projects / GitHub / MotorolaMobilityLLC / kernel-slsi.git / blob
? search:


7fbe478f83ca4adc6ffca52eb23610feca0a4c32
[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / net / xfrm / xfrm_state.c
1 /*
2 * xfrm_state.c
3 *
4 * Changes:
5 * Mitsuru KANDA @USAGI
6 * Kazunori MIYAZAWA @USAGI
7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com>
8 * IPv6 support
9 * YOSHIFUJI Hideaki @USAGI
10 * Split up af-specific functions
11 * Derek Atkins <derek@ihtfp.com>
12 * Add UDP Encapsulation
13 *
14 */
15
16 #include <linux/workqueue.h>
17 #include <net/xfrm.h>
18 #include <linux/pfkeyv2.h>
19 #include <linux/ipsec.h>
20 #include <linux/module.h>
21 #include <linux/cache.h>
22 #include <linux/audit.h>
23 #include <linux/uaccess.h>
24 #include <linux/ktime.h>
25 #include <linux/slab.h>
26 #include <linux/interrupt.h>
27 #include <linux/kernel.h>
28
29 #include "xfrm_hash.h"
30
31 #define xfrm_state_deref_prot(table, net) \
32 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
33
34 static void xfrm_state_gc_task(struct work_struct *work);
35
36 /* Each xfrm_state may be linked to two tables:
37
38 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
39 2. Hash table by (daddr,family,reqid) to find what SAs exist for given
40 destination/tunnel endpoint. (output)
41 */
42
43 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
44 static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
45
46 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
47 static HLIST_HEAD(xfrm_state_gc_list);
48
49 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
50 {
51 return refcount_inc_not_zero(&x->refcnt);
52 }
53
54 static inline unsigned int xfrm_dst_hash(struct net *net,
55 const xfrm_address_t *daddr,
56 const xfrm_address_t *saddr,
57 u32 reqid,
58 unsigned short family)
59 {
60 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
61 }
62
63 static inline unsigned int xfrm_src_hash(struct net *net,
64 const xfrm_address_t *daddr,
65 const xfrm_address_t *saddr,
66 unsigned short family)
67 {
68 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
69 }
70
71 static inline unsigned int
72 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
73 __be32 spi, u8 proto, unsigned short family)
74 {
75 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
76 }
77
78 static void xfrm_hash_transfer(struct hlist_head *list,
79 struct hlist_head *ndsttable,
80 struct hlist_head *nsrctable,
81 struct hlist_head *nspitable,
82 unsigned int nhashmask)
83 {
84 struct hlist_node *tmp;
85 struct xfrm_state *x;
86
87 hlist_for_each_entry_safe(x, tmp, list, bydst) {
88 unsigned int h;
89
90 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
91 x->props.reqid, x->props.family,
92 nhashmask);
93 hlist_add_head_rcu(&x->bydst, ndsttable + h);
94
95 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
96 x->props.family,
97 nhashmask);
98 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
99
100 if (x->id.spi) {
101 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
102 x->id.proto, x->props.family,
103 nhashmask);
104 hlist_add_head_rcu(&x->byspi, nspitable + h);
105 }
106 }
107 }
108
109 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
110 {
111 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
112 }
113
114 static void xfrm_hash_resize(struct work_struct *work)
115 {
116 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
117 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
118 unsigned long nsize, osize;
119 unsigned int nhashmask, ohashmask;
120 int i;
121
122 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
123 ndst = xfrm_hash_alloc(nsize);
124 if (!ndst)
125 return;
126 nsrc = xfrm_hash_alloc(nsize);
127 if (!nsrc) {
128 xfrm_hash_free(ndst, nsize);
129 return;
130 }
131 nspi = xfrm_hash_alloc(nsize);
132 if (!nspi) {
133 xfrm_hash_free(ndst, nsize);
134 xfrm_hash_free(nsrc, nsize);
135 return;
136 }
137
138 spin_lock_bh(&net->xfrm.xfrm_state_lock);
139 write_seqcount_begin(&xfrm_state_hash_generation);
140
141 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
142 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
143 for (i = net->xfrm.state_hmask; i >= 0; i--)
144 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
145
146 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
147 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
148 ohashmask = net->xfrm.state_hmask;
149
150 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
151 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
152 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
153 net->xfrm.state_hmask = nhashmask;
154
155 write_seqcount_end(&xfrm_state_hash_generation);
156 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
157
158 osize = (ohashmask + 1) * sizeof(struct hlist_head);
159
160 synchronize_rcu();
161
162 xfrm_hash_free(odst, osize);
163 xfrm_hash_free(osrc, osize);
164 xfrm_hash_free(ospi, osize);
165 }
166
167 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
168 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
169
170 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
171
172 int __xfrm_state_delete(struct xfrm_state *x);
173
174 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
175 bool km_is_alive(const struct km_event *c);
176 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
177
178 static DEFINE_SPINLOCK(xfrm_type_lock);
179 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
180 {
181 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
182 const struct xfrm_type **typemap;
183 int err = 0;
184
185 if (unlikely(afinfo == NULL))
186 return -EAFNOSUPPORT;
187 typemap = afinfo->type_map;
188 spin_lock_bh(&xfrm_type_lock);
189
190 if (likely(typemap[type->proto] == NULL))
191 typemap[type->proto] = type;
192 else
193 err = -EEXIST;
194 spin_unlock_bh(&xfrm_type_lock);
195 rcu_read_unlock();
196 return err;
197 }
198 EXPORT_SYMBOL(xfrm_register_type);
199
200 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
201 {
202 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
203 const struct xfrm_type **typemap;
204 int err = 0;
205
206 if (unlikely(afinfo == NULL))
207 return -EAFNOSUPPORT;
208 typemap = afinfo->type_map;
209 spin_lock_bh(&xfrm_type_lock);
210
211 if (unlikely(typemap[type->proto] != type))
212 err = -ENOENT;
213 else
214 typemap[type->proto] = NULL;
215 spin_unlock_bh(&xfrm_type_lock);
216 rcu_read_unlock();
217 return err;
218 }
219 EXPORT_SYMBOL(xfrm_unregister_type);
220
221 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
222 {
223 struct xfrm_state_afinfo *afinfo;
224 const struct xfrm_type **typemap;
225 const struct xfrm_type *type;
226 int modload_attempted = 0;
227
228 retry:
229 afinfo = xfrm_state_get_afinfo(family);
230 if (unlikely(afinfo == NULL))
231 return NULL;
232 typemap = afinfo->type_map;
233
234 type = READ_ONCE(typemap[proto]);
235 if (unlikely(type && !try_module_get(type->owner)))
236 type = NULL;
237
238 rcu_read_unlock();
239
240 if (!type && !modload_attempted) {
241 request_module("xfrm-type-%d-%d", family, proto);
242 modload_attempted = 1;
243 goto retry;
244 }
245
246 return type;
247 }
248
249 static void xfrm_put_type(const struct xfrm_type *type)
250 {
251 module_put(type->owner);
252 }
253
254 static DEFINE_SPINLOCK(xfrm_type_offload_lock);
255 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
256 unsigned short family)
257 {
258 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
259 const struct xfrm_type_offload **typemap;
260 int err = 0;
261
262 if (unlikely(afinfo == NULL))
263 return -EAFNOSUPPORT;
264 typemap = afinfo->type_offload_map;
265 spin_lock_bh(&xfrm_type_offload_lock);
266
267 if (likely(typemap[type->proto] == NULL))
268 typemap[type->proto] = type;
269 else
270 err = -EEXIST;
271 spin_unlock_bh(&xfrm_type_offload_lock);
272 rcu_read_unlock();
273 return err;
274 }
275 EXPORT_SYMBOL(xfrm_register_type_offload);
276
277 int xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
278 unsigned short family)
279 {
280 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
281 const struct xfrm_type_offload **typemap;
282 int err = 0;
283
284 if (unlikely(afinfo == NULL))
285 return -EAFNOSUPPORT;
286 typemap = afinfo->type_offload_map;
287 spin_lock_bh(&xfrm_type_offload_lock);
288
289 if (unlikely(typemap[type->proto] != type))
290 err = -ENOENT;
291 else
292 typemap[type->proto] = NULL;
293 spin_unlock_bh(&xfrm_type_offload_lock);
294 rcu_read_unlock();
