?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf tools: Streamline bpf examples and headers installation
[linux/fpc-iii.git] / net / xfrm / xfrm_state.c
blob8308281f32530bd9103e7f7bd06472824600c526
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 static struct kmem_cache *xfrm_state_cache __ro_after_init;
46
47 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
48 static HLIST_HEAD(xfrm_state_gc_list);
49
50 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
51 {
52 return refcount_inc_not_zero(&x->refcnt);
53 }
54
55 static inline unsigned int xfrm_dst_hash(struct net *net,
56 const xfrm_address_t *daddr,
57 const xfrm_address_t *saddr,
58 u32 reqid,
59 unsigned short family)
60 {
61 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
62 }
63
64 static inline unsigned int xfrm_src_hash(struct net *net,
65 const xfrm_address_t *daddr,
66 const xfrm_address_t *saddr,
67 unsigned short family)
68 {
69 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
70 }
71
72 static inline unsigned int
73 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
74 __be32 spi, u8 proto, unsigned short family)
75 {
76 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
77 }
78
79 static void xfrm_hash_transfer(struct hlist_head *list,
80 struct hlist_head *ndsttable,
81 struct hlist_head *nsrctable,
82 struct hlist_head *nspitable,
83 unsigned int nhashmask)
84 {
85 struct hlist_node *tmp;
86 struct xfrm_state *x;
87
88 hlist_for_each_entry_safe(x, tmp, list, bydst) {
89 unsigned int h;
90
91 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
92 x->props.reqid, x->props.family,
93 nhashmask);
94 hlist_add_head_rcu(&x->bydst, ndsttable + h);
95
96 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
97 x->props.family,
98 nhashmask);
99 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
100
101 if (x->id.spi) {
102 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
103 x->id.proto, x->props.family,
104 nhashmask);
105 hlist_add_head_rcu(&x->byspi, nspitable + h);
106 }
107 }
108 }
109
110 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
111 {
112 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
113 }
114
115 static void xfrm_hash_resize(struct work_struct *work)
116 {
117 struct net *net = container_of(work, struct net, xfrm.state_hash_work);
118 struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
119 unsigned long nsize, osize;
120 unsigned int nhashmask, ohashmask;
121 int i;
122
123 nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
124 ndst = xfrm_hash_alloc(nsize);
125 if (!ndst)
126 return;
127 nsrc = xfrm_hash_alloc(nsize);
128 if (!nsrc) {
129 xfrm_hash_free(ndst, nsize);
130 return;
131 }
132 nspi = xfrm_hash_alloc(nsize);
133 if (!nspi) {
134 xfrm_hash_free(ndst, nsize);
135 xfrm_hash_free(nsrc, nsize);
136 return;
137 }
138
139 spin_lock_bh(&net->xfrm.xfrm_state_lock);
140 write_seqcount_begin(&xfrm_state_hash_generation);
141
142 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
143 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
144 for (i = net->xfrm.state_hmask; i >= 0; i--)
145 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
146
147 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
148 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
149 ohashmask = net->xfrm.state_hmask;
150
151 rcu_assign_pointer(net->xfrm.state_bydst, ndst);
152 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
153 rcu_assign_pointer(net->xfrm.state_byspi, nspi);
154 net->xfrm.state_hmask = nhashmask;
155
156 write_seqcount_end(&xfrm_state_hash_generation);
157 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
158
159 osize = (ohashmask + 1) * sizeof(struct hlist_head);
160
161 synchronize_rcu();
162
163 xfrm_hash_free(odst, osize);
164 xfrm_hash_free(osrc, osize);
165 xfrm_hash_free(ospi, osize);
166 }
167
168 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
169 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
170
171 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
172
173 int __xfrm_state_delete(struct xfrm_state *x);
174
175 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
176 bool km_is_alive(const struct km_event *c);
177 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
178
179 static DEFINE_SPINLOCK(xfrm_type_lock);
180 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
181 {
182 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
183 const struct xfrm_type **typemap;
184 int err = 0;
185
186 if (unlikely(afinfo == NULL))
187 return -EAFNOSUPPORT;
188 typemap = afinfo->type_map;
189 spin_lock_bh(&xfrm_type_lock);
190
191 if (likely(typemap[type->proto] == NULL))
192 typemap[type->proto] = type;
193 else
194 err = -EEXIST;
195 spin_unlock_bh(&xfrm_type_lock);
196 rcu_read_unlock();
197 return err;
198 }
199 EXPORT_SYMBOL(xfrm_register_type);
200
201 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
202 {
203 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
204 const struct xfrm_type **typemap;
205 int err = 0;
206
207 if (unlikely(afinfo == NULL))
208 return -EAFNOSUPPORT;
209 typemap = afinfo->type_map;
210 spin_lock_bh(&xfrm_type_lock);
211
212 if (unlikely(typemap[type->proto] != type))
213 err = -ENOENT;
214 else
215 typemap[type->proto] = NULL;
216 spin_unlock_bh(&xfrm_type_lock);
217 rcu_read_unlock();
218 return err;
219 }
220 EXPORT_SYMBOL(xfrm_unregister_type);
221
222 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
223 {
224 struct xfrm_state_afinfo *afinfo;
225 const struct xfrm_type **typemap;
226 const struct xfrm_type *type;
227 int modload_attempted = 0;
228
229 retry:
230 afinfo = xfrm_state_get_afinfo(family);
231 if (unlikely(afinfo == NULL))
232 return NULL;
233 typemap = afinfo->type_map;
234
235 type = READ_ONCE(typemap[proto]);
236 if (unlikely(type && !try_module_get(type->owner)))
237 type = NULL;
238
239 rcu_read_unlock();
240
241 if (!type && !modload_attempted) {
242 request_module("xfrm-type-%d-%d", family, proto);
243 modload_attempted = 1;
244 goto retry;
245 }
246
247 return type;
248 }
249
250 static void xfrm_put_type(const struct xfrm_type *type)
251 {
252 module_put(type->owner);
253 }
254
