinfo(afinfo);
204 return err;
205}
206EXPORT_SYMBOL(xfrm_register_type);
207
208int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
209{
210 struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
211 const struct xfrm_type **typemap;
212 int err = 0;
213
214 if (unlikely(afinfo == NULL))
215 return -EAFNOSUPPORT;
216 typemap = afinfo->type_map;
217
218 if (unlikely(typemap[type->proto] != type))
219 err = -ENOENT;
220 else
221 typemap[type->proto] = NULL;
222 xfrm_state_unlock_afinfo(afinfo);
223 return err;
224}
225EXPORT_SYMBOL(xfrm_unregister_type);
226
227static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
228{
229 struct xfrm_state_afinfo *afinfo;
230 const struct xfrm_type **typemap;
231 const struct xfrm_type *type;
232 int modload_attempted = 0;
233
234retry:
235 afinfo = xfrm_state_get_afinfo(family);
236 if (unlikely(afinfo == NULL))
237 return NULL;
238 typemap = afinfo->type_map;
239
240 type = typemap[proto];
241 if (unlikely(type && !try_module_get(type->owner)))
242 type = NULL;
243 if (!type && !modload_attempted) {
244 xfrm_state_put_afinfo(afinfo);
245 request_module("xfrm-type-%d-%d", family, proto);
246 modload_attempted = 1;
247 goto retry;
248 }
249
250 xfrm_state_put_afinfo(afinfo);
251 return type;
252}
253
254static void xfrm_put_type(const struct xfrm_type *type)
255{
256 module_put(type->owner);
257}
258
259int xfrm_register_mode(struct xfrm_mode *mode, int family)
260{
261 struct xfrm_state_afinfo *afinfo;
262 struct xfrm_mode **modemap;
263 int err;
264
265 if (unlikely(mode->encap >= XFRM_MODE_MAX))
266 return -EINVAL;
267
268 afinfo = xfrm_state_lock_afinfo(family);
269 if (unlikely(afinfo == NULL))
270 return -EAFNOSUPPORT;
271
272 err = -EEXIST;
273 modemap = afinfo->mode_map;
274 if (modemap[mode->encap])
275 goto out;
276
277 err = -ENOENT;
278 if (!try_module_get(afinfo->owner))
279 goto out;
280
281 mode->afinfo = afinfo;
282 modemap[mode->encap] = mode;
283 err = 0;
284
285out:
286 xfrm_state_unlock_afinfo(afinfo);
287 return err;
288}
289EXPORT_SYMBOL(xfrm_register_mode);
290
291int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
292{
293 struct xfrm_state_afinfo *afinfo;
294 struct xfrm_mode **modemap;
295 int err;
296
297 if (unlikely(mode->encap >= XFRM_MODE_MAX))
298 return -EINVAL;
299
300 afinfo = xfrm_state_lock_afinfo(family);
301 if (unlikely(afinfo == NULL))
302 return -EAFNOSUPPORT;
303
304 err = -ENOENT;
305 modemap = afinfo->mode_map;
306 if (likely(modemap[mode->encap] == mode)) {
307 modemap[mode->encap] = NULL;
308 module_put(mode->afinfo->owner);
309 err = 0;
310 }
311
312 xfrm_state_unlock_afinfo(afinfo);
313 return err;
314}
315EXPORT_SYMBOL(xfrm_unregister_mode);
316
317static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
318{
319 struct xfrm_state_afinfo *afinfo;
320 struct xfrm_mode *mode;
321 int modload_attempted = 0;
322
323 if (unlikely(encap >= XFRM_MODE_MAX))
324 return NULL;
325
326retry:
327 afinfo = xfrm_state_get_afinfo(family);
328 if (unlikely(afinfo == NULL))
329 return NULL;
330
331 mode = afinfo->mode_map[encap];
332 if (unlikely(mode && !try_module_get(mode->owner)))
333 mode = NULL;
334 if (!mode && !modload_attempted) {
335 xfrm_state_put_afinfo(afinfo);
336 request_module("xfrm-mode-%d-%d", family, encap);
337 modload_attempted = 1;
338 goto retry;
339 }
340
341 xfrm_state_put_afinfo(afinfo);
342 return mode;
343}
344
345static void xfrm_put_mode(struct xfrm_mode *mode)
346{
347 module_put(mode->owner);
348}
349
350static void xfrm_state_gc_destroy(struct xfrm_state *x)
351{
352 tasklet_hrtimer_cancel(&x->mtimer);
353 del_timer_sync(&x->rtimer);
354 kfree(x->aalg);
355 kfree(x->ealg);
356 kfree(x->calg);
357 kfree(x->encap);
358 kfree(x->coaddr);
359 kfree(x->replay_esn);
360 kfree(x->preplay_esn);
361 if (x->inner_mode)
362 xfrm_put_mode(x->inner_mode);
363 if (x->inner_mode_iaf)
364 xfrm_put_mode(x->inner_mode_iaf);
365 if (x->outer_mode)
366 xfrm_put_mode(x->outer_mode);
367 if (x->type) {
368 x->type->destructor(x);
369 xfrm_put_type(x->type);
370 }
371 security_xfrm_state_free(x);
372 kfree(x);
373}
374
375static void xfrm_state_gc_task(struct work_struct *work)
376{
377 struct net *net = container_of(work, struct net, xfrm.state_gc_work);
378 struct xfrm_state *x;
379 struct hlist_node *entry, *tmp;
380 struct hlist_head gc_list;
381
382 spin_lock_bh(&xfrm_state_gc_lock);
383 hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
384 spin_unlock_bh(&xfrm_state_gc_lock);
385
386 hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
387 xfrm_state_gc_destroy(x);
388
389 wake_up(&net->xfrm.km_waitq);
390}
391
392static inline unsigned long make_jiffies(long secs)
393{
394 if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
395 return MAX_SCHEDULE_TIMEOUT-1;
396 else
397 return secs*HZ;
398}
399
400static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
401{
402 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
403 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
404 struct net *net = xs_net(x);
405 unsigned long now = get_seconds();
406 long next = LONG_MAX;
407 int warn = 0;
408 int err = 0;
409
410 spin_lock(&x->lock);
411 if (x->km.state == XFRM_STATE_DEAD)
412 goto out;
413 if (x->km.state == XFRM_STATE_EXPIRED)
414 goto expired;
415 if (x->lft.hard_add_expires_seconds) {
416 long tmo = x->lft.hard_add_expires_seconds +
417 x->curlft.add_time - now;
418 if (tmo <= 0)
419 goto expired;
420 if (tmo < next)
421 next = tmo;
422 }
423 if (x->lft.hard_use_expires_seconds) {
424 long tmo = x->lft.hard_use_expires_seconds +
425 (x->curlft.use_time ? : now) - now;
426 if (tmo <= 0)
427 goto expired;
428 if (tmo < next)
429 next = tmo;
430 }
431 if (x->km.dying)
432 goto resched;
433 if (x->lft.soft_add_expires_seconds) {
434 long tmo = x->lft.soft_add_expires_seconds +
435 x->curlft.add_time - now;
436 if (tmo <= 0)
437 warn = 1;
438 else if (tmo < next)
439 next = tmo;
440 }
441 if (x->lft.soft_use_expires_seconds) {
442 long tmo = x->lft.soft_use_expires_seconds +
443 (x->curlft.use_time ? : now) - now;
444 if (tmo <= 0)
445 warn = 1;
446 else if (tmo < next)
447 next = tmo;
448 }
449
450 x->km.dying = warn;
451 if (warn)
452 km_state_expired(x, 0, 0);
453resched:
454 if (next != LONG_MAX){
455 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
456 }
457
458 goto out;
459
460expired:
461 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
462 x->km.state = XFRM_STATE_EXPIRED;
463 wake_up(&net->xfrm.km_waitq);
464 next = 2;
465 goto resched;
466 }
467
468 err = __xfrm_state_delete(x);
469 if (!err && x->id.spi)
470 km_state_expired(x, 1, 0);
471
472 xfrm_audit_state_delete(x, err ? 0 : 1,
473 audit_get_loginuid(current),
474 audit_get_sessionid(current), 0);
475
476out:
477 spin_unlock(&x->lock);
478 return HRTIMER_NORESTART;
479}
480
481static void xfrm_replay_timer_handler(unsigned long data);
482
483struct xfrm_state *xfrm_state_alloc(struct net *net)
484{
485 struct xfrm_state *x;
486
487 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
488
489 if (x) {
490 write_pnet(&x->xs_net, net);
491 atomic_set(&x->refcnt, 1);
492 atomic_set(&x->tunnel_users, 0);
493 INIT_LIST_HEAD(&x->km.all);
494 INIT_HLIST_NODE(&x->bydst);
495 INIT_HLIST_NODE(&x->bysrc);
496 INIT_HLIST_NODE(&x->byspi);
497 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
498 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
499 (unsigned long)x);
500 x->curlft.add_time = get_seconds();
501 x->lft.soft_byte_limit = XFRM_INF;
502 x->lft.soft_packet_limit = XFRM_INF;
503 x->lft.hard_byte_limit = XFRM_INF;
504 x->lft.hard_packet_limit = XFRM_INF;
505 x->replay_maxage = 0;
506 x->replay_maxdiff = 0;
507 x->inner_mode = NULL;
508 x->inner_mode_iaf = NULL;
509 spin_lock_init(&x->lock);
510 }
511 return x;
512}
513EXPORT_SYMBOL(xfrm_state_alloc);
514
515void __xfrm_state_destroy(struct xfrm_state *x)
516{
517 struct net *net = xs_net(x);
518
519 WARN_ON(x->km.state != XFRM_STATE_DEAD);
520
521 spin_lock_bh(&xfrm_state_gc_lock);
522 hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
523 spin_unlock_bh(&xfrm_state_gc_lock);
524 schedule_work(&net->xfrm.state_gc_work);
525}
526EXPORT_SYMBOL(__xfrm_state_destroy);
527
528int __xfrm_state_delete(struct xfrm_state *x)
529{
530 struct net *net = xs_net(x);
531 int err = -ESRCH;
532
533 if (x->km.state != XFRM_STATE_DEAD) {
534 x->km.state = XFRM_STATE_DEAD;
535 spin_lock(&xfrm_state_lock);
536 list_del(&x->km.all);
537 hlist_del(&x->bydst);
538 hlist_del(&x->bysrc);
539 if (x->id.spi)
540 hlist_del(&x->byspi);
541 net->xfrm.state_num--;
542 spin_unlock(&xfrm_state_lock);
543
544 /* All xfrm_state objects are created by xfrm_state_alloc.
545 * The xfrm_state_alloc call gives a reference, and that
546 * is what we are dropping here.
547 */
548 xfrm_state_put(x);
549 err = 0;
550 }
551
552 return err;
553}
554EXPORT_SYMBOL(__xfrm_state_delete);
555
556int xfrm_state_delete(struct xfrm_state *x)
557{
558 int err;
559
560 spin_lock_bh(&x->lock);
561 err = __xfrm_state_delete(x);
562 spin_unlock_bh(&x->lock);
563
564 return err;
565}
566EXPORT_SYMBOL(xfrm_state_delete);
567
568#ifdef CONFIG_SECURITY_NETWORK_XFRM
569static inline int
570xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
571{
572 int i, err = 0;
573
574 for (i = 0; i <= net->xfrm.state_hmask; i++) {
575 struct hlist_node *entry;
576 struct xfrm_state *x;
577
578 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
579 if (xfrm_id_proto_match(x->id.proto, proto) &&
580 (err = security_xfrm_state_delete(x)) != 0) {
581 xfrm_audit_state_delete(x, 0,
582 audit_info->loginuid,
583 audit_info->sessionid,
584 audit_info->secid);
585 return err;
586 }
587 }
588 }
589
590 return err;
591}
592#else
593static inline int
594xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
595{
596 return 0;
597}
598#endif
599
600int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
601{
602 int i, err = 0, cnt = 0;
603
604 spin_lock_bh(&xfrm_state_lock);
605 err = xfrm_state_flush_secctx_check(net, proto, audit_info);
606 if (err)
607 goto out;
608
609 err = -ESRCH;
610 for (i = 0; i <= net->xfrm.state_hmask; i++) {
611 struct hlist_node *entry;
612 struct xfrm_state *x;
613restart:
614 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
615 if (!xfrm_state_kern(x) &&
616 xfrm_id_proto_match(x->id.proto, proto)) {
617 xfrm_state_hold(x);
618 spin_unlock_bh(&xfrm_state_lock);
619
620 err = xfrm_state_delete(x);
621 xfrm_audit_state_delete(x, err ? 0 : 1,
622 audit_info->loginuid,
623 audit_info->sessionid,
624 audit_info->secid);
625 xfrm_state_put(x);
626 if (!err)
627 cnt++;
628
629 spin_lock_bh(&xfrm_state_lock);
630 goto restart;
631 }
632 }
633 }
634 if (cnt)
635 err = 0;
636
637out:
638 spin_unlock_bh(&xfrm_state_lock);
639 wake_up(&net->xfrm.km_waitq);
640 return err;
641}
642EXPORT_SYMBOL(xfrm_state_flush);
643
644void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
645{
646 spin_lock_bh(&xfrm_state_lock);
647 si->sadcnt = net->xfrm.state_num;
648 si->sadhcnt = net->xfrm.state_hmask;
649 si->sadhmcnt = xfrm_state_hashmax;
650 spin_unlock_bh(&xfrm_state_lock);
651}
652EXPORT_SYMBOL(xfrm_sad_getinfo);
653
654static int
655xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
656 const struct xfrm_tmpl *tmpl,
657 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
658 unsigned short family)
659{
660 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
661 if (!afinfo)
662 return -1;
663 afinfo->init_tempsel(&x->sel, fl);
664
665 if (family != tmpl->encap_family) {
666 xfrm_state_put_afinfo(afinfo);
667 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
668 if (!afinfo)
669 return -1;
670 }
671 afinfo->init_temprop(x, tmpl, daddr, saddr);
672 xfrm_state_put_afinfo(afinfo);
673 return 0;
674}
675
676static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
677 const xfrm_address_t *daddr,
678 __be32 spi, u8 proto,
679 unsigned short family)
680{
681 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
682 struct xfrm_state *x;
683 struct hlist_node *entry;
684
685 hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
686 if (x->props.family != family ||
687 x->id.spi != spi ||
688 x->id.proto != proto ||
689 xfrm_addr_cmp(&x->id.daddr, daddr, family))
690 continue;
691
692 if ((mark & x->mark.m) != x->mark.v)
693 continue;
694 xfrm_state_hold(x);
695 return x;
696 }
697
698 return NULL;
699}
700
701static 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)
705{
706 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
707 struct xfrm_state *x;
708 struct hlist_node *entry;
709
710 hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
711 if (x->props.family != family ||
712 x->id.proto != proto ||
713 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
714 xfrm_addr_cmp(&x->props.saddr, saddr, family))
715 continue;
716
717 if ((mark & x->mark.m) != x->mark.v)
718 continue;
719 xfrm_state_hold(x);
720 return x;
721 }
722
723 return NULL;
724}
725
726static inline struct xfrm_state *
727__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
728{
729 struct net *net = xs_net(x);
730 u32 mark = x->mark.v & x->mark.m;
731
732 if (use_spi)
733 return __xfrm_state_lookup(net, mark, &x->id.daddr,
734 x->id.spi, x->id.proto, family);
735 else
736 return __xfrm_state_lookup_byaddr(net, mark,
737 &x->id.daddr,
738 &x->props.saddr,
739 x->id.proto, family);
740}
741
742static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
743{
744 if (have_hash_collision &&
745 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
746 net->xfrm.state_num > net->xfrm.state_hmask)
747 schedule_work(&net->xfrm.state_hash_work);
748}
749
750static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
751 const struct flowi *fl, unsigned short family,
752 struct xfrm_state **best, int *acq_in_progress,
753 int *error)
754{
755 /* Resolution logic:
756 * 1. There is a valid state with matching selector. Done.
757 * 2. Valid state with inappropriate selector. Skip.
758 *
759 * Entering area of "sysdeps".
760 *
761 * 3. If state is not valid, selector is temporary, it selects
762 * only session which triggered previous resolution. Key
763 * manager will do something to install a state with proper
764 * selector.
