Loading...
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/compat.h>
18#include <linux/workqueue.h>
19#include <net/xfrm.h>
20#include <linux/pfkeyv2.h>
21#include <linux/ipsec.h>
22#include <linux/module.h>
23#include <linux/cache.h>
24#include <linux/audit.h>
25#include <linux/uaccess.h>
26#include <linux/ktime.h>
27#include <linux/slab.h>
28#include <linux/interrupt.h>
29#include <linux/kernel.h>
30
31#include <crypto/aead.h>
32
33#include "xfrm_hash.h"
34
35#define xfrm_state_deref_prot(table, net) \
36 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
37#define xfrm_state_deref_check(table, net) \
38 rcu_dereference_check((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
39
40static void xfrm_state_gc_task(struct work_struct *work);
41
42/* Each xfrm_state may be linked to two tables:
43
44 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
45 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
46 destination/tunnel endpoint. (output)
47 */
48
49static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
50static struct kmem_cache *xfrm_state_cache __ro_after_init;
51
52static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
53static HLIST_HEAD(xfrm_state_gc_list);
54static HLIST_HEAD(xfrm_state_dev_gc_list);
55
56static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
57{
58 return refcount_inc_not_zero(&x->refcnt);
59}
60
61static inline unsigned int xfrm_dst_hash(struct net *net,
62 const xfrm_address_t *daddr,
63 const xfrm_address_t *saddr,
64 u32 reqid,
65 unsigned short family)
66{
67 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
68
69 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
70}
71
72static inline unsigned int xfrm_src_hash(struct net *net,
73 const xfrm_address_t *daddr,
74 const xfrm_address_t *saddr,
75 unsigned short family)
76{
77 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
78
79 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
80}
81
82static inline unsigned int
83xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
84 __be32 spi, u8 proto, unsigned short family)
85{
86 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
87
88 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
89}
90
91static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
92{
93 lockdep_assert_held(&net->xfrm.xfrm_state_lock);
94
95 return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
96}
97
98#define XFRM_STATE_INSERT(by, _n, _h, _type) \
99 { \
100 struct xfrm_state *_x = NULL; \
101 \
102 if (_type != XFRM_DEV_OFFLOAD_PACKET) { \
103 hlist_for_each_entry_rcu(_x, _h, by) { \
104 if (_x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
105 continue; \
106 break; \
107 } \
108 } \
109 \
110 if (!_x || _x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
111 /* SAD is empty or consist from HW SAs only */ \
112 hlist_add_head_rcu(_n, _h); \
113 else \
114 hlist_add_before_rcu(_n, &_x->by); \
115 }
116
117static void xfrm_hash_transfer(struct hlist_head *list,
118 struct hlist_head *ndsttable,
119 struct hlist_head *nsrctable,
120 struct hlist_head *nspitable,
121 struct hlist_head *nseqtable,
122 unsigned int nhashmask)
123{
124 struct hlist_node *tmp;
125 struct xfrm_state *x;
126
127 hlist_for_each_entry_safe(x, tmp, list, bydst) {
128 unsigned int h;
129
130 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
131 x->props.reqid, x->props.family,
132 nhashmask);
133 XFRM_STATE_INSERT(bydst, &x->bydst, ndsttable + h, x->xso.type);
134
135 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
136 x->props.family,
137 nhashmask);
138 XFRM_STATE_INSERT(bysrc, &x->bysrc, nsrctable + h, x->xso.type);
139
140 if (x->id.spi) {
141 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
142 x->id.proto, x->props.family,
143 nhashmask);
144 XFRM_STATE_INSERT(byspi, &x->byspi, nspitable + h,
145 x->xso.type);
146 }
147
148 if (x->km.seq) {
149 h = __xfrm_seq_hash(x->km.seq, nhashmask);
150 XFRM_STATE_INSERT(byseq, &x->byseq, nseqtable + h,
151 x->xso.type);
152 }
153 }
154}
155
156static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
157{
158 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
159}
160
161static void xfrm_hash_resize(struct work_struct *work)
162{
163 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
164 struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
165 unsigned long nsize, osize;
166 unsigned int nhashmask, ohashmask;
167 int i;
168
169 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
170 ndst = xfrm_hash_alloc(nsize);
171 if (!ndst)
172 return;
173 nsrc = xfrm_hash_alloc(nsize);
174 if (!nsrc) {
175 xfrm_hash_free(ndst, nsize);
176 return;
177 }
178 nspi = xfrm_hash_alloc(nsize);
179 if (!nspi) {
180 xfrm_hash_free(ndst, nsize);
181 xfrm_hash_free(nsrc, nsize);
182 return;
183 }
184 nseq = xfrm_hash_alloc(nsize);
185 if (!nseq) {
186 xfrm_hash_free(ndst, nsize);
187 xfrm_hash_free(nsrc, nsize);
188 xfrm_hash_free(nspi, nsize);
189 return;
190 }
191
192 spin_lock_bh(&net->xfrm.xfrm_state_lock);
193 write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
194
195 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
196 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
197 for (i = net->xfrm.state_hmask; i >= 0; i--)
198 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
199
200 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
201 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
202 oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
203 ohashmask = net->xfrm.state_hmask;
204
205 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
206 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
207 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
208 rcu_assign_pointer(net->xfrm.state_byseq, nseq);
209 net->xfrm.state_hmask = nhashmask;
210
211 write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
212 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
213
214 osize = (ohashmask + 1) * sizeof(struct hlist_head);
215
216 synchronize_rcu();
217
218 xfrm_hash_free(odst, osize);
219 xfrm_hash_free(osrc, osize);
220 xfrm_hash_free(ospi, osize);
221 xfrm_hash_free(oseq, osize);
222}
223
224static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
225static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
226
227static DEFINE_SPINLOCK(xfrm_state_gc_lock);
228static DEFINE_SPINLOCK(xfrm_state_dev_gc_lock);
229
230int __xfrm_state_delete(struct xfrm_state *x);
231
232int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
233static bool km_is_alive(const struct km_event *c);
234void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
235
236int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
237{
238 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
239 int err = 0;
240
241 if (!afinfo)
242 return -EAFNOSUPPORT;
243
244#define X(afi, T, name) do { \
245 WARN_ON((afi)->type_ ## name); \
246 (afi)->type_ ## name = (T); \
247 } while (0)
248
249 switch (type->proto) {
250 case IPPROTO_COMP:
251 X(afinfo, type, comp);
252 break;
253 case IPPROTO_AH:
254 X(afinfo, type, ah);
255 break;
256 case IPPROTO_ESP:
257 X(afinfo, type, esp);
258 break;
259 case IPPROTO_IPIP:
260 X(afinfo, type, ipip);
261 break;
262 case IPPROTO_DSTOPTS:
263 X(afinfo, type, dstopts);
264 break;
265 case IPPROTO_ROUTING:
266 X(afinfo, type, routing);
267 break;
268 case IPPROTO_IPV6:
269 X(afinfo, type, ipip6);
270 break;
271 default:
272 WARN_ON(1);
273 err = -EPROTONOSUPPORT;
274 break;
275 }
276#undef X
277 rcu_read_unlock();
278 return err;
279}
280EXPORT_SYMBOL(xfrm_register_type);
281
282void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
283{
284 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
285
286 if (unlikely(afinfo == NULL))
287 return;
288
289#define X(afi, T, name) do { \
290 WARN_ON((afi)->type_ ## name != (T)); \
291 (afi)->type_ ## name = NULL; \
292 } while (0)
293
294 switch (type->proto) {
295 case IPPROTO_COMP:
296 X(afinfo, type, comp);
297 break;
298 case IPPROTO_AH:
299 X(afinfo, type, ah);
300 break;
301 case IPPROTO_ESP:
302 X(afinfo, type, esp);
303 break;
304 case IPPROTO_IPIP:
305 X(afinfo, type, ipip);
306 break;
307 case IPPROTO_DSTOPTS:
308 X(afinfo, type, dstopts);
309 break;
310 case IPPROTO_ROUTING:
311 X(afinfo, type, routing);
312 break;
313 case IPPROTO_IPV6:
314 X(afinfo, type, ipip6);
315 break;
316 default:
317 WARN_ON(1);
318 break;
319 }
320#undef X
321 rcu_read_unlock();
322}
323EXPORT_SYMBOL(xfrm_unregister_type);
324
325static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
326{
327 const struct xfrm_type *type = NULL;
328 struct xfrm_state_afinfo *afinfo;
329 int modload_attempted = 0;
330
331retry:
332 afinfo = xfrm_state_get_afinfo(family);
333 if (unlikely(afinfo == NULL))
334 return NULL;
335
336 switch (proto) {
337 case IPPROTO_COMP:
338 type = afinfo->type_comp;
339 break;
340 case IPPROTO_AH:
341 type = afinfo->type_ah;
342 break;
343 case IPPROTO_ESP:
344 type = afinfo->type_esp;
345 break;
346 case IPPROTO_IPIP:
347 type = afinfo->type_ipip;
348 break;
349 case IPPROTO_DSTOPTS:
350 type = afinfo->type_dstopts;
351 break;
352 case IPPROTO_ROUTING:
353 type = afinfo->type_routing;
354 break;
355 case IPPROTO_IPV6:
356 type = afinfo->type_ipip6;
357 break;
358 default:
359 break;
360 }
361
362 if (unlikely(type && !try_module_get(type->owner)))
363 type = NULL;
364
365 rcu_read_unlock();
366
367 if (!type && !modload_attempted) {
368 request_module("xfrm-type-%d-%d", family, proto);
369 modload_attempted = 1;
370 goto retry;
371 }
372
373 return type;
374}
375
376static void xfrm_put_type(const struct xfrm_type *type)
377{
378 module_put(type->owner);
379}
380
381int xfrm_register_type_offload(const struct xfrm_type_offload *type,
382 unsigned short family)
383{
384 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
385 int err = 0;
386
387 if (unlikely(afinfo == NULL))
388 return -EAFNOSUPPORT;
389
390 switch (type->proto) {
391 case IPPROTO_ESP:
392 WARN_ON(afinfo->type_offload_esp);
393 afinfo->type_offload_esp = type;
394 break;
395 default:
396 WARN_ON(1);
397 err = -EPROTONOSUPPORT;
398 break;
399 }
400
401 rcu_read_unlock();
402 return err;
403}
404EXPORT_SYMBOL(xfrm_register_type_offload);
405
406void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
407 unsigned short family)
408{
409 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
410
411 if (unlikely(afinfo == NULL))
412 return;
413
414 switch (type->proto) {
415 case IPPROTO_ESP:
416 WARN_ON(afinfo->type_offload_esp != type);
417 afinfo->type_offload_esp = NULL;
418 break;
419 default:
420 WARN_ON(1);
421 break;
422 }
423 rcu_read_unlock();
424}
425EXPORT_SYMBOL(xfrm_unregister_type_offload);
426
427static const struct xfrm_type_offload *
428xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
429{
430 const struct xfrm_type_offload *type = NULL;
431 struct xfrm_state_afinfo *afinfo;
432
433retry:
434 afinfo = xfrm_state_get_afinfo(family);
435 if (unlikely(afinfo == NULL))
436 return NULL;
437
438 switch (proto) {
439 case IPPROTO_ESP:
440 type = afinfo->type_offload_esp;
441 break;
442 default:
443 break;
444 }
445
446 if ((type && !try_module_get(type->owner)))
447 type = NULL;
448
449 rcu_read_unlock();
450
451 if (!type && try_load) {
452 request_module("xfrm-offload-%d-%d", family, proto);
453 try_load = false;
454 goto retry;
455 }
456
457 return type;
458}
459
460static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
461{
462 module_put(type->owner);
463}
464
465static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
466 [XFRM_MODE_BEET] = {
467 .encap = XFRM_MODE_BEET,
468 .flags = XFRM_MODE_FLAG_TUNNEL,
469 .family = AF_INET,
470 },
471 [XFRM_MODE_TRANSPORT] = {
472 .encap = XFRM_MODE_TRANSPORT,
473 .family = AF_INET,
474 },
475 [XFRM_MODE_TUNNEL] = {
476 .encap = XFRM_MODE_TUNNEL,
477 .flags = XFRM_MODE_FLAG_TUNNEL,
478 .family = AF_INET,
479 },
480};
481
482static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
483 [XFRM_MODE_BEET] = {
484 .encap = XFRM_MODE_BEET,
485 .flags = XFRM_MODE_FLAG_TUNNEL,
486 .family = AF_INET6,
487 },
488 [XFRM_MODE_ROUTEOPTIMIZATION] = {
489 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
490 .family = AF_INET6,
491 },
492 [XFRM_MODE_TRANSPORT] = {
493 .encap = XFRM_MODE_TRANSPORT,
494 .family = AF_INET6,
495 },
496 [XFRM_MODE_TUNNEL] = {
497 .encap = XFRM_MODE_TUNNEL,
498 .flags = XFRM_MODE_FLAG_TUNNEL,
499 .family = AF_INET6,
500 },
501};
502
503static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
504{
505 const struct xfrm_mode *mode;
506
507 if (unlikely(encap >= XFRM_MODE_MAX))
508 return NULL;
509
510 switch (family) {
511 case AF_INET:
512 mode = &xfrm4_mode_map[encap];
513 if (mode->family == family)
514 return mode;
515 break;
516 case AF_INET6:
517 mode = &xfrm6_mode_map[encap];
518 if (mode->family == family)
519 return mode;
520 break;
521 default:
522 break;
523 }
524
525 return NULL;
526}
527
528void xfrm_state_free(struct xfrm_state *x)
529{
530 kmem_cache_free(xfrm_state_cache, x);
531}
532EXPORT_SYMBOL(xfrm_state_free);
533
534static void ___xfrm_state_destroy(struct xfrm_state *x)
535{
536 hrtimer_cancel(&x->mtimer);
537 del_timer_sync(&x->rtimer);
538 kfree(x->aead);
539 kfree(x->aalg);
540 kfree(x->ealg);
541 kfree(x->calg);
542 kfree(x->encap);
543 kfree(x->coaddr);
544 kfree(x->replay_esn);
545 kfree(x->preplay_esn);
546 if (x->type_offload)
547 xfrm_put_type_offload(x->type_offload);
548 if (x->type) {
549 x->type->destructor(x);
550 xfrm_put_type(x->type);
551 }
552 if (x->xfrag.page)
553 put_page(x->xfrag.page);
554 xfrm_dev_state_free(x);
555 security_xfrm_state_free(x);
556 xfrm_state_free(x);
557}
558
559static void xfrm_state_gc_task(struct work_struct *work)
560{
561 struct xfrm_state *x;
562 struct hlist_node *tmp;
563 struct hlist_head gc_list;
564
565 spin_lock_bh(&xfrm_state_gc_lock);
566 hlist_move_list(&xfrm_state_gc_list, &gc_list);
567 spin_unlock_bh(&xfrm_state_gc_lock);
568
569 synchronize_rcu();
570
571 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
572 ___xfrm_state_destroy(x);
573}
574
575static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
576{
577 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
578 enum hrtimer_restart ret = HRTIMER_NORESTART;
579 time64_t now = ktime_get_real_seconds();
580 time64_t next = TIME64_MAX;
581 int warn = 0;
582 int err = 0;
583
584 spin_lock(&x->lock);
585 xfrm_dev_state_update_stats(x);
586
587 if (x->km.state == XFRM_STATE_DEAD)
588 goto out;
589 if (x->km.state == XFRM_STATE_EXPIRED)
590 goto expired;
591 if (x->lft.hard_add_expires_seconds) {
592 time64_t tmo = x->lft.hard_add_expires_seconds +
593 x->curlft.add_time - now;
594 if (tmo <= 0) {
595 if (x->xflags & XFRM_SOFT_EXPIRE) {
596 /* enter hard expire without soft expire first?!
597 * setting a new date could trigger this.