255 static DEFINE_SPINLOCK(xfrm_type_offload_lock);
256 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
257 unsigned short family)
258 {
259 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
260 const struct xfrm_type_offload **typemap;
261 int err = 0;
262
263 if (unlikely(afinfo == NULL))
264 return -EAFNOSUPPORT;
265 typemap = afinfo->type_offload_map;
266 spin_lock_bh(&xfrm_type_offload_lock);
267
268 if (likely(typemap[type->proto] == NULL))
269 typemap[type->proto] = type;
270 else
271 err = -EEXIST;
272 spin_unlock_bh(&xfrm_type_offload_lock);
273 rcu_read_unlock();
274 return err;
275 }
276 EXPORT_SYMBOL(xfrm_register_type_offload);
277
278 int xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
279 unsigned short family)
280 {
281 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
282 const struct xfrm_type_offload **typemap;
283 int err = 0;
284
285 if (unlikely(afinfo == NULL))
286 return -EAFNOSUPPORT;
287 typemap = afinfo->type_offload_map;
288 spin_lock_bh(&xfrm_type_offload_lock);
289
290 if (unlikely(typemap[type->proto] != type))
291 err = -ENOENT;
292 else
293 typemap[type->proto] = NULL;
294 spin_unlock_bh(&xfrm_type_offload_lock);
295 rcu_read_unlock();
296 return err;
297 }
298 EXPORT_SYMBOL(xfrm_unregister_type_offload);
299
300 static const struct xfrm_type_offload *
301 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
302 {
303 struct xfrm_state_afinfo *afinfo;
304 const struct xfrm_type_offload **typemap;
305 const struct xfrm_type_offload *type;
306
307 retry:
308 afinfo = xfrm_state_get_afinfo(family);
309 if (unlikely(afinfo == NULL))
310 return NULL;
311 typemap = afinfo->type_offload_map;
312
313 type = typemap[proto];
314 if ((type && !try_module_get(type->owner)))
315 type = NULL;
316
317 rcu_read_unlock();
318
319 if (!type && try_load) {
320 request_module("xfrm-offload-%d-%d", family, proto);
321 try_load = false;
322 goto retry;
323 }
324
325 return type;
326 }
327
328 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
329 {
330 module_put(type->owner);
331 }
332
333 static DEFINE_SPINLOCK(xfrm_mode_lock);
334 int xfrm_register_mode(struct xfrm_mode *mode, int family)
335 {
336 struct xfrm_state_afinfo *afinfo;
337 struct xfrm_mode **modemap;
338 int err;
339
340 if (unlikely(mode->encap >= XFRM_MODE_MAX))
341 return -EINVAL;
342
343 afinfo = xfrm_state_get_afinfo(family);
344 if (unlikely(afinfo == NULL))
345 return -EAFNOSUPPORT;
346
347 err = -EEXIST;
348 modemap = afinfo->mode_map;
349 spin_lock_bh(&xfrm_mode_lock);
350 if (modemap[mode->encap])
351 goto out;
352
353 err = -ENOENT;
354 if (!try_module_get(afinfo->owner))
355 goto out;
356
357 mode->afinfo = afinfo;
358 modemap[mode->encap] = mode;
359 err = 0;
360
361 out:
362 spin_unlock_bh(&xfrm_mode_lock);
363 rcu_read_unlock();
364 return err;
365 }
366 EXPORT_SYMBOL(xfrm_register_mode);
367
368 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
369 {
370 struct xfrm_state_afinfo *afinfo;
371 struct xfrm_mode **modemap;
372 int err;
373
374 if (unlikely(mode->encap >= XFRM_MODE_MAX))
375 return -EINVAL;
376
377 afinfo = xfrm_state_get_afinfo(family);
378 if (unlikely(afinfo == NULL))
379 return -EAFNOSUPPORT;
380
381 err = -ENOENT;
382 modemap = afinfo->mode_map;
383 spin_lock_bh(&xfrm_mode_lock);
384 if (likely(modemap[mode->encap] == mode)) {
385 modemap[mode->encap] = NULL;
386 module_put(mode->afinfo->owner);
387 err = 0;
388 }
389
390 spin_unlock_bh(&xfrm_mode_lock);
391 rcu_read_unlock();
392 return err;
393 }
394 EXPORT_SYMBOL(xfrm_unregister_mode);
395
396 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
397 {
398 struct xfrm_state_afinfo *afinfo;
399 struct xfrm_mode *mode;
400 int modload_attempted = 0;
401
402 if (unlikely(encap >= XFRM_MODE_MAX))
403 return NULL;
404
405 retry:
406 afinfo = xfrm_state_get_afinfo(family);
407 if (unlikely(afinfo == NULL))
408 return NULL;
409
410 mode = READ_ONCE(afinfo->mode_map[encap]);
411 if (unlikely(mode && !try_module_get(mode->owner)))
412 mode = NULL;
413
414 rcu_read_unlock();
415 if (!mode && !modload_attempted) {
416 request_module("xfrm-mode-%d-%d", family, encap);
417 modload_attempted = 1;
418 goto retry;
419 }
420
421 return mode;
422 }
423
424 static void xfrm_put_mode(struct xfrm_mode *mode)
425 {
426 module_put(mode->owner);
427 }
428
429 static void xfrm_state_gc_destroy(struct xfrm_state *x)
430 {
431 tasklet_hrtimer_cancel(&x->mtimer);
432 del_timer_sync(&x->rtimer);
433 kfree(x->aead);
434 kfree(x->aalg);
435 kfree(x->ealg);
436 kfree(x->calg);
437 kfree(x->encap);
438 kfree(x->coaddr);
439 kfree(x->replay_esn);
440 kfree(x->preplay_esn);
441 if (x->inner_mode)
442 xfrm_put_mode(x->inner_mode);
443 if (x->inner_mode_iaf)
444 xfrm_put_mode(x->inner_mode_iaf);
445 if (x->outer_mode)
446 xfrm_put_mode(x->outer_mode);
447 if (x->type_offload)
448 xfrm_put_type_offload(x->type_offload);
449 if (x->type) {
450 x->type->destructor(x);
451 xfrm_put_type(x->type);
452 }
453 xfrm_dev_state_free(x);
454 security_xfrm_state_free(x);
455 kmem_cache_free(xfrm_state_cache, x);
456 }
457
458 static void xfrm_state_gc_task(struct work_struct *work)
459 {
460 struct xfrm_state *x;
461 struct hlist_node *tmp;
462 struct hlist_head gc_list;
463
464 spin_lock_bh(&xfrm_state_gc_lock);
465 hlist_move_list(&xfrm_state_gc_list, &gc_list);
466 spin_unlock_bh(&xfrm_state_gc_lock);
467
468 synchronize_rcu();
469
470 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
471 xfrm_state_gc_destroy(x);
472 }
473
474 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
475 {
476 struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
477 struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
478 unsigned long now = get_seconds();
479 long next = LONG_MAX;
480 int warn = 0;
481 int err = 0;
482
483 spin_lock(&x->lock);
484 if (x->km.state == XFRM_STATE_DEAD)
485 goto out;
486 if (x->km.state == XFRM_STATE_EXPIRED)
487 goto expired;
488 if (x->lft.hard_add_expires_seconds) {
489 long tmo = x->lft.hard_add_expires_seconds +
490 x->curlft.add_time - now;
491 if (tmo <= 0) {
492 if (x->xflags & XFRM_SOFT_EXPIRE) {
493 /* enter hard expire without soft expire first?!
494 * setting a new date could trigger this.
