xfrm: Assign the inner mode output function to the dst entry
[linux-2.6.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 /* Each xfrm_state may be linked to two tables:
32
33    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
34    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
35       destination/tunnel endpoint. (output)
36  */
37
38 static DEFINE_SPINLOCK(xfrm_state_lock);
39
40 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
41
42 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
44
45 static inline unsigned int xfrm_dst_hash(struct net *net,
46                                          const xfrm_address_t *daddr,
47                                          const xfrm_address_t *saddr,
48                                          u32 reqid,
49                                          unsigned short family)
50 {
51         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
52 }
53
54 static inline unsigned int xfrm_src_hash(struct net *net,
55                                          const xfrm_address_t *daddr,
56                                          const xfrm_address_t *saddr,
57                                          unsigned short family)
58 {
59         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
60 }
61
62 static inline unsigned int
63 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
64               __be32 spi, u8 proto, unsigned short family)
65 {
66         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
67 }
68
69 static void xfrm_hash_transfer(struct hlist_head *list,
70                                struct hlist_head *ndsttable,
71                                struct hlist_head *nsrctable,
72                                struct hlist_head *nspitable,
73                                unsigned int nhashmask)
74 {
75         struct hlist_node *entry, *tmp;
76         struct xfrm_state *x;
77
78         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
79                 unsigned int h;
80
81                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
82                                     x->props.reqid, x->props.family,
83                                     nhashmask);
84                 hlist_add_head(&x->bydst, ndsttable+h);
85
86                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
87                                     x->props.family,
88                                     nhashmask);
89                 hlist_add_head(&x->bysrc, nsrctable+h);
90
91                 if (x->id.spi) {
92                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
93                                             x->id.proto, x->props.family,
94                                             nhashmask);
95                         hlist_add_head(&x->byspi, nspitable+h);
96                 }
97         }
98 }
99
100 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
101 {
102         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
103 }
104
105 static DEFINE_MUTEX(hash_resize_mutex);
106
107 static void xfrm_hash_resize(struct work_struct *work)
108 {
109         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
110         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
111         unsigned long nsize, osize;
112         unsigned int nhashmask, ohashmask;
113         int i;
114
115         mutex_lock(&hash_resize_mutex);
116
117         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
118         ndst = xfrm_hash_alloc(nsize);
119         if (!ndst)
120                 goto out_unlock;
121         nsrc = xfrm_hash_alloc(nsize);
122         if (!nsrc) {
123                 xfrm_hash_free(ndst, nsize);
124                 goto out_unlock;
125         }
126         nspi = xfrm_hash_alloc(nsize);
127         if (!nspi) {
128                 xfrm_hash_free(ndst, nsize);
129                 xfrm_hash_free(nsrc, nsize);
130                 goto out_unlock;
131         }
132
133         spin_lock_bh(&xfrm_state_lock);
134
135         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
136         for (i = net->xfrm.state_hmask; i >= 0; i--)
137                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
138                                    nhashmask);
139
140         odst = net->xfrm.state_bydst;
141         osrc = net->xfrm.state_bysrc;
142         ospi = net->xfrm.state_byspi;
143         ohashmask = net->xfrm.state_hmask;
144
145         net->xfrm.state_bydst = ndst;
146         net->xfrm.state_bysrc = nsrc;
147         net->xfrm.state_byspi = nspi;
148         net->xfrm.state_hmask = nhashmask;
149
150         spin_unlock_bh(&xfrm_state_lock);
151
152         osize = (ohashmask + 1) * sizeof(struct hlist_head);
153         xfrm_hash_free(odst, osize);
154         xfrm_hash_free(osrc, osize);
155         xfrm_hash_free(ospi, osize);
156
157 out_unlock:
158         mutex_unlock(&hash_resize_mutex);
159 }
160
161 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
162 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
163
164 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
165
166 int __xfrm_state_delete(struct xfrm_state *x);
167
168 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
169 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
170
171 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
172 {
173         struct xfrm_state_afinfo *afinfo;
174         if (unlikely(family >= NPROTO))
175                 return NULL;
176         write_lock_bh(&xfrm_state_afinfo_lock);
177         afinfo = xfrm_state_afinfo[family];
178         if (unlikely(!afinfo))
179                 write_unlock_bh(&xfrm_state_afinfo_lock);
180         return afinfo;
181 }
182
183 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
184         __releases(xfrm_state_afinfo_lock)
185 {
186         write_unlock_bh(&xfrm_state_afinfo_lock);
187 }
188
189 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
190 {
191         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
192         const struct xfrm_type **typemap;
193         int err = 0;
194
195         if (unlikely(afinfo == NULL))
196                 return -EAFNOSUPPORT;
197         typemap = afinfo->type_map;
198
199         if (likely(typemap[type->proto] == NULL))
200                 typemap[type->proto] = type;
201         else
202                 err = -EEXIST;
203         xfrm_state_unlock_afinfo(afinfo);
204         return err;
205 }
206 EXPORT_SYMBOL(xfrm_register_type);
207
208 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
209 {
210         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
211         const struct xfrm_type **typemap;
212         int err = 0;
213
214         if (unlikely(afinfo == NULL))
215                 return -EAFNOSUPPORT;
216         typemap = afinfo->type_map;
217
218         if (unlikely(typemap[type->proto] != type))
219                 err = -ENOENT;
220         else
221                 typemap[type->proto] = NULL;
222         xfrm_state_unlock_afinfo(afinfo);
223         return err;
224 }
225 EXPORT_SYMBOL(xfrm_unregister_type);
226
227 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
228 {
229         struct xfrm_state_afinfo *afinfo;
230         const struct xfrm_type **typemap;
231         const struct xfrm_type *type;
232         int modload_attempted = 0;
233
234 retry:
235         afinfo = xfrm_state_get_afinfo(family);
236         if (unlikely(afinfo == NULL))
237                 return NULL;
238         typemap = afinfo->type_map;
239
240         type = typemap[proto];
241         if (unlikely(type && !try_module_get(type->owner)))
242                 type = NULL;
243         if (!type && !modload_attempted) {
244                 xfrm_state_put_afinfo(afinfo);
245                 request_module("xfrm-type-%d-%d", family, proto);
246                 modload_attempted = 1;
247                 goto retry;
248         }
249
250         xfrm_state_put_afinfo(afinfo);
251         return type;
252 }
253
254 static void xfrm_put_type(const struct xfrm_type *type)
255 {
256         module_put(type->owner);
257 }
258
259 int xfrm_register_mode(struct xfrm_mode *mode, int family)
260 {
261         struct xfrm_state_afinfo *afinfo;
262         struct xfrm_mode **modemap;
263         int err;
264
265         if (unlikely(mode->encap >= XFRM_MODE_MAX))
266                 return -EINVAL;
267
268         afinfo = xfrm_state_lock_afinfo(family);
269         if (unlikely(afinfo == NULL))
270                 return -EAFNOSUPPORT;
271
272         err = -EEXIST;
273         modemap = afinfo->mode_map;
274         if (modemap[mode->encap])
275                 goto out;
276
277         err = -ENOENT;
278         if (!try_module_get(afinfo->owner))
279                 goto out;
280
281         mode->afinfo = afinfo;
282         modemap[mode->encap] = mode;
283         err = 0;
284
285 out:
286         xfrm_state_unlock_afinfo(afinfo);
287         return err;
288 }
289 EXPORT_SYMBOL(xfrm_register_mode);
290