598 * workaround: fix x->curflt.add_time by below:
599 */
600 x->curlft.add_time = now - x->saved_tmo - 1;
601 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
602 } else
603 goto expired;
604 }
605 if (tmo < next)
606 next = tmo;
607 }
608 if (x->lft.hard_use_expires_seconds) {
609 time64_t tmo = x->lft.hard_use_expires_seconds +
610 (READ_ONCE(x->curlft.use_time) ? : now) - now;
611 if (tmo <= 0)
612 goto expired;
613 if (tmo < next)
614 next = tmo;
615 }
616 if (x->km.dying)
617 goto resched;
618 if (x->lft.soft_add_expires_seconds) {
619 time64_t tmo = x->lft.soft_add_expires_seconds +
620 x->curlft.add_time - now;
621 if (tmo <= 0) {
622 warn = 1;
623 x->xflags &= ~XFRM_SOFT_EXPIRE;
624 } else if (tmo < next) {
625 next = tmo;
626 x->xflags |= XFRM_SOFT_EXPIRE;
627 x->saved_tmo = tmo;
628 }
629 }
630 if (x->lft.soft_use_expires_seconds) {
631 time64_t tmo = x->lft.soft_use_expires_seconds +
632 (READ_ONCE(x->curlft.use_time) ? : now) - now;
633 if (tmo <= 0)
634 warn = 1;
635 else if (tmo < next)
636 next = tmo;
637 }
638
639 x->km.dying = warn;
640 if (warn)
641 km_state_expired(x, 0, 0);
642resched:
643 if (next != TIME64_MAX) {
644 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
645 ret = HRTIMER_RESTART;
646 }
647
648 goto out;
649
650expired:
651 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
652 x->km.state = XFRM_STATE_EXPIRED;
653
654 err = __xfrm_state_delete(x);
655 if (!err)
656 km_state_expired(x, 1, 0);
657
658 xfrm_audit_state_delete(x, err ? 0 : 1, true);
659
660out:
661 spin_unlock(&x->lock);
662 return ret;
663}
664
665static void xfrm_replay_timer_handler(struct timer_list *t);
666
667struct xfrm_state *xfrm_state_alloc(struct net *net)
668{
669 struct xfrm_state *x;
670
671 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
672
673 if (x) {
674 write_pnet(&x->xs_net, net);
675 refcount_set(&x->refcnt, 1);
676 atomic_set(&x->tunnel_users, 0);
677 INIT_LIST_HEAD(&x->km.all);
678 INIT_HLIST_NODE(&x->state_cache);
679 INIT_HLIST_NODE(&x->bydst);
680 INIT_HLIST_NODE(&x->bysrc);
681 INIT_HLIST_NODE(&x->byspi);
682 INIT_HLIST_NODE(&x->byseq);
683 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
684 x->mtimer.function = xfrm_timer_handler;
685 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
686 x->curlft.add_time = ktime_get_real_seconds();
687 x->lft.soft_byte_limit = XFRM_INF;
688 x->lft.soft_packet_limit = XFRM_INF;
689 x->lft.hard_byte_limit = XFRM_INF;
690 x->lft.hard_packet_limit = XFRM_INF;
691 x->replay_maxage = 0;
692 x->replay_maxdiff = 0;
693 x->pcpu_num = UINT_MAX;
694 spin_lock_init(&x->lock);
695 }
696 return x;
697}
698EXPORT_SYMBOL(xfrm_state_alloc);
699
700#ifdef CONFIG_XFRM_OFFLOAD
701void xfrm_dev_state_delete(struct xfrm_state *x)
702{
703 struct xfrm_dev_offload *xso = &x->xso;
704 struct net_device *dev = READ_ONCE(xso->dev);
705
706 if (dev) {
707 dev->xfrmdev_ops->xdo_dev_state_delete(x);
708 spin_lock_bh(&xfrm_state_dev_gc_lock);
709 hlist_add_head(&x->dev_gclist, &xfrm_state_dev_gc_list);
710 spin_unlock_bh(&xfrm_state_dev_gc_lock);
711 }
712}
713EXPORT_SYMBOL_GPL(xfrm_dev_state_delete);
714
715void xfrm_dev_state_free(struct xfrm_state *x)
716{
717 struct xfrm_dev_offload *xso = &x->xso;
718 struct net_device *dev = READ_ONCE(xso->dev);
719
720 if (dev && dev->xfrmdev_ops) {
721 spin_lock_bh(&xfrm_state_dev_gc_lock);
722 if (!hlist_unhashed(&x->dev_gclist))
723 hlist_del(&x->dev_gclist);
724 spin_unlock_bh(&xfrm_state_dev_gc_lock);
725
726 if (dev->xfrmdev_ops->xdo_dev_state_free)
727 dev->xfrmdev_ops->xdo_dev_state_free(x);
728 WRITE_ONCE(xso->dev, NULL);
729 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
730 netdev_put(dev, &xso->dev_tracker);
731 }
732}
733#endif
734
735void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
736{
737 WARN_ON(x->km.state != XFRM_STATE_DEAD);
738
739 if (sync) {
740 synchronize_rcu();
741 ___xfrm_state_destroy(x);
742 } else {
743 spin_lock_bh(&xfrm_state_gc_lock);
744 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
745 spin_unlock_bh(&xfrm_state_gc_lock);
746 schedule_work(&xfrm_state_gc_work);
747 }
748}
749EXPORT_SYMBOL(__xfrm_state_destroy);
750
751int __xfrm_state_delete(struct xfrm_state *x)
752{
753 struct net *net = xs_net(x);
754 int err = -ESRCH;
755
756 if (x->km.state != XFRM_STATE_DEAD) {
757 x->km.state = XFRM_STATE_DEAD;
758
759 spin_lock(&net->xfrm.xfrm_state_lock);
760 list_del(&x->km.all);
761 hlist_del_rcu(&x->bydst);
762 hlist_del_rcu(&x->bysrc);
763 if (x->km.seq)
764 hlist_del_rcu(&x->byseq);
765 if (!hlist_unhashed(&x->state_cache))
766 hlist_del_rcu(&x->state_cache);
767 if (!hlist_unhashed(&x->state_cache_input))
768 hlist_del_rcu(&x->state_cache_input);
769
770 if (x->id.spi)
771 hlist_del_rcu(&x->byspi);
772 net->xfrm.state_num--;
773 xfrm_nat_keepalive_state_updated(x);
774 spin_unlock(&net->xfrm.xfrm_state_lock);
775
776 if (x->encap_sk)
777 sock_put(rcu_dereference_raw(x->encap_sk));
778
779 xfrm_dev_state_delete(x);
780
781 /* All xfrm_state objects are created by xfrm_state_alloc.
782 * The xfrm_state_alloc call gives a reference, and that
783 * is what we are dropping here.
784 */
785 xfrm_state_put(x);
786 err = 0;
787 }
788
789 return err;
790}
791EXPORT_SYMBOL(__xfrm_state_delete);
792
793int xfrm_state_delete(struct xfrm_state *x)
794{
795 int err;
796
797 spin_lock_bh(&x->lock);
798 err = __xfrm_state_delete(x);
799 spin_unlock_bh(&x->lock);
800
801 return err;
802}
803EXPORT_SYMBOL(xfrm_state_delete);
804
805#ifdef CONFIG_SECURITY_NETWORK_XFRM
806static inline int
807xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
808{
809 int i, err = 0;
810
811 for (i = 0; i <= net->xfrm.state_hmask; i++) {
812 struct xfrm_state *x;
813
814 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
815 if (xfrm_id_proto_match(x->id.proto, proto) &&
816 (err = security_xfrm_state_delete(x)) != 0) {
817 xfrm_audit_state_delete(x, 0, task_valid);
818 return err;
819 }
820 }
821 }
822
823 return err;
824}
825
826static inline int
827xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
828{
829 int i, err = 0;
830
831 for (i = 0; i <= net->xfrm.state_hmask; i++) {
832 struct xfrm_state *x;
833 struct xfrm_dev_offload *xso;
834
835 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
836 xso = &x->xso;
837
838 if (xso->dev == dev &&
839 (err = security_xfrm_state_delete(x)) != 0) {
840 xfrm_audit_state_delete(x, 0, task_valid);
841 return err;
842 }
843 }
844 }
845
846 return err;
847}
848#else
849static inline int
850xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
851{
852 return 0;
853}
854
855static inline int
856xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
857{
858 return 0;
859}
860#endif
861
862int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
863{
864 int i, err = 0, cnt = 0;
865
866 spin_lock_bh(&net->xfrm.xfrm_state_lock);
867 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
868 if (err)
869 goto out;
870
871 err = -ESRCH;
872 for (i = 0; i <= net->xfrm.state_hmask; i++) {
873 struct xfrm_state *x;
874restart:
875 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
876 if (!xfrm_state_kern(x) &&
877 xfrm_id_proto_match(x->id.proto, proto)) {
878 xfrm_state_hold(x);
879 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
880
881 err = xfrm_state_delete(x);
882 xfrm_audit_state_delete(x, err ? 0 : 1,
883 task_valid);
884 if (sync)
885 xfrm_state_put_sync(x);
886 else
887 xfrm_state_put(x);
888 if (!err)
889 cnt++;
890
891 spin_lock_bh(&net->xfrm.xfrm_state_lock);
892 goto restart;
893 }
894 }
895 }
896out:
897 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
898 if (cnt)
899 err = 0;
900
901 return err;
902}
903EXPORT_SYMBOL(xfrm_state_flush);
904
905int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
906{
907 struct xfrm_state *x;
908 struct hlist_node *tmp;
909 struct xfrm_dev_offload *xso;
910 int i, err = 0, cnt = 0;
911
912 spin_lock_bh(&net->xfrm.xfrm_state_lock);
913 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
914 if (err)
915 goto out;
916
917 err = -ESRCH;
918 for (i = 0; i <= net->xfrm.state_hmask; i++) {
919restart:
920 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
921 xso = &x->xso;
922
923 if (!xfrm_state_kern(x) && xso->dev == dev) {
924 xfrm_state_hold(x);
925 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
926
927 err = xfrm_state_delete(x);
928 xfrm_dev_state_free(x);
929
930 xfrm_audit_state_delete(x, err ? 0 : 1,
931 task_valid);
932 xfrm_state_put(x);
933 if (!err)
934 cnt++;
935
936 spin_lock_bh(&net->xfrm.xfrm_state_lock);
937 goto restart;
938 }
939 }
940 }
941 if (cnt)
942 err = 0;
943
944out:
945 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
946
947 spin_lock_bh(&xfrm_state_dev_gc_lock);
948restart_gc:
949 hlist_for_each_entry_safe(x, tmp, &xfrm_state_dev_gc_list, dev_gclist) {
950 xso = &x->xso;
951
952 if (xso->dev == dev) {
953 spin_unlock_bh(&xfrm_state_dev_gc_lock);
954 xfrm_dev_state_free(x);
955 spin_lock_bh(&xfrm_state_dev_gc_lock);
956 goto restart_gc;
957 }
958
959 }
960 spin_unlock_bh(&xfrm_state_dev_gc_lock);
961
962 xfrm_flush_gc();
963
964 return err;
965}
966EXPORT_SYMBOL(xfrm_dev_state_flush);
967
968void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
969{
970 spin_lock_bh(&net->xfrm.xfrm_state_lock);
971 si->sadcnt = net->xfrm.state_num;
972 si->sadhcnt = net->xfrm.state_hmask + 1;
973 si->sadhmcnt = xfrm_state_hashmax;
974 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
975}
976EXPORT_SYMBOL(xfrm_sad_getinfo);
977
978static void
979__xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
980{
981 const struct flowi4 *fl4 = &fl->u.ip4;
982
983 sel->daddr.a4 = fl4->daddr;
984 sel->saddr.a4 = fl4->saddr;
985 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
986 sel->dport_mask = htons(0xffff);
987 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
988 sel->sport_mask = htons(0xffff);
989 sel->family = AF_INET;
990 sel->prefixlen_d = 32;
991 sel->prefixlen_s = 32;
992 sel->proto = fl4->flowi4_proto;
993 sel->ifindex = fl4->flowi4_oif;
994}
995
996static void
997__xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
998{
999 const struct flowi6 *fl6 = &fl->u.ip6;
1000
1001 /* Initialize temporary selector matching only to current session. */
1002 *(struct in6_addr *)&sel->daddr = fl6->daddr;
1003 *(struct in6_addr *)&sel->saddr = fl6->saddr;
1004 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
1005 sel->dport_mask = htons(0xffff);
1006 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
1007 sel->sport_mask = htons(0xffff);
1008 sel->family = AF_INET6;
1009 sel->prefixlen_d = 128;
1010 sel->prefixlen_s = 128;
1011 sel->proto = fl6->flowi6_proto;
1012 sel->ifindex = fl6->flowi6_oif;
1013}
1014
1015static void
1016xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
1017 const struct xfrm_tmpl *tmpl,
1018 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1019 unsigned short family)
1020{
1021 switch (family) {
1022 case AF_INET:
1023 __xfrm4_init_tempsel(&x->sel, fl);
1024 break;
1025 case AF_INET6:
1026 __xfrm6_init_tempsel(&x->sel, fl);
1027 break;
1028 }
1029
1030 x->id = tmpl->id;
1031
1032 switch (tmpl->encap_family) {
1033 case AF_INET:
1034 if (x->id.daddr.a4 == 0)
1035 x->id.daddr.a4 = daddr->a4;
1036 x->props.saddr = tmpl->saddr;
1037 if (x->props.saddr.a4 == 0)
1038 x->props.saddr.a4 = saddr->a4;
1039 break;
1040 case AF_INET6:
1041 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
1042 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
1043 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
1044 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
1045 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
1046 break;
1047 }
1048
1049 x->props.mode = tmpl->mode;
1050 x->props.reqid = tmpl->reqid;
1051 x->props.family = tmpl->encap_family;
1052}
1053
1054struct xfrm_hash_state_ptrs {
1055 const struct hlist_head *bydst;
1056 const struct hlist_head *bysrc;
1057 const struct hlist_head *byspi;
1058 unsigned int hmask;
1059};
1060
1061static void xfrm_hash_ptrs_get(const struct net *net, struct xfrm_hash_state_ptrs *ptrs)
1062{
1063 unsigned int sequence;
1064
1065 do {
1066 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1067
1068 ptrs->bydst = xfrm_state_deref_check(net->xfrm.state_bydst, net);
1069 ptrs->bysrc = xfrm_state_deref_check(net->xfrm.state_bysrc, net);
1070 ptrs->byspi = xfrm_state_deref_check(net->xfrm.state_byspi, net);
1071 ptrs->hmask = net->xfrm.state_hmask;
1072 } while (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence));
1073}
1074
1075static struct xfrm_state *__xfrm_state_lookup_all(const struct xfrm_hash_state_ptrs *state_ptrs,
1076 u32 mark,
1077 const xfrm_address_t *daddr,
1078 __be32 spi, u8 proto,
1079 unsigned short family,
1080 struct xfrm_dev_offload *xdo)
1081{
1082 unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
1083 struct xfrm_state *x;
1084
1085 hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
1086#ifdef CONFIG_XFRM_OFFLOAD
1087 if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) {
1088 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1089 /* HW states are in the head of list, there is
1090 * no need to iterate further.
1091 */
1092 break;
1093
1094 /* Packet offload: both policy and SA should
1095 * have same device.