495 * workaround: fix x->curflt.add_time by below:
496 */
497 x->curlft.add_time = now - x->saved_tmo - 1;
498 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
499 } else
500 goto expired;
501 }
502 if (tmo < next)
503 next = tmo;
504 }
505 if (x->lft.hard_use_expires_seconds) {
506 long tmo = x->lft.hard_use_expires_seconds +
507 (x->curlft.use_time ? : now) - now;
508 if (tmo <= 0)
509 goto expired;
510 if (tmo < next)
511 next = tmo;
512 }
513 if (x->km.dying)
514 goto resched;
515 if (x->lft.soft_add_expires_seconds) {
516 long tmo = x->lft.soft_add_expires_seconds +
517 x->curlft.add_time - now;
518 if (tmo <= 0) {
519 warn = 1;
520 x->xflags &= ~XFRM_SOFT_EXPIRE;
521 } else if (tmo < next) {
522 next = tmo;
523 x->xflags |= XFRM_SOFT_EXPIRE;
524 x->saved_tmo = tmo;
525 }
526 }
527 if (x->lft.soft_use_expires_seconds) {
528 long tmo = x->lft.soft_use_expires_seconds +
529 (x->curlft.use_time ? : now) - now;
530 if (tmo <= 0)
531 warn = 1;
532 else if (tmo < next)
533 next = tmo;
534 }
535
536 x->km.dying = warn;
537 if (warn)
538 km_state_expired(x, 0, 0);
539 resched:
540 if (next != LONG_MAX) {
541 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
542 }
543
544 goto out;
545
546 expired:
547 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
548 x->km.state = XFRM_STATE_EXPIRED;
549
550 err = __xfrm_state_delete(x);
551 if (!err)
552 km_state_expired(x, 1, 0);
553
554 xfrm_audit_state_delete(x, err ? 0 : 1, true);
555
556 out:
557 spin_unlock(&x->lock);
558 return HRTIMER_NORESTART;
559 }
560
561 static void xfrm_replay_timer_handler(struct timer_list *t);
562
563 struct xfrm_state *xfrm_state_alloc(struct net *net)
564 {
565 struct xfrm_state *x;
566
567 x = kmem_cache_alloc(xfrm_state_cache, GFP_ATOMIC | __GFP_ZERO);
568
569 if (x) {
570 write_pnet(&x->xs_net, net);
571 refcount_set(&x->refcnt, 1);
572 atomic_set(&x->tunnel_users, 0);
573 INIT_LIST_HEAD(&x->km.all);
574 INIT_HLIST_NODE(&x->bydst);
575 INIT_HLIST_NODE(&x->bysrc);
576 INIT_HLIST_NODE(&x->byspi);
577 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
578 CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
579 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
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;
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)
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->props.flags & XFRM_STATE_WILDRECV) &&
959 xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
960 tmpl->mode == x->props.mode &&
961 tmpl->id.proto == x->id.proto &&
962 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
963 xfrm_state_look_at(pol, x, fl, encap_family,
964 &best, &acquire_in_progress, &error);
965 }
966 if (best || acquire_in_progress)
967 goto found;
968
969 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
970 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
971 if (x->props.family == encap_family &&
972 x->props.reqid == tmpl->reqid &&
973 (mark & x->mark.m) == x->mark.v &&
974 !(x->props.flags & XFRM_STATE_WILDRECV) &&
975 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
976 tmpl->mode == x->props.mode &&
977 tmpl->id.proto == x->id.proto &&
978 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
979 xfrm_state_look_at(pol, x, fl, encap_family,
980 &best, &acquire_in_progress, &error);
981 }
982
983 found:
984 x = best;
985 if (!x && !error && !acquire_in_progress) {
986 if (tmpl->id.spi &&
987 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
988 tmpl->id.proto, encap_family)) != NULL) {
989 to_put = x0;
990 error = -EEXIST;
991 goto out;
992 }
993
994 c.net = net;
995 /* If the KMs have no listeners (yet...), avoid allocating an SA
996 * for each and every packet - garbage collection might not
997 * handle the flood.
998 */
999 if (!km_is_alive(&c)) {
1000 error = -ESRCH;
1001 goto out;
1002 }
1003
1004 x = xfrm_state_alloc(net);
1005 if (x == NULL) {
1006 error = -ENOMEM;
1007 goto out;
1008 }
1009 /* Initialize temporary state matching only
1010 * to current session. */
1011 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1012 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1013
1014 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1015 if (error) {
1016 x->km.state = XFRM_STATE_DEAD;
1017 to_put = x;
1018 x = NULL;
1019 goto out;
1020 }
1021
1022 if (km_query(x, tmpl, pol) == 0) {
1023 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1024 x->km.state = XFRM_STATE_ACQ;
1025 list_add(&x->km.all, &net->xfrm.state_all);
1026 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1027 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1028 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1029 if (x->id.spi) {
1030 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1031 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1032 }
1033 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1034 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1035 net->xfrm.state_num++;
1036 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1037 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1038 } else {
1039 x->km.state = XFRM_STATE_DEAD;
1040 to_put = x;
1041 x = NULL;
1042 error = -ESRCH;
1043 }
1044 }
1045 out:
1046 if (x) {
1047 if (!xfrm_state_hold_rcu(x)) {
1048 *err = -EAGAIN;
1049 x = NULL;
1050 }
1051 } else {
1052 *err = acquire_in_progress ? -EAGAIN : error;
1053 }
1054 rcu_read_unlock();
1055 if (to_put)
1056 xfrm_state_put(to_put);
1057
1058 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1059 *err = -EAGAIN;
1060 if (x) {
1061 xfrm_state_put(x);
1062 x = NULL;
1063 }
1064 }
1065
1066 return x;
1067 }
1068
1069 struct xfrm_state *
1070 xfrm_stateonly_find(struct net *net, u32 mark,
1071 xfrm_address_t *daddr, xfrm_address_t *saddr,
1072 unsigned short family, u8 mode, u8 proto, u32 reqid)
1073 {
1074 unsigned int h;
1075 struct xfrm_state *rx = NULL, *x = NULL;
1076
1077 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1078 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1079 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1080 if (x->props.family == family &&
1081 x->props.reqid == reqid &&
1082 (mark & x->mark.m) == x->mark.v &&
1083 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1084 xfrm_state_addr_check(x, daddr, saddr, family) &&
1085 mode == x->props.mode &&
1086 proto == x->id.proto &&
1087 x->km.state == XFRM_STATE_VALID) {
1088 rx = x;
1089 break;
1090 }
1091 }
1092
1093 if (rx)
1094 xfrm_state_hold(rx);
1095 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1096
1097
1098 return rx;
1099 }
1100 EXPORT_SYMBOL(xfrm_stateonly_find);
1101
1102 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1103 unsigned short family)
1104 {
1105 struct xfrm_state *x;
1106 struct xfrm_state_walk *w;
1107
1108 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1109 list_for_each_entry(w, &net->xfrm.state_all, all) {
1110 x = container_of(w, struct xfrm_state, km);
1111 if (x->props.family != family ||
1112 x->id.spi != spi)
1113 continue;
1114
1115 xfrm_state_hold(x);
1116 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1117 return x;
1118 }
1119 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1120 return NULL;
1121 }
1122 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1123
1124 static void __xfrm_state_insert(struct xfrm_state *x)
1125 {
1126 struct net *net = xs_net(x);
1127 unsigned int h;
1128
