?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf tools: Don't clone maps from parent when synthesizing forks
[linux/fpc-iii.git] / net / xfrm / xfrm_state.c
blobb669262682c9763e7c863d6bb77f44ed34402cce
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 time64_t now = ktime_get_real_seconds();
479 time64_t next = TIME64_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 != TIME64_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 = ktime_get_real_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
741 return err;
742 }
743 EXPORT_SYMBOL(xfrm_state_flush);
744
745 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
746 {
747 int i, err = 0, cnt = 0;
748
749 spin_lock_bh(&net->xfrm.xfrm_state_lock);
750 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
751 if (err)
752 goto out;
753
754 err = -ESRCH;
755 for (i = 0; i <= net->xfrm.state_hmask; i++) {
756 struct xfrm_state *x;
757 struct xfrm_state_offload *xso;
758 restart:
759 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
760 xso = &x->xso;
761
762 if (!xfrm_state_kern(x) && xso->dev == dev) {
763 xfrm_state_hold(x);
764 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
765
766 err = xfrm_state_delete(x);
767 xfrm_audit_state_delete(x, err ? 0 : 1,
768 task_valid);
769 xfrm_state_put(x);
770 if (!err)
771 cnt++;
772
773 spin_lock_bh(&net->xfrm.xfrm_state_lock);
774 goto restart;
775 }
776 }
777 }
778 if (cnt)
779 err = 0;
780
781 out:
782 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
783 return err;
784 }
785 EXPORT_SYMBOL(xfrm_dev_state_flush);
786
787 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
788 {
789 spin_lock_bh(&net->xfrm.xfrm_state_lock);
790 si->sadcnt = net->xfrm.state_num;
791 si->sadhcnt = net->xfrm.state_hmask;
792 si->sadhmcnt = xfrm_state_hashmax;
793 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
794 }
795 EXPORT_SYMBOL(xfrm_sad_getinfo);
796
797 static void
798 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
799 const struct xfrm_tmpl *tmpl,
800 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
801 unsigned short family)
802 {
803 struct xfrm_state_afinfo *afinfo = xfrm_state_afinfo_get_rcu(family);
804
805 if (!afinfo)
806 return;
807
808 afinfo->init_tempsel(&x->sel, fl);
809
810 if (family != tmpl->encap_family) {
811 afinfo = xfrm_state_afinfo_get_rcu(tmpl->encap_family);
812 if (!afinfo)
813 return;
814 }
815 afinfo->init_temprop(x, tmpl, daddr, saddr);
816 }
817
818 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
819 const xfrm_address_t *daddr,
820 __be32 spi, u8 proto,
821 unsigned short family)
822 {
823 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
824 struct xfrm_state *x;
825
826 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
827 if (x->props.family != family ||
828 x->id.spi != spi ||
829 x->id.proto != proto ||
830 !xfrm_addr_equal(&x->id.daddr, daddr, family))
831 continue;
832
833 if ((mark & x->mark.m) != x->mark.v)
834 continue;
835 if (!xfrm_state_hold_rcu(x))
836 continue;
837 return x;
838 }
839
840 return NULL;
841 }
842
843 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
844 const xfrm_address_t *daddr,
845 const xfrm_address_t *saddr,
846 u8 proto, unsigned short family)
847 {
848 unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
849 struct xfrm_state *x;
850
851 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
852 if (x->props.family != family ||
853 x->id.proto != proto ||
854 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
855 !xfrm_addr_equal(&x->props.saddr, saddr, family))
856 continue;
857
858 if ((mark & x->mark.m) != x->mark.v)
859 continue;
860 if (!xfrm_state_hold_rcu(x))
861 continue;
862 return x;
863 }
864
865 return NULL;
866 }
867
868 static inline struct xfrm_state *
869 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
870 {
871 struct net *net = xs_net(x);
872 u32 mark = x->mark.v & x->mark.m;
873
874 if (use_spi)
875 return __xfrm_state_lookup(net, mark, &x->id.daddr,
876 x->id.spi, x->id.proto, family);
877 else
878 return __xfrm_state_lookup_byaddr(net, mark,
879 &x->id.daddr,
880 &x->props.saddr,
881 x->id.proto, family);
882 }
883
884 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
885 {
886 if (have_hash_collision &&
887 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
888 net->xfrm.state_num > net->xfrm.state_hmask)
889 schedule_work(&net->xfrm.state_hash_work);
890 }
891
892 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
893 const struct flowi *fl, unsigned short family,
894 struct xfrm_state **best, int *acq_in_progress,
895 int *error)
896 {
897 /* Resolution logic:
898 * 1. There is a valid state with matching selector. Done.
899 * 2. Valid state with inappropriate selector. Skip.
900 *
901 * Entering area of "sysdeps".
902 *
903 * 3. If state is not valid, selector is temporary, it selects
904 * only session which triggered previous resolution. Key
905 * manager will do something to install a state with proper
906 * selector.
907 */
908 if (x->km.state == XFRM_STATE_VALID) {
909 if ((x->sel.family &&
910 !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
911 !security_xfrm_state_pol_flow_match(x, pol, fl))
912 return;
913
914 if (!*best ||
915 (*best)->km.dying > x->km.dying ||
916 ((*best)->km.dying == x->km.dying &&
917 (*best)->curlft.add_time < x->curlft.add_time))
918 *best = x;
919 } else if (x->km.state == XFRM_STATE_ACQ) {
920 *acq_in_progress = 1;
921 } else if (x->km.state == XFRM_STATE_ERROR ||
922 x->km.state == XFRM_STATE_EXPIRED) {
923 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
924 security_xfrm_state_pol_flow_match(x, pol, fl))
925 *error = -ESRCH;
926 }
927 }
928
929 struct xfrm_state *
930 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
931 const struct flowi *fl, struct xfrm_tmpl *tmpl,
932 struct xfrm_policy *pol, int *err,
933 unsigned short family, u32 if_id)
934 {
935 static xfrm_address_t saddr_wildcard = { };
936 struct net *net = xp_net(pol);
937 unsigned int h, h_wildcard;
938 struct xfrm_state *x, *x0, *to_put;
939 int acquire_in_progress = 0;
940 int error = 0;
941 struct xfrm_state *best = NULL;
942 u32 mark = pol->mark.v & pol->mark.m;
943 unsigned short encap_family = tmpl->encap_family;
944 unsigned int sequence;
945 struct km_event c;
946
947 to_put = NULL;
948
949 sequence = read_seqcount_begin(&xfrm_state_hash_generation);
950
951 rcu_read_lock();
952 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
953 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
954 if (x->props.family == encap_family &&
955 x->props.reqid == tmpl->reqid &&
956 (mark & x->mark.m) == x->mark.v &&
957 x->if_id == if_id &&
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->if_id == if_id &&
975 !(x->props.flags & XFRM_STATE_WILDRECV) &&
976 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
977 tmpl->mode == x->props.mode &&
978 tmpl->id.proto == x->id.proto &&
979 (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
980 xfrm_state_look_at(pol, x, fl, encap_family,
981 &best, &acquire_in_progress, &error);
982 }
983
984 found:
985 x = best;
986 if (!x && !error && !acquire_in_progress) {
987 if (tmpl->id.spi &&
988 (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
989 tmpl->id.proto, encap_family)) != NULL) {
990 to_put = x0;
991 error = -EEXIST;
992 goto out;
993 }
994
995 c.net = net;
996 /* If the KMs have no listeners (yet...), avoid allocating an SA
997 * for each and every packet - garbage collection might not
998 * handle the flood.