1096 */
1097 if (xdo->dev != x->xso.dev)
1098 continue;
1099 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1100 /* Skip HW policy for SW lookups */
1101 continue;
1102#endif
1103 if (x->props.family != family ||
1104 x->id.spi != spi ||
1105 x->id.proto != proto ||
1106 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1107 continue;
1108
1109 if ((mark & x->mark.m) != x->mark.v)
1110 continue;
1111 if (!xfrm_state_hold_rcu(x))
1112 continue;
1113 return x;
1114 }
1115
1116 return NULL;
1117}
1118
1119static struct xfrm_state *__xfrm_state_lookup(const struct xfrm_hash_state_ptrs *state_ptrs,
1120 u32 mark,
1121 const xfrm_address_t *daddr,
1122 __be32 spi, u8 proto,
1123 unsigned short family)
1124{
1125 unsigned int h = __xfrm_spi_hash(daddr, spi, proto, family, state_ptrs->hmask);
1126 struct xfrm_state *x;
1127
1128 hlist_for_each_entry_rcu(x, state_ptrs->byspi + h, byspi) {
1129 if (x->props.family != family ||
1130 x->id.spi != spi ||
1131 x->id.proto != proto ||
1132 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1133 continue;
1134
1135 if ((mark & x->mark.m) != x->mark.v)
1136 continue;
1137 if (!xfrm_state_hold_rcu(x))
1138 continue;
1139 return x;
1140 }
1141
1142 return NULL;
1143}
1144
1145struct xfrm_state *xfrm_input_state_lookup(struct net *net, u32 mark,
1146 const xfrm_address_t *daddr,
1147 __be32 spi, u8 proto,
1148 unsigned short family)
1149{
1150 struct xfrm_hash_state_ptrs state_ptrs;
1151 struct hlist_head *state_cache_input;
1152 struct xfrm_state *x = NULL;
1153
1154 state_cache_input = raw_cpu_ptr(net->xfrm.state_cache_input);
1155
1156 rcu_read_lock();
1157 hlist_for_each_entry_rcu(x, state_cache_input, state_cache_input) {
1158 if (x->props.family != family ||
1159 x->id.spi != spi ||
1160 x->id.proto != proto ||
1161 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1162 continue;
1163
1164 if ((mark & x->mark.m) != x->mark.v)
1165 continue;
1166 if (!xfrm_state_hold_rcu(x))
1167 continue;
1168 goto out;
1169 }
1170
1171 xfrm_hash_ptrs_get(net, &state_ptrs);
1172
1173 x = __xfrm_state_lookup(&state_ptrs, mark, daddr, spi, proto, family);
1174
1175 if (x && x->km.state == XFRM_STATE_VALID) {
1176 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1177 if (hlist_unhashed(&x->state_cache_input)) {
1178 hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
1179 } else {
1180 hlist_del_rcu(&x->state_cache_input);
1181 hlist_add_head_rcu(&x->state_cache_input, state_cache_input);
1182 }
1183 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1184 }
1185
1186out:
1187 rcu_read_unlock();
1188 return x;
1189}
1190EXPORT_SYMBOL(xfrm_input_state_lookup);
1191
1192static struct xfrm_state *__xfrm_state_lookup_byaddr(const struct xfrm_hash_state_ptrs *state_ptrs,
1193 u32 mark,
1194 const xfrm_address_t *daddr,
1195 const xfrm_address_t *saddr,
1196 u8 proto, unsigned short family)
1197{
1198 unsigned int h = __xfrm_src_hash(daddr, saddr, family, state_ptrs->hmask);
1199 struct xfrm_state *x;
1200
1201 hlist_for_each_entry_rcu(x, state_ptrs->bysrc + h, bysrc) {
1202 if (x->props.family != family ||
1203 x->id.proto != proto ||
1204 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1205 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1206 continue;
1207
1208 if ((mark & x->mark.m) != x->mark.v)
1209 continue;
1210 if (!xfrm_state_hold_rcu(x))
1211 continue;
1212 return x;
1213 }
1214
1215 return NULL;
1216}
1217
1218static inline struct xfrm_state *
1219__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1220{
1221 struct xfrm_hash_state_ptrs state_ptrs;
1222 struct net *net = xs_net(x);
1223 u32 mark = x->mark.v & x->mark.m;
1224
1225 xfrm_hash_ptrs_get(net, &state_ptrs);
1226
1227 if (use_spi)
1228 return __xfrm_state_lookup(&state_ptrs, mark, &x->id.daddr,
1229 x->id.spi, x->id.proto, family);
1230 else
1231 return __xfrm_state_lookup_byaddr(&state_ptrs, mark,
1232 &x->id.daddr,
1233 &x->props.saddr,
1234 x->id.proto, family);
1235}
1236
1237static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1238{
1239 if (have_hash_collision &&
1240 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1241 net->xfrm.state_num > net->xfrm.state_hmask)
1242 schedule_work(&net->xfrm.state_hash_work);
1243}
1244
1245static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1246 const struct flowi *fl, unsigned short family,
1247 struct xfrm_state **best, int *acq_in_progress,
1248 int *error)
1249{
1250 /* We need the cpu id just as a lookup key,
1251 * we don't require it to be stable.
1252 */
1253 unsigned int pcpu_id = get_cpu();
1254 put_cpu();
1255
1256 /* Resolution logic:
1257 * 1. There is a valid state with matching selector. Done.
1258 * 2. Valid state with inappropriate selector. Skip.
1259 *
1260 * Entering area of "sysdeps".
1261 *
1262 * 3. If state is not valid, selector is temporary, it selects
1263 * only session which triggered previous resolution. Key
1264 * manager will do something to install a state with proper
1265 * selector.
1266 */
1267 if (x->km.state == XFRM_STATE_VALID) {
1268 if ((x->sel.family &&
1269 (x->sel.family != family ||
1270 !xfrm_selector_match(&x->sel, fl, family))) ||
1271 !security_xfrm_state_pol_flow_match(x, pol,
1272 &fl->u.__fl_common))
1273 return;
1274
1275 if (x->pcpu_num != UINT_MAX && x->pcpu_num != pcpu_id)
1276 return;
1277
1278 if (!*best ||
1279 ((*best)->pcpu_num == UINT_MAX && x->pcpu_num == pcpu_id) ||
1280 (*best)->km.dying > x->km.dying ||
1281 ((*best)->km.dying == x->km.dying &&
1282 (*best)->curlft.add_time < x->curlft.add_time))
1283 *best = x;
1284 } else if (x->km.state == XFRM_STATE_ACQ) {
1285 if (!*best || x->pcpu_num == pcpu_id)
1286 *acq_in_progress = 1;
1287 } else if (x->km.state == XFRM_STATE_ERROR ||
1288 x->km.state == XFRM_STATE_EXPIRED) {
1289 if ((!x->sel.family ||
1290 (x->sel.family == family &&
1291 xfrm_selector_match(&x->sel, fl, family))) &&
1292 security_xfrm_state_pol_flow_match(x, pol,
1293 &fl->u.__fl_common))
1294 *error = -ESRCH;
1295 }
1296}
1297
1298struct xfrm_state *
1299xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1300 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1301 struct xfrm_policy *pol, int *err,
1302 unsigned short family, u32 if_id)
1303{
1304 static xfrm_address_t saddr_wildcard = { };
1305 struct xfrm_hash_state_ptrs state_ptrs;
1306 struct net *net = xp_net(pol);
1307 unsigned int h, h_wildcard;
1308 struct xfrm_state *x, *x0, *to_put;
1309 int acquire_in_progress = 0;
1310 int error = 0;
1311 struct xfrm_state *best = NULL;
1312 u32 mark = pol->mark.v & pol->mark.m;
1313 unsigned short encap_family = tmpl->encap_family;
1314 unsigned int sequence;
1315 struct km_event c;
1316 unsigned int pcpu_id;
1317 bool cached = false;
1318
1319 /* We need the cpu id just as a lookup key,
1320 * we don't require it to be stable.
1321 */
1322 pcpu_id = get_cpu();
1323 put_cpu();
1324
1325 to_put = NULL;
1326
1327 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1328
1329 rcu_read_lock();
1330 hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
1331 if (x->props.family == encap_family &&
1332 x->props.reqid == tmpl->reqid &&
1333 (mark & x->mark.m) == x->mark.v &&
1334 x->if_id == if_id &&
1335 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1336 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1337 tmpl->mode == x->props.mode &&
1338 tmpl->id.proto == x->id.proto &&
1339 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1340 xfrm_state_look_at(pol, x, fl, encap_family,
1341 &best, &acquire_in_progress, &error);
1342 }
1343
1344 if (best)
1345 goto cached;
1346
1347 hlist_for_each_entry_rcu(x, &pol->state_cache_list, state_cache) {
1348 if (x->props.family == encap_family &&
1349 x->props.reqid == tmpl->reqid &&
1350 (mark & x->mark.m) == x->mark.v &&
1351 x->if_id == if_id &&
1352 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1353 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1354 tmpl->mode == x->props.mode &&
1355 tmpl->id.proto == x->id.proto &&
1356 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1357 xfrm_state_look_at(pol, x, fl, family,
1358 &best, &acquire_in_progress, &error);
1359 }
1360
1361cached:
1362 cached = true;
1363 if (best)
1364 goto found;
1365 else if (error)
1366 best = NULL;
1367 else if (acquire_in_progress) /* XXX: acquire_in_progress should not happen */
1368 WARN_ON(1);
1369
1370 xfrm_hash_ptrs_get(net, &state_ptrs);
1371
1372 h = __xfrm_dst_hash(daddr, saddr, tmpl->reqid, encap_family, state_ptrs.hmask);
1373 hlist_for_each_entry_rcu(x, state_ptrs.bydst + h, bydst) {
1374#ifdef CONFIG_XFRM_OFFLOAD
1375 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1376 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1377 /* HW states are in the head of list, there is
1378 * no need to iterate further.
1379 */
1380 break;
1381
1382 /* Packet offload: both policy and SA should
1383 * have same device.
1384 */
1385 if (pol->xdo.dev != x->xso.dev)
1386 continue;
1387 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1388 /* Skip HW policy for SW lookups */
1389 continue;
1390#endif
1391 if (x->props.family == encap_family &&
1392 x->props.reqid == tmpl->reqid &&
1393 (mark & x->mark.m) == x->mark.v &&
1394 x->if_id == if_id &&
1395 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1396 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1397 tmpl->mode == x->props.mode &&
1398 tmpl->id.proto == x->id.proto &&
1399 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1400 xfrm_state_look_at(pol, x, fl, family,
1401 &best, &acquire_in_progress, &error);
1402 }
1403 if (best || acquire_in_progress)
1404 goto found;
1405
1406 h_wildcard = __xfrm_dst_hash(daddr, &saddr_wildcard, tmpl->reqid,
1407 encap_family, state_ptrs.hmask);
1408 hlist_for_each_entry_rcu(x, state_ptrs.bydst + h_wildcard, bydst) {
1409#ifdef CONFIG_XFRM_OFFLOAD
1410 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1411 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1412 /* HW states are in the head of list, there is
1413 * no need to iterate further.
1414 */
1415 break;
1416
1417 /* Packet offload: both policy and SA should
1418 * have same device.
1419 */
1420 if (pol->xdo.dev != x->xso.dev)
1421 continue;
1422 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1423 /* Skip HW policy for SW lookups */
1424 continue;
1425#endif
1426 if (x->props.family == encap_family &&
1427 x->props.reqid == tmpl->reqid &&
1428 (mark & x->mark.m) == x->mark.v &&
1429 x->if_id == if_id &&
1430 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1431 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1432 tmpl->mode == x->props.mode &&
1433 tmpl->id.proto == x->id.proto &&
1434 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1435 xfrm_state_look_at(pol, x, fl, family,
1436 &best, &acquire_in_progress, &error);
1437 }
1438
1439found:
1440 if (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) ||
1441 (best && (best->pcpu_num == pcpu_id)))
1442 x = best;
1443
1444 if (!x && !error && !acquire_in_progress) {
1445 if (tmpl->id.spi &&
1446 (x0 = __xfrm_state_lookup_all(&state_ptrs, mark, daddr,
1447 tmpl->id.spi, tmpl->id.proto,
1448 encap_family,
1449 &pol->xdo)) != NULL) {
1450 to_put = x0;
1451 error = -EEXIST;
1452 goto out;
1453 }
1454
1455 c.net = net;
1456 /* If the KMs have no listeners (yet...), avoid allocating an SA
1457 * for each and every packet - garbage collection might not
1458 * handle the flood.
1459 */
1460 if (!km_is_alive(&c)) {
1461 error = -ESRCH;
1462 goto out;
1463 }
1464
1465 x = xfrm_state_alloc(net);
1466 if (x == NULL) {
1467 error = -ENOMEM;
1468 goto out;
1469 }
1470 /* Initialize temporary state matching only
1471 * to current session. */
1472 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1473 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1474 x->if_id = if_id;
1475 if ((pol->flags & XFRM_POLICY_CPU_ACQUIRE) && best)
1476 x->pcpu_num = pcpu_id;
1477
1478 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1479 if (error) {
1480 x->km.state = XFRM_STATE_DEAD;
1481 to_put = x;
1482 x = NULL;
1483 goto out;
1484 }
1485#ifdef CONFIG_XFRM_OFFLOAD
1486 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1487 struct xfrm_dev_offload *xdo = &pol->xdo;
1488 struct xfrm_dev_offload *xso = &x->xso;
1489
1490 xso->type = XFRM_DEV_OFFLOAD_PACKET;
1491 xso->dir = xdo->dir;
1492 xso->dev = xdo->dev;
1493 xso->real_dev = xdo->real_dev;
1494 xso->flags = XFRM_DEV_OFFLOAD_FLAG_ACQ;
1495 netdev_hold(xso->dev, &xso->dev_tracker, GFP_ATOMIC);
1496 error = xso->dev->xfrmdev_ops->xdo_dev_state_add(x, NULL);
1497 if (error) {
1498 xso->dir = 0;
1499 netdev_put(xso->dev, &xso->dev_tracker);
1500 xso->dev = NULL;
1501 xso->real_dev = NULL;
1502 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
1503 x->km.state = XFRM_STATE_DEAD;
1504 to_put = x;
1505 x = NULL;
1506 goto out;
1507 }
1508 }
1509#endif
1510 if (km_query(x, tmpl, pol) == 0) {
1511 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1512 x->km.state = XFRM_STATE_ACQ;
1513 x->dir = XFRM_SA_DIR_OUT;
1514 list_add(&x->km.all, &net->xfrm.state_all);
1515 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1516 XFRM_STATE_INSERT(bydst, &x->bydst,
1517 net->xfrm.state_bydst + h,
1518 x->xso.type);
1519 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1520 XFRM_STATE_INSERT(bysrc, &x->bysrc,
1521 net->xfrm.state_bysrc + h,
1522 x->xso.type);
1523 INIT_HLIST_NODE(&x->state_cache);
1524 if (x->id.spi) {
1525 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1526 XFRM_STATE_INSERT(byspi, &x->byspi,
1527 net->xfrm.state_byspi + h,
1528 x->xso.type);
1529 }
1530 if (x->km.seq) {
1531 h = xfrm_seq_hash(net, x->km.seq);
1532 XFRM_STATE_INSERT(byseq, &x->byseq,
1533 net->xfrm.state_byseq + h,
1534 x->xso.type);
1535 }
1536 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1537 hrtimer_start(&x->mtimer,
1538 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1539 HRTIMER_MODE_REL_SOFT);
1540 net->xfrm.state_num++;
1541 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1542 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1543 } else {
1544#ifdef CONFIG_XFRM_OFFLOAD
1545 struct xfrm_dev_offload *xso = &x->xso;
1546
1547 if (xso->type == XFRM_DEV_OFFLOAD_PACKET) {
1548 xfrm_dev_state_delete(x);
1549 xfrm_dev_state_free(x);
1550 }
1551#endif
1552 x->km.state = XFRM_STATE_DEAD;
1553 to_put = x;
1554 x = NULL;
1555 error = -ESRCH;
1556 }
1557
1558 /* Use the already installed 'fallback' while the CPU-specific
1559 * SA acquire is handled*/
1560 if (best)
1561 x = best;
1562 }
1563out:
1564 if (x) {
1565 if (!xfrm_state_hold_rcu(x)) {
1566 *err = -EAGAIN;
1567 x = NULL;
1568 }
1569 } else {
1570 *err = acquire_in_progress ? -EAGAIN : error;
1571 }
1572
1573 if (x && x->km.state == XFRM_STATE_VALID && !cached &&
1574 (!(pol->flags & XFRM_POLICY_CPU_ACQUIRE) || x->pcpu_num == pcpu_id)) {
1575 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1576 if (hlist_unhashed(&x->state_cache))
1577 hlist_add_head_rcu(&x->state_cache, &pol->state_cache_list);
1578 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1579 }
1580
1581 rcu_read_unlock();
1582 if (to_put)
1583 xfrm_state_put(to_put);
1584
1585 if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1586 *err = -EAGAIN;
1587 if (x) {
1588 xfrm_state_put(x);
1589 x = NULL;
1590 }
1591 }
1592
1593 return x;
1594}
1595
1596struct xfrm_state *
1597xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1598 xfrm_address_t *daddr, xfrm_address_t *saddr,
1599 unsigned short family, u8 mode, u8 proto, u32 reqid)
1600{
1601 unsigned int h;
1602 struct xfrm_state *rx = NULL, *x = NULL;
1603
1604 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1605 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1606 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1607 if (x->props.family == family &&