295 return err;
296 }
297 EXPORT_SYMBOL(xfrm_unregister_type_offload);
298
299 static const struct xfrm_type_offload *
300 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
301 {
302 struct xfrm_state_afinfo *afinfo;
303 const struct xfrm_type_offload **typemap;
304 const struct xfrm_type_offload *type;
305
306 retry:
307 afinfo = xfrm_state_get_afinfo(family);
308 if (unlikely(afinfo == NULL))
309 return NULL;
310 typemap = afinfo->type_offload_map;
311
312 type = typemap[proto];
313 if ((type && !try_module_get(type->owner)))
314 type = NULL;
315
316 rcu_read_unlock();
317
318 if (!type && try_load) {
319 request_module("xfrm-offload-%d-%d", family, proto);
320 try_load = 0;
321 goto retry;
322 }
323
324 return type;
325 }
326
327 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
328 {
329 module_put(type->owner);
330 }
331
332 static DEFINE_SPINLOCK(xfrm_mode_lock);
333 int xfrm_register_mode(struct xfrm_mode *mode, int family)
334 {
335 struct xfrm_state_afinfo *afinfo;
336 struct xfrm_mode **modemap;
337 int err;
338
339 if (unlikely(mode->encap >= XFRM_MODE_MAX))
340 return -EINVAL;
341
342 afinfo = xfrm_state_get_afinfo(family);
343 if (unlikely(afinfo == NULL))
344 return -EAFNOSUPPORT;
345
346 err = -EEXIST;
347 modemap = afinfo->mode_map;
348 spin_lock_bh(&xfrm_mode_lock);
349 if (modemap[mode->encap])
350 goto out;
351
352 err = -ENOENT;
353 if (!try_module_get(afinfo->owner))
354 goto out;
355
356 mode->afinfo = afinfo;
357 modemap[mode->encap] = mode;
358 err = 0;
359
360 out:
361 spin_unlock_bh(&xfrm_mode_lock);
362 rcu_read_unlock();
363 return err;
364 }
365 EXPORT_SYMBOL(xfrm_register_mode);
366
367 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
368 {
369 struct xfrm_state_afinfo *afinfo;
370 struct xfrm_mode **modemap;
371 int err;
372
373 if (unlikely(mode->encap >= XFRM_MODE_MAX))
374 return -EINVAL;
375
376 afinfo = xfrm_state_get_afinfo(family);
377 if (unlikely(afinfo == NULL))
378 return -EAFNOSUPPORT;
379
380 err = -ENOENT;
381 modemap = afinfo->mode_map;
382 spin_lock_bh(&xfrm_mode_lock);
383 if (likely(modemap[mode->encap] == mode)) {
384 modemap[mode->encap] = NULL;
385 module_put(mode->afinfo->owner);
386 err = 0;
387 }
388
389 spin_unlock_bh(&xfrm_mode_lock);
390 rcu_read_unlock();
391 return err;
392 }
393 EXPORT_SYMBOL(xfrm_unregister_mode);
394
395 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
396 {
397 struct xfrm_state_afinfo *afinfo;
398 struct xfrm_mode *mode;
399 int modload_attempted = 0;
400
401 if (unlikely(encap >= XFRM_MODE_MAX))
402 return NULL;
403
404 retry:
405 afinfo = xfrm_state_get_afinfo(family);
406 if (unlikely(afinfo == NULL))
407 return NULL;
408
409 mode = READ_ONCE(afinfo->mode_map[encap]);
410 if (unlikely(mode && !try_module_get(mode->owner)))
411 mode = NULL;
412
413 rcu_read_unlock();
414 if (!mode && !modload_attempted) {
415 request_module("xfrm-mode-%d-%d", family, encap);
416 modload_attempted = 1;
417 goto retry;
418 }
419
420 return mode;
421 }
422
423 static void xfrm_put_mode(struct xfrm_mode *mode)
424 {
425 module_put(mode->owner);
426 }
427
428 static void xfrm_state_gc_destroy(struct xfrm_state *x)
429 {
430 tasklet_hrtimer_cancel(&x->mtimer);
431 del_timer_sync(&x->rtimer);
432 kfree(x->aead);
433 kfree(x->aalg);
434 kfree(x->ealg);
435 kfree(x->calg);
436 kfree(x->encap);
437 kfree(x->coaddr);
438 kfree(x->replay_esn);
439 kfree(x->preplay_esn);
440 if (x->inner_mode)
441 xfrm_put_mode(x->inner_mode);
442 if (x->inner_mode_iaf)
443 xfrm_put_mode(x->inner_mode_iaf);
444 if (x->outer_mode)
445 xfrm_put_mode(x->outer_mode);
446 if (x->type_offload)
447 xfrm_put_type_offload(x->type_offload);
448 if (x->type) {
449 x->type->destructor(x);
450 xfrm_put_type(x->type);
451 }
452 xfrm_dev_state_free(x);
453 security_xfrm_state_free(x);
454 kfree(x);
455 }
456
457 static void xfrm_state_gc_task(struct work_struct *work)
458 {
459 struct xfrm_state *x;
460 struct hlist_node *tmp;
461 struct hlist_head gc_list;
462
463 spin_lock_bh(&xfrm_state_gc_lock);
464 hlist_move_list(&xfrm_state_gc_list, &gc_list);
465 spin_unlock_bh(&xfrm_state_gc_lock);
466
467 synchronize_rcu();
468
469 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
470 xfrm_state_gc_destroy(x);
471 }
472
473 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
474 {
475 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
476 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
477 unsigned long now = get_seconds();
478 long next = LONG_MAX;
479 int warn = 0;
480 int err = 0;
481
482 spin_lock(&x->lock);
483 if (x->km.state == XFRM_STATE_DEAD)
484 goto out;
485 if (x->km.state == XFRM_STATE_EXPIRED)
486 goto expired;
487 if (x->lft.hard_add_expires_seconds) {
488 long tmo = x->lft.hard_add_expires_seconds +
489 x->curlft.add_time - now;
490 if (tmo <= 0) {
491 if (x->xflags & XFRM_SOFT_EXPIRE) {
492 /* enter hard expire without soft expire first?!
493 * setting a new date could trigger this.
494 * workaround: fix x->curflt.add_time by below:
495 */
496 x->curlft.add_time = now - x->saved_tmo - 1;
497 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
498 } else
499 goto expired;
500 }
501 if (tmo < next)
502 next = tmo;
503 }
504 if (x->lft.hard_use_expires_seconds) {
505 long tmo = x->lft.hard_use_expires_seconds +
506 (x->curlft.use_time ? : now) - now;
507 if (tmo <= 0)
508 goto expired;
509 if (tmo < next)
510 next = tmo;
511 }
512 if (x->km.dying)
513 goto resched;
514 if (x->lft.soft_add_expires_seconds) {
515 long tmo = x->lft.soft_add_expires_seconds +
516 x->curlft.add_time - now;
517 if (tmo <= 0) {
518 warn = 1;
519 x->xflags &= ~XFRM_SOFT_EXPIRE;
520 } else if (tmo < next) {
521 next = tmo;
522 x->xflags |= XFRM_SOFT_EXPIRE;
523 x->saved_tmo = tmo;
524 }
525 }
526 if (x->lft.soft_use_expires_seconds) {
527 long tmo = x->lft.soft_use_expires_seconds +
528 (x->curlft.use_time ? : now) - now;
529 if (tmo <= 0)
530 warn = 1;
531 else if (tmo < next)
532 next = tmo;
533 }
534
535 x->km.dying = warn;
536 if (warn)
537 km_state_expired(x, 0, 0);
538 resched:
539 if (next != LONG_MAX) {
540 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
541 }
542
543 goto out;
544
545 expired:
546 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
547 x->km.state = XFRM_STATE_EXPIRED;
548
549 err = __xfrm_state_delete(x);
550 if (!err)
551 km_state_expired(x, 1, 0);
552
553 xfrm_audit_state_delete(x, err ? 0 : 1, true);
554
555 out:
556 spin_unlock(&x->lock);
557 return HRTIMER_NORESTART;
558 }
559
560 static void xfrm_replay_timer_handler(unsigned long data);
561
562 struct xfrm_state *xfrm_state_alloc(struct net *net)
563 {
564 struct xfrm_state *x;
565
566 x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
567
568 if (x) {
569 write_pnet(&x->xs_net, net);
570 refcount_set(&x->refcnt, 1);
571 atomic_set(&x->tunnel_users, 0);
572 INIT_LIST_HEAD(&x->km.all);
573 INIT_HLIST_NODE(&x->bydst);
574 INIT_HLIST_NODE(&x->bysrc);
575 INIT_HLIST_NODE(&x->byspi);
576 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
577 CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
578 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
579 (unsigned long)x);
580 x->curlft.add_time = get_seconds();
581 x->lft.soft_byte_limit = XFRM_INF;
582 x->lft.soft_packet_limit = XFRM_INF;
583 x->lft.hard_byte_limit = XFRM_INF;
584 x->lft.hard_packet_limit = XFRM_INF;
585 x->replay_maxage = 0;
586 x->replay_maxdiff = 0;
587 x->inner_mode = NULL;
588 x->inner_mode_iaf = NULL;
589 spin_lock_init(&x->lock);
590 }
591 return x;
592 }
593 EXPORT_SYMBOL(xfrm_state_alloc);
594
595 void __xfrm_state_destroy(struct xfrm_state *x)
596 {
597 WARN_ON(x->km.state != XFRM_STATE_DEAD);
598
599 spin_lock_bh(&xfrm_state_gc_lock);
600 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
601 spin_unlock_bh(&xfrm_state_gc_lock);
602 schedule_work(&xfrm_state_gc_work);
603 }
604 EXPORT_SYMBOL(__xfrm_state_destroy);
605
606 int __xfrm_state_delete(struct xfrm_state *x)
607 {
608 struct net *net = xs_net(x);
609 int err = -ESRCH;
610
611 if (x->km.state != XFRM_STATE_DEAD) {
612 x->km.state = XFRM_STATE_DEAD;
613 spin_lock(&net->xfrm.xfrm_state_lock);
614 list_del(&x->km.all);
615 hlist_del_rcu(&x->bydst);
616 hlist_del_rcu(&x->bysrc);
617 if (x->id.spi)
618 hlist_del_rcu(&x->byspi);
619 net->xfrm.state_num--;
620 spin_unlock(&net->xfrm.xfrm_state_lock);
621
622 xfrm_dev_state_delete(x);
623
624 /* All xfrm_state objects are created by xfrm_state_alloc.