765 */
766 if (x->km.state == XFRM_STATE_VALID) {
767 if ((x->sel.family &&
768 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
769 !security_xfrm_state_pol_flow_match(x, pol, fl))
770 return;
771
772 if (!*best ||
773 (*best)->km.dying > x->km.dying ||
774 ((*best)->km.dying == x->km.dying &&
775 (*best)->curlft.add_time < x->curlft.add_time))
776 *best = x;
777 } else if (x->km.state == XFRM_STATE_ACQ) {
778 *acq_in_progress = 1;
779 } else if (x->km.state == XFRM_STATE_ERROR ||
780 x->km.state == XFRM_STATE_EXPIRED) {
781 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
782 security_xfrm_state_pol_flow_match(x, pol, fl))
783 *error = -ESRCH;
784 }
785}
786
787struct xfrm_state *
788xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
789 const struct flowi *fl, struct xfrm_tmpl *tmpl,
790 struct xfrm_policy *pol, int *err,
791 unsigned short family)
792{
793 static xfrm_address_t saddr_wildcard = { };
794 struct net *net = xp_net(pol);
795 unsigned int h, h_wildcard;
796 struct hlist_node *entry;
797 struct xfrm_state *x, *x0, *to_put;
798 int acquire_in_progress = 0;
799 int error = 0;
800 struct xfrm_state *best = NULL;
801 u32 mark = pol->mark.v & pol->mark.m;
802 unsigned short encap_family = tmpl->encap_family;
803
804 to_put = NULL;
805
806 spin_lock_bh(&xfrm_state_lock);
807 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
808 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
809 if (x->props.family == encap_family &&
810 x->props.reqid == tmpl->reqid &&
811 (mark & x->mark.m) == x->mark.v &&
812 !(x->props.flags & XFRM_STATE_WILDRECV) &&
813 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
814 tmpl->mode == x->props.mode &&
815 tmpl->id.proto == x->id.proto &&
816 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
817 xfrm_state_look_at(pol, x, fl, encap_family,
818 &best, &acquire_in_progress, &error);
819 }
820 if (best)
821 goto found;
822
823 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
824 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
825 if (x->props.family == encap_family &&
826 x->props.reqid == tmpl->reqid &&
827 (mark & x->mark.m) == x->mark.v &&
828 !(x->props.flags & XFRM_STATE_WILDRECV) &&
829 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
830 tmpl->mode == x->props.mode &&
831 tmpl->id.proto == x->id.proto &&
832 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
833 xfrm_state_look_at(pol, x, fl, encap_family,
834 &best, &acquire_in_progress, &error);
835 }
836
837found:
838 x = best;
839 if (!x && !error && !acquire_in_progress) {
840 if (tmpl->id.spi &&
841 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
842 tmpl->id.proto, encap_family)) != NULL) {
843 to_put = x0;
844 error = -EEXIST;
845 goto out;
846 }
847 x = xfrm_state_alloc(net);
848 if (x == NULL) {
849 error = -ENOMEM;
850 goto out;
851 }
852 /* Initialize temporary state matching only
853 * to current session. */
854 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
855 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
856
857 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
858 if (error) {
859 x->km.state = XFRM_STATE_DEAD;
860 to_put = x;
861 x = NULL;
862 goto out;
863 }
864
865 if (km_query(x, tmpl, pol) == 0) {
866 x->km.state = XFRM_STATE_ACQ;
867 list_add(&x->km.all, &net->xfrm.state_all);
868 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
869 h = xfrm_src_hash(net, daddr, saddr, encap_family);
870 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
871 if (x->id.spi) {
872 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
873 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
874 }
875 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
876 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
877 net->xfrm.state_num++;
878 xfrm_hash_grow_check(net, x->bydst.next != NULL);
879 } else {
880 x->km.state = XFRM_STATE_DEAD;
881 to_put = x;
882 x = NULL;
883 error = -ESRCH;
884 }
885 }
886out:
887 if (x)
888 xfrm_state_hold(x);
889 else
890 *err = acquire_in_progress ? -EAGAIN : error;
891 spin_unlock_bh(&xfrm_state_lock);
892 if (to_put)
893 xfrm_state_put(to_put);
894 return x;
895}
896
897struct xfrm_state *
898xfrm_stateonly_find(struct net *net, u32 mark,
899 xfrm_address_t *daddr, xfrm_address_t *saddr,
900 unsigned short family, u8 mode, u8 proto, u32 reqid)
901{
902 unsigned int h;
903 struct xfrm_state *rx = NULL, *x = NULL;
904 struct hlist_node *entry;
905
906 spin_lock(&xfrm_state_lock);
907 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
908 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
909 if (x->props.family == family &&
910 x->props.reqid == reqid &&
911 (mark & x->mark.m) == x->mark.v &&
912 !(x->props.flags & XFRM_STATE_WILDRECV) &&
913 xfrm_state_addr_check(x, daddr, saddr, family) &&
914 mode == x->props.mode &&
915 proto == x->id.proto &&
916 x->km.state == XFRM_STATE_VALID) {
917 rx = x;
918 break;
919 }
920 }
921
922 if (rx)
923 xfrm_state_hold(rx);
924 spin_unlock(&xfrm_state_lock);
925
926
927 return rx;
928}
929EXPORT_SYMBOL(xfrm_stateonly_find);
930
931static void __xfrm_state_insert(struct xfrm_state *x)
932{
933 struct net *net = xs_net(x);
934 unsigned int h;
935
936 list_add(&x->km.all, &net->xfrm.state_all);
937
938 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
939 x->props.reqid, x->props.family);
940 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
941
942 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
943 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
944
945 if (x->id.spi) {
946 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
947 x->props.family);
948
949 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
950 }
951
952 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
953 if (x->replay_maxage)
954 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
955
956 wake_up(&net->xfrm.km_waitq);
957
958 net->xfrm.state_num++;
959
960 xfrm_hash_grow_check(net, x->bydst.next != NULL);
961}
962
963/* xfrm_state_lock is held */
964static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
965{
966 struct net *net = xs_net(xnew);
967 unsigned short family = xnew->props.family;
968 u32 reqid = xnew->props.reqid;
969 struct xfrm_state *x;
970 struct hlist_node *entry;
971 unsigned int h;
972 u32 mark = xnew->mark.v & xnew->mark.m;
973
974 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
975 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
976 if (x->props.family == family &&
977 x->props.reqid == reqid &&
978 (mark & x->mark.m) == x->mark.v &&
979 !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
980 !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
981 x->genid++;
982 }
983}
984
985void xfrm_state_insert(struct xfrm_state *x)
986{
987 spin_lock_bh(&xfrm_state_lock);
988 __xfrm_state_bump_genids(x);
989 __xfrm_state_insert(x);
990 spin_unlock_bh(&xfrm_state_lock);
991}
992EXPORT_SYMBOL(xfrm_state_insert);
993
994/* xfrm_state_lock is held */
995static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
996 unsigned short family, u8 mode,
997 u32 reqid, u8 proto,
998 const xfrm_address_t *daddr,
999 const xfrm_address_t *saddr, int create)
1000{
1001 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1002 struct hlist_node *entry;
1003 struct xfrm_state *x;
1004 u32 mark = m->v & m->m;
1005
1006 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1007 if (x->props.reqid != reqid ||
1008 x->props.mode != mode ||
1009 x->props.family != family ||
1010 x->km.state != XFRM_STATE_ACQ ||
1011 x->id.spi != 0 ||
1012 x->id.proto != proto ||
1013 (mark & x->mark.m) != x->mark.v ||
1014 xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1015 xfrm_addr_cmp(&x->props.saddr, saddr, family))
1016 continue;
1017
1018 xfrm_state_hold(x);
1019 return x;
1020 }
1021
1022 if (!create)
1023 return NULL;
1024
1025 x = xfrm_state_alloc(net);
1026 if (likely(x)) {
1027 switch (family) {
1028 case AF_INET:
1029 x->sel.daddr.a4 = daddr->a4;
1030 x->sel.saddr.a4 = saddr->a4;
1031 x->sel.prefixlen_d = 32;
1032 x->sel.prefixlen_s = 32;
1033 x->props.saddr.a4 = saddr->a4;
1034 x->id.daddr.a4 = daddr->a4;
1035 break;
1036
1037 case AF_INET6:
1038 *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1039 *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1040 x->sel.prefixlen_d = 128;
1041 x->sel.prefixlen_s = 128;
1042 *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1043 *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1044 break;
1045 }
1046
1047 x->km.state = XFRM_STATE_ACQ;
1048 x->id.proto = proto;
1049 x->props.family = family;
1050 x->props.mode = mode;
1051 x->props.reqid = reqid;
1052 x->mark.v = m->v;
1053 x->mark.m = m->m;
1054 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1055 xfrm_state_hold(x);
1056 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1057 list_add(&x->km.all, &net->xfrm.state_all);
1058 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1059 h = xfrm_src_hash(net, daddr, saddr, family);
1060 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1061
1062 net->xfrm.state_num++;
1063
1064 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1065 }
1066
1067 return x;
1068}
1069
1070static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1071
1072int xfrm_state_add(struct xfrm_state *x)
1073{
1074 struct net *net = xs_net(x);
1075 struct xfrm_state *x1, *to_put;
1076 int family;
1077 int err;
1078 u32 mark = x->mark.v & x->mark.m;
1079 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1080
1081 family = x->props.family;
1082
1083 to_put = NULL;
1084
1085 spin_lock_bh(&xfrm_state_lock);
1086
1087 x1 = __xfrm_state_locate(x, use_spi, family);
1088 if (x1) {
1089 to_put = x1;
1090 x1 = NULL;
1091 err = -EEXIST;
1092 goto out;
1093 }
1094
1095 if (use_spi && x->km.seq) {
1096 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1097 if (x1 && ((x1->id.proto != x->id.proto) ||
1098 xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1099 to_put = x1;
1100 x1 = NULL;
1101 }
1102 }
1103
1104 if (use_spi && !x1)
1105 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1106 x->props.reqid, x->id.proto,
1107 &x->id.daddr, &x->props.saddr, 0);
1108
1109 __xfrm_state_bump_genids(x);
1110 __xfrm_state_insert(x);
1111 err = 0;
1112
1113out:
1114 spin_unlock_bh(&xfrm_state_lock);
1115
1116 if (x1) {
1117 xfrm_state_delete(x1);
1118 xfrm_state_put(x1);
1119 }
1120
1121 if (to_put)
1122 xfrm_state_put(to_put);
1123
1124 return err;
1125}
1126EXPORT_SYMBOL(xfrm_state_add);
1127
1128#ifdef CONFIG_XFRM_MIGRATE
1129static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1130{
1131 struct net *net = xs_net(orig);
1132 int err = -ENOMEM;
1133 struct xfrm_state *x = xfrm_state_alloc(net);
1134 if (!x)
1135 goto out;
1136
1137 memcpy(&x->id, &orig->id, sizeof(x->id));
1138 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1139 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1140 x->props.mode = orig->props.mode;
1141 x->props.replay_window = orig->props.replay_window;
1142 x->props.reqid = orig->props.reqid;
1143 x->props.family = orig->props.family;
1144 x->props.saddr = orig->props.saddr;
1145
1146 if (orig->aalg) {
1147 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1148 if (!x->aalg)
1149 goto error;
1150 }
1151 x->props.aalgo = orig->props.aalgo;
1152
1153 if (orig->ealg) {
1154 x->ealg = xfrm_algo_clone(orig->ealg);
1155 if (!x->ealg)
1156 goto error;
1157 }
1158 x->props.ealgo = orig->props.ealgo;
1159
1160 if (orig->calg) {
1161 x->calg = xfrm_algo_clone(orig->calg);
1162 if (!x->calg)
1163 goto error;
1164 }
1165 x->props.calgo = orig->props.calgo;
1166
1167 if (orig->encap) {
1168 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1169 if (!x->encap)
1170 goto error;
1171 }
1172
1173 if (orig->coaddr) {
1174 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1175 GFP_KERNEL);
1176 if (!x->coaddr)
1177 goto error;
1178 }
1179
1180 if (orig->replay_esn) {
1181 err = xfrm_replay_clone(x, orig);
1182 if (err)
1183 goto error;
1184 }
1185
1186 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1187
1188 err = xfrm_init_state(x);
1189 if (err)
1190 goto error;
1191
1192 x->props.flags = orig->props.flags;
1193
1194 x->curlft.add_time = orig->curlft.add_time;
1195 x->km.state = orig->km.state;
1196 x->km.seq = orig->km.seq;
1197
1198 return x;
1199
1200 error:
1201 xfrm_state_put(x);
1202out:
1203 if (errp)
1204 *errp = err;
1205 return NULL;
1206}
1207
1208/* xfrm_state_lock is held */
1209struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1210{
1211 unsigned int h;
1212 struct xfrm_state *x;
1213 struct hlist_node *entry;
1214
1215 if (m->reqid) {
1216 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1217 m->reqid, m->old_family);
1218 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1219 if (x->props.mode != m->mode ||
1220 x->id.proto != m->proto)
1221 continue;
1222 if (m->reqid && x->props.reqid != m->reqid)
1223 continue;
1224 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1225 m->old_family) ||
1226 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1227 m->old_family))
1228 continue;
1229 xfrm_state_hold(x);
1230 return x;
1231 }
1232 } else {
1233 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1234 m->old_family);
1235 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1236 if (x->props.mode != m->mode ||
1237 x->id.proto != m->proto)
1238 continue;
1239 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1240 m->old_family) ||
1241 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1242 m->old_family))
1243 continue;
1244 xfrm_state_hold(x);
1245 return x;
1246 }
1247 }
1248
1249 return NULL;
1250}
1251EXPORT_SYMBOL(xfrm_migrate_state_find);
1252
1253struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1254 struct xfrm_migrate *m)
1255{
1256 struct xfrm_state *xc;
1257 int err;
1258
1259 xc = xfrm_state_clone(x, &err);
1260 if (!xc)
1261 return NULL;
1262
1263 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1264 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1265
1266 /* add state */
1267 if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1268 /* a care is needed when the destination address of the
1269 state is to be updated as it is a part of triplet */
1270 xfrm_state_insert(xc);
1271 } else {
1272 if ((err = xfrm_state_add(xc)) < 0)
1273 goto error;
1274 }
1275
1276 return xc;
1277error:
1278 xfrm_state_put(xc);
1279 return NULL;
1280}
1281EXPORT_SYMBOL(xfrm_state_migrate);
1282#endif
1283
1284int xfrm_state_update(struct xfrm_state *x)
1285{
1286 struct xfrm_state *x1, *to_put;
1287 int err;
1288 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1289
1290 to_put = NULL;
1291
1292 spin_lock_bh(&xfrm_state_lock);
1293 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1294
1295 err = -ESRCH;
1296 if (!x1)
1297 goto out;
1298
1299 if (xfrm_state_kern(x1)) {
1300 to_put = x1;
1301 err = -EEXIST;
1302 goto out;
1303 }
1304
1305 if (x1->km.state == XFRM_STATE_ACQ) {
1306 __xfrm_state_insert(x);
1307 x = NULL;
1308 }
1309 err = 0;
1310
1311out:
1312 spin_unlock_bh(&xfrm_state_lock);
1313
1314 if (to_put)
1315 xfrm_state_put(to_put);
1316
1317 if (err)
1318 return err;
1319
1320 if (!x) {
1321 xfrm_state_delete(x1);
1322 xfrm_state_put(x1);
1323 return 0;
1324 }
1325
1326 err = -EINVAL;
1327 spin_lock_bh(&x1->lock);
1328 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1329 if (x->encap && x1->encap)
1330 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1331 if (x->coaddr && x1->coaddr) {
1332 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1333 }
1334 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1335 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1336 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1337 x1->km.dying = 0;
1338
1339 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1340 if (x1->curlft.use_time)
1341 xfrm_state_check_expire(x1);
1342
1343 err = 0;
1344 x->km.state = XFRM_STATE_DEAD;
1345 __xfrm_state_put(x);
1346 }
1347 spin_unlock_bh(&x1->lock);
1348
1349 xfrm_state_put(x1);
1350
1351 return err;
1352}
1353EXPORT_SYMBOL(xfrm_state_update);
1354
1355int xfrm_state_check_expire(struct xfrm_state *x)
1356{
1357 if (!x->curlft.use_time)