1129 list_add(&x->km.all, &net->xfrm.state_all);
1130
1131 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1132 x->props.reqid, x->props.family);
1133 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1134
1135 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1136 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1137
1138 if (x->id.spi) {
1139 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1140 x->props.family);
1141
1142 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1143 }
1144
1145 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1146 if (x->replay_maxage)
1147 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1148
1149 net->xfrm.state_num++;
1150
1151 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1152 }
1153
1154 /* net->xfrm.xfrm_state_lock is held */
1155 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1156 {
1157 struct net *net = xs_net(xnew);
1158 unsigned short family = xnew->props.family;
1159 u32 reqid = xnew->props.reqid;
1160 struct xfrm_state *x;
1161 unsigned int h;
1162 u32 mark = xnew->mark.v & xnew->mark.m;
1163
1164 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1165 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1166 if (x->props.family == family &&
1167 x->props.reqid == reqid &&
1168 (mark & x->mark.m) == x->mark.v &&
1169 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1170 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1171 x->genid++;
1172 }
1173 }
1174
1175 void xfrm_state_insert(struct xfrm_state *x)
1176 {
1177 struct net *net = xs_net(x);
1178
1179 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1180 __xfrm_state_bump_genids(x);
1181 __xfrm_state_insert(x);
1182 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1183 }
1184 EXPORT_SYMBOL(xfrm_state_insert);
1185
1186 /* net->xfrm.xfrm_state_lock is held */
1187 static struct xfrm_state *__find_acq_core(struct net *net,
1188 const struct xfrm_mark *m,
1189 unsigned short family, u8 mode,
1190 u32 reqid, u8 proto,
1191 const xfrm_address_t *daddr,
1192 const xfrm_address_t *saddr,
1193 int create)
1194 {
1195 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1196 struct xfrm_state *x;
1197 u32 mark = m->v & m->m;
1198
1199 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1200 if (x->props.reqid != reqid ||
1201 x->props.mode != mode ||
1202 x->props.family != family ||
1203 x->km.state != XFRM_STATE_ACQ ||
1204 x->id.spi != 0 ||
1205 x->id.proto != proto ||
1206 (mark & x->mark.m) != x->mark.v ||
1207 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1208 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1209 continue;
1210
1211 xfrm_state_hold(x);
1212 return x;
1213 }
1214
1215 if (!create)
1216 return NULL;
1217
1218 x = xfrm_state_alloc(net);
1219 if (likely(x)) {
1220 switch (family) {
1221 case AF_INET:
1222 x->sel.daddr.a4 = daddr->a4;
1223 x->sel.saddr.a4 = saddr->a4;
1224 x->sel.prefixlen_d = 32;
1225 x->sel.prefixlen_s = 32;
1226 x->props.saddr.a4 = saddr->a4;
1227 x->id.daddr.a4 = daddr->a4;
1228 break;
1229
1230 case AF_INET6:
1231 x->sel.daddr.in6 = daddr->in6;
1232 x->sel.saddr.in6 = saddr->in6;
1233 x->sel.prefixlen_d = 128;
1234 x->sel.prefixlen_s = 128;
1235 x->props.saddr.in6 = saddr->in6;
1236 x->id.daddr.in6 = daddr->in6;
1237 break;
1238 }
1239
1240 x->km.state = XFRM_STATE_ACQ;
1241 x->id.proto = proto;
1242 x->props.family = family;
1243 x->props.mode = mode;
1244 x->props.reqid = reqid;
1245 x->mark.v = m->v;
1246 x->mark.m = m->m;
1247 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1248 xfrm_state_hold(x);
1249 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1250 list_add(&x->km.all, &net->xfrm.state_all);
1251 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1252 h = xfrm_src_hash(net, daddr, saddr, family);
1253 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1254
1255 net->xfrm.state_num++;
1256
1257 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1258 }
1259
1260 return x;
1261 }
1262
1263 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1264
1265 int xfrm_state_add(struct xfrm_state *x)
1266 {
1267 struct net *net = xs_net(x);
1268 struct xfrm_state *x1, *to_put;
1269 int family;
1270 int err;
1271 u32 mark = x->mark.v & x->mark.m;
1272 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1273
1274 family = x->props.family;
1275
1276 to_put = NULL;
1277
1278 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1279
1280 x1 = __xfrm_state_locate(x, use_spi, family);
1281 if (x1) {
1282 to_put = x1;
1283 x1 = NULL;
1284 err = -EEXIST;
1285 goto out;
1286 }
1287
1288 if (use_spi && x->km.seq) {
1289 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1290 if (x1 && ((x1->id.proto != x->id.proto) ||
1291 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1292 to_put = x1;
1293 x1 = NULL;
1294 }
1295 }
1296
1297 if (use_spi && !x1)
1298 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1299 x->props.reqid, x->id.proto,
1300 &x->id.daddr, &x->props.saddr, 0);
1301
1302 __xfrm_state_bump_genids(x);
1303 __xfrm_state_insert(x);
1304 err = 0;
1305
1306 out:
1307 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1308
1309 if (x1) {
1310 xfrm_state_delete(x1);
1311 xfrm_state_put(x1);
1312 }
1313
1314 if (to_put)
1315 xfrm_state_put(to_put);
1316
1317 return err;
1318 }
1319 EXPORT_SYMBOL(xfrm_state_add);
1320
1321 #ifdef CONFIG_XFRM_MIGRATE
1322 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1323 struct xfrm_encap_tmpl *encap)
1324 {
1325 struct net *net = xs_net(orig);
1326 struct xfrm_state *x = xfrm_state_alloc(net);
1327 if (!x)
1328 goto out;
1329
1330 memcpy(&x->id, &orig->id, sizeof(x->id));
1331 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1332 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1333 x->props.mode = orig->props.mode;
1334 x->props.replay_window = orig->props.replay_window;
1335 x->props.reqid = orig->props.reqid;
1336 x->props.family = orig->props.family;
1337 x->props.saddr = orig->props.saddr;
1338
1339 if (orig->aalg) {
1340 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1341 if (!x->aalg)
1342 goto error;
1343 }
1344 x->props.aalgo = orig->props.aalgo;
1345
1346 if (orig->aead) {
1347 x->aead = xfrm_algo_aead_clone(orig->aead);
1348 x->geniv = orig->geniv;
1349 if (!x->aead)
1350 goto error;
1351 }
1352 if (orig->ealg) {
1353 x->ealg = xfrm_algo_clone(orig->ealg);
1354 if (!x->ealg)
1355 goto error;
1356 }
1357 x->props.ealgo = orig->props.ealgo;
1358
1359 if (orig->calg) {
1360 x->calg = xfrm_algo_clone(orig->calg);
1361 if (!x->calg)
1362 goto error;
1363 }
1364 x->props.calgo = orig->props.calgo;
1365
1366 if (encap || orig->encap) {
1367 if (encap)
1368 x->encap = kmemdup(encap, sizeof(*x->encap),
1369 GFP_KERNEL);
1370 else
1371 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1372 GFP_KERNEL);
1373
1374 if (!x->encap)
1375 goto error;
1376 }
1377
1378 if (orig->coaddr) {
1379 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1380 GFP_KERNEL);
1381 if (!x->coaddr)
1382 goto error;
1383 }
1384
1385 if (orig->replay_esn) {
1386 if (xfrm_replay_clone(x, orig))
1387 goto error;
1388 }
1389
1390 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1391
1392 if (xfrm_init_state(x) < 0)
1393 goto error;
1394
1395 x->props.flags = orig->props.flags;
1396 x->props.extra_flags = orig->props.extra_flags;
1397
1398 x->tfcpad = orig->tfcpad;
1399 x->replay_maxdiff = orig->replay_maxdiff;