999 */
1000 if (!km_is_alive(&c)) {
1001 error = -ESRCH;
1002 goto out;
1003 }
1004
1005 x = xfrm_state_alloc(net);
1006 if (x == NULL) {
1007 error = -ENOMEM;
1008 goto out;
1009 }
1010 /* Initialize temporary state matching only
1011 * to current session. */
1012 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1013 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1014 x->if_id = if_id;
1015
1016 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1017 if (error) {
1018 x->km.state = XFRM_STATE_DEAD;
1019 to_put = x;
1020 x = NULL;
1021 goto out;
1022 }
1023
1024 if (km_query(x, tmpl, pol) == 0) {
1025 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1026 x->km.state = XFRM_STATE_ACQ;
1027 list_add(&x->km.all, &net->xfrm.state_all);
1028 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1029 h = xfrm_src_hash(net, daddr, saddr, encap_family);
1030 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1031 if (x->id.spi) {
1032 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1033 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1034 }
1035 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1036 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1037 net->xfrm.state_num++;
1038 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1039 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1040 } else {
1041 x->km.state = XFRM_STATE_DEAD;
1042 to_put = x;
1043 x = NULL;
1044 error = -ESRCH;
1045 }
1046 }
1047 out:
1048 if (x) {
1049 if (!xfrm_state_hold_rcu(x)) {
1050 *err = -EAGAIN;
1051 x = NULL;
1052 }
1053 } else {
1054 *err = acquire_in_progress ? -EAGAIN : error;
1055 }
1056 rcu_read_unlock();
1057 if (to_put)
1058 xfrm_state_put(to_put);
1059
1060 if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
1061 *err = -EAGAIN;
1062 if (x) {
1063 xfrm_state_put(x);
1064 x = NULL;
1065 }
1066 }
1067
1068 return x;
1069 }
1070
1071 struct xfrm_state *
1072 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1073 xfrm_address_t *daddr, xfrm_address_t *saddr,
1074 unsigned short family, u8 mode, u8 proto, u32 reqid)
1075 {
1076 unsigned int h;
1077 struct xfrm_state *rx = NULL, *x = NULL;
1078
1079 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1080 h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1081 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1082 if (x->props.family == family &&
1083 x->props.reqid == reqid &&
1084 (mark & x->mark.m) == x->mark.v &&
1085 x->if_id == if_id &&
1086 !(x->props.flags & XFRM_STATE_WILDRECV) &&
1087 xfrm_state_addr_check(x, daddr, saddr, family) &&
1088 mode == x->props.mode &&
1089 proto == x->id.proto &&
1090 x->km.state == XFRM_STATE_VALID) {
1091 rx = x;
1092 break;
1093 }
1094 }
1095
1096 if (rx)
1097 xfrm_state_hold(rx);
1098 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1099
1100
1101 return rx;
1102 }
1103 EXPORT_SYMBOL(xfrm_stateonly_find);
1104
1105 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1106 unsigned short family)
1107 {
1108 struct xfrm_state *x;
1109 struct xfrm_state_walk *w;
1110
1111 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1112 list_for_each_entry(w, &net->xfrm.state_all, all) {
1113 x = container_of(w, struct xfrm_state, km);
1114 if (x->props.family != family ||
1115 x->id.spi != spi)
1116 continue;
1117
1118 xfrm_state_hold(x);
1119 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1120 return x;
1121 }
1122 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1123 return NULL;
1124 }
1125 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1126
1127 static void __xfrm_state_insert(struct xfrm_state *x)
1128 {
1129 struct net *net = xs_net(x);
1130 unsigned int h;
1131
1132 list_add(&x->km.all, &net->xfrm.state_all);
1133
1134 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1135 x->props.reqid, x->props.family);
1136 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1137
1138 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1139 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1140
1141 if (x->id.spi) {
1142 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1143 x->props.family);
1144
1145 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1146 }
1147
1148 tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1149 if (x->replay_maxage)
1150 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1151
1152 net->xfrm.state_num++;
1153
1154 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1155 }
1156
1157 /* net->xfrm.xfrm_state_lock is held */
1158 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1159 {
1160 struct net *net = xs_net(xnew);
1161 unsigned short family = xnew->props.family;
1162 u32 reqid = xnew->props.reqid;
1163 struct xfrm_state *x;
1164 unsigned int h;
1165 u32 mark = xnew->mark.v & xnew->mark.m;
1166 u32 if_id = xnew->if_id;
1167
1168 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1169 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1170 if (x->props.family == family &&
1171 x->props.reqid == reqid &&
1172 x->if_id == if_id &&
1173 (mark & x->mark.m) == x->mark.v &&
1174 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1175 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1176 x->genid++;
1177 }
1178 }
1179
1180 void xfrm_state_insert(struct xfrm_state *x)
1181 {
1182 struct net *net = xs_net(x);
1183
1184 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1185 __xfrm_state_bump_genids(x);
1186 __xfrm_state_insert(x);
1187 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1188 }
1189 EXPORT_SYMBOL(xfrm_state_insert);
1190
1191 /* net->xfrm.xfrm_state_lock is held */
1192 static struct xfrm_state *__find_acq_core(struct net *net,
1193 const struct xfrm_mark *m,
1194 unsigned short family, u8 mode,
1195 u32 reqid, u32 if_id, u8 proto,
1196 const xfrm_address_t *daddr,
1197 const xfrm_address_t *saddr,
1198 int create)
1199 {
1200 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1201 struct xfrm_state *x;