1608 x->props.reqid == reqid &&
1609 (mark & x->mark.m) == x->mark.v &&
1610 x->if_id == if_id &&
1611 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1612 xfrm_state_addr_check(x, daddr, saddr, family) &&
1613 mode == x->props.mode &&
1614 proto == x->id.proto &&
1615 x->km.state == XFRM_STATE_VALID) {
1616 rx = x;
1617 break;
1618 }
1619 }
1620
1621 if (rx)
1622 xfrm_state_hold(rx);
1623 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1624
1625
1626 return rx;
1627}
1628EXPORT_SYMBOL(xfrm_stateonly_find);
1629
1630struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1631 unsigned short family)
1632{
1633 struct xfrm_state *x;
1634 struct xfrm_state_walk *w;
1635
1636 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1637 list_for_each_entry(w, &net->xfrm.state_all, all) {
1638 x = container_of(w, struct xfrm_state, km);
1639 if (x->props.family != family ||
1640 x->id.spi != spi)
1641 continue;
1642
1643 xfrm_state_hold(x);
1644 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1645 return x;
1646 }
1647 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1648 return NULL;
1649}
1650EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1651
1652static void __xfrm_state_insert(struct xfrm_state *x)
1653{
1654 struct net *net = xs_net(x);
1655 unsigned int h;
1656
1657 list_add(&x->km.all, &net->xfrm.state_all);
1658
1659 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1660 x->props.reqid, x->props.family);
1661 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1662 x->xso.type);
1663
1664 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1665 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1666 x->xso.type);
1667
1668 if (x->id.spi) {
1669 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1670 x->props.family);
1671
1672 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
1673 x->xso.type);
1674 }
1675
1676 if (x->km.seq) {
1677 h = xfrm_seq_hash(net, x->km.seq);
1678
1679 XFRM_STATE_INSERT(byseq, &x->byseq, net->xfrm.state_byseq + h,
1680 x->xso.type);
1681 }
1682
1683 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1684 if (x->replay_maxage)
1685 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1686
1687 net->xfrm.state_num++;
1688
1689 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1690 xfrm_nat_keepalive_state_updated(x);
1691}
1692
1693/* net->xfrm.xfrm_state_lock is held */
1694static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1695{
1696 struct net *net = xs_net(xnew);
1697 unsigned short family = xnew->props.family;
1698 u32 reqid = xnew->props.reqid;
1699 struct xfrm_state *x;
1700 unsigned int h;
1701 u32 mark = xnew->mark.v & xnew->mark.m;
1702 u32 if_id = xnew->if_id;
1703 u32 cpu_id = xnew->pcpu_num;
1704
1705 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1706 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1707 if (x->props.family == family &&
1708 x->props.reqid == reqid &&
1709 x->if_id == if_id &&
1710 x->pcpu_num == cpu_id &&
1711 (mark & x->mark.m) == x->mark.v &&
1712 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1713 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1714 x->genid++;
1715 }
1716}
1717
1718void xfrm_state_insert(struct xfrm_state *x)
1719{
1720 struct net *net = xs_net(x);
1721
1722 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1723 __xfrm_state_bump_genids(x);
1724 __xfrm_state_insert(x);
1725 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1726}
1727EXPORT_SYMBOL(xfrm_state_insert);
1728
1729/* net->xfrm.xfrm_state_lock is held */
1730static struct xfrm_state *__find_acq_core(struct net *net,
1731 const struct xfrm_mark *m,
1732 unsigned short family, u8 mode,
1733 u32 reqid, u32 if_id, u32 pcpu_num, u8 proto,
1734 const xfrm_address_t *daddr,
1735 const xfrm_address_t *saddr,
1736 int create)
1737{
1738 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1739 struct xfrm_state *x;
1740 u32 mark = m->v & m->m;
1741
1742 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1743 if (x->props.reqid != reqid ||
1744 x->props.mode != mode ||
1745 x->props.family != family ||
1746 x->km.state != XFRM_STATE_ACQ ||
1747 x->id.spi != 0 ||
1748 x->id.proto != proto ||
1749 (mark & x->mark.m) != x->mark.v ||
1750 x->pcpu_num != pcpu_num ||
1751 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1752 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1753 continue;
1754
1755 xfrm_state_hold(x);
1756 return x;
1757 }
1758
1759 if (!create)
1760 return NULL;
1761
1762 x = xfrm_state_alloc(net);
1763 if (likely(x)) {
1764 switch (family) {
1765 case AF_INET:
1766 x->sel.daddr.a4 = daddr->a4;
1767 x->sel.saddr.a4 = saddr->a4;
1768 x->sel.prefixlen_d = 32;
1769 x->sel.prefixlen_s = 32;
1770 x->props.saddr.a4 = saddr->a4;
1771 x->id.daddr.a4 = daddr->a4;
1772 break;
1773
1774 case AF_INET6:
1775 x->sel.daddr.in6 = daddr->in6;
1776 x->sel.saddr.in6 = saddr->in6;
1777 x->sel.prefixlen_d = 128;
1778 x->sel.prefixlen_s = 128;
1779 x->props.saddr.in6 = saddr->in6;
1780 x->id.daddr.in6 = daddr->in6;
1781 break;
1782 }
1783
1784 x->pcpu_num = pcpu_num;
1785 x->km.state = XFRM_STATE_ACQ;
1786 x->id.proto = proto;
1787 x->props.family = family;
1788 x->props.mode = mode;
1789 x->props.reqid = reqid;
1790 x->if_id = if_id;
1791 x->mark.v = m->v;
1792 x->mark.m = m->m;
1793 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1794 xfrm_state_hold(x);
1795 hrtimer_start(&x->mtimer,
1796 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1797 HRTIMER_MODE_REL_SOFT);
1798 list_add(&x->km.all, &net->xfrm.state_all);
1799 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1800 x->xso.type);
1801 h = xfrm_src_hash(net, daddr, saddr, family);
1802 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1803 x->xso.type);
1804
1805 net->xfrm.state_num++;
1806
1807 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1808 }
1809
1810 return x;
1811}
1812
1813static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num);
1814
1815int xfrm_state_add(struct xfrm_state *x)
1816{
1817 struct net *net = xs_net(x);
1818 struct xfrm_state *x1, *to_put;
1819 int family;
1820 int err;
1821 u32 mark = x->mark.v & x->mark.m;
1822 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1823
1824 family = x->props.family;
1825
1826 to_put = NULL;
1827
1828 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1829
1830 x1 = __xfrm_state_locate(x, use_spi, family);
1831 if (x1) {
1832 to_put = x1;
1833 x1 = NULL;
1834 err = -EEXIST;
1835 goto out;
1836 }
1837
1838 if (use_spi && x->km.seq) {
1839 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq, x->pcpu_num);
1840 if (x1 && ((x1->id.proto != x->id.proto) ||
1841 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1842 to_put = x1;
1843 x1 = NULL;
1844 }
1845 }
1846
1847 if (use_spi && !x1)
1848 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1849 x->props.reqid, x->if_id, x->pcpu_num, x->id.proto,
1850 &x->id.daddr, &x->props.saddr, 0);
1851
1852 __xfrm_state_bump_genids(x);
1853 __xfrm_state_insert(x);
1854 err = 0;
1855
1856out:
1857 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1858
1859 if (x1) {
1860 xfrm_state_delete(x1);
1861 xfrm_state_put(x1);
1862 }
1863
1864 if (to_put)
1865 xfrm_state_put(to_put);
1866
1867 return err;
1868}
1869EXPORT_SYMBOL(xfrm_state_add);
1870
1871#ifdef CONFIG_XFRM_MIGRATE
1872static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1873{
1874 struct xfrm_user_sec_ctx *uctx;
1875 int size = sizeof(*uctx) + security->ctx_len;
1876 int err;
1877
1878 uctx = kmalloc(size, GFP_KERNEL);
1879 if (!uctx)
1880 return -ENOMEM;
1881
1882 uctx->exttype = XFRMA_SEC_CTX;
1883 uctx->len = size;
1884 uctx->ctx_doi = security->ctx_doi;
1885 uctx->ctx_alg = security->ctx_alg;
1886 uctx->ctx_len = security->ctx_len;
1887 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1888 err = security_xfrm_state_alloc(x, uctx);
1889 kfree(uctx);
1890 if (err)
1891 return err;
1892
1893 return 0;
1894}
1895
1896static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1897 struct xfrm_encap_tmpl *encap)
1898{
1899 struct net *net = xs_net(orig);
1900 struct xfrm_state *x = xfrm_state_alloc(net);
1901 if (!x)
1902 goto out;
1903
1904 memcpy(&x->id, &orig->id, sizeof(x->id));
1905 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1906 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1907 x->props.mode = orig->props.mode;
1908 x->props.replay_window = orig->props.replay_window;
1909 x->props.reqid = orig->props.reqid;
1910 x->props.family = orig->props.family;
1911 x->props.saddr = orig->props.saddr;
1912
1913 if (orig->aalg) {
1914 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1915 if (!x->aalg)
1916 goto error;
1917 }
1918 x->props.aalgo = orig->props.aalgo;
1919
1920 if (orig->aead) {
1921 x->aead = xfrm_algo_aead_clone(orig->aead);
1922 x->geniv = orig->geniv;
1923 if (!x->aead)
1924 goto error;
1925 }
1926 if (orig->ealg) {
1927 x->ealg = xfrm_algo_clone(orig->ealg);
1928 if (!x->ealg)
1929 goto error;
1930 }
1931 x->props.ealgo = orig->props.ealgo;
1932
1933 if (orig->calg) {
1934 x->calg = xfrm_algo_clone(orig->calg);
1935 if (!x->calg)
1936 goto error;
1937 }
1938 x->props.calgo = orig->props.calgo;
1939
1940 if (encap || orig->encap) {
1941 if (encap)
1942 x->encap = kmemdup(encap, sizeof(*x->encap),
1943 GFP_KERNEL);
1944 else
1945 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1946 GFP_KERNEL);
1947
1948 if (!x->encap)
1949 goto error;
1950 }
1951
1952 if (orig->security)
1953 if (clone_security(x, orig->security))
1954 goto error;
1955
1956 if (orig->coaddr) {
1957 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1958 GFP_KERNEL);
1959 if (!x->coaddr)
1960 goto error;
1961 }
1962
1963 if (orig->replay_esn) {
1964 if (xfrm_replay_clone(x, orig))
1965 goto error;
1966 }
1967
1968 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1969 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1970
1971 x->props.flags = orig->props.flags;
1972 x->props.extra_flags = orig->props.extra_flags;
1973
1974 x->pcpu_num = orig->pcpu_num;
1975 x->if_id = orig->if_id;
1976 x->tfcpad = orig->tfcpad;
1977 x->replay_maxdiff = orig->replay_maxdiff;
1978 x->replay_maxage = orig->replay_maxage;
1979 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1980 x->km.state = orig->km.state;
1981 x->km.seq = orig->km.seq;
1982 x->replay = orig->replay;
1983 x->preplay = orig->preplay;
1984 x->mapping_maxage = orig->mapping_maxage;
1985 x->lastused = orig->lastused;
1986 x->new_mapping = 0;
1987 x->new_mapping_sport = 0;
1988 x->dir = orig->dir;
1989
1990 return x;
1991
1992 error:
1993 xfrm_state_put(x);
1994out:
1995 return NULL;
1996}
1997
1998struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
1999 u32 if_id)
2000{
2001 unsigned int h;
2002 struct xfrm_state *x = NULL;
2003
2004 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2005
2006 if (m->reqid) {
2007 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
2008 m->reqid, m->old_family);
2009 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
2010 if (x->props.mode != m->mode ||
2011 x->id.proto != m->proto)
2012 continue;
2013 if (m->reqid && x->props.reqid != m->reqid)
2014 continue;
2015 if (if_id != 0 && x->if_id != if_id)
2016 continue;
2017 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
2018 m->old_family) ||
2019 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
2020 m->old_family))
2021 continue;
2022 xfrm_state_hold(x);
2023 break;
2024 }
2025 } else {
2026 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
2027 m->old_family);
2028 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
2029 if (x->props.mode != m->mode ||
2030 x->id.proto != m->proto)
2031 continue;
2032 if (if_id != 0 && x->if_id != if_id)
2033 continue;
2034 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
2035 m->old_family) ||
2036 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
2037 m->old_family))
2038 continue;
2039 xfrm_state_hold(x);
2040 break;
2041 }
2042 }
2043
2044 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2045
2046 return x;
2047}
2048EXPORT_SYMBOL(xfrm_migrate_state_find);
2049
2050struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
2051 struct xfrm_migrate *m,
2052 struct xfrm_encap_tmpl *encap)
2053{
2054 struct xfrm_state *xc;
2055
2056 xc = xfrm_state_clone(x, encap);
2057 if (!xc)
2058 return NULL;
2059
2060 xc->props.family = m->new_family;
2061
2062 if (xfrm_init_state(xc) < 0)
2063 goto error;
2064
2065 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
2066 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
2067
2068 /* add state */
2069 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
2070 /* a care is needed when the destination address of the
2071 state is to be updated as it is a part of triplet */
2072 xfrm_state_insert(xc);
2073 } else {
2074 if (xfrm_state_add(xc) < 0)
2075 goto error;
2076 }
2077
2078 return xc;
2079error:
2080 xfrm_state_put(xc);
2081 return NULL;
2082}
2083EXPORT_SYMBOL(xfrm_state_migrate);
2084#endif
2085
2086int xfrm_state_update(struct xfrm_state *x)
2087{
2088 struct xfrm_state *x1, *to_put;
2089 int err;
2090 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
2091 struct net *net = xs_net(x);
2092
2093 to_put = NULL;
2094
2095 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2096 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
2097
2098 err = -ESRCH;
2099 if (!x1)
2100 goto out;
2101
2102 if (xfrm_state_kern(x1)) {
2103 to_put = x1;
2104 err = -EEXIST;
2105 goto out;
2106 }
2107
2108 if (x1->km.state == XFRM_STATE_ACQ) {
2109 if (x->dir && x1->dir != x->dir)
2110 goto out;
2111
2112 __xfrm_state_insert(x);
2113 x = NULL;
2114 } else {
2115 if (x1->dir != x->dir)
2116 goto out;
2117 }
2118 err = 0;
2119
2120out:
2121 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2122
2123 if (to_put)
2124 xfrm_state_put(to_put);
2125
2126 if (err)
2127 return err;
2128
2129 if (!x) {
2130 xfrm_state_delete(x1);
2131 xfrm_state_put(x1);
2132 return 0;
2133 }
2134
2135 err = -EINVAL;
2136 spin_lock_bh(&x1->lock);
2137 if (likely(x1->km.state == XFRM_STATE_VALID)) {
2138 if (x->encap && x1->encap &&
2139 x->encap->encap_type == x1->encap->encap_type)
2140 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
2141 else if (x->encap || x1->encap)
2142 goto fail;
2143
2144 if (x->coaddr && x1->coaddr) {
2145 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
2146 }
2147 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
2148 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
2149 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
2150 x1->km.dying = 0;
2151
2152 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
2153 HRTIMER_MODE_REL_SOFT);
2154 if (READ_ONCE(x1->curlft.use_time))
2155 xfrm_state_check_expire(x1);
2156
2157 if (x->props.smark.m || x->props.smark.v || x->if_id) {
2158 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2159
2160 if (x->props.smark.m || x->props.smark.v)
2161 x1->props.smark = x->props.smark;
2162
2163 if (x->if_id)
2164 x1->if_id = x->if_id;
2165
2166 __xfrm_state_bump_genids(x1);
2167 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2168 }
2169
2170 err = 0;
2171 x->km.state = XFRM_STATE_DEAD;
2172 __xfrm_state_put(x);
2173 }
2174
2175fail:
2176 spin_unlock_bh(&x1->lock);
2177
2178 xfrm_state_put(x1);
2179
2180 return err;
2181}
2182EXPORT_SYMBOL(xfrm_state_update);
2183
2184int xfrm_state_check_expire(struct xfrm_state *x)
2185{
2186 xfrm_dev_state_update_stats(x);
2187
2188 if (!READ_ONCE(x->curlft.use_time))
2189 WRITE_ONCE(x->curlft.use_time, ktime_get_real_seconds());
2190
2191 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
2192 x->curlft.packets >= x->lft.hard_packet_limit) {
2193 x->km.state = XFRM_STATE_EXPIRED;
2194 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
