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
38static void xfrm_state_gc_task(struct work_struct *work);
39
40/* Each xfrm_state may be linked to two tables:
41
42 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
43 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
44 destination/tunnel endpoint. (output)
45 */
46
47static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
48static struct kmem_cache *xfrm_state_cache __ro_after_init;
49
50static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
51static HLIST_HEAD(xfrm_state_gc_list);
52
53static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
54{
55 return refcount_inc_not_zero(&x->refcnt);
56}
57
58static inline unsigned int xfrm_dst_hash(struct net *net,
59 const xfrm_address_t *daddr,
60 const xfrm_address_t *saddr,
61 u32 reqid,
62 unsigned short family)
63{
64 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
65}
66
67static inline unsigned int xfrm_src_hash(struct net *net,
68 const xfrm_address_t *daddr,
69 const xfrm_address_t *saddr,
70 unsigned short family)
71{
72 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
73}
74
75static inline unsigned int
76xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
77 __be32 spi, u8 proto, unsigned short family)
78{
79 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
80}
81
82static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
83{
84 return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
85}
86
87#define XFRM_STATE_INSERT(by, _n, _h, _type) \
88 { \
89 struct xfrm_state *_x = NULL; \
90 \
91 if (_type != XFRM_DEV_OFFLOAD_PACKET) { \
92 hlist_for_each_entry_rcu(_x, _h, by) { \
93 if (_x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
94 continue; \
95 break; \
96 } \
97 } \
98 \
99 if (!_x || _x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \
100 /* SAD is empty or consist from HW SAs only */ \
101 hlist_add_head_rcu(_n, _h); \
102 else \
103 hlist_add_before_rcu(_n, &_x->by); \
104 }
105
106static void xfrm_hash_transfer(struct hlist_head *list,
107 struct hlist_head *ndsttable,
108 struct hlist_head *nsrctable,
109 struct hlist_head *nspitable,
110 struct hlist_head *nseqtable,
111 unsigned int nhashmask)
112{
113 struct hlist_node *tmp;
114 struct xfrm_state *x;
115
116 hlist_for_each_entry_safe(x, tmp, list, bydst) {
117 unsigned int h;
118
119 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
120 x->props.reqid, x->props.family,
121 nhashmask);
122 XFRM_STATE_INSERT(bydst, &x->bydst, ndsttable + h, x->xso.type);
123
124 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
125 x->props.family,
126 nhashmask);
127 XFRM_STATE_INSERT(bysrc, &x->bysrc, nsrctable + h, x->xso.type);
128
129 if (x->id.spi) {
130 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
131 x->id.proto, x->props.family,
132 nhashmask);
133 XFRM_STATE_INSERT(byspi, &x->byspi, nspitable + h,
134 x->xso.type);
135 }
136
137 if (x->km.seq) {
138 h = __xfrm_seq_hash(x->km.seq, nhashmask);
139 XFRM_STATE_INSERT(byseq, &x->byseq, nseqtable + h,
140 x->xso.type);
141 }
142 }
143}
144
145static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
146{
147 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
148}
149
150static void xfrm_hash_resize(struct work_struct *work)
151{
152 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
153 struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
154 unsigned long nsize, osize;
155 unsigned int nhashmask, ohashmask;
156 int i;
157
158 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
159 ndst = xfrm_hash_alloc(nsize);
160 if (!ndst)
161 return;
162 nsrc = xfrm_hash_alloc(nsize);
163 if (!nsrc) {
164 xfrm_hash_free(ndst, nsize);
165 return;
166 }
167 nspi = xfrm_hash_alloc(nsize);
168 if (!nspi) {
169 xfrm_hash_free(ndst, nsize);
170 xfrm_hash_free(nsrc, nsize);
171 return;
172 }
173 nseq = xfrm_hash_alloc(nsize);
174 if (!nseq) {
175 xfrm_hash_free(ndst, nsize);
176 xfrm_hash_free(nsrc, nsize);
177 xfrm_hash_free(nspi, nsize);
178 return;
179 }
180
181 spin_lock_bh(&net->xfrm.xfrm_state_lock);
182 write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
183
184 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
185 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
186 for (i = net->xfrm.state_hmask; i >= 0; i--)
187 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
188
189 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
190 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
191 oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
192 ohashmask = net->xfrm.state_hmask;
193
194 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
195 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
196 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
197 rcu_assign_pointer(net->xfrm.state_byseq, nseq);
198 net->xfrm.state_hmask = nhashmask;
199
200 write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
201 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
202
203 osize = (ohashmask + 1) * sizeof(struct hlist_head);
204
205 synchronize_rcu();
206
207 xfrm_hash_free(odst, osize);
208 xfrm_hash_free(osrc, osize);
209 xfrm_hash_free(ospi, osize);
210 xfrm_hash_free(oseq, osize);
211}
212
213static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
214static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
215
216static DEFINE_SPINLOCK(xfrm_state_gc_lock);
217
218int __xfrm_state_delete(struct xfrm_state *x);
219
220int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
221static bool km_is_alive(const struct km_event *c);
222void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
223
224int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
225{
226 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
227 int err = 0;
228
229 if (!afinfo)
230 return -EAFNOSUPPORT;
231
232#define X(afi, T, name) do { \
233 WARN_ON((afi)->type_ ## name); \
234 (afi)->type_ ## name = (T); \
235 } while (0)
236
237 switch (type->proto) {
238 case IPPROTO_COMP:
239 X(afinfo, type, comp);
240 break;
241 case IPPROTO_AH:
242 X(afinfo, type, ah);
243 break;
244 case IPPROTO_ESP:
245 X(afinfo, type, esp);
246 break;
247 case IPPROTO_IPIP:
248 X(afinfo, type, ipip);
249 break;
250 case IPPROTO_DSTOPTS:
251 X(afinfo, type, dstopts);
252 break;
253 case IPPROTO_ROUTING:
254 X(afinfo, type, routing);
255 break;
256 case IPPROTO_IPV6:
257 X(afinfo, type, ipip6);
258 break;
259 default:
260 WARN_ON(1);
261 err = -EPROTONOSUPPORT;
262 break;
263 }
264#undef X
265 rcu_read_unlock();
266 return err;
267}
268EXPORT_SYMBOL(xfrm_register_type);
269
270void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
271{
272 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
273
274 if (unlikely(afinfo == NULL))
275 return;
276
277#define X(afi, T, name) do { \
278 WARN_ON((afi)->type_ ## name != (T)); \
279 (afi)->type_ ## name = NULL; \
280 } while (0)
281
282 switch (type->proto) {
283 case IPPROTO_COMP:
284 X(afinfo, type, comp);
285 break;
286 case IPPROTO_AH:
287 X(afinfo, type, ah);
288 break;
289 case IPPROTO_ESP:
290 X(afinfo, type, esp);
291 break;
292 case IPPROTO_IPIP:
293 X(afinfo, type, ipip);
294 break;
295 case IPPROTO_DSTOPTS:
296 X(afinfo, type, dstopts);
297 break;
298 case IPPROTO_ROUTING:
299 X(afinfo, type, routing);
300 break;
301 case IPPROTO_IPV6:
302 X(afinfo, type, ipip6);
303 break;
304 default:
305 WARN_ON(1);
306 break;
307 }
308#undef X
309 rcu_read_unlock();
310}
311EXPORT_SYMBOL(xfrm_unregister_type);
312
313static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
314{
315 const struct xfrm_type *type = NULL;
316 struct xfrm_state_afinfo *afinfo;
317 int modload_attempted = 0;
318
319retry:
320 afinfo = xfrm_state_get_afinfo(family);
321 if (unlikely(afinfo == NULL))
322 return NULL;
323
324 switch (proto) {
325 case IPPROTO_COMP:
326 type = afinfo->type_comp;
327 break;
328 case IPPROTO_AH:
329 type = afinfo->type_ah;
330 break;
331 case IPPROTO_ESP:
332 type = afinfo->type_esp;
333 break;
334 case IPPROTO_IPIP:
335 type = afinfo->type_ipip;
336 break;
337 case IPPROTO_DSTOPTS:
338 type = afinfo->type_dstopts;
339 break;
340 case IPPROTO_ROUTING:
341 type = afinfo->type_routing;
342 break;
343 case IPPROTO_IPV6:
344 type = afinfo->type_ipip6;
345 break;
346 default:
347 break;
348 }
349
350 if (unlikely(type && !try_module_get(type->owner)))
351 type = NULL;
352
353 rcu_read_unlock();
354
355 if (!type && !modload_attempted) {
356 request_module("xfrm-type-%d-%d", family, proto);
357 modload_attempted = 1;
358 goto retry;
359 }
360
361 return type;
362}
363
364static void xfrm_put_type(const struct xfrm_type *type)
365{
366 module_put(type->owner);
367}
368
369int xfrm_register_type_offload(const struct xfrm_type_offload *type,
370 unsigned short family)
371{
372 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