1358 x->curlft.use_time = get_seconds();
1359
1360 if (x->km.state != XFRM_STATE_VALID)
1361 return -EINVAL;
1362
1363 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1364 x->curlft.packets >= x->lft.hard_packet_limit) {
1365 x->km.state = XFRM_STATE_EXPIRED;
1366 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1367 return -EINVAL;
1368 }
1369
1370 if (!x->km.dying &&
1371 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1372 x->curlft.packets >= x->lft.soft_packet_limit)) {
1373 x->km.dying = 1;
1374 km_state_expired(x, 0, 0);
1375 }
1376 return 0;
1377}
1378EXPORT_SYMBOL(xfrm_state_check_expire);
1379
1380struct xfrm_state *
1381xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1382 u8 proto, unsigned short family)
1383{
1384 struct xfrm_state *x;
1385
1386 spin_lock_bh(&xfrm_state_lock);
1387 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1388 spin_unlock_bh(&xfrm_state_lock);
1389 return x;
1390}
1391EXPORT_SYMBOL(xfrm_state_lookup);
1392
1393struct xfrm_state *
1394xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1395 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1396 u8 proto, unsigned short family)
1397{
1398 struct xfrm_state *x;
1399
1400 spin_lock_bh(&xfrm_state_lock);
1401 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1402 spin_unlock_bh(&xfrm_state_lock);
1403 return x;
1404}
1405EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1406
1407struct xfrm_state *
1408xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1409 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1410 int create, unsigned short family)
1411{
1412 struct xfrm_state *x;
1413
1414 spin_lock_bh(&xfrm_state_lock);
1415 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1416 spin_unlock_bh(&xfrm_state_lock);
1417
1418 return x;
1419}
1420EXPORT_SYMBOL(xfrm_find_acq);
1421
1422#ifdef CONFIG_XFRM_SUB_POLICY
1423int
1424xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1425 unsigned short family)
1426{
1427 int err = 0;
1428 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1429 if (!afinfo)
1430 return -EAFNOSUPPORT;
1431
1432 spin_lock_bh(&xfrm_state_lock);
1433 if (afinfo->tmpl_sort)
1434 err = afinfo->tmpl_sort(dst, src, n);
1435 spin_unlock_bh(&xfrm_state_lock);
1436 xfrm_state_put_afinfo(afinfo);
1437 return err;
1438}
1439EXPORT_SYMBOL(xfrm_tmpl_sort);
1440
1441int
1442xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1443 unsigned short family)
1444{
1445 int err = 0;
1446 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1447 if (!afinfo)
1448 return -EAFNOSUPPORT;
1449
1450 spin_lock_bh(&xfrm_state_lock);
1451 if (afinfo->state_sort)
1452 err = afinfo->state_sort(dst, src, n);
1453 spin_unlock_bh(&xfrm_state_lock);
1454 xfrm_state_put_afinfo(afinfo);
1455 return err;
1456}
1457EXPORT_SYMBOL(xfrm_state_sort);
1458#endif
1459
1460/* Silly enough, but I'm lazy to build resolution list */
1461
1462static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1463{
1464 int i;
1465
1466 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1467 struct hlist_node *entry;
1468 struct xfrm_state *x;
1469
1470 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1471 if (x->km.seq == seq &&
1472 (mark & x->mark.m) == x->mark.v &&
1473 x->km.state == XFRM_STATE_ACQ) {
1474 xfrm_state_hold(x);
1475 return x;
1476 }
1477 }
1478 }
1479 return NULL;
1480}
1481
1482struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1483{
1484 struct xfrm_state *x;
1485
1486 spin_lock_bh(&xfrm_state_lock);
1487 x = __xfrm_find_acq_byseq(net, mark, seq);
1488 spin_unlock_bh(&xfrm_state_lock);
1489 return x;
1490}
1491EXPORT_SYMBOL(xfrm_find_acq_byseq);
1492
1493u32 xfrm_get_acqseq(void)
1494{
1495 u32 res;
1496 static atomic_t acqseq;
1497
1498 do {
1499 res = atomic_inc_return(&acqseq);
1500 } while (!res);
1501
1502 return res;
1503}
1504EXPORT_SYMBOL(xfrm_get_acqseq);
1505
1506int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1507{
1508 struct net *net = xs_net(x);
1509 unsigned int h;
1510 struct xfrm_state *x0;
1511 int err = -ENOENT;
1512 __be32 minspi = htonl(low);
1513 __be32 maxspi = htonl(high);
1514 u32 mark = x->mark.v & x->mark.m;
1515
1516 spin_lock_bh(&x->lock);
1517 if (x->km.state == XFRM_STATE_DEAD)
1518 goto unlock;
1519
1520 err = 0;
1521 if (x->id.spi)
1522 goto unlock;
1523
1524 err = -ENOENT;
1525
1526 if (minspi == maxspi) {
1527 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1528 if (x0) {
1529 xfrm_state_put(x0);
1530 goto unlock;
1531 }
1532 x->id.spi = minspi;
1533 } else {
1534 u32 spi = 0;
1535 for (h=0; h<high-low+1; h++) {
1536 spi = low + net_random()%(high-low+1);
1537 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1538 if (x0 == NULL) {
1539 x->id.spi = htonl(spi);
1540 break;
1541 }
1542 xfrm_state_put(x0);
1543 }
1544 }
1545 if (x->id.spi) {
1546 spin_lock_bh(&xfrm_state_lock);
1547 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1548 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1549 spin_unlock_bh(&xfrm_state_lock);
1550
1551 err = 0;
1552 }
1553
1554unlock:
1555 spin_unlock_bh(&x->lock);
1556
1557 return err;
1558}
1559EXPORT_SYMBOL(xfrm_alloc_spi);
1560
1561int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1562 int (*func)(struct xfrm_state *, int, void*),
1563 void *data)
1564{
1565 struct xfrm_state *state;
1566 struct xfrm_state_walk *x;
1567 int err = 0;
1568
1569 if (walk->seq != 0 && list_empty(&walk->all))
1570 return 0;
1571
1572 spin_lock_bh(&xfrm_state_lock);
1573 if (list_empty(&walk->all))
1574 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1575 else
1576 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1577 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1578 if (x->state == XFRM_STATE_DEAD)
1579 continue;
1580 state = container_of(x, struct xfrm_state, km);
1581 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1582 continue;
1583 err = func(state, walk->seq, data);
1584 if (err) {
1585 list_move_tail(&walk->all, &x->all);
1586 goto out;
1587 }
1588 walk->seq++;
1589 }
1590 if (walk->seq == 0) {
1591 err = -ENOENT;
1592 goto out;
1593 }
1594 list_del_init(&walk->all);
1595out:
1596 spin_unlock_bh(&xfrm_state_lock);
1597 return err;
1598}
1599EXPORT_SYMBOL(xfrm_state_walk);
1600
1601void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1602{
1603 INIT_LIST_HEAD(&walk->all);
1604 walk->proto = proto;
1605 walk->state = XFRM_STATE_DEAD;
1606 walk->seq = 0;
1607}
1608EXPORT_SYMBOL(xfrm_state_walk_init);
1609
1610void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1611{
1612 if (list_empty(&walk->all))
1613 return;
1614
1615 spin_lock_bh(&xfrm_state_lock);
1616 list_del(&walk->all);
1617 spin_unlock_bh(&xfrm_state_lock);
1618}
1619EXPORT_SYMBOL(xfrm_state_walk_done);
1620
1621static void xfrm_replay_timer_handler(unsigned long data)
1622{
1623 struct xfrm_state *x = (struct xfrm_state*)data;
1624
1625 spin_lock(&x->lock);
1626
1627 if (x->km.state == XFRM_STATE_VALID) {
1628 if (xfrm_aevent_is_on(xs_net(x)))
1629 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1630 else
1631 x->xflags |= XFRM_TIME_DEFER;
1632 }
1633
1634 spin_unlock(&x->lock);
1635}
1636
1637static LIST_HEAD(xfrm_km_list);
1638static DEFINE_RWLOCK(xfrm_km_lock);
1639
1640void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1641{
1642 struct xfrm_mgr *km;
1643
1644 read_lock(&xfrm_km_lock);
1645 list_for_each_entry(km, &xfrm_km_list, list)
1646 if (km->notify_policy)
1647 km->notify_policy(xp, dir, c);
1648 read_unlock(&xfrm_km_lock);
1649}
1650
1651void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1652{
1653 struct xfrm_mgr *km;
1654 read_lock(&xfrm_km_lock);
1655 list_for_each_entry(km, &xfrm_km_list, list)
1656 if (km->notify)
1657 km->notify(x, c);
1658 read_unlock(&xfrm_km_lock);
1659}
1660
1661EXPORT_SYMBOL(km_policy_notify);
1662EXPORT_SYMBOL(km_state_notify);
1663
1664void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1665{
1666 struct net *net = xs_net(x);
1667 struct km_event c;
1668
1669 c.data.hard = hard;
1670 c.pid = pid;
1671 c.event = XFRM_MSG_EXPIRE;
1672 km_state_notify(x, &c);
1673
1674 if (hard)
1675 wake_up(&net->xfrm.km_waitq);
1676}
1677
1678EXPORT_SYMBOL(km_state_expired);
1679/*
1680 * We send to all registered managers regardless of failure
1681 * We are happy with one success
1682*/
1683int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1684{
1685 int err = -EINVAL, acqret;
1686 struct xfrm_mgr *km;
1687
1688 read_lock(&xfrm_km_lock);
1689 list_for_each_entry(km, &xfrm_km_list, list) {
1690 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1691 if (!acqret)
1692 err = acqret;
1693 }
1694 read_unlock(&xfrm_km_lock);
1695 return err;
1696}
1697EXPORT_SYMBOL(km_query);
1698
1699int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1700{
1701 int err = -EINVAL;
1702 struct xfrm_mgr *km;
1703
1704 read_lock(&xfrm_km_lock);
1705 list_for_each_entry(km, &xfrm_km_list, list) {
1706 if (km->new_mapping)
1707 err = km->new_mapping(x, ipaddr, sport);
1708 if (!err)
1709 break;
1710 }
1711 read_unlock(&xfrm_km_lock);
1712 return err;
1713}
1714EXPORT_SYMBOL(km_new_mapping);
1715
1716void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1717{
1718 struct net *net = xp_net(pol);
1719 struct km_event c;
1720
1721 c.data.hard = hard;
1722 c.pid = pid;
1723 c.event = XFRM_MSG_POLEXPIRE;
1724 km_policy_notify(pol, dir, &c);
1725
1726 if (hard)
1727 wake_up(&net->xfrm.km_waitq);
1728}
1729EXPORT_SYMBOL(km_policy_expired);
1730
1731#ifdef CONFIG_XFRM_MIGRATE
1732int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1733 const struct xfrm_migrate *m, int num_migrate,
1734 const struct xfrm_kmaddress *k)
1735{
1736 int err = -EINVAL;
1737 int ret;
1738 struct xfrm_mgr *km;
1739
1740 read_lock(&xfrm_km_lock);
1741 list_for_each_entry(km, &xfrm_km_list, list) {
1742 if (km->migrate) {
1743 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1744 if (!ret)
1745 err = ret;
1746 }
1747 }
1748 read_unlock(&xfrm_km_lock);
1749 return err;
1750}
1751EXPORT_SYMBOL(km_migrate);
1752#endif
1753
1754int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1755{
1756 int err = -EINVAL;
1757 int ret;
1758 struct xfrm_mgr *km;
1759
1760 read_lock(&xfrm_km_lock);
1761 list_for_each_entry(km, &xfrm_km_list, list) {
1762 if (km->report) {
1763 ret = km->report(net, proto, sel, addr);
1764 if (!ret)
1765 err = ret;
1766 }
1767 }
1768 read_unlock(&xfrm_km_lock);
1769 return err;
1770}
1771EXPORT_SYMBOL(km_report);
1772
1773int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1774{
1775 int err;
1776 u8 *data;
1777 struct xfrm_mgr *km;
1778 struct xfrm_policy *pol = NULL;
1779
1780 if (optlen <= 0 || optlen > PAGE_SIZE)
1781 return -EMSGSIZE;
1782
1783 data = kmalloc(optlen, GFP_KERNEL);
1784 if (!data)
1785 return -ENOMEM;
1786
1787 err = -EFAULT;
1788 if (copy_from_user(data, optval, optlen))
1789 goto out;
1790
1791 err = -EINVAL;
1792 read_lock(&xfrm_km_lock);
1793 list_for_each_entry(km, &xfrm_km_list, list) {
1794 pol = km->compile_policy(sk, optname, data,
1795 optlen, &err);
1796 if (err >= 0)
1797 break;
1798 }
1799 read_unlock(&xfrm_km_lock);
1800
1801 if (err >= 0) {
1802 xfrm_sk_policy_insert(sk, err, pol);
1803 xfrm_pol_put(pol);
1804 err = 0;
1805 }
1806
1807out:
1808 kfree(data);
1809 return err;
1810}
1811EXPORT_SYMBOL(xfrm_user_policy);
1812
1813int xfrm_register_km(struct xfrm_mgr *km)
1814{
1815 write_lock_bh(&xfrm_km_lock);
1816 list_add_tail(&km->list, &xfrm_km_list);
1817 write_unlock_bh(&xfrm_km_lock);
1818 return 0;
1819}
1820EXPORT_SYMBOL(xfrm_register_km);
1821
1822int xfrm_unregister_km(struct xfrm_mgr *km)
1823{
1824 write_lock_bh(&xfrm_km_lock);
1825 list_del(&km->list);
1826 write_unlock_bh(&xfrm_km_lock);
1827 return 0;
1828}
1829EXPORT_SYMBOL(xfrm_unregister_km);
1830
1831int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1832{
1833 int err = 0;
1834 if (unlikely(afinfo == NULL))
1835 return -EINVAL;
1836 if (unlikely(afinfo->family >= NPROTO))
1837 return -EAFNOSUPPORT;
1838 write_lock_bh(&xfrm_state_afinfo_lock);
1839 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1840 err = -ENOBUFS;
1841 else
1842 xfrm_state_afinfo[afinfo->family] = afinfo;
1843 write_unlock_bh(&xfrm_state_afinfo_lock);
1844 return err;
1845}
1846EXPORT_SYMBOL(xfrm_state_register_afinfo);
1847
1848int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1849{
1850 int err = 0;
1851 if (unlikely(afinfo == NULL))
1852 return -EINVAL;
1853 if (unlikely(afinfo->family >= NPROTO))
1854 return -EAFNOSUPPORT;
1855 write_lock_bh(&xfrm_state_afinfo_lock);
1856 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1857 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1858 err = -EINVAL;
1859 else
1860 xfrm_state_afinfo[afinfo->family] = NULL;
1861 }
1862 write_unlock_bh(&xfrm_state_afinfo_lock);
1863 return err;
1864}
1865EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1866
1867static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1868{
1869 struct xfrm_state_afinfo *afinfo;
1870 if (unlikely(family >= NPROTO))
1871 return NULL;
1872 read_lock(&xfrm_state_afinfo_lock);
1873 afinfo = xfrm_state_afinfo[family];
1874 if (unlikely(!afinfo))
1875 read_unlock(&xfrm_state_afinfo_lock);
1876 return afinfo;
1877}
1878
1879static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1880 __releases(xfrm_state_afinfo_lock)
1881{
1882 read_unlock(&xfrm_state_afinfo_lock);
1883}
1884
1885/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1886void xfrm_state_delete_tunnel(struct xfrm_state *x)
1887{
1888 if (x->tunnel) {
1889 struct xfrm_state *t = x->tunnel;
1890
1891 if (atomic_read(&t->tunnel_users) == 2)
1892 xfrm_state_delete(t);
1893 atomic_dec(&t->tunnel_users);
1894 xfrm_state_put(t);
1895 x->tunnel = NULL;
1896 }
1897}
1898EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1899
1900int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1901{
1902 int res;
1903
1904 spin_lock_bh(&x->lock);
1905 if (x->km.state == XFRM_STATE_VALID &&
1906 x->type && x->type->get_mtu)
1907 res = x->type->get_mtu(x, mtu);
1908 else
1909 res = mtu - x->props.header_len;
1910 spin_unlock_bh(&x->lock);
1911 return res;
1912}
1913
1914int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1915{
1916 struct xfrm_state_afinfo *afinfo;
1917 struct xfrm_mode *inner_mode;
1918 int family = x->props.family;
1919 int err;
1920
1921 err = -EAFNOSUPPORT;
1922 afinfo = xfrm_state_get_afinfo(family);
1923 if (!afinfo)
1924 goto error;
1925
1926 err = 0;
1927 if (afinfo->init_flags)
1928 err = afinfo->init_flags(x);
1929
1930 xfrm_state_put_afinfo(afinfo);
1931
1932 if (err)
1933 goto error;
1934
1935 err = -EPROTONOSUPPORT;
1936
1937 if (x->sel.family != AF_UNSPEC) {
1938 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1939 if (inner_mode == NULL)
1940 goto error;
1941
1942 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1943 family != x->sel.family) {
1944 xfrm_put_mode(inner_mode);
1945 goto error;
1946 }
1947
1948 x->inner_mode = inner_mode;
1949 } else {
1950 struct xfrm_mode *inner_mode_iaf;
1951 int iafamily = AF_INET;
1952
1953 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1954 if (inner_mode == NULL)
1955 goto error;
1956
1957 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1958 xfrm_put_mode(inner_mode);
1959 goto error;
1960 }
1961 x->inner_mode = inner_mode;
1962
1963 if (x->props.family == AF_INET)
1964 iafamily = AF_INET6;
1965
1966 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1967 if (inner_mode_iaf) {
1968 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1969 x->inner_mode_iaf = inner_mode_iaf;
1970 else
1971 xfrm_put_mode(inner_mode_iaf);
1972 }
1973 }
1974
1975 x->type = xfrm_get_type(x->id.proto, family);
1976 if (x->type == NULL)
1977 goto error;
1978
1979 err = x->type->init_state(x);
1980 if (err)
1981 goto error;
1982
1983 x->outer_mode = xfrm_get_mode(x->props.mode, family);
1984 if (x->outer_mode == NULL)
1985 goto error;
1986
1987 if (init_replay) {
1988 err = xfrm_init_replay(x);
1989 if (err)
1990 goto error;
1991 }
1992
1993 x->km.state = XFRM_STATE_VALID;
1994
1995error:
1996 return err;
1997}
1998
1999EXPORT_SYMBOL(__xfrm_init_state);
2000
2001int xfrm_init_state(struct xfrm_state *x)
2002{