2195 return -EINVAL;
2196 }
2197
2198 if (!x->km.dying &&
2199 (x->curlft.bytes >= x->lft.soft_byte_limit ||
2200 x->curlft.packets >= x->lft.soft_packet_limit)) {
2201 x->km.dying = 1;
2202 km_state_expired(x, 0, 0);
2203 }
2204 return 0;
2205}
2206EXPORT_SYMBOL(xfrm_state_check_expire);
2207
2208void xfrm_state_update_stats(struct net *net)
2209{
2210 struct xfrm_state *x;
2211 int i;
2212
2213 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2214 for (i = 0; i <= net->xfrm.state_hmask; i++) {
2215 hlist_for_each_entry(x, net->xfrm.state_bydst + i, bydst)
2216 xfrm_dev_state_update_stats(x);
2217 }
2218 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2219}
2220
2221struct xfrm_state *
2222xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
2223 u8 proto, unsigned short family)
2224{
2225 struct xfrm_hash_state_ptrs state_ptrs;
2226 struct xfrm_state *x;
2227
2228 rcu_read_lock();
2229 xfrm_hash_ptrs_get(net, &state_ptrs);
2230
2231 x = __xfrm_state_lookup(&state_ptrs, mark, daddr, spi, proto, family);
2232 rcu_read_unlock();
2233 return x;
2234}
2235EXPORT_SYMBOL(xfrm_state_lookup);
2236
2237struct xfrm_state *
2238xfrm_state_lookup_byaddr(struct net *net, u32 mark,
2239 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
2240 u8 proto, unsigned short family)
2241{
2242 struct xfrm_hash_state_ptrs state_ptrs;
2243 struct xfrm_state *x;
2244
2245 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2246
2247 xfrm_hash_ptrs_get(net, &state_ptrs);
2248
2249 x = __xfrm_state_lookup_byaddr(&state_ptrs, mark, daddr, saddr, proto, family);
2250 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2251 return x;
2252}
2253EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
2254
2255struct xfrm_state *
2256xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
2257 u32 if_id, u32 pcpu_num, u8 proto, const xfrm_address_t *daddr,
2258 const xfrm_address_t *saddr, int create, unsigned short family)
2259{
2260 struct xfrm_state *x;
2261
2262 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2263 x = __find_acq_core(net, mark, family, mode, reqid, if_id, pcpu_num,
2264 proto, daddr, saddr, create);
2265 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2266
2267 return x;
2268}
2269EXPORT_SYMBOL(xfrm_find_acq);
2270
2271#ifdef CONFIG_XFRM_SUB_POLICY
2272#if IS_ENABLED(CONFIG_IPV6)
2273/* distribution counting sort function for xfrm_state and xfrm_tmpl */
2274static void
2275__xfrm6_sort(void **dst, void **src, int n,
2276 int (*cmp)(const void *p), int maxclass)
2277{
2278 int count[XFRM_MAX_DEPTH] = { };
2279 int class[XFRM_MAX_DEPTH];
2280 int i;
2281
2282 for (i = 0; i < n; i++) {
2283 int c = cmp(src[i]);
2284
2285 class[i] = c;
2286 count[c]++;
2287 }
2288
2289 for (i = 2; i < maxclass; i++)
2290 count[i] += count[i - 1];
2291
2292 for (i = 0; i < n; i++) {
2293 dst[count[class[i] - 1]++] = src[i];
2294 src[i] = NULL;
2295 }
2296}
2297
2298/* Rule for xfrm_state:
2299 *
2300 * rule 1: select IPsec transport except AH
2301 * rule 2: select MIPv6 RO or inbound trigger
2302 * rule 3: select IPsec transport AH
2303 * rule 4: select IPsec tunnel
2304 * rule 5: others
2305 */
2306static int __xfrm6_state_sort_cmp(const void *p)
2307{
2308 const struct xfrm_state *v = p;
2309
2310 switch (v->props.mode) {
2311 case XFRM_MODE_TRANSPORT:
2312 if (v->id.proto != IPPROTO_AH)
2313 return 1;
2314 else
2315 return 3;
2316#if IS_ENABLED(CONFIG_IPV6_MIP6)
2317 case XFRM_MODE_ROUTEOPTIMIZATION:
2318 case XFRM_MODE_IN_TRIGGER:
2319 return 2;
2320#endif
2321 case XFRM_MODE_TUNNEL:
2322 case XFRM_MODE_BEET:
2323 return 4;
2324 }
2325 return 5;
2326}
2327
2328/* Rule for xfrm_tmpl:
2329 *
2330 * rule 1: select IPsec transport
2331 * rule 2: select MIPv6 RO or inbound trigger
2332 * rule 3: select IPsec tunnel
2333 * rule 4: others
2334 */
2335static int __xfrm6_tmpl_sort_cmp(const void *p)
2336{
2337 const struct xfrm_tmpl *v = p;
2338
2339 switch (v->mode) {
2340 case XFRM_MODE_TRANSPORT:
2341 return 1;
2342#if IS_ENABLED(CONFIG_IPV6_MIP6)
2343 case XFRM_MODE_ROUTEOPTIMIZATION:
2344 case XFRM_MODE_IN_TRIGGER:
2345 return 2;
2346#endif
2347 case XFRM_MODE_TUNNEL:
2348 case XFRM_MODE_BEET:
2349 return 3;
2350 }
2351 return 4;
2352}
2353#else
2354static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
2355static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
2356
2357static inline void
2358__xfrm6_sort(void **dst, void **src, int n,
2359 int (*cmp)(const void *p), int maxclass)
2360{
2361 int i;
2362
2363 for (i = 0; i < n; i++)
2364 dst[i] = src[i];
2365}
2366#endif /* CONFIG_IPV6 */
2367
2368void
2369xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
2370 unsigned short family)
2371{
2372 int i;
2373
2374 if (family == AF_INET6)
2375 __xfrm6_sort((void **)dst, (void **)src, n,
2376 __xfrm6_tmpl_sort_cmp, 5);
2377 else
2378 for (i = 0; i < n; i++)
2379 dst[i] = src[i];
2380}
2381
2382void
2383xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
2384 unsigned short family)
2385{
2386 int i;
2387
2388 if (family == AF_INET6)
2389 __xfrm6_sort((void **)dst, (void **)src, n,
2390 __xfrm6_state_sort_cmp, 6);
2391 else
2392 for (i = 0; i < n; i++)
2393 dst[i] = src[i];
2394}
2395#endif
2396
2397/* Silly enough, but I'm lazy to build resolution list */
2398
2399static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
2400{
2401 unsigned int h = xfrm_seq_hash(net, seq);
2402 struct xfrm_state *x;
2403
2404 hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
2405 if (x->km.seq == seq &&
2406 (mark & x->mark.m) == x->mark.v &&
2407 x->pcpu_num == pcpu_num &&
2408 x->km.state == XFRM_STATE_ACQ) {
2409 xfrm_state_hold(x);
2410 return x;
2411 }
2412 }
2413
2414 return NULL;
2415}
2416
2417struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq, u32 pcpu_num)
2418{
2419 struct xfrm_state *x;
2420
2421 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2422 x = __xfrm_find_acq_byseq(net, mark, seq, pcpu_num);
2423 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2424 return x;
2425}
2426EXPORT_SYMBOL(xfrm_find_acq_byseq);
2427
2428u32 xfrm_get_acqseq(void)
2429{
2430 u32 res;
2431 static atomic_t acqseq;
2432
2433 do {
2434 res = atomic_inc_return(&acqseq);
2435 } while (!res);
2436
2437 return res;
2438}
2439EXPORT_SYMBOL(xfrm_get_acqseq);
2440
2441int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack)
2442{
2443 switch (proto) {
2444 case IPPROTO_AH:
2445 case IPPROTO_ESP:
2446 break;
2447
2448 case IPPROTO_COMP:
2449 /* IPCOMP spi is 16-bits. */
2450 if (max >= 0x10000) {
2451 NL_SET_ERR_MSG(extack, "IPCOMP SPI must be <= 65535");
2452 return -EINVAL;
2453 }
2454 break;
2455
2456 default:
2457 NL_SET_ERR_MSG(extack, "Invalid protocol, must be one of AH, ESP, IPCOMP");
2458 return -EINVAL;
2459 }
2460
2461 if (min > max) {
2462 NL_SET_ERR_MSG(extack, "Invalid SPI range: min > max");
2463 return -EINVAL;
2464 }
2465
2466 return 0;
2467}
2468EXPORT_SYMBOL(verify_spi_info);
2469
2470int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high,
2471 struct netlink_ext_ack *extack)
2472{
2473 struct net *net = xs_net(x);
2474 unsigned int h;
2475 struct xfrm_state *x0;
2476 int err = -ENOENT;
2477 __be32 minspi = htonl(low);
2478 __be32 maxspi = htonl(high);
2479 __be32 newspi = 0;
2480 u32 mark = x->mark.v & x->mark.m;
2481
2482 spin_lock_bh(&x->lock);
2483 if (x->km.state == XFRM_STATE_DEAD) {
2484 NL_SET_ERR_MSG(extack, "Target ACQUIRE is in DEAD state");
2485 goto unlock;
2486 }
2487
2488 err = 0;
2489 if (x->id.spi)
2490 goto unlock;
2491
2492 err = -ENOENT;
2493
2494 if (minspi == maxspi) {
2495 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2496 if (x0) {
2497 NL_SET_ERR_MSG(extack, "Requested SPI is already in use");
2498 xfrm_state_put(x0);
2499 goto unlock;
2500 }
2501 newspi = minspi;
2502 } else {
2503 u32 spi = 0;
2504 for (h = 0; h < high-low+1; h++) {
2505 spi = get_random_u32_inclusive(low, high);
2506 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2507 if (x0 == NULL) {
2508 newspi = htonl(spi);
2509 break;
2510 }
2511 xfrm_state_put(x0);
2512 }
2513 }
2514 if (newspi) {
2515 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2516 x->id.spi = newspi;
2517 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2518 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
2519 x->xso.type);
2520 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2521
2522 err = 0;
2523 } else {
2524 NL_SET_ERR_MSG(extack, "No SPI available in the requested range");
2525 }
2526
2527unlock:
2528 spin_unlock_bh(&x->lock);
2529
2530 return err;
2531}
2532EXPORT_SYMBOL(xfrm_alloc_spi);
2533
2534static bool __xfrm_state_filter_match(struct xfrm_state *x,
2535 struct xfrm_address_filter *filter)
2536{
2537 if (filter) {
2538 if ((filter->family == AF_INET ||
2539 filter->family == AF_INET6) &&
2540 x->props.family != filter->family)
2541 return false;
2542
2543 return addr_match(&x->props.saddr, &filter->saddr,
2544 filter->splen) &&
2545 addr_match(&x->id.daddr, &filter->daddr,
2546 filter->dplen);
2547 }
2548 return true;
2549}
2550
2551int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2552 int (*func)(struct xfrm_state *, int, void*),
2553 void *data)
2554{
2555 struct xfrm_state *state;
2556 struct xfrm_state_walk *x;
2557 int err = 0;
2558
2559 if (walk->seq != 0 && list_empty(&walk->all))
2560 return 0;
2561
2562 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2563 if (list_empty(&walk->all))
2564 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2565 else
2566 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2567 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2568 if (x->state == XFRM_STATE_DEAD)
2569 continue;
2570 state = container_of(x, struct xfrm_state, km);
2571 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2572 continue;
2573 if (!__xfrm_state_filter_match(state, walk->filter))
2574 continue;
2575 err = func(state, walk->seq, data);
2576 if (err) {
2577 list_move_tail(&walk->all, &x->all);
2578 goto out;
2579 }
2580 walk->seq++;
2581 }
2582 if (walk->seq == 0) {
2583 err = -ENOENT;
2584 goto out;
2585 }
2586 list_del_init(&walk->all);
2587out:
2588 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2589 return err;
2590}
2591EXPORT_SYMBOL(xfrm_state_walk);
2592
2593void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2594 struct xfrm_address_filter *filter)
2595{
2596 INIT_LIST_HEAD(&walk->all);
2597 walk->proto = proto;
2598 walk->state = XFRM_STATE_DEAD;
2599 walk->seq = 0;
2600 walk->filter = filter;
2601}
2602EXPORT_SYMBOL(xfrm_state_walk_init);
2603
2604void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2605{
2606 kfree(walk->filter);
2607
2608 if (list_empty(&walk->all))
2609 return;
2610
2611 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2612 list_del(&walk->all);
2613 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2614}
2615EXPORT_SYMBOL(xfrm_state_walk_done);
2616
2617static void xfrm_replay_timer_handler(struct timer_list *t)
2618{
2619 struct xfrm_state *x = from_timer(x, t, rtimer);
2620
2621 spin_lock(&x->lock);
2622
2623 if (x->km.state == XFRM_STATE_VALID) {
2624 if (xfrm_aevent_is_on(xs_net(x)))
2625 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2626 else
2627 x->xflags |= XFRM_TIME_DEFER;
2628 }
2629
2630 spin_unlock(&x->lock);
2631}
2632
2633static LIST_HEAD(xfrm_km_list);
2634
2635void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2636{
2637 struct xfrm_mgr *km;
2638
2639 rcu_read_lock();
2640 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2641 if (km->notify_policy)
2642 km->notify_policy(xp, dir, c);
2643 rcu_read_unlock();
2644}
2645
2646void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2647{
2648 struct xfrm_mgr *km;
2649 rcu_read_lock();
2650 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2651 if (km->notify)
2652 km->notify(x, c);
2653 rcu_read_unlock();
2654}
2655
2656EXPORT_SYMBOL(km_policy_notify);
2657EXPORT_SYMBOL(km_state_notify);
2658
2659void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2660{
2661 struct km_event c;
2662
2663 c.data.hard = hard;
2664 c.portid = portid;
2665 c.event = XFRM_MSG_EXPIRE;
2666 km_state_notify(x, &c);
2667}
2668
2669EXPORT_SYMBOL(km_state_expired);
2670/*
2671 * We send to all registered managers regardless of failure
2672 * We are happy with one success
2673*/
2674int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2675{
2676 int err = -EINVAL, acqret;
2677 struct xfrm_mgr *km;
2678
2679 rcu_read_lock();
2680 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2681 acqret = km->acquire(x, t, pol);
2682 if (!acqret)
2683 err = acqret;
2684 }
2685 rcu_read_unlock();
2686 return err;
2687}
2688EXPORT_SYMBOL(km_query);
2689
2690static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2691{
2692 int err = -EINVAL;
2693 struct xfrm_mgr *km;
2694
2695 rcu_read_lock();
2696 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2697 if (km->new_mapping)
2698 err = km->new_mapping(x, ipaddr, sport);
2699 if (!err)
2700 break;
2701 }
2702 rcu_read_unlock();
2703 return err;
2704}
2705
2706int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2707{
2708 int ret = 0;
2709
2710 if (x->mapping_maxage) {
2711 if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage ||
2712 x->new_mapping_sport != sport) {
2713 x->new_mapping_sport = sport;
2714 x->new_mapping = jiffies / HZ;
2715 ret = __km_new_mapping(x, ipaddr, sport);
2716 }
2717 } else {
2718 ret = __km_new_mapping(x, ipaddr, sport);
2719 }
2720
2721 return ret;
2722}
2723EXPORT_SYMBOL(km_new_mapping);
2724
2725void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2726{
2727 struct km_event c;
2728
2729 c.data.hard = hard;
2730 c.portid = portid;
2731 c.event = XFRM_MSG_POLEXPIRE;
2732 km_policy_notify(pol, dir, &c);
2733}
2734EXPORT_SYMBOL(km_policy_expired);
2735
2736#ifdef CONFIG_XFRM_MIGRATE
2737int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2738 const struct xfrm_migrate *m, int num_migrate,
2739 const struct xfrm_kmaddress *k,
2740 const struct xfrm_encap_tmpl *encap)
2741{
2742 int err = -EINVAL;
2743 int ret;
2744 struct xfrm_mgr *km;
2745
2746 rcu_read_lock();
2747 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2748 if (km->migrate) {
2749 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2750 encap);
2751 if (!ret)
2752 err = ret;
2753 }
2754 }
2755 rcu_read_unlock();
2756 return err;
2757}
2758EXPORT_SYMBOL(km_migrate);
2759#endif
2760
2761int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2762{
2763 int err = -EINVAL;
2764 int ret;
2765 struct xfrm_mgr *km;
2766
2767 rcu_read_lock();
2768 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2769 if (km->report) {
2770 ret = km->report(net, proto, sel, addr);
2771 if (!ret)
2772 err = ret;
2773 }
2774 }
2775 rcu_read_unlock();
2776 return err;
2777}
2778EXPORT_SYMBOL(km_report);
2779
2780static bool km_is_alive(const struct km_event *c)
2781{
2782 struct xfrm_mgr *km;
2783 bool is_alive = false;
2784
2785 rcu_read_lock();
2786 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2787 if (km->is_alive && km->is_alive(c)) {
2788 is_alive = true;
2789 break;
2790 }
2791 }
2792 rcu_read_unlock();
2793
2794 return is_alive;
2795}
2796
2797#if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2798static DEFINE_SPINLOCK(xfrm_translator_lock);
2799static struct xfrm_translator __rcu *xfrm_translator;
2800
2801struct xfrm_translator *xfrm_get_translator(void)
2802{
2803 struct xfrm_translator *xtr;
2804
2805 rcu_read_lock();
2806 xtr = rcu_dereference(xfrm_translator);
2807 if (unlikely(!xtr))
2808 goto out;
2809 if (!try_module_get(xtr->owner))
2810 xtr = NULL;
2811out:
2812 rcu_read_unlock();
2813 return xtr;
2814}
2815EXPORT_SYMBOL_GPL(xfrm_get_translator);
2816
2817void xfrm_put_translator(struct xfrm_translator *xtr)
2818{