373 int err = 0;
374
375 if (unlikely(afinfo == NULL))
376 return -EAFNOSUPPORT;
377
378 switch (type->proto) {
379 case IPPROTO_ESP:
380 WARN_ON(afinfo->type_offload_esp);
381 afinfo->type_offload_esp = type;
382 break;
383 default:
384 WARN_ON(1);
385 err = -EPROTONOSUPPORT;
386 break;
387 }
388
389 rcu_read_unlock();
390 return err;
391}
392EXPORT_SYMBOL(xfrm_register_type_offload);
393
394void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
395 unsigned short family)
396{
397 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
398
399 if (unlikely(afinfo == NULL))
400 return;
401
402 switch (type->proto) {
403 case IPPROTO_ESP:
404 WARN_ON(afinfo->type_offload_esp != type);
405 afinfo->type_offload_esp = NULL;
406 break;
407 default:
408 WARN_ON(1);
409 break;
410 }
411 rcu_read_unlock();
412}
413EXPORT_SYMBOL(xfrm_unregister_type_offload);
414
415static const struct xfrm_type_offload *
416xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
417{
418 const struct xfrm_type_offload *type = NULL;
419 struct xfrm_state_afinfo *afinfo;
420
421retry:
422 afinfo = xfrm_state_get_afinfo(family);
423 if (unlikely(afinfo == NULL))
424 return NULL;
425
426 switch (proto) {
427 case IPPROTO_ESP:
428 type = afinfo->type_offload_esp;
429 break;
430 default:
431 break;
432 }
433
434 if ((type && !try_module_get(type->owner)))
435 type = NULL;
436
437 rcu_read_unlock();
438
439 if (!type && try_load) {
440 request_module("xfrm-offload-%d-%d", family, proto);
441 try_load = false;
442 goto retry;
443 }
444
445 return type;
446}
447
448static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
449{
450 module_put(type->owner);
451}
452
453static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
454 [XFRM_MODE_BEET] = {
455 .encap = XFRM_MODE_BEET,
456 .flags = XFRM_MODE_FLAG_TUNNEL,
457 .family = AF_INET,
458 },
459 [XFRM_MODE_TRANSPORT] = {
460 .encap = XFRM_MODE_TRANSPORT,
461 .family = AF_INET,
462 },
463 [XFRM_MODE_TUNNEL] = {
464 .encap = XFRM_MODE_TUNNEL,
465 .flags = XFRM_MODE_FLAG_TUNNEL,
466 .family = AF_INET,
467 },
468};
469
470static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
471 [XFRM_MODE_BEET] = {
472 .encap = XFRM_MODE_BEET,
473 .flags = XFRM_MODE_FLAG_TUNNEL,
474 .family = AF_INET6,
475 },
476 [XFRM_MODE_ROUTEOPTIMIZATION] = {
477 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
478 .family = AF_INET6,
479 },
480 [XFRM_MODE_TRANSPORT] = {
481 .encap = XFRM_MODE_TRANSPORT,
482 .family = AF_INET6,
483 },
484 [XFRM_MODE_TUNNEL] = {
485 .encap = XFRM_MODE_TUNNEL,
486 .flags = XFRM_MODE_FLAG_TUNNEL,
487 .family = AF_INET6,
488 },
489};
490
491static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
492{
493 const struct xfrm_mode *mode;
494
495 if (unlikely(encap >= XFRM_MODE_MAX))
496 return NULL;
497
498 switch (family) {
499 case AF_INET:
500 mode = &xfrm4_mode_map[encap];
501 if (mode->family == family)
502 return mode;
503 break;
504 case AF_INET6:
505 mode = &xfrm6_mode_map[encap];
506 if (mode->family == family)
507 return mode;
508 break;
509 default:
510 break;
511 }
512
513 return NULL;
514}
515
516void xfrm_state_free(struct xfrm_state *x)
517{
518 kmem_cache_free(xfrm_state_cache, x);
519}
520EXPORT_SYMBOL(xfrm_state_free);
521
522static void ___xfrm_state_destroy(struct xfrm_state *x)
523{
524 hrtimer_cancel(&x->mtimer);
525 del_timer_sync(&x->rtimer);
526 kfree(x->aead);
527 kfree(x->aalg);
528 kfree(x->ealg);
529 kfree(x->calg);
530 kfree(x->encap);
531 kfree(x->coaddr);
532 kfree(x->replay_esn);
533 kfree(x->preplay_esn);
534 if (x->type_offload)
535 xfrm_put_type_offload(x->type_offload);
536 if (x->type) {
537 x->type->destructor(x);
538 xfrm_put_type(x->type);
539 }
540 if (x->xfrag.page)
541 put_page(x->xfrag.page);
542 xfrm_dev_state_free(x);
543 security_xfrm_state_free(x);
544 xfrm_state_free(x);
545}
546
547static void xfrm_state_gc_task(struct work_struct *work)
548{
549 struct xfrm_state *x;
550 struct hlist_node *tmp;
551 struct hlist_head gc_list;
552
553 spin_lock_bh(&xfrm_state_gc_lock);
554 hlist_move_list(&xfrm_state_gc_list, &gc_list);
555 spin_unlock_bh(&xfrm_state_gc_lock);
556
557 synchronize_rcu();
558
559 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
560 ___xfrm_state_destroy(x);
561}
562
563static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
564{
565 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
566 enum hrtimer_restart ret = HRTIMER_NORESTART;
567 time64_t now = ktime_get_real_seconds();
568 time64_t next = TIME64_MAX;
569 int warn = 0;
570 int err = 0;
571
572 spin_lock(&x->lock);
573 xfrm_dev_state_update_curlft(x);
574
575 if (x->km.state == XFRM_STATE_DEAD)
576 goto out;
577 if (x->km.state == XFRM_STATE_EXPIRED)
578 goto expired;
579 if (x->lft.hard_add_expires_seconds) {
580 time64_t tmo = x->lft.hard_add_expires_seconds +
581 x->curlft.add_time - now;
582 if (tmo <= 0) {
583 if (x->xflags & XFRM_SOFT_EXPIRE) {
584 /* enter hard expire without soft expire first?!
585 * setting a new date could trigger this.
586 * workaround: fix x->curflt.add_time by below:
587 */
588 x->curlft.add_time = now - x->saved_tmo - 1;
589 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
590 } else
591 goto expired;
592 }
593 if (tmo < next)
594 next = tmo;
595 }
596 if (x->lft.hard_use_expires_seconds) {
597 time64_t tmo = x->lft.hard_use_expires_seconds +
598 (READ_ONCE(x->curlft.use_time) ? : now) - now;
599 if (tmo <= 0)
600 goto expired;
601 if (tmo < next)
602 next = tmo;
603 }
604 if (x->km.dying)
605 goto resched;
606 if (x->lft.soft_add_expires_seconds) {
607 time64_t tmo = x->lft.soft_add_expires_seconds +
608 x->curlft.add_time - now;
609 if (tmo <= 0) {
610 warn = 1;
611 x->xflags &= ~XFRM_SOFT_EXPIRE;
612 } else if (tmo < next) {
613 next = tmo;
614 x->xflags |= XFRM_SOFT_EXPIRE;
615 x->saved_tmo = tmo;
616 }
617 }
618 if (x->lft.soft_use_expires_seconds) {
619 time64_t tmo = x->lft.soft_use_expires_seconds +
620 (READ_ONCE(x->curlft.use_time) ? : now) - now;
621 if (tmo <= 0)
622 warn = 1;
623 else if (tmo < next)
624 next = tmo;
625 }
626
627 x->km.dying = warn;
628 if (warn)
629 km_state_expired(x, 0, 0);
630resched:
631 if (next != TIME64_MAX) {
632 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
633 ret = HRTIMER_RESTART;
634 }
635
636 goto out;
637
638expired:
639 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
640 x->km.state = XFRM_STATE_EXPIRED;
641
642 err = __xfrm_state_delete(x);
643 if (!err)
644 km_state_expired(x, 1, 0);
645
646 xfrm_audit_state_delete(x, err ? 0 : 1, true);
647
648out:
649 spin_unlock(&x->lock);
650 return ret;
651}
652
653static void xfrm_replay_timer_handler(struct timer_list *t);
654
655struct xfrm_state *xfrm_state_alloc(struct net *net)
656{
657 struct xfrm_state *x;
658
659 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
660
661 if (x) {
662 write_pnet(&x->xs_net, net);
663 refcount_set(&x->refcnt, 1);
664 atomic_set(&x->tunnel_users, 0);
665 INIT_LIST_HEAD(&x->km.all);
666 INIT_HLIST_NODE(&x->bydst);
667 INIT_HLIST_NODE(&x->bysrc);
668 INIT_HLIST_NODE(&x->byspi);
669 INIT_HLIST_NODE(&x->byseq);
670 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
671 x->mtimer.function = xfrm_timer_handler;
672 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
673 x->curlft.add_time = ktime_get_real_seconds();
674 x->lft.soft_byte_limit = XFRM_INF;
675 x->lft.soft_packet_limit = XFRM_INF;
676 x->lft.hard_byte_limit = XFRM_INF;
677 x->lft.hard_packet_limit = XFRM_INF;
678 x->replay_maxage = 0;
679 x->replay_maxdiff = 0;
680 spin_lock_init(&x->lock);
681 }
682 return x;
683}
684EXPORT_SYMBOL(xfrm_state_alloc);
685
686void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
687{
688 WARN_ON(x->km.state != XFRM_STATE_DEAD);
689
690 if (sync) {
691 synchronize_rcu();
692 ___xfrm_state_destroy(x);
693 } else {
694 spin_lock_bh(&xfrm_state_gc_lock);
695 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
696 spin_unlock_bh(&xfrm_state_gc_lock);
697 schedule_work(&xfrm_state_gc_work);
698 }
699}
700EXPORT_SYMBOL(__xfrm_state_destroy);
701
702int __xfrm_state_delete(struct xfrm_state *x)
703{
704 struct net *net = xs_net(x);
705 int err = -ESRCH;
706
707 if (x->km.state != XFRM_STATE_DEAD) {
708 x->km.state = XFRM_STATE_DEAD;
709 spin_lock(&net->xfrm.xfrm_state_lock);
710 list_del(&x->km.all);
711 hlist_del_rcu(&x->bydst);
712 hlist_del_rcu(&x->bysrc);
713 if (x->km.seq)
714 hlist_del_rcu(&x->byseq);
715 if (x->id.spi)
716 hlist_del_rcu(&x->byspi);
717 net->xfrm.state_num--;
718 spin_unlock(&net->xfrm.xfrm_state_lock);
719
720 if (x->encap_sk)
721 sock_put(rcu_dereference_raw(x->encap_sk));
722
723 xfrm_dev_state_delete(x);
724
725 /* All xfrm_state objects are created by xfrm_state_alloc.
726 * The xfrm_state_alloc call gives a reference, and that
727 * is what we are dropping here.