291 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
292 {
293         struct xfrm_state_afinfo *afinfo;
294         struct xfrm_mode **modemap;
295         int err;
296
297         if (unlikely(mode->encap >= XFRM_MODE_MAX))
298                 return -EINVAL;
299
300         afinfo = xfrm_state_lock_afinfo(family);
301         if (unlikely(afinfo == NULL))
302                 return -EAFNOSUPPORT;
303
304         err = -ENOENT;
305         modemap = afinfo->mode_map;
306         if (likely(modemap[mode->encap] == mode)) {
307                 modemap[mode->encap] = NULL;
308                 module_put(mode->afinfo->owner);
309                 err = 0;
310         }
311
312         xfrm_state_unlock_afinfo(afinfo);
313         return err;
314 }
315 EXPORT_SYMBOL(xfrm_unregister_mode);
316
317 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
318 {
319         struct xfrm_state_afinfo *afinfo;
320         struct xfrm_mode *mode;
321         int modload_attempted = 0;
322
323         if (unlikely(encap >= XFRM_MODE_MAX))
324                 return NULL;
325
326 retry:
327         afinfo = xfrm_state_get_afinfo(family);
328         if (unlikely(afinfo == NULL))
329                 return NULL;
330
331         mode = afinfo->mode_map[encap];
332         if (unlikely(mode && !try_module_get(mode->owner)))
333                 mode = NULL;
334         if (!mode && !modload_attempted) {
335                 xfrm_state_put_afinfo(afinfo);
336                 request_module("xfrm-mode-%d-%d", family, encap);
337                 modload_attempted = 1;
338                 goto retry;
339         }
340
341         xfrm_state_put_afinfo(afinfo);
342         return mode;
343 }
344
345 static void xfrm_put_mode(struct xfrm_mode *mode)
346 {
347         module_put(mode->owner);
348 }
349
350 static void xfrm_state_gc_destroy(struct xfrm_state *x)
351 {
352         tasklet_hrtimer_cancel(&x->mtimer);
353         del_timer_sync(&x->rtimer);
354         kfree(x->aalg);
355         kfree(x->ealg);
356         kfree(x->calg);
357         kfree(x->encap);
358         kfree(x->coaddr);
359         kfree(x->replay_esn);
360         kfree(x->preplay_esn);
361         if (x->inner_mode)
362                 xfrm_put_mode(x->inner_mode);
363         if (x->inner_mode_iaf)
364                 xfrm_put_mode(x->inner_mode_iaf);
365         if (x->outer_mode)
366                 xfrm_put_mode(x->outer_mode);
367         if (x->type) {
368                 x->type->destructor(x);
369                 xfrm_put_type(x->type);
370         }
371         security_xfrm_state_free(x);
372         kfree(x);
373 }
374
375 static void xfrm_state_gc_task(struct work_struct *work)
376 {
377         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
378         struct xfrm_state *x;
379         struct hlist_node *entry, *tmp;
380         struct hlist_head gc_list;
381
382         spin_lock_bh(&xfrm_state_gc_lock);
383         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
384         spin_unlock_bh(&xfrm_state_gc_lock);
385
386         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
387                 xfrm_state_gc_destroy(x);
388
389         wake_up(&net->xfrm.km_waitq);
390 }
391
392 static inline unsigned long make_jiffies(long secs)
393 {
394         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
395                 return MAX_SCHEDULE_TIMEOUT-1;
396         else
397                 return secs*HZ;
398 }
399
400 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
401 {
402         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
403         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
404         struct net *net = xs_net(x);
405         unsigned long now = get_seconds();
406         long next = LONG_MAX;
407         int warn = 0;
408         int err = 0;
409
410         spin_lock(&x->lock);
411         if (x->km.state == XFRM_STATE_DEAD)
412                 goto out;
413         if (x->km.state == XFRM_STATE_EXPIRED)
414                 goto expired;
415         if (x->lft.hard_add_expires_seconds) {
416                 long tmo = x->lft.hard_add_expires_seconds +
417                         x->curlft.add_time - now;
418                 if (tmo <= 0)
419                         goto expired;
420                 if (tmo < next)
421                         next = tmo;
422         }
423         if (x->lft.hard_use_expires_seconds) {
424                 long tmo = x->lft.hard_use_expires_seconds +
425                         (x->curlft.use_time ? : now) - now;
426                 if (tmo <= 0)
427                         goto expired;
428                 if (tmo < next)
429                         next = tmo;
430         }
431         if (x->km.dying)
432                 goto resched;
433         if (x->lft.soft_add_expires_seconds) {
434                 long tmo = x->lft.soft_add_expires_seconds +
435                         x->curlft.add_time - now;
436                 if (tmo <= 0)
437                         warn = 1;
438                 else if (tmo < next)
439                         next = tmo;
440         }
441         if (x->lft.soft_use_expires_seconds) {
442                 long tmo = x->lft.soft_use_expires_seconds +
443                         (x->curlft.use_time ? : now) - now;
444                 if (tmo <= 0)
445                         warn = 1;
446                 else if (tmo < next)
447                         next = tmo;
448         }
449
450         x->km.dying = warn;
451         if (warn)
452                 km_state_expired(x, 0, 0);
453 resched:
454         if (next != LONG_MAX){
455                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
456         }
457
458         goto out;
459
460 expired:
461         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
462                 x->km.state = XFRM_STATE_EXPIRED;
463                 wake_up(&net->xfrm.km_waitq);
464                 next = 2;
465                 goto resched;
466         }
467
468         err = __xfrm_state_delete(x);
469         if (!err && x->id.spi)
470                 km_state_expired(x, 1, 0);
471
472         xfrm_audit_state_delete(x, err ? 0 : 1,
473                                 audit_get_loginuid(current),
474                                 audit_get_sessionid(current), 0);
475
476 out:
477         spin_unlock(&x->lock);
478         return HRTIMER_NORESTART;
479 }
480
481 static void xfrm_replay_timer_handler(unsigned long data);
482
483 struct xfrm_state *xfrm_state_alloc(struct net *net)
484 {
485         struct xfrm_state *x;
486
487         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
488
489         if (x) {
490                 write_pnet(&x->xs_net, net);
491                 atomic_set(&x->refcnt, 1);
492                 atomic_set(&x->tunnel_users, 0);
493                 INIT_LIST_HEAD(&x->km.all);
494                 INIT_HLIST_NODE(&x->bydst);
495                 INIT_HLIST_NODE(&x->bysrc);
496                 INIT_HLIST_NODE(&x->byspi);
497                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
498                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
499                                 (unsigned long)x);
500                 x->curlft.add_time = get_seconds();
501                 x->lft.soft_byte_limit = XFRM_INF;
502                 x->lft.soft_packet_limit = XFRM_INF;
503                 x->lft.hard_byte_limit = XFRM_INF;
504                 x->lft.hard_packet_limit = XFRM_INF;
505                 x->replay_maxage = 0;
506                 x->replay_maxdiff = 0;
507                 x->inner_mode = NULL;
508                 x->inner_mode_iaf = NULL;
509                 spin_lock_init(&x->lock);
510         }
511         return x;
512 }
513 EXPORT_SYMBOL(xfrm_state_alloc);
514
515 void __xfrm_state_destroy(struct xfrm_state *x)
516 {
517         struct net *net = xs_net(x);
518
519         WARN_ON(x->km.state != XFRM_STATE_DEAD);
520
521         spin_lock_bh(&xfrm_state_gc_lock);
522         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
523         spin_unlock_bh(&xfrm_state_gc_lock);
524         schedule_work(&net->xfrm.state_gc_work);
525 }
526 EXPORT_SYMBOL(__xfrm_state_destroy);
527
528 int __xfrm_state_delete(struct xfrm_state *x)
529 {
530         struct net *net = xs_net(x);
531         int err = -ESRCH;
532
533         if (x->km.state != XFRM_STATE_DEAD) {
534                 x->km.state = XFRM_STATE_DEAD;
535                 spin_lock(&xfrm_state_lock);
536                 list_del(&x->km.all);
537                 hlist_del(&x->bydst);
538                 hlist_del(&x->bysrc);
539                 if (x->id.spi)
540                         hlist_del(&x->byspi);
541                 net->xfrm.state_num--;
542                 spin_unlock(&xfrm_state_lock);
543
544                 /* All xfrm_state objects are created by xfrm_state_alloc.
545                  * The xfrm_state_alloc call gives a reference, and that
546                  * is what we are dropping here.