2003 return __xfrm_init_state(x, true);
2004}
2005
2006EXPORT_SYMBOL(xfrm_init_state);
2007
2008int __net_init xfrm_state_init(struct net *net)
2009{
2010 unsigned int sz;
2011
2012 INIT_LIST_HEAD(&net->xfrm.state_all);
2013
2014 sz = sizeof(struct hlist_head) * 8;
2015
2016 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2017 if (!net->xfrm.state_bydst)
2018 goto out_bydst;
2019 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2020 if (!net->xfrm.state_bysrc)
2021 goto out_bysrc;
2022 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2023 if (!net->xfrm.state_byspi)
2024 goto out_byspi;
2025 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2026
2027 net->xfrm.state_num = 0;
2028 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2029 INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2030 INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2031 init_waitqueue_head(&net->xfrm.km_waitq);
2032 return 0;
2033
2034out_byspi:
2035 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2036out_bysrc:
2037 xfrm_hash_free(net->xfrm.state_bydst, sz);
2038out_bydst:
2039 return -ENOMEM;
2040}
2041
2042void xfrm_state_fini(struct net *net)
2043{
2044 struct xfrm_audit audit_info;
2045 unsigned int sz;
2046
2047 flush_work(&net->xfrm.state_hash_work);
2048 audit_info.loginuid = -1;
2049 audit_info.sessionid = -1;
2050 audit_info.secid = 0;
2051 xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2052 flush_work(&net->xfrm.state_gc_work);
2053
2054 WARN_ON(!list_empty(&net->xfrm.state_all));
2055
2056 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2057 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2058 xfrm_hash_free(net->xfrm.state_byspi, sz);
2059 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2060 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2061 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2062 xfrm_hash_free(net->xfrm.state_bydst, sz);
2063}
2064
2065#ifdef CONFIG_AUDITSYSCALL
2066static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2067 struct audit_buffer *audit_buf)
2068{
2069 struct xfrm_sec_ctx *ctx = x->security;
2070 u32 spi = ntohl(x->id.spi);
2071
2072 if (ctx)
2073 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2074 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2075
2076 switch(x->props.family) {
2077 case AF_INET:
2078 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2079 &x->props.saddr.a4, &x->id.daddr.a4);
2080 break;
2081 case AF_INET6:
2082 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2083 x->props.saddr.a6, x->id.daddr.a6);
2084 break;
2085 }
2086
2087 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2088}
2089
2090static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2091 struct audit_buffer *audit_buf)
2092{
2093 const struct iphdr *iph4;
2094 const struct ipv6hdr *iph6;
2095
2096 switch (family) {
2097 case AF_INET:
2098 iph4 = ip_hdr(skb);
2099 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2100 &iph4->saddr, &iph4->daddr);
2101 break;
2102 case AF_INET6:
2103 iph6 = ipv6_hdr(skb);
2104 audit_log_format(audit_buf,
2105 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2106 &iph6->saddr,&iph6->daddr,
2107 iph6->flow_lbl[0] & 0x0f,
2108 iph6->flow_lbl[1],
2109 iph6->flow_lbl[2]);
2110 break;
2111 }
2112}
2113
2114void xfrm_audit_state_add(struct xfrm_state *x, int result,
2115 uid_t auid, u32 sessionid, u32 secid)
2116{
2117 struct audit_buffer *audit_buf;
2118
2119 audit_buf = xfrm_audit_start("SAD-add");
2120 if (audit_buf == NULL)
2121 return;
2122 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2123 xfrm_audit_helper_sainfo(x, audit_buf);
2124 audit_log_format(audit_buf, " res=%u", result);
2125 audit_log_end(audit_buf);
2126}
2127EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2128
2129void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2130 uid_t auid, u32 sessionid, u32 secid)
2131{
2132 struct audit_buffer *audit_buf;
2133
2134 audit_buf = xfrm_audit_start("SAD-delete");
2135 if (audit_buf == NULL)
2136 return;
2137 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2138 xfrm_audit_helper_sainfo(x, audit_buf);
2139 audit_log_format(audit_buf, " res=%u", result);
2140 audit_log_end(audit_buf);
2141}
2142EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2143
2144void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2145 struct sk_buff *skb)
2146{
2147 struct audit_buffer *audit_buf;
2148 u32 spi;
2149
2150 audit_buf = xfrm_audit_start("SA-replay-overflow");
2151 if (audit_buf == NULL)
2152 return;
2153 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2154 /* don't record the sequence number because it's inherent in this kind
2155 * of audit message */
2156 spi = ntohl(x->id.spi);
2157 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158 audit_log_end(audit_buf);
2159}
2160EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2161
2162void xfrm_audit_state_replay(struct xfrm_state *x,
2163 struct sk_buff *skb, __be32 net_seq)
2164{
2165 struct audit_buffer *audit_buf;
2166 u32 spi;
2167
2168 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2169 if (audit_buf == NULL)
2170 return;
2171 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2172 spi = ntohl(x->id.spi);
2173 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2174 spi, spi, ntohl(net_seq));
2175 audit_log_end(audit_buf);
2176}
2177EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2178
2179void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2180{
2181 struct audit_buffer *audit_buf;
2182
2183 audit_buf = xfrm_audit_start("SA-notfound");
2184 if (audit_buf == NULL)
2185 return;
2186 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2187 audit_log_end(audit_buf);
2188}
2189EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2190
2191void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2192 __be32 net_spi, __be32 net_seq)
2193{
2194 struct audit_buffer *audit_buf;
2195 u32 spi;
2196
2197 audit_buf = xfrm_audit_start("SA-notfound");
2198 if (audit_buf == NULL)
2199 return;
2200 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2201 spi = ntohl(net_spi);
2202 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2203 spi, spi, ntohl(net_seq));
2204 audit_log_end(audit_buf);
2205}
2206EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2207
2208void xfrm_audit_state_icvfail(struct xfrm_state *x,
2209 struct sk_buff *skb, u8 proto)
2210{
2211 struct audit_buffer *audit_buf;
2212 __be32 net_spi;
2213 __be32 net_seq;
2214
2215 audit_buf = xfrm_audit_start("SA-icv-failure");
2216 if (audit_buf == NULL)
2217 return;
2218 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2219 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2220 u32 spi = ntohl(net_spi);
2221 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2222 spi, spi, ntohl(net_seq));
2223 }
2224 audit_log_end(audit_buf);
2225}
2226EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2227#endif /* CONFIG_AUDITSYSCALL */
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * xfrm_state.c
4 *
5 * Changes:
6 * Mitsuru KANDA @USAGI
7 * Kazunori MIYAZAWA @USAGI
8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * IPv6 support
10 * YOSHIFUJI Hideaki @USAGI
11 * Split up af-specific functions
12 * Derek Atkins <derek@ihtfp.com>
13 * Add UDP Encapsulation
14 *
15 */
16
17#include <linux/workqueue.h>
18#include <net/xfrm.h>
19#include <linux/pfkeyv2.h>
20#include <linux/ipsec.h>
21#include <linux/module.h>
22#include <linux/cache.h>
23#include <linux/audit.h>
24#include <linux/uaccess.h>
25#include <linux/ktime.h>
26#include <linux/slab.h>
27#include <linux/interrupt.h>
28#include <linux/kernel.h>
29
30#include <crypto/aead.h>
31
32#include "xfrm_hash.h"
33
34#define xfrm_state_deref_prot(table, net) \
35 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
36
37static void xfrm_state_gc_task(struct work_struct *work);
38
39/* Each xfrm_state may be linked to two tables:
40
41 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
42 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
43 destination/tunnel endpoint. (output)
44 */
45
46static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
47static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
48static struct kmem_cache *xfrm_state_cache __ro_after_init;
49
50static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
51static HLIST_HEAD(xfrm_state_gc_list);
52
53static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54{
55 return refcount_inc_not_zero(&x->refcnt);
56}
57
58static inline unsigned int xfrm_dst_hash(struct net *net,
59 const xfrm_address_t *daddr,
60 const xfrm_address_t *saddr,
61 u32 reqid,
62 unsigned short family)
63{
64 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
65}
66
67static inline unsigned int xfrm_src_hash(struct net *net,
68 const xfrm_address_t *daddr,
69 const xfrm_address_t *saddr,
70 unsigned short family)
71{
72 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
73}
74
75static inline unsigned int
76xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
77 __be32 spi, u8 proto, unsigned short family)
78{
79 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
80}
81
82static void xfrm_hash_transfer(struct hlist_head *list,
83 struct hlist_head *ndsttable,
84 struct hlist_head *nsrctable,
85 struct hlist_head *nspitable,
86 unsigned int nhashmask)
87{
88 struct hlist_node *tmp;
89 struct xfrm_state *x;
90
91 hlist_for_each_entry_safe(x, tmp, list, bydst) {
92 unsigned int h;
93
94 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
95 x->props.reqid, x->props.family,
96 nhashmask);
97 hlist_add_head_rcu(&x->bydst, ndsttable + h);
98
99 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
100 x->props.family,
101 nhashmask);
102 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
103
104 if (x->id.spi) {
105 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
106 x->id.proto, x->props.family,
107 nhashmask);
108 hlist_add_head_rcu(&x->byspi, nspitable + h);
109 }
110 }
111}
112
113static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
114{
115 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
116}
117
118static void xfrm_hash_resize(struct work_struct *work)
119{
120 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
121 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
122 unsigned long nsize, osize;
123 unsigned int nhashmask, ohashmask;
124 int i;
125
126 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
127 ndst = xfrm_hash_alloc(nsize);
128 if (!ndst)
129 return;
130 nsrc = xfrm_hash_alloc(nsize);
131 if (!nsrc) {
132 xfrm_hash_free(ndst, nsize);
133 return;
134 }
135 nspi = xfrm_hash_alloc(nsize);
136 if (!nspi) {
137 xfrm_hash_free(ndst, nsize);
138 xfrm_hash_free(nsrc, nsize);
139 return;
140 }
141
142 spin_lock_bh(&net->xfrm.xfrm_state_lock);
143 write_seqcount_begin(&xfrm_state_hash_generation);
144
145 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
146 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
147 for (i = net->xfrm.state_hmask; i >= 0; i--)
148 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
149
150 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
151 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
152 ohashmask = net->xfrm.state_hmask;
153
154 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
155 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
156 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
157 net->xfrm.state_hmask = nhashmask;
158
159 write_seqcount_end(&xfrm_state_hash_generation);
160 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
161
162 osize = (ohashmask + 1) * sizeof(struct hlist_head);
163
164 synchronize_rcu();
165
166 xfrm_hash_free(odst, osize);
167 xfrm_hash_free(osrc, osize);
168 xfrm_hash_free(ospi, osize);
169}
170
171static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
172static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
173
174static DEFINE_SPINLOCK(xfrm_state_gc_lock);
175
176int __xfrm_state_delete(struct xfrm_state *x);
177
178int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
179static bool km_is_alive(const struct km_event *c);
180void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
181
182int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
183{
184 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
185 int err = 0;
186
187 if (!afinfo)
188 return -EAFNOSUPPORT;
189
190#define X(afi, T, name) do { \
191 WARN_ON((afi)->type_ ## name); \
192 (afi)->type_ ## name = (T); \
193 } while (0)
194
195 switch (type->proto) {
196 case IPPROTO_COMP:
197 X(afinfo, type, comp);
198 break;
199 case IPPROTO_AH:
200 X(afinfo, type, ah);
201 break;
202 case IPPROTO_ESP:
203 X(afinfo, type, esp);
204 break;
205 case IPPROTO_IPIP:
206 X(afinfo, type, ipip);
207 break;
208 case IPPROTO_DSTOPTS:
209 X(afinfo, type, dstopts);
210 break;
211 case IPPROTO_ROUTING:
212 X(afinfo, type, routing);
213 break;
214 case IPPROTO_IPV6:
215 X(afinfo, type, ipip6);
216 break;
217 default:
218 WARN_ON(1);
219 err = -EPROTONOSUPPORT;
220 break;
221 }
222#undef X
223 rcu_read_unlock();
224 return err;
225}
226EXPORT_SYMBOL(xfrm_register_type);
227
228void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
229{
230 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
231
232 if (unlikely(afinfo == NULL))
233 return;
234
235#define X(afi, T, name) do { \
236 WARN_ON((afi)->type_ ## name != (T)); \
237 (afi)->type_ ## name = NULL; \
238 } while (0)
239
240 switch (type->proto) {
241 case IPPROTO_COMP:
242 X(afinfo, type, comp);
243 break;
244 case IPPROTO_AH:
245 X(afinfo, type, ah);
246 break;
247 case IPPROTO_ESP:
248 X(afinfo, type, esp);
249 break;
250 case IPPROTO_IPIP:
251 X(afinfo, type, ipip);
252 break;
253 case IPPROTO_DSTOPTS:
254 X(afinfo, type, dstopts);
255 break;
256 case IPPROTO_ROUTING:
257 X(afinfo, type, routing);
258 break;
259 case IPPROTO_IPV6:
260 X(afinfo, type, ipip6);
261 break;
262 default:
263 WARN_ON(1);
264 break;
265 }
266#undef X
267 rcu_read_unlock();
268}
269EXPORT_SYMBOL(xfrm_unregister_type);
270
271static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
272{
273 const struct xfrm_type *type = NULL;
274 struct xfrm_state_afinfo *afinfo;
275 int modload_attempted = 0;
276
277retry:
278 afinfo = xfrm_state_get_afinfo(family);
279 if (unlikely(afinfo == NULL))
280 return NULL;
281
282 switch (proto) {
283 case IPPROTO_COMP:
284 type = afinfo->type_comp;
285 break;
286 case IPPROTO_AH:
287 type = afinfo->type_ah;
288 break;
289 case IPPROTO_ESP:
290 type = afinfo->type_esp;
291 break;
292 case IPPROTO_IPIP:
293 type = afinfo->type_ipip;
294 break;
295 case IPPROTO_DSTOPTS:
296 type = afinfo->type_dstopts;
297 break;
298 case IPPROTO_ROUTING:
299 type = afinfo->type_routing;
300 break;
301 case IPPROTO_IPV6:
302 type = afinfo->type_ipip6;
303 break;
304 default:
305 break;
306 }
307
308 if (unlikely(type && !try_module_get(type->owner)))
309 type = NULL;
310
311 rcu_read_unlock();
312
313 if (!type && !modload_attempted) {
314 request_module("xfrm-type-%d-%d", family, proto);
315 modload_attempted = 1;
316 goto retry;
317 }
318
319 return type;
320}
321
322static void xfrm_put_type(const struct xfrm_type *type)
323{
324 module_put(type->owner);
325}
326
327int xfrm_register_type_offload(const struct xfrm_type_offload *type,
328 unsigned short family)
329{
330 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
331 int err = 0;
332
333 if (unlikely(afinfo == NULL))
334 return -EAFNOSUPPORT;
335
336 switch (type->proto) {
337 case IPPROTO_ESP:
338 WARN_ON(afinfo->type_offload_esp);
339 afinfo->type_offload_esp = type;
340 break;
341 default:
342 WARN_ON(1);
343 err = -EPROTONOSUPPORT;
344 break;
345 }
346
347 rcu_read_unlock();
348 return err;
349}
350EXPORT_SYMBOL(xfrm_register_type_offload);
351
352void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
353 unsigned short family)
354{
355 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
356
357 if (unlikely(afinfo == NULL))
358 return;
359
360 switch (type->proto) {
361 case IPPROTO_ESP:
362 WARN_ON(afinfo->type_offload_esp != type);
363 afinfo->type_offload_esp = NULL;
364 break;
365 default:
366 WARN_ON(1);
367 break;
368 }
369 rcu_read_unlock();
370}