728 */
729 xfrm_state_put(x);
730 err = 0;
731 }
732
733 return err;
734}
735EXPORT_SYMBOL(__xfrm_state_delete);
736
737int xfrm_state_delete(struct xfrm_state *x)
738{
739 int err;
740
741 spin_lock_bh(&x->lock);
742 err = __xfrm_state_delete(x);
743 spin_unlock_bh(&x->lock);
744
745 return err;
746}
747EXPORT_SYMBOL(xfrm_state_delete);
748
749#ifdef CONFIG_SECURITY_NETWORK_XFRM
750static inline int
751xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
752{
753 int i, err = 0;
754
755 for (i = 0; i <= net->xfrm.state_hmask; i++) {
756 struct xfrm_state *x;
757
758 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
759 if (xfrm_id_proto_match(x->id.proto, proto) &&
760 (err = security_xfrm_state_delete(x)) != 0) {
761 xfrm_audit_state_delete(x, 0, task_valid);
762 return err;
763 }
764 }
765 }
766
767 return err;
768}
769
770static inline int
771xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
772{
773 int i, err = 0;
774
775 for (i = 0; i <= net->xfrm.state_hmask; i++) {
776 struct xfrm_state *x;
777 struct xfrm_dev_offload *xso;
778
779 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
780 xso = &x->xso;
781
782 if (xso->dev == dev &&
783 (err = security_xfrm_state_delete(x)) != 0) {
784 xfrm_audit_state_delete(x, 0, task_valid);
785 return err;
786 }
787 }
788 }
789
790 return err;
791}
792#else
793static inline int
794xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
795{
796 return 0;
797}
798
799static inline int
800xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
801{
802 return 0;
803}
804#endif
805
806int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
807{
808 int i, err = 0, cnt = 0;
809
810 spin_lock_bh(&net->xfrm.xfrm_state_lock);
811 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
812 if (err)
813 goto out;
814
815 err = -ESRCH;
816 for (i = 0; i <= net->xfrm.state_hmask; i++) {
817 struct xfrm_state *x;
818restart:
819 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
820 if (!xfrm_state_kern(x) &&
821 xfrm_id_proto_match(x->id.proto, proto)) {
822 xfrm_state_hold(x);
823 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
824
825 err = xfrm_state_delete(x);
826 xfrm_audit_state_delete(x, err ? 0 : 1,
827 task_valid);
828 if (sync)
829 xfrm_state_put_sync(x);
830 else
831 xfrm_state_put(x);
832 if (!err)
833 cnt++;
834
835 spin_lock_bh(&net->xfrm.xfrm_state_lock);
836 goto restart;
837 }
838 }
839 }
840out:
841 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
842 if (cnt)
843 err = 0;
844
845 return err;
846}
847EXPORT_SYMBOL(xfrm_state_flush);
848
849int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
850{
851 int i, err = 0, cnt = 0;
852
853 spin_lock_bh(&net->xfrm.xfrm_state_lock);
854 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
855 if (err)
856 goto out;
857
858 err = -ESRCH;
859 for (i = 0; i <= net->xfrm.state_hmask; i++) {
860 struct xfrm_state *x;
861 struct xfrm_dev_offload *xso;
862restart:
863 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
864 xso = &x->xso;
865
866 if (!xfrm_state_kern(x) && xso->dev == dev) {
867 xfrm_state_hold(x);
868 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
869
870 err = xfrm_state_delete(x);
871 xfrm_audit_state_delete(x, err ? 0 : 1,
872 task_valid);
873 xfrm_state_put(x);
874 if (!err)
875 cnt++;
876
877 spin_lock_bh(&net->xfrm.xfrm_state_lock);
878 goto restart;
879 }
880 }
881 }
882 if (cnt)
883 err = 0;
884
885out:
886 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
887 return err;
888}
889EXPORT_SYMBOL(xfrm_dev_state_flush);
890
891void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
892{
893 spin_lock_bh(&net->xfrm.xfrm_state_lock);
894 si->sadcnt = net->xfrm.state_num;
895 si->sadhcnt = net->xfrm.state_hmask + 1;
896 si->sadhmcnt = xfrm_state_hashmax;
897 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
898}
899EXPORT_SYMBOL(xfrm_sad_getinfo);
900
901static void
902__xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
903{
904 const struct flowi4 *fl4 = &fl->u.ip4;
905
906 sel->daddr.a4 = fl4->daddr;
907 sel->saddr.a4 = fl4->saddr;
908 sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
909 sel->dport_mask = htons(0xffff);
910 sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
911 sel->sport_mask = htons(0xffff);
912 sel->family = AF_INET;
913 sel->prefixlen_d = 32;
914 sel->prefixlen_s = 32;
915 sel->proto = fl4->flowi4_proto;
916 sel->ifindex = fl4->flowi4_oif;
917}
918
919static void
920__xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
921{
922 const struct flowi6 *fl6 = &fl->u.ip6;
923
924 /* Initialize temporary selector matching only to current session. */
925 *(struct in6_addr *)&sel->daddr = fl6->daddr;
926 *(struct in6_addr *)&sel->saddr = fl6->saddr;
927 sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
928 sel->dport_mask = htons(0xffff);
929 sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
930 sel->sport_mask = htons(0xffff);
931 sel->family = AF_INET6;
932 sel->prefixlen_d = 128;
933 sel->prefixlen_s = 128;
934 sel->proto = fl6->flowi6_proto;
935 sel->ifindex = fl6->flowi6_oif;
936}
937
938static void
939xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
940 const struct xfrm_tmpl *tmpl,
941 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
942 unsigned short family)
943{
944 switch (family) {
945 case AF_INET:
946 __xfrm4_init_tempsel(&x->sel, fl);
947 break;
948 case AF_INET6:
949 __xfrm6_init_tempsel(&x->sel, fl);
950 break;
951 }
952
953 x->id = tmpl->id;
954
955 switch (tmpl->encap_family) {
956 case AF_INET:
957 if (x->id.daddr.a4 == 0)
958 x->id.daddr.a4 = daddr->a4;
959 x->props.saddr = tmpl->saddr;
960 if (x->props.saddr.a4 == 0)
961 x->props.saddr.a4 = saddr->a4;
962 break;
963 case AF_INET6:
964 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
965 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
966 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
967 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
968 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
969 break;
970 }
971
972 x->props.mode = tmpl->mode;
973 x->props.reqid = tmpl->reqid;
974 x->props.family = tmpl->encap_family;
975}
976
977static struct xfrm_state *__xfrm_state_lookup_all(struct net *net, u32 mark,
978 const xfrm_address_t *daddr,
979 __be32 spi, u8 proto,
980 unsigned short family,
981 struct xfrm_dev_offload *xdo)
982{
983 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
984 struct xfrm_state *x;
985
986 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
987#ifdef CONFIG_XFRM_OFFLOAD
988 if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) {
989 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
990 /* HW states are in the head of list, there is
991 * no need to iterate further.
992 */
993 break;
994
995 /* Packet offload: both policy and SA should
996 * have same device.
997 */
998 if (xdo->dev != x->xso.dev)
999 continue;
1000 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1001 /* Skip HW policy for SW lookups */
1002 continue;
1003#endif
1004 if (x->props.family != family ||
1005 x->id.spi != spi ||
1006 x->id.proto != proto ||
1007 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1008 continue;
1009
1010 if ((mark & x->mark.m) != x->mark.v)
1011 continue;
1012 if (!xfrm_state_hold_rcu(x))
1013 continue;
1014 return x;
1015 }
1016
1017 return NULL;
1018}
1019
1020static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
1021 const xfrm_address_t *daddr,
1022 __be32 spi, u8 proto,
1023 unsigned short family)
1024{
1025 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
1026 struct xfrm_state *x;
1027
1028 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
1029 if (x->props.family != family ||
1030 x->id.spi != spi ||
1031 x->id.proto != proto ||
1032 !xfrm_addr_equal(&x->id.daddr, daddr, family))
1033 continue;
1034
1035 if ((mark & x->mark.m) != x->mark.v)
1036 continue;
1037 if (!xfrm_state_hold_rcu(x))
1038 continue;
1039 return x;
1040 }
1041
1042 return NULL;
1043}
1044
1045static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1046 const xfrm_address_t *daddr,
1047 const xfrm_address_t *saddr,
1048 u8 proto, unsigned short family)
1049{
1050 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
1051 struct xfrm_state *x;
1052
1053 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
1054 if (x->props.family != family ||
1055 x->id.proto != proto ||
1056 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1057 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1058 continue;
1059
1060 if ((mark & x->mark.m) != x->mark.v)
1061 continue;
1062 if (!xfrm_state_hold_rcu(x))
1063 continue;
1064 return x;
1065 }
1066
1067 return NULL;
1068}
1069
1070static inline struct xfrm_state *
1071__xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1072{
1073 struct net *net = xs_net(x);
1074 u32 mark = x->mark.v & x->mark.m;
1075
1076 if (use_spi)
1077 return __xfrm_state_lookup(net, mark, &x->id.daddr,
1078 x->id.spi, x->id.proto, family);
1079 else
1080 return __xfrm_state_lookup_byaddr(net, mark,
1081 &x->id.daddr,
1082 &x->props.saddr,
1083 x->id.proto, family);
1084}
1085
1086static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1087{
1088 if (have_hash_collision &&
1089 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1090 net->xfrm.state_num > net->xfrm.state_hmask)
1091 schedule_work(&net->xfrm.state_hash_work);
1092}
1093
1094static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1095 const struct flowi *fl, unsigned short family,
1096 struct xfrm_state **best, int *acq_in_progress,
1097 int *error)
1098{
1099 /* Resolution logic:
1100 * 1. There is a valid state with matching selector. Done.
1101 * 2. Valid state with inappropriate selector. Skip.
1102 *
1103 * Entering area of "sysdeps".
1104 *
1105 * 3. If state is not valid, selector is temporary, it selects
1106 * only session which triggered previous resolution. Key
1107 * manager will do something to install a state with proper
1108 * selector.
1109 */
1110 if (x->km.state == XFRM_STATE_VALID) {
1111 if ((x->sel.family &&
1112 (x->sel.family != family ||
1113 !xfrm_selector_match(&x->sel, fl, family))) ||
1114 !security_xfrm_state_pol_flow_match(x, pol,
1115 &fl->u.__fl_common))
1116 return;
1117
1118 if (!*best ||
1119 (*best)->km.dying > x->km.dying ||
1120 ((*best)->km.dying == x->km.dying &&
1121 (*best)->curlft.add_time < x->curlft.add_time))
1122 *best = x;
1123 } else if (x->km.state == XFRM_STATE_ACQ) {
1124 *acq_in_progress = 1;
1125 } else if (x->km.state == XFRM_STATE_ERROR ||
1126 x->km.state == XFRM_STATE_EXPIRED) {
1127 if ((!x->sel.family ||
1128 (x->sel.family == family &&
1129 xfrm_selector_match(&x->sel, fl, family))) &&
1130 security_xfrm_state_pol_flow_match(x, pol,
1131 &fl->u.__fl_common))
1132 *error = -ESRCH;
1133 }
1134}
1135
1136struct xfrm_state *
1137xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1138 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1139 struct xfrm_policy *pol, int *err,
1140 unsigned short family, u32 if_id)
1141{
1142 static xfrm_address_t saddr_wildcard = { };
1143 struct net *net = xp_net(pol);
1144 unsigned int h, h_wildcard;
1145 struct xfrm_state *x, *x0, *to_put;
1146 int acquire_in_progress = 0;
1147 int error = 0;
1148 struct xfrm_state *best = NULL;
1149 u32 mark = pol->mark.v & pol->mark.m;
1150 unsigned short encap_family = tmpl->encap_family;
1151 unsigned int sequence;
1152 struct km_event c;
1153
1154 to_put = NULL;
1155
1156 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1157
1158 rcu_read_lock();
1159 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1160 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1161#ifdef CONFIG_XFRM_OFFLOAD
1162 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1163 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1164 /* HW states are in the head of list, there is
1165 * no need to iterate further.
1166 */
1167 break;
1168
1169 /* Packet offload: both policy and SA should
1170 * have same device.
1171 */
1172 if (pol->xdo.dev != x->xso.dev)
1173 continue;
1174 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1175 /* Skip HW policy for SW lookups */
1176 continue;
1177#endif
1178 if (x->props.family == encap_family &&
1179 x->props.reqid == tmpl->reqid &&
1180 (mark & x->mark.m) == x->mark.v &&
1181 x->if_id == if_id &&
1182 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1183 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1184 tmpl->mode == x->props.mode &&
1185 tmpl->id.proto == x->id.proto &&
1186 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1187 xfrm_state_look_at(pol, x, fl, family,
1188 &best, &acquire_in_progress, &error);
1189 }
1190 if (best || acquire_in_progress)
1191 goto found;
1192
1193 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1194 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1195#ifdef CONFIG_XFRM_OFFLOAD
1196 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1197 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET)
1198 /* HW states are in the head of list, there is
1199 * no need to iterate further.