625 * The xfrm_state_alloc call gives a reference, and that
626 * is what we are dropping here.
627 */
628 xfrm_state_put(x);
629 err = 0;
630 }
631
632 return err;
633 }
634 EXPORT_SYMBOL(__xfrm_state_delete);
635
636 int xfrm_state_delete(struct xfrm_state *x)
637 {
638 int err;
639
640 spin_lock_bh(&x->lock);
641 err = __xfrm_state_delete(x);
642 spin_unlock_bh(&x->lock);
643
644 return err;
645 }
646 EXPORT_SYMBOL(xfrm_state_delete);
647
648 #ifdef CONFIG_SECURITY_NETWORK_XFRM
649 static inline int
650 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
651 {
652 int i, err = 0;
653
654 for (i = 0; i <= net->xfrm.state_hmask; i++) {
655 struct xfrm_state *x;
656
657 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
658 if (xfrm_id_proto_match(x->id.proto, proto) &&
659 (err = security_xfrm_state_delete(x)) != 0) {
660 xfrm_audit_state_delete(x, 0, task_valid);
661 return err;
662 }
663 }
664 }
665
666 return err;
667 }
668
669 static inline int
670 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
671 {
672 int i, err = 0;
673
674 for (i = 0; i <= net->xfrm.state_hmask; i++) {
675 struct xfrm_state *x;
676 struct xfrm_state_offload *xso;
677
678 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
679 xso = &x->xso;
680
681 if (xso->dev == dev &&
682 (err = security_xfrm_state_delete(x)) != 0) {
683 xfrm_audit_state_delete(x, 0, task_valid);
684 return err;
685 }
686 }
687 }
688
689 return err;
690 }
691 #else
692 static inline int
693 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
694 {
695 return 0;
696 }
697
698 static inline int
699 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
700 {
701 return 0;
702 }
703 #endif
704
705 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
706 {
707 int i, err = 0, cnt = 0;
708
709 spin_lock_bh(&net->xfrm.xfrm_state_lock);
710 err = xfrm_state_flush_secctx_check(net, proto, task_valid);
711 if (err)
712 goto out;
713
714 err = -ESRCH;
715 for (i = 0; i <= net->xfrm.state_hmask; i++) {
716 struct xfrm_state *x;
717 restart:
718 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
719 if (!xfrm_state_kern(x) &&
720 xfrm_id_proto_match(x->id.proto, proto)) {
721 xfrm_state_hold(x);
722 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
723
724 err = xfrm_state_delete(x);
725 xfrm_audit_state_delete(x, err ? 0 : 1,
726 task_valid);
727 xfrm_state_put(x);
728 if (!err)
729 cnt++;
730
731 spin_lock_bh(&net->xfrm.xfrm_state_lock);
732 goto restart;
733 }
734 }
735 }
736 out:
737 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
738 if (cnt) {
739 err = 0;
740 xfrm_policy_cache_flush();
741 }
742 return err;
743 }
744 EXPORT_SYMBOL(xfrm_state_flush);
745
746 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
747 {
748 int i, err = 0, cnt = 0;
749
750 spin_lock_bh(&net->xfrm.xfrm_state_lock);
751 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
752 if (err)
753 goto out;
754
755 err = -ESRCH;
756 for (i = 0; i <= net->xfrm.state_hmask; i++) {
757 struct xfrm_state *x;
758 struct xfrm_state_offload *xso;
759 restart:
760 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
761 xso = &x->xso;
762
763 if (!xfrm_state_kern(x) && xso->dev == dev) {
764 xfrm_state_hold(x);
765 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
766
767 err = xfrm_state_delete(x);
768 xfrm_audit_state_delete(x, err ? 0 : 1,
769 task_valid);
770 xfrm_state_put(x);
771 if (!err)
772 cnt++;
773
774 spin_lock_bh(&net->xfrm.xfrm_state_lock);
775 goto restart;
776 }
777 }
778 }
779 if (cnt)
780 err = 0;
781
782 out:
783 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
784 return err;
785 }
786 EXPORT_SYMBOL(xfrm_dev_state_flush);
787
788 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
789 {
790 spin_lock_bh(&net->xfrm.xfrm_state_lock);
791 si->sadcnt = net->xfrm.state_num;
792 si->sadhcnt = net->xfrm.state_hmask + 1;
793 si->sadhmcnt = xfrm_state_hashmax;
794 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
795 }
796 EXPORT_SYMBOL(xfrm_sad_getinfo);
797
798 static void
799 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
800 const struct xfrm_tmpl *tmpl,
801 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
802 unsigned short family)
803 {
804 struct xfrm_state_afinfo *afinfo = xfrm_state_afinfo_get_rcu(family);
805
806 if (!afinfo)
807 return;
808
809 afinfo->init_tempsel(&x->sel, fl);
810
811 if (family != tmpl->encap_family) {
812 afinfo = xfrm_state_afinfo_get_rcu(tmpl->encap_family);
813 if (!afinfo)
814 return;
815 }
816 afinfo->init_temprop(x, tmpl, daddr, saddr);
817 }
818
819 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
820 const xfrm_address_t *daddr,
821 __be32 spi, u8 proto,
822 unsigned short family)
823 {
824 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
825 struct xfrm_state *x;
826
827 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
828 if (x->props.family != family ||
829 x->id.spi != spi ||
830 x->id.proto != proto ||
831 !xfrm_addr_equal(&x->id.daddr, daddr, family))
832 continue;
833
834 if ((mark & x->mark.m) != x->mark.v)
835 continue;
836 if (!xfrm_state_hold_rcu(x))
837 continue;
838 return x;
839 }
840
841 return NULL;
842 }
843
844 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
845 const xfrm_address_t *daddr,
846 const xfrm_address_t *saddr,
847 u8 proto, unsigned short family)
848 {
849 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
850 struct xfrm_state *x;
851
852 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
853 if (x->props.family != family ||
854 x->id.proto != proto ||
855 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
856 !xfrm_addr_equal(&x->props.saddr, saddr, family))
857 continue;
858
859 if ((mark & x->mark.m) != x->mark.v)
860 continue;
861 if (!xfrm_state_hold_rcu(x))
862 continue;
863 return x;
864 }
865
866 return NULL;
867 }
868
869 static inline struct xfrm_state *
870 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
871 {
872 struct net *net = xs_net(x);
873 u32 mark = x->mark.v & x->mark.m;
874
875 if (use_spi)
876 return __xfrm_state_lookup(net, mark, &x->id.daddr,
877 x->id.spi, x->id.proto, family);
878 else
879 return __xfrm_state_lookup_byaddr(net, mark,
880 &x->id.daddr,
881 &x->props.saddr,
882 x->id.proto, family);
883 }
884
885 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
886 {
887 if (have_hash_collision &&
888 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
889 net->xfrm.state_num > net->xfrm.state_hmask)
890 schedule_work(&net->xfrm.state_hash_work);
891 }
892
893 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
894 const struct flowi *fl, unsigned short family,
895 struct xfrm_state **best, int *acq_in_progress,
896 int *error)
897 {
898 /* Resolution logic:
899 * 1. There is a valid state with matching selector. Done.
900 * 2. Valid state with inappropriate selector. Skip.
901 *
902 * Entering area of "sysdeps".
903 *
904 * 3. If state is not valid, selector is temporary, it selects
905 * only session which triggered previous resolution. Key
906 * manager will do something to install a state with proper
907 * selector.
