Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
[linux-2.6] / 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
25 #include "xfrm_hash.h"
26
27 struct sock *xfrm_nl;
28 EXPORT_SYMBOL(xfrm_nl);
29
30 u32 sysctl_xfrm_aevent_etime __read_mostly = XFRM_AE_ETIME;
31 EXPORT_SYMBOL(sysctl_xfrm_aevent_etime);
32
33 u32 sysctl_xfrm_aevent_rseqth __read_mostly = XFRM_AE_SEQT_SIZE;
34 EXPORT_SYMBOL(sysctl_xfrm_aevent_rseqth);
35
36 u32 sysctl_xfrm_acq_expires __read_mostly = 30;
37
38 /* Each xfrm_state may be linked to two tables:
39
40    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
41    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
42       destination/tunnel endpoint. (output)
43  */
44
45 static DEFINE_SPINLOCK(xfrm_state_lock);
46
47 /* Hash table to find appropriate SA towards given target (endpoint
48  * of tunnel or destination of transport mode) allowed by selector.
49  *
50  * Main use is finding SA after policy selected tunnel or transport mode.
51  * Also, it can be used by ah/esp icmp error handler to find offending SA.
52  */
53 static LIST_HEAD(xfrm_state_all);
54 static struct hlist_head *xfrm_state_bydst __read_mostly;
55 static struct hlist_head *xfrm_state_bysrc __read_mostly;
56 static struct hlist_head *xfrm_state_byspi __read_mostly;
57 static unsigned int xfrm_state_hmask __read_mostly;
58 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
59 static unsigned int xfrm_state_num;
60 static unsigned int xfrm_state_genid;
61
62 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
63 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
64
65 #ifdef CONFIG_AUDITSYSCALL
66 static void xfrm_audit_state_replay(struct xfrm_state *x,
67                                     struct sk_buff *skb, __be32 net_seq);
68 #else
69 #define xfrm_audit_state_replay(x, s, sq)       do { ; } while (0)
70 #endif /* CONFIG_AUDITSYSCALL */
71
72 static inline unsigned int xfrm_dst_hash(xfrm_address_t *daddr,
73                                          xfrm_address_t *saddr,
74                                          u32 reqid,
75                                          unsigned short family)
76 {
77         return __xfrm_dst_hash(daddr, saddr, reqid, family, xfrm_state_hmask);
78 }
79
80 static inline unsigned int xfrm_src_hash(xfrm_address_t *daddr,
81                                          xfrm_address_t *saddr,
82                                          unsigned short family)
83 {
84         return __xfrm_src_hash(daddr, saddr, family, xfrm_state_hmask);
85 }
86
87 static inline unsigned int
88 xfrm_spi_hash(xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
89 {
90         return __xfrm_spi_hash(daddr, spi, proto, family, xfrm_state_hmask);
91 }
92
93 static void xfrm_hash_transfer(struct hlist_head *list,
94                                struct hlist_head *ndsttable,
95                                struct hlist_head *nsrctable,
96                                struct hlist_head *nspitable,
97                                unsigned int nhashmask)
98 {
99         struct hlist_node *entry, *tmp;
100         struct xfrm_state *x;
101
102         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
103                 unsigned int h;
104
105                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
106                                     x->props.reqid, x->props.family,
107                                     nhashmask);
108                 hlist_add_head(&x->bydst, ndsttable+h);
109
110                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
111                                     x->props.family,
112                                     nhashmask);
113                 hlist_add_head(&x->bysrc, nsrctable+h);
114
115                 if (x->id.spi) {
116                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
117                                             x->id.proto, x->props.family,
118                                             nhashmask);
119                         hlist_add_head(&x->byspi, nspitable+h);
120                 }
121         }
122 }
123
124 static unsigned long xfrm_hash_new_size(void)
125 {
126         return ((xfrm_state_hmask + 1) << 1) *
127                 sizeof(struct hlist_head);
128 }
129
130 static DEFINE_MUTEX(hash_resize_mutex);
131
132 static void xfrm_hash_resize(struct work_struct *__unused)
133 {
134         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
135         unsigned long nsize, osize;
136         unsigned int nhashmask, ohashmask;
137         int i;
138
139         mutex_lock(&hash_resize_mutex);
140
141         nsize = xfrm_hash_new_size();
142         ndst = xfrm_hash_alloc(nsize);
143         if (!ndst)
144                 goto out_unlock;
145         nsrc = xfrm_hash_alloc(nsize);
146         if (!nsrc) {
147                 xfrm_hash_free(ndst, nsize);
148                 goto out_unlock;
149         }
150         nspi = xfrm_hash_alloc(nsize);
151         if (!nspi) {
152                 xfrm_hash_free(ndst, nsize);
153                 xfrm_hash_free(nsrc, nsize);
154                 goto out_unlock;
155         }
156
157         spin_lock_bh(&xfrm_state_lock);
158
159         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
160         for (i = xfrm_state_hmask; i >= 0; i--)
161                 xfrm_hash_transfer(xfrm_state_bydst+i, ndst, nsrc, nspi,
162                                    nhashmask);
163
164         odst = xfrm_state_bydst;
165         osrc = xfrm_state_bysrc;
166         ospi = xfrm_state_byspi;
167         ohashmask = xfrm_state_hmask;
168
169         xfrm_state_bydst = ndst;
170         xfrm_state_bysrc = nsrc;
171         xfrm_state_byspi = nspi;
172         xfrm_state_hmask = nhashmask;
173
174         spin_unlock_bh(&xfrm_state_lock);
175
176         osize = (ohashmask + 1) * sizeof(struct hlist_head);
177         xfrm_hash_free(odst, osize);
178         xfrm_hash_free(osrc, osize);
179         xfrm_hash_free(ospi, osize);
180
181 out_unlock:
182         mutex_unlock(&hash_resize_mutex);
183 }
184
185 static DECLARE_WORK(xfrm_hash_work, xfrm_hash_resize);
186
187 DECLARE_WAIT_QUEUE_HEAD(km_waitq);
188 EXPORT_SYMBOL(km_waitq);
189
190 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