1200 */
1201 break;
1202
1203 /* Packet offload: both policy and SA should
1204 * have same device.
1205 */
1206 if (pol->xdo.dev != x->xso.dev)
1207 continue;
1208 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET)
1209 /* Skip HW policy for SW lookups */
1210 continue;
1211#endif
1212 if (x->props.family == encap_family &&
1213 x->props.reqid == tmpl->reqid &&
1214 (mark & x->mark.m) == x->mark.v &&
1215 x->if_id == if_id &&
1216 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1217 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1218 tmpl->mode == x->props.mode &&
1219 tmpl->id.proto == x->id.proto &&
1220 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1221 xfrm_state_look_at(pol, x, fl, family,
1222 &best, &acquire_in_progress, &error);
1223 }
1224
1225found:
1226 x = best;
1227 if (!x && !error && !acquire_in_progress) {
1228 if (tmpl->id.spi &&
1229 (x0 = __xfrm_state_lookup_all(net, mark, daddr,
1230 tmpl->id.spi, tmpl->id.proto,
1231 encap_family,
1232 &pol->xdo)) != NULL) {
1233 to_put = x0;
1234 error = -EEXIST;
1235 goto out;
1236 }
1237
1238 c.net = net;
1239 /* If the KMs have no listeners (yet...), avoid allocating an SA
1240 * for each and every packet - garbage collection might not
1241 * handle the flood.
1242 */
1243 if (!km_is_alive(&c)) {
1244 error = -ESRCH;
1245 goto out;
1246 }
1247
1248 x = xfrm_state_alloc(net);
1249 if (x == NULL) {
1250 error = -ENOMEM;
1251 goto out;
1252 }
1253 /* Initialize temporary state matching only
1254 * to current session. */
1255 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1256 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1257 x->if_id = if_id;
1258
1259 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1260 if (error) {
1261 x->km.state = XFRM_STATE_DEAD;
1262 to_put = x;
1263 x = NULL;
1264 goto out;
1265 }
1266#ifdef CONFIG_XFRM_OFFLOAD
1267 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) {
1268 struct xfrm_dev_offload *xdo = &pol->xdo;
1269 struct xfrm_dev_offload *xso = &x->xso;
1270
1271 xso->type = XFRM_DEV_OFFLOAD_PACKET;
1272 xso->dir = xdo->dir;
1273 xso->dev = xdo->dev;
1274 xso->real_dev = xdo->real_dev;
1275 xso->flags = XFRM_DEV_OFFLOAD_FLAG_ACQ;
1276 netdev_tracker_alloc(xso->dev, &xso->dev_tracker,
1277 GFP_ATOMIC);
1278 error = xso->dev->xfrmdev_ops->xdo_dev_state_add(x, NULL);
1279 if (error) {
1280 xso->dir = 0;
1281 netdev_put(xso->dev, &xso->dev_tracker);
1282 xso->dev = NULL;
1283 xso->real_dev = NULL;
1284 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED;
1285 x->km.state = XFRM_STATE_DEAD;
1286 to_put = x;
1287 x = NULL;
1288 goto out;
1289 }
1290 }
1291#endif
1292 if (km_query(x, tmpl, pol) == 0) {
1293 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1294 x->km.state = XFRM_STATE_ACQ;
1295 list_add(&x->km.all, &net->xfrm.state_all);
1296 XFRM_STATE_INSERT(bydst, &x->bydst,
1297 net->xfrm.state_bydst + h,
1298 x->xso.type);
1299 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1300 XFRM_STATE_INSERT(bysrc, &x->bysrc,
1301 net->xfrm.state_bysrc + h,
1302 x->xso.type);
1303 if (x->id.spi) {
1304 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1305 XFRM_STATE_INSERT(byspi, &x->byspi,
1306 net->xfrm.state_byspi + h,
1307 x->xso.type);
1308 }
1309 if (x->km.seq) {
1310 h = xfrm_seq_hash(net, x->km.seq);
1311 XFRM_STATE_INSERT(byseq, &x->byseq,
1312 net->xfrm.state_byseq + h,
1313 x->xso.type);
1314 }
1315 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1316 hrtimer_start(&x->mtimer,
1317 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1318 HRTIMER_MODE_REL_SOFT);
1319 net->xfrm.state_num++;
1320 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1321 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1322 } else {
1323#ifdef CONFIG_XFRM_OFFLOAD
1324 struct xfrm_dev_offload *xso = &x->xso;
1325
1326 if (xso->type == XFRM_DEV_OFFLOAD_PACKET) {
1327 xfrm_dev_state_delete(x);
1328 xfrm_dev_state_free(x);
1329 }
1330#endif
1331 x->km.state = XFRM_STATE_DEAD;
1332 to_put = x;
1333 x = NULL;
1334 error = -ESRCH;
1335 }
1336 }
1337out:
1338 if (x) {
1339 if (!xfrm_state_hold_rcu(x)) {
1340 *err = -EAGAIN;
1341 x = NULL;
1342 }
1343 } else {
1344 *err = acquire_in_progress ? -EAGAIN : error;
1345 }
1346 rcu_read_unlock();
1347 if (to_put)
1348 xfrm_state_put(to_put);
1349
1350 if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1351 *err = -EAGAIN;
1352 if (x) {
1353 xfrm_state_put(x);
1354 x = NULL;
1355 }
1356 }
1357
1358 return x;
1359}
1360
1361struct xfrm_state *
1362xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1363 xfrm_address_t *daddr, xfrm_address_t *saddr,
1364 unsigned short family, u8 mode, u8 proto, u32 reqid)
1365{
1366 unsigned int h;
1367 struct xfrm_state *rx = NULL, *x = NULL;
1368
1369 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1370 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1371 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1372 if (x->props.family == family &&
1373 x->props.reqid == reqid &&
1374 (mark & x->mark.m) == x->mark.v &&
1375 x->if_id == if_id &&
1376 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1377 xfrm_state_addr_check(x, daddr, saddr, family) &&
1378 mode == x->props.mode &&
1379 proto == x->id.proto &&
1380 x->km.state == XFRM_STATE_VALID) {
1381 rx = x;
1382 break;
1383 }
1384 }
1385
1386 if (rx)
1387 xfrm_state_hold(rx);
1388 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1389
1390
1391 return rx;
1392}
1393EXPORT_SYMBOL(xfrm_stateonly_find);
1394
1395struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1396 unsigned short family)
1397{
1398 struct xfrm_state *x;
1399 struct xfrm_state_walk *w;
1400
1401 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1402 list_for_each_entry(w, &net->xfrm.state_all, all) {
1403 x = container_of(w, struct xfrm_state, km);
1404 if (x->props.family != family ||
1405 x->id.spi != spi)
1406 continue;
1407
1408 xfrm_state_hold(x);
1409 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1410 return x;
1411 }
1412 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1413 return NULL;
1414}
1415EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1416
1417static void __xfrm_state_insert(struct xfrm_state *x)
1418{
1419 struct net *net = xs_net(x);
1420 unsigned int h;
1421
1422 list_add(&x->km.all, &net->xfrm.state_all);
1423
1424 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1425 x->props.reqid, x->props.family);
1426 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1427 x->xso.type);
1428
1429 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1430 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1431 x->xso.type);
1432
1433 if (x->id.spi) {
1434 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1435 x->props.family);
1436
1437 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
1438 x->xso.type);
1439 }
1440
1441 if (x->km.seq) {
1442 h = xfrm_seq_hash(net, x->km.seq);
1443
1444 XFRM_STATE_INSERT(byseq, &x->byseq, net->xfrm.state_byseq + h,
1445 x->xso.type);
1446 }
1447
1448 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1449 if (x->replay_maxage)
1450 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1451
1452 net->xfrm.state_num++;
1453
1454 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1455}
1456
1457/* net->xfrm.xfrm_state_lock is held */
1458static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1459{
1460 struct net *net = xs_net(xnew);
1461 unsigned short family = xnew->props.family;
1462 u32 reqid = xnew->props.reqid;
1463 struct xfrm_state *x;
1464 unsigned int h;
1465 u32 mark = xnew->mark.v & xnew->mark.m;
1466 u32 if_id = xnew->if_id;
1467
1468 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1469 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1470 if (x->props.family == family &&
1471 x->props.reqid == reqid &&
1472 x->if_id == if_id &&
1473 (mark & x->mark.m) == x->mark.v &&
1474 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1475 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1476 x->genid++;
1477 }
1478}
1479
1480void xfrm_state_insert(struct xfrm_state *x)
1481{
1482 struct net *net = xs_net(x);
1483
1484 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1485 __xfrm_state_bump_genids(x);
1486 __xfrm_state_insert(x);
1487 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1488}
1489EXPORT_SYMBOL(xfrm_state_insert);
1490
1491/* net->xfrm.xfrm_state_lock is held */
1492static struct xfrm_state *__find_acq_core(struct net *net,
1493 const struct xfrm_mark *m,
1494 unsigned short family, u8 mode,
1495 u32 reqid, u32 if_id, u8 proto,
1496 const xfrm_address_t *daddr,
1497 const xfrm_address_t *saddr,
1498 int create)
1499{
1500 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1501 struct xfrm_state *x;
1502 u32 mark = m->v & m->m;
1503
1504 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1505 if (x->props.reqid != reqid ||
1506 x->props.mode != mode ||
1507 x->props.family != family ||
1508 x->km.state != XFRM_STATE_ACQ ||
1509 x->id.spi != 0 ||
1510 x->id.proto != proto ||
1511 (mark & x->mark.m) != x->mark.v ||
1512 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1513 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1514 continue;
1515
1516 xfrm_state_hold(x);
1517 return x;
1518 }
1519
1520 if (!create)
1521 return NULL;
1522
1523 x = xfrm_state_alloc(net);
1524 if (likely(x)) {
1525 switch (family) {
1526 case AF_INET:
1527 x->sel.daddr.a4 = daddr->a4;
1528 x->sel.saddr.a4 = saddr->a4;
1529 x->sel.prefixlen_d = 32;
1530 x->sel.prefixlen_s = 32;
1531 x->props.saddr.a4 = saddr->a4;
1532 x->id.daddr.a4 = daddr->a4;
1533 break;
1534
1535 case AF_INET6:
1536 x->sel.daddr.in6 = daddr->in6;
1537 x->sel.saddr.in6 = saddr->in6;
1538 x->sel.prefixlen_d = 128;
1539 x->sel.prefixlen_s = 128;
1540 x->props.saddr.in6 = saddr->in6;
1541 x->id.daddr.in6 = daddr->in6;
1542 break;
1543 }
1544
1545 x->km.state = XFRM_STATE_ACQ;
1546 x->id.proto = proto;
1547 x->props.family = family;
1548 x->props.mode = mode;
1549 x->props.reqid = reqid;
1550 x->if_id = if_id;
1551 x->mark.v = m->v;
1552 x->mark.m = m->m;
1553 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1554 xfrm_state_hold(x);
1555 hrtimer_start(&x->mtimer,
1556 ktime_set(net->xfrm.sysctl_acq_expires, 0),