1202 u32 mark = m->v & m->m;
1203
1204 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1205 if (x->props.reqid != reqid ||
1206 x->props.mode != mode ||
1207 x->props.family != family ||
1208 x->km.state != XFRM_STATE_ACQ ||
1209 x->id.spi != 0 ||
1210 x->id.proto != proto ||
1211 (mark & x->mark.m) != x->mark.v ||
1212 !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1213 !xfrm_addr_equal(&x->props.saddr, saddr, family))
1214 continue;
1215
1216 xfrm_state_hold(x);
1217 return x;
1218 }
1219
1220 if (!create)
1221 return NULL;
1222
1223 x = xfrm_state_alloc(net);
1224 if (likely(x)) {
1225 switch (family) {
1226 case AF_INET:
1227 x->sel.daddr.a4 = daddr->a4;
1228 x->sel.saddr.a4 = saddr->a4;
1229 x->sel.prefixlen_d = 32;
1230 x->sel.prefixlen_s = 32;
1231 x->props.saddr.a4 = saddr->a4;
1232 x->id.daddr.a4 = daddr->a4;
1233 break;
1234
1235 case AF_INET6:
1236 x->sel.daddr.in6 = daddr->in6;
1237 x->sel.saddr.in6 = saddr->in6;
1238 x->sel.prefixlen_d = 128;
1239 x->sel.prefixlen_s = 128;
1240 x->props.saddr.in6 = saddr->in6;
1241 x->id.daddr.in6 = daddr->in6;
1242 break;
1243 }
1244
1245 x->km.state = XFRM_STATE_ACQ;
1246 x->id.proto = proto;
1247 x->props.family = family;
1248 x->props.mode = mode;
1249 x->props.reqid = reqid;
1250 x->if_id = if_id;
1251 x->mark.v = m->v;
1252 x->mark.m = m->m;
1253 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1254 xfrm_state_hold(x);
1255 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1256 list_add(&x->km.all, &net->xfrm.state_all);
1257 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1258 h = xfrm_src_hash(net, daddr, saddr, family);
1259 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1260
1261 net->xfrm.state_num++;
1262
1263 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1264 }
1265
1266 return x;
1267 }
1268
1269 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1270
1271 int xfrm_state_add(struct xfrm_state *x)
1272 {
1273 struct net *net = xs_net(x);
1274 struct xfrm_state *x1, *to_put;
1275 int family;
1276 int err;
1277 u32 mark = x->mark.v & x->mark.m;
1278 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1279
1280 family = x->props.family;
1281
1282 to_put = NULL;
1283
1284 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1285
1286 x1 = __xfrm_state_locate(x, use_spi, family);
1287 if (x1) {
1288 to_put = x1;
1289 x1 = NULL;
1290 err = -EEXIST;
1291 goto out;
1292 }
1293
1294 if (use_spi && x->km.seq) {
1295 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1296 if (x1 && ((x1->id.proto != x->id.proto) ||
1297 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1298 to_put = x1;
1299 x1 = NULL;
1300 }
1301 }
1302
1303 if (use_spi && !x1)
1304 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1305 x->props.reqid, x->if_id, x->id.proto,
1306 &x->id.daddr, &x->props.saddr, 0);
1307
1308 __xfrm_state_bump_genids(x);
1309 __xfrm_state_insert(x);
1310 err = 0;
1311
1312 out:
1313 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1314
1315 if (x1) {
1316 xfrm_state_delete(x1);
1317 xfrm_state_put(x1);
1318 }
1319
1320 if (to_put)
1321 xfrm_state_put(to_put);
1322
1323 return err;
1324 }
1325 EXPORT_SYMBOL(xfrm_state_add);
1326
1327 #ifdef CONFIG_XFRM_MIGRATE
1328 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1329 struct xfrm_encap_tmpl *encap)
1330 {
1331 struct net *net = xs_net(orig);
1332 struct xfrm_state *x = xfrm_state_alloc(net);
1333 if (!x)
1334 goto out;
1335
1336 memcpy(&x->id, &orig->id, sizeof(x->id));
1337 memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1338 memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1339 x->props.mode = orig->props.mode;
1340 x->props.replay_window = orig->props.replay_window;
1341 x->props.reqid = orig->props.reqid;
1342 x->props.family = orig->props.family;
1343 x->props.saddr = orig->props.saddr;
1344
1345 if (orig->aalg) {
1346 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1347 if (!x->aalg)
1348 goto error;
1349 }
1350 x->props.aalgo = orig->props.aalgo;
1351
1352 if (orig->aead) {
1353 x->aead = xfrm_algo_aead_clone(orig->aead);
1354 x->geniv = orig->geniv;
1355 if (!x->aead)
1356 goto error;
1357 }
1358 if (orig->ealg) {
1359 x->ealg = xfrm_algo_clone(orig->ealg);
1360 if (!x->ealg)
1361 goto error;
1362 }
1363 x->props.ealgo = orig->props.ealgo;
1364
1365 if (orig->calg) {
1366 x->calg = xfrm_algo_clone(orig->calg);
1367 if (!x->calg)
1368 goto error;
1369 }
1370 x->props.calgo = orig->props.calgo;
1371
1372 if (encap || orig->encap) {
1373 if (encap)
1374 x->encap = kmemdup(encap, sizeof(*x->encap),
1375 GFP_KERNEL);
1376 else
1377 x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1378 GFP_KERNEL);
1379
1380 if (!x->encap)
1381 goto error;
1382 }
1383
1384 if (orig->coaddr) {
1385 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1386 GFP_KERNEL);
1387 if (!x->coaddr)
1388 goto error;
1389 }
1390
1391 if (orig->replay_esn) {
1392 if (xfrm_replay_clone(x, orig))
1393 goto error;
1394 }
1395
1396 memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1397
1398 if (xfrm_init_state(x) < 0)
1399 goto error;
1400
1401 x->props.flags = orig->props.flags;
1402 x->props.extra_flags = orig->props.extra_flags;
1403
1404 x->if_id = orig->if_id;
1405 x->tfcpad = orig->tfcpad;
1406 x->replay_maxdiff = orig->replay_maxdiff;
1407 x->replay_maxage = orig->replay_maxage;
1408 x->curlft.add_time = orig->curlft.add_time;
1409 x->km.state = orig->km.state;
1410 x->km.seq = orig->km.seq;
1411 x->replay = orig->replay;
1412 x->preplay = orig->preplay;
1413
1414 return x;
1415
1416 error:
1417 xfrm_state_put(x);
1418 out:
1419 return NULL;
1420 }
1421
1422 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1423 {
1424 unsigned int h;
1425 struct xfrm_state *x = NULL;
1426
1427 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1428
1429 if (m->reqid) {