547                  */
548                 xfrm_state_put(x);
549                 err = 0;
550         }
551
552         return err;
553 }
554 EXPORT_SYMBOL(__xfrm_state_delete);
555
556 int xfrm_state_delete(struct xfrm_state *x)
557 {
558         int err;
559
560         spin_lock_bh(&x->lock);
561         err = __xfrm_state_delete(x);
562         spin_unlock_bh(&x->lock);
563
564         return err;
565 }
566 EXPORT_SYMBOL(xfrm_state_delete);
567
568 #ifdef CONFIG_SECURITY_NETWORK_XFRM
569 static inline int
570 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
571 {
572         int i, err = 0;
573
574         for (i = 0; i <= net->xfrm.state_hmask; i++) {
575                 struct hlist_node *entry;
576                 struct xfrm_state *x;
577
578                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
579                         if (xfrm_id_proto_match(x->id.proto, proto) &&
580                            (err = security_xfrm_state_delete(x)) != 0) {
581                                 xfrm_audit_state_delete(x, 0,
582                                                         audit_info->loginuid,
583                                                         audit_info->sessionid,
584                                                         audit_info->secid);
585                                 return err;
586                         }
587                 }
588         }
589
590         return err;
591 }
592 #else
593 static inline int
594 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
595 {
596         return 0;
597 }
598 #endif
599
600 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
601 {
602         int i, err = 0, cnt = 0;
603
604         spin_lock_bh(&xfrm_state_lock);
605         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
606         if (err)
607                 goto out;
608
609         err = -ESRCH;
610         for (i = 0; i <= net->xfrm.state_hmask; i++) {
611                 struct hlist_node *entry;
612                 struct xfrm_state *x;
613 restart:
614                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
615                         if (!xfrm_state_kern(x) &&
616                             xfrm_id_proto_match(x->id.proto, proto)) {
617                                 xfrm_state_hold(x);
618                                 spin_unlock_bh(&xfrm_state_lock);
619
620                                 err = xfrm_state_delete(x);
621                                 xfrm_audit_state_delete(x, err ? 0 : 1,
622                                                         audit_info->loginuid,
623                                                         audit_info->sessionid,
624                                                         audit_info->secid);
625                                 xfrm_state_put(x);
626                                 if (!err)
627                                         cnt++;
628
629                                 spin_lock_bh(&xfrm_state_lock);
630                                 goto restart;
631                         }
632                 }
633         }
634         if (cnt)
635                 err = 0;
636
637 out:
638         spin_unlock_bh(&xfrm_state_lock);
639         wake_up(&net->xfrm.km_waitq);
640         return err;
641 }
642 EXPORT_SYMBOL(xfrm_state_flush);
643
644 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
645 {
646         spin_lock_bh(&xfrm_state_lock);
647         si->sadcnt = net->xfrm.state_num;
648         si->sadhcnt = net->xfrm.state_hmask;
649         si->sadhmcnt = xfrm_state_hashmax;
650         spin_unlock_bh(&xfrm_state_lock);
651 }
652 EXPORT_SYMBOL(xfrm_sad_getinfo);
653
654 static int
655 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
656                     const struct xfrm_tmpl *tmpl,
657                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
658                     unsigned short family)
659 {
660         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
661         if (!afinfo)
662                 return -1;
663         afinfo->init_tempsel(&x->sel, fl);
664
665         if (family != tmpl->encap_family) {
666                 xfrm_state_put_afinfo(afinfo);
667                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
668                 if (!afinfo)
669                         return -1;
670         }
671         afinfo->init_temprop(x, tmpl, daddr, saddr);
672         xfrm_state_put_afinfo(afinfo);
673         return 0;
674 }
675
676 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
677                                               const xfrm_address_t *daddr,
678                                               __be32 spi, u8 proto,
679                                               unsigned short family)
680 {
681         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
682         struct xfrm_state *x;
683         struct hlist_node *entry;
684
685         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
686                 if (x->props.family != family ||
687                     x->id.spi       != spi ||
688                     x->id.proto     != proto ||
689                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
690                         continue;
691
692                 if ((mark & x->mark.m) != x->mark.v)
693                         continue;
694                 xfrm_state_hold(x);
695                 return x;
696         }
697
698         return NULL;
699 }
700
701 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
702                                                      const xfrm_address_t *daddr,
703                                                      const xfrm_address_t *saddr,
704                                                      u8 proto, unsigned short family)
705 {
706         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
707         struct xfrm_state *x;
708         struct hlist_node *entry;
709
710         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
711                 if (x->props.family != family ||
712                     x->id.proto     != proto ||
713                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
714                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
715                         continue;
716
717                 if ((mark & x->mark.m) != x->mark.v)
718                         continue;
719                 xfrm_state_hold(x);
720                 return x;
721         }
722
723         return NULL;
724 }
725
726 static inline struct xfrm_state *
727 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
728 {
729         struct net *net = xs_net(x);
730         u32 mark = x->mark.v & x->mark.m;
731
732         if (use_spi)
733                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
734                                            x->id.spi, x->id.proto, family);
735         else
736                 return __xfrm_state_lookup_byaddr(net, mark,
737                                                   &x->id.daddr,
738                                                   &x->props.saddr,
739                                                   x->id.proto, family);
740 }
741
742 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
743 {
744         if (have_hash_collision &&
745             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
746             net->xfrm.state_num > net->xfrm.state_hmask)
747                 schedule_work(&net->xfrm.state_hash_work);
748 }
749
750 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
751                                const struct flowi *fl, unsigned short family,
752                                struct xfrm_state **best, int *acq_in_progress,
753                                int *error)
754 {
755         /* Resolution logic:
756          * 1. There is a valid state with matching selector. Done.
757          * 2. Valid state with inappropriate selector. Skip.
758          *
759          * Entering area of "sysdeps".
760          *
761          * 3. If state is not valid, selector is temporary, it selects
762          *    only session which triggered previous resolution. Key
763          *    manager will do something to install a state with proper
764          *    selector.