2819 module_put(xtr->owner);
2820}
2821EXPORT_SYMBOL_GPL(xfrm_put_translator);
2822
2823int xfrm_register_translator(struct xfrm_translator *xtr)
2824{
2825 int err = 0;
2826
2827 spin_lock_bh(&xfrm_translator_lock);
2828 if (unlikely(xfrm_translator != NULL))
2829 err = -EEXIST;
2830 else
2831 rcu_assign_pointer(xfrm_translator, xtr);
2832 spin_unlock_bh(&xfrm_translator_lock);
2833
2834 return err;
2835}
2836EXPORT_SYMBOL_GPL(xfrm_register_translator);
2837
2838int xfrm_unregister_translator(struct xfrm_translator *xtr)
2839{
2840 int err = 0;
2841
2842 spin_lock_bh(&xfrm_translator_lock);
2843 if (likely(xfrm_translator != NULL)) {
2844 if (rcu_access_pointer(xfrm_translator) != xtr)
2845 err = -EINVAL;
2846 else
2847 RCU_INIT_POINTER(xfrm_translator, NULL);
2848 }
2849 spin_unlock_bh(&xfrm_translator_lock);
2850 synchronize_rcu();
2851
2852 return err;
2853}
2854EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2855#endif
2856
2857int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2858{
2859 int err;
2860 u8 *data;
2861 struct xfrm_mgr *km;
2862 struct xfrm_policy *pol = NULL;
2863
2864 if (sockptr_is_null(optval) && !optlen) {
2865 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2866 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2867 __sk_dst_reset(sk);
2868 return 0;
2869 }
2870
2871 if (optlen <= 0 || optlen > PAGE_SIZE)
2872 return -EMSGSIZE;
2873
2874 data = memdup_sockptr(optval, optlen);
2875 if (IS_ERR(data))
2876 return PTR_ERR(data);
2877
2878 if (in_compat_syscall()) {
2879 struct xfrm_translator *xtr = xfrm_get_translator();
2880
2881 if (!xtr) {
2882 kfree(data);
2883 return -EOPNOTSUPP;
2884 }
2885
2886 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2887 xfrm_put_translator(xtr);
2888 if (err) {
2889 kfree(data);
2890 return err;
2891 }
2892 }
2893
2894 err = -EINVAL;
2895 rcu_read_lock();
2896 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2897 pol = km->compile_policy(sk, optname, data,
2898 optlen, &err);
2899 if (err >= 0)
2900 break;
2901 }
2902 rcu_read_unlock();
2903
2904 if (err >= 0) {
2905 xfrm_sk_policy_insert(sk, err, pol);
2906 xfrm_pol_put(pol);
2907 __sk_dst_reset(sk);
2908 err = 0;
2909 }
2910
2911 kfree(data);
2912 return err;
2913}
2914EXPORT_SYMBOL(xfrm_user_policy);
2915
2916static DEFINE_SPINLOCK(xfrm_km_lock);
2917
2918void xfrm_register_km(struct xfrm_mgr *km)
2919{
2920 spin_lock_bh(&xfrm_km_lock);
2921 list_add_tail_rcu(&km->list, &xfrm_km_list);
2922 spin_unlock_bh(&xfrm_km_lock);
2923}
2924EXPORT_SYMBOL(xfrm_register_km);
2925
2926void xfrm_unregister_km(struct xfrm_mgr *km)
2927{
2928 spin_lock_bh(&xfrm_km_lock);
2929 list_del_rcu(&km->list);
2930 spin_unlock_bh(&xfrm_km_lock);
2931 synchronize_rcu();
2932}
2933EXPORT_SYMBOL(xfrm_unregister_km);
2934
2935int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2936{
2937 int err = 0;
2938
2939 if (WARN_ON(afinfo->family >= NPROTO))
2940 return -EAFNOSUPPORT;
2941
2942 spin_lock_bh(&xfrm_state_afinfo_lock);
2943 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2944 err = -EEXIST;
2945 else
2946 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2947 spin_unlock_bh(&xfrm_state_afinfo_lock);
2948 return err;
2949}
2950EXPORT_SYMBOL(xfrm_state_register_afinfo);
2951
2952int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2953{
2954 int err = 0, family = afinfo->family;
2955
2956 if (WARN_ON(family >= NPROTO))
2957 return -EAFNOSUPPORT;
2958
2959 spin_lock_bh(&xfrm_state_afinfo_lock);
2960 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2961 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2962 err = -EINVAL;
2963 else
2964 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2965 }
2966 spin_unlock_bh(&xfrm_state_afinfo_lock);
2967 synchronize_rcu();
2968 return err;
2969}
2970EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2971
2972struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2973{
2974 if (unlikely(family >= NPROTO))
2975 return NULL;
2976
2977 return rcu_dereference(xfrm_state_afinfo[family]);
2978}
2979EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2980
2981struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2982{
2983 struct xfrm_state_afinfo *afinfo;
2984 if (unlikely(family >= NPROTO))
2985 return NULL;
2986 rcu_read_lock();
2987 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2988 if (unlikely(!afinfo))
2989 rcu_read_unlock();
2990 return afinfo;
2991}
2992
2993void xfrm_flush_gc(void)
2994{
2995 flush_work(&xfrm_state_gc_work);
2996}
2997EXPORT_SYMBOL(xfrm_flush_gc);
2998
2999/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
3000void xfrm_state_delete_tunnel(struct xfrm_state *x)
3001{
3002 if (x->tunnel) {
3003 struct xfrm_state *t = x->tunnel;
3004
3005 if (atomic_read(&t->tunnel_users) == 2)
3006 xfrm_state_delete(t);
3007 atomic_dec(&t->tunnel_users);
3008 xfrm_state_put_sync(t);
3009 x->tunnel = NULL;
3010 }
3011}
3012EXPORT_SYMBOL(xfrm_state_delete_tunnel);
3013
3014u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
3015{
3016 const struct xfrm_type *type = READ_ONCE(x->type);
3017 struct crypto_aead *aead;
3018 u32 blksize, net_adj = 0;
3019
3020 if (x->km.state != XFRM_STATE_VALID ||
3021 !type || type->proto != IPPROTO_ESP)
3022 return mtu - x->props.header_len;
3023
3024 aead = x->data;
3025 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
3026
3027 switch (x->props.mode) {
3028 case XFRM_MODE_TRANSPORT:
3029 case XFRM_MODE_BEET:
3030 if (x->props.family == AF_INET)
3031 net_adj = sizeof(struct iphdr);
3032 else if (x->props.family == AF_INET6)
3033 net_adj = sizeof(struct ipv6hdr);
3034 break;
3035 case XFRM_MODE_TUNNEL:
3036 break;
3037 default:
3038 WARN_ON_ONCE(1);
3039 break;
3040 }
3041
3042 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
3043 net_adj) & ~(blksize - 1)) + net_adj - 2;
3044}
3045EXPORT_SYMBOL_GPL(xfrm_state_mtu);
3046
3047int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload,
3048 struct netlink_ext_ack *extack)
3049{
3050 const struct xfrm_mode *inner_mode;
3051 const struct xfrm_mode *outer_mode;
3052 int family = x->props.family;
3053 int err;
3054
3055 if (family == AF_INET &&
3056 READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
3057 x->props.flags |= XFRM_STATE_NOPMTUDISC;
3058
3059 err = -EPROTONOSUPPORT;
3060
3061 if (x->sel.family != AF_UNSPEC) {
3062 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
3063 if (inner_mode == NULL) {
3064 NL_SET_ERR_MSG(extack, "Requested mode not found");
3065 goto error;
3066 }
3067
3068 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
3069 family != x->sel.family) {
3070 NL_SET_ERR_MSG(extack, "Only tunnel modes can accommodate a change of family");
3071 goto error;
3072 }
3073
3074 x->inner_mode = *inner_mode;
3075 } else {
3076 const struct xfrm_mode *inner_mode_iaf;
3077 int iafamily = AF_INET;
3078
3079 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
3080 if (inner_mode == NULL) {
3081 NL_SET_ERR_MSG(extack, "Requested mode not found");
3082 goto error;
3083 }
3084
3085 x->inner_mode = *inner_mode;
3086
3087 if (x->props.family == AF_INET)
3088 iafamily = AF_INET6;
3089
3090 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
3091 if (inner_mode_iaf) {
3092 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
3093 x->inner_mode_iaf = *inner_mode_iaf;
3094 }
3095 }
3096
3097 x->type = xfrm_get_type(x->id.proto, family);
3098 if (x->type == NULL) {
3099 NL_SET_ERR_MSG(extack, "Requested type not found");
3100 goto error;
3101 }
3102
3103 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
3104
3105 err = x->type->init_state(x, extack);
3106 if (err)
3107 goto error;
3108
3109 outer_mode = xfrm_get_mode(x->props.mode, family);
3110 if (!outer_mode) {
3111 NL_SET_ERR_MSG(extack, "Requested mode not found");
3112 err = -EPROTONOSUPPORT;
3113 goto error;
3114 }
3115
3116 x->outer_mode = *outer_mode;
3117 if (init_replay) {
3118 err = xfrm_init_replay(x, extack);
3119 if (err)
3120 goto error;
3121 }
3122
3123 if (x->nat_keepalive_interval) {
3124 if (x->dir != XFRM_SA_DIR_OUT) {
3125 NL_SET_ERR_MSG(extack, "NAT keepalive is only supported for outbound SAs");
3126 err = -EINVAL;
3127 goto error;
3128 }
3129
3130 if (!x->encap || x->encap->encap_type != UDP_ENCAP_ESPINUDP) {
3131 NL_SET_ERR_MSG(extack,
3132 "NAT keepalive is only supported for UDP encapsulation");
3133 err = -EINVAL;
3134 goto error;
3135 }
3136 }
3137
3138error:
3139 return err;
3140}
3141
3142EXPORT_SYMBOL(__xfrm_init_state);
3143
3144int xfrm_init_state(struct xfrm_state *x)
3145{
3146 int err;
3147
3148 err = __xfrm_init_state(x, true, false, NULL);
3149 if (!err)
3150 x->km.state = XFRM_STATE_VALID;
3151
3152 return err;
3153}
3154
3155EXPORT_SYMBOL(xfrm_init_state);
3156
3157int __net_init xfrm_state_init(struct net *net)
3158{
3159 unsigned int sz;
3160
3161 if (net_eq(net, &init_net))
3162 xfrm_state_cache = KMEM_CACHE(xfrm_state,
3163 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3164
3165 INIT_LIST_HEAD(&net->xfrm.state_all);
3166
3167 sz = sizeof(struct hlist_head) * 8;
3168
3169 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
3170 if (!net->xfrm.state_bydst)
3171 goto out_bydst;
3172 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
3173 if (!net->xfrm.state_bysrc)
3174 goto out_bysrc;
3175 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
3176 if (!net->xfrm.state_byspi)
3177 goto out_byspi;
3178 net->xfrm.state_byseq = xfrm_hash_alloc(sz);
3179 if (!net->xfrm.state_byseq)
3180 goto out_byseq;
3181
3182 net->xfrm.state_cache_input = alloc_percpu(struct hlist_head);
3183 if (!net->xfrm.state_cache_input)
3184 goto out_state_cache_input;
3185
3186 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
3187
3188 net->xfrm.state_num = 0;
3189 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
3190 spin_lock_init(&net->xfrm.xfrm_state_lock);
3191 seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
3192 &net->xfrm.xfrm_state_lock);
3193 return 0;
3194
3195out_state_cache_input:
3196 xfrm_hash_free(net->xfrm.state_byseq, sz);
3197out_byseq:
3198 xfrm_hash_free(net->xfrm.state_byspi, sz);
3199out_byspi:
3200 xfrm_hash_free(net->xfrm.state_bysrc, sz);
3201out_bysrc:
3202 xfrm_hash_free(net->xfrm.state_bydst, sz);
3203out_bydst:
3204 return -ENOMEM;
3205}
3206
3207void xfrm_state_fini(struct net *net)
3208{
3209 unsigned int sz;
3210
3211 flush_work(&net->xfrm.state_hash_work);
3212 flush_work(&xfrm_state_gc_work);
3213 xfrm_state_flush(net, 0, false, true);
3214
3215 WARN_ON(!list_empty(&net->xfrm.state_all));
3216
3217 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
3218 WARN_ON(!hlist_empty(net->xfrm.state_byseq));
3219 xfrm_hash_free(net->xfrm.state_byseq, sz);
3220 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
3221 xfrm_hash_free(net->xfrm.state_byspi, sz);
3222 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
3223 xfrm_hash_free(net->xfrm.state_bysrc, sz);
3224 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
3225 xfrm_hash_free(net->xfrm.state_bydst, sz);
3226 free_percpu(net->xfrm.state_cache_input);
3227}
3228
3229#ifdef CONFIG_AUDITSYSCALL
3230static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
3231 struct audit_buffer *audit_buf)
3232{
3233 struct xfrm_sec_ctx *ctx = x->security;
3234 u32 spi = ntohl(x->id.spi);
3235
3236 if (ctx)
3237 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
3238 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
3239
3240 switch (x->props.family) {
3241 case AF_INET:
3242 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
3243 &x->props.saddr.a4, &x->id.daddr.a4);
3244 break;
3245 case AF_INET6:
3246 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
3247 x->props.saddr.a6, x->id.daddr.a6);
3248 break;
3249 }
3250
3251 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3252}
3253
3254static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
3255 struct audit_buffer *audit_buf)
3256{
3257 const struct iphdr *iph4;
3258 const struct ipv6hdr *iph6;
3259
3260 switch (family) {
3261 case AF_INET:
3262 iph4 = ip_hdr(skb);
3263 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
3264 &iph4->saddr, &iph4->daddr);
3265 break;
3266 case AF_INET6:
3267 iph6 = ipv6_hdr(skb);
3268 audit_log_format(audit_buf,
3269 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
3270 &iph6->saddr, &iph6->daddr,
3271 iph6->flow_lbl[0] & 0x0f,
3272 iph6->flow_lbl[1],
3273 iph6->flow_lbl[2]);
3274 break;
3275 }
3276}
3277
3278void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
3279{
3280 struct audit_buffer *audit_buf;
3281
3282 audit_buf = xfrm_audit_start("SAD-add");
3283 if (audit_buf == NULL)
3284 return;
3285 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3286 xfrm_audit_helper_sainfo(x, audit_buf);
3287 audit_log_format(audit_buf, " res=%u", result);
3288 audit_log_end(audit_buf);
3289}
3290EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
3291
3292void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
3293{
3294 struct audit_buffer *audit_buf;
3295
3296 audit_buf = xfrm_audit_start("SAD-delete");
3297 if (audit_buf == NULL)
3298 return;
3299 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3300 xfrm_audit_helper_sainfo(x, audit_buf);
3301 audit_log_format(audit_buf, " res=%u", result);
3302 audit_log_end(audit_buf);
3303}
3304EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
3305
3306void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
3307 struct sk_buff *skb)
3308{
3309 struct audit_buffer *audit_buf;
3310 u32 spi;
3311
3312 audit_buf = xfrm_audit_start("SA-replay-overflow");
3313 if (audit_buf == NULL)
3314 return;
3315 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3316 /* don't record the sequence number because it's inherent in this kind
3317 * of audit message */
3318 spi = ntohl(x->id.spi);
3319 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3320 audit_log_end(audit_buf);
3321}
3322EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
3323
3324void xfrm_audit_state_replay(struct xfrm_state *x,
3325 struct sk_buff *skb, __be32 net_seq)
3326{
3327 struct audit_buffer *audit_buf;
3328 u32 spi;
3329
3330 audit_buf = xfrm_audit_start("SA-replayed-pkt");
3331 if (audit_buf == NULL)
3332 return;
3333 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3334 spi = ntohl(x->id.spi);
3335 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3336 spi, spi, ntohl(net_seq));
3337 audit_log_end(audit_buf);
3338}
3339EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
3340
3341void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
3342{
3343 struct audit_buffer *audit_buf;
3344
3345 audit_buf = xfrm_audit_start("SA-notfound");
3346 if (audit_buf == NULL)
3347 return;
3348 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3349 audit_log_end(audit_buf);
3350}
3351EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
3352
3353void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
3354 __be32 net_spi, __be32 net_seq)
3355{
3356 struct audit_buffer *audit_buf;
3357 u32 spi;
3358
3359 audit_buf = xfrm_audit_start("SA-notfound");
3360 if (audit_buf == NULL)
3361 return;
3362 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3363 spi = ntohl(net_spi);
3364 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3365 spi, spi, ntohl(net_seq));
3366 audit_log_end(audit_buf);
3367}
3368EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
3369
3370void xfrm_audit_state_icvfail(struct xfrm_state *x,
3371 struct sk_buff *skb, u8 proto)
3372{
3373 struct audit_buffer *audit_buf;
3374 __be32 net_spi;
3375 __be32 net_seq;
3376
3377 audit_buf = xfrm_audit_start("SA-icv-failure");
3378 if (audit_buf == NULL)
3379 return;
3380 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3381 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
3382 u32 spi = ntohl(net_spi);