191 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
192
193 static struct work_struct xfrm_state_gc_work;
194 static HLIST_HEAD(xfrm_state_gc_list);
195 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
196
197 int __xfrm_state_delete(struct xfrm_state *x);
198
199 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
200 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
201
202 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
203 {
204         struct xfrm_state_afinfo *afinfo;
205         if (unlikely(family >= NPROTO))
206                 return NULL;
207         write_lock_bh(&xfrm_state_afinfo_lock);
208         afinfo = xfrm_state_afinfo[family];
209         if (unlikely(!afinfo))
210                 write_unlock_bh(&xfrm_state_afinfo_lock);
211         return afinfo;
212 }
213
214 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
215         __releases(xfrm_state_afinfo_lock)
216 {
217         write_unlock_bh(&xfrm_state_afinfo_lock);
218 }
219
220 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
221 {
222         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
223         const struct xfrm_type **typemap;
224         int err = 0;
225
226         if (unlikely(afinfo == NULL))
227                 return -EAFNOSUPPORT;
228         typemap = afinfo->type_map;
229
230         if (likely(typemap[type->proto] == NULL))
231                 typemap[type->proto] = type;
232         else
233                 err = -EEXIST;
234         xfrm_state_unlock_afinfo(afinfo);
235         return err;
236 }
237 EXPORT_SYMBOL(xfrm_register_type);
238
239 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
240 {
241         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
242         const struct xfrm_type **typemap;
243         int err = 0;
244
245         if (unlikely(afinfo == NULL))
246                 return -EAFNOSUPPORT;
247         typemap = afinfo->type_map;
248
249         if (unlikely(typemap[type->proto] != type))
250                 err = -ENOENT;
251         else
252                 typemap[type->proto] = NULL;
253         xfrm_state_unlock_afinfo(afinfo);
254         return err;
255 }
256 EXPORT_SYMBOL(xfrm_unregister_type);
257
258 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
259 {
260         struct xfrm_state_afinfo *afinfo;
261         const struct xfrm_type **typemap;
262         const struct xfrm_type *type;
263         int modload_attempted = 0;
264
265 retry:
266         afinfo = xfrm_state_get_afinfo(family);
267         if (unlikely(afinfo == NULL))
268                 return NULL;
269         typemap = afinfo->type_map;
270
271         type = typemap[proto];
272         if (unlikely(type && !try_module_get(type->owner)))
273                 type = NULL;
274         if (!type && !modload_attempted) {
275                 xfrm_state_put_afinfo(afinfo);
276                 request_module("xfrm-type-%d-%d", family, proto);
277                 modload_attempted = 1;
278                 goto retry;
279         }
280
281         xfrm_state_put_afinfo(afinfo);
282         return type;
283 }
284
285 static void xfrm_put_type(const struct xfrm_type *type)
286 {
287         module_put(type->owner);
288 }
289
290 int xfrm_register_mode(struct xfrm_mode *mode, int family)
291 {
292         struct xfrm_state_afinfo *afinfo;
293         struct xfrm_mode **modemap;
294         int err;
295
296         if (unlikely(mode->encap >= XFRM_MODE_MAX))
297                 return -EINVAL;
298
299         afinfo = xfrm_state_lock_afinfo(family);
300         if (unlikely(afinfo == NULL))
301                 return -EAFNOSUPPORT;
302
303         err = -EEXIST;
304         modemap = afinfo->mode_map;
305         if (modemap[mode->encap])
306                 goto out;
307
308         err = -ENOENT;
309         if (!try_module_get(afinfo->owner))
310                 goto out;
311
312         mode->afinfo = afinfo;
313         modemap[mode->encap] = mode;
314         err = 0;
315
316 out:
317         xfrm_state_unlock_afinfo(afinfo);
318         return err;
319 }
320 EXPORT_SYMBOL(xfrm_register_mode);
321
322 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
323 {
324         struct xfrm_state_afinfo *afinfo;
325         struct xfrm_mode **modemap;
326         int err;
327
328         if (unlikely(mode->encap >= XFRM_MODE_MAX))
329                 return -EINVAL;
330
331         afinfo = xfrm_state_lock_afinfo(family);
332         if (unlikely(afinfo == NULL))
333                 return -EAFNOSUPPORT;
334
335         err = -ENOENT;
336         modemap = afinfo->mode_map;
337         if (likely(modemap[mode->encap] == mode)) {
338                 modemap[mode->encap] = NULL;
339                 module_put(mode->afinfo->owner);
340                 err = 0;
341         }
342
343         xfrm_state_unlock_afinfo(afinfo);
344         return err;
345 }
346 EXPORT_SYMBOL(xfrm_unregister_mode);
347
348 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
349 {
350         struct xfrm_state_afinfo *afinfo;
351         struct xfrm_mode *mode;
352         int modload_attempted = 0;
353
354         if (unlikely(encap >= XFRM_MODE_MAX))
355                 return NULL;
356
357 retry:
358         afinfo = xfrm_state_get_afinfo(family);
359         if (unlikely(afinfo == NULL))
360                 return NULL;
361
362         mode = afinfo->mode_map[encap];
363         if (unlikely(mode && !try_module_get(mode->owner)))
364                 mode = NULL;
365         if (!mode && !modload_attempted) {
366                 xfrm_state_put_afinfo(afinfo);
367                 request_module("xfrm-mode-%d-%d", family, encap);
368                 modload_attempted = 1;
369                 goto retry;
370         }
371
372         xfrm_state_put_afinfo(afinfo);
373         return mode;
374 }
375
376 static void xfrm_put_mode(struct xfrm_mode *mode)
377 {
378         module_put(mode->owner);
379 }
380
381 static void xfrm_state_gc_destroy(struct xfrm_state *x)
382 {
383         del_timer_sync(&x->timer);
384         del_timer_sync(&x->rtimer);
385         kfree(x->aalg);
386         kfree(x->ealg);
387         kfree(x->calg);
388         kfree(x->encap);
389         kfree(x->coaddr);
390         if (x->inner_mode)