1557 HRTIMER_MODE_REL_SOFT);
1558 list_add(&x->km.all, &net->xfrm.state_all);
1559 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h,
1560 x->xso.type);
1561 h = xfrm_src_hash(net, daddr, saddr, family);
1562 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h,
1563 x->xso.type);
1564
1565 net->xfrm.state_num++;
1566
1567 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1568 }
1569
1570 return x;
1571}
1572
1573static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1574
1575int xfrm_state_add(struct xfrm_state *x)
1576{
1577 struct net *net = xs_net(x);
1578 struct xfrm_state *x1, *to_put;
1579 int family;
1580 int err;
1581 u32 mark = x->mark.v & x->mark.m;
1582 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1583
1584 family = x->props.family;
1585
1586 to_put = NULL;
1587
1588 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1589
1590 x1 = __xfrm_state_locate(x, use_spi, family);
1591 if (x1) {
1592 to_put = x1;
1593 x1 = NULL;
1594 err = -EEXIST;
1595 goto out;
1596 }
1597
1598 if (use_spi && x->km.seq) {
1599 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1600 if (x1 && ((x1->id.proto != x->id.proto) ||
1601 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1602 to_put = x1;
1603 x1 = NULL;
1604 }
1605 }
1606
1607 if (use_spi && !x1)
1608 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1609 x->props.reqid, x->if_id, x->id.proto,
1610 &x->id.daddr, &x->props.saddr, 0);
1611
1612 __xfrm_state_bump_genids(x);
1613 __xfrm_state_insert(x);
1614 err = 0;
1615
1616out:
1617 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1618
1619 if (x1) {
1620 xfrm_state_delete(x1);
1621 xfrm_state_put(x1);
1622 }
1623
1624 if (to_put)
1625 xfrm_state_put(to_put);
1626
1627 return err;
1628}
1629EXPORT_SYMBOL(xfrm_state_add);
1630
1631#ifdef CONFIG_XFRM_MIGRATE
1632static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1633{
1634 struct xfrm_user_sec_ctx *uctx;
1635 int size = sizeof(*uctx) + security->ctx_len;
1636 int err;
1637
1638 uctx = kmalloc(size, GFP_KERNEL);
1639 if (!uctx)
1640 return -ENOMEM;
1641
1642 uctx->exttype = XFRMA_SEC_CTX;
1643 uctx->len = size;
1644 uctx->ctx_doi = security->ctx_doi;
1645 uctx->ctx_alg = security->ctx_alg;
1646 uctx->ctx_len = security->ctx_len;
1647 memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1648 err = security_xfrm_state_alloc(x, uctx);
1649 kfree(uctx);
1650 if (err)
1651 return err;
1652
1653 return 0;
1654}
1655
1656static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1657 struct xfrm_encap_tmpl *encap)
1658{
1659 struct net *net = xs_net(orig);
1660 struct xfrm_state *x = xfrm_state_alloc(net);
1661 if (!x)
1662 goto out;
1663
1664 memcpy(&x->id, &orig->id, sizeof(x->id));
1665 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1666 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1667 x->props.mode = orig->props.mode;
1668 x->props.replay_window = orig->props.replay_window;
1669 x->props.reqid = orig->props.reqid;
1670 x->props.family = orig->props.family;
1671 x->props.saddr = orig->props.saddr;
1672
1673 if (orig->aalg) {
1674 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1675 if (!x->aalg)
1676 goto error;
1677 }
1678 x->props.aalgo = orig->props.aalgo;
1679
1680 if (orig->aead) {
1681 x->aead = xfrm_algo_aead_clone(orig->aead);
1682 x->geniv = orig->geniv;
1683 if (!x->aead)
1684 goto error;
1685 }
1686 if (orig->ealg) {
1687 x->ealg = xfrm_algo_clone(orig->ealg);
1688 if (!x->ealg)
1689 goto error;
1690 }
1691 x->props.ealgo = orig->props.ealgo;
1692
1693 if (orig->calg) {
1694 x->calg = xfrm_algo_clone(orig->calg);
1695 if (!x->calg)
1696 goto error;
1697 }
1698 x->props.calgo = orig->props.calgo;
1699
1700 if (encap || orig->encap) {
1701 if (encap)
1702 x->encap = kmemdup(encap, sizeof(*x->encap),
1703 GFP_KERNEL);
1704 else
1705 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1706 GFP_KERNEL);
1707
1708 if (!x->encap)
1709 goto error;
1710 }
1711
1712 if (orig->security)
1713 if (clone_security(x, orig->security))
1714 goto error;
1715
1716 if (orig->coaddr) {
1717 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1718 GFP_KERNEL);
1719 if (!x->coaddr)
1720 goto error;
1721 }
1722
1723 if (orig->replay_esn) {
1724 if (xfrm_replay_clone(x, orig))
1725 goto error;
1726 }
1727
1728 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1729 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1730
1731 x->props.flags = orig->props.flags;
1732 x->props.extra_flags = orig->props.extra_flags;
1733
1734 x->if_id = orig->if_id;
1735 x->tfcpad = orig->tfcpad;
1736 x->replay_maxdiff = orig->replay_maxdiff;
1737 x->replay_maxage = orig->replay_maxage;
1738 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1739 x->km.state = orig->km.state;
1740 x->km.seq = orig->km.seq;
1741 x->replay = orig->replay;
1742 x->preplay = orig->preplay;
1743 x->mapping_maxage = orig->mapping_maxage;
1744 x->lastused = orig->lastused;
1745 x->new_mapping = 0;
1746 x->new_mapping_sport = 0;
1747
1748 return x;
1749
1750 error:
1751 xfrm_state_put(x);
1752out:
1753 return NULL;
1754}
1755
1756struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
1757 u32 if_id)
1758{
1759 unsigned int h;
1760 struct xfrm_state *x = NULL;
1761
1762 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1763
1764 if (m->reqid) {
1765 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1766 m->reqid, m->old_family);
1767 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1768 if (x->props.mode != m->mode ||
1769 x->id.proto != m->proto)
1770 continue;
1771 if (m->reqid && x->props.reqid != m->reqid)
1772 continue;
1773 if (if_id != 0 && x->if_id != if_id)
1774 continue;
1775 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1776 m->old_family) ||
1777 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1778 m->old_family))
1779 continue;
1780 xfrm_state_hold(x);
1781 break;
1782 }
1783 } else {
1784 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1785 m->old_family);
1786 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1787 if (x->props.mode != m->mode ||
1788 x->id.proto != m->proto)
1789 continue;
1790 if (if_id != 0 && x->if_id != if_id)
1791 continue;
1792 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1793 m->old_family) ||
1794 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1795 m->old_family))
1796 continue;
1797 xfrm_state_hold(x);
1798 break;
1799 }
1800 }
1801
1802 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1803
1804 return x;
1805}
1806EXPORT_SYMBOL(xfrm_migrate_state_find);
1807
1808struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1809 struct xfrm_migrate *m,
1810 struct xfrm_encap_tmpl *encap)
1811{
1812 struct xfrm_state *xc;
1813
1814 xc = xfrm_state_clone(x, encap);
1815 if (!xc)
1816 return NULL;
1817
1818 xc->props.family = m->new_family;
1819
1820 if (xfrm_init_state(xc) < 0)
1821 goto error;
1822
1823 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1824 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1825
1826 /* add state */
1827 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1828 /* a care is needed when the destination address of the
1829 state is to be updated as it is a part of triplet */
1830 xfrm_state_insert(xc);
1831 } else {
1832 if (xfrm_state_add(xc) < 0)
1833 goto error;
1834 }
1835
1836 return xc;
1837error:
1838 xfrm_state_put(xc);
1839 return NULL;
1840}
1841EXPORT_SYMBOL(xfrm_state_migrate);
1842#endif
1843
1844int xfrm_state_update(struct xfrm_state *x)
1845{
1846 struct xfrm_state *x1, *to_put;
1847 int err;
1848 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1849 struct net *net = xs_net(x);
1850
1851 to_put = NULL;
1852
1853 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1854 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1855
1856 err = -ESRCH;
1857 if (!x1)
1858 goto out;
1859
1860 if (xfrm_state_kern(x1)) {
1861 to_put = x1;
1862 err = -EEXIST;
1863 goto out;
1864 }
1865
1866 if (x1->km.state == XFRM_STATE_ACQ) {
1867 __xfrm_state_insert(x);
1868 x = NULL;
1869 }
1870 err = 0;
1871
1872out:
1873 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1874
1875 if (to_put)
1876 xfrm_state_put(to_put);
1877
1878 if (err)
1879 return err;
1880
1881 if (!x) {
1882 xfrm_state_delete(x1);
1883 xfrm_state_put(x1);
1884 return 0;
1885 }
1886
1887 err = -EINVAL;
1888 spin_lock_bh(&x1->lock);
1889 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1890 if (x->encap && x1->encap &&
1891 x->encap->encap_type == x1->encap->encap_type)
1892 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1893 else if (x->encap || x1->encap)
1894 goto fail;
1895
1896 if (x->coaddr && x1->coaddr) {
1897 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1898 }
1899 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1900 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1901 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1902 x1->km.dying = 0;
1903
1904 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1905 HRTIMER_MODE_REL_SOFT);
1906 if (READ_ONCE(x1->curlft.use_time))
1907 xfrm_state_check_expire(x1);
1908
1909 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1910 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1911
1912 if (x->props.smark.m || x->props.smark.v)
1913 x1->props.smark = x->props.smark;
1914
1915 if (x->if_id)
1916 x1->if_id = x->if_id;
1917
1918 __xfrm_state_bump_genids(x1);
1919 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1920 }
1921
1922 err = 0;
1923 x->km.state = XFRM_STATE_DEAD;
1924 __xfrm_state_put(x);
1925 }
1926
1927fail:
1928 spin_unlock_bh(&x1->lock);
1929
1930 xfrm_state_put(x1);
1931
1932 return err;
1933}
1934EXPORT_SYMBOL(xfrm_state_update);
1935
1936int xfrm_state_check_expire(struct xfrm_state *x)
1937{
1938 xfrm_dev_state_update_curlft(x);
1939
1940 if (!READ_ONCE(x->curlft.use_time))
1941 WRITE_ONCE(x->curlft.use_time, ktime_get_real_seconds());
1942
1943 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1944 x->curlft.packets >= x->lft.hard_packet_limit) {
1945 x->km.state = XFRM_STATE_EXPIRED;
1946 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1947 return -EINVAL;
1948 }
1949
1950 if (!x->km.dying &&
1951 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1952 x->curlft.packets >= x->lft.soft_packet_limit)) {