765          */
766         if (x->km.state == XFRM_STATE_VALID) {
767                 if ((x->sel.family &&
768                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
769                     !security_xfrm_state_pol_flow_match(x, pol, fl))
770                         return;
771
772                 if (!*best ||
773                     (*best)->km.dying > x->km.dying ||
774                     ((*best)->km.dying == x->km.dying &&
775                      (*best)->curlft.add_time < x->curlft.add_time))
776                         *best = x;
777         } else if (x->km.state == XFRM_STATE_ACQ) {
778                 *acq_in_progress = 1;
779         } else if (x->km.state == XFRM_STATE_ERROR ||
780                    x->km.state == XFRM_STATE_EXPIRED) {
781                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
782                     security_xfrm_state_pol_flow_match(x, pol, fl))
783                         *error = -ESRCH;
784         }
785 }
786
787 struct xfrm_state *
788 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
789                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
790                 struct xfrm_policy *pol, int *err,
791                 unsigned short family)
792 {
793         static xfrm_address_t saddr_wildcard = { };
794         struct net *net = xp_net(pol);
795         unsigned int h, h_wildcard;
796         struct hlist_node *entry;
797         struct xfrm_state *x, *x0, *to_put;
798         int acquire_in_progress = 0;
799         int error = 0;
800         struct xfrm_state *best = NULL;
801         u32 mark = pol->mark.v & pol->mark.m;
802         unsigned short encap_family = tmpl->encap_family;
803
804         to_put = NULL;
805
806         spin_lock_bh(&xfrm_state_lock);
807         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
808         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
809                 if (x->props.family == encap_family &&
810                     x->props.reqid == tmpl->reqid &&
811                     (mark & x->mark.m) == x->mark.v &&
812                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
813                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
814                     tmpl->mode == x->props.mode &&
815                     tmpl->id.proto == x->id.proto &&
816                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
817                         xfrm_state_look_at(pol, x, fl, encap_family,
818                                            &best, &acquire_in_progress, &error);
819         }
820         if (best)
821                 goto found;
822
823         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
824         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
825                 if (x->props.family == encap_family &&
826                     x->props.reqid == tmpl->reqid &&
827                     (mark & x->mark.m) == x->mark.v &&
828                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
829                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
830                     tmpl->mode == x->props.mode &&
831                     tmpl->id.proto == x->id.proto &&
832                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
833                         xfrm_state_look_at(pol, x, fl, encap_family,
834                                            &best, &acquire_in_progress, &error);
835         }
836
837 found:
838         x = best;
839         if (!x && !error && !acquire_in_progress) {
840                 if (tmpl->id.spi &&
841                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
842                                               tmpl->id.proto, encap_family)) != NULL) {
843                         to_put = x0;
844                         error = -EEXIST;
845                         goto out;
846                 }
847                 x = xfrm_state_alloc(net);
848                 if (x == NULL) {
849                         error = -ENOMEM;
850                         goto out;
851                 }
852                 /* Initialize temporary state matching only
853                  * to current session. */
854                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
855                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
856
857                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
858                 if (error) {
859                         x->km.state = XFRM_STATE_DEAD;
860                         to_put = x;
861                         x = NULL;
862                         goto out;
863                 }
864
865                 if (km_query(x, tmpl, pol) == 0) {
866                         x->km.state = XFRM_STATE_ACQ;
867                         list_add(&x->km.all, &net->xfrm.state_all);
868                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
869                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
870                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
871                         if (x->id.spi) {
872                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
873                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
874                         }
875                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
876                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
877                         net->xfrm.state_num++;
878                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
879                 } else {
880                         x->km.state = XFRM_STATE_DEAD;
881                         to_put = x;
882                         x = NULL;
883                         error = -ESRCH;
884                 }
885         }
886 out:
887         if (x)
888                 xfrm_state_hold(x);
889         else
890                 *err = acquire_in_progress ? -EAGAIN : error;
891         spin_unlock_bh(&xfrm_state_lock);
892         if (to_put)
893                 xfrm_state_put(to_put);
894         return x;
895 }
896
897 struct xfrm_state *
898 xfrm_stateonly_find(struct net *net, u32 mark,
899                     xfrm_address_t *daddr, xfrm_address_t *saddr,
900                     unsigned short family, u8 mode, u8 proto, u32 reqid)
901 {
902         unsigned int h;
903         struct xfrm_state *rx = NULL, *x = NULL;
904         struct hlist_node *entry;
905
906         spin_lock(&xfrm_state_lock);
907         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
908         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
909                 if (x->props.family == family &&
910                     x->props.reqid == reqid &&
911                     (mark & x->mark.m) == x->mark.v &&
912                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
913                     xfrm_state_addr_check(x, daddr, saddr, family) &&
914                     mode == x->props.mode &&
915                     proto == x->id.proto &&
916                     x->km.state == XFRM_STATE_VALID) {
917                         rx = x;
918                         break;
919                 }
920         }
921
922         if (rx)
923                 xfrm_state_hold(rx);
924         spin_unlock(&xfrm_state_lock);
925
926
927         return rx;
928 }
929 EXPORT_SYMBOL(xfrm_stateonly_find);
930
931 static void __xfrm_state_insert(struct xfrm_state *x)
932 {
933         struct net *net = xs_net(x);
934         unsigned int h;
935
936         list_add(&x->km.all, &net->xfrm.state_all);
937
938         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
939                           x->props.reqid, x->props.family);
940         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
941
942         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
943         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
944
945         if (x->id.spi) {
946                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
947                                   x->props.family);
948
949                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
950         }
951
952         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
953         if (x->replay_maxage)
954                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
955
956         wake_up(&net->xfrm.km_waitq);
957
958         net->xfrm.state_num++;
959
960         xfrm_hash_grow_check(net, x->bydst.next != NULL);
961 }
962
963 /* xfrm_state_lock is held */
964 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
965 {
966         struct net *net = xs_net(xnew);
967         unsigned short family = xnew->props.family;
968         u32 reqid = xnew->props.reqid;
969         struct xfrm_state *x;
970         struct hlist_node *entry;
971         unsigned int h;
972         u32 mark = xnew->mark.v & xnew->mark.m;
973
974         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
975         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
976                 if (x->props.family     == family &&
977                     x->props.reqid      == reqid &&
978                     (mark & x->mark.m) == x->mark.v &&
979                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
980                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
981                         x->genid++;
982         }
983 }
984
985 void xfrm_state_insert(struct xfrm_state *x)
986 {
987         spin_lock_bh(&xfrm_state_lock);
988         __xfrm_state_bump_genids(x);
989         __xfrm_state_insert(x);
990         spin_unlock_bh(&xfrm_state_lock);
991 }
992 EXPORT_SYMBOL(xfrm_state_insert);
993
994 /* xfrm_state_lock is held */
995 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
996                                           unsigned short family, u8 mode,
997                                           u32 reqid, u8 proto,
998                                           const xfrm_address_t *daddr,
999                                           const xfrm_address_t *saddr, int create)
1000 {
1001         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1002         struct hlist_node *entry;
1003         struct xfrm_state *x;
1004         u32 mark = m->v & m->m;
1005
1006         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1007                 if (x->props.reqid  != reqid ||
1008                     x->props.mode   != mode ||
1009                     x->props.family != family ||
1010                     x->km.state     != XFRM_STATE_ACQ ||
1011                     x->id.spi       != 0 ||
1012                     x->id.proto     != proto ||