391                 xfrm_put_mode(x->inner_mode);
392         if (x->outer_mode)
393                 xfrm_put_mode(x->outer_mode);
394         if (x->type) {
395                 x->type->destructor(x);
396                 xfrm_put_type(x->type);
397         }
398         security_xfrm_state_free(x);
399         kfree(x);
400 }
401
402 static void xfrm_state_gc_task(struct work_struct *data)
403 {
404         struct xfrm_state *x;
405         struct hlist_node *entry, *tmp;
406         struct hlist_head gc_list;
407
408         spin_lock_bh(&xfrm_state_gc_lock);
409         gc_list.first = xfrm_state_gc_list.first;
410         INIT_HLIST_HEAD(&xfrm_state_gc_list);
411         spin_unlock_bh(&xfrm_state_gc_lock);
412
413         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, bydst)
414                 xfrm_state_gc_destroy(x);
415
416         wake_up(&km_waitq);
417 }
418
419 static inline unsigned long make_jiffies(long secs)
420 {
421         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
422                 return MAX_SCHEDULE_TIMEOUT-1;
423         else
424                 return secs*HZ;
425 }
426
427 static void xfrm_timer_handler(unsigned long data)
428 {
429         struct xfrm_state *x = (struct xfrm_state*)data;
430         unsigned long now = get_seconds();
431         long next = LONG_MAX;
432         int warn = 0;
433         int err = 0;
434
435         spin_lock(&x->lock);
436         if (x->km.state == XFRM_STATE_DEAD)
437                 goto out;
438         if (x->km.state == XFRM_STATE_EXPIRED)
439                 goto expired;
440         if (x->lft.hard_add_expires_seconds) {
441                 long tmo = x->lft.hard_add_expires_seconds +
442                         x->curlft.add_time - now;
443                 if (tmo <= 0)
444                         goto expired;
445                 if (tmo < next)
446                         next = tmo;
447         }
448         if (x->lft.hard_use_expires_seconds) {
449                 long tmo = x->lft.hard_use_expires_seconds +
450                         (x->curlft.use_time ? : now) - now;
451                 if (tmo <= 0)
452                         goto expired;
453                 if (tmo < next)
454                         next = tmo;
455         }
456         if (x->km.dying)
457                 goto resched;
458         if (x->lft.soft_add_expires_seconds) {
459                 long tmo = x->lft.soft_add_expires_seconds +
460                         x->curlft.add_time - now;
461                 if (tmo <= 0)
462                         warn = 1;
463                 else if (tmo < next)
464                         next = tmo;
465         }
466         if (x->lft.soft_use_expires_seconds) {
467                 long tmo = x->lft.soft_use_expires_seconds +
468                         (x->curlft.use_time ? : now) - now;
469                 if (tmo <= 0)
470                         warn = 1;
471                 else if (tmo < next)
472                         next = tmo;
473         }
474
475         x->km.dying = warn;
476         if (warn)
477                 km_state_expired(x, 0, 0);
478 resched:
479         if (next != LONG_MAX)
480                 mod_timer(&x->timer, jiffies + make_jiffies(next));
481
482         goto out;
483
484 expired:
485         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
486                 x->km.state = XFRM_STATE_EXPIRED;
487                 wake_up(&km_waitq);
488                 next = 2;
489                 goto resched;
490         }
491
492         err = __xfrm_state_delete(x);
493         if (!err && x->id.spi)
494                 km_state_expired(x, 1, 0);
495
496         xfrm_audit_state_delete(x, err ? 0 : 1,
497                                 audit_get_loginuid(current), 0);
498
499 out:
500         spin_unlock(&x->lock);
501 }
502
503 static void xfrm_replay_timer_handler(unsigned long data);
504
505 struct xfrm_state *xfrm_state_alloc(void)
506 {
507         struct xfrm_state *x;
508
509         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
510
511         if (x) {
512                 atomic_set(&x->refcnt, 1);
513                 atomic_set(&x->tunnel_users, 0);
514                 INIT_LIST_HEAD(&x->all);
515                 INIT_HLIST_NODE(&x->bydst);
516                 INIT_HLIST_NODE(&x->bysrc);
517                 INIT_HLIST_NODE(&x->byspi);
518                 setup_timer(&x->timer, xfrm_timer_handler, (unsigned long)x);
519                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
520                                 (unsigned long)x);
521                 x->curlft.add_time = get_seconds();
522                 x->lft.soft_byte_limit = XFRM_INF;
523                 x->lft.soft_packet_limit = XFRM_INF;
524                 x->lft.hard_byte_limit = XFRM_INF;
525                 x->lft.hard_packet_limit = XFRM_INF;
526                 x->replay_maxage = 0;
527                 x->replay_maxdiff = 0;
528                 spin_lock_init(&x->lock);
529         }
530         return x;
531 }
532 EXPORT_SYMBOL(xfrm_state_alloc);
533
534 void __xfrm_state_destroy(struct xfrm_state *x)
535 {
536         BUG_TRAP(x->km.state == XFRM_STATE_DEAD);
537
538         spin_lock_bh(&xfrm_state_lock);
539         list_del(&x->all);
540         spin_unlock_bh(&xfrm_state_lock);
541
542         spin_lock_bh(&xfrm_state_gc_lock);
543         hlist_add_head(&x->bydst, &xfrm_state_gc_list);
544         spin_unlock_bh(&xfrm_state_gc_lock);
545         schedule_work(&xfrm_state_gc_work);
546 }
547 EXPORT_SYMBOL(__xfrm_state_destroy);
548
549 int __xfrm_state_delete(struct xfrm_state *x)
550 {
551         int err = -ESRCH;
552
553         if (x->km.state != XFRM_STATE_DEAD) {
554                 x->km.state = XFRM_STATE_DEAD;
555                 spin_lock(&xfrm_state_lock);
556                 hlist_del(&x->bydst);
557                 hlist_del(&x->bysrc);
558                 if (x->id.spi)
559                         hlist_del(&x->byspi);
560                 xfrm_state_num--;
561                 spin_unlock(&xfrm_state_lock);
562
563                 /* All xfrm_state objects are created by xfrm_state_alloc.
564                  * The xfrm_state_alloc call gives a reference, and that
565                  * is what we are dropping here.