908 */
909 if (x->km.state == XFRM_STATE_VALID) {
910 if ((x->sel.family &&
911 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
912 !security_xfrm_state_pol_flow_match(x, pol, fl))
913 return;
914
915 if (!*best ||
916 (*best)->km.dying > x->km.dying ||
917 ((*best)->km.dying == x->km.dying &&
918 (*best)->curlft.add_time < x->curlft.add_time))
919 *best = x;
920 } else if (x->km.state == XFRM_STATE_ACQ) {
921 *acq_in_progress = 1;
922 } else if (x->km.state == XFRM_STATE_ERROR ||
923 x->km.state == XFRM_STATE_EXPIRED) {
924 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
925 security_xfrm_state_pol_flow_match(x, pol, fl))
926 *error = -ESRCH;
927 }
928 }
929
930 struct xfrm_state *
931 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
932 const struct flowi *fl, struct xfrm_tmpl *tmpl,
933 struct xfrm_policy *pol, int *err,
934 unsigned short family, u32 if_id)
935 {
936 static xfrm_address_t saddr_wildcard = { };
937 struct net *net = xp_net(pol);
938 unsigned int h, h_wildcard;
939 struct xfrm_state *x, *x0, *to_put;
940 int acquire_in_progress = 0;
941 int error = 0;
942 struct xfrm_state *best = NULL;
943 u32 mark = pol->mark.v & pol->mark.m;
944 unsigned short encap_family = tmpl->encap_family;
945 unsigned int sequence;
946 struct km_event c;
947
948 to_put = NULL;
949
950 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
951
952 rcu_read_lock();
953 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
954 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
955 if (x->props.family == encap_family &&
956 x->props.reqid == tmpl->reqid &&
957 (mark & x->mark.m) == x->mark.v &&
958 x->if_id == if_id &&
959 !(x->props.flags & XFRM_STATE_WILDRECV) &&
960 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
961 tmpl->mode == x->props.mode &&
962 tmpl->id.proto == x->id.proto &&
963 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
964 xfrm_state_look_at(pol, x, fl, encap_family,
965 &best, &acquire_in_progress, &error);
966 }
967 if (best || acquire_in_progress)
968 goto found;
969
970 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
971 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
972 if (x->props.family == encap_family &&
973 x->props.reqid == tmpl->reqid &&
974 (mark & x->mark.m) == x->mark.v &&
975 x->if_id == if_id &&
976 !(x->props.flags & XFRM_STATE_WILDRECV) &&
977 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
978 tmpl->mode == x->props.mode &&
979 tmpl->id.proto == x->id.proto &&
980 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
981 xfrm_state_look_at(pol, x, fl, encap_family,
982 &best, &acquire_in_progress, &error);
983 }
984
985 found:
986 x = best;
987 if (!x && !error && !acquire_in_progress) {
988 if (tmpl->id.spi &&
989 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
990 tmpl->id.proto, encap_family)) != NULL) {
991 to_put = x0;
992 error = -EEXIST;
993 goto out;
994 }
995
996 c.net = net;
997 /* If the KMs have no listeners (yet...), avoid allocating an SA
998 * for each and every packet - garbage collection might not
999 * handle the flood.
1000 */
1001 if (!km_is_alive(&c)) {
1002 error = -ESRCH;
1003 goto out;
1004 }
1005
1006 x = xfrm_state_alloc(net);
1007 if (x == NULL) {
1008 error = -ENOMEM;
1009 goto out;
1010 }
1011 /* Initialize temporary state matching only
1012 * to current session. */
1013 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1014 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1015 x->if_id = if_id;
1016
1017 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1018 if (error) {
1019 x->km.state = XFRM_STATE_DEAD;
1020 to_put = x;
1021 x = NULL;
1022 goto out;
1023 }
1024
1025 if (km_query(x, tmpl, pol) == 0) {
1026 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1027 x->km.state = XFRM_STATE_ACQ;
1028 list_add(&x->km.all, &net->xfrm.state_all);
1029 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1030 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1031 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1032 if (x->id.spi) {
1033 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1034 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1035 }
1036 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1037 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1038 net->xfrm.state_num++;
1039 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1040 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1041 } else {
1042 x->km.state = XFRM_STATE_DEAD;
1043 to_put = x;
1044 x = NULL;
1045 error = -ESRCH;
1046 }
1047 }
1048 out:
1049 if (x) {
1050 if (!xfrm_state_hold_rcu(x)) {
1051 *err = -EAGAIN;
1052 x = NULL;
1053 }
1054 } else {
1055 *err = acquire_in_progress ? -EAGAIN : error;
1056 }
1057 rcu_read_unlock();
1058 if (to_put)
1059 xfrm_state_put(to_put);
1060
1061 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1062 *err = -EAGAIN;
1063 if (x) {
1064 xfrm_state_put(x);
1065 x = NULL;
1066 }
1067 }
1068
1069 return x;
1070 }
1071
1072 struct xfrm_state *
1073 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1074 xfrm_address_t *daddr, xfrm_address_t *saddr,
1075 unsigned short family, u8 mode, u8 proto, u32 reqid)
1076 {
1077 unsigned int h;
1078 struct xfrm_state *rx = NULL, *x = NULL;
1079
1080 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1081 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1082 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1083 if (x->props.family == family &&
1084 x->props.reqid == reqid &&
1085 (mark & x->mark.m) == x->mark.v &&
1086 x->if_id == if_id &&
1087 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1088 xfrm_state_addr_check(x, daddr, saddr, family) &&
1089 mode == x->props.mode &&
1090 proto == x->id.proto &&
1091 x->km.state == XFRM_STATE_VALID) {
1092 rx = x;
1093 break;
1094 }
1095 }
1096
1097 if (rx)
1098 xfrm_state_hold(rx);
1099 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1100
1101
1102 return rx;
1103 }
1104 EXPORT_SYMBOL(xfrm_stateonly_find);
1105
1106 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1107 unsigned short family)
1108 {
1109 struct xfrm_state *x;
1110 struct xfrm_state_walk *w;
1111
1112 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1113 list_for_each_entry(w, &net->xfrm.state_all, all) {
1114 x = container_of(w, struct xfrm_state, km);
1115 if (x->props.family != family ||
1116 x->id.spi != spi)
1117 continue;
1118
1119 xfrm_state_hold(x);
1120 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1121 return x;
1122 }
1123 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1124 return NULL;
1125 }
1126 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1127
1128 static void __xfrm_state_insert(struct xfrm_state *x)
1129 {
1130 struct net *net = xs_net(x);
1131 unsigned int h;
1132
1133 list_add(&x->km.all, &net->xfrm.state_all);
1134
1135 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1136 x->props.reqid, x->props.family);
1137 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1138
1139 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1140 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1141
1142 if (x->id.spi) {
1143 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1144 x->props.family);
1145
1146 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1147 }
1148
1149 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1150 if (x->replay_maxage)
1151 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1152
1153 net->xfrm.state_num++;
1154
1155 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1156 }
1157
1158 /* net->xfrm.xfrm_state_lock is held */
1159 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1160 {
1161 struct net *net = xs_net(xnew);
1162 unsigned short family = xnew->props.family;
1163 u32 reqid = xnew->props.reqid;
1164 struct xfrm_state *x;
1165 unsigned int h;
1166 u32 mark = xnew->mark.v & xnew->mark.m;
1167 u32 if_id = xnew->if_id;
1168
1169 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1170 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1171 if (x->props.family == family &&
1172 x->props.reqid == reqid &&
1173 x->if_id == if_id &&
1174 (mark & x->mark.m) == x->mark.v &&
1175 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1176 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1177 x->genid++;
1178 }
1179 }
1180
1181 void xfrm_state_insert(struct xfrm_state *x)
1182 {
1183 struct net *net = xs_net(x);
1184
1185 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1186 __xfrm_state_bump_genids(x);
1187 __xfrm_state_insert(x);
1188 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1189 }
1190 EXPORT_SYMBOL(xfrm_state_insert);
1191
1192 /* net->xfrm.xfrm_state_lock is held */
1193 static struct xfrm_state *__find_acq_core(struct net *net,
1194 const struct xfrm_mark *m,