1430 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1431 m->reqid, m->old_family);
1432 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1433 if (x->props.mode != m->mode ||
1434 x->id.proto != m->proto)
1435 continue;
1436 if (m->reqid && x->props.reqid != m->reqid)
1437 continue;
1438 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1439 m->old_family) ||
1440 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1441 m->old_family))
1442 continue;
1443 xfrm_state_hold(x);
1444 break;
1445 }
1446 } else {
1447 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1448 m->old_family);
1449 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1450 if (x->props.mode != m->mode ||
1451 x->id.proto != m->proto)
1452 continue;
1453 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1454 m->old_family) ||
1455 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1456 m->old_family))
1457 continue;
1458 xfrm_state_hold(x);
1459 break;
1460 }
1461 }
1462
1463 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1464
1465 return x;
1466 }
1467 EXPORT_SYMBOL(xfrm_migrate_state_find);
1468
1469 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1470 struct xfrm_migrate *m,
1471 struct xfrm_encap_tmpl *encap)
1472 {
1473 struct xfrm_state *xc;
1474
1475 xc = xfrm_state_clone(x, encap);
1476 if (!xc)
1477 return NULL;
1478
1479 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1480 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1481
1482 /* add state */
1483 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1484 /* a care is needed when the destination address of the
1485 state is to be updated as it is a part of triplet */
1486 xfrm_state_insert(xc);
1487 } else {
1488 if (xfrm_state_add(xc) < 0)
1489 goto error;
1490 }
1491
1492 return xc;
1493 error:
1494 xfrm_state_put(xc);
1495 return NULL;
1496 }
1497 EXPORT_SYMBOL(xfrm_state_migrate);
1498 #endif
1499
1500 int xfrm_state_update(struct xfrm_state *x)
1501 {
1502 struct xfrm_state *x1, *to_put;
1503 int err;
1504 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1505 struct net *net = xs_net(x);
1506
1507 to_put = NULL;
1508
1509 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1510 x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1511
1512 err = -ESRCH;
1513 if (!x1)
1514 goto out;
1515
1516 if (xfrm_state_kern(x1)) {
1517 to_put = x1;
1518 err = -EEXIST;
1519 goto out;
1520 }
1521
1522 if (x1->km.state == XFRM_STATE_ACQ) {
1523 __xfrm_state_insert(x);
1524 x = NULL;
1525 }
1526 err = 0;
1527
1528 out:
1529 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1530
1531 if (to_put)
1532 xfrm_state_put(to_put);
1533
1534 if (err)
1535 return err;
1536
1537 if (!x) {
1538 xfrm_state_delete(x1);
1539 xfrm_state_put(x1);
1540 return 0;
1541 }
1542
1543 err = -EINVAL;
1544 spin_lock_bh(&x1->lock);
1545 if (likely(x1->km.state == XFRM_STATE_VALID)) {
1546 if (x->encap && x1->encap &&
1547 x->encap->encap_type == x1->encap->encap_type)
1548 memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1549 else if (x->encap || x1->encap)
1550 goto fail;
1551
1552 if (x->coaddr && x1->coaddr) {
1553 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1554 }
1555 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1556 memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1557 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1558 x1->km.dying = 0;
1559
1560 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1561 if (x1->curlft.use_time)
1562 xfrm_state_check_expire(x1);
1563
1564 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1565 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1566
1567 if (x->props.smark.m || x->props.smark.v)
1568 x1->props.smark = x->props.smark;
1569
1570 if (x->if_id)
1571 x1->if_id = x->if_id;
1572
1573 __xfrm_state_bump_genids(x1);
1574 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1575 }
1576
1577 err = 0;
1578 x->km.state = XFRM_STATE_DEAD;
1579 __xfrm_state_put(x);
1580 }
1581
1582 fail:
1583 spin_unlock_bh(&x1->lock);
1584
1585 xfrm_state_put(x1);
1586
1587 return err;
1588 }
1589 EXPORT_SYMBOL(xfrm_state_update);
1590
1591 int xfrm_state_check_expire(struct xfrm_state *x)
1592 {
1593 if (!x->curlft.use_time)
1594 x->curlft.use_time = ktime_get_real_seconds();
1595
1596 if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1597 x->curlft.packets >= x->lft.hard_packet_limit) {
1598 x->km.state = XFRM_STATE_EXPIRED;
1599 tasklet_hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL);
1600 return -EINVAL;
1601 }
1602
1603 if (!x->km.dying &&
1604 (x->curlft.bytes >= x->lft.soft_byte_limit ||
1605 x->curlft.packets >= x->lft.soft_packet_limit)) {
1606 x->km.dying = 1;
1607 km_state_expired(x, 0, 0);
1608 }
1609 return 0;
1610 }
1611 EXPORT_SYMBOL(xfrm_state_check_expire);
1612
1613 struct xfrm_state *
1614 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1615 u8 proto, unsigned short family)
1616 {
1617 struct xfrm_state *x;
1618
1619 rcu_read_lock();
1620 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1621 rcu_read_unlock();
1622 return x;
1623 }
1624 EXPORT_SYMBOL(xfrm_state_lookup);
1625
1626 struct xfrm_state *
1627 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1628 const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1629 u8 proto, unsigned short family)
1630 {
1631 struct xfrm_state *x;
1632
1633 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1634 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1635 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1636 return x;
1637 }
1638 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1639
1640 struct xfrm_state *
1641 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1642 u32 if_id, u8 proto, const xfrm_address_t *daddr,
1643 const xfrm_address_t *saddr, int create, unsigned short family)
1644 {
1645 struct xfrm_state *x;
1646