566                  */
567                 xfrm_state_put(x);
568                 err = 0;
569         }
570
571         return err;
572 }
573 EXPORT_SYMBOL(__xfrm_state_delete);
574
575 int xfrm_state_delete(struct xfrm_state *x)
576 {
577         int err;
578
579         spin_lock_bh(&x->lock);
580         err = __xfrm_state_delete(x);
581         spin_unlock_bh(&x->lock);
582
583         return err;
584 }
585 EXPORT_SYMBOL(xfrm_state_delete);
586
587 #ifdef CONFIG_SECURITY_NETWORK_XFRM
588 static inline int
589 xfrm_state_flush_secctx_check(u8 proto, struct xfrm_audit *audit_info)
590 {
591         int i, err = 0;
592
593         for (i = 0; i <= xfrm_state_hmask; i++) {
594                 struct hlist_node *entry;
595                 struct xfrm_state *x;
596
597                 hlist_for_each_entry(x, entry, xfrm_state_bydst+i, bydst) {
598                         if (xfrm_id_proto_match(x->id.proto, proto) &&
599                            (err = security_xfrm_state_delete(x)) != 0) {
600                                 xfrm_audit_state_delete(x, 0,
601                                                         audit_info->loginuid,
602                                                         audit_info->secid);
603                                 return err;
604                         }
605                 }
606         }
607
608         return err;
609 }
610 #else
611 static inline int
612 xfrm_state_flush_secctx_check(u8 proto, struct xfrm_audit *audit_info)
613 {
614         return 0;
615 }
616 #endif
617
618 int xfrm_state_flush(u8 proto, struct xfrm_audit *audit_info)
619 {
620         int i, err = 0;
621
622         spin_lock_bh(&xfrm_state_lock);
623         err = xfrm_state_flush_secctx_check(proto, audit_info);
624         if (err)
625                 goto out;
626
627         for (i = 0; i <= xfrm_state_hmask; i++) {
628                 struct hlist_node *entry;
629                 struct xfrm_state *x;
630 restart:
631                 hlist_for_each_entry(x, entry, xfrm_state_bydst+i, bydst) {
632                         if (!xfrm_state_kern(x) &&
633                             xfrm_id_proto_match(x->id.proto, proto)) {
634                                 xfrm_state_hold(x);
635                                 spin_unlock_bh(&xfrm_state_lock);
636
637                                 err = xfrm_state_delete(x);
638                                 xfrm_audit_state_delete(x, err ? 0 : 1,
639                                                         audit_info->loginuid,
640                                                         audit_info->secid);
641                                 xfrm_state_put(x);
642
643                                 spin_lock_bh(&xfrm_state_lock);
644                                 goto restart;
645                         }
646                 }
647         }
648         err = 0;
649
650 out:
651         spin_unlock_bh(&xfrm_state_lock);
652         wake_up(&km_waitq);
653         return err;
654 }
655 EXPORT_SYMBOL(xfrm_state_flush);
656
657 void xfrm_sad_getinfo(struct xfrmk_sadinfo *si)
658 {
659         spin_lock_bh(&xfrm_state_lock);
660         si->sadcnt = xfrm_state_num;
661         si->sadhcnt = xfrm_state_hmask;
662         si->sadhmcnt = xfrm_state_hashmax;
663         spin_unlock_bh(&xfrm_state_lock);
664 }
665 EXPORT_SYMBOL(xfrm_sad_getinfo);
666
667 static int
668 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
669                   struct xfrm_tmpl *tmpl,
670                   xfrm_address_t *daddr, xfrm_address_t *saddr,
671                   unsigned short family)
672 {
673         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
674         if (!afinfo)
675                 return -1;
676         afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
677         xfrm_state_put_afinfo(afinfo);
678         return 0;
679 }
680
681 static struct xfrm_state *__xfrm_state_lookup(xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
682 {
683         unsigned int h = xfrm_spi_hash(daddr, spi, proto, family);
684         struct xfrm_state *x;
685         struct hlist_node *entry;
686
687         hlist_for_each_entry(x, entry, xfrm_state_byspi+h, byspi) {
688                 if (x->props.family != family ||
689                     x->id.spi       != spi ||
690                     x->id.proto     != proto)
691                         continue;
692
693                 switch (family) {
694                 case AF_INET:
695                         if (x->id.daddr.a4 != daddr->a4)
696                                 continue;
697                         break;
698                 case AF_INET6:
699                         if (!ipv6_addr_equal((struct in6_addr *)daddr,
700                                              (struct in6_addr *)
701                                              x->id.daddr.a6))
702                                 continue;
703                         break;
704                 }
705
706                 xfrm_state_hold(x);
707                 return x;
708         }
709
710         return NULL;
711 }
712
713 static struct xfrm_state *__xfrm_state_lookup_byaddr(xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
714 {
715         unsigned int h = xfrm_src_hash(daddr, saddr, family);
716         struct xfrm_state *x;
717         struct hlist_node *entry;
718
719         hlist_for_each_entry(x, entry, xfrm_state_bysrc+h, bysrc) {
720                 if (x->props.family != family ||
721                     x->id.proto     != proto)
722                         continue;
723
724                 switch (family) {
725                 case AF_INET:
726                         if (x->id.daddr.a4 != daddr->a4 ||
727                             x->props.saddr.a4 != saddr->a4)
728                                 continue;
729                         break;
730                 case AF_INET6:
731                         if (!ipv6_addr_equal((struct in6_addr *)daddr,
732                                              (struct in6_addr *)
733                                              x->id.daddr.a6) ||
734                             !ipv6_addr_equal((struct in6_addr *)saddr,
735                                              (struct in6_addr *)
736                                              x->props.saddr.a6))
737                                 continue;
738                         break;
739                 }
740
741                 xfrm_state_hold(x);
742                 return x;
743         }
744
745         return NULL;
746 }
747
748 static inline struct xfrm_state *
749 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
750 {
751         if (use_spi)
752                 return __xfrm_state_lookup(&x->id.daddr, x->id.spi,
753                                            x->id.proto, family);
754         else
755                 return __xfrm_state_lookup_byaddr(&x->id.daddr,
756                                                   &x->props.saddr,
757                                                   x->id.proto, family);
758 }
759
760 static void xfrm_hash_grow_check(int have_hash_collision)
761 {
762         if (have_hash_collision &&
763             (xfrm_state_hmask + 1) < xfrm_state_hashmax &&
764             xfrm_state_num > xfrm_state_hmask)
765                 schedule_work(&xfrm_hash_work);
766 }
767
768 struct xfrm_state *
769 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
770                 struct flowi *fl, struct xfrm_tmpl *tmpl,
771                 struct xfrm_policy *pol, int *err,
772                 unsigned short family)
773 {
774         unsigned int h;
775         struct hlist_node *entry;
776         struct xfrm_state *x, *x0;
777         int acquire_in_progress = 0;
778         int error = 0;
779         struct xfrm_state *best = NULL;
780
781         spin_lock_bh(&xfrm_state_lock);
782         h = xfrm_dst_hash(daddr, saddr, tmpl->reqid, family);
783         hlist_for_each_entry(x, entry, xfrm_state_bydst+h, bydst) {
784                 if (x->props.family == family &&
785                     x->props.reqid == tmpl->reqid &&
786                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
787                     xfrm_state_addr_check(x, daddr, saddr, family) &&
788                     tmpl->mode == x->props.mode &&
789                     tmpl->id.proto == x->id.proto &&
790                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi)) {
791                         /* Resolution logic:
792                            1. There is a valid state with matching selector.