1953 x->km.dying = 1;
1954 km_state_expired(x, 0, 0);
1955 }
1956 return 0;
1957}
1958EXPORT_SYMBOL(xfrm_state_check_expire);
1959
1960struct xfrm_state *
1961xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1962 u8 proto, unsigned short family)
1963{
1964 struct xfrm_state *x;
1965
1966 rcu_read_lock();
1967 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1968 rcu_read_unlock();
1969 return x;
1970}
1971EXPORT_SYMBOL(xfrm_state_lookup);
1972
1973struct xfrm_state *
1974xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1975 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1976 u8 proto, unsigned short family)
1977{
1978 struct xfrm_state *x;
1979
1980 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1981 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1982 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1983 return x;
1984}
1985EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1986
1987struct xfrm_state *
1988xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1989 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1990 const xfrm_address_t *saddr, int create, unsigned short family)
1991{
1992 struct xfrm_state *x;
1993
1994 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1995 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1996 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1997
1998 return x;
1999}
2000EXPORT_SYMBOL(xfrm_find_acq);
2001
2002#ifdef CONFIG_XFRM_SUB_POLICY
2003#if IS_ENABLED(CONFIG_IPV6)
2004/* distribution counting sort function for xfrm_state and xfrm_tmpl */
2005static void
2006__xfrm6_sort(void **dst, void **src, int n,
2007 int (*cmp)(const void *p), int maxclass)
2008{
2009 int count[XFRM_MAX_DEPTH] = { };
2010 int class[XFRM_MAX_DEPTH];
2011 int i;
2012
2013 for (i = 0; i < n; i++) {
2014 int c = cmp(src[i]);
2015
2016 class[i] = c;
2017 count[c]++;
2018 }
2019
2020 for (i = 2; i < maxclass; i++)
2021 count[i] += count[i - 1];
2022
2023 for (i = 0; i < n; i++) {
2024 dst[count[class[i] - 1]++] = src[i];
2025 src[i] = NULL;
2026 }
2027}
2028
2029/* Rule for xfrm_state:
2030 *
2031 * rule 1: select IPsec transport except AH
2032 * rule 2: select MIPv6 RO or inbound trigger
2033 * rule 3: select IPsec transport AH
2034 * rule 4: select IPsec tunnel
2035 * rule 5: others
2036 */
2037static int __xfrm6_state_sort_cmp(const void *p)
2038{
2039 const struct xfrm_state *v = p;
2040
2041 switch (v->props.mode) {
2042 case XFRM_MODE_TRANSPORT:
2043 if (v->id.proto != IPPROTO_AH)
2044 return 1;
2045 else
2046 return 3;
2047#if IS_ENABLED(CONFIG_IPV6_MIP6)
2048 case XFRM_MODE_ROUTEOPTIMIZATION:
2049 case XFRM_MODE_IN_TRIGGER:
2050 return 2;
2051#endif
2052 case XFRM_MODE_TUNNEL:
2053 case XFRM_MODE_BEET:
2054 return 4;
2055 }
2056 return 5;
2057}
2058
2059/* Rule for xfrm_tmpl:
2060 *
2061 * rule 1: select IPsec transport
2062 * rule 2: select MIPv6 RO or inbound trigger
2063 * rule 3: select IPsec tunnel
2064 * rule 4: others
2065 */
2066static int __xfrm6_tmpl_sort_cmp(const void *p)
2067{
2068 const struct xfrm_tmpl *v = p;
2069
2070 switch (v->mode) {
2071 case XFRM_MODE_TRANSPORT:
2072 return 1;
2073#if IS_ENABLED(CONFIG_IPV6_MIP6)
2074 case XFRM_MODE_ROUTEOPTIMIZATION:
2075 case XFRM_MODE_IN_TRIGGER:
2076 return 2;
2077#endif
2078 case XFRM_MODE_TUNNEL:
2079 case XFRM_MODE_BEET:
2080 return 3;
2081 }
2082 return 4;
2083}
2084#else
2085static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
2086static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
2087
2088static inline void
2089__xfrm6_sort(void **dst, void **src, int n,
2090 int (*cmp)(const void *p), int maxclass)
2091{
2092 int i;
2093
2094 for (i = 0; i < n; i++)
2095 dst[i] = src[i];
2096}
2097#endif /* CONFIG_IPV6 */
2098
2099void
2100xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
2101 unsigned short family)
2102{
2103 int i;
2104
2105 if (family == AF_INET6)
2106 __xfrm6_sort((void **)dst, (void **)src, n,
2107 __xfrm6_tmpl_sort_cmp, 5);
2108 else
2109 for (i = 0; i < n; i++)
2110 dst[i] = src[i];
2111}
2112
2113void
2114xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
2115 unsigned short family)
2116{
2117 int i;
2118
2119 if (family == AF_INET6)
2120 __xfrm6_sort((void **)dst, (void **)src, n,
2121 __xfrm6_state_sort_cmp, 6);
2122 else
2123 for (i = 0; i < n; i++)
2124 dst[i] = src[i];
2125}
2126#endif
2127
2128/* Silly enough, but I'm lazy to build resolution list */
2129
2130static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
2131{
2132 unsigned int h = xfrm_seq_hash(net, seq);
2133 struct xfrm_state *x;
2134
2135 hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
2136 if (x->km.seq == seq &&
2137 (mark & x->mark.m) == x->mark.v &&
2138 x->km.state == XFRM_STATE_ACQ) {
2139 xfrm_state_hold(x);
2140 return x;
2141 }
2142 }
2143
2144 return NULL;
2145}
2146
2147struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
2148{
2149 struct xfrm_state *x;
2150
2151 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2152 x = __xfrm_find_acq_byseq(net, mark, seq);
2153 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2154 return x;
2155}
2156EXPORT_SYMBOL(xfrm_find_acq_byseq);
2157
2158u32 xfrm_get_acqseq(void)
2159{
2160 u32 res;
2161 static atomic_t acqseq;
2162
2163 do {
2164 res = atomic_inc_return(&acqseq);
2165 } while (!res);
2166
2167 return res;
2168}
2169EXPORT_SYMBOL(xfrm_get_acqseq);
2170
2171int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack)
2172{
2173 switch (proto) {
2174 case IPPROTO_AH:
2175 case IPPROTO_ESP:
2176 break;
2177
2178 case IPPROTO_COMP:
2179 /* IPCOMP spi is 16-bits. */
2180 if (max >= 0x10000) {
2181 NL_SET_ERR_MSG(extack, "IPCOMP SPI must be <= 65535");
2182 return -EINVAL;
2183 }
2184 break;
2185
2186 default:
2187 NL_SET_ERR_MSG(extack, "Invalid protocol, must be one of AH, ESP, IPCOMP");
2188 return -EINVAL;
2189 }
2190
2191 if (min > max) {
2192 NL_SET_ERR_MSG(extack, "Invalid SPI range: min > max");
2193 return -EINVAL;
2194 }
2195
2196 return 0;
2197}
2198EXPORT_SYMBOL(verify_spi_info);
2199
2200int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high,
2201 struct netlink_ext_ack *extack)
2202{
2203 struct net *net = xs_net(x);
2204 unsigned int h;
2205 struct xfrm_state *x0;
2206 int err = -ENOENT;
2207 __be32 minspi = htonl(low);
2208 __be32 maxspi = htonl(high);
2209 __be32 newspi = 0;
2210 u32 mark = x->mark.v & x->mark.m;
2211
2212 spin_lock_bh(&x->lock);
2213 if (x->km.state == XFRM_STATE_DEAD) {
2214 NL_SET_ERR_MSG(extack, "Target ACQUIRE is in DEAD state");
2215 goto unlock;
2216 }
2217
2218 err = 0;
2219 if (x->id.spi)
2220 goto unlock;
2221
2222 err = -ENOENT;
2223
2224 if (minspi == maxspi) {
2225 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2226 if (x0) {
2227 NL_SET_ERR_MSG(extack, "Requested SPI is already in use");
2228 xfrm_state_put(x0);
2229 goto unlock;
2230 }
2231 newspi = minspi;
2232 } else {
2233 u32 spi = 0;
2234 for (h = 0; h < high-low+1; h++) {
2235 spi = get_random_u32_inclusive(low, high);
2236 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2237 if (x0 == NULL) {
2238 newspi = htonl(spi);
2239 break;
2240 }
2241 xfrm_state_put(x0);
2242 }
2243 }
2244 if (newspi) {
2245 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2246 x->id.spi = newspi;
2247 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2248 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h,
2249 x->xso.type);
2250 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2251
2252 err = 0;
2253 } else {
2254 NL_SET_ERR_MSG(extack, "No SPI available in the requested range");
2255 }
2256
2257unlock:
2258 spin_unlock_bh(&x->lock);
2259
2260 return err;
2261}
2262EXPORT_SYMBOL(xfrm_alloc_spi);
2263
2264static bool __xfrm_state_filter_match(struct xfrm_state *x,
2265 struct xfrm_address_filter *filter)
2266{
2267 if (filter) {
2268 if ((filter->family == AF_INET ||
2269 filter->family == AF_INET6) &&
2270 x->props.family != filter->family)
2271 return false;
2272
2273 return addr_match(&x->props.saddr, &filter->saddr,
2274 filter->splen) &&
2275 addr_match(&x->id.daddr, &filter->daddr,
2276 filter->dplen);
2277 }
2278 return true;
2279}
2280
2281int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2282 int (*func)(struct xfrm_state *, int, void*),
2283 void *data)
2284{
2285 struct xfrm_state *state;
2286 struct xfrm_state_walk *x;
2287 int err = 0;
2288
2289 if (walk->seq != 0 && list_empty(&walk->all))
2290 return 0;
2291
2292 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2293 if (list_empty(&walk->all))
2294 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2295 else
2296 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2297 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2298 if (x->state == XFRM_STATE_DEAD)
2299 continue;
2300 state = container_of(x, struct xfrm_state, km);
2301 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2302 continue;
2303 if (!__xfrm_state_filter_match(state, walk->filter))
2304 continue;
2305 err = func(state, walk->seq, data);
2306 if (err) {
2307 list_move_tail(&walk->all, &x->all);
2308 goto out;
2309 }
2310 walk->seq++;
2311 }
2312 if (walk->seq == 0) {
2313 err = -ENOENT;
2314 goto out;
2315 }
2316 list_del_init(&walk->all);
2317out:
2318 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2319 return err;
2320}
2321EXPORT_SYMBOL(xfrm_state_walk);
2322
2323void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2324 struct xfrm_address_filter *filter)
2325{
2326 INIT_LIST_HEAD(&walk->all);
2327 walk->proto = proto;
2328 walk->state = XFRM_STATE_DEAD;
2329 walk->seq = 0;
2330 walk->filter = filter;
2331}
2332EXPORT_SYMBOL(xfrm_state_walk_init);
2333
2334void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2335{
2336 kfree(walk->filter);
2337
2338 if (list_empty(&walk->all))
2339 return;
2340
2341 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2342 list_del(&walk->all);
2343 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2344}
2345EXPORT_SYMBOL(xfrm_state_walk_done);
2346
2347static void xfrm_replay_timer_handler(struct timer_list *t)
2348{
2349 struct xfrm_state *x = from_timer(x, t, rtimer);
2350
2351 spin_lock(&x->lock);
2352
2353 if (x->km.state == XFRM_STATE_VALID) {