1400 x->replay_maxage = orig->replay_maxage;
1401 x->curlft.add_time = orig->curlft.add_time;
1402 x->km.state = orig->km.state;
1403 x->km.seq = orig->km.seq;
1404 x->replay = orig->replay;
1405 x->preplay = orig->preplay;
1406
1407 return x;
1408
1409 error:
1410 xfrm_state_put(x);
1411 out:
1412 return NULL;
1413 }
1414
1415 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1416 {
1417 unsigned int h;
1418 struct xfrm_state *x = NULL;
1419
1420 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1421
1422 if (m->reqid) {
1423 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1424 m->reqid, m->old_family);
1425 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1426 if (x->props.mode != m->mode ||
1427 x->id.proto != m->proto)
1428 continue;
1429 if (m->reqid && x->props.reqid != m->reqid)
1430 continue;
1431 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1432 m->old_family) ||
1433 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1434 m->old_family))
1435 continue;
1436 xfrm_state_hold(x);
1437 break;
1438 }
1439 } else {
1440 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1441 m->old_family);
1442 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1443 if (x->props.mode != m->mode ||
1444 x->id.proto != m->proto)
1445 continue;
1446 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1447 m->old_family) ||
1448 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1449 m->old_family))
1450 continue;
1451 xfrm_state_hold(x);
1452 break;
1453 }
1454 }
1455
1456 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1457
1458 return x;
1459 }
1460 EXPORT_SYMBOL(xfrm_migrate_state_find);
1461
1462 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1463 struct xfrm_migrate *m,
1464 struct xfrm_encap_tmpl *encap)
1465 {
1466 struct xfrm_state *xc;
1467
1468 xc = xfrm_state_clone(x, encap);
1469 if (!xc)
1470 return NULL;
1471
1472 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1473 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1474
1475 /* add state */
1476 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1477 /* a care is needed when the destination address of the
1478 state is to be updated as it is a part of triplet */
1479 xfrm_state_insert(xc);
1480 } else {
1481 if (xfrm_state_add(xc) < 0)
1482 goto error;
1483 }
1484
1485 return xc;
1486 error:
1487 xfrm_state_put(xc);
1488 return NULL;
1489 }
1490 EXPORT_SYMBOL(xfrm_state_migrate);
1491 #endif
1492
1493 int xfrm_state_update(struct xfrm_state *x)
1494 {
1495 struct xfrm_state *x1, *to_put;
1496 int err;
1497 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1498 struct net *net = xs_net(x);
1499
1500 to_put = NULL;
1501
1502 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1503 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1504
1505 err = -ESRCH;
1506 if (!x1)
1507 goto out;
1508
1509 if (xfrm_state_kern(x1)) {
1510 to_put = x1;
1511 err = -EEXIST;
1512 goto out;
1513 }
1514
1515 if (x1->km.state == XFRM_STATE_ACQ) {
1516 __xfrm_state_insert(x);
1517 x = NULL;
1518 }
1519 err = 0;
1520
1521 out:
1522 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1523
1524 if (to_put)
1525 xfrm_state_put(to_put);
1526
1527 if (err)
1528 return err;
1529
1530 if (!x) {
1531 xfrm_state_delete(x1);
1532 xfrm_state_put(x1);
1533 return 0;
1534 }
1535
1536 err = -EINVAL;
1537 spin_lock_bh(&x1->lock);
1538 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1539 if (x->encap && x1->encap &&
1540 x->encap->encap_type == x1->encap->encap_type)
1541 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1542 else if (x->encap || x1->encap)
1543 goto fail;
1544
1545 if (x->coaddr && x1->coaddr) {
1546 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1547 }
1548 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1549 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1550 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1551 x1->km.dying = 0;
1552
1553 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1554 if (x1->curlft.use_time)
1555 xfrm_state_check_expire(x1);
1556
1557 err = 0;
1558 x->km.state = XFRM_STATE_DEAD;
1559 __xfrm_state_put(x);
1560 }
1561
1562 fail:
1563 spin_unlock_bh(&x1->lock);
1564
1565 xfrm_state_put(x1);
1566
1567 return err;
1568 }
1569 EXPORT_SYMBOL(xfrm_state_update);
1570
1571 int xfrm_state_check_expire(struct xfrm_state *x)
1572 {
1573 if (!x->curlft.use_time)
1574 x->curlft.use_time = get_seconds();
1575
1576 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1577 x->curlft.packets >= x->lft.hard_packet_limit) {
1578 x->km.state = XFRM_STATE_EXPIRED;
1579 tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1580 return -EINVAL;
1581 }
1582
1583 if (!x->km.dying &&
1584 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1585 x->curlft.packets >= x->lft.soft_packet_limit)) {
1586 x->km.dying = 1;
1587 km_state_expired(x, 0, 0);
1588 }
1589 return 0;
1590 }
1591 EXPORT_SYMBOL(xfrm_state_check_expire);
1592
1593 struct xfrm_state *
1594 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1595 u8 proto, unsigned short family)
1596 {
1597 struct xfrm_state *x;
1598
1599 rcu_read_lock();
1600 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1601 rcu_read_unlock();
1602 return x;
1603 }
1604 EXPORT_SYMBOL(xfrm_state_lookup);
1605
1606 struct xfrm_state *
1607 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1608 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1609 u8 proto, unsigned short family)
1610 {
1611 struct xfrm_state *x;
1612
1613 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1614 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1615 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1616 return x;
1617 }
1618 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1619
1620 struct xfrm_state *
1621 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1622 u8 proto, const xfrm_address_t *daddr,
1623 const xfrm_address_t *saddr, int create, unsigned short family)
1624 {
1625 struct xfrm_state *x;
1626
1627 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1628 x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1629 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1630
1631 return x;
1632 }
1633 EXPORT_SYMBOL(xfrm_find_acq);
1634
1635 #ifdef CONFIG_XFRM_SUB_POLICY
1636 int
1637 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1638 unsigned short family, struct net *net)
1639 {
1640 int i;
1641 int err = 0;
1642 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1643 if (!afinfo)
1644 return -EAFNOSUPPORT;
1645
1646 spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1647 if (afinfo->tmpl_sort)
1648 err = afinfo->tmpl_sort(dst, src, n);
1649 else
1650 for (i = 0; i < n; i++)
1651 dst[i] = src[i];
1652 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1653 rcu_read_unlock();
1654 return err;
1655 }
1656 EXPORT_SYMBOL(xfrm_tmpl_sort);
1657
1658 int
1659 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1660 unsigned short family)
1661 {
1662 int i;
1663 int err = 0;
1664 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1665 struct net *net = xs_net(*src);
1666
1667 if (!afinfo)
1668 return -EAFNOSUPPORT;
1669
1670 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1671 if (afinfo->state_sort)
1672 err = afinfo->state_sort(dst, src, n);
1673 else
1674 for (i = 0; i < n; i++)
1675 dst[i] = src[i];
1676 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1677 rcu_read_unlock();