371EXPORT_SYMBOL(xfrm_unregister_type_offload);
372
373static const struct xfrm_type_offload *
374xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
375{
376 const struct xfrm_type_offload *type = NULL;
377 struct xfrm_state_afinfo *afinfo;
378
379retry:
380 afinfo = xfrm_state_get_afinfo(family);
381 if (unlikely(afinfo == NULL))
382 return NULL;
383
384 switch (proto) {
385 case IPPROTO_ESP:
386 type = afinfo->type_offload_esp;
387 break;
388 default:
389 break;
390 }
391
392 if ((type && !try_module_get(type->owner)))
393 type = NULL;
394
395 rcu_read_unlock();
396
397 if (!type && try_load) {
398 request_module("xfrm-offload-%d-%d", family, proto);
399 try_load = false;
400 goto retry;
401 }
402
403 return type;
404}
405
406static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
407{
408 module_put(type->owner);
409}
410
411static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
412 [XFRM_MODE_BEET] = {
413 .encap = XFRM_MODE_BEET,
414 .flags = XFRM_MODE_FLAG_TUNNEL,
415 .family = AF_INET,
416 },
417 [XFRM_MODE_TRANSPORT] = {
418 .encap = XFRM_MODE_TRANSPORT,
419 .family = AF_INET,
420 },
421 [XFRM_MODE_TUNNEL] = {
422 .encap = XFRM_MODE_TUNNEL,
423 .flags = XFRM_MODE_FLAG_TUNNEL,
424 .family = AF_INET,
425 },
426};
427
428static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
429 [XFRM_MODE_BEET] = {
430 .encap = XFRM_MODE_BEET,
431 .flags = XFRM_MODE_FLAG_TUNNEL,
432 .family = AF_INET6,
433 },
434 [XFRM_MODE_ROUTEOPTIMIZATION] = {
435 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
436 .family = AF_INET6,
437 },
438 [XFRM_MODE_TRANSPORT] = {
439 .encap = XFRM_MODE_TRANSPORT,
440 .family = AF_INET6,
441 },
442 [XFRM_MODE_TUNNEL] = {
443 .encap = XFRM_MODE_TUNNEL,
444 .flags = XFRM_MODE_FLAG_TUNNEL,
445 .family = AF_INET6,
446 },
447};
448
449static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
450{
451 const struct xfrm_mode *mode;
452
453 if (unlikely(encap >= XFRM_MODE_MAX))
454 return NULL;
455
456 switch (family) {
457 case AF_INET:
458 mode = &xfrm4_mode_map[encap];
459 if (mode->family == family)
460 return mode;
461 break;
462 case AF_INET6:
463 mode = &xfrm6_mode_map[encap];
464 if (mode->family == family)
465 return mode;
466 break;
467 default:
468 break;
469 }
470
471 return NULL;
472}
473
474void xfrm_state_free(struct xfrm_state *x)
475{
476 kmem_cache_free(xfrm_state_cache, x);
477}
478EXPORT_SYMBOL(xfrm_state_free);
479
480static void ___xfrm_state_destroy(struct xfrm_state *x)
481{
482 hrtimer_cancel(&x->mtimer);
483 del_timer_sync(&x->rtimer);
484 kfree(x->aead);
485 kfree(x->aalg);
486 kfree(x->ealg);
487 kfree(x->calg);
488 kfree(x->encap);
489 kfree(x->coaddr);
490 kfree(x->replay_esn);
491 kfree(x->preplay_esn);
492 if (x->type_offload)
493 xfrm_put_type_offload(x->type_offload);
494 if (x->type) {
495 x->type->destructor(x);
496 xfrm_put_type(x->type);
497 }
498 if (x->xfrag.page)
499 put_page(x->xfrag.page);
500 xfrm_dev_state_free(x);
501 security_xfrm_state_free(x);
502 xfrm_state_free(x);
503}
504
505static void xfrm_state_gc_task(struct work_struct *work)
506{
507 struct xfrm_state *x;
508 struct hlist_node *tmp;
509 struct hlist_head gc_list;
510
511 spin_lock_bh(&xfrm_state_gc_lock);
512 hlist_move_list(&xfrm_state_gc_list, &gc_list);
513 spin_unlock_bh(&xfrm_state_gc_lock);
514
515 synchronize_rcu();
516
517 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
518 ___xfrm_state_destroy(x);
519}
520
521static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
522{
523 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
524 enum hrtimer_restart ret = HRTIMER_NORESTART;
525 time64_t now = ktime_get_real_seconds();
526 time64_t next = TIME64_MAX;
527 int warn = 0;
528 int err = 0;
529
530 spin_lock(&x->lock);
531 if (x->km.state == XFRM_STATE_DEAD)
532 goto out;
533 if (x->km.state == XFRM_STATE_EXPIRED)
534 goto expired;
535 if (x->lft.hard_add_expires_seconds) {
536 long tmo = x->lft.hard_add_expires_seconds +
537 x->curlft.add_time - now;
538 if (tmo <= 0) {
539 if (x->xflags & XFRM_SOFT_EXPIRE) {
540 /* enter hard expire without soft expire first?!
541 * setting a new date could trigger this.
542 * workaround: fix x->curflt.add_time by below:
543 */
544 x->curlft.add_time = now - x->saved_tmo - 1;
545 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
546 } else
547 goto expired;
548 }
549 if (tmo < next)
550 next = tmo;
551 }
552 if (x->lft.hard_use_expires_seconds) {
553 long tmo = x->lft.hard_use_expires_seconds +
554 (x->curlft.use_time ? : now) - now;
555 if (tmo <= 0)
556 goto expired;
557 if (tmo < next)
558 next = tmo;
559 }
560 if (x->km.dying)
561 goto resched;
562 if (x->lft.soft_add_expires_seconds) {
563 long tmo = x->lft.soft_add_expires_seconds +
564 x->curlft.add_time - now;
565 if (tmo <= 0) {
566 warn = 1;
567 x->xflags &= ~XFRM_SOFT_EXPIRE;
568 } else if (tmo < next) {
569 next = tmo;
570 x->xflags |= XFRM_SOFT_EXPIRE;
571 x->saved_tmo = tmo;
572 }
573 }
574 if (x->lft.soft_use_expires_seconds) {
575 long tmo = x->lft.soft_use_expires_seconds +
576 (x->curlft.use_time ? : now) - now;
577 if (tmo <= 0)
578 warn = 1;
579 else if (tmo < next)
580 next = tmo;
581 }
582
583 x->km.dying = warn;
584 if (warn)
585 km_state_expired(x, 0, 0);
586resched:
587 if (next != TIME64_MAX) {
588 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
589 ret = HRTIMER_RESTART;
590 }
591
592 goto out;
593
594expired:
595 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
596 x->km.state = XFRM_STATE_EXPIRED;
597
598 err = __xfrm_state_delete(x);
599 if (!err)
600 km_state_expired(x, 1, 0);
601
602 xfrm_audit_state_delete(x, err ? 0 : 1, true);
603
604out:
605 spin_unlock(&x->lock);
606 return ret;
607}
608
609static void xfrm_replay_timer_handler(struct timer_list *t);
610
611struct xfrm_state *xfrm_state_alloc(struct net *net)
612{
613 struct xfrm_state *x;
614
615 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
616
617 if (x) {
618 write_pnet(&x->xs_net, net);
619 refcount_set(&x->refcnt, 1);
620 atomic_set(&x->tunnel_users, 0);
621 INIT_LIST_HEAD(&x->km.all);
622 INIT_HLIST_NODE(&x->bydst);
623 INIT_HLIST_NODE(&x->bysrc);
624 INIT_HLIST_NODE(&x->byspi);
625 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
626 x->mtimer.function = xfrm_timer_handler;
627 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
628 x->curlft.add_time = ktime_get_real_seconds();
629 x->lft.soft_byte_limit = XFRM_INF;
630 x->lft.soft_packet_limit = XFRM_INF;
631 x->lft.hard_byte_limit = XFRM_INF;
632 x->lft.hard_packet_limit = XFRM_INF;
633 x->replay_maxage = 0;
634 x->replay_maxdiff = 0;
635 spin_lock_init(&x->lock);
636 }
637 return x;
638}
639EXPORT_SYMBOL(xfrm_state_alloc);
640
641void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
642{
643 WARN_ON(x->km.state != XFRM_STATE_DEAD);
644
645 if (sync) {
646 synchronize_rcu();
647 ___xfrm_state_destroy(x);
648 } else {
649 spin_lock_bh(&xfrm_state_gc_lock);
650 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
651 spin_unlock_bh(&xfrm_state_gc_lock);
652 schedule_work(&xfrm_state_gc_work);
653 }
654}
655EXPORT_SYMBOL(__xfrm_state_destroy);
656
657int __xfrm_state_delete(struct xfrm_state *x)
658{
659 struct net *net = xs_net(x);
660 int err = -ESRCH;
661
662 if (x->km.state != XFRM_STATE_DEAD) {
663 x->km.state = XFRM_STATE_DEAD;
664 spin_lock(&net->xfrm.xfrm_state_lock);
665 list_del(&x->km.all);
666 hlist_del_rcu(&x->bydst);
667 hlist_del_rcu(&x->bysrc);
668 if (x->id.spi)
669 hlist_del_rcu(&x->byspi);
670 net->xfrm.state_num--;
671 spin_unlock(&net->xfrm.xfrm_state_lock);
672
673 if (x->encap_sk)
674 sock_put(rcu_dereference_raw(x->encap_sk));
675
676 xfrm_dev_state_delete(x);
677
678 /* All xfrm_state objects are created by xfrm_state_alloc.
679 * The xfrm_state_alloc call gives a reference, and that
680 * is what we are dropping here.
681 */
682 xfrm_state_put(x);
683 err = 0;
684 }
685
686 return err;
687}
688EXPORT_SYMBOL(__xfrm_state_delete);
689
690int xfrm_state_delete(struct xfrm_state *x)
691{
692 int err;
693
694 spin_lock_bh(&x->lock);
695 err = __xfrm_state_delete(x);
696 spin_unlock_bh(&x->lock);
697
698 return err;
699}
700EXPORT_SYMBOL(xfrm_state_delete);
701
702#ifdef CONFIG_SECURITY_NETWORK_XFRM
703static inline int
704xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
705{
706 int i, err = 0;
707
708 for (i = 0; i <= net->xfrm.state_hmask; i++) {
709 struct xfrm_state *x;
710
711 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
712 if (xfrm_id_proto_match(x->id.proto, proto) &&
713 (err = security_xfrm_state_delete(x)) != 0) {
714 xfrm_audit_state_delete(x, 0, task_valid);
715 return err;
716 }
717 }
718 }
719
720 return err;
721}
722
723static inline int
724xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
725{
726 int i, err = 0;
727
728 for (i = 0; i <= net->xfrm.state_hmask; i++) {
729 struct xfrm_state *x;
730 struct xfrm_state_offload *xso;
731
732 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
733 xso = &x->xso;
734
735 if (xso->dev == dev &&
736 (err = security_xfrm_state_delete(x)) != 0) {
737 xfrm_audit_state_delete(x, 0, task_valid);
738 return err;
739 }
740 }
741 }
742
743 return err;
744}
745#else
746static inline int
747xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
748{
749 return 0;
750}
751
752static inline int
753xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
754{
755 return 0;
756}
757#endif
758
759int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
760{
761 int i, err = 0, cnt = 0;
762
763 spin_lock_bh(&net->xfrm.xfrm_state_lock);
764 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
765 if (err)
766 goto out;
767
768 err = -ESRCH;
769 for (i = 0; i <= net->xfrm.state_hmask; i++) {
770 struct xfrm_state *x;
771restart:
772 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
773 if (!xfrm_state_kern(x) &&
774 xfrm_id_proto_match(x->id.proto, proto)) {
775 xfrm_state_hold(x);
776 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
777
778 err = xfrm_state_delete(x);
779 xfrm_audit_state_delete(x, err ? 0 : 1,
780 task_valid);
781 if (sync)
782 xfrm_state_put_sync(x);
783 else
784 xfrm_state_put(x);
785 if (!err)
786 cnt++;
787
788 spin_lock_bh(&net->xfrm.xfrm_state_lock);
789 goto restart;
790 }
791 }
792 }
793out:
794 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
795 if (cnt)
796 err = 0;
797
798 return err;
799}
800EXPORT_SYMBOL(xfrm_state_flush);
801
802int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
803{
804 int i, err = 0, cnt = 0;
805
806 spin_lock_bh(&net->xfrm.xfrm_state_lock);
807 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
808 if (err)
809 goto out;
810
811 err = -ESRCH;
812 for (i = 0; i <= net->xfrm.state_hmask; i++) {
813 struct xfrm_state *x;
814 struct xfrm_state_offload *xso;
815restart:
816 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
817 xso = &x->xso;
818
819 if (!xfrm_state_kern(x) && xso->dev == dev) {
820 xfrm_state_hold(x);
821 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
822
823 err = xfrm_state_delete(x);
824 xfrm_audit_state_delete(x, err ? 0 : 1,
825 task_valid);
826 xfrm_state_put(x);
827 if (!err)
828 cnt++;
829
830 spin_lock_bh(&net->xfrm.xfrm_state_lock);
831 goto restart;
832 }
833 }
834 }
835 if (cnt)
836 err = 0;
837
838out:
839 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
840 return err;
841}
842EXPORT_SYMBOL(xfrm_dev_state_flush);
843
844void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
845{
846 spin_lock_bh(&net->xfrm.xfrm_state_lock);
847 si->sadcnt = net->xfrm.state_num;
848 si->sadhcnt = net->xfrm.state_hmask + 1;
849 si->sadhmcnt = xfrm_state_hashmax;
850 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
851}
852EXPORT_SYMBOL(xfrm_sad_getinfo);
853
854static void
855__xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
856{
857 const struct flowi4 *fl4 = &fl->u.ip4;
858
859 sel->daddr.a4 = fl4->daddr;
860 sel->saddr.a4 = fl4->saddr;
861 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
862 sel->dport_mask = htons(0xffff);
863 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
864 sel->sport_mask = htons(0xffff);
865 sel->family = AF_INET;
866 sel->prefixlen_d = 32;
867 sel->prefixlen_s = 32;
868 sel->proto = fl4->flowi4_proto;
869 sel->ifindex = fl4->flowi4_oif;
870}
871
872static void
873__xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
874{
875 const struct flowi6 *fl6 = &fl->u.ip6;
876
877 /* Initialize temporary selector matching only to current session. */
878 *(struct in6_addr *)&sel->daddr = fl6->daddr;
879 *(struct in6_addr *)&sel->saddr = fl6->saddr;
880 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
881 sel->dport_mask = htons(0xffff);
882 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
883 sel->sport_mask = htons(0xffff);
884 sel->family = AF_INET6;
885 sel->prefixlen_d = 128;
886 sel->prefixlen_s = 128;
887 sel->proto = fl6->flowi6_proto;
888 sel->ifindex = fl6->flowi6_oif;
889}
890
891static void
892xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
893 const struct xfrm_tmpl *tmpl,
894 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
895 unsigned short family)
896{
897 switch (family) {
898 case AF_INET:
899 __xfrm4_init_tempsel(&x->sel, fl);
900 break;
901 case AF_INET6:
902 __xfrm6_init_tempsel(&x->sel, fl);
903 break;
904 }
905
906 x->id = tmpl->id;
907
908 switch (tmpl->encap_family) {
909 case AF_INET:
910 if (x->id.daddr.a4 == 0)
911 x->id.daddr.a4 = daddr->a4;
912 x->props.saddr = tmpl->saddr;
913 if (x->props.saddr.a4 == 0)
914 x->props.saddr.a4 = saddr->a4;
915 break;
916 case AF_INET6:
917 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
918 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
919 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
920 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
921 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
922 break;
923 }
924
925 x->props.mode = tmpl->mode;
926 x->props.reqid = tmpl->reqid;
927 x->props.family = tmpl->encap_family;
928}
929
930static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
931 const xfrm_address_t *daddr,
932 __be32 spi, u8 proto,
933 unsigned short family)
934{
935 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
936 struct xfrm_state *x;
937
938 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
939 if (x->props.family != family ||
940 x->id.spi != spi ||
941 x->id.proto != proto ||
942 !xfrm_addr_equal(&x->id.daddr, daddr, family))
943 continue;
944
945 if ((mark & x->mark.m) != x->mark.v)
946 continue;
947 if (!xfrm_state_hold_rcu(x))
948 continue;
949 return x;
950 }
951
952 return NULL;
953}
954
955static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
956 const xfrm_address_t *daddr,
957 const xfrm_address_t *saddr,
958 u8 proto, unsigned short family)
959{
960 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
961 struct xfrm_state *x;
962
963 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
964 if (x->props.family != family ||
965 x->id.proto != proto ||
966 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
967 !xfrm_addr_equal(&x->props.saddr, saddr, family))
968 continue;
969
970 if ((mark & x->mark.m) != x->mark.v)
971 continue;
972 if (!xfrm_state_hold_rcu(x))
973 continue;
974 return x;
975 }
976
977 return NULL;
978}
979
980static inline struct xfrm_state *
981__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
982{
983 struct net *net = xs_net(x);
984 u32 mark = x->mark.v & x->mark.m;
985
986 if (use_spi)
987 return __xfrm_state_lookup(net, mark, &x->id.daddr,
988 x->id.spi, x->id.proto, family);
989 else
990 return __xfrm_state_lookup_byaddr(net, mark,
991 &x->id.daddr,
992 &x->props.saddr,
993 x->id.proto, family);
994}
995
996static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
997{
998 if (have_hash_collision &&
999 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1000 net->xfrm.state_num > net->xfrm.state_hmask)
1001 schedule_work(&net->xfrm.state_hash_work);
1002}
1003
1004static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1005 const struct flowi *fl, unsigned short family,
1006 struct xfrm_state **best, int *acq_in_progress,
1007 int *error)
1008{
1009 /* Resolution logic:
1010 * 1. There is a valid state with matching selector. Done.