1013                     (mark & x->mark.m) != x->mark.v ||
1014                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1015                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1016                         continue;
1017
1018                 xfrm_state_hold(x);
1019                 return x;
1020         }
1021
1022         if (!create)
1023                 return NULL;
1024
1025         x = xfrm_state_alloc(net);
1026         if (likely(x)) {
1027                 switch (family) {
1028                 case AF_INET:
1029                         x->sel.daddr.a4 = daddr->a4;
1030                         x->sel.saddr.a4 = saddr->a4;
1031                         x->sel.prefixlen_d = 32;
1032                         x->sel.prefixlen_s = 32;
1033                         x->props.saddr.a4 = saddr->a4;
1034                         x->id.daddr.a4 = daddr->a4;
1035                         break;
1036
1037                 case AF_INET6:
1038                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1039                                        (struct in6_addr *)daddr);
1040                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1041                                        (struct in6_addr *)saddr);
1042                         x->sel.prefixlen_d = 128;
1043                         x->sel.prefixlen_s = 128;
1044                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1045                                        (struct in6_addr *)saddr);
1046                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1047                                        (struct in6_addr *)daddr);
1048                         break;
1049                 }
1050
1051                 x->km.state = XFRM_STATE_ACQ;
1052                 x->id.proto = proto;
1053                 x->props.family = family;
1054                 x->props.mode = mode;
1055                 x->props.reqid = reqid;
1056                 x->mark.v = m->v;
1057                 x->mark.m = m->m;
1058                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1059                 xfrm_state_hold(x);
1060                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1061                 list_add(&x->km.all, &net->xfrm.state_all);
1062                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1063                 h = xfrm_src_hash(net, daddr, saddr, family);
1064                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1065
1066                 net->xfrm.state_num++;
1067
1068                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1069         }
1070
1071         return x;
1072 }
1073
1074 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1075
1076 int xfrm_state_add(struct xfrm_state *x)
1077 {
1078         struct net *net = xs_net(x);
1079         struct xfrm_state *x1, *to_put;
1080         int family;
1081         int err;
1082         u32 mark = x->mark.v & x->mark.m;
1083         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1084
1085         family = x->props.family;
1086
1087         to_put = NULL;
1088
1089         spin_lock_bh(&xfrm_state_lock);
1090
1091         x1 = __xfrm_state_locate(x, use_spi, family);
1092         if (x1) {
1093                 to_put = x1;
1094                 x1 = NULL;
1095                 err = -EEXIST;
1096                 goto out;
1097         }
1098
1099         if (use_spi && x->km.seq) {
1100                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1101                 if (x1 && ((x1->id.proto != x->id.proto) ||
1102                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1103                         to_put = x1;
1104                         x1 = NULL;
1105                 }
1106         }
1107
1108         if (use_spi && !x1)
1109                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1110                                      x->props.reqid, x->id.proto,
1111                                      &x->id.daddr, &x->props.saddr, 0);
1112
1113         __xfrm_state_bump_genids(x);
1114         __xfrm_state_insert(x);
1115         err = 0;
1116
1117 out:
1118         spin_unlock_bh(&xfrm_state_lock);
1119
1120         if (x1) {
1121                 xfrm_state_delete(x1);
1122                 xfrm_state_put(x1);
1123         }
1124
1125         if (to_put)
1126                 xfrm_state_put(to_put);
1127
1128         return err;
1129 }
1130 EXPORT_SYMBOL(xfrm_state_add);
1131
1132 #ifdef CONFIG_XFRM_MIGRATE
1133 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1134 {
1135         struct net *net = xs_net(orig);
1136         int err = -ENOMEM;
1137         struct xfrm_state *x = xfrm_state_alloc(net);
1138         if (!x)
1139                 goto out;
1140
1141         memcpy(&x->id, &orig->id, sizeof(x->id));
1142         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1143         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1144         x->props.mode = orig->props.mode;
1145         x->props.replay_window = orig->props.replay_window;
1146         x->props.reqid = orig->props.reqid;
1147         x->props.family = orig->props.family;
1148         x->props.saddr = orig->props.saddr;
1149
1150         if (orig->aalg) {
1151                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1152                 if (!x->aalg)
1153                         goto error;
1154         }
1155         x->props.aalgo = orig->props.aalgo;
1156
1157         if (orig->ealg) {
1158                 x->ealg = xfrm_algo_clone(orig->ealg);
1159                 if (!x->ealg)
1160                         goto error;
1161         }
1162         x->props.ealgo = orig->props.ealgo;
1163
1164         if (orig->calg) {
1165                 x->calg = xfrm_algo_clone(orig->calg);
1166                 if (!x->calg)
1167                         goto error;
1168         }
1169         x->props.calgo = orig->props.calgo;
1170
1171         if (orig->encap) {
1172                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1173                 if (!x->encap)
1174                         goto error;
1175         }
1176
1177         if (orig->coaddr) {
1178                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1179                                     GFP_KERNEL);
1180                 if (!x->coaddr)
1181                         goto error;
1182         }
1183
1184         if (orig->replay_esn) {
1185                 err = xfrm_replay_clone(x, orig);
1186                 if (err)
1187                         goto error;
1188         }
1189
1190         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1191
1192         err = xfrm_init_state(x);
1193         if (err)
1194                 goto error;
1195
1196         x->props.flags = orig->props.flags;
1197
1198         x->curlft.add_time = orig->curlft.add_time;
1199         x->km.state = orig->km.state;
1200         x->km.seq = orig->km.seq;
1201
1202         return x;
1203
1204  error:
1205         xfrm_state_put(x);
1206 out:
1207         if (errp)
1208                 *errp = err;
1209         return NULL;
1210 }
1211
1212 /* xfrm_state_lock is held */
1213 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1214 {
1215         unsigned int h;
1216         struct xfrm_state *x;
1217         struct hlist_node *entry;
1218
1219         if (m->reqid) {
1220                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1221                                   m->reqid, m->old_family);
1222                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1223                         if (x->props.mode != m->mode ||
1224                             x->id.proto != m->proto)
1225                                 continue;
1226                         if (m->reqid && x->props.reqid != m->reqid)
1227                                 continue;
1228                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1229                                           m->old_family) ||
1230                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1231                                           m->old_family))
1232                                 continue;
1233                         xfrm_state_hold(x);
1234                         return x;
1235                 }
1236         } else {
1237                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1238                                   m->old_family);
1239                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1240                         if (x->props.mode != m->mode ||
1241                             x->id.proto != m->proto)
1242                                 continue;
1243                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1244                                           m->old_family) ||
1245                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1246                                           m->old_family))
1247                                 continue;
1248                         xfrm_state_hold(x);
1249                         return x;
1250                 }
1251         }
1252
1253         return NULL;
1254 }
1255 EXPORT_SYMBOL(xfrm_migrate_state_find);
1256
1257 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1258                                        struct xfrm_migrate *m)
1259 {
1260         struct xfrm_state *xc;
1261         int err;
1262
1263         xc = xfrm_state_clone(x, &err);
1264         if (!xc)
1265                 return NULL;
1266
1267         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1268         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1269
1270         /* add state */
1271         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1272                 /* a care is needed when the destination address of the
1273                    state is to be updated as it is a part of triplet */
1274                 xfrm_state_insert(xc);
1275         } else {
1276                 if ((err = xfrm_state_add(xc)) < 0)
1277                         goto error;
1278         }
1279
1280         return xc;
1281 error:
1282         xfrm_state_put(xc);
1283         return NULL;
1284 }
1285 EXPORT_SYMBOL(xfrm_state_migrate);
1286 #endif
1287
1288 int xfrm_state_update(struct xfrm_state *x)
1289 {
1290         struct xfrm_state *x1, *to_put;
1291         int err;
1292         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1293
1294         to_put = NULL;
1295
1296         spin_lock_bh(&xfrm_state_lock);
1297         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1298
1299         err = -ESRCH;