2354 if (xfrm_aevent_is_on(xs_net(x)))
2355 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2356 else
2357 x->xflags |= XFRM_TIME_DEFER;
2358 }
2359
2360 spin_unlock(&x->lock);
2361}
2362
2363static LIST_HEAD(xfrm_km_list);
2364
2365void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2366{
2367 struct xfrm_mgr *km;
2368
2369 rcu_read_lock();
2370 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2371 if (km->notify_policy)
2372 km->notify_policy(xp, dir, c);
2373 rcu_read_unlock();
2374}
2375
2376void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2377{
2378 struct xfrm_mgr *km;
2379 rcu_read_lock();
2380 list_for_each_entry_rcu(km, &xfrm_km_list, list)
2381 if (km->notify)
2382 km->notify(x, c);
2383 rcu_read_unlock();
2384}
2385
2386EXPORT_SYMBOL(km_policy_notify);
2387EXPORT_SYMBOL(km_state_notify);
2388
2389void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2390{
2391 struct km_event c;
2392
2393 c.data.hard = hard;
2394 c.portid = portid;
2395 c.event = XFRM_MSG_EXPIRE;
2396 km_state_notify(x, &c);
2397}
2398
2399EXPORT_SYMBOL(km_state_expired);
2400/*
2401 * We send to all registered managers regardless of failure
2402 * We are happy with one success
2403*/
2404int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2405{
2406 int err = -EINVAL, acqret;
2407 struct xfrm_mgr *km;
2408
2409 rcu_read_lock();
2410 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2411 acqret = km->acquire(x, t, pol);
2412 if (!acqret)
2413 err = acqret;
2414 }
2415 rcu_read_unlock();
2416 return err;
2417}
2418EXPORT_SYMBOL(km_query);
2419
2420static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2421{
2422 int err = -EINVAL;
2423 struct xfrm_mgr *km;
2424
2425 rcu_read_lock();
2426 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2427 if (km->new_mapping)
2428 err = km->new_mapping(x, ipaddr, sport);
2429 if (!err)
2430 break;
2431 }
2432 rcu_read_unlock();
2433 return err;
2434}
2435
2436int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2437{
2438 int ret = 0;
2439
2440 if (x->mapping_maxage) {
2441 if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage ||
2442 x->new_mapping_sport != sport) {
2443 x->new_mapping_sport = sport;
2444 x->new_mapping = jiffies / HZ;
2445 ret = __km_new_mapping(x, ipaddr, sport);
2446 }
2447 } else {
2448 ret = __km_new_mapping(x, ipaddr, sport);
2449 }
2450
2451 return ret;
2452}
2453EXPORT_SYMBOL(km_new_mapping);
2454
2455void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2456{
2457 struct km_event c;
2458
2459 c.data.hard = hard;
2460 c.portid = portid;
2461 c.event = XFRM_MSG_POLEXPIRE;
2462 km_policy_notify(pol, dir, &c);
2463}
2464EXPORT_SYMBOL(km_policy_expired);
2465
2466#ifdef CONFIG_XFRM_MIGRATE
2467int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2468 const struct xfrm_migrate *m, int num_migrate,
2469 const struct xfrm_kmaddress *k,
2470 const struct xfrm_encap_tmpl *encap)
2471{
2472 int err = -EINVAL;
2473 int ret;
2474 struct xfrm_mgr *km;
2475
2476 rcu_read_lock();
2477 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2478 if (km->migrate) {
2479 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2480 encap);
2481 if (!ret)
2482 err = ret;
2483 }
2484 }
2485 rcu_read_unlock();
2486 return err;
2487}
2488EXPORT_SYMBOL(km_migrate);
2489#endif
2490
2491int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2492{
2493 int err = -EINVAL;
2494 int ret;
2495 struct xfrm_mgr *km;
2496
2497 rcu_read_lock();
2498 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2499 if (km->report) {
2500 ret = km->report(net, proto, sel, addr);
2501 if (!ret)
2502 err = ret;
2503 }
2504 }
2505 rcu_read_unlock();
2506 return err;
2507}
2508EXPORT_SYMBOL(km_report);
2509
2510static bool km_is_alive(const struct km_event *c)
2511{
2512 struct xfrm_mgr *km;
2513 bool is_alive = false;
2514
2515 rcu_read_lock();
2516 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2517 if (km->is_alive && km->is_alive(c)) {
2518 is_alive = true;
2519 break;
2520 }
2521 }
2522 rcu_read_unlock();
2523
2524 return is_alive;
2525}
2526
2527#if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2528static DEFINE_SPINLOCK(xfrm_translator_lock);
2529static struct xfrm_translator __rcu *xfrm_translator;
2530
2531struct xfrm_translator *xfrm_get_translator(void)
2532{
2533 struct xfrm_translator *xtr;
2534
2535 rcu_read_lock();
2536 xtr = rcu_dereference(xfrm_translator);
2537 if (unlikely(!xtr))
2538 goto out;
2539 if (!try_module_get(xtr->owner))
2540 xtr = NULL;
2541out:
2542 rcu_read_unlock();
2543 return xtr;
2544}
2545EXPORT_SYMBOL_GPL(xfrm_get_translator);
2546
2547void xfrm_put_translator(struct xfrm_translator *xtr)
2548{
2549 module_put(xtr->owner);
2550}
2551EXPORT_SYMBOL_GPL(xfrm_put_translator);
2552
2553int xfrm_register_translator(struct xfrm_translator *xtr)
2554{
2555 int err = 0;
2556
2557 spin_lock_bh(&xfrm_translator_lock);
2558 if (unlikely(xfrm_translator != NULL))
2559 err = -EEXIST;
2560 else
2561 rcu_assign_pointer(xfrm_translator, xtr);
2562 spin_unlock_bh(&xfrm_translator_lock);
2563
2564 return err;
2565}
2566EXPORT_SYMBOL_GPL(xfrm_register_translator);
2567
2568int xfrm_unregister_translator(struct xfrm_translator *xtr)
2569{
2570 int err = 0;
2571
2572 spin_lock_bh(&xfrm_translator_lock);
2573 if (likely(xfrm_translator != NULL)) {
2574 if (rcu_access_pointer(xfrm_translator) != xtr)
2575 err = -EINVAL;
2576 else
2577 RCU_INIT_POINTER(xfrm_translator, NULL);
2578 }
2579 spin_unlock_bh(&xfrm_translator_lock);
2580 synchronize_rcu();
2581
2582 return err;
2583}
2584EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2585#endif
2586
2587int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2588{
2589 int err;
2590 u8 *data;
2591 struct xfrm_mgr *km;
2592 struct xfrm_policy *pol = NULL;
2593
2594 if (sockptr_is_null(optval) && !optlen) {
2595 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2596 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2597 __sk_dst_reset(sk);
2598 return 0;
2599 }
2600
2601 if (optlen <= 0 || optlen > PAGE_SIZE)
2602 return -EMSGSIZE;
2603
2604 data = memdup_sockptr(optval, optlen);
2605 if (IS_ERR(data))
2606 return PTR_ERR(data);
2607
2608 if (in_compat_syscall()) {
2609 struct xfrm_translator *xtr = xfrm_get_translator();
2610
2611 if (!xtr) {
2612 kfree(data);
2613 return -EOPNOTSUPP;
2614 }
2615
2616 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2617 xfrm_put_translator(xtr);
2618 if (err) {
2619 kfree(data);
2620 return err;
2621 }
2622 }
2623
2624 err = -EINVAL;
2625 rcu_read_lock();
2626 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2627 pol = km->compile_policy(sk, optname, data,
2628 optlen, &err);
2629 if (err >= 0)
2630 break;
2631 }
2632 rcu_read_unlock();
2633
2634 if (err >= 0) {
2635 xfrm_sk_policy_insert(sk, err, pol);
2636 xfrm_pol_put(pol);
2637 __sk_dst_reset(sk);
2638 err = 0;
2639 }
2640
2641 kfree(data);
2642 return err;
2643}
2644EXPORT_SYMBOL(xfrm_user_policy);
2645
2646static DEFINE_SPINLOCK(xfrm_km_lock);
2647
2648void xfrm_register_km(struct xfrm_mgr *km)
2649{
2650 spin_lock_bh(&xfrm_km_lock);
2651 list_add_tail_rcu(&km->list, &xfrm_km_list);
2652 spin_unlock_bh(&xfrm_km_lock);
2653}
2654EXPORT_SYMBOL(xfrm_register_km);
2655
2656void xfrm_unregister_km(struct xfrm_mgr *km)
2657{
2658 spin_lock_bh(&xfrm_km_lock);
2659 list_del_rcu(&km->list);
2660 spin_unlock_bh(&xfrm_km_lock);
2661 synchronize_rcu();
2662}
2663EXPORT_SYMBOL(xfrm_unregister_km);
2664
2665int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2666{
2667 int err = 0;
2668
2669 if (WARN_ON(afinfo->family >= NPROTO))
2670 return -EAFNOSUPPORT;
2671
2672 spin_lock_bh(&xfrm_state_afinfo_lock);
2673 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2674 err = -EEXIST;
2675 else
2676 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2677 spin_unlock_bh(&xfrm_state_afinfo_lock);
2678 return err;
2679}
2680EXPORT_SYMBOL(xfrm_state_register_afinfo);
2681
2682int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2683{
2684 int err = 0, family = afinfo->family;
2685
2686 if (WARN_ON(family >= NPROTO))
2687 return -EAFNOSUPPORT;
2688
2689 spin_lock_bh(&xfrm_state_afinfo_lock);
2690 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2691 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2692 err = -EINVAL;
2693 else
2694 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2695 }
2696 spin_unlock_bh(&xfrm_state_afinfo_lock);
2697 synchronize_rcu();
2698 return err;
2699}
2700EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2701
2702struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2703{
2704 if (unlikely(family >= NPROTO))
2705 return NULL;
2706
2707 return rcu_dereference(xfrm_state_afinfo[family]);
2708}
2709EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2710
2711struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2712{
2713 struct xfrm_state_afinfo *afinfo;
2714 if (unlikely(family >= NPROTO))
2715 return NULL;
2716 rcu_read_lock();
2717 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2718 if (unlikely(!afinfo))
2719 rcu_read_unlock();
2720 return afinfo;
2721}
2722
2723void xfrm_flush_gc(void)
2724{
2725 flush_work(&xfrm_state_gc_work);
2726}
2727EXPORT_SYMBOL(xfrm_flush_gc);
2728
2729/* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2730void xfrm_state_delete_tunnel(struct xfrm_state *x)
2731{
2732 if (x->tunnel) {
2733 struct xfrm_state *t = x->tunnel;
2734
2735 if (atomic_read(&t->tunnel_users) == 2)
2736 xfrm_state_delete(t);
2737 atomic_dec(&t->tunnel_users);
2738 xfrm_state_put_sync(t);
2739 x->tunnel = NULL;
2740 }
2741}
2742EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2743
2744u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2745{
2746 const struct xfrm_type *type = READ_ONCE(x->type);
2747 struct crypto_aead *aead;
2748 u32 blksize, net_adj = 0;
2749
2750 if (x->km.state != XFRM_STATE_VALID ||
2751 !type || type->proto != IPPROTO_ESP)
2752 return mtu - x->props.header_len;
2753
2754 aead = x->data;
2755 blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2756
2757 switch (x->props.mode) {
2758 case XFRM_MODE_TRANSPORT:
2759 case XFRM_MODE_BEET:
2760 if (x->props.family == AF_INET)
2761 net_adj = sizeof(struct iphdr);
2762 else if (x->props.family == AF_INET6)