1011 * 2. Valid state with inappropriate selector. Skip.
1012 *
1013 * Entering area of "sysdeps".
1014 *
1015 * 3. If state is not valid, selector is temporary, it selects
1016 * only session which triggered previous resolution. Key
1017 * manager will do something to install a state with proper
1018 * selector.
1019 */
1020 if (x->km.state == XFRM_STATE_VALID) {
1021 if ((x->sel.family &&
1022 (x->sel.family != family ||
1023 !xfrm_selector_match(&x->sel, fl, family))) ||
1024 !security_xfrm_state_pol_flow_match(x, pol, fl))
1025 return;
1026
1027 if (!*best ||
1028 (*best)->km.dying > x->km.dying ||
1029 ((*best)->km.dying == x->km.dying &&
1030 (*best)->curlft.add_time < x->curlft.add_time))
1031 *best = x;
1032 } else if (x->km.state == XFRM_STATE_ACQ) {
1033 *acq_in_progress = 1;
1034 } else if (x->km.state == XFRM_STATE_ERROR ||
1035 x->km.state == XFRM_STATE_EXPIRED) {
1036 if ((!x->sel.family ||
1037 (x->sel.family == family &&
1038 xfrm_selector_match(&x->sel, fl, family))) &&
1039 security_xfrm_state_pol_flow_match(x, pol, fl))
1040 *error = -ESRCH;
1041 }
1042}
1043
1044struct xfrm_state *
1045xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1046 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1047 struct xfrm_policy *pol, int *err,
1048 unsigned short family, u32 if_id)
1049{
1050 static xfrm_address_t saddr_wildcard = { };
1051 struct net *net = xp_net(pol);
1052 unsigned int h, h_wildcard;
1053 struct xfrm_state *x, *x0, *to_put;
1054 int acquire_in_progress = 0;
1055 int error = 0;
1056 struct xfrm_state *best = NULL;
1057 u32 mark = pol->mark.v & pol->mark.m;
1058 unsigned short encap_family = tmpl->encap_family;
1059 unsigned int sequence;
1060 struct km_event c;
1061
1062 to_put = NULL;
1063
1064 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
1065
1066 rcu_read_lock();
1067 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1068 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1069 if (x->props.family == encap_family &&
1070 x->props.reqid == tmpl->reqid &&
1071 (mark & x->mark.m) == x->mark.v &&
1072 x->if_id == if_id &&
1073 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1074 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1075 tmpl->mode == x->props.mode &&
1076 tmpl->id.proto == x->id.proto &&
1077 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1078 xfrm_state_look_at(pol, x, fl, family,
1079 &best, &acquire_in_progress, &error);
1080 }
1081 if (best || acquire_in_progress)
1082 goto found;
1083
1084 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1085 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1086 if (x->props.family == encap_family &&
1087 x->props.reqid == tmpl->reqid &&
1088 (mark & x->mark.m) == x->mark.v &&
1089 x->if_id == if_id &&
1090 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1091 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1092 tmpl->mode == x->props.mode &&
1093 tmpl->id.proto == x->id.proto &&
1094 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1095 xfrm_state_look_at(pol, x, fl, family,
1096 &best, &acquire_in_progress, &error);
1097 }
1098
1099found:
1100 x = best;
1101 if (!x && !error && !acquire_in_progress) {
1102 if (tmpl->id.spi &&
1103 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1104 tmpl->id.proto, encap_family)) != NULL) {
1105 to_put = x0;
1106 error = -EEXIST;
1107 goto out;
1108 }
1109
1110 c.net = net;
1111 /* If the KMs have no listeners (yet...), avoid allocating an SA
1112 * for each and every packet - garbage collection might not
1113 * handle the flood.
1114 */
1115 if (!km_is_alive(&c)) {
1116 error = -ESRCH;
1117 goto out;
1118 }
1119
1120 x = xfrm_state_alloc(net);
1121 if (x == NULL) {
1122 error = -ENOMEM;
1123 goto out;
1124 }
1125 /* Initialize temporary state matching only
1126 * to current session. */
1127 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1128 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1129 x->if_id = if_id;
1130
1131 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1132 if (error) {
1133 x->km.state = XFRM_STATE_DEAD;
1134 to_put = x;
1135 x = NULL;
1136 goto out;
1137 }
1138
1139 if (km_query(x, tmpl, pol) == 0) {
1140 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1141 x->km.state = XFRM_STATE_ACQ;
1142 list_add(&x->km.all, &net->xfrm.state_all);
1143 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1144 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1145 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1146 if (x->id.spi) {
1147 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1148 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1149 }
1150 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1151 hrtimer_start(&x->mtimer,
1152 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1153 HRTIMER_MODE_REL_SOFT);
1154 net->xfrm.state_num++;
1155 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1156 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1157 } else {
1158 x->km.state = XFRM_STATE_DEAD;
1159 to_put = x;
1160 x = NULL;
1161 error = -ESRCH;
1162 }
1163 }
1164out:
1165 if (x) {
1166 if (!xfrm_state_hold_rcu(x)) {
1167 *err = -EAGAIN;
1168 x = NULL;
1169 }
1170 } else {
1171 *err = acquire_in_progress ? -EAGAIN : error;
1172 }
1173 rcu_read_unlock();
1174 if (to_put)
1175 xfrm_state_put(to_put);
1176
1177 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1178 *err = -EAGAIN;
1179 if (x) {
1180 xfrm_state_put(x);
1181 x = NULL;
1182 }
1183 }
1184
1185 return x;
1186}
1187
1188struct xfrm_state *
1189xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1190 xfrm_address_t *daddr, xfrm_address_t *saddr,
1191 unsigned short family, u8 mode, u8 proto, u32 reqid)
1192{
1193 unsigned int h;
1194 struct xfrm_state *rx = NULL, *x = NULL;
1195
1196 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1197 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1198 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1199 if (x->props.family == family &&
1200 x->props.reqid == reqid &&
1201 (mark & x->mark.m) == x->mark.v &&
1202 x->if_id == if_id &&
1203 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1204 xfrm_state_addr_check(x, daddr, saddr, family) &&
1205 mode == x->props.mode &&
1206 proto == x->id.proto &&
1207 x->km.state == XFRM_STATE_VALID) {
1208 rx = x;
1209 break;
1210 }
1211 }
1212
1213 if (rx)
1214 xfrm_state_hold(rx);
1215 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1216
1217
1218 return rx;
1219}
1220EXPORT_SYMBOL(xfrm_stateonly_find);
1221
1222struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1223 unsigned short family)
1224{
1225 struct xfrm_state *x;
1226 struct xfrm_state_walk *w;
1227
1228 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1229 list_for_each_entry(w, &net->xfrm.state_all, all) {
1230 x = container_of(w, struct xfrm_state, km);
1231 if (x->props.family != family ||
1232 x->id.spi != spi)
1233 continue;
1234
1235 xfrm_state_hold(x);
1236 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1237 return x;
1238 }
1239 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1240 return NULL;
1241}
1242EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1243
1244static void __xfrm_state_insert(struct xfrm_state *x)
1245{
1246 struct net *net = xs_net(x);
1247 unsigned int h;
1248
1249 list_add(&x->km.all, &net->xfrm.state_all);
1250
1251 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1252 x->props.reqid, x->props.family);
1253 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1254
1255 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1256 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1257
1258 if (x->id.spi) {
1259 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1260 x->props.family);
1261
1262 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1263 }
1264
1265 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1266 if (x->replay_maxage)
1267 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1268
1269 net->xfrm.state_num++;
1270
1271 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1272}
1273
1274/* net->xfrm.xfrm_state_lock is held */
1275static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1276{
1277 struct net *net = xs_net(xnew);
1278 unsigned short family = xnew->props.family;
1279 u32 reqid = xnew->props.reqid;
1280 struct xfrm_state *x;
1281 unsigned int h;
1282 u32 mark = xnew->mark.v & xnew->mark.m;
1283 u32 if_id = xnew->if_id;
1284
1285 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1286 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1287 if (x->props.family == family &&
1288 x->props.reqid == reqid &&
1289 x->if_id == if_id &&
1290 (mark & x->mark.m) == x->mark.v &&
1291 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1292 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1293 x->genid++;
1294 }
1295}
1296
1297void xfrm_state_insert(struct xfrm_state *x)
1298{
1299 struct net *net = xs_net(x);
1300
1301 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1302 __xfrm_state_bump_genids(x);
1303 __xfrm_state_insert(x);
1304 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1305}
1306EXPORT_SYMBOL(xfrm_state_insert);
1307
1308/* net->xfrm.xfrm_state_lock is held */
1309static struct xfrm_state *__find_acq_core(struct net *net,
1310 const struct xfrm_mark *m,
1311 unsigned short family, u8 mode,
1312 u32 reqid, u32 if_id, u8 proto,
1313 const xfrm_address_t *daddr,
1314 const xfrm_address_t *saddr,
1315 int create)
1316{
1317 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1318 struct xfrm_state *x;
1319 u32 mark = m->v & m->m;
1320
1321 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1322 if (x->props.reqid != reqid ||
1323 x->props.mode != mode ||
1324 x->props.family != family ||
1325 x->km.state != XFRM_STATE_ACQ ||
1326 x->id.spi != 0 ||
1327 x->id.proto != proto ||
1328 (mark & x->mark.m) != x->mark.v ||
1329 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1330 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1331 continue;
1332
1333 xfrm_state_hold(x);
1334 return x;
1335 }
1336
1337 if (!create)
1338 return NULL;
1339
1340 x = xfrm_state_alloc(net);
1341 if (likely(x)) {
1342 switch (family) {
1343 case AF_INET:
1344 x->sel.daddr.a4 = daddr->a4;
1345 x->sel.saddr.a4 = saddr->a4;
1346 x->sel.prefixlen_d = 32;
1347 x->sel.prefixlen_s = 32;
1348 x->props.saddr.a4 = saddr->a4;
1349 x->id.daddr.a4 = daddr->a4;
1350 break;
1351
1352 case AF_INET6:
1353 x->sel.daddr.in6 = daddr->in6;
1354 x->sel.saddr.in6 = saddr->in6;
1355 x->sel.prefixlen_d = 128;
1356 x->sel.prefixlen_s = 128;
1357 x->props.saddr.in6 = saddr->in6;
1358 x->id.daddr.in6 = daddr->in6;
1359 break;
1360 }
1361
1362 x->km.state = XFRM_STATE_ACQ;
1363 x->id.proto = proto;
1364 x->props.family = family;
1365 x->props.mode = mode;
1366 x->props.reqid = reqid;
1367 x->if_id = if_id;
1368 x->mark.v = m->v;
1369 x->mark.m = m->m;
1370 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1371 xfrm_state_hold(x);
1372 hrtimer_start(&x->mtimer,
1373 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1374 HRTIMER_MODE_REL_SOFT);
1375 list_add(&x->km.all, &net->xfrm.state_all);
1376 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1377 h = xfrm_src_hash(net, daddr, saddr, family);
1378 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1379
1380 net->xfrm.state_num++;
1381
1382 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1383 }
1384
1385 return x;
1386}
1387
1388static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1389
1390int xfrm_state_add(struct xfrm_state *x)
1391{
1392 struct net *net = xs_net(x);
1393 struct xfrm_state *x1, *to_put;
1394 int family;
1395 int err;
1396 u32 mark = x->mark.v & x->mark.m;
1397 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1398
1399 family = x->props.family;
1400
1401 to_put = NULL;
1402
1403 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1404
1405 x1 = __xfrm_state_locate(x, use_spi, family);
1406 if (x1) {
1407 to_put = x1;
1408 x1 = NULL;
1409 err = -EEXIST;
1410 goto out;
1411 }
1412
1413 if (use_spi && x->km.seq) {
1414 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1415 if (x1 && ((x1->id.proto != x->id.proto) ||
1416 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1417 to_put = x1;
1418 x1 = NULL;
1419 }
1420 }
1421
1422 if (use_spi && !x1)
1423 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1424 x->props.reqid, x->if_id, x->id.proto,
1425 &x->id.daddr, &x->props.saddr, 0);
1426
1427 __xfrm_state_bump_genids(x);
1428 __xfrm_state_insert(x);
1429 err = 0;
1430
1431out:
1432 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1433
1434 if (x1) {
1435 xfrm_state_delete(x1);
1436 xfrm_state_put(x1);
1437 }
1438
1439 if (to_put)
1440 xfrm_state_put(to_put);
1441
1442 return err;
1443}
1444EXPORT_SYMBOL(xfrm_state_add);
1445
1446#ifdef CONFIG_XFRM_MIGRATE
1447static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1448{
1449 struct xfrm_user_sec_ctx *uctx;
1450 int size = sizeof(*uctx) + security->ctx_len;
1451 int err;
1452
1453 uctx = kmalloc(size, GFP_KERNEL);
1454 if (!uctx)
1455 return -ENOMEM;
1456
1457 uctx->exttype = XFRMA_SEC_CTX;
1458 uctx->len = size;
1459 uctx->ctx_doi = security->ctx_doi;
1460 uctx->ctx_alg = security->ctx_alg;
1461 uctx->ctx_len = security->ctx_len;
1462 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1463 err = security_xfrm_state_alloc(x, uctx);
1464 kfree(uctx);
1465 if (err)
1466 return err;
1467
1468 return 0;
1469}
1470
1471static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1472 struct xfrm_encap_tmpl *encap)
1473{
1474 struct net *net = xs_net(orig);
1475 struct xfrm_state *x = xfrm_state_alloc(net);
1476 if (!x)
1477 goto out;
1478
1479 memcpy(&x->id, &orig->id, sizeof(x->id));
1480 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1481 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1482 x->props.mode = orig->props.mode;
1483 x->props.replay_window = orig->props.replay_window;
1484 x->props.reqid = orig->props.reqid;
1485 x->props.family = orig->props.family;
1486 x->props.saddr = orig->props.saddr;
1487
1488 if (orig->aalg) {
1489 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1490 if (!x->aalg)
1491 goto error;
1492 }
1493 x->props.aalgo = orig->props.aalgo;
1494
1495 if (orig->aead) {
1496 x->aead = xfrm_algo_aead_clone(orig->aead);
1497 x->geniv = orig->geniv;
1498 if (!x->aead)
1499 goto error;
1500 }
1501 if (orig->ealg) {
1502 x->ealg = xfrm_algo_clone(orig->ealg);
1503 if (!x->ealg)
1504 goto error;
1505 }
1506 x->props.ealgo = orig->props.ealgo;
1507
1508 if (orig->calg) {
1509 x->calg = xfrm_algo_clone(orig->calg);
1510 if (!x->calg)
1511 goto error;
1512 }
1513 x->props.calgo = orig->props.calgo;
1514
1515 if (encap || orig->encap) {
1516 if (encap)
1517 x->encap = kmemdup(encap, sizeof(*x->encap),
1518 GFP_KERNEL);
1519 else
1520 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1521 GFP_KERNEL);
1522
1523 if (!x->encap)
1524 goto error;
1525 }
1526
1527 if (orig->security)
1528 if (clone_security(x, orig->security))
1529 goto error;
1530
1531 if (orig->coaddr) {
1532 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1533 GFP_KERNEL);
1534 if (!x->coaddr)
1535 goto error;
1536 }
1537
1538 if (orig->replay_esn) {
1539 if (xfrm_replay_clone(x, orig))
1540 goto error;
1541 }
1542
1543 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1544 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1545
1546 if (xfrm_init_state(x) < 0)
1547 goto error;
1548
1549 x->props.flags = orig->props.flags;
1550 x->props.extra_flags = orig->props.extra_flags;
1551
1552 x->if_id = orig->if_id;
1553 x->tfcpad = orig->tfcpad;
1554 x->replay_maxdiff = orig->replay_maxdiff;
1555 x->replay_maxage = orig->replay_maxage;
1556 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1557 x->km.state = orig->km.state;
1558 x->km.seq = orig->km.seq;
1559 x->replay = orig->replay;