793                               Done.
794                            2. Valid state with inappropriate selector. Skip.
795
796                            Entering area of "sysdeps".
797
798                            3. If state is not valid, selector is temporary,
799                               it selects only session which triggered
800                               previous resolution. Key manager will do
801                               something to install a state with proper
802                               selector.
803                          */
804                         if (x->km.state == XFRM_STATE_VALID) {
805                                 if (!xfrm_selector_match(&x->sel, fl, x->sel.family) ||
806                                     !security_xfrm_state_pol_flow_match(x, pol, fl))
807                                         continue;
808                                 if (!best ||
809                                     best->km.dying > x->km.dying ||
810                                     (best->km.dying == x->km.dying &&
811                                      best->curlft.add_time < x->curlft.add_time))
812                                         best = x;
813                         } else if (x->km.state == XFRM_STATE_ACQ) {
814                                 acquire_in_progress = 1;
815                         } else if (x->km.state == XFRM_STATE_ERROR ||
816                                    x->km.state == XFRM_STATE_EXPIRED) {
817                                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
818                                     security_xfrm_state_pol_flow_match(x, pol, fl))
819                                         error = -ESRCH;
820                         }
821                 }
822         }
823
824         x = best;
825         if (!x && !error && !acquire_in_progress) {
826                 if (tmpl->id.spi &&
827                     (x0 = __xfrm_state_lookup(daddr, tmpl->id.spi,
828                                               tmpl->id.proto, family)) != NULL) {
829                         xfrm_state_put(x0);
830                         error = -EEXIST;
831                         goto out;
832                 }
833                 x = xfrm_state_alloc();
834                 if (x == NULL) {
835                         error = -ENOMEM;
836                         goto out;
837                 }
838                 /* Initialize temporary selector matching only
839                  * to current session. */
840                 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
841
842                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
843                 if (error) {
844                         x->km.state = XFRM_STATE_DEAD;
845                         xfrm_state_put(x);
846                         x = NULL;
847                         goto out;
848                 }
849
850                 if (km_query(x, tmpl, pol) == 0) {
851                         x->km.state = XFRM_STATE_ACQ;
852                         hlist_add_head(&x->bydst, xfrm_state_bydst+h);
853                         h = xfrm_src_hash(daddr, saddr, family);
854                         hlist_add_head(&x->bysrc, xfrm_state_bysrc+h);
855                         if (x->id.spi) {
856                                 h = xfrm_spi_hash(&x->id.daddr, x->id.spi, x->id.proto, family);
857                                 hlist_add_head(&x->byspi, xfrm_state_byspi+h);
858                         }
859                         x->lft.hard_add_expires_seconds = sysctl_xfrm_acq_expires;
860                         x->timer.expires = jiffies + sysctl_xfrm_acq_expires*HZ;
861                         add_timer(&x->timer);
862                         xfrm_state_num++;
863                         xfrm_hash_grow_check(x->bydst.next != NULL);
864                 } else {
865                         x->km.state = XFRM_STATE_DEAD;
866                         xfrm_state_put(x);
867                         x = NULL;
868                         error = -ESRCH;
869                 }
870         }
871 out:
872         if (x)
873                 xfrm_state_hold(x);
874         else
875                 *err = acquire_in_progress ? -EAGAIN : error;
876         spin_unlock_bh(&xfrm_state_lock);
877         return x;
878 }
879
880 struct xfrm_state *
881 xfrm_stateonly_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
882                     unsigned short family, u8 mode, u8 proto, u32 reqid)
883 {
884         unsigned int h;
885         struct xfrm_state *rx = NULL, *x = NULL;
886         struct hlist_node *entry;
887
888         spin_lock(&xfrm_state_lock);
889         h = xfrm_dst_hash(daddr, saddr, reqid, family);
890         hlist_for_each_entry(x, entry, xfrm_state_bydst+h, bydst) {
891                 if (x->props.family == family &&
892                     x->props.reqid == reqid &&
893                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
894                     xfrm_state_addr_check(x, daddr, saddr, family) &&
895                     mode == x->props.mode &&
896                     proto == x->id.proto &&
897                     x->km.state == XFRM_STATE_VALID) {
898                         rx = x;
899                         break;
900                 }
901         }
902
903         if (rx)
904                 xfrm_state_hold(rx);
905         spin_unlock(&xfrm_state_lock);
906
907
908         return rx;
909 }
910 EXPORT_SYMBOL(xfrm_stateonly_find);
911
912 static void __xfrm_state_insert(struct xfrm_state *x)
913 {
914         unsigned int h;
915
916         x->genid = ++xfrm_state_genid;
917
918         list_add_tail(&x->all, &xfrm_state_all);
919
920         h = xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
921                           x->props.reqid, x->props.family);
922         hlist_add_head(&x->bydst, xfrm_state_bydst+h);
923
924         h = xfrm_src_hash(&x->id.daddr, &x->props.saddr, x->props.family);
925         hlist_add_head(&x->bysrc, xfrm_state_bysrc+h);
926
927         if (x->id.spi) {
928                 h = xfrm_spi_hash(&x->id.daddr, x->id.spi, x->id.proto,
929                                   x->props.family);
930
931                 hlist_add_head(&x->byspi, xfrm_state_byspi+h);
932         }
933
934         mod_timer(&x->timer, jiffies + HZ);
935         if (x->replay_maxage)
936                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
937
938         wake_up(&km_waitq);
939
940         xfrm_state_num++;
941
942         xfrm_hash_grow_check(x->bydst.next != NULL);
943 }
944
945 /* xfrm_state_lock is held */
