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