1678 return err;
1679 }
1680 EXPORT_SYMBOL(xfrm_state_sort);
1681 #endif
1682
1683 /* Silly enough, but I'm lazy to build resolution list */
1684
1685 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1686 {
1687 int i;
1688
1689 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1690 struct xfrm_state *x;
1691
1692 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1693 if (x->km.seq == seq &&
1694 (mark & x->mark.m) == x->mark.v &&
1695 x->km.state == XFRM_STATE_ACQ) {
1696 xfrm_state_hold(x);
1697 return x;
1698 }
1699 }
1700 }
1701 return NULL;
1702 }
1703
1704 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1705 {
1706 struct xfrm_state *x;
1707
1708 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1709 x = __xfrm_find_acq_byseq(net, mark, seq);
1710 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1711 return x;
1712 }
1713 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1714
1715 u32 xfrm_get_acqseq(void)
1716 {
1717 u32 res;
1718 static atomic_t acqseq;
1719
1720 do {
1721 res = atomic_inc_return(&acqseq);
1722 } while (!res);
1723
1724 return res;
1725 }
1726 EXPORT_SYMBOL(xfrm_get_acqseq);
1727
1728 int verify_spi_info(u8 proto, u32 min, u32 max)
1729 {
1730 switch (proto) {
1731 case IPPROTO_AH:
1732 case IPPROTO_ESP:
1733 break;
1734
1735 case IPPROTO_COMP:
1736 /* IPCOMP spi is 16-bits. */
1737 if (max >= 0x10000)
1738 return -EINVAL;
1739 break;
1740
1741 default:
1742 return -EINVAL;
1743 }
1744
1745 if (min > max)
1746 return -EINVAL;
1747
1748 return 0;
1749 }
1750 EXPORT_SYMBOL(verify_spi_info);
1751
1752 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1753 {
1754 struct net *net = xs_net(x);
1755 unsigned int h;
1756 struct xfrm_state *x0;
1757 int err = -ENOENT;
1758 __be32 minspi = htonl(low);
1759 __be32 maxspi = htonl(high);
1760 u32 mark = x->mark.v & x->mark.m;
1761
1762 spin_lock_bh(&x->lock);
1763 if (x->km.state == XFRM_STATE_DEAD)
1764 goto unlock;
1765
1766 err = 0;
1767 if (x->id.spi)
1768 goto unlock;
1769
1770 err = -ENOENT;
1771
1772 if (minspi == maxspi) {
1773 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1774 if (x0) {
1775 xfrm_state_put(x0);
1776 goto unlock;
1777 }
1778 x->id.spi = minspi;
1779 } else {
1780 u32 spi = 0;
1781 for (h = 0; h < high-low+1; h++) {
1782 spi = low + prandom_u32()%(high-low+1);
1783 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1784 if (x0 == NULL) {
1785 x->id.spi = htonl(spi);
1786 break;
1787 }
1788 xfrm_state_put(x0);
1789 }
1790 }
1791 if (x->id.spi) {
1792 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1793 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1794 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1795 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1796
1797 err = 0;
1798 }
1799
1800 unlock:
1801 spin_unlock_bh(&x->lock);
1802
1803 return err;
1804 }
1805 EXPORT_SYMBOL(xfrm_alloc_spi);
1806
1807 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1808 struct xfrm_address_filter *filter)
1809 {
1810 if (filter) {
1811 if ((filter->family == AF_INET ||
1812 filter->family == AF_INET6) &&
1813 x->props.family != filter->family)
1814 return false;
1815
1816 return addr_match(&x->props.saddr, &filter->saddr,
1817 filter->splen) &&
1818 addr_match(&x->id.daddr, &filter->daddr,
1819 filter->dplen);
1820 }
1821 return true;
1822 }
1823
1824 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1825 int (*func)(struct xfrm_state *, int, void*),
1826 void *data)
1827 {
1828 struct xfrm_state *state;
1829 struct xfrm_state_walk *x;
1830 int err = 0;
1831
1832 if (walk->seq != 0 && list_empty(&walk->all))
1833 return 0;
1834
1835 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1836 if (list_empty(&walk->all))
1837 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1838 else
1839 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1840 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1841 if (x->state == XFRM_STATE_DEAD)
1842 continue;
1843 state = container_of(x, struct xfrm_state, km);
1844 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1845 continue;
1846 if (!__xfrm_state_filter_match(state, walk->filter))
1847 continue;
1848 err = func(state, walk->seq, data);
1849 if (err) {
1850 list_move_tail(&walk->all, &x->all);
1851 goto out;
1852 }
1853 walk->seq++;
1854 }
1855 if (walk->seq == 0) {
1856 err = -ENOENT;
1857 goto out;
1858 }
1859 list_del_init(&walk->all);
1860 out:
1861 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1862 return err;
1863 }
1864 EXPORT_SYMBOL(xfrm_state_walk);
1865
1866 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1867 struct xfrm_address_filter *filter)
1868 {
1869 INIT_LIST_HEAD(&walk->all);
1870 walk->proto = proto;
1871 walk->state = XFRM_STATE_DEAD;
1872 walk->seq = 0;
1873 walk->filter = filter;
1874 }
1875 EXPORT_SYMBOL(xfrm_state_walk_init);
1876
1877 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1878 {
1879 kfree(walk->filter);
1880
1881 if (list_empty(&walk->all))
1882 return;
1883
1884 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1885 list_del(&walk->all);
1886 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1887 }
1888 EXPORT_SYMBOL(xfrm_state_walk_done);
1889
1890 static void xfrm_replay_timer_handler(struct timer_list *t)
1891 {
1892 struct xfrm_state *x = from_timer(x, t, rtimer);
1893
1894 spin_lock(&x->lock);
1895
1896 if (x->km.state == XFRM_STATE_VALID) {
1897 if (xfrm_aevent_is_on(xs_net(x)))
1898 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1899 else
1900 x->xflags |= XFRM_TIME_DEFER;
1901 }
1902
1903 spin_unlock(&x->lock);
1904 }
1905
1906 static LIST_HEAD(xfrm_km_list);
1907
1908 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1909 {
1910 struct xfrm_mgr *km;
1911
1912 rcu_read_lock();
1913 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1914 if (km->notify_policy)
1915 km->notify_policy(xp, dir, c);
1916 rcu_read_unlock();
1917 }
1918
1919 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1920 {
1921 struct xfrm_mgr *km;
1922 rcu_read_lock();
1923 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1924 if (km->notify)
1925 km->notify(x, c);
1926 rcu_read_unlock();
1927 }
1928
1929 EXPORT_SYMBOL(km_policy_notify);
1930 EXPORT_SYMBOL(km_state_notify);
1931
1932 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1933 {
1934 struct km_event c;
1935
1936 c.data.hard = hard;
1937 c.portid = portid;
1938 c.event = XFRM_MSG_EXPIRE;
1939 km_state_notify(x, &c);
1940 }
1941
1942 EXPORT_SYMBOL(km_state_expired);
1943 /*
1944 * We send to all registered managers regardless of failure
1945 * We are happy with one success
1946 */
1947 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1948 {
1949 int err = -EINVAL, acqret;
1950 struct xfrm_mgr *km;
1951
1952 rcu_read_lock();
1953 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1954 acqret = km->acquire(x, t, pol);
1955 if (!acqret)
1956 err = acqret;
1957 }
1958 rcu_read_unlock();
1959 return err;
1960 }
1961 EXPORT_SYMBOL(km_query);
1962
1963 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1964 {
1965 int err = -EINVAL;
1966 struct xfrm_mgr *km;
1967
1968 rcu_read_lock();
1969 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1970 if (km->new_mapping)
1971 err = km->new_mapping(x, ipaddr, sport);
1972 if (!err)
1973 break;
1974 }
1975 rcu_read_unlock();
1976 return err;
1977 }