2763 net_adj = sizeof(struct ipv6hdr);
2764 break;
2765 case XFRM_MODE_TUNNEL:
2766 break;
2767 default:
2768 WARN_ON_ONCE(1);
2769 break;
2770 }
2771
2772 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2773 net_adj) & ~(blksize - 1)) + net_adj - 2;
2774}
2775EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2776
2777int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload,
2778 struct netlink_ext_ack *extack)
2779{
2780 const struct xfrm_mode *inner_mode;
2781 const struct xfrm_mode *outer_mode;
2782 int family = x->props.family;
2783 int err;
2784
2785 if (family == AF_INET &&
2786 READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
2787 x->props.flags |= XFRM_STATE_NOPMTUDISC;
2788
2789 err = -EPROTONOSUPPORT;
2790
2791 if (x->sel.family != AF_UNSPEC) {
2792 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2793 if (inner_mode == NULL) {
2794 NL_SET_ERR_MSG(extack, "Requested mode not found");
2795 goto error;
2796 }
2797
2798 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2799 family != x->sel.family) {
2800 NL_SET_ERR_MSG(extack, "Only tunnel modes can accommodate a change of family");
2801 goto error;
2802 }
2803
2804 x->inner_mode = *inner_mode;
2805 } else {
2806 const struct xfrm_mode *inner_mode_iaf;
2807 int iafamily = AF_INET;
2808
2809 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2810 if (inner_mode == NULL) {
2811 NL_SET_ERR_MSG(extack, "Requested mode not found");
2812 goto error;
2813 }
2814
2815 x->inner_mode = *inner_mode;
2816
2817 if (x->props.family == AF_INET)
2818 iafamily = AF_INET6;
2819
2820 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2821 if (inner_mode_iaf) {
2822 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2823 x->inner_mode_iaf = *inner_mode_iaf;
2824 }
2825 }
2826
2827 x->type = xfrm_get_type(x->id.proto, family);
2828 if (x->type == NULL) {
2829 NL_SET_ERR_MSG(extack, "Requested type not found");
2830 goto error;
2831 }
2832
2833 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2834
2835 err = x->type->init_state(x, extack);
2836 if (err)
2837 goto error;
2838
2839 outer_mode = xfrm_get_mode(x->props.mode, family);
2840 if (!outer_mode) {
2841 NL_SET_ERR_MSG(extack, "Requested mode not found");
2842 err = -EPROTONOSUPPORT;
2843 goto error;
2844 }
2845
2846 x->outer_mode = *outer_mode;
2847 if (init_replay) {
2848 err = xfrm_init_replay(x, extack);
2849 if (err)
2850 goto error;
2851 }
2852
2853error:
2854 return err;
2855}
2856
2857EXPORT_SYMBOL(__xfrm_init_state);
2858
2859int xfrm_init_state(struct xfrm_state *x)
2860{
2861 int err;
2862
2863 err = __xfrm_init_state(x, true, false, NULL);
2864 if (!err)
2865 x->km.state = XFRM_STATE_VALID;
2866
2867 return err;
2868}
2869
2870EXPORT_SYMBOL(xfrm_init_state);
2871
2872int __net_init xfrm_state_init(struct net *net)
2873{
2874 unsigned int sz;
2875
2876 if (net_eq(net, &init_net))
2877 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2878 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2879
2880 INIT_LIST_HEAD(&net->xfrm.state_all);
2881
2882 sz = sizeof(struct hlist_head) * 8;
2883
2884 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2885 if (!net->xfrm.state_bydst)
2886 goto out_bydst;
2887 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2888 if (!net->xfrm.state_bysrc)
2889 goto out_bysrc;
2890 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2891 if (!net->xfrm.state_byspi)
2892 goto out_byspi;
2893 net->xfrm.state_byseq = xfrm_hash_alloc(sz);
2894 if (!net->xfrm.state_byseq)
2895 goto out_byseq;
2896 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2897
2898 net->xfrm.state_num = 0;
2899 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2900 spin_lock_init(&net->xfrm.xfrm_state_lock);
2901 seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2902 &net->xfrm.xfrm_state_lock);
2903 return 0;
2904
2905out_byseq:
2906 xfrm_hash_free(net->xfrm.state_byspi, sz);
2907out_byspi:
2908 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2909out_bysrc:
2910 xfrm_hash_free(net->xfrm.state_bydst, sz);
2911out_bydst:
2912 return -ENOMEM;
2913}
2914
2915void xfrm_state_fini(struct net *net)
2916{
2917 unsigned int sz;
2918
2919 flush_work(&net->xfrm.state_hash_work);
2920 flush_work(&xfrm_state_gc_work);
2921 xfrm_state_flush(net, 0, false, true);
2922
2923 WARN_ON(!list_empty(&net->xfrm.state_all));
2924
2925 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2926 WARN_ON(!hlist_empty(net->xfrm.state_byseq));
2927 xfrm_hash_free(net->xfrm.state_byseq, sz);
2928 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2929 xfrm_hash_free(net->xfrm.state_byspi, sz);
2930 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2931 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2932 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2933 xfrm_hash_free(net->xfrm.state_bydst, sz);
2934}
2935
2936#ifdef CONFIG_AUDITSYSCALL
2937static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2938 struct audit_buffer *audit_buf)
2939{
2940 struct xfrm_sec_ctx *ctx = x->security;
2941 u32 spi = ntohl(x->id.spi);
2942
2943 if (ctx)
2944 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2945 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2946
2947 switch (x->props.family) {
2948 case AF_INET:
2949 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2950 &x->props.saddr.a4, &x->id.daddr.a4);
2951 break;
2952 case AF_INET6:
2953 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2954 x->props.saddr.a6, x->id.daddr.a6);
2955 break;
2956 }
2957
2958 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2959}
2960
2961static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2962 struct audit_buffer *audit_buf)
2963{
2964 const struct iphdr *iph4;
2965 const struct ipv6hdr *iph6;
2966
2967 switch (family) {
2968 case AF_INET:
2969 iph4 = ip_hdr(skb);
2970 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2971 &iph4->saddr, &iph4->daddr);
2972 break;
2973 case AF_INET6:
2974 iph6 = ipv6_hdr(skb);
2975 audit_log_format(audit_buf,
2976 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2977 &iph6->saddr, &iph6->daddr,
2978 iph6->flow_lbl[0] & 0x0f,
2979 iph6->flow_lbl[1],
2980 iph6->flow_lbl[2]);
2981 break;
2982 }
2983}
2984
2985void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2986{
2987 struct audit_buffer *audit_buf;
2988
2989 audit_buf = xfrm_audit_start("SAD-add");
2990 if (audit_buf == NULL)
2991 return;
2992 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2993 xfrm_audit_helper_sainfo(x, audit_buf);
2994 audit_log_format(audit_buf, " res=%u", result);
2995 audit_log_end(audit_buf);
2996}
2997EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2998
2999void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
3000{
3001 struct audit_buffer *audit_buf;
3002
3003 audit_buf = xfrm_audit_start("SAD-delete");
3004 if (audit_buf == NULL)
3005 return;
3006 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
3007 xfrm_audit_helper_sainfo(x, audit_buf);
3008 audit_log_format(audit_buf, " res=%u", result);
3009 audit_log_end(audit_buf);
3010}
3011EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
3012
3013void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
3014 struct sk_buff *skb)
3015{
3016 struct audit_buffer *audit_buf;
3017 u32 spi;
3018
3019 audit_buf = xfrm_audit_start("SA-replay-overflow");
3020 if (audit_buf == NULL)
3021 return;
3022 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3023 /* don't record the sequence number because it's inherent in this kind
3024 * of audit message */
3025 spi = ntohl(x->id.spi);
3026 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
3027 audit_log_end(audit_buf);
3028}
3029EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
3030
3031void xfrm_audit_state_replay(struct xfrm_state *x,
3032 struct sk_buff *skb, __be32 net_seq)
3033{
3034 struct audit_buffer *audit_buf;
3035 u32 spi;
3036
3037 audit_buf = xfrm_audit_start("SA-replayed-pkt");
3038 if (audit_buf == NULL)
3039 return;
3040 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3041 spi = ntohl(x->id.spi);
3042 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3043 spi, spi, ntohl(net_seq));
3044 audit_log_end(audit_buf);
3045}
3046EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
3047
3048void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
3049{
3050 struct audit_buffer *audit_buf;
3051
3052 audit_buf = xfrm_audit_start("SA-notfound");
3053 if (audit_buf == NULL)
3054 return;
3055 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3056 audit_log_end(audit_buf);
3057}
3058EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
3059
3060void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
3061 __be32 net_spi, __be32 net_seq)
3062{
3063 struct audit_buffer *audit_buf;
3064 u32 spi;
3065
3066 audit_buf = xfrm_audit_start("SA-notfound");
3067 if (audit_buf == NULL)
3068 return;
3069 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
3070 spi = ntohl(net_spi);
3071 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3072 spi, spi, ntohl(net_seq));
3073 audit_log_end(audit_buf);
3074}
3075EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
3076
3077void xfrm_audit_state_icvfail(struct xfrm_state *x,
3078 struct sk_buff *skb, u8 proto)
3079{
3080 struct audit_buffer *audit_buf;
3081 __be32 net_spi;
3082 __be32 net_seq;
3083
3084 audit_buf = xfrm_audit_start("SA-icv-failure");
3085 if (audit_buf == NULL)
3086 return;
3087 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
3088 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
3089 u32 spi = ntohl(net_spi);
3090 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
3091 spi, spi, ntohl(net_seq));
3092 }
3093 audit_log_end(audit_buf);
3094}
3095EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
3096#endif /* CONFIG_AUDITSYSCALL */
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * xfrm_state.c
4 *
5 * Changes:
6 * Mitsuru KANDA @USAGI
7 * Kazunori MIYAZAWA @USAGI
8 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
9 * IPv6 support
10 * YOSHIFUJI Hideaki @USAGI
11 * Split up af-specific functions
12 * Derek Atkins <derek@ihtfp.com>
13 * Add UDP Encapsulation
14 *
15 */
16
17#include <linux/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 */