1560 x->preplay = orig->preplay;
1561
1562 return x;
1563
1564 error:
1565 xfrm_state_put(x);
1566out:
1567 return NULL;
1568}
1569
1570struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1571{
1572 unsigned int h;
1573 struct xfrm_state *x = NULL;
1574
1575 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1576
1577 if (m->reqid) {
1578 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1579 m->reqid, m->old_family);
1580 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1581 if (x->props.mode != m->mode ||
1582 x->id.proto != m->proto)
1583 continue;
1584 if (m->reqid && x->props.reqid != m->reqid)
1585 continue;
1586 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1587 m->old_family) ||
1588 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1589 m->old_family))
1590 continue;
1591 xfrm_state_hold(x);
1592 break;
1593 }
1594 } else {
1595 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1596 m->old_family);
1597 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1598 if (x->props.mode != m->mode ||
1599 x->id.proto != m->proto)
1600 continue;
1601 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1602 m->old_family) ||
1603 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1604 m->old_family))
1605 continue;
1606 xfrm_state_hold(x);
1607 break;
1608 }
1609 }
1610
1611 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1612
1613 return x;
1614}
1615EXPORT_SYMBOL(xfrm_migrate_state_find);
1616
1617struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1618 struct xfrm_migrate *m,
1619 struct xfrm_encap_tmpl *encap)
1620{
1621 struct xfrm_state *xc;
1622
1623 xc = xfrm_state_clone(x, encap);
1624 if (!xc)
1625 return NULL;
1626
1627 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1628 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1629
1630 /* add state */
1631 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1632 /* a care is needed when the destination address of the
1633 state is to be updated as it is a part of triplet */
1634 xfrm_state_insert(xc);
1635 } else {
1636 if (xfrm_state_add(xc) < 0)
1637 goto error;
1638 }
1639
1640 return xc;
1641error:
1642 xfrm_state_put(xc);
1643 return NULL;
1644}
1645EXPORT_SYMBOL(xfrm_state_migrate);
1646#endif
1647
1648int xfrm_state_update(struct xfrm_state *x)
1649{
1650 struct xfrm_state *x1, *to_put;
1651 int err;
1652 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1653 struct net *net = xs_net(x);
1654
1655 to_put = NULL;
1656
1657 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1658 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1659
1660 err = -ESRCH;
1661 if (!x1)
1662 goto out;
1663
1664 if (xfrm_state_kern(x1)) {
1665 to_put = x1;
1666 err = -EEXIST;
1667 goto out;
1668 }
1669
1670 if (x1->km.state == XFRM_STATE_ACQ) {
1671 __xfrm_state_insert(x);
1672 x = NULL;
1673 }
1674 err = 0;
1675
1676out:
1677 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1678
1679 if (to_put)
1680 xfrm_state_put(to_put);
1681
1682 if (err)
1683 return err;
1684
1685 if (!x) {
1686 xfrm_state_delete(x1);
1687 xfrm_state_put(x1);
1688 return 0;
1689 }
1690
1691 err = -EINVAL;
1692 spin_lock_bh(&x1->lock);
1693 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1694 if (x->encap && x1->encap &&
1695 x->encap->encap_type == x1->encap->encap_type)
1696 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1697 else if (x->encap || x1->encap)
1698 goto fail;
1699
1700 if (x->coaddr && x1->coaddr) {
1701 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1702 }
1703 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1704 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1705 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1706 x1->km.dying = 0;
1707
1708 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1709 HRTIMER_MODE_REL_SOFT);
1710 if (x1->curlft.use_time)
1711 xfrm_state_check_expire(x1);
1712
1713 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1714 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1715
1716 if (x->props.smark.m || x->props.smark.v)
1717 x1->props.smark = x->props.smark;
1718
1719 if (x->if_id)
1720 x1->if_id = x->if_id;
1721
1722 __xfrm_state_bump_genids(x1);
1723 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1724 }
1725
1726 err = 0;
1727 x->km.state = XFRM_STATE_DEAD;
1728 __xfrm_state_put(x);
1729 }
1730
1731fail:
1732 spin_unlock_bh(&x1->lock);
1733
1734 xfrm_state_put(x1);
1735
1736 return err;
1737}
1738EXPORT_SYMBOL(xfrm_state_update);
1739
1740int xfrm_state_check_expire(struct xfrm_state *x)
1741{
1742 if (!x->curlft.use_time)
1743 x->curlft.use_time = ktime_get_real_seconds();
1744
1745 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1746 x->curlft.packets >= x->lft.hard_packet_limit) {
1747 x->km.state = XFRM_STATE_EXPIRED;
1748 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1749 return -EINVAL;
1750 }
1751
1752 if (!x->km.dying &&
1753 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1754 x->curlft.packets >= x->lft.soft_packet_limit)) {
1755 x->km.dying = 1;
1756 km_state_expired(x, 0, 0);
1757 }
1758 return 0;
1759}
1760EXPORT_SYMBOL(xfrm_state_check_expire);
1761
1762struct xfrm_state *
1763xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1764 u8 proto, unsigned short family)
1765{
1766 struct xfrm_state *x;
1767
1768 rcu_read_lock();
1769 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1770 rcu_read_unlock();
1771 return x;
1772}
1773EXPORT_SYMBOL(xfrm_state_lookup);
1774
1775struct xfrm_state *
1776xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1777 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1778 u8 proto, unsigned short family)
1779{
1780 struct xfrm_state *x;
1781
1782 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1783 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1784 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1785 return x;
1786}
1787EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1788
1789struct xfrm_state *
1790xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1791 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1792 const xfrm_address_t *saddr, int create, unsigned short family)
1793{
1794 struct xfrm_state *x;
1795
1796 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1797 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1798 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1799
1800 return x;
1801}
1802EXPORT_SYMBOL(xfrm_find_acq);
1803
1804#ifdef CONFIG_XFRM_SUB_POLICY
1805#if IS_ENABLED(CONFIG_IPV6)
1806/* distribution counting sort function for xfrm_state and xfrm_tmpl */
1807static void
1808__xfrm6_sort(void **dst, void **src, int n,
1809 int (*cmp)(const void *p), int maxclass)
1810{
1811 int count[XFRM_MAX_DEPTH] = { };
1812 int class[XFRM_MAX_DEPTH];
1813 int i;
1814
1815 for (i = 0; i < n; i++) {
1816 int c = cmp(src[i]);
1817
1818 class[i] = c;
1819 count[c]++;
1820 }
1821
1822 for (i = 2; i < maxclass; i++)
1823 count[i] += count[i - 1];
1824
1825 for (i = 0; i < n; i++) {
1826 dst[count[class[i] - 1]++] = src[i];
1827 src[i] = NULL;
1828 }
1829}
1830
1831/* Rule for xfrm_state:
1832 *
1833 * rule 1: select IPsec transport except AH
1834 * rule 2: select MIPv6 RO or inbound trigger
1835 * rule 3: select IPsec transport AH
1836 * rule 4: select IPsec tunnel
1837 * rule 5: others
1838 */
1839static int __xfrm6_state_sort_cmp(const void *p)
1840{
1841 const struct xfrm_state *v = p;
1842
1843 switch (v->props.mode) {
1844 case XFRM_MODE_TRANSPORT:
1845 if (v->id.proto != IPPROTO_AH)
1846 return 1;
1847 else
1848 return 3;
1849#if IS_ENABLED(CONFIG_IPV6_MIP6)
1850 case XFRM_MODE_ROUTEOPTIMIZATION:
1851 case XFRM_MODE_IN_TRIGGER:
1852 return 2;
1853#endif
1854 case XFRM_MODE_TUNNEL:
1855 case XFRM_MODE_BEET:
1856 return 4;
1857 }
1858 return 5;
1859}
1860
1861/* Rule for xfrm_tmpl:
1862 *
1863 * rule 1: select IPsec transport
1864 * rule 2: select MIPv6 RO or inbound trigger
1865 * rule 3: select IPsec tunnel
1866 * rule 4: others
1867 */
1868static int __xfrm6_tmpl_sort_cmp(const void *p)
1869{
1870 const struct xfrm_tmpl *v = p;
1871
1872 switch (v->mode) {
1873 case XFRM_MODE_TRANSPORT:
1874 return 1;
1875#if IS_ENABLED(CONFIG_IPV6_MIP6)
1876 case XFRM_MODE_ROUTEOPTIMIZATION:
1877 case XFRM_MODE_IN_TRIGGER:
1878 return 2;
1879#endif
1880 case XFRM_MODE_TUNNEL:
1881 case XFRM_MODE_BEET:
1882 return 3;
1883 }
1884 return 4;
1885}
1886#else
1887static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1888static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1889
1890static inline void
1891__xfrm6_sort(void **dst, void **src, int n,
1892 int (*cmp)(const void *p), int maxclass)
1893{
1894 int i;
1895
1896 for (i = 0; i < n; i++)
1897 dst[i] = src[i];
1898}
1899#endif /* CONFIG_IPV6 */
1900
1901void
1902xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1903 unsigned short family)
1904{
1905 int i;
1906
1907 if (family == AF_INET6)
1908 __xfrm6_sort((void **)dst, (void **)src, n,
1909 __xfrm6_tmpl_sort_cmp, 5);
1910 else
1911 for (i = 0; i < n; i++)
1912 dst[i] = src[i];
1913}
1914
1915void
1916xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1917 unsigned short family)
1918{
1919 int i;
1920
1921 if (family == AF_INET6)
1922 __xfrm6_sort((void **)dst, (void **)src, n,
1923 __xfrm6_state_sort_cmp, 6);
1924 else
1925 for (i = 0; i < n; i++)
1926 dst[i] = src[i];
1927}
1928#endif
1929
1930/* Silly enough, but I'm lazy to build resolution list */
1931
1932static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1933{
1934 int i;
1935
1936 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1937 struct xfrm_state *x;
1938
1939 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1940 if (x->km.seq == seq &&
1941 (mark & x->mark.m) == x->mark.v &&
1942 x->km.state == XFRM_STATE_ACQ) {
1943 xfrm_state_hold(x);
1944 return x;
1945 }
1946 }
1947 }
1948 return NULL;
1949}
1950
1951struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1952{
1953 struct xfrm_state *x;
1954
1955 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1956 x = __xfrm_find_acq_byseq(net, mark, seq);
1957 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1958 return x;
1959}
1960EXPORT_SYMBOL(xfrm_find_acq_byseq);
1961
1962u32 xfrm_get_acqseq(void)
1963{
1964 u32 res;
1965 static atomic_t acqseq;
1966
1967 do {
1968 res = atomic_inc_return(&acqseq);
1969 } while (!res);
1970
1971 return res;
1972}
1973EXPORT_SYMBOL(xfrm_get_acqseq);
1974
1975int verify_spi_info(u8 proto, u32 min, u32 max)
1976{
1977 switch (proto) {
1978 case IPPROTO_AH:
1979 case IPPROTO_ESP:
1980 break;
1981
1982 case IPPROTO_COMP:
1983 /* IPCOMP spi is 16-bits. */
1984 if (max >= 0x10000)
1985 return -EINVAL;
1986 break;
1987
1988 default:
1989 return -EINVAL;
1990 }
1991
1992 if (min > max)
1993 return -EINVAL;
1994
1995 return 0;
1996}
1997EXPORT_SYMBOL(verify_spi_info);
1998
1999int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2000{
2001 struct net *net = xs_net(x);
2002 unsigned int h;
2003 struct xfrm_state *x0;
2004 int err = -ENOENT;
2005 __be32 minspi = htonl(low);
2006 __be32 maxspi = htonl(high);
2007 u32 mark = x->mark.v & x->mark.m;
2008
2009 spin_lock_bh(&x->lock);
2010 if (x->km.state == XFRM_STATE_DEAD)
2011 goto unlock;
2012
2013 err = 0;
2014 if (x->id.spi)
2015 goto unlock;
2016
2017 err = -ENOENT;
2018
2019 if (minspi == maxspi) {
2020 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2021 if (x0) {
2022 xfrm_state_put(x0);
2023 goto unlock;
2024 }
2025 x->id.spi = minspi;
2026 } else {
2027 u32 spi = 0;
2028 for (h = 0; h < high-low+1; h++) {
2029 spi = low + prandom_u32()%(high-low+1);
2030 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2031 if (x0 == NULL) {
2032 x->id.spi = htonl(spi);
2033 break;
2034 }
2035 xfrm_state_put(x0);
2036 }
2037 }
2038 if (x->id.spi) {
2039 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2040 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2041 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2042 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2043
2044 err = 0;
2045 }
2046
2047unlock:
2048 spin_unlock_bh(&x->lock);
2049
2050 return err;
2051}
2052EXPORT_SYMBOL(xfrm_alloc_spi);
2053
2054static bool __xfrm_state_filter_match(struct xfrm_state *x,
2055 struct xfrm_address_filter *filter)
2056{
2057 if (filter) {
2058 if ((filter->family == AF_INET ||
2059 filter->family == AF_INET6) &&
2060 x->props.family != filter->family)
2061 return false;
2062
2063 return addr_match(&x->props.saddr, &filter->saddr,
2064 filter->splen) &&
2065 addr_match(&x->id.daddr, &filter->daddr,
2066 filter->dplen);
2067 }
2068 return true;
2069}
2070
2071int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2072 int (*func)(struct xfrm_state *, int, void*),
2073 void *data)
2074{
2075 struct xfrm_state *state;
2076 struct xfrm_state_walk *x;
2077 int err = 0;
2078
2079 if (walk->seq != 0 && list_empty(&walk->all))
2080 return 0;
2081
2082 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2083 if (list_empty(&walk->all))
2084 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2085 else
2086 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2087 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2088 if (x->state == XFRM_STATE_DEAD)
2089 continue;
2090 state = container_of(x, struct xfrm_state, km);
2091 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2092 continue;
2093 if (!__xfrm_state_filter_match(state, walk->filter))
2094 continue;
2095 err = func(state, walk->seq, data);
2096 if (err) {
2097 list_move_tail(&walk->all, &x->all);
2098 goto out;
2099 }
2100 walk->seq++;
2101 }
2102 if (walk->seq == 0) {
2103 err = -ENOENT;
2104 goto out;
2105 }
2106 list_del_init(&walk->all);
2107out:
2108 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2109 return err;
2110}
2111EXPORT_SYMBOL(xfrm_state_walk);
2112
2113void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2114 struct xfrm_address_filter *filter)
2115{
2116 INIT_LIST_HEAD(&walk->all);
2117 walk->proto = proto;
2118 walk->state = XFRM_STATE_DEAD;
2119 walk->seq = 0;
2120 walk->filter = filter;
2121}
2122EXPORT_SYMBOL(xfrm_state_walk_init);
2123
2124void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2125{
2126 kfree(walk->filter);
2127
2128 if (list_empty(&walk->all))
2129 return;
2130
2131 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2132 list_del(&walk->all);
2133 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2134}
2135EXPORT_SYMBOL(xfrm_state_walk_done);
2136
2137static void xfrm_replay_timer_handler(struct timer_list *t)
2138{
2139 struct xfrm_state *x = from_timer(x, t, rtimer);
2140
2141 spin_lock(&x->lock);
2142
2143 if (x->km.state == XFRM_STATE_VALID) {
2144 if (xfrm_aevent_is_on(xs_net(x)))
2145 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
2146 else
2147 x->xflags |= XFRM_TIME_DEFER;
2148 }
2149
2150 spin_unlock(&x->lock);
2151}
2152
2153static LIST_HEAD(xfrm_km_list);
2154
2155void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2156{
2157 struct xfrm_mgr *km;
2158
2159 rcu_read_lock();
2160 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2161 if (km->notify_policy)
2162 km->notify_policy(xp, dir, c);
2163 rcu_read_unlock();
2164}
2165
2166void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2167{
2168 struct xfrm_mgr *km;
2169 rcu_read_lock();
2170 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2171 if (km->notify)
2172 km->notify(x, c);
2173 rcu_read_unlock();
2174}
2175
2176EXPORT_SYMBOL(km_policy_notify);
2177EXPORT_SYMBOL(km_state_notify);
2178
2179void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2180{
2181 struct km_event c;
2182
2183 c.data.hard = hard;
2184 c.portid = portid;
2185 c.event = XFRM_MSG_EXPIRE;
2186 km_state_notify(x, &c);
2187}
2188
2189EXPORT_SYMBOL(km_state_expired);
2190/*
2191 * We send to all registered managers regardless of failure
2192 * We are happy with one success
2193*/