1300         if (!x1)
1301                 goto out;
1302
1303         if (xfrm_state_kern(x1)) {
1304                 to_put = x1;
1305                 err = -EEXIST;
1306                 goto out;
1307         }
1308
1309         if (x1->km.state == XFRM_STATE_ACQ) {
1310                 __xfrm_state_insert(x);
1311                 x = NULL;
1312         }
1313         err = 0;
1314
1315 out:
1316         spin_unlock_bh(&xfrm_state_lock);
1317
1318         if (to_put)
1319                 xfrm_state_put(to_put);
1320
1321         if (err)
1322                 return err;
1323
1324         if (!x) {
1325                 xfrm_state_delete(x1);
1326                 xfrm_state_put(x1);
1327                 return 0;
1328         }
1329
1330         err = -EINVAL;
1331         spin_lock_bh(&x1->lock);
1332         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1333                 if (x->encap && x1->encap)
1334                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1335                 if (x->coaddr && x1->coaddr) {
1336                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1337                 }
1338                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1339                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1340                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1341                 x1->km.dying = 0;
1342
1343                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1344                 if (x1->curlft.use_time)
1345                         xfrm_state_check_expire(x1);
1346
1347                 err = 0;
1348         }
1349         spin_unlock_bh(&x1->lock);
1350
1351         xfrm_state_put(x1);
1352
1353         return err;
1354 }
1355 EXPORT_SYMBOL(xfrm_state_update);
1356
1357 int xfrm_state_check_expire(struct xfrm_state *x)
1358 {
1359         if (!x->curlft.use_time)
1360                 x->curlft.use_time = get_seconds();
1361
1362         if (x->km.state != XFRM_STATE_VALID)
1363                 return -EINVAL;
1364
1365         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1366             x->curlft.packets >= x->lft.hard_packet_limit) {
1367                 x->km.state = XFRM_STATE_EXPIRED;
1368                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1369                 return -EINVAL;
1370         }
1371
1372         if (!x->km.dying &&
1373             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1374              x->curlft.packets >= x->lft.soft_packet_limit)) {
1375                 x->km.dying = 1;
1376                 km_state_expired(x, 0, 0);
1377         }
1378         return 0;
1379 }
1380 EXPORT_SYMBOL(xfrm_state_check_expire);
1381
1382 struct xfrm_state *
1383 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1384                   u8 proto, unsigned short family)
1385 {
1386         struct xfrm_state *x;
1387
1388         spin_lock_bh(&xfrm_state_lock);
1389         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1390         spin_unlock_bh(&xfrm_state_lock);
1391         return x;
1392 }
1393 EXPORT_SYMBOL(xfrm_state_lookup);
1394
1395 struct xfrm_state *
1396 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1397                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1398                          u8 proto, unsigned short family)
1399 {
1400         struct xfrm_state *x;
1401
1402         spin_lock_bh(&xfrm_state_lock);
1403         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1404         spin_unlock_bh(&xfrm_state_lock);
1405         return x;
1406 }
1407 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1408
1409 struct xfrm_state *
1410 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1411               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1412               int create, unsigned short family)
1413 {
1414         struct xfrm_state *x;
1415
1416         spin_lock_bh(&xfrm_state_lock);
1417         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1418         spin_unlock_bh(&xfrm_state_lock);
1419
1420         return x;
1421 }
1422 EXPORT_SYMBOL(xfrm_find_acq);
1423
1424 #ifdef CONFIG_XFRM_SUB_POLICY
1425 int
1426 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1427                unsigned short family)
1428 {
1429         int err = 0;
1430         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1431         if (!afinfo)
1432                 return -EAFNOSUPPORT;
1433
1434         spin_lock_bh(&xfrm_state_lock);
1435         if (afinfo->tmpl_sort)
1436                 err = afinfo->tmpl_sort(dst, src, n);
1437         spin_unlock_bh(&xfrm_state_lock);
1438         xfrm_state_put_afinfo(afinfo);
1439         return err;
1440 }
1441 EXPORT_SYMBOL(xfrm_tmpl_sort);
1442
1443 int
1444 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1445                 unsigned short family)
1446 {
1447         int err = 0;
1448         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1449         if (!afinfo)
1450                 return -EAFNOSUPPORT;
1451
1452         spin_lock_bh(&xfrm_state_lock);
1453         if (afinfo->state_sort)
1454                 err = afinfo->state_sort(dst, src, n);
1455         spin_unlock_bh(&xfrm_state_lock);
1456         xfrm_state_put_afinfo(afinfo);
1457         return err;
1458 }
1459 EXPORT_SYMBOL(xfrm_state_sort);
1460 #endif
1461
1462 /* Silly enough, but I'm lazy to build resolution list */
1463
1464 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1465 {
1466         int i;
1467
1468         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1469                 struct hlist_node *entry;
1470                 struct xfrm_state *x;
1471
1472                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1473                         if (x->km.seq == seq &&
1474                             (mark & x->mark.m) == x->mark.v &&
1475                             x->km.state == XFRM_STATE_ACQ) {
1476                                 xfrm_state_hold(x);
1477                                 return x;
1478                         }
1479                 }
1480         }
1481         return NULL;
1482 }
1483
1484 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1485 {
1486         struct xfrm_state *x;
1487
1488         spin_lock_bh(&xfrm_state_lock);
1489         x = __xfrm_find_acq_byseq(net, mark, seq);
1490         spin_unlock_bh(&xfrm_state_lock);
1491         return x;
1492 }
1493 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1494
1495 u32 xfrm_get_acqseq(void)
1496 {
1497         u32 res;
1498         static atomic_t acqseq;
1499
1500         do {
1501                 res = atomic_inc_return(&acqseq);
1502         } while (!res);
1503
1504         return res;
1505 }
1506 EXPORT_SYMBOL(xfrm_get_acqseq);
1507
1508 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1509 {
1510         struct net *net = xs_net(x);
1511         unsigned int h;
1512         struct xfrm_state *x0;
1513         int err = -ENOENT;
1514         __be32 minspi = htonl(low);
1515         __be32 maxspi = htonl(high);
1516         u32 mark = x->mark.v & x->mark.m;
1517
1518         spin_lock_bh(&x->lock);
1519         if (x->km.state == XFRM_STATE_DEAD)
1520                 goto unlock;
1521
1522         err = 0;
1523         if (x->id.spi)
1524                 goto unlock;
1525
1526         err = -ENOENT;
1527
1528         if (minspi == maxspi) {
1529                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1530                 if (x0) {
1531                         xfrm_state_put(x0);
1532                         goto unlock;
1533                 }
1534                 x->id.spi = minspi;
1535         } else {
1536                 u32 spi = 0;
1537                 for (h=0; h<high-low+1; h++) {
1538                         spi = low + net_random()%(high-low+1);
1539                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1540                         if (x0 == NULL) {
1541                                 x->id.spi = htonl(spi);
1542                                 break;
1543                         }
1544                         xfrm_state_put(x0);
1545                 }
1546         }
1547         if (x->id.spi) {
1548                 spin_lock_bh(&xfrm_state_lock);
1549                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1550                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1551                 spin_unlock_bh(&xfrm_state_lock);
1552
1553                 err = 0;
1554         }
1555
1556 unlock:
1557         spin_unlock_bh(&x->lock);
1558
1559         return err;
1560 }
1561 EXPORT_SYMBOL(xfrm_alloc_spi);
1562
1563 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1564                     int (*func)(struct xfrm_state *, int, void*),
1565                     void *data)
1566 {
1567         struct xfrm_state *state;
1568         struct xfrm_state_walk *x;
1569         int err = 0;
1570
1571         if (walk->seq != 0 && list_empty(&walk->all))
1572                 return 0;
1573
1574         spin_lock_bh(&xfrm_state_lock);
1575         if (list_empty(&walk->all))
1576                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1577         else
1578                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1579         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1580                 if (x->state == XFRM_STATE_DEAD)
1581                         continue;
1582                 state = container_of(x, struct xfrm_state, km);
1583                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1584                         continue;
1585                 err = func(state, walk->seq, data);
1586                 if (err) {
1587                         list_move_tail(&walk->all, &x->all);
1588                         goto out;
1589                 }
1590                 walk->seq++;
1591         }
1592         if (walk->seq == 0) {
1593                 err = -ENOENT;
1594                 goto out;
1595         }
1596         list_del_init(&walk->all);
1597 out:
1598         spin_unlock_bh(&xfrm_state_lock);
1599         return err;
1600 }
1601 EXPORT_SYMBOL(xfrm_state_walk);
1602
1603 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1604 {
1605         INIT_LIST_HEAD(&walk->all);
1606         walk->proto = proto;
1607         walk->state = XFRM_STATE_DEAD;
1608         walk->seq = 0;
1609 }
1610 EXPORT_SYMBOL(xfrm_state_walk_init);
1611
1612 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1613 {
1614         if (list_empty(&walk->all))
1615                 return;
1616
1617         spin_lock_bh(&xfrm_state_lock);