1647 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1648 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1649 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1650
1651 return x;
1652 }
1653 EXPORT_SYMBOL(xfrm_find_acq);
1654
1655 #ifdef CONFIG_XFRM_SUB_POLICY
1656 int
1657 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1658 unsigned short family, struct net *net)
1659 {
1660 int i;
1661 int err = 0;
1662 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1663 if (!afinfo)
1664 return -EAFNOSUPPORT;
1665
1666 spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1667 if (afinfo->tmpl_sort)
1668 err = afinfo->tmpl_sort(dst, src, n);
1669 else
1670 for (i = 0; i < n; i++)
1671 dst[i] = src[i];
1672 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1673 rcu_read_unlock();
1674 return err;
1675 }
1676 EXPORT_SYMBOL(xfrm_tmpl_sort);
1677
1678 int
1679 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1680 unsigned short family)
1681 {
1682 int i;
1683 int err = 0;
1684 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1685 struct net *net = xs_net(*src);
1686
1687 if (!afinfo)
1688 return -EAFNOSUPPORT;
1689
1690 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1691 if (afinfo->state_sort)
1692 err = afinfo->state_sort(dst, src, n);
1693 else
1694 for (i = 0; i < n; i++)
1695 dst[i] = src[i];
1696 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1697 rcu_read_unlock();
1698 return err;
1699 }
1700 EXPORT_SYMBOL(xfrm_state_sort);
1701 #endif
1702
1703 /* Silly enough, but I'm lazy to build resolution list */
1704
1705 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1706 {
1707 int i;
1708
1709 for (i = 0; i <= net->xfrm.state_hmask; i++) {
1710 struct xfrm_state *x;
1711
1712 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1713 if (x->km.seq == seq &&
1714 (mark & x->mark.m) == x->mark.v &&
1715 x->km.state == XFRM_STATE_ACQ) {
1716 xfrm_state_hold(x);
1717 return x;
1718 }
1719 }
1720 }
1721 return NULL;
1722 }
1723
1724 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1725 {
1726 struct xfrm_state *x;
1727
1728 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1729 x = __xfrm_find_acq_byseq(net, mark, seq);
1730 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1731 return x;
1732 }
1733 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1734
1735 u32 xfrm_get_acqseq(void)
1736 {
1737 u32 res;
1738 static atomic_t acqseq;
1739
1740 do {
1741 res = atomic_inc_return(&acqseq);
1742 } while (!res);
1743
1744 return res;
1745 }
1746 EXPORT_SYMBOL(xfrm_get_acqseq);
1747
1748 int verify_spi_info(u8 proto, u32 min, u32 max)
1749 {
1750 switch (proto) {
1751 case IPPROTO_AH:
1752 case IPPROTO_ESP:
1753 break;
1754
1755 case IPPROTO_COMP:
1756 /* IPCOMP spi is 16-bits. */
1757 if (max >= 0x10000)
1758 return -EINVAL;
1759 break;
1760
1761 default:
1762 return -EINVAL;
1763 }
1764
1765 if (min > max)
1766 return -EINVAL;
1767
1768 return 0;
1769 }
1770 EXPORT_SYMBOL(verify_spi_info);
1771
1772 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1773 {
1774 struct net *net = xs_net(x);
1775 unsigned int h;
1776 struct xfrm_state *x0;
1777 int err = -ENOENT;
1778 __be32 minspi = htonl(low);
1779 __be32 maxspi = htonl(high);
1780 u32 mark = x->mark.v & x->mark.m;
1781
1782 spin_lock_bh(&x->lock);
1783 if (x->km.state == XFRM_STATE_DEAD)
1784 goto unlock;
1785
1786 err = 0;
1787 if (x->id.spi)
1788 goto unlock;
1789
1790 err = -ENOENT;
1791
1792 if (minspi == maxspi) {
1793 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1794 if (x0) {
1795 xfrm_state_put(x0);
1796 goto unlock;
1797 }
1798 x->id.spi = minspi;
1799 } else {
1800 u32 spi = 0;
1801 for (h = 0; h < high-low+1; h++) {
1802 spi = low + prandom_u32()%(high-low+1);
1803 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1804 if (x0 == NULL) {
1805 x->id.spi = htonl(spi);
1806 break;
1807 }
1808 xfrm_state_put(x0);
1809 }
1810 }
1811 if (x->id.spi) {
1812 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1813 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1814 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1815 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1816
1817 err = 0;
1818 }
1819
1820 unlock:
1821 spin_unlock_bh(&x->lock);
1822
1823 return err;
1824 }
1825 EXPORT_SYMBOL(xfrm_alloc_spi);
1826
1827 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1828 struct xfrm_address_filter *filter)
1829 {
1830 if (filter) {
1831 if ((filter->family == AF_INET ||
1832 filter->family == AF_INET6) &&
1833 x->props.family != filter->family)
1834 return false;
1835
1836 return addr_match(&x->props.saddr, &filter->saddr,
1837 filter->splen) &&
1838 addr_match(&x->id.daddr, &filter->daddr,
1839 filter->dplen);
1840 }
1841 return true;
1842 }
1843
1844 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1845 int (*func)(struct xfrm_state *, int, void*),
1846 void *data)
1847 {
1848 struct xfrm_state *state;
1849 struct xfrm_state_walk *x;
1850 int err = 0;
1851
1852 if (walk->seq != 0 && list_empty(&walk->all))
1853 return 0;
1854
1855 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1856 if (list_empty(&walk->all))
1857 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1858 else
1859 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1860 list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1861 if (x->state == XFRM_STATE_DEAD)
1862 continue;
1863 state = container_of(x, struct xfrm_state, km);
1864 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1865 continue;
1866 if (!__xfrm_state_filter_match(state, walk->filter))
1867 continue;
1868 err = func(state, walk->seq, data);
1869 if (err) {
1870 list_move_tail(&walk->all, &x->all);
1871 goto out;
1872 }
1873 walk->seq++;
1874 }
1875 if (walk->seq == 0) {
1876 err = -ENOENT;
1877 goto out;
1878 }
1879 list_del_init(&walk->all);
1880 out:
1881 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1882 return err;