946 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
947 {
948         unsigned short family = xnew->props.family;
949         u32 reqid = xnew->props.reqid;
950         struct xfrm_state *x;
951         struct hlist_node *entry;
952         unsigned int h;
953
954         h = xfrm_dst_hash(&xnew->id.daddr, &xnew->props.saddr, reqid, family);
955         hlist_for_each_entry(x, entry, xfrm_state_bydst+h, bydst) {
956                 if (x->props.family     == family &&
957                     x->props.reqid      == reqid &&
958                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
959                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
960                         x->genid = xfrm_state_genid;
961         }
962 }
963
964 void xfrm_state_insert(struct xfrm_state *x)
965 {
966         spin_lock_bh(&xfrm_state_lock);
967         __xfrm_state_bump_genids(x);
968         __xfrm_state_insert(x);
969         spin_unlock_bh(&xfrm_state_lock);
970 }
971 EXPORT_SYMBOL(xfrm_state_insert);
972
973 /* xfrm_state_lock is held */
974 static struct xfrm_state *__find_acq_core(unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
975 {
976         unsigned int h = xfrm_dst_hash(daddr, saddr, reqid, family);
977         struct hlist_node *entry;
978         struct xfrm_state *x;
979
980         hlist_for_each_entry(x, entry, xfrm_state_bydst+h, bydst) {
981                 if (x->props.reqid  != reqid ||
982                     x->props.mode   != mode ||
983                     x->props.family != family ||
984                     x->km.state     != XFRM_STATE_ACQ ||
985                     x->id.spi       != 0 ||
986                     x->id.proto     != proto)
987                         continue;
988
989                 switch (family) {
990                 case AF_INET:
991                         if (x->id.daddr.a4    != daddr->a4 ||
992                             x->props.saddr.a4 != saddr->a4)
993                                 continue;
994                         break;
995                 case AF_INET6:
996                         if (!ipv6_addr_equal((struct in6_addr *)x->id.daddr.a6,
997                                              (struct in6_addr *)daddr) ||
998                             !ipv6_addr_equal((struct in6_addr *)
999                                              x->props.saddr.a6,
1000                                              (struct in6_addr *)saddr))
1001                                 continue;
1002                         break;
1003                 }
1004
1005                 xfrm_state_hold(x);
1006                 return x;
1007         }
1008
1009         if (!create)
1010                 return NULL;
1011
1012         x = xfrm_state_alloc();
1013         if (likely(x)) {
1014                 switch (family) {
1015                 case AF_INET:
1016                         x->sel.daddr.a4 = daddr->a4;
1017                         x->sel.saddr.a4 = saddr->a4;
1018                         x->sel.prefixlen_d = 32;
1019                         x->sel.prefixlen_s = 32;
1020                         x->props.saddr.a4 = saddr->a4;
1021                         x->id.daddr.a4 = daddr->a4;
1022                         break;
1023
1024                 case AF_INET6:
1025                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1026                                        (struct in6_addr *)daddr);
1027                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1028                                        (struct in6_addr *)saddr);
1029                         x->sel.prefixlen_d = 128;
1030                         x->sel.prefixlen_s = 128;
1031                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1032                                        (struct in6_addr *)saddr);
1033                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1034                                        (struct in6_addr *)daddr);
1035                         break;
1036                 }
1037
1038                 x->km.state = XFRM_STATE_ACQ;
1039                 x->id.proto = proto;
1040                 x->props.family = family;
1041                 x->props.mode = mode;
1042                 x->props.reqid = reqid;
1043                 x->lft.hard_add_expires_seconds = sysctl_xfrm_acq_expires;
1044                 xfrm_state_hold(x);
1045                 x->timer.expires = jiffies + sysctl_xfrm_acq_expires*HZ;
1046                 add_timer(&x->timer);
1047                 hlist_add_head(&x->bydst, xfrm_state_bydst+h);
1048                 h = xfrm_src_hash(daddr, saddr, family);
1049                 hlist_add_head(&x->bysrc, xfrm_state_bysrc+h);
1050
1051                 xfrm_state_num++;
1052
1053                 xfrm_hash_grow_check(x->bydst.next != NULL);
1054         }
1055
1056         return x;
1057 }
1058
1059 static struct xfrm_state *__xfrm_find_acq_byseq(u32 seq);
1060
1061 int xfrm_state_add(struct xfrm_state *x)
1062 {
1063         struct xfrm_state *x1;
1064         int family;
1065         int err;
1066         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1067
1068         family = x->props.family;
1069
1070         spin_lock_bh(&xfrm_state_lock);
1071
1072         x1 = __xfrm_state_locate(x, use_spi, family);
1073         if (x1) {
1074                 xfrm_state_put(x1);
1075                 x1 = NULL;
1076                 err = -EEXIST;
1077                 goto out;
1078         }
1079
1080         if (use_spi && x->km.seq) {
1081                 x1 = __xfrm_find_acq_byseq(x->km.seq);
1082                 if (x1 && ((x1->id.proto != x->id.proto) ||
1083                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1084                         xfrm_state_put(x1);
1085                         x1 = NULL;
1086                 }
1087         }
1088
1089         if (use_spi && !x1)
1090                 x1 = __find_acq_core(family, x->props.mode, x->props.reqid,
1091                                      x->id.proto,
1092                                      &x->id.daddr, &x->props.saddr, 0);
1093
1094         __xfrm_state_bump_genids(x);
1095         __xfrm_state_insert(x);
1096         err = 0;
1097
1098 out:
1099         spin_unlock_bh(&xfrm_state_lock);
1100
1101         if (x1) {
1102                 xfrm_state_delete(x1);
1103                 xfrm_state_put(x1);
1104         }
1105
1106         return err;
1107 }
1108 EXPORT_SYMBOL(xfrm_state_add);
1109
1110 #ifdef CONFIG_XFRM_MIGRATE
1111 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1112 {
1113         int err = -ENOMEM;