1978 EXPORT_SYMBOL(km_new_mapping);
1979
1980 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1981 {
1982 struct km_event c;
1983
1984 c.data.hard = hard;
1985 c.portid = portid;
1986 c.event = XFRM_MSG_POLEXPIRE;
1987 km_policy_notify(pol, dir, &c);
1988 }
1989 EXPORT_SYMBOL(km_policy_expired);
1990
1991 #ifdef CONFIG_XFRM_MIGRATE
1992 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1993 const struct xfrm_migrate *m, int num_migrate,
1994 const struct xfrm_kmaddress *k,
1995 const struct xfrm_encap_tmpl *encap)
1996 {
1997 int err = -EINVAL;
1998 int ret;
1999 struct xfrm_mgr *km;
2000
2001 rcu_read_lock();
2002 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2003 if (km->migrate) {
2004 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2005 encap);
2006 if (!ret)
2007 err = ret;
2008 }
2009 }
2010 rcu_read_unlock();
2011 return err;
2012 }
2013 EXPORT_SYMBOL(km_migrate);
2014 #endif
2015
2016 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2017 {
2018 int err = -EINVAL;
2019 int ret;
2020 struct xfrm_mgr *km;
2021
2022 rcu_read_lock();
2023 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2024 if (km->report) {
2025 ret = km->report(net, proto, sel, addr);
2026 if (!ret)
2027 err = ret;
2028 }
2029 }
2030 rcu_read_unlock();
2031 return err;
2032 }
2033 EXPORT_SYMBOL(km_report);
2034
2035 bool km_is_alive(const struct km_event *c)
2036 {
2037 struct xfrm_mgr *km;
2038 bool is_alive = false;
2039
2040 rcu_read_lock();
2041 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2042 if (km->is_alive && km->is_alive(c)) {
2043 is_alive = true;
2044 break;
2045 }
2046 }
2047 rcu_read_unlock();
2048
2049 return is_alive;
2050 }
2051 EXPORT_SYMBOL(km_is_alive);
2052
2053 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2054 {
2055 int err;
2056 u8 *data;
2057 struct xfrm_mgr *km;
2058 struct xfrm_policy *pol = NULL;
2059
2060 #ifdef CONFIG_COMPAT
2061 if (in_compat_syscall())
2062 return -EOPNOTSUPP;
2063 #endif
2064
2065 if (!optval && !optlen) {
2066 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2067 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2068 __sk_dst_reset(sk);
2069 return 0;
2070 }
2071
2072 if (optlen <= 0 || optlen > PAGE_SIZE)
2073 return -EMSGSIZE;
2074
2075 data = memdup_user(optval, optlen);
2076 if (IS_ERR(data))
2077 return PTR_ERR(data);
2078
2079 err = -EINVAL;
2080 rcu_read_lock();
2081 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2082 pol = km->compile_policy(sk, optname, data,
2083 optlen, &err);
2084 if (err >= 0)
2085 break;
2086 }
2087 rcu_read_unlock();
2088
2089 if (err >= 0) {
2090 xfrm_sk_policy_insert(sk, err, pol);
2091 xfrm_pol_put(pol);
2092 __sk_dst_reset(sk);
2093 err = 0;
2094 }
2095
2096 kfree(data);
2097 return err;
2098 }
2099 EXPORT_SYMBOL(xfrm_user_policy);
2100
2101 static DEFINE_SPINLOCK(xfrm_km_lock);
2102
2103 int xfrm_register_km(struct xfrm_mgr *km)
2104 {
2105 spin_lock_bh(&xfrm_km_lock);
2106 list_add_tail_rcu(&km->list, &xfrm_km_list);
2107 spin_unlock_bh(&xfrm_km_lock);
2108 return 0;
2109 }
2110 EXPORT_SYMBOL(xfrm_register_km);
2111
2112 int xfrm_unregister_km(struct xfrm_mgr *km)
2113 {
2114 spin_lock_bh(&xfrm_km_lock);
2115 list_del_rcu(&km->list);
2116 spin_unlock_bh(&xfrm_km_lock);
2117 synchronize_rcu();
2118 return 0;
2119 }
2120 EXPORT_SYMBOL(xfrm_unregister_km);
2121
2122 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2123 {
2124 int err = 0;
2125
2126 if (WARN_ON(afinfo->family >= NPROTO))
2127 return -EAFNOSUPPORT;
2128
2129 spin_lock_bh(&xfrm_state_afinfo_lock);
2130 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2131 err = -EEXIST;
2132 else
2133 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2134 spin_unlock_bh(&xfrm_state_afinfo_lock);
2135 return err;
2136 }
2137 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2138
2139 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2140 {
2141 int err = 0, family = afinfo->family;
2142
2143 if (WARN_ON(family >= NPROTO))
2144 return -EAFNOSUPPORT;
2145
2146 spin_lock_bh(&xfrm_state_afinfo_lock);
2147 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2148 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2149 err = -EINVAL;
2150 else
2151 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2152 }
2153 spin_unlock_bh(&xfrm_state_afinfo_lock);
2154 synchronize_rcu();
2155 return err;
2156 }
2157 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2158
2159 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2160 {
2161 if (unlikely(family >= NPROTO))
2162 return NULL;
2163
2164 return rcu_dereference(xfrm_state_afinfo[family]);
2165 }
2166
2167 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2168 {
2169 struct xfrm_state_afinfo *afinfo;
2170 if (unlikely(family >= NPROTO))
2171 return NULL;
2172 rcu_read_lock();
2173 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2174 if (unlikely(!afinfo))
2175 rcu_read_unlock();
2176 return afinfo;
2177 }
2178
2179 void xfrm_flush_gc(void)
2180 {
2181 flush_work(&xfrm_state_gc_work);
2182 }
2183 EXPORT_SYMBOL(xfrm_flush_gc);
2184
2185 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2186 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2187 {
2188 if (x->tunnel) {
2189 struct xfrm_state *t = x->tunnel;
2190
2191 if (atomic_read(&t->tunnel_users) == 2)
2192 xfrm_state_delete(t);
2193 atomic_dec(&t->tunnel_users);
2194 xfrm_state_put(t);
2195 x->tunnel = NULL;
2196 }
2197 }
2198 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2199
2200 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2201 {
2202 const struct xfrm_type *type = READ_ONCE(x->type);
2203
2204 if (x->km.state == XFRM_STATE_VALID &&
2205 type && type->get_mtu)
2206 return type->get_mtu(x, mtu);
2207
2208 return mtu - x->props.header_len;
2209 }
2210
2211 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2212 {
2213 struct xfrm_state_afinfo *afinfo;
2214 struct xfrm_mode *inner_mode;
2215 int family = x->props.family;
2216 int err;
2217
2218 err = -EAFNOSUPPORT;
2219 afinfo = xfrm_state_get_afinfo(family);
2220 if (!afinfo)
2221 goto error;
2222
2223 err = 0;
2224 if (afinfo->init_flags)
2225 err = afinfo->init_flags(x);
2226
2227 rcu_read_unlock();
2228
2229 if (err)
2230 goto error;
2231
2232 err = -EPROTONOSUPPORT;
2233
2234 if (x->sel.family != AF_UNSPEC) {
2235 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2236 if (inner_mode == NULL)
2237 goto error;
2238
2239 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2240 family != x->sel.family) {
2241 xfrm_put_mode(inner_mode);
2242 goto error;
2243 }
2244
2245 x->inner_mode = inner_mode;
2246 } else {
2247 struct xfrm_mode *inner_mode_iaf;
2248 int iafamily = AF_INET;
2249
2250 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2251 if (inner_mode == NULL)
2252 goto error;
2253
2254 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2255 xfrm_put_mode(inner_mode);
2256 goto error;
2257 }
2258 x->inner_mode = inner_mode;
2259
2260 if (x->props.family == AF_INET)
2261 iafamily = AF_INET6;
2262
2263 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2264 if (inner_mode_iaf) {
2265 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2266 x->inner_mode_iaf = inner_mode_iaf;
2267 else
2268 xfrm_put_mode(inner_mode_iaf);
2269 }
2270 }
2271
2272 x->type = xfrm_get_type(x->id.proto, family);
2273 if (x->type == NULL)
2274 goto error;
2275
2276 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2277