1883 }
1884 EXPORT_SYMBOL(xfrm_state_walk);
1885
1886 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1887 struct xfrm_address_filter *filter)
1888 {
1889 INIT_LIST_HEAD(&walk->all);
1890 walk->proto = proto;
1891 walk->state = XFRM_STATE_DEAD;
1892 walk->seq = 0;
1893 walk->filter = filter;
1894 }
1895 EXPORT_SYMBOL(xfrm_state_walk_init);
1896
1897 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1898 {
1899 kfree(walk->filter);
1900
1901 if (list_empty(&walk->all))
1902 return;
1903
1904 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1905 list_del(&walk->all);
1906 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1907 }
1908 EXPORT_SYMBOL(xfrm_state_walk_done);
1909
1910 static void xfrm_replay_timer_handler(struct timer_list *t)
1911 {
1912 struct xfrm_state *x = from_timer(x, t, rtimer);
1913
1914 spin_lock(&x->lock);
1915
1916 if (x->km.state == XFRM_STATE_VALID) {
1917 if (xfrm_aevent_is_on(xs_net(x)))
1918 x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1919 else
1920 x->xflags |= XFRM_TIME_DEFER;
1921 }
1922
1923 spin_unlock(&x->lock);
1924 }
1925
1926 static LIST_HEAD(xfrm_km_list);
1927
1928 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1929 {
1930 struct xfrm_mgr *km;
1931
1932 rcu_read_lock();
1933 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1934 if (km->notify_policy)
1935 km->notify_policy(xp, dir, c);
1936 rcu_read_unlock();
1937 }
1938
1939 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1940 {
1941 struct xfrm_mgr *km;
1942 rcu_read_lock();
1943 list_for_each_entry_rcu(km, &xfrm_km_list, list)
1944 if (km->notify)
1945 km->notify(x, c);
1946 rcu_read_unlock();
1947 }
1948
1949 EXPORT_SYMBOL(km_policy_notify);
1950 EXPORT_SYMBOL(km_state_notify);
1951
1952 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1953 {
1954 struct km_event c;
1955
1956 c.data.hard = hard;
1957 c.portid = portid;
1958 c.event = XFRM_MSG_EXPIRE;
1959 km_state_notify(x, &c);
1960 }
1961
1962 EXPORT_SYMBOL(km_state_expired);
1963 /*
1964 * We send to all registered managers regardless of failure
1965 * We are happy with one success
1966 */
1967 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1968 {
1969 int err = -EINVAL, acqret;
1970 struct xfrm_mgr *km;
1971
1972 rcu_read_lock();
1973 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1974 acqret = km->acquire(x, t, pol);
1975 if (!acqret)
1976 err = acqret;
1977 }
1978 rcu_read_unlock();
1979 return err;
1980 }
1981 EXPORT_SYMBOL(km_query);
1982
1983 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1984 {
1985 int err = -EINVAL;
1986 struct xfrm_mgr *km;
1987
1988 rcu_read_lock();
1989 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1990 if (km->new_mapping)
1991 err = km->new_mapping(x, ipaddr, sport);
1992 if (!err)
1993 break;
1994 }
1995 rcu_read_unlock();
1996 return err;
1997 }
1998 EXPORT_SYMBOL(km_new_mapping);
1999
2000 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2001 {
2002 struct km_event c;
2003
2004 c.data.hard = hard;
2005 c.portid = portid;
2006 c.event = XFRM_MSG_POLEXPIRE;
2007 km_policy_notify(pol, dir, &c);
2008 }
2009 EXPORT_SYMBOL(km_policy_expired);
2010
2011 #ifdef CONFIG_XFRM_MIGRATE
2012 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2013 const struct xfrm_migrate *m, int num_migrate,
2014 const struct xfrm_kmaddress *k,
2015 const struct xfrm_encap_tmpl *encap)
2016 {
2017 int err = -EINVAL;
2018 int ret;
2019 struct xfrm_mgr *km;
2020
2021 rcu_read_lock();
2022 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2023 if (km->migrate) {
2024 ret = km->migrate(sel, dir, type, m, num_migrate, k,
2025 encap);
2026 if (!ret)
2027 err = ret;
2028 }
2029 }
2030 rcu_read_unlock();
2031 return err;
2032 }
2033 EXPORT_SYMBOL(km_migrate);
2034 #endif
2035
2036 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2037 {
2038 int err = -EINVAL;
2039 int ret;
2040 struct xfrm_mgr *km;
2041
2042 rcu_read_lock();
2043 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2044 if (km->report) {
2045 ret = km->report(net, proto, sel, addr);
2046 if (!ret)
2047 err = ret;
2048 }
2049 }
2050 rcu_read_unlock();
2051 return err;
2052 }
2053 EXPORT_SYMBOL(km_report);
2054
2055 bool km_is_alive(const struct km_event *c)
2056 {
2057 struct xfrm_mgr *km;
2058 bool is_alive = false;
2059
2060 rcu_read_lock();
2061 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2062 if (km->is_alive && km->is_alive(c)) {
2063 is_alive = true;
2064 break;
2065 }
2066 }
2067 rcu_read_unlock();
2068
2069 return is_alive;
2070 }
2071 EXPORT_SYMBOL(km_is_alive);
2072
2073 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
2074 {
2075 int err;
2076 u8 *data;
2077 struct xfrm_mgr *km;
2078 struct xfrm_policy *pol = NULL;
2079
2080 #ifdef CONFIG_COMPAT
2081 if (in_compat_syscall())
2082 return -EOPNOTSUPP;
2083 #endif
2084
2085 if (!optval && !optlen) {
2086 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2087 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2088 __sk_dst_reset(sk);
2089 return 0;
2090 }
2091
2092 if (optlen <= 0 || optlen > PAGE_SIZE)
2093 return -EMSGSIZE;
2094
2095 data = memdup_user(optval, optlen);
2096 if (IS_ERR(data))
2097 return PTR_ERR(data);
2098
2099 err = -EINVAL;
2100 rcu_read_lock();
2101 list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2102 pol = km->compile_policy(sk, optname, data,
2103 optlen, &err);
2104 if (err >= 0)
2105 break;
2106 }
2107 rcu_read_unlock();
2108
2109 if (err >= 0) {
2110 xfrm_sk_policy_insert(sk, err, pol);
2111 xfrm_pol_put(pol);
2112 __sk_dst_reset(sk);
2113 err = 0;
2114 }
2115
2116 kfree(data);
2117 return err;
2118 }
2119 EXPORT_SYMBOL(xfrm_user_policy);
2120
2121 static DEFINE_SPINLOCK(xfrm_km_lock);
2122
2123 int xfrm_register_km(struct xfrm_mgr *km)
2124 {
2125 spin_lock_bh(&xfrm_km_lock);
2126 list_add_tail_rcu(&km->list, &xfrm_km_list);
2127 spin_unlock_bh(&xfrm_km_lock);