1114         struct xfrm_state *x = xfrm_state_alloc();
1115         if (!x)
1116                 goto error;
1117
1118         memcpy(&x->id, &orig->id, sizeof(x->id));
1119         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1120         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1121         x->props.mode = orig->props.mode;
1122         x->props.replay_window = orig->props.replay_window;
1123         x->props.reqid = orig->props.reqid;
1124         x->props.family = orig->props.family;
1125         x->props.saddr = orig->props.saddr;
1126
1127         if (orig->aalg) {
1128                 x->aalg = xfrm_algo_clone(orig->aalg);
1129                 if (!x->aalg)
1130                         goto error;
1131         }
1132         x->props.aalgo = orig->props.aalgo;
1133
1134         if (orig->ealg) {
1135                 x->ealg = xfrm_algo_clone(orig->ealg);
1136                 if (!x->ealg)
1137                         goto error;
1138         }
1139         x->props.ealgo = orig->props.ealgo;
1140
1141         if (orig->calg) {
1142                 x->calg = xfrm_algo_clone(orig->calg);
1143                 if (!x->calg)
1144                         goto error;
1145         }
1146         x->props.calgo = orig->props.calgo;
1147
1148         if (orig->encap) {
1149                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1150                 if (!x->encap)
1151                         goto error;
1152         }
1153
1154         if (orig->coaddr) {
1155                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1156                                     GFP_KERNEL);
1157                 if (!x->coaddr)
1158                         goto error;
1159         }
1160
1161         err = xfrm_init_state(x);
1162         if (err)
1163                 goto error;
1164
1165         x->props.flags = orig->props.flags;
1166
1167         x->curlft.add_time = orig->curlft.add_time;
1168         x->km.state = orig->km.state;
1169         x->km.seq = orig->km.seq;
1170
1171         return x;
1172
1173  error:
1174         if (errp)
1175                 *errp = err;
1176         if (x) {
1177                 kfree(x->aalg);
1178                 kfree(x->ealg);
1179                 kfree(x->calg);
1180                 kfree(x->encap);
1181                 kfree(x->coaddr);
1182         }
1183         kfree(x);
1184         return NULL;
1185 }
1186
1187 /* xfrm_state_lock is held */
1188 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1189 {
1190         unsigned int h;
1191         struct xfrm_state *x;
1192         struct hlist_node *entry;
1193
1194         if (m->reqid) {
1195                 h = xfrm_dst_hash(&m->old_daddr, &m->old_saddr,
1196                                   m->reqid, m->old_family);
1197                 hlist_for_each_entry(x, entry, xfrm_state_bydst+h, bydst) {
1198                         if (x->props.mode != m->mode ||
1199                             x->id.proto != m->proto)
1200                                 continue;
1201                         if (m->reqid && x->props.reqid != m->reqid)
1202                                 continue;
1203                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1204                                           m->old_family) ||
1205                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1206                                           m->old_family))
1207                                 continue;
1208                         xfrm_state_hold(x);
1209                         return x;
1210                 }
1211         } else {
1212                 h = xfrm_src_hash(&m->old_daddr, &m->old_saddr,
1213                                   m->old_family);
1214                 hlist_for_each_entry(x, entry, xfrm_state_bysrc+h, bysrc) {
1215                         if (x->props.mode != m->mode ||
1216                             x->id.proto != m->proto)
1217                                 continue;
1218                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1219                                           m->old_family) ||
1220                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1221                                           m->old_family))
1222                                 continue;
1223                         xfrm_state_hold(x);
1224                         return x;
1225                 }
1226         }
1227
1228         return NULL;
1229 }
1230 EXPORT_SYMBOL(xfrm_migrate_state_find);
1231
1232 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1233                                        struct xfrm_migrate *m)
1234 {
1235         struct xfrm_state *xc;
1236         int err;
1237
1238         xc = xfrm_state_clone(x, &err);
1239         if (!xc)
1240                 return NULL;
1241
1242         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1243         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1244
1245         /* add state */
1246         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1247                 /* a care is needed when the destination address of the
1248                    state is to be updated as it is a part of triplet */
1249                 xfrm_state_insert(xc);
1250         } else {
1251                 if ((err = xfrm_state_add(xc)) < 0)
1252                         goto error;
1253         }
1254
1255         return xc;
1256 error:
1257         kfree(xc);
1258         return NULL;
1259 }
1260 EXPORT_SYMBOL(xfrm_state_migrate);
1261 #endif
1262
1263 int xfrm_state_update(struct xfrm_state *x)
1264 {
1265         struct xfrm_state *x1;
1266         int err;
1267         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1268
1269         spin_lock_bh(&xfrm_state_lock);
1270         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1271
1272         err = -ESRCH;
1273         if (!x1)
1274                 goto out;
1275
1276         if (xfrm_state_kern(x1)) {
1277                 xfrm_state_put(x1);
1278                 err = -EEXIST;
1279                 goto out;
1280         }
1281
1282         if (x1->km.state == XFRM_STATE_ACQ) {
1283                 __xfrm_state_insert(x);
1284                 x = NULL;
1285         }
1286         err = 0;
1287
1288 out:
1289         spin_unlock_bh(&xfrm_state_lock);
1290
1291         if (err)
1292                 return err;
1293
1294         if (!x) {
1295                 xfrm_state_delete(x1);
1296                 xfrm_state_put(x1);
1297                 return 0;
1298         }
1299
1300         err = -EINVAL;
1301         spin_lock_bh(&x1->lock);