3383 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3384 spi, spi, ntohl(net_seq));
3385 }
3386 audit_log_end(audit_buf);
3387}
3388EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
3389#endif /* CONFIG_AUDITSYSCALL */
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 ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1039 (const struct in6_addr *)daddr);
1040 ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1041 (const struct in6_addr *)saddr);
1042 x->sel.prefixlen_d = 128;
1043 x->sel.prefixlen_s = 128;
1044 ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1045 (const struct in6_addr *)saddr);
1046 ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1047 (const struct in6_addr *)daddr);
1048 break;
1049 }
1050
1051 x->km.state = XFRM_STATE_ACQ;
1052 x->id.proto = proto;
1053 x->props.family = family;
1054 x->props.mode = mode;
1055 x->props.reqid = reqid;
1056 x->mark.v = m->v;
1057 x->mark.m = m->m;
1058 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1059 xfrm_state_hold(x);
1060 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1061 list_add(&x->km.all, &net->xfrm.state_all);
1062 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1063 h = xfrm_src_hash(net, daddr, saddr, family);
1064 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1065
1066 net->xfrm.state_num++;
1067
1068 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1069 }
1070
1071 return x;
1072}
1073
1074static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1075
1076int xfrm_state_add(struct xfrm_state *x)
1077{
1078 struct net *net = xs_net(x);
1079 struct xfrm_state *x1, *to_put;
1080 int family;
1081 int err;
1082 u32 mark = x->mark.v & x->mark.m;
1083 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1084
1085 family = x->props.family;
1086
1087 to_put = NULL;
1088
1089 spin_lock_bh(&xfrm_state_lock);
1090
1091 x1 = __xfrm_state_locate(x, use_spi, family);
1092 if (x1) {
1093 to_put = x1;
1094 x1 = NULL;
1095 err = -EEXIST;
1096 goto out;
1097 }
1098
1099 if (use_spi && x->km.seq) {
1100 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1101 if (x1 && ((x1->id.proto != x->id.proto) ||
1102 xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1103 to_put = x1;
1104 x1 = NULL;
1105 }
1106 }
1107
1108 if (use_spi && !x1)
1109 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1110 x->props.reqid, x->id.proto,
1111 &x->id.daddr, &x->props.saddr, 0);
1112
1113 __xfrm_state_bump_genids(x);
1114 __xfrm_state_insert(x);
1115 err = 0;
1116
1117out:
1118 spin_unlock_bh(&xfrm_state_lock);
1119
1120 if (x1) {
1121 xfrm_state_delete(x1);
1122 xfrm_state_put(x1);
1123 }
1124
1125 if (to_put)
1126 xfrm_state_put(to_put);
1127
1128 return err;
1129}
1130EXPORT_SYMBOL(xfrm_state_add);
1131
1132#ifdef CONFIG_XFRM_MIGRATE
1133static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1134{
1135 struct net *net = xs_net(orig);
1136 int err = -ENOMEM;
1137 struct xfrm_state *x = xfrm_state_alloc(net);
1138 if (!x)
1139 goto out;
1140
1141 memcpy(&x->id, &orig->id, sizeof(x->id));
1142 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1143 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1144 x->props.mode = orig->props.mode;
1145 x->props.replay_window = orig->props.replay_window;
1146 x->props.reqid = orig->props.reqid;
1147 x->props.family = orig->props.family;
1148 x->props.saddr = orig->props.saddr;
1149
1150 if (orig->aalg) {
1151 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1152 if (!x->aalg)
1153 goto error;
1154 }
1155 x->props.aalgo = orig->props.aalgo;
1156
1157 if (orig->ealg) {
1158 x->ealg = xfrm_algo_clone(orig->ealg);
1159 if (!x->ealg)
1160 goto error;
1161 }
1162 x->props.ealgo = orig->props.ealgo;
1163
1164 if (orig->calg) {
1165 x->calg = xfrm_algo_clone(orig->calg);
1166 if (!x->calg)
1167 goto error;
1168 }
1169 x->props.calgo = orig->props.calgo;
1170
1171 if (orig->encap) {
1172 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1173 if (!x->encap)
1174 goto error;
1175 }
1176
1177 if (orig->coaddr) {
1178 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1179 GFP_KERNEL);
1180 if (!x->coaddr)
1181 goto error;
1182 }
1183
1184 if (orig->replay_esn) {
1185 err = xfrm_replay_clone(x, orig);
1186 if (err)
1187 goto error;
1188 }
1189
1190 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1191
1192 err = xfrm_init_state(x);
1193 if (err)
1194 goto error;
1195
1196 x->props.flags = orig->props.flags;
1197
1198 x->curlft.add_time = orig->curlft.add_time;
1199 x->km.state = orig->km.state;
1200 x->km.seq = orig->km.seq;
1201
1202 return x;
1203
1204 error:
1205 xfrm_state_put(x);
1206out:
1207 if (errp)
1208 *errp = err;
1209 return NULL;
1210}
1211
1212/* xfrm_state_lock is held */
1213struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1214{
1215 unsigned int h;
1216 struct xfrm_state *x;
1217 struct hlist_node *entry;
1218
1219 if (m->reqid) {
1220 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1221 m->reqid, m->old_family);
1222 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1223 if (x->props.mode != m->mode ||
1224 x->id.proto != m->proto)
1225 continue;
1226 if (m->reqid && x->props.reqid != m->reqid)
1227 continue;
1228 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1229 m->old_family) ||
1230 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1231 m->old_family))
1232 continue;
1233 xfrm_state_hold(x);
1234 return x;
1235 }
1236 } else {
1237 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1238 m->old_family);
1239 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1240 if (x->props.mode != m->mode ||
1241 x->id.proto != m->proto)
1242 continue;
1243 if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1244 m->old_family) ||
1245 xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1246 m->old_family))
1247 continue;
1248 xfrm_state_hold(x);
1249 return x;
1250 }
1251 }
1252
1253 return NULL;
1254}
1255EXPORT_SYMBOL(xfrm_migrate_state_find);
1256
1257struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1258 struct xfrm_migrate *m)
1259{
1260 struct xfrm_state *xc;
1261 int err;
1262
1263 xc = xfrm_state_clone(x, &err);
1264 if (!xc)
1265 return NULL;
1266
1267 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1268 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1269
1270 /* add state */
1271 if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1272 /* a care is needed when the destination address of the
1273 state is to be updated as it is a part of triplet */
1274 xfrm_state_insert(xc);
1275 } else {
1276 if ((err = xfrm_state_add(xc)) < 0)
1277 goto error;
1278 }
1279
1280 return xc;
1281error:
1282 xfrm_state_put(xc);
1283 return NULL;
1284}
1285EXPORT_SYMBOL(xfrm_state_migrate);
1286#endif
1287
1288int xfrm_state_update(struct xfrm_state *x)
1289{
1290 struct xfrm_state *x1, *to_put;
1291 int err;
1292 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1293
1294 to_put = NULL;
1295
1296 spin_lock_bh(&xfrm_state_lock);
1297 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1298
1299 err = -ESRCH;
1300 if (!x1)
1301 goto out;
1302
1303 if (xfrm_state_kern(x1)) {
1304 to_put = x1;
1305 err = -EEXIST;
1306 goto out;
1307 }
1308
1309 if (x1->km.state == XFRM_STATE_ACQ) {
1310 __xfrm_state_insert(x);
1311 x = NULL;
1312 }
1313 err = 0;
1314
1315out:
1316 spin_unlock_bh(&xfrm_state_lock);
1317
1318 if (to_put)
1319 xfrm_state_put(to_put);
1320
1321 if (err)
1322 return err;
1323
1324 if (!x) {
1325 xfrm_state_delete(x1);
1326 xfrm_state_put(x1);
1327 return 0;
1328 }
1329
1330 err = -EINVAL;
1331 spin_lock_bh(&x1->lock);
1332 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1333 if (x->encap && x1->encap)
1334 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1335 if (x->coaddr && x1->coaddr) {
1336 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1337 }
1338 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1339 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1340 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1341 x1->km.dying = 0;
1342
1343 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1344 if (x1->curlft.use_time)
1345 xfrm_state_check_expire(x1);
1346
1347 err = 0;
1348 x->km.state = XFRM_STATE_DEAD;
1349 __xfrm_state_put(x);
1350 }
1351 spin_unlock_bh(&x1->lock);
1352
1353 xfrm_state_put(x1);
1354
1355 return err;
1356}
1357EXPORT_SYMBOL(xfrm_state_update);
1358
1359int xfrm_state_check_expire(struct xfrm_state *x)
1360{
1361 if (!x->curlft.use_time)
1362 x->curlft.use_time = get_seconds();
1363
1364 if (x->km.state != XFRM_STATE_VALID)
1365 return -EINVAL;
1366
1367 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1368 x->curlft.packets >= x->lft.hard_packet_limit) {
1369 x->km.state = XFRM_STATE_EXPIRED;
1370 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1371 return -EINVAL;
1372 }
1373
1374 if (!x->km.dying &&
1375 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1376 x->curlft.packets >= x->lft.soft_packet_limit)) {
1377 x->km.dying = 1;
1378 km_state_expired(x, 0, 0);
1379 }
1380 return 0;
1381}
1382EXPORT_SYMBOL(xfrm_state_check_expire);
1383
1384struct xfrm_state *
1385xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1386 u8 proto, unsigned short family)
1387{
1388 struct xfrm_state *x;
1389
1390 spin_lock_bh(&xfrm_state_lock);
1391 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1392 spin_unlock_bh(&xfrm_state_lock);
1393 return x;
1394}
1395EXPORT_SYMBOL(xfrm_state_lookup);
1396
1397struct xfrm_state *
1398xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1399 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1400 u8 proto, unsigned short family)
1401{
1402 struct xfrm_state *x;
1403
1404 spin_lock_bh(&xfrm_state_lock);
1405 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1406 spin_unlock_bh(&xfrm_state_lock);
1407 return x;
1408}
1409EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1410
1411struct xfrm_state *
1412xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1413 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1414 int create, unsigned short family)
1415{
1416 struct xfrm_state *x;
1417
1418 spin_lock_bh(&xfrm_state_lock);
1419 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1420 spin_unlock_bh(&xfrm_state_lock);
1421
1422 return x;
1423}
1424EXPORT_SYMBOL(xfrm_find_acq);
1425
1426#ifdef CONFIG_XFRM_SUB_POLICY
1427int
1428xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1429 unsigned short family)
1430{
1431 int err = 0;
1432 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1433 if (!afinfo)
1434 return -EAFNOSUPPORT;
1435
1436 spin_lock_bh(&xfrm_state_lock);
1437 if (afinfo->tmpl_sort)
1438 err = afinfo->tmpl_sort(dst, src, n);
1439 spin_unlock_bh(&xfrm_state_lock);
1440 xfrm_state_put_afinfo(afinfo);
1441 return err;
1442}
1443EXPORT_SYMBOL(xfrm_tmpl_sort);
1444
1445int
1446xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1447 unsigned short family)
1448{
1449 int err = 0;
1450 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1451 if (!afinfo)
1452 return -EAFNOSUPPORT;
1453
1454 spin_lock_bh(&xfrm_state_lock);
1455 if (afinfo->state_sort)
1456 err = afinfo->state_sort(dst, src, n);
1457 spin_unlock_bh(&xfrm_state_lock);
1458 xfrm_state_put_afinfo(afinfo);
1459 return err;
1460}
1461EXPORT_SYMBOL(xfrm_state_sort);
1462#endif
1463
1464/* Silly enough, but I'm lazy to build resolution list */
1465
1466static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1467{
1468 int i;
1469
1470 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1471 struct hlist_node *entry;
1472 struct xfrm_state *x;
1473
1474 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1475 if (x->km.seq == seq &&
1476 (mark & x->mark.m) == x->mark.v &&
1477 x->km.state == XFRM_STATE_ACQ) {
1478 xfrm_state_hold(x);
1479 return x;
1480 }
1481 }
1482 }
1483 return NULL;
1484}
1485
1486struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1487{
1488 struct xfrm_state *x;
1489
1490 spin_lock_bh(&xfrm_state_lock);
1491 x = __xfrm_find_acq_byseq(net, mark, seq);
1492 spin_unlock_bh(&xfrm_state_lock);
1493 return x;
1494}
1495EXPORT_SYMBOL(xfrm_find_acq_byseq);
1496
1497u32 xfrm_get_acqseq(void)
1498{
1499 u32 res;
1500 static atomic_t acqseq;
1501
1502 do {
1503 res = atomic_inc_return(&acqseq);
1504 } while (!res);
1505
1506 return res;
1507}
1508EXPORT_SYMBOL(xfrm_get_acqseq);
1509
1510int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1511{
1512 struct net *net = xs_net(x);
1513 unsigned int h;
1514 struct xfrm_state *x0;
1515 int err = -ENOENT;
1516 __be32 minspi = htonl(low);
1517 __be32 maxspi = htonl(high);
1518 u32 mark = x->mark.v & x->mark.m;
1519
1520 spin_lock_bh(&x->lock);
1521 if (x->km.state == XFRM_STATE_DEAD)
1522 goto unlock;
1523
1524 err = 0;
1525 if (x->id.spi)
1526 goto unlock;
1527
1528 err = -ENOENT;
1529
1530 if (minspi == maxspi) {
1531 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1532 if (x0) {
1533 xfrm_state_put(x0);
1534 goto unlock;
1535 }
1536 x->id.spi = minspi;
1537 } else {
1538 u32 spi = 0;
1539 for (h=0; h<high-low+1; h++) {
1540 spi = low + net_random()%(high-low+1);
1541 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1542 if (x0 == NULL) {
1543 x->id.spi = htonl(spi);
1544 break;
1545 }
1546 xfrm_state_put(x0);
1547 }
1548 }
1549 if (x->id.spi) {
1550 spin_lock_bh(&xfrm_state_lock);
1551 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1552 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1553 spin_unlock_bh(&xfrm_state_lock);
1554
1555 err = 0;
1556 }
1557
1558unlock:
1559 spin_unlock_bh(&x->lock);
1560
1561 return err;
1562}
1563EXPORT_SYMBOL(xfrm_alloc_spi);
1564
1565int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1566 int (*func)(struct xfrm_state *, int, void*),
1567 void *data)
1568{
1569 struct xfrm_state *state;
1570 struct xfrm_state_walk *x;
1571 int err = 0;
1572
1573 if (walk->seq != 0 && list_empty(&walk->all))
1574 return 0;
1575
1576 spin_lock_bh(&xfrm_state_lock);
1577 if (list_empty(&walk->all))
1578 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1579 else
1580 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1581 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1582 if (x->state == XFRM_STATE_DEAD)
1583 continue;
1584 state = container_of(x, struct xfrm_state, km);
1585 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1586 continue;
1587 err = func(state, walk->seq, data);
1588 if (err) {
1589 list_move_tail(&walk->all, &x->all);
1590 goto out;
1591 }
1592 walk->seq++;
1593 }
1594 if (walk->seq == 0) {
1595 err = -ENOENT;
1596 goto out;
1597 }
1598 list_del_init(&walk->all);
1599out:
1600 spin_unlock_bh(&xfrm_state_lock);
1601 return err;
1602}
1603EXPORT_SYMBOL(xfrm_state_walk);
1604
1605void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1606{
1607 INIT_LIST_HEAD(&walk->all);
1608 walk->proto = proto;
1609 walk->state = XFRM_STATE_DEAD;
1610 walk->seq = 0;
1611}
1612EXPORT_SYMBOL(xfrm_state_walk_init);
1613
1614void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1615{
1616 if (list_empty(&walk->all))
1617 return;
1618
1619 spin_lock_bh(&xfrm_state_lock);
1620 list_del(&walk->all);
1621 spin_unlock_bh(&xfrm_state_lock);
1622}
1623EXPORT_SYMBOL(xfrm_state_walk_done);
1624
1625static void xfrm_replay_timer_handler(unsigned long data)
1626{
1627 struct xfrm_state *x = (struct xfrm_state*)data;
1628
1629 spin_lock(&x->lock);
1630
1631 if (x->km.state == XFRM_STATE_VALID) {
1632 if (xfrm_aevent_is_on(xs_net(x)))
1633 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1634 else
1635 x->xflags |= XFRM_TIME_DEFER;
1636 }
1637
1638 spin_unlock(&x->lock);
1639}
1640
1641static LIST_HEAD(xfrm_km_list);
1642static DEFINE_RWLOCK(xfrm_km_lock);
1643
1644void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1645{
1646 struct xfrm_mgr *km;
1647
1648 read_lock(&xfrm_km_lock);