1195 unsigned short family, u8 mode,
1196 u32 reqid, u32 if_id, u8 proto,
1197 const xfrm_address_t *daddr,
1198 const xfrm_address_t *saddr,
1199 int create)
1200 {
1201 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1202 struct xfrm_state *x;
1203 u32 mark = m->v & m->m;
1204
1205 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1206 if (x->props.reqid != reqid ||
1207 x->props.mode != mode ||
1208 x->props.family != family ||
1209 x->km.state != XFRM_STATE_ACQ ||
1210 x->id.spi != 0 ||
1211 x->id.proto != proto ||
1212 (mark & x->mark.m) != x->mark.v ||
1213 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1214 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1215 continue;
1216
1217 xfrm_state_hold(x);
1218 return x;
1219 }
1220
1221 if (!create)
1222 return NULL;
1223
1224 x = xfrm_state_alloc(net);
1225 if (likely(x)) {
1226 switch (family) {
1227 case AF_INET:
1228 x->sel.daddr.a4 = daddr->a4;
1229 x->sel.saddr.a4 = saddr->a4;
1230 x->sel.prefixlen_d = 32;
1231 x->sel.prefixlen_s = 32;
1232 x->props.saddr.a4 = saddr->a4;
1233 x->id.daddr.a4 = daddr->a4;
1234 break;
1235
1236 case AF_INET6:
1237 x->sel.daddr.in6 = daddr->in6;
1238 x->sel.saddr.in6 = saddr->in6;
1239 x->sel.prefixlen_d = 128;
1240 x->sel.prefixlen_s = 128;
1241 x->props.saddr.in6 = saddr->in6;
1242 x->id.daddr.in6 = daddr->in6;
1243 break;
1244 }
1245
1246 x->km.state = XFRM_STATE_ACQ;
1247 x->id.proto = proto;
1248 x->props.family = family;
1249 x->props.mode = mode;
1250 x->props.reqid = reqid;
1251 x->if_id = if_id;
1252 x->mark.v = m->v;
1253 x->mark.m = m->m;
1254 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1255 xfrm_state_hold(x);
1256 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1257 list_add(&x->km.all, &net->xfrm.state_all);
1258 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1259 h = xfrm_src_hash(net, daddr, saddr, family);
1260 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1261
1262 net->xfrm.state_num++;
1263
1264 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1265 }
1266
1267 return x;
1268 }
1269
1270 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1271
1272 int xfrm_state_add(struct xfrm_state *x)
1273 {
1274 struct net *net = xs_net(x);
1275 struct xfrm_state *x1, *to_put;
1276 int family;
1277 int err;
1278 u32 mark = x->mark.v & x->mark.m;
1279 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1280
1281 family = x->props.family;
1282
1283 to_put = NULL;
1284
1285 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1286
1287 x1 = __xfrm_state_locate(x, use_spi, family);
1288 if (x1) {
1289 to_put = x1;
1290 x1 = NULL;
1291 err = -EEXIST;
1292 goto out;
1293 }
1294
1295 if (use_spi && x->km.seq) {
1296 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1297 if (x1 && ((x1->id.proto != x->id.proto) ||
1298 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1299 to_put = x1;
1300 x1 = NULL;
1301 }
1302 }
1303
1304 if (use_spi && !x1)
1305 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1306 x->props.reqid, x->if_id, x->id.proto,
1307 &x->id.daddr, &x->props.saddr, 0);
1308
1309 __xfrm_state_bump_genids(x);
1310 __xfrm_state_insert(x);
1311 err = 0;
1312
1313 out:
1314 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1315
1316 if (x1) {
1317 xfrm_state_delete(x1);
1318 xfrm_state_put(x1);
1319 }
1320
1321 if (to_put)
1322 xfrm_state_put(to_put);
1323
1324 return err;
1325 }
1326 EXPORT_SYMBOL(xfrm_state_add);
1327
1328 #ifdef CONFIG_XFRM_MIGRATE
1329 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1330 struct xfrm_encap_tmpl *encap)
1331 {
1332 struct net *net = xs_net(orig);
1333 struct xfrm_state *x = xfrm_state_alloc(net);
1334 if (!x)
1335 goto out;
1336
1337 memcpy(&x->id, &orig->id, sizeof(x->id));
1338 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1339 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1340 x->props.mode = orig->props.mode;
1341 x->props.replay_window = orig->props.replay_window;
1342 x->props.reqid = orig->props.reqid;
1343 x->props.family = orig->props.family;
1344 x->props.saddr = orig->props.saddr;
1345
1346 if (orig->aalg) {
1347 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1348 if (!x->aalg)
1349 goto error;
1350 }
1351 x->props.aalgo = orig->props.aalgo;
1352
1353 if (orig->aead) {
1354 x->aead = xfrm_algo_aead_clone(orig->aead);
1355 x->geniv = orig->geniv;
1356 if (!x->aead)
1357 goto error;
1358 }
1359 if (orig->ealg) {
1360 x->ealg = xfrm_algo_clone(orig->ealg);
1361 if (!x->ealg)
1362 goto error;
1363 }
1364 x->props.ealgo = orig->props.ealgo;
1365
1366 if (orig->calg) {
1367 x->calg = xfrm_algo_clone(orig->calg);
1368 if (!x->calg)
1369 goto error;
1370 }
1371 x->props.calgo = orig->props.calgo;
1372
1373 if (encap || orig->encap) {
1374 if (encap)
1375 x->encap = kmemdup(encap, sizeof(*x->encap),
1376 GFP_KERNEL);
1377 else
1378 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1379 GFP_KERNEL);
1380
1381 if (!x->encap)
1382 goto error;
1383 }
1384
1385 if (orig->coaddr) {
1386 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1387 GFP_KERNEL);
1388 if (!x->coaddr)
1389 goto error;
1390 }
1391
1392 if (orig->replay_esn) {
1393 if (xfrm_replay_clone(x, orig))
1394 goto error;
1395 }
1396
1397 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1398
1399 if (xfrm_init_state(x) < 0)
1400 goto error;
1401
1402 x->props.flags = orig->props.flags;
1403 x->props.extra_flags = orig->props.extra_flags;
1404
1405 x->if_id = orig->if_id;
1406 x->tfcpad = orig->tfcpad;
1407 x->replay_maxdiff = orig->replay_maxdiff;
1408 x->replay_maxage = orig->replay_maxage;
1409 x->curlft.add_time = orig->curlft.add_time;
1410 x->km.state = orig->km.state;
1411 x->km.seq = orig->km.seq;
1412 x->replay = orig->replay;
1413 x->preplay = orig->preplay;
1414
1415 return x;
1416
1417 error:
1418 xfrm_state_put(x);
1419 out:
1420 return NULL;
1421 }
1422
1423 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1424 {
1425 unsigned int h;
1426 struct xfrm_state *x = NULL;
1427
1428 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1429
1430 if (m->reqid) {
1431 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1432 m->reqid, m->old_family);
1433 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1434 if (x->props.mode != m->mode ||
1435 x->id.proto != m->proto)
1436 continue;
1437 if (m->reqid && x->props.reqid != m->reqid)
1438 continue;
1439 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1440 m->old_family) ||
1441 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1442 m->old_family))
1443 continue;
1444 xfrm_state_hold(x);
1445 break;
1446 }
1447 } else {
1448 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1449 m->old_family);
1450 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1451 if (x->props.mode != m->mode ||
1452 x->id.proto != m->proto)
1453 continue;
1454 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1455 m->old_family) ||
1456 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1457 m->old_family))
1458 continue;
1459 xfrm_state_hold(x);
1460 break;
1461 }
1462 }
1463
1464 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1465
1466 return x;
1467 }
1468 EXPORT_SYMBOL(xfrm_migrate_state_find);
1469
1470 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1471 struct xfrm_migrate *m,
1472 struct xfrm_encap_tmpl *encap)
1473 {
1474 struct xfrm_state *xc;
1475
1476 xc = xfrm_state_clone(x, encap);
1477 if (!xc)
1478 return NULL;
1479
1480 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1481 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1482
1483 /* add state */
1484 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1485 /* a care is needed when the destination address of the
1486 state is to be updated as it is a part of triplet */
1487 xfrm_state_insert(xc);
1488 } else {
1489 if (xfrm_state_add(xc) < 0)
1490 goto error;
1491 }
1492
1493 return xc;
1494 error:
1495 xfrm_state_put(xc);
1496 return NULL;
1497 }
1498 EXPORT_SYMBOL(xfrm_state_migrate);
1499 #endif
1500
1501 int xfrm_state_update(struct xfrm_state *x)
1502 {
1503 struct xfrm_state *x1, *to_put;
1504 int err;
1505 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1506 struct net *net = xs_net(x);
1507
1508 to_put = NULL;
1509
1510 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1511 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1512
1513 err = -ESRCH;
1514 if (!x1)
1515 goto out;
1516
1517 if (xfrm_state_kern(x1)) {
1518 to_put = x1;
1519 err = -EEXIST;
1520 goto out;
1521 }
1522
1523 if (x1->km.state == XFRM_STATE_ACQ) {
1524 __xfrm_state_insert(x);
1525 x = NULL;
1526 }
1527 err = 0;
1528
1529 out:
1530 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1531
1532 if (to_put)
1533 xfrm_state_put(to_put);
1534
1535 if (err)
1536 return err;
1537