2128 return 0;
2129 }
2130 EXPORT_SYMBOL(xfrm_register_km);
2131
2132 int xfrm_unregister_km(struct xfrm_mgr *km)
2133 {
2134 spin_lock_bh(&xfrm_km_lock);
2135 list_del_rcu(&km->list);
2136 spin_unlock_bh(&xfrm_km_lock);
2137 synchronize_rcu();
2138 return 0;
2139 }
2140 EXPORT_SYMBOL(xfrm_unregister_km);
2141
2142 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2143 {
2144 int err = 0;
2145
2146 if (WARN_ON(afinfo->family >= NPROTO))
2147 return -EAFNOSUPPORT;
2148
2149 spin_lock_bh(&xfrm_state_afinfo_lock);
2150 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2151 err = -EEXIST;
2152 else
2153 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2154 spin_unlock_bh(&xfrm_state_afinfo_lock);
2155 return err;
2156 }
2157 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2158
2159 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2160 {
2161 int err = 0, family = afinfo->family;
2162
2163 if (WARN_ON(family >= NPROTO))
2164 return -EAFNOSUPPORT;
2165
2166 spin_lock_bh(&xfrm_state_afinfo_lock);
2167 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2168 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2169 err = -EINVAL;
2170 else
2171 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2172 }
2173 spin_unlock_bh(&xfrm_state_afinfo_lock);
2174 synchronize_rcu();
2175 return err;
2176 }
2177 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2178
2179 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2180 {
2181 if (unlikely(family >= NPROTO))
2182 return NULL;
2183
2184 return rcu_dereference(xfrm_state_afinfo[family]);
2185 }
2186
2187 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2188 {
2189 struct xfrm_state_afinfo *afinfo;
2190 if (unlikely(family >= NPROTO))
2191 return NULL;
2192 rcu_read_lock();
2193 afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2194 if (unlikely(!afinfo))
2195 rcu_read_unlock();
2196 return afinfo;
2197 }
2198
2199 void xfrm_flush_gc(void)
2200 {
2201 flush_work(&xfrm_state_gc_work);
2202 }
2203 EXPORT_SYMBOL(xfrm_flush_gc);
2204
2205 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2206 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2207 {
2208 if (x->tunnel) {
2209 struct xfrm_state *t = x->tunnel;
2210
2211 if (atomic_read(&t->tunnel_users) == 2)
2212 xfrm_state_delete(t);
2213 atomic_dec(&t->tunnel_users);
2214 xfrm_state_put(t);
2215 x->tunnel = NULL;
2216 }
2217 }
2218 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2219
2220 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2221 {
2222 const struct xfrm_type *type = READ_ONCE(x->type);
2223
2224 if (x->km.state == XFRM_STATE_VALID &&
2225 type && type->get_mtu)
2226 return type->get_mtu(x, mtu);
2227
2228 return mtu - x->props.header_len;
2229 }
2230
2231 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2232 {
2233 struct xfrm_state_afinfo *afinfo;
2234 struct xfrm_mode *inner_mode;
2235 int family = x->props.family;
2236 int err;
2237
2238 err = -EAFNOSUPPORT;
2239 afinfo = xfrm_state_get_afinfo(family);
2240 if (!afinfo)
2241 goto error;
2242
2243 err = 0;
2244 if (afinfo->init_flags)
2245 err = afinfo->init_flags(x);
2246
2247 rcu_read_unlock();
2248
2249 if (err)
2250 goto error;
2251
2252 err = -EPROTONOSUPPORT;
2253
2254 if (x->sel.family != AF_UNSPEC) {
2255 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2256 if (inner_mode == NULL)
2257 goto error;
2258
2259 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2260 family != x->sel.family) {
2261 xfrm_put_mode(inner_mode);
2262 goto error;
2263 }
2264
2265 x->inner_mode = inner_mode;
2266 } else {
2267 struct xfrm_mode *inner_mode_iaf;
2268 int iafamily = AF_INET;
2269
2270 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2271 if (inner_mode == NULL)
2272 goto error;
2273
2274 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2275 xfrm_put_mode(inner_mode);
2276 goto error;
2277 }
2278 x->inner_mode = inner_mode;
2279
2280 if (x->props.family == AF_INET)
2281 iafamily = AF_INET6;
2282
2283 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2284 if (inner_mode_iaf) {
2285 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2286 x->inner_mode_iaf = inner_mode_iaf;
2287 else
2288 xfrm_put_mode(inner_mode_iaf);
2289 }
2290 }
2291
2292 x->type = xfrm_get_type(x->id.proto, family);
2293 if (x->type == NULL)
2294 goto error;
2295
2296 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2297
2298 err = x->type->init_state(x);
2299 if (err)
2300 goto error;
2301
2302 x->outer_mode = xfrm_get_mode(x->props.mode, family);
2303 if (x->outer_mode == NULL) {
2304 err = -EPROTONOSUPPORT;
2305 goto error;
2306 }
2307
2308 if (init_replay) {
2309 err = xfrm_init_replay(x);
2310 if (err)
2311 goto error;
2312 }
2313
2314 error:
2315 return err;
2316 }
2317
2318 EXPORT_SYMBOL(__xfrm_init_state);
2319
2320 int xfrm_init_state(struct xfrm_state *x)
2321 {
2322 int err;
2323
2324 err = __xfrm_init_state(x, true, false);
2325 if (!err)
2326 x->km.state = XFRM_STATE_VALID;
2327
2328 return err;
2329 }
2330
2331 EXPORT_SYMBOL(xfrm_init_state);
2332
2333 int __net_init xfrm_state_init(struct net *net)
2334 {
2335 unsigned int sz;
2336
2337 if (net_eq(net, &init_net))
2338 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2339 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2340
2341 INIT_LIST_HEAD(&net->xfrm.state_all);
2342
2343 sz = sizeof(struct hlist_head) * 8;
2344
2345 net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2346 if (!net->xfrm.state_bydst)
2347 goto out_bydst;
2348 net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2349 if (!net->xfrm.state_bysrc)
2350 goto out_bysrc;
2351 net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2352 if (!net->xfrm.state_byspi)
2353 goto out_byspi;
2354 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2355
2356 net->xfrm.state_num = 0;
2357 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2358 spin_lock_init(&net->xfrm.xfrm_state_lock);