1618         list_del(&walk->all);
1619         spin_unlock_bh(&xfrm_state_lock);
1620 }
1621 EXPORT_SYMBOL(xfrm_state_walk_done);
1622
1623 static void xfrm_replay_timer_handler(unsigned long data)
1624 {
1625         struct xfrm_state *x = (struct xfrm_state*)data;
1626
1627         spin_lock(&x->lock);
1628
1629         if (x->km.state == XFRM_STATE_VALID) {
1630                 if (xfrm_aevent_is_on(xs_net(x)))
1631                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1632                 else
1633                         x->xflags |= XFRM_TIME_DEFER;
1634         }
1635
1636         spin_unlock(&x->lock);
1637 }
1638
1639 static LIST_HEAD(xfrm_km_list);
1640 static DEFINE_RWLOCK(xfrm_km_lock);
1641
1642 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1643 {
1644         struct xfrm_mgr *km;
1645
1646         read_lock(&xfrm_km_lock);
1647         list_for_each_entry(km, &xfrm_km_list, list)
1648                 if (km->notify_policy)
1649                         km->notify_policy(xp, dir, c);
1650         read_unlock(&xfrm_km_lock);
1651 }
1652
1653 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1654 {
1655         struct xfrm_mgr *km;
1656         read_lock(&xfrm_km_lock);
1657         list_for_each_entry(km, &xfrm_km_list, list)
1658                 if (km->notify)
1659                         km->notify(x, c);
1660         read_unlock(&xfrm_km_lock);
1661 }
1662
1663 EXPORT_SYMBOL(km_policy_notify);
1664 EXPORT_SYMBOL(km_state_notify);
1665
1666 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1667 {
1668         struct net *net = xs_net(x);
1669         struct km_event c;
1670
1671         c.data.hard = hard;
1672         c.pid = pid;
1673         c.event = XFRM_MSG_EXPIRE;
1674         km_state_notify(x, &c);
1675
1676         if (hard)
1677                 wake_up(&net->xfrm.km_waitq);
1678 }
1679
1680 EXPORT_SYMBOL(km_state_expired);
1681 /*
1682  * We send to all registered managers regardless of failure
1683  * We are happy with one success
1684 */
1685 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1686 {
1687         int err = -EINVAL, acqret;
1688         struct xfrm_mgr *km;
1689
1690         read_lock(&xfrm_km_lock);
1691         list_for_each_entry(km, &xfrm_km_list, list) {
1692                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1693                 if (!acqret)
1694                         err = acqret;
1695         }
1696         read_unlock(&xfrm_km_lock);
1697         return err;
1698 }
1699 EXPORT_SYMBOL(km_query);
1700
1701 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1702 {
1703         int err = -EINVAL;
1704         struct xfrm_mgr *km;
1705
1706         read_lock(&xfrm_km_lock);
1707         list_for_each_entry(km, &xfrm_km_list, list) {
1708                 if (km->new_mapping)
1709                         err = km->new_mapping(x, ipaddr, sport);
1710                 if (!err)
1711                         break;
1712         }
1713         read_unlock(&xfrm_km_lock);
1714         return err;
1715 }
1716 EXPORT_SYMBOL(km_new_mapping);
1717
1718 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1719 {
1720         struct net *net = xp_net(pol);
1721         struct km_event c;
1722
1723         c.data.hard = hard;
1724         c.pid = pid;
1725         c.event = XFRM_MSG_POLEXPIRE;
1726         km_policy_notify(pol, dir, &c);
1727
1728         if (hard)
1729                 wake_up(&net->xfrm.km_waitq);
1730 }
1731 EXPORT_SYMBOL(km_policy_expired);
1732
1733 #ifdef CONFIG_XFRM_MIGRATE
1734 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1735                const struct xfrm_migrate *m, int num_migrate,
1736                const struct xfrm_kmaddress *k)
1737 {
1738         int err = -EINVAL;
1739         int ret;
1740         struct xfrm_mgr *km;
1741
1742         read_lock(&xfrm_km_lock);
1743         list_for_each_entry(km, &xfrm_km_list, list) {
1744                 if (km->migrate) {
1745                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1746                         if (!ret)
1747                                 err = ret;
1748                 }
1749         }
1750         read_unlock(&xfrm_km_lock);
1751         return err;
1752 }
1753 EXPORT_SYMBOL(km_migrate);
1754 #endif
1755
1756 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1757 {
1758         int err = -EINVAL;
1759         int ret;
1760         struct xfrm_mgr *km;
1761
1762         read_lock(&xfrm_km_lock);
1763         list_for_each_entry(km, &xfrm_km_list, list) {
1764                 if (km->report) {
1765                         ret = km->report(net, proto, sel, addr);
1766                         if (!ret)
1767                                 err = ret;
1768                 }
1769         }
1770         read_unlock(&xfrm_km_lock);
1771         return err;
1772 }
1773 EXPORT_SYMBOL(km_report);
1774
1775 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1776 {
1777         int err;
1778         u8 *data;
1779         struct xfrm_mgr *km;
1780         struct xfrm_policy *pol = NULL;
1781
1782         if (optlen <= 0 || optlen > PAGE_SIZE)
1783                 return -EMSGSIZE;
1784
1785         data = kmalloc(optlen, GFP_KERNEL);
1786         if (!data)
1787                 return -ENOMEM;
1788
1789         err = -EFAULT;
1790         if (copy_from_user(data, optval, optlen))
1791                 goto out;
1792
1793         err = -EINVAL;
1794         read_lock(&xfrm_km_lock);
1795         list_for_each_entry(km, &xfrm_km_list, list) {
1796                 pol = km->compile_policy(sk, optname, data,
1797                                          optlen, &err);
1798                 if (err >= 0)
1799                         break;
1800         }
1801         read_unlock(&xfrm_km_lock);
1802
1803         if (err >= 0) {
1804                 xfrm_sk_policy_insert(sk, err, pol);
1805                 xfrm_pol_put(pol);
1806                 err = 0;
1807         }
1808
1809 out:
1810         kfree(data);
1811         return err;
1812 }
1813 EXPORT_SYMBOL(xfrm_user_policy);
1814
1815 int xfrm_register_km(struct xfrm_mgr *km)
1816 {
1817         write_lock_bh(&xfrm_km_lock);
1818         list_add_tail(&km->list, &xfrm_km_list);
1819         write_unlock_bh(&xfrm_km_lock);
1820         return 0;
1821 }
1822 EXPORT_SYMBOL(xfrm_register_km);
1823
1824 int xfrm_unregister_km(struct xfrm_mgr *km)
1825 {
1826         write_lock_bh(&xfrm_km_lock);
1827         list_del(&km->list);
1828         write_unlock_bh(&xfrm_km_lock);
1829         return 0;
1830 }
1831 EXPORT_SYMBOL(xfrm_unregister_km);
1832
1833 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1834 {
1835         int err = 0;
1836         if (unlikely(afinfo == NULL))
1837                 return -EINVAL;
1838         if (unlikely(afinfo->family >= NPROTO))
1839                 return -EAFNOSUPPORT;
1840         write_lock_bh(&xfrm_state_afinfo_lock);
1841         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1842                 err = -ENOBUFS;
1843         else
1844                 xfrm_state_afinfo[afinfo->family] = afinfo;
1845         write_unlock_bh(&xfrm_state_afinfo_lock);
1846         return err;
1847 }
1848 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1849
1850 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1851 {
1852         int err = 0;
1853         if (unlikely(afinfo == NULL))
1854                 return -EINVAL;
1855         if (unlikely(afinfo->family >= NPROTO))
1856                 return -EAFNOSUPPORT;
1857         write_lock_bh(&xfrm_state_afinfo_lock);
1858         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1859                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1860                         err = -EINVAL;
1861                 else
1862                         xfrm_state_afinfo[afinfo->family] = NULL;
1863         }
1864         write_unlock_bh(&xfrm_state_afinfo_lock);
1865         return err;
1866 }
1867 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1868
1869 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1870 {
1871         struct xfrm_state_afinfo *afinfo;
1872         if (unlikely(family >= NPROTO))
1873                 return NULL;
1874         read_lock(&xfrm_state_afinfo_lock);
1875         afinfo = xfrm_state_afinfo[family];
1876         if (unlikely(!afinfo))
1877                 read_unlock(&xfrm_state_afinfo_lock);
1878         return afinfo;
1879 }
1880
1881 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1882         __releases(xfrm_state_afinfo_lock)
1883 {
1884         read_unlock(&xfrm_state_afinfo_lock);
1885 }
1886
1887 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1888 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1889 {
1890         if (x->tunnel) {
1891                 struct xfrm_state *t = x->tunnel;
1892
1893                 if (atomic_read(&t->tunnel_users) == 2)
1894                         xfrm_state_delete(t);
1895                 atomic_dec(&t->tunnel_users);
1896                 xfrm_state_put(t);
1897                 x->tunnel = NULL;
1898         }
1899 }
1900 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1901
1902 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1903 {
1904         int res;
1905
1906         spin_lock_bh(&x->lock);
1907         if (x->km.state == XFRM_STATE_VALID &&
1908             x->type && x->type->get_mtu)
1909                 res = x->type->get_mtu(x, mtu);
1910         else
1911                 res = mtu - x->props.header_len;
1912         spin_unlock_bh(&x->lock);
1913         return res;
1914 }
1915
1916 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1917 {
1918         struct xfrm_state_afinfo *afinfo;
1919         struct xfrm_mode *inner_mode;
1920         int family = x->props.family;
1921         int err;
1922
1923         err = -EAFNOSUPPORT;
1924         afinfo = xfrm_state_get_afinfo(family);
1925         if (!afinfo)
1926                 goto error;
1927
1928         err = 0;
1929         if (afinfo->init_flags)
1930                 err = afinfo->init_flags(x);
1931
1932         xfrm_state_put_afinfo(afinfo);
1933
1934         if (err)
1935                 goto error;
1936
1937         err = -EPROTONOSUPPORT;
1938
1939         if (x->sel.family != AF_UNSPEC) {
1940                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1941                 if (inner_mode == NULL)
1942                         goto error;
1943
1944                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1945                     family != x->sel.family) {
1946                         xfrm_put_mode(inner_mode);