1538 if (!x) {
1539 xfrm_state_delete(x1);
1540 xfrm_state_put(x1);
1541 return 0;
1542 }
1543
1544 err = -EINVAL;
1545 spin_lock_bh(&x1->lock);
1546 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1547 if (x->encap && x1->encap)
1548 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1549 if (x->coaddr && x1->coaddr) {
1550 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1551 }
1552 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1553 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1554 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1555 x1->km.dying = 0;
1556
1557 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1558 if (x1->curlft.use_time)
1559 xfrm_state_check_expire(x1);
1560
1561 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1562 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1563
1564 if (x->props.smark.m || x->props.smark.v)
1565 x1->props.smark = x->props.smark;
1566
1567 if (x->if_id)
1568 x1->if_id = x->if_id;
1569
1570 __xfrm_state_bump_genids(x1);
1571 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1572 }
1573
1574 err = 0;
1575 x->km.state = XFRM_STATE_DEAD;
1576 __xfrm_state_put(x);
1577 }
1578 spin_unlock_bh(&x1->lock);
1579
1580 xfrm_state_put(x1);
1581
1582 return err;
1583 }
1584 EXPORT_SYMBOL(xfrm_state_update);
1585
1586 int xfrm_state_check_expire(struct xfrm_state *x)
1587 {
1588 if (!x->curlft.use_time)
1589 x->curlft.use_time = get_seconds();
1590
1591 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1592 x->curlft.packets >= x->lft.hard_packet_limit) {
1593 x->km.state = XFRM_STATE_EXPIRED;
1594 tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1595 return -EINVAL;
1596 }
1597
1598 if (!x->km.dying &&
1599 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1600 x->curlft.packets >= x->lft.soft_packet_limit)) {
1601 x->km.dying = 1;
1602 km_state_expired(x, 0, 0);
1603 }
1604 return 0;
1605 }
1606 EXPORT_SYMBOL(xfrm_state_check_expire);
1607
1608 struct xfrm_state *
1609 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1610 u8 proto, unsigned short family)
1611 {
1612 struct xfrm_state *x;
1613
1614 rcu_read_lock();
1615 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1616 rcu_read_unlock();
1617 return x;
1618 }
1619 EXPORT_SYMBOL(xfrm_state_lookup);
1620
1621 struct xfrm_state *
1622 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1623 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1624 u8 proto, unsigned short family)
1625 {
1626 struct xfrm_state *x;
1627
1628 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1629 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1630 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1631 return x;
1632 }
1633 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1634
1635 struct xfrm_state *
1636 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1637 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1638 const xfrm_address_t *saddr, int create, unsigned short family)
1639 {
1640 struct xfrm_state *x;
1641
1642 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1643 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1644 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1645
1646 return x;
1647 }
1648 EXPORT_SYMBOL(xfrm_find_acq);
1649
1650 #ifdef CONFIG_XFRM_SUB_POLICY
1651 int
1652 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1653 unsigned short family, struct net *net)
1654 {
1655 int i;
1656 int err = 0;
1657 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1658 if (!afinfo)
1659 return -EAFNOSUPPORT;
1660
1661 spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1662 if (afinfo->tmpl_sort)
1663 err = afinfo->tmpl_sort(dst, src, n);
1664 else
1665 for (i = 0; i < n; i++)
1666 dst[i] = src[i];
1667 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1668 rcu_read_unlock();
1669 return err;
1670 }
1671 EXPORT_SYMBOL(xfrm_tmpl_sort);
1672
1673 int
1674 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1675 unsigned short family)
1676 {
1677 int i;
1678 int err = 0;
1679 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1680 struct net *net = xs_net(*src);
1681
1682 if (!afinfo)
1683 return -EAFNOSUPPORT;
1684
1685 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1686 if (afinfo->state_sort)
1687 err = afinfo->state_sort(dst, src, n);
1688 else
1689 for (i = 0; i < n; i++)
1690 dst[i] = src[i];
1691 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1692 rcu_read_unlock();
1693 return err;
1694 }
1695 EXPORT_SYMBOL(xfrm_state_sort);
1696 #endif
1697
1698 /* Silly enough, but I'm lazy to build resolution list */
1699
1700 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1701 {
1702 int i;
1703
1704 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1705 struct xfrm_state *x;
1706
1707 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1708 if (x->km.seq == seq &&
1709 (mark & x->mark.m) == x->mark.v &&
1710 x->km.state == XFRM_STATE_ACQ) {
1711 xfrm_state_hold(x);
1712 return x;
1713 }
1714 }
1715 }
1716 return NULL;
1717 }
1718
1719 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1720 {
1721 struct xfrm_state *x;
1722
1723 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1724 x = __xfrm_find_acq_byseq(net, mark, seq);
1725 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1726 return x;
1727 }
1728 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1729
1730 u32 xfrm_get_acqseq(void)
1731 {
1732 u32 res;
1733 static atomic_t acqseq;
1734
1735 do {
1736 res = atomic_inc_return(&acqseq);
1737 } while (!res);
1738
1739 return res;
1740 }
1741 EXPORT_SYMBOL(xfrm_get_acqseq);
1742
1743 int verify_spi_info(u8 proto, u32 min, u32 max)
1744 {
1745 switch (proto) {
1746 case IPPROTO_AH:
1747 case IPPROTO_ESP:
1748 break;
1749
1750 case IPPROTO_COMP:
1751 /* IPCOMP spi is 16-bits. */
1752 if (max >= 0x10000)
1753 return -EINVAL;
1754 break;
1755
1756 default:
1757 return -EINVAL;
1758 }
1759
1760 if (min > max)
1761 return -EINVAL;
1762
1763 return 0;
1764 }
1765 EXPORT_SYMBOL(verify_spi_info);
1766
1767 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1768 {
1769 struct net *net = xs_net(x);
1770 unsigned int h;
1771 struct xfrm_state *x0;
1772 int err = -ENOENT;
1773 __be32 minspi = htonl(low);
1774 __be32 maxspi = htonl(high);
1775 u32 mark = x->mark.v & x->mark.m;
1776
1777 spin_lock_bh(&x->lock);
1778 if (x->km.state == XFRM_STATE_DEAD)
1779 goto unlock;
1780
1781 err = 0;
1782 if (x->id.spi)
1783 goto unlock;
1784
1785 err = -ENOENT;
1786
1787 if (minspi == maxspi) {
1788 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1789 if (x0) {
1790 xfrm_state_put(x0);
1791 goto unlock;
1792 }
1793 x->id.spi = minspi;
1794 } else {
1795 u32 spi = 0;
1796 for (h = 0; h < high-low+1; h++) {
1797 spi = low + prandom_u32()%(high-low+1);
1798 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1799 if (x0 == NULL) {
1800 x->id.spi = htonl(spi);
1801 break;
1802 }
1803 xfrm_state_put(x0);
1804 }
1805 }
1806 if (x->id.spi) {
1807 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1808 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1809 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1810 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1811
1812 err = 0;
1813 }
1814
1815 unlock:
1816 spin_unlock_bh(&x->lock);
1817
1818 return err;
1819 }
1820 EXPORT_SYMBOL(xfrm_alloc_spi);
1821
1822 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1823 struct xfrm_address_filter *filter)
1824 {
1825 if (filter) {
1826 if ((filter->family == AF_INET ||
1827 filter->family == AF_INET6) &&
1828 x->props.family != filter->family)
1829 return false;
1830
1831 return addr_match(&x->props.saddr, &filter->saddr,
1832 filter->splen) &&
1833 addr_match(&x->id.daddr, &filter->daddr,
1834 filter->dplen);
1835 }
1836 return true;
1837 }
1838
1839 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1840 int (*func)(struct xfrm_state *, int, void*),
1841 void *data)
1842 {
1843 struct xfrm_state *state;
1844 struct xfrm_state_walk *x;
1845 int err = 0;
1846
1847 if (walk->seq != 0 && list_empty(&walk->all))
1848 return 0;
1849
1850 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1851 if (list_empty(&walk->all))
1852 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1853 else
1854 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1855 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1856 if (x->state == XFRM_STATE_DEAD)
1857 continue;
1858 state = container_of(x, struct xfrm_state, km);
1859 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1860 continue;
1861 if (!__xfrm_state_filter_match(state, walk->filter))
1862 continue;
1863 err = func(state, walk->seq, data);
1864 if (err) {
1865 list_move_tail(&walk->all, &x->all);
1866 goto out;
1867 }
1868 walk->seq++;
1869 }