2359 return 0;
2360
2361 out_byspi:
2362 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2363 out_bysrc:
2364 xfrm_hash_free(net->xfrm.state_bydst, sz);
2365 out_bydst:
2366 return -ENOMEM;
2367 }
2368
2369 void xfrm_state_fini(struct net *net)
2370 {
2371 unsigned int sz;
2372
2373 flush_work(&net->xfrm.state_hash_work);
2374 xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2375 flush_work(&xfrm_state_gc_work);
2376
2377 WARN_ON(!list_empty(&net->xfrm.state_all));
2378
2379 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2380 WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2381 xfrm_hash_free(net->xfrm.state_byspi, sz);
2382 WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2383 xfrm_hash_free(net->xfrm.state_bysrc, sz);
2384 WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2385 xfrm_hash_free(net->xfrm.state_bydst, sz);
2386 }
2387
2388 #ifdef CONFIG_AUDITSYSCALL
2389 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2390 struct audit_buffer *audit_buf)
2391 {
2392 struct xfrm_sec_ctx *ctx = x->security;
2393 u32 spi = ntohl(x->id.spi);
2394
2395 if (ctx)
2396 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2397 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2398
2399 switch (x->props.family) {
2400 case AF_INET:
2401 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2402 &x->props.saddr.a4, &x->id.daddr.a4);
2403 break;
2404 case AF_INET6:
2405 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2406 x->props.saddr.a6, x->id.daddr.a6);
2407 break;
2408 }
2409
2410 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2411 }
2412
2413 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2414 struct audit_buffer *audit_buf)
2415 {
2416 const struct iphdr *iph4;
2417 const struct ipv6hdr *iph6;
2418
2419 switch (family) {
2420 case AF_INET:
2421 iph4 = ip_hdr(skb);
2422 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2423 &iph4->saddr, &iph4->daddr);
2424 break;
2425 case AF_INET6:
2426 iph6 = ipv6_hdr(skb);
2427 audit_log_format(audit_buf,
2428 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2429 &iph6->saddr, &iph6->daddr,
2430 iph6->flow_lbl[0] & 0x0f,
2431 iph6->flow_lbl[1],
2432 iph6->flow_lbl[2]);
2433 break;
2434 }
2435 }
2436
2437 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2438 {
2439 struct audit_buffer *audit_buf;
2440
2441 audit_buf = xfrm_audit_start("SAD-add");
2442 if (audit_buf == NULL)
2443 return;
2444 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2445 xfrm_audit_helper_sainfo(x, audit_buf);
2446 audit_log_format(audit_buf, " res=%u", result);
2447 audit_log_end(audit_buf);
2448 }
2449 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2450
2451 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2452 {
2453 struct audit_buffer *audit_buf;
2454
2455 audit_buf = xfrm_audit_start("SAD-delete");
2456 if (audit_buf == NULL)
2457 return;
2458 xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2459 xfrm_audit_helper_sainfo(x, audit_buf);
2460 audit_log_format(audit_buf, " res=%u", result);
2461 audit_log_end(audit_buf);
2462 }
2463 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2464
2465 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2466 struct sk_buff *skb)
2467 {
2468 struct audit_buffer *audit_buf;
2469 u32 spi;
2470
2471 audit_buf = xfrm_audit_start("SA-replay-overflow");
2472 if (audit_buf == NULL)
2473 return;
2474 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2475 /* don't record the sequence number because it's inherent in this kind
2476 * of audit message */
2477 spi = ntohl(x->id.spi);
2478 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2479 audit_log_end(audit_buf);
2480 }
2481 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2482
2483 void xfrm_audit_state_replay(struct xfrm_state *x,
2484 struct sk_buff *skb, __be32 net_seq)
2485 {
2486 struct audit_buffer *audit_buf;
2487 u32 spi;
2488
2489 audit_buf = xfrm_audit_start("SA-replayed-pkt");
2490 if (audit_buf == NULL)
2491 return;
2492 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2493 spi = ntohl(x->id.spi);
2494 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2495 spi, spi, ntohl(net_seq));
2496 audit_log_end(audit_buf);
2497 }
2498 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2499
2500 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2501 {
2502 struct audit_buffer *audit_buf;
2503
2504 audit_buf = xfrm_audit_start("SA-notfound");
2505 if (audit_buf == NULL)
2506 return;
2507 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2508 audit_log_end(audit_buf);
2509 }
2510 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2511
2512 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2513 __be32 net_spi, __be32 net_seq)
2514 {
2515 struct audit_buffer *audit_buf;
2516 u32 spi;
2517
2518 audit_buf = xfrm_audit_start("SA-notfound");
2519 if (audit_buf == NULL)
2520 return;
2521 xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2522 spi = ntohl(net_spi);
2523 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2524 spi, spi, ntohl(net_seq));
2525 audit_log_end(audit_buf);
2526 }
2527 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2528
2529 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2530 struct sk_buff *skb, u8 proto)
2531 {
2532 struct audit_buffer *audit_buf;
2533 __be32 net_spi;
2534 __be32 net_seq;
2535
2536 audit_buf = xfrm_audit_start("SA-icv-failure");
2537 if (audit_buf == NULL)
2538 return;
2539 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2540 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2541 u32 spi = ntohl(net_spi);
2542 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2543 spi, spi, ntohl(net_seq));
2544 }
2545 audit_log_end(audit_buf);
2546 }
2547 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2548 #endif /* CONFIG_AUDITSYSCALL */
Linux with added FPC-III support