2278 err = x->type->init_state(x);
2279 if (err)
2280 goto error;
2281
2282 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2283 if (x->outer_mode == NULL) {
2284 err = -EPROTONOSUPPORT;
2285 goto error;
2286 }
2287
2288 if (init_replay) {
2289 err = xfrm_init_replay(x);
2290 if (err)
2291 goto error;
2292 }
2293
2294 error:
2295 return err;
2296 }
2297
2298 EXPORT_SYMBOL(__xfrm_init_state);
2299
2300 int xfrm_init_state(struct xfrm_state *x)
2301 {
2302 int err;
2303
2304 err = __xfrm_init_state(x, true, false);
2305 if (!err)
2306 x->km.state = XFRM_STATE_VALID;
2307
2308 return err;
2309 }
2310
2311 EXPORT_SYMBOL(xfrm_init_state);
2312
2313 int __net_init xfrm_state_init(struct net *net)
2314 {
2315 unsigned int sz;
2316
2317 if (net_eq(net, &init_net))
2318 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2319 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2320
2321 INIT_LIST_HEAD(&net->xfrm.state_all);
2322
2323 sz = sizeof(struct hlist_head) * 8;
2324
2325 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2326 if (!net->xfrm.state_bydst)
2327 goto out_bydst;
2328 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2329 if (!net->xfrm.state_bysrc)
2330 goto out_bysrc;
2331 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2332 if (!net->xfrm.state_byspi)
2333 goto out_byspi;
2334 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2335
2336 net->xfrm.state_num = 0;
2337 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2338 spin_lock_init(&net->xfrm.xfrm_state_lock);
2339 return 0;
2340
2341 out_byspi:
2342 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2343 out_bysrc:
2344 xfrm_hash_free(net->xfrm.state_bydst, sz);
2345 out_bydst:
2346 return -ENOMEM;
2347 }
2348
2349 void xfrm_state_fini(struct net *net)
2350 {
2351 unsigned int sz;
2352
2353 flush_work(&net->xfrm.state_hash_work);
2354 xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2355 flush_work(&xfrm_state_gc_work);
2356
2357 WARN_ON(!list_empty(&net->xfrm.state_all));
2358
2359 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2360 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2361 xfrm_hash_free(net->xfrm.state_byspi, sz);
2362 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2363 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2364 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2365 xfrm_hash_free(net->xfrm.state_bydst, sz);
2366 }
2367
2368 #ifdef CONFIG_AUDITSYSCALL
2369 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2370 struct audit_buffer *audit_buf)
2371 {
2372 struct xfrm_sec_ctx *ctx = x->security;
2373 u32 spi = ntohl(x->id.spi);
2374
2375 if (ctx)
2376 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2377 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2378
2379 switch (x->props.family) {
2380 case AF_INET:
2381 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2382 &x->props.saddr.a4, &x->id.daddr.a4);
2383 break;
2384 case AF_INET6:
2385 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2386 x->props.saddr.a6, x->id.daddr.a6);
2387 break;
2388 }
2389
2390 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2391 }
2392
2393 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2394 struct audit_buffer *audit_buf)
2395 {
2396 const struct iphdr *iph4;
2397 const struct ipv6hdr *iph6;
2398
2399 switch (family) {
2400 case AF_INET:
2401 iph4 = ip_hdr(skb);
2402 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2403 &iph4->saddr, &iph4->daddr);
2404 break;
2405 case AF_INET6:
2406 iph6 = ipv6_hdr(skb);
2407 audit_log_format(audit_buf,
2408 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2409 &iph6->saddr, &iph6->daddr,
2410 iph6->flow_lbl[0] & 0x0f,
2411 iph6->flow_lbl[1],
2412 iph6->flow_lbl[2]);
2413 break;
2414 }
2415 }
2416
2417 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2418 {
2419 struct audit_buffer *audit_buf;
2420
2421 audit_buf = xfrm_audit_start("SAD-add");
2422 if (audit_buf == NULL)
2423 return;
2424 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2425 xfrm_audit_helper_sainfo(x, audit_buf);
2426 audit_log_format(audit_buf, " res=%u", result);
2427 audit_log_end(audit_buf);
2428 }
2429 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2430
2431 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2432 {
2433 struct audit_buffer *audit_buf;
2434
2435 audit_buf = xfrm_audit_start("SAD-delete");
2436 if (audit_buf == NULL)
2437 return;
2438 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2439 xfrm_audit_helper_sainfo(x, audit_buf);
2440 audit_log_format(audit_buf, " res=%u", result);
2441 audit_log_end(audit_buf);
2442 }
2443 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2444
2445 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2446 struct sk_buff *skb)
2447 {
2448 struct audit_buffer *audit_buf;
2449 u32 spi;
2450
2451 audit_buf = xfrm_audit_start("SA-replay-overflow");
2452 if (audit_buf == NULL)
2453 return;
2454 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2455 /* don't record the sequence number because it's inherent in this kind
2456 * of audit message */
2457 spi = ntohl(x->id.spi);
2458 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2459 audit_log_end(audit_buf);
2460 }
2461 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2462
2463 void xfrm_audit_state_replay(struct xfrm_state *x,
2464 struct sk_buff *skb, __be32 net_seq)
2465 {
2466 struct audit_buffer *audit_buf;
2467 u32 spi;
2468
2469 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2470 if (audit_buf == NULL)
2471 return;
2472 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2473 spi = ntohl(x->id.spi);
2474 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2475 spi, spi, ntohl(net_seq));
2476 audit_log_end(audit_buf);
2477 }
2478 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2479
2480 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2481 {
2482 struct audit_buffer *audit_buf;
2483
2484 audit_buf = xfrm_audit_start("SA-notfound");
2485 if (audit_buf == NULL)
2486 return;
2487 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2488 audit_log_end(audit_buf);
2489 }
2490 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2491
2492 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2493 __be32 net_spi, __be32 net_seq)
2494 {
2495 struct audit_buffer *audit_buf;
2496 u32 spi;
2497
2498 audit_buf = xfrm_audit_start("SA-notfound");
2499 if (audit_buf == NULL)
2500 return;
2501 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2502 spi = ntohl(net_spi);
2503 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2504 spi, spi, ntohl(net_seq));
2505 audit_log_end(audit_buf);
2506 }
2507 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2508
2509 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2510 struct sk_buff *skb, u8 proto)
2511 {
2512 struct audit_buffer *audit_buf;
2513 __be32 net_spi;
2514 __be32 net_seq;
2515
2516 audit_buf = xfrm_audit_start("SA-icv-failure");
2517 if (audit_buf == NULL)
2518 return;
2519 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2520 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2521 u32 spi = ntohl(net_spi);
2522 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2523 spi, spi, ntohl(net_seq));
2524 }
2525 audit_log_end(audit_buf);
2526 }
2527 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2528 #endif /* CONFIG_AUDITSYSCALL */
Linux with added FPC-III support