1870 if (walk->seq == 0) {
1871 err = -ENOENT;
1872 goto out;
1873 }
1874 list_del_init(&walk->all);
1875 out:
1876 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1877 return err;
1878 }
1879 EXPORT_SYMBOL(xfrm_state_walk);
1880
1881 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1882 struct xfrm_address_filter *filter)
1883 {
1884 INIT_LIST_HEAD(&walk->all);
1885 walk->proto = proto;
1886 walk->state = XFRM_STATE_DEAD;
1887 walk->seq = 0;
1888 walk->filter = filter;
1889 }
1890 EXPORT_SYMBOL(xfrm_state_walk_init);
1891
1892 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1893 {
1894 kfree(walk->filter);
1895
1896 if (list_empty(&walk->all))
1897 return;
1898
1899 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1900 list_del(&walk->all);
1901 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1902 }
1903 EXPORT_SYMBOL(xfrm_state_walk_done);
1904
1905 static void xfrm_replay_timer_handler(unsigned long data)
1906 {
1907 struct xfrm_state *x = (struct xfrm_state *)data;
1908
1909 spin_lock(&x->lock);
1910
1911 if (x->km.state == XFRM_STATE_VALID) {
1912 if (xfrm_aevent_is_on(xs_net(x)))
1913 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1914 else
1915 x->xflags |= XFRM_TIME_DEFER;
1916 }
1917
1918 spin_unlock(&x->lock);
1919 }
1920
1921 static LIST_HEAD(xfrm_km_list);
1922
1923 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1924 {
1925 struct xfrm_mgr *km;
1926
1927 rcu_read_lock();
1928 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1929 if (km->notify_policy)
1930 km->notify_policy(xp, dir, c);
1931 rcu_read_unlock();
1932 }
1933
1934 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1935 {
1936 struct xfrm_mgr *km;
1937 rcu_read_lock();
1938 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1939 if (km->notify)
1940 km->notify(x, c);
1941 rcu_read_unlock();
1942 }
1943
1944 EXPORT_SYMBOL(km_policy_notify);
1945 EXPORT_SYMBOL(km_state_notify);
1946
1947 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1948 {
1949 struct km_event c;
1950
1951 c.data.hard = hard;
1952 c.portid = portid;
1953 c.event = XFRM_MSG_EXPIRE;
1954 km_state_notify(x, &c);
1955 }
1956
1957 EXPORT_SYMBOL(km_state_expired);
1958 /*
1959 * We send to all registered managers regardless of failure
1960 * We are happy with one success
1961 */
1962 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1963 {
1964 int err = -EINVAL, acqret;
1965 struct xfrm_mgr *km;
1966
1967 rcu_read_lock();
1968 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1969 acqret = km->acquire(x, t, pol);
1970 if (!acqret)
1971 err = acqret;
1972 }
1973 rcu_read_unlock();
1974 return err;
1975 }
1976 EXPORT_SYMBOL(km_query);
1977
1978 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1979 {
1980 int err = -EINVAL;
1981 struct xfrm_mgr *km;
1982
1983 rcu_read_lock();
1984 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1985 if (km->new_mapping)
1986 err = km->new_mapping(x, ipaddr, sport);
1987 if (!err)
1988 break;
1989 }
1990 rcu_read_unlock();
1991 return err;
1992 }
1993 EXPORT_SYMBOL(km_new_mapping);
1994
1995 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1996 {
1997 struct km_event c;
1998
1999 c.data.hard = hard;
2000 c.portid = portid;
2001 c.event = XFRM_MSG_POLEXPIRE;
2002 km_policy_notify(pol, dir, &c);
2003 }
2004 EXPORT_SYMBOL(km_policy_expired);
2005
2006 #ifdef CONFIG_XFRM_MIGRATE
2007 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2008 const struct xfrm_migrate *m, int num_migrate,
2009 const struct xfrm_kmaddress *k,
2010 const struct xfrm_encap_tmpl *encap)
2011 {
2012 int err = -EINVAL;
2013 int ret;
2014 struct xfrm_mgr *km;
2015
2016 rcu_read_lock();
2017 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2018 if (km->migrate) {
2019 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2020 encap);
2021 if (!ret)
2022 err = ret;
2023 }
2024 }
2025 rcu_read_unlock();
2026 return err;
2027 }
2028 EXPORT_SYMBOL(km_migrate);
2029 #endif
2030
2031 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2032 {
2033 int err = -EINVAL;
2034 int ret;
2035 struct xfrm_mgr *km;
2036
2037 rcu_read_lock();
2038 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2039 if (km->report) {
2040 ret = km->report(net, proto, sel, addr);
2041 if (!ret)
2042 err = ret;
2043 }
2044 }
2045 rcu_read_unlock();
2046 return err;
2047 }
2048 EXPORT_SYMBOL(km_report);
2049
2050 bool km_is_alive(const struct km_event *c)
2051 {
2052 struct xfrm_mgr *km;
2053 bool is_alive = false;
2054
2055 rcu_read_lock();
2056 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2057 if (km->is_alive && km->is_alive(c)) {
2058 is_alive = true;
2059 break;
2060 }
2061 }
2062 rcu_read_unlock();
2063
2064 return is_alive;
2065 }
2066 EXPORT_SYMBOL(km_is_alive);
2067
2068 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2069 {
2070 int err;
2071 u8 *data;
2072 struct xfrm_mgr *km;
2073 struct xfrm_policy *pol = NULL;
2074
2075 #ifdef SUPPORT_STRONGSWAN_5X
2076 #ifdef CONFIG_COMPAT
2077 if (in_compat_syscall())
2078 return -EOPNOTSUPP;
2079 #endif
2080 #endif
2081
2082 if (!optval && !optlen) {
2083 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2084 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2085 __sk_dst_reset(sk);
2086 return 0;
2087 }
2088
2089 if (optlen <= 0 || optlen > PAGE_SIZE)
2090 return -EMSGSIZE;
2091
2092 data = memdup_user(optval, optlen);
2093 if (IS_ERR(data))
2094 return PTR_ERR(data);
2095
2096 err = -EINVAL;
2097 rcu_read_lock();
2098 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2099 pol = km->compile_policy(sk, optname, data,
2100 optlen, &err);
2101 if (err >= 0)
2102 break;
2103 }
2104 rcu_read_unlock();
2105
2106 if (err >= 0) {
2107 xfrm_sk_policy_insert(sk, err, pol);
2108 xfrm_pol_put(pol);
2109 __sk_dst_reset(sk);
2110 err = 0;
2111 }
2112
2113 kfree(data);
2114 return err;
2115 }
2116 EXPORT_SYMBOL(xfrm_user_policy);
2117
2118 static DEFINE_SPINLOCK(xfrm_km_lock);
2119
2120 int xfrm_register_km(struct xfrm_mgr *km)
2121 {
2122 spin_lock_bh(&xfrm_km_lock);
2123 list_add_tail_rcu(&km->list, &xfrm_km_list);
2124 spin_unlock_bh(&xfrm_km_lock);
2125 return 0;
2126 }
2127 EXPORT_SYMBOL(xfrm_register_km);
2128
2129 int xfrm_unregister_km(struct xfrm_mgr *km)
2130 {
2131 spin_lock_bh(&xfrm_km_lock);
2132 list_del_rcu(&km->list);
2133 spin_unlock_bh(&xfrm_km_lock);
2134 synchronize_rcu();
2135 return 0;
2136 }
2137 EXPORT_SYMBOL(xfrm_unregister_km);
2138
2139 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2140 {
2141 int err = 0;
2142
2143 if (WARN_ON(afinfo->family >= NPROTO))
2144 return -EAFNOSUPPORT;
2145
2146 spin_lock_bh(&xfrm_state_afinfo_lock);
2147 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2148 err = -EEXIST;
2149 else
2150 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2151 spin_unlock_bh(&xfrm_state_afinfo_lock);
2152 return err;
2153 }
2154 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2155
2156 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2157 {
2158 int err = 0, family = afinfo->family;
2159
2160 if (WARN_ON(family >= NPROTO))
2161 return -EAFNOSUPPORT;
2162
2163 spin_lock_bh(&xfrm_state_afinfo_lock);
2164 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2165 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2166 err = -EINVAL;
2167 else
2168 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2169 }
2170 spin_unlock_bh(&xfrm_state_afinfo_lock);
2171 synchronize_rcu();
2172 return err;
2173 }
2174 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2175
2176 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2177 {
2178 if (unlikely(family >= NPROTO))
2179 return NULL;
2180
2181 return rcu_dereference(xfrm_state_afinfo[family]);
2182 }
2183
2184 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2185 {
2186 struct xfrm_state_afinfo *afinfo;
2187 if (unlikely(family >= NPROTO))
2188 return NULL;
2189 rcu_read_lock();
2190 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2191 if (unlikely(!afinfo))
2192 rcu_read_unlock();
2193 return afinfo;
2194 }
2195
2196 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2197 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2198 {
2199 if (x->tunnel) {
2200 struct xfrm_state *t = x->tunnel;
2201
2202 if (atomic_read(&t->tunnel_users) == 2)
2203 xfrm_state_delete(t);
2204 atomic_dec(&t->tunnel_users);
2205 xfrm_state_put(t);
2206 x->tunnel = NULL;
2207 }
2208 }
2209 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2210
2211 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2212 {
2213 const struct xfrm_type *type = READ_ONCE(x->type);
2214
2215 if (x->km.state == XFRM_STATE_VALID &&
2216 type && type->get_mtu)
2217 return type->get_mtu(x, mtu);
2218
2219 return mtu - x->props.header_len;
2220 }
2221
2222 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2223 {