1947                         goto error;
1948                 }
1949
1950                 x->inner_mode = inner_mode;
1951         } else {
1952                 struct xfrm_mode *inner_mode_iaf;
1953                 int iafamily = AF_INET;
1954
1955                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1956                 if (inner_mode == NULL)
1957                         goto error;
1958
1959                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1960                         xfrm_put_mode(inner_mode);
1961                         goto error;
1962                 }
1963                 x->inner_mode = inner_mode;
1964
1965                 if (x->props.family == AF_INET)
1966                         iafamily = AF_INET6;
1967
1968                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1969                 if (inner_mode_iaf) {
1970                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1971                                 x->inner_mode_iaf = inner_mode_iaf;
1972                         else
1973                                 xfrm_put_mode(inner_mode_iaf);
1974                 }
1975         }
1976
1977         x->type = xfrm_get_type(x->id.proto, family);
1978         if (x->type == NULL)
1979                 goto error;
1980
1981         err = x->type->init_state(x);
1982         if (err)
1983                 goto error;
1984
1985         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1986         if (x->outer_mode == NULL)
1987                 goto error;
1988
1989         if (init_replay) {
1990                 err = xfrm_init_replay(x);
1991                 if (err)
1992                         goto error;
1993         }
1994
1995         x->km.state = XFRM_STATE_VALID;
1996
1997 error:
1998         return err;
1999 }
2000
2001 EXPORT_SYMBOL(__xfrm_init_state);
2002
2003 int xfrm_init_state(struct xfrm_state *x)
2004 {
2005         return __xfrm_init_state(x, true);
2006 }
2007
2008 EXPORT_SYMBOL(xfrm_init_state);
2009
2010 int __net_init xfrm_state_init(struct net *net)
2011 {
2012         unsigned int sz;
2013
2014         INIT_LIST_HEAD(&net->xfrm.state_all);
2015
2016         sz = sizeof(struct hlist_head) * 8;
2017
2018         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2019         if (!net->xfrm.state_bydst)
2020                 goto out_bydst;
2021         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2022         if (!net->xfrm.state_bysrc)
2023                 goto out_bysrc;
2024         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2025         if (!net->xfrm.state_byspi)
2026                 goto out_byspi;
2027         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2028
2029         net->xfrm.state_num = 0;
2030         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2031         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2032         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2033         init_waitqueue_head(&net->xfrm.km_waitq);
2034         return 0;
2035
2036 out_byspi:
2037         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2038 out_bysrc:
2039         xfrm_hash_free(net->xfrm.state_bydst, sz);
2040 out_bydst:
2041         return -ENOMEM;
2042 }
2043
2044 void xfrm_state_fini(struct net *net)
2045 {
2046         struct xfrm_audit audit_info;
2047         unsigned int sz;
2048
2049         flush_work(&net->xfrm.state_hash_work);
2050         audit_info.loginuid = -1;
2051         audit_info.sessionid = -1;
2052         audit_info.secid = 0;
2053         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2054         flush_work(&net->xfrm.state_gc_work);
2055
2056         WARN_ON(!list_empty(&net->xfrm.state_all));
2057
2058         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2059         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2060         xfrm_hash_free(net->xfrm.state_byspi, sz);
2061         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2062         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2063         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2064         xfrm_hash_free(net->xfrm.state_bydst, sz);
2065 }
2066
2067 #ifdef CONFIG_AUDITSYSCALL
2068 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2069                                      struct audit_buffer *audit_buf)
2070 {
2071         struct xfrm_sec_ctx *ctx = x->security;
2072         u32 spi = ntohl(x->id.spi);
2073
2074         if (ctx)
2075                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2076                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2077
2078         switch(x->props.family) {
2079         case AF_INET:
2080                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2081                                  &x->props.saddr.a4, &x->id.daddr.a4);
2082                 break;
2083         case AF_INET6:
2084                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2085                                  x->props.saddr.a6, x->id.daddr.a6);
2086                 break;
2087         }
2088
2089         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2090 }
2091
2092 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2093                                       struct audit_buffer *audit_buf)
2094 {
2095         struct iphdr *iph4;
2096         struct ipv6hdr *iph6;
2097
2098         switch (family) {
2099         case AF_INET:
2100                 iph4 = ip_hdr(skb);
2101                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2102                                  &iph4->saddr, &iph4->daddr);
2103                 break;
2104         case AF_INET6:
2105                 iph6 = ipv6_hdr(skb);
2106                 audit_log_format(audit_buf,
2107                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2108                                  &iph6->saddr,&iph6->daddr,
2109                                  iph6->flow_lbl[0] & 0x0f,
2110                                  iph6->flow_lbl[1],
2111                                  iph6->flow_lbl[2]);
2112                 break;
2113         }
2114 }
2115
2116 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2117                           uid_t auid, u32 sessionid, u32 secid)
2118 {
2119         struct audit_buffer *audit_buf;
2120
2121         audit_buf = xfrm_audit_start("SAD-add");
2122         if (audit_buf == NULL)
2123                 return;
2124         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2125         xfrm_audit_helper_sainfo(x, audit_buf);
2126         audit_log_format(audit_buf, " res=%u", result);
2127         audit_log_end(audit_buf);
2128 }
2129 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2130
2131 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2132                              uid_t auid, u32 sessionid, u32 secid)
2133 {
2134         struct audit_buffer *audit_buf;
2135
2136         audit_buf = xfrm_audit_start("SAD-delete");
2137         if (audit_buf == NULL)
2138                 return;
2139         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2140         xfrm_audit_helper_sainfo(x, audit_buf);
2141         audit_log_format(audit_buf, " res=%u", result);
2142         audit_log_end(audit_buf);
2143 }
2144 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2145
2146 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2147                                       struct sk_buff *skb)
2148 {
2149         struct audit_buffer *audit_buf;
2150         u32 spi;
2151
2152         audit_buf = xfrm_audit_start("SA-replay-overflow");
2153         if (audit_buf == NULL)
2154                 return;
2155         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2156         /* don't record the sequence number because it's inherent in this kind
2157          * of audit message */
2158         spi = ntohl(x->id.spi);
2159         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2160         audit_log_end(audit_buf);
2161 }
2162 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2163
2164 void xfrm_audit_state_replay(struct xfrm_state *x,
2165                              struct sk_buff *skb, __be32 net_seq)
2166 {
2167         struct audit_buffer *audit_buf;
2168         u32 spi;
2169
2170         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2171         if (audit_buf == NULL)
2172                 return;
2173         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2174         spi = ntohl(x->id.spi);
2175         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2176                          spi, spi, ntohl(net_seq));
2177         audit_log_end(audit_buf);
2178 }
2179 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2180
2181 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2182 {
2183         struct audit_buffer *audit_buf;
2184
2185         audit_buf = xfrm_audit_start("SA-notfound");
2186         if (audit_buf == NULL)
2187                 return;
2188         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2189         audit_log_end(audit_buf);
2190 }
2191 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2192
2193 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2194                                __be32 net_spi, __be32 net_seq)
2195 {
2196         struct audit_buffer *audit_buf;
2197         u32 spi;
2198
2199         audit_buf = xfrm_audit_start("SA-notfound");
2200         if (audit_buf == NULL)
2201                 return;
2202         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2203         spi = ntohl(net_spi);
2204         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2205                          spi, spi, ntohl(net_seq));
2206         audit_log_end(audit_buf);
2207 }
2208 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2209
2210 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2211                               struct sk_buff *skb, u8 proto)
2212 {
2213         struct audit_buffer *audit_buf;
2214         __be32 net_spi;
2215         __be32 net_seq;
2216
2217         audit_buf = xfrm_audit_start("SA-icv-failure");
2218         if (audit_buf == NULL)
2219                 return;
2220         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2221         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2222                 u32 spi = ntohl(net_spi);
2223                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2224                                  spi, spi, ntohl(net_seq));
2225         }
2226         audit_log_end(audit_buf);
2227 }
2228 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2229 #endif /* CONFIG_AUDITSYSCALL */