2194int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2195{
2196 int err = -EINVAL, acqret;
2197 struct xfrm_mgr *km;
2198
2199 rcu_read_lock();
2200 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2201 acqret = km->acquire(x, t, pol);
2202 if (!acqret)
2203 err = acqret;
2204 }
2205 rcu_read_unlock();
2206 return err;
2207}
2208EXPORT_SYMBOL(km_query);
2209
2210int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2211{
2212 int err = -EINVAL;
2213 struct xfrm_mgr *km;
2214
2215 rcu_read_lock();
2216 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2217 if (km->new_mapping)
2218 err = km->new_mapping(x, ipaddr, sport);
2219 if (!err)
2220 break;
2221 }
2222 rcu_read_unlock();
2223 return err;
2224}
2225EXPORT_SYMBOL(km_new_mapping);
2226
2227void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2228{
2229 struct km_event c;
2230
2231 c.data.hard = hard;
2232 c.portid = portid;
2233 c.event = XFRM_MSG_POLEXPIRE;
2234 km_policy_notify(pol, dir, &c);
2235}
2236EXPORT_SYMBOL(km_policy_expired);
2237
2238#ifdef CONFIG_XFRM_MIGRATE
2239int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2240 const struct xfrm_migrate *m, int num_migrate,
2241 const struct xfrm_kmaddress *k,
2242 const struct xfrm_encap_tmpl *encap)
2243{
2244 int err = -EINVAL;
2245 int ret;
2246 struct xfrm_mgr *km;
2247
2248 rcu_read_lock();
2249 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2250 if (km->migrate) {
2251 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2252 encap);
2253 if (!ret)
2254 err = ret;
2255 }
2256 }
2257 rcu_read_unlock();
2258 return err;
2259}
2260EXPORT_SYMBOL(km_migrate);
2261#endif
2262
2263int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2264{
2265 int err = -EINVAL;
2266 int ret;
2267 struct xfrm_mgr *km;
2268
2269 rcu_read_lock();
2270 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2271 if (km->report) {
2272 ret = km->report(net, proto, sel, addr);
2273 if (!ret)
2274 err = ret;
2275 }
2276 }
2277 rcu_read_unlock();
2278 return err;
2279}
2280EXPORT_SYMBOL(km_report);
2281
2282static bool km_is_alive(const struct km_event *c)
2283{
2284 struct xfrm_mgr *km;
2285 bool is_alive = false;
2286
2287 rcu_read_lock();
2288 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2289 if (km->is_alive && km->is_alive(c)) {
2290 is_alive = true;
2291 break;
2292 }
2293 }
2294 rcu_read_unlock();
2295
2296 return is_alive;
2297}
2298
2299int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2300{
2301 int err;
2302 u8 *data;
2303 struct xfrm_mgr *km;
2304 struct xfrm_policy *pol = NULL;
2305
2306 if (in_compat_syscall())
2307 return -EOPNOTSUPP;
2308
2309 if (sockptr_is_null(optval) && !optlen) {
2310 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2311 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2312 __sk_dst_reset(sk);
2313 return 0;
2314 }
2315
2316 if (optlen <= 0 || optlen > PAGE_SIZE)
2317 return -EMSGSIZE;
2318
2319 data = memdup_sockptr(optval, optlen);
2320 if (IS_ERR(data))
2321 return PTR_ERR(data);
2322
2323 err = -EINVAL;
2324 rcu_read_lock();
2325 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2326 pol = km->compile_policy(sk, optname, data,
2327 optlen, &err);
2328 if (err >= 0)
2329 break;
2330 }
2331 rcu_read_unlock();
2332
2333 if (err >= 0) {
2334 xfrm_sk_policy_insert(sk, err, pol);
2335 xfrm_pol_put(pol);
2336 __sk_dst_reset(sk);
2337 err = 0;
2338 }
2339
2340 kfree(data);
2341 return err;
2342}
2343EXPORT_SYMBOL(xfrm_user_policy);
2344
2345static DEFINE_SPINLOCK(xfrm_km_lock);
2346
2347int xfrm_register_km(struct xfrm_mgr *km)
2348{
2349 spin_lock_bh(&xfrm_km_lock);
2350 list_add_tail_rcu(&km->list, &xfrm_km_list);
2351 spin_unlock_bh(&xfrm_km_lock);
2352 return 0;
2353}
2354EXPORT_SYMBOL(xfrm_register_km);
2355
2356int xfrm_unregister_km(struct xfrm_mgr *km)
2357{
2358 spin_lock_bh(&xfrm_km_lock);
2359 list_del_rcu(&km->list);
2360 spin_unlock_bh(&xfrm_km_lock);
2361 synchronize_rcu();
2362 return 0;
2363}
2364EXPORT_SYMBOL(xfrm_unregister_km);
2365
2366int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2367{
2368 int err = 0;
2369
2370 if (WARN_ON(afinfo->family >= NPROTO))
2371 return -EAFNOSUPPORT;
2372
2373 spin_lock_bh(&xfrm_state_afinfo_lock);
2374 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2375 err = -EEXIST;
2376 else
2377 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2378 spin_unlock_bh(&xfrm_state_afinfo_lock);
2379 return err;
2380}
2381EXPORT_SYMBOL(xfrm_state_register_afinfo);
2382
2383int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2384{
2385 int err = 0, family = afinfo->family;
2386
2387 if (WARN_ON(family >= NPROTO))
2388 return -EAFNOSUPPORT;
2389
2390 spin_lock_bh(&xfrm_state_afinfo_lock);
2391 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2392 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2393 err = -EINVAL;
2394 else
2395 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2396 }
2397 spin_unlock_bh(&xfrm_state_afinfo_lock);
2398 synchronize_rcu();
2399 return err;
2400}
2401EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2402
2403struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2404{
2405 if (unlikely(family >= NPROTO))
2406 return NULL;
2407
2408 return rcu_dereference(xfrm_state_afinfo[family]);
2409}
2410EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2411
2412struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2413{
2414 struct xfrm_state_afinfo *afinfo;
2415 if (unlikely(family >= NPROTO))
2416 return NULL;
2417 rcu_read_lock();
2418 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2419 if (unlikely(!afinfo))
2420 rcu_read_unlock();
2421 return afinfo;
2422}
2423
2424void xfrm_flush_gc(void)
2425{
2426 flush_work(&xfrm_state_gc_work);
2427}
2428EXPORT_SYMBOL(xfrm_flush_gc);
2429
2430/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2431void xfrm_state_delete_tunnel(struct xfrm_state *x)
2432{
2433 if (x->tunnel) {
2434 struct xfrm_state *t = x->tunnel;
2435
2436 if (atomic_read(&t->tunnel_users) == 2)
2437 xfrm_state_delete(t);
2438 atomic_dec(&t->tunnel_users);
2439 xfrm_state_put_sync(t);
2440 x->tunnel = NULL;
2441 }
2442}
2443EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2444
2445u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2446{
2447 const struct xfrm_type *type = READ_ONCE(x->type);
2448 struct crypto_aead *aead;
2449 u32 blksize, net_adj = 0;
2450
2451 if (x->km.state != XFRM_STATE_VALID ||
2452 !type || type->proto != IPPROTO_ESP)
2453 return mtu - x->props.header_len;
2454
2455 aead = x->data;
2456 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2457
2458 switch (x->props.mode) {
2459 case XFRM_MODE_TRANSPORT:
2460 case XFRM_MODE_BEET:
2461 if (x->props.family == AF_INET)
2462 net_adj = sizeof(struct iphdr);
2463 else if (x->props.family == AF_INET6)
2464 net_adj = sizeof(struct ipv6hdr);
2465 break;
2466 case XFRM_MODE_TUNNEL:
2467 break;
2468 default:
2469 WARN_ON_ONCE(1);
2470 break;
2471 }
2472
2473 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2474 net_adj) & ~(blksize - 1)) + net_adj - 2;
2475}
2476EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2477
2478int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2479{
2480 const struct xfrm_mode *inner_mode;
2481 const struct xfrm_mode *outer_mode;
2482 int family = x->props.family;
2483 int err;
2484
2485 if (family == AF_INET &&
2486 xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)
2487 x->props.flags |= XFRM_STATE_NOPMTUDISC;
2488
2489 err = -EPROTONOSUPPORT;
2490
2491 if (x->sel.family != AF_UNSPEC) {
2492 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2493 if (inner_mode == NULL)
2494 goto error;
2495
2496 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2497 family != x->sel.family)
2498 goto error;
2499
2500 x->inner_mode = *inner_mode;
2501 } else {
2502 const struct xfrm_mode *inner_mode_iaf;
2503 int iafamily = AF_INET;
2504
2505 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2506 if (inner_mode == NULL)
2507 goto error;
2508
2509 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL))
2510 goto error;
2511
2512 x->inner_mode = *inner_mode;
2513
2514 if (x->props.family == AF_INET)
2515 iafamily = AF_INET6;
2516
2517 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2518 if (inner_mode_iaf) {
2519 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2520 x->inner_mode_iaf = *inner_mode_iaf;
2521 }
2522 }
2523
2524 x->type = xfrm_get_type(x->id.proto, family);
2525 if (x->type == NULL)
2526 goto error;
2527
2528 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2529
2530 err = x->type->init_state(x);
2531 if (err)
2532 goto error;
2533
2534 outer_mode = xfrm_get_mode(x->props.mode, family);
2535 if (!outer_mode) {
2536 err = -EPROTONOSUPPORT;
2537 goto error;
2538 }
2539
2540 x->outer_mode = *outer_mode;
2541 if (init_replay) {
2542 err = xfrm_init_replay(x);
2543 if (err)
2544 goto error;
2545 }
2546
2547error:
2548 return err;
2549}
2550
2551EXPORT_SYMBOL(__xfrm_init_state);
2552
2553int xfrm_init_state(struct xfrm_state *x)
2554{
2555 int err;
2556
2557 err = __xfrm_init_state(x, true, false);
2558 if (!err)
2559 x->km.state = XFRM_STATE_VALID;
2560
2561 return err;
2562}
2563
2564EXPORT_SYMBOL(xfrm_init_state);
2565
2566int __net_init xfrm_state_init(struct net *net)
2567{
2568 unsigned int sz;
2569
2570 if (net_eq(net, &init_net))
2571 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2572 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2573
2574 INIT_LIST_HEAD(&net->xfrm.state_all);
2575
2576 sz = sizeof(struct hlist_head) * 8;
2577
2578 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2579 if (!net->xfrm.state_bydst)
2580 goto out_bydst;
2581 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2582 if (!net->xfrm.state_bysrc)
2583 goto out_bysrc;
2584 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2585 if (!net->xfrm.state_byspi)
2586 goto out_byspi;
2587 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2588
2589 net->xfrm.state_num = 0;
2590 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2591 spin_lock_init(&net->xfrm.xfrm_state_lock);
2592 return 0;
2593
2594out_byspi:
2595 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2596out_bysrc:
2597 xfrm_hash_free(net->xfrm.state_bydst, sz);
2598out_bydst:
2599 return -ENOMEM;
2600}
2601
2602void xfrm_state_fini(struct net *net)
2603{
2604 unsigned int sz;
2605
2606 flush_work(&net->xfrm.state_hash_work);
2607 flush_work(&xfrm_state_gc_work);
2608 xfrm_state_flush(net, 0, false, true);
2609
2610 WARN_ON(!list_empty(&net->xfrm.state_all));
2611
2612 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2613 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2614 xfrm_hash_free(net->xfrm.state_byspi, sz);
2615 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2616 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2617 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2618 xfrm_hash_free(net->xfrm.state_bydst, sz);
2619}
2620
2621#ifdef CONFIG_AUDITSYSCALL
2622static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2623 struct audit_buffer *audit_buf)
2624{
2625 struct xfrm_sec_ctx *ctx = x->security;
2626 u32 spi = ntohl(x->id.spi);
2627
2628 if (ctx)
2629 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2630 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2631
2632 switch (x->props.family) {
2633 case AF_INET:
2634 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2635 &x->props.saddr.a4, &x->id.daddr.a4);
2636 break;
2637 case AF_INET6:
2638 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2639 x->props.saddr.a6, x->id.daddr.a6);
2640 break;
2641 }
2642
2643 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2644}
2645
2646static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2647 struct audit_buffer *audit_buf)
2648{
2649 const struct iphdr *iph4;
2650 const struct ipv6hdr *iph6;
2651
2652 switch (family) {
2653 case AF_INET:
2654 iph4 = ip_hdr(skb);
2655 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2656 &iph4->saddr, &iph4->daddr);
2657 break;
2658 case AF_INET6:
2659 iph6 = ipv6_hdr(skb);
2660 audit_log_format(audit_buf,
2661 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2662 &iph6->saddr, &iph6->daddr,
2663 iph6->flow_lbl[0] & 0x0f,
2664 iph6->flow_lbl[1],
2665 iph6->flow_lbl[2]);
2666 break;
2667 }
2668}
2669
2670void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2671{
2672 struct audit_buffer *audit_buf;
2673
2674 audit_buf = xfrm_audit_start("SAD-add");
2675 if (audit_buf == NULL)
2676 return;
2677 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2678 xfrm_audit_helper_sainfo(x, audit_buf);
2679 audit_log_format(audit_buf, " res=%u", result);
2680 audit_log_end(audit_buf);
2681}
2682EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2683
2684void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2685{
2686 struct audit_buffer *audit_buf;
2687
2688 audit_buf = xfrm_audit_start("SAD-delete");
2689 if (audit_buf == NULL)
2690 return;
2691 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2692 xfrm_audit_helper_sainfo(x, audit_buf);
2693 audit_log_format(audit_buf, " res=%u", result);
2694 audit_log_end(audit_buf);
2695}
2696EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2697
2698void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2699 struct sk_buff *skb)
2700{
2701 struct audit_buffer *audit_buf;
2702 u32 spi;
2703
2704 audit_buf = xfrm_audit_start("SA-replay-overflow");
2705 if (audit_buf == NULL)
2706 return;
2707 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2708 /* don't record the sequence number because it's inherent in this kind
2709 * of audit message */
2710 spi = ntohl(x->id.spi);
2711 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2712 audit_log_end(audit_buf);
2713}
2714EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2715
2716void xfrm_audit_state_replay(struct xfrm_state *x,
2717 struct sk_buff *skb, __be32 net_seq)
2718{
2719 struct audit_buffer *audit_buf;
2720 u32 spi;
2721
2722 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2723 if (audit_buf == NULL)
2724 return;
2725 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2726 spi = ntohl(x->id.spi);
2727 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2728 spi, spi, ntohl(net_seq));
2729 audit_log_end(audit_buf);
2730}
2731EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2732
2733void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2734{
2735 struct audit_buffer *audit_buf;
2736
2737 audit_buf = xfrm_audit_start("SA-notfound");
2738 if (audit_buf == NULL)
2739 return;
2740 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2741 audit_log_end(audit_buf);
2742}
2743EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2744
2745void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2746 __be32 net_spi, __be32 net_seq)
2747{
2748 struct audit_buffer *audit_buf;
2749 u32 spi;
2750
2751 audit_buf = xfrm_audit_start("SA-notfound");
2752 if (audit_buf == NULL)
2753 return;
2754 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2755 spi = ntohl(net_spi);
2756 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2757 spi, spi, ntohl(net_seq));
2758 audit_log_end(audit_buf);
2759}
2760EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2761
2762void xfrm_audit_state_icvfail(struct xfrm_state *x,
2763 struct sk_buff *skb, u8 proto)
2764{
2765 struct audit_buffer *audit_buf;
2766 __be32 net_spi;
2767 __be32 net_seq;
2768
2769 audit_buf = xfrm_audit_start("SA-icv-failure");
2770 if (audit_buf == NULL)
2771 return;
2772 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2773 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2774 u32 spi = ntohl(net_spi);
2775 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2776 spi, spi, ntohl(net_seq));
2777 }
2778 audit_log_end(audit_buf);
2779}
2780EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2781#endif /* CONFIG_AUDITSYSCALL */