2224 struct xfrm_state_afinfo *afinfo;
2225 struct xfrm_mode *inner_mode;
2226 int family = x->props.family;
2227 int err;
2228
2229 err = -EAFNOSUPPORT;
2230 afinfo = xfrm_state_get_afinfo(family);
2231 if (!afinfo)
2232 goto error;
2233
2234 err = 0;
2235 if (afinfo->init_flags)
2236 err = afinfo->init_flags(x);
2237
2238 rcu_read_unlock();
2239
2240 if (err)
2241 goto error;
2242
2243 err = -EPROTONOSUPPORT;
2244
2245 if (x->sel.family != AF_UNSPEC) {
2246 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2247 if (inner_mode == NULL)
2248 goto error;
2249
2250 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2251 family != x->sel.family) {
2252 xfrm_put_mode(inner_mode);
2253 goto error;
2254 }
2255
2256 x->inner_mode = inner_mode;
2257 } else {
2258 struct xfrm_mode *inner_mode_iaf;
2259 int iafamily = AF_INET;
2260
2261 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2262 if (inner_mode == NULL)
2263 goto error;
2264
2265 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2266 xfrm_put_mode(inner_mode);
2267 goto error;
2268 }
2269 x->inner_mode = inner_mode;
2270
2271 if (x->props.family == AF_INET)
2272 iafamily = AF_INET6;
2273
2274 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2275 if (inner_mode_iaf) {
2276 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2277 x->inner_mode_iaf = inner_mode_iaf;
2278 else
2279 xfrm_put_mode(inner_mode_iaf);
2280 }
2281 }
2282
2283 x->type = xfrm_get_type(x->id.proto, family);
2284 if (x->type == NULL)
2285 goto error;
2286
2287 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2288
2289 err = x->type->init_state(x);
2290 if (err)
2291 goto error;
2292
2293 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2294 if (x->outer_mode == NULL) {
2295 err = -EPROTONOSUPPORT;
2296 goto error;
2297 }
2298
2299 if (init_replay) {
2300 err = xfrm_init_replay(x);
2301 if (err)
2302 goto error;
2303 }
2304
2305 x->km.state = XFRM_STATE_VALID;
2306
2307 error:
2308 return err;
2309 }
2310
2311 EXPORT_SYMBOL(__xfrm_init_state);
2312
2313 int xfrm_init_state(struct xfrm_state *x)
2314 {
2315 return __xfrm_init_state(x, true, false);
2316 }
2317
2318 EXPORT_SYMBOL(xfrm_init_state);
2319
2320 int __net_init xfrm_state_init(struct net *net)
2321 {
2322 unsigned int sz;
2323
2324 INIT_LIST_HEAD(&net->xfrm.state_all);
2325
2326 sz = sizeof(struct hlist_head) * 8;
2327
2328 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2329 if (!net->xfrm.state_bydst)
2330 goto out_bydst;
2331 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2332 if (!net->xfrm.state_bysrc)
2333 goto out_bysrc;
2334 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2335 if (!net->xfrm.state_byspi)
2336 goto out_byspi;
2337 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2338
2339 net->xfrm.state_num = 0;
2340 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2341 spin_lock_init(&net->xfrm.xfrm_state_lock);
2342 return 0;
2343
2344 out_byspi:
2345 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2346 out_bysrc:
2347 xfrm_hash_free(net->xfrm.state_bydst, sz);
2348 out_bydst:
2349 return -ENOMEM;
2350 }
2351
2352 void xfrm_state_fini(struct net *net)
2353 {
2354 unsigned int sz;
2355
2356 flush_work(&net->xfrm.state_hash_work);
2357 xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2358 flush_work(&xfrm_state_gc_work);
2359
2360 WARN_ON(!list_empty(&net->xfrm.state_all));
2361
2362 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2363 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2364 xfrm_hash_free(net->xfrm.state_byspi, sz);
2365 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2366 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2367 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2368 xfrm_hash_free(net->xfrm.state_bydst, sz);
2369 }
2370
2371 #ifdef CONFIG_AUDITSYSCALL
2372 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2373 struct audit_buffer *audit_buf)
2374 {
2375 struct xfrm_sec_ctx *ctx = x->security;
2376 u32 spi = ntohl(x->id.spi);
2377
2378 if (ctx)
2379 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2380 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2381
2382 switch (x->props.family) {
2383 case AF_INET:
2384 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2385 &x->props.saddr.a4, &x->id.daddr.a4);
2386 break;
2387 case AF_INET6:
2388 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2389 x->props.saddr.a6, x->id.daddr.a6);
2390 break;
2391 }
2392
2393 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2394 }
2395
2396 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2397 struct audit_buffer *audit_buf)
2398 {
2399 const struct iphdr *iph4;
2400 const struct ipv6hdr *iph6;
2401
2402 switch (family) {
2403 case AF_INET:
2404 iph4 = ip_hdr(skb);
2405 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2406 &iph4->saddr, &iph4->daddr);
2407 break;
2408 case AF_INET6:
2409 iph6 = ipv6_hdr(skb);
2410 audit_log_format(audit_buf,
2411 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2412 &iph6->saddr, &iph6->daddr,
2413 iph6->flow_lbl[0] & 0x0f,
2414 iph6->flow_lbl[1],
2415 iph6->flow_lbl[2]);
2416 break;
2417 }
2418 }
2419
2420 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2421 {
2422 struct audit_buffer *audit_buf;
2423
2424 audit_buf = xfrm_audit_start("SAD-add");
2425 if (audit_buf == NULL)
2426 return;
2427 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2428 xfrm_audit_helper_sainfo(x, audit_buf);
2429 audit_log_format(audit_buf, " res=%u", result);
2430 audit_log_end(audit_buf);
2431 }
2432 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2433
2434 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2435 {
2436 struct audit_buffer *audit_buf;
2437
2438 audit_buf = xfrm_audit_start("SAD-delete");
2439 if (audit_buf == NULL)
2440 return;
2441 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2442 xfrm_audit_helper_sainfo(x, audit_buf);
2443 audit_log_format(audit_buf, " res=%u", result);
2444 audit_log_end(audit_buf);
2445 }
2446 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2447
2448 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2449 struct sk_buff *skb)
2450 {
2451 struct audit_buffer *audit_buf;
2452 u32 spi;
2453
2454 audit_buf = xfrm_audit_start("SA-replay-overflow");
2455 if (audit_buf == NULL)
2456 return;
2457 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2458 /* don't record the sequence number because it's inherent in this kind
2459 * of audit message */
2460 spi = ntohl(x->id.spi);
2461 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2462 audit_log_end(audit_buf);
2463 }
2464 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2465
2466 void xfrm_audit_state_replay(struct xfrm_state *x,
2467 struct sk_buff *skb, __be32 net_seq)
2468 {
2469 struct audit_buffer *audit_buf;
2470 u32 spi;
2471
2472 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2473 if (audit_buf == NULL)
2474 return;
2475 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2476 spi = ntohl(x->id.spi);
2477 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2478 spi, spi, ntohl(net_seq));
2479 audit_log_end(audit_buf);
2480 }
2481 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2482
2483 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2484 {
2485 struct audit_buffer *audit_buf;
2486
2487 audit_buf = xfrm_audit_start("SA-notfound");
2488 if (audit_buf == NULL)
2489 return;
2490 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2491 audit_log_end(audit_buf);
2492 }
2493 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2494
2495 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2496 __be32 net_spi, __be32 net_seq)
2497 {
2498 struct audit_buffer *audit_buf;
2499 u32 spi;
2500
2501 audit_buf = xfrm_audit_start("SA-notfound");
2502 if (audit_buf == NULL)
2503 return;
2504 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2505 spi = ntohl(net_spi);
2506 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2507 spi, spi, ntohl(net_seq));
2508 audit_log_end(audit_buf);
2509 }
2510 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2511
2512 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2513 struct sk_buff *skb, u8 proto)
2514 {
2515 struct audit_buffer *audit_buf;
2516 __be32 net_spi;
2517 __be32 net_seq;
2518
2519 audit_buf = xfrm_audit_start("SA-icv-failure");
2520 if (audit_buf == NULL)
2521 return;
2522 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2523 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2524 u32 spi = ntohl(net_spi);
2525 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2526 spi, spi, ntohl(net_seq));
2527 }
2528 audit_log_end(audit_buf);
2529 }
2530 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2531 #endif /* CONFIG_AUDITSYSCALL */
Motorola kernel-slsi