1649 list_for_each_entry(km, &xfrm_km_list, list)
1650 if (km->notify_policy)
1651 km->notify_policy(xp, dir, c);
1652 read_unlock(&xfrm_km_lock);
1653}
1654
1655void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1656{
1657 struct xfrm_mgr *km;
1658 read_lock(&xfrm_km_lock);
1659 list_for_each_entry(km, &xfrm_km_list, list)
1660 if (km->notify)
1661 km->notify(x, c);
1662 read_unlock(&xfrm_km_lock);
1663}
1664
1665EXPORT_SYMBOL(km_policy_notify);
1666EXPORT_SYMBOL(km_state_notify);
1667
1668void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1669{
1670 struct net *net = xs_net(x);
1671 struct km_event c;
1672
1673 c.data.hard = hard;
1674 c.pid = pid;
1675 c.event = XFRM_MSG_EXPIRE;
1676 km_state_notify(x, &c);
1677
1678 if (hard)
1679 wake_up(&net->xfrm.km_waitq);
1680}
1681
1682EXPORT_SYMBOL(km_state_expired);
1683/*
1684 * We send to all registered managers regardless of failure
1685 * We are happy with one success
1686*/
1687int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1688{
1689 int err = -EINVAL, acqret;
1690 struct xfrm_mgr *km;
1691
1692 read_lock(&xfrm_km_lock);
1693 list_for_each_entry(km, &xfrm_km_list, list) {
1694 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1695 if (!acqret)
1696 err = acqret;
1697 }
1698 read_unlock(&xfrm_km_lock);
1699 return err;
1700}
1701EXPORT_SYMBOL(km_query);
1702
1703int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1704{
1705 int err = -EINVAL;
1706 struct xfrm_mgr *km;
1707
1708 read_lock(&xfrm_km_lock);
1709 list_for_each_entry(km, &xfrm_km_list, list) {
1710 if (km->new_mapping)
1711 err = km->new_mapping(x, ipaddr, sport);
1712 if (!err)
1713 break;
1714 }
1715 read_unlock(&xfrm_km_lock);
1716 return err;
1717}
1718EXPORT_SYMBOL(km_new_mapping);
1719
1720void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1721{
1722 struct net *net = xp_net(pol);
1723 struct km_event c;
1724
1725 c.data.hard = hard;
1726 c.pid = pid;
1727 c.event = XFRM_MSG_POLEXPIRE;
1728 km_policy_notify(pol, dir, &c);
1729
1730 if (hard)
1731 wake_up(&net->xfrm.km_waitq);
1732}
1733EXPORT_SYMBOL(km_policy_expired);
1734
1735#ifdef CONFIG_XFRM_MIGRATE
1736int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1737 const struct xfrm_migrate *m, int num_migrate,
1738 const struct xfrm_kmaddress *k)
1739{
1740 int err = -EINVAL;
1741 int ret;
1742 struct xfrm_mgr *km;
1743
1744 read_lock(&xfrm_km_lock);
1745 list_for_each_entry(km, &xfrm_km_list, list) {
1746 if (km->migrate) {
1747 ret = km->migrate(sel, dir, type, m, num_migrate, k);
1748 if (!ret)
1749 err = ret;
1750 }
1751 }
1752 read_unlock(&xfrm_km_lock);
1753 return err;
1754}
1755EXPORT_SYMBOL(km_migrate);
1756#endif
1757
1758int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1759{
1760 int err = -EINVAL;
1761 int ret;
1762 struct xfrm_mgr *km;
1763
1764 read_lock(&xfrm_km_lock);
1765 list_for_each_entry(km, &xfrm_km_list, list) {
1766 if (km->report) {
1767 ret = km->report(net, proto, sel, addr);
1768 if (!ret)
1769 err = ret;
1770 }
1771 }
1772 read_unlock(&xfrm_km_lock);
1773 return err;
1774}
1775EXPORT_SYMBOL(km_report);
1776
1777int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1778{
1779 int err;
1780 u8 *data;
1781 struct xfrm_mgr *km;
1782 struct xfrm_policy *pol = NULL;
1783
1784 if (optlen <= 0 || optlen > PAGE_SIZE)
1785 return -EMSGSIZE;
1786
1787 data = kmalloc(optlen, GFP_KERNEL);
1788 if (!data)
1789 return -ENOMEM;
1790
1791 err = -EFAULT;
1792 if (copy_from_user(data, optval, optlen))
1793 goto out;
1794
1795 err = -EINVAL;
1796 read_lock(&xfrm_km_lock);
1797 list_for_each_entry(km, &xfrm_km_list, list) {
1798 pol = km->compile_policy(sk, optname, data,
1799 optlen, &err);
1800 if (err >= 0)
1801 break;
1802 }
1803 read_unlock(&xfrm_km_lock);
1804
1805 if (err >= 0) {
1806 xfrm_sk_policy_insert(sk, err, pol);
1807 xfrm_pol_put(pol);
1808 err = 0;
1809 }
1810
1811out:
1812 kfree(data);
1813 return err;
1814}
1815EXPORT_SYMBOL(xfrm_user_policy);
1816
1817int xfrm_register_km(struct xfrm_mgr *km)
1818{
1819 write_lock_bh(&xfrm_km_lock);
1820 list_add_tail(&km->list, &xfrm_km_list);
1821 write_unlock_bh(&xfrm_km_lock);
1822 return 0;
1823}
1824EXPORT_SYMBOL(xfrm_register_km);
1825
1826int xfrm_unregister_km(struct xfrm_mgr *km)
1827{
1828 write_lock_bh(&xfrm_km_lock);
1829 list_del(&km->list);
1830 write_unlock_bh(&xfrm_km_lock);
1831 return 0;
1832}
1833EXPORT_SYMBOL(xfrm_unregister_km);
1834
1835int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1836{
1837 int err = 0;
1838 if (unlikely(afinfo == NULL))
1839 return -EINVAL;
1840 if (unlikely(afinfo->family >= NPROTO))
1841 return -EAFNOSUPPORT;
1842 write_lock_bh(&xfrm_state_afinfo_lock);
1843 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1844 err = -ENOBUFS;
1845 else
1846 xfrm_state_afinfo[afinfo->family] = afinfo;
1847 write_unlock_bh(&xfrm_state_afinfo_lock);
1848 return err;
1849}
1850EXPORT_SYMBOL(xfrm_state_register_afinfo);
1851
1852int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1853{
1854 int err = 0;
1855 if (unlikely(afinfo == NULL))
1856 return -EINVAL;
1857 if (unlikely(afinfo->family >= NPROTO))
1858 return -EAFNOSUPPORT;
1859 write_lock_bh(&xfrm_state_afinfo_lock);
1860 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1861 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1862 err = -EINVAL;
1863 else
1864 xfrm_state_afinfo[afinfo->family] = NULL;
1865 }
1866 write_unlock_bh(&xfrm_state_afinfo_lock);
1867 return err;
1868}
1869EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1870
1871static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1872{
1873 struct xfrm_state_afinfo *afinfo;
1874 if (unlikely(family >= NPROTO))
1875 return NULL;
1876 read_lock(&xfrm_state_afinfo_lock);
1877 afinfo = xfrm_state_afinfo[family];
1878 if (unlikely(!afinfo))
1879 read_unlock(&xfrm_state_afinfo_lock);
1880 return afinfo;
1881}
1882
1883static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1884 __releases(xfrm_state_afinfo_lock)
1885{
1886 read_unlock(&xfrm_state_afinfo_lock);
1887}
1888
1889/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1890void xfrm_state_delete_tunnel(struct xfrm_state *x)
1891{
1892 if (x->tunnel) {
1893 struct xfrm_state *t = x->tunnel;
1894
1895 if (atomic_read(&t->tunnel_users) == 2)
1896 xfrm_state_delete(t);
1897 atomic_dec(&t->tunnel_users);
1898 xfrm_state_put(t);
1899 x->tunnel = NULL;
1900 }
1901}
1902EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1903
1904int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1905{
1906 int res;
1907
1908 spin_lock_bh(&x->lock);
1909 if (x->km.state == XFRM_STATE_VALID &&
1910 x->type && x->type->get_mtu)
1911 res = x->type->get_mtu(x, mtu);
1912 else
1913 res = mtu - x->props.header_len;
1914 spin_unlock_bh(&x->lock);
1915 return res;
1916}
1917
1918int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1919{
1920 struct xfrm_state_afinfo *afinfo;
1921 struct xfrm_mode *inner_mode;
1922 int family = x->props.family;
1923 int err;
1924
1925 err = -EAFNOSUPPORT;
1926 afinfo = xfrm_state_get_afinfo(family);
1927 if (!afinfo)
1928 goto error;
1929
1930 err = 0;
1931 if (afinfo->init_flags)
1932 err = afinfo->init_flags(x);
1933
1934 xfrm_state_put_afinfo(afinfo);
1935
1936 if (err)
1937 goto error;
1938
1939 err = -EPROTONOSUPPORT;
1940
1941 if (x->sel.family != AF_UNSPEC) {
1942 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1943 if (inner_mode == NULL)
1944 goto error;
1945
1946 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1947 family != x->sel.family) {
1948 xfrm_put_mode(inner_mode);
1949 goto error;
1950 }
1951
1952 x->inner_mode = inner_mode;
1953 } else {
1954 struct xfrm_mode *inner_mode_iaf;
1955 int iafamily = AF_INET;
1956
1957 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1958 if (inner_mode == NULL)
1959 goto error;
1960
1961 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1962 xfrm_put_mode(inner_mode);
1963 goto error;
1964 }
1965 x->inner_mode = inner_mode;
1966
1967 if (x->props.family == AF_INET)
1968 iafamily = AF_INET6;
1969
1970 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1971 if (inner_mode_iaf) {
1972 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1973 x->inner_mode_iaf = inner_mode_iaf;
1974 else
1975 xfrm_put_mode(inner_mode_iaf);
1976 }
1977 }
1978
1979 x->type = xfrm_get_type(x->id.proto, family);
1980 if (x->type == NULL)
1981 goto error;
1982
1983 err = x->type->init_state(x);
1984 if (err)
1985 goto error;
1986
1987 x->outer_mode = xfrm_get_mode(x->props.mode, family);
1988 if (x->outer_mode == NULL)
1989 goto error;
1990
1991 if (init_replay) {
1992 err = xfrm_init_replay(x);
1993 if (err)
1994 goto error;
1995 }
1996
1997 x->km.state = XFRM_STATE_VALID;
1998
1999error:
2000 return err;
2001}
2002
2003EXPORT_SYMBOL(__xfrm_init_state);
2004
2005int xfrm_init_state(struct xfrm_state *x)
2006{
2007 return __xfrm_init_state(x, true);
2008}
2009
2010EXPORT_SYMBOL(xfrm_init_state);
2011
2012int __net_init xfrm_state_init(struct net *net)
2013{
2014 unsigned int sz;
2015
2016 INIT_LIST_HEAD(&net->xfrm.state_all);
2017
2018 sz = sizeof(struct hlist_head) * 8;
2019
2020 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2021 if (!net->xfrm.state_bydst)
2022 goto out_bydst;
2023 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2024 if (!net->xfrm.state_bysrc)
2025 goto out_bysrc;
2026 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2027 if (!net->xfrm.state_byspi)
2028 goto out_byspi;
2029 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2030
2031 net->xfrm.state_num = 0;
2032 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2033 INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2034 INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2035 init_waitqueue_head(&net->xfrm.km_waitq);
2036 return 0;
2037
2038out_byspi:
2039 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2040out_bysrc:
2041 xfrm_hash_free(net->xfrm.state_bydst, sz);
2042out_bydst:
2043 return -ENOMEM;
2044}
2045
2046void xfrm_state_fini(struct net *net)
2047{
2048 struct xfrm_audit audit_info;
2049 unsigned int sz;
2050
2051 flush_work(&net->xfrm.state_hash_work);
2052 audit_info.loginuid = -1;
2053 audit_info.sessionid = -1;
2054 audit_info.secid = 0;
2055 xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2056 flush_work(&net->xfrm.state_gc_work);
2057
2058 WARN_ON(!list_empty(&net->xfrm.state_all));
2059
2060 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2061 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2062 xfrm_hash_free(net->xfrm.state_byspi, sz);
2063 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2064 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2065 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2066 xfrm_hash_free(net->xfrm.state_bydst, sz);
2067}
2068
2069#ifdef CONFIG_AUDITSYSCALL
2070static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2071 struct audit_buffer *audit_buf)
2072{
2073 struct xfrm_sec_ctx *ctx = x->security;
2074 u32 spi = ntohl(x->id.spi);
2075
2076 if (ctx)
2077 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2078 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2079
2080 switch(x->props.family) {
2081 case AF_INET:
2082 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2083 &x->props.saddr.a4, &x->id.daddr.a4);
2084 break;
2085 case AF_INET6:
2086 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2087 x->props.saddr.a6, x->id.daddr.a6);
2088 break;
2089 }
2090
2091 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2092}
2093
2094static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2095 struct audit_buffer *audit_buf)
2096{
2097 const struct iphdr *iph4;
2098 const struct ipv6hdr *iph6;
2099
2100 switch (family) {
2101 case AF_INET:
2102 iph4 = ip_hdr(skb);
2103 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2104 &iph4->saddr, &iph4->daddr);
2105 break;
2106 case AF_INET6:
2107 iph6 = ipv6_hdr(skb);
2108 audit_log_format(audit_buf,
2109 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2110 &iph6->saddr,&iph6->daddr,
2111 iph6->flow_lbl[0] & 0x0f,
2112 iph6->flow_lbl[1],
2113 iph6->flow_lbl[2]);
2114 break;
2115 }
2116}
2117
2118void xfrm_audit_state_add(struct xfrm_state *x, int result,
2119 uid_t auid, u32 sessionid, u32 secid)
2120{
2121 struct audit_buffer *audit_buf;
2122
2123 audit_buf = xfrm_audit_start("SAD-add");
2124 if (audit_buf == NULL)
2125 return;
2126 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2127 xfrm_audit_helper_sainfo(x, audit_buf);
2128 audit_log_format(audit_buf, " res=%u", result);
2129 audit_log_end(audit_buf);
2130}
2131EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2132
2133void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2134 uid_t auid, u32 sessionid, u32 secid)
2135{
2136 struct audit_buffer *audit_buf;
2137
2138 audit_buf = xfrm_audit_start("SAD-delete");
2139 if (audit_buf == NULL)
2140 return;
2141 xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2142 xfrm_audit_helper_sainfo(x, audit_buf);
2143 audit_log_format(audit_buf, " res=%u", result);
2144 audit_log_end(audit_buf);
2145}
2146EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2147
2148void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2149 struct sk_buff *skb)
2150{
2151 struct audit_buffer *audit_buf;
2152 u32 spi;
2153
2154 audit_buf = xfrm_audit_start("SA-replay-overflow");
2155 if (audit_buf == NULL)
2156 return;
2157 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2158 /* don't record the sequence number because it's inherent in this kind
2159 * of audit message */
2160 spi = ntohl(x->id.spi);
2161 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2162 audit_log_end(audit_buf);
2163}
2164EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2165
2166void xfrm_audit_state_replay(struct xfrm_state *x,
2167 struct sk_buff *skb, __be32 net_seq)
2168{
2169 struct audit_buffer *audit_buf;
2170 u32 spi;
2171
2172 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2173 if (audit_buf == NULL)
2174 return;
2175 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2176 spi = ntohl(x->id.spi);
2177 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2178 spi, spi, ntohl(net_seq));
2179 audit_log_end(audit_buf);
2180}
2181EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2182
2183void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2184{
2185 struct audit_buffer *audit_buf;
2186
2187 audit_buf = xfrm_audit_start("SA-notfound");
2188 if (audit_buf == NULL)
2189 return;
2190 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2191 audit_log_end(audit_buf);
2192}
2193EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2194
2195void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2196 __be32 net_spi, __be32 net_seq)
2197{
2198 struct audit_buffer *audit_buf;
2199 u32 spi;
2200
2201 audit_buf = xfrm_audit_start("SA-notfound");
2202 if (audit_buf == NULL)
2203 return;
2204 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2205 spi = ntohl(net_spi);
2206 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2207 spi, spi, ntohl(net_seq));
2208 audit_log_end(audit_buf);
2209}
2210EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2211
2212void xfrm_audit_state_icvfail(struct xfrm_state *x,
2213 struct sk_buff *skb, u8 proto)
2214{
2215 struct audit_buffer *audit_buf;
2216 __be32 net_spi;
2217 __be32 net_seq;
2218
2219 audit_buf = xfrm_audit_start("SA-icv-failure");
2220 if (audit_buf == NULL)
2221 return;
2222 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2223 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2224 u32 spi = ntohl(net_spi);
2225 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2226 spi, spi, ntohl(net_seq));
2227 }
2228 audit_log_end(audit_buf);
2229}
2230EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2231#endif /* CONFIG_AUDITSYSCALL */