1302         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1303                 if (x->encap && x1->encap)
1304                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1305                 if (x->coaddr && x1->coaddr) {
1306                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1307                 }
1308                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1309                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1310                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1311                 x1->km.dying = 0;
1312
1313                 mod_timer(&x1->timer, jiffies + HZ);
1314                 if (x1->curlft.use_time)
1315                         xfrm_state_check_expire(x1);
1316
1317                 err = 0;
1318         }
1319         spin_unlock_bh(&x1->lock);
1320
1321         xfrm_state_put(x1);
1322
1323         return err;
1324 }
1325 EXPORT_SYMBOL(xfrm_state_update);
1326
1327 int xfrm_state_check_expire(struct xfrm_state *x)
1328 {
1329         if (!x->curlft.use_time)
1330                 x->curlft.use_time = get_seconds();
1331
1332         if (x->km.state != XFRM_STATE_VALID)
1333                 return -EINVAL;
1334
1335         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1336             x->curlft.packets >= x->lft.hard_packet_limit) {
1337                 x->km.state = XFRM_STATE_EXPIRED;
1338                 mod_timer(&x->timer, jiffies);
1339                 return -EINVAL;
1340         }
1341
1342         if (!x->km.dying &&
1343             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1344              x->curlft.packets >= x->lft.soft_packet_limit)) {
1345                 x->km.dying = 1;
1346                 km_state_expired(x, 0, 0);
1347         }
1348         return 0;
1349 }
1350 EXPORT_SYMBOL(xfrm_state_check_expire);
1351
1352 struct xfrm_state *
1353 xfrm_state_lookup(xfrm_address_t *daddr, __be32 spi, u8 proto,
1354                   unsigned short family)
1355 {
1356         struct xfrm_state *x;
1357
1358         spin_lock_bh(&xfrm_state_lock);
1359         x = __xfrm_state_lookup(daddr, spi, proto, family);
1360         spin_unlock_bh(&xfrm_state_lock);
1361         return x;
1362 }
1363 EXPORT_SYMBOL(xfrm_state_lookup);
1364
1365 struct xfrm_state *
1366 xfrm_state_lookup_byaddr(xfrm_address_t *daddr, xfrm_address_t *saddr,
1367                          u8 proto, unsigned short family)
1368 {
1369         struct xfrm_state *x;
1370
1371         spin_lock_bh(&xfrm_state_lock);
1372         x = __xfrm_state_lookup_byaddr(daddr, saddr, proto, family);
1373         spin_unlock_bh(&xfrm_state_lock);
1374         return x;
1375 }
1376 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1377
1378 struct xfrm_state *
1379 xfrm_find_acq(u8 mode, u32 reqid, u8 proto,
1380               xfrm_address_t *daddr, xfrm_address_t *saddr,
1381               int create, unsigned short family)
1382 {
1383         struct xfrm_state *x;
1384
1385         spin_lock_bh(&xfrm_state_lock);
1386         x = __find_acq_core(family, mode, reqid, proto, daddr, saddr, create);
1387         spin_unlock_bh(&xfrm_state_lock);
1388
1389         return x;
1390 }
1391 EXPORT_SYMBOL(xfrm_find_acq);
1392
1393 #ifdef CONFIG_XFRM_SUB_POLICY
1394 int
1395 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1396                unsigned short family)
1397 {
1398         int err = 0;
1399         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1400         if (!afinfo)
1401                 return -EAFNOSUPPORT;
1402
1403         spin_lock_bh(&xfrm_state_lock);
1404         if (afinfo->tmpl_sort)
1405                 err = afinfo->tmpl_sort(dst, src, n);
1406         spin_unlock_bh(&xfrm_state_lock);
1407         xfrm_state_put_afinfo(afinfo);
1408         return err;
1409 }
1410 EXPORT_SYMBOL(xfrm_tmpl_sort);
1411
1412 int
1413 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1414                 unsigned short family)
1415 {
1416         int err = 0;
1417         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1418         if (!afinfo)
1419                 return -EAFNOSUPPORT;
1420
1421         spin_lock_bh(&xfrm_state_lock);
1422         if (afinfo->state_sort)
1423                 err = afinfo->state_sort(dst, src, n);
1424         spin_unlock_bh(&xfrm_state_lock);
1425         xfrm_state_put_afinfo(afinfo);
1426         return err;
1427 }
1428 EXPORT_SYMBOL(xfrm_state_sort);
1429 #endif
1430
1431 /* Silly enough, but I'm lazy to build resolution list */
1432
1433 static struct xfrm_state *__xfrm_find_acq_byseq(u32 seq)
1434 {
1435         int i;
1436
1437         for (i = 0; i <= xfrm_state_hmask; i++) {
1438                 struct hlist_node *entry;
1439                 struct xfrm_state *x;
1440
1441                 hlist_for_each_entry(x, entry, xfrm_state_bydst+i, bydst) {
1442                         if (x->km.seq == seq &&
1443                             x->km.state == XFRM_STATE_ACQ) {
1444                                 xfrm_state_hold(x);
1445                                 return x;
1446                         }
1447                 }
1448         }
1449         return NULL;
1450 }
1451
1452 struct xfrm_state *xfrm_find_acq_byseq(u32 seq)
1453 {
1454         struct xfrm_state *x;
1455
1456         spin_lock_bh(&xfrm_state_lock);
1457         x = __xfrm_find_acq_byseq(seq);
1458         spin_unlock_bh(&xfrm_state_lock);
1459         return x;
1460 }
1461 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1462
1463 u32 xfrm_get_acqseq(void)
1464 {
1465         u32 res;
1466         static u32 acqseq;
1467         static DEFINE_SPINLOCK(acqseq_lock);
1468
1469         spin_lock_bh(&acqseq_lock);
1470         res = (++acqseq ? : ++acqseq);
1471         spin_unlock_bh(&acqseq_lock);
1472         return res;
1473 }
1474 EXPORT_SYMBOL(xfrm_get_acqseq);
1475
1476 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1477 {
1478         unsigned int h;
1479         struct xfrm_state *x0;
1480         int err = -ENOENT;
1481         __be32 minspi = htonl(low);
1482         __be32 maxspi = htonl(high);
1483
1484         spin_lock_bh(&x->lock);
1485         if (x->km.state == XFRM_STATE_DEAD)
1486                 goto unlock;
1487
1488         err = 0;
1489         if (x->id.spi)
1490                 goto unlock;
1491
1492         err = -ENOENT;
1493
1494         if (minspi == maxspi) {
1495                 x0 = xfrm_state_lookup(&x->id.daddr, minspi, x->id.proto, x->props.family);
1496                 if (x0) {
1497                         xfrm_state_put(x0);
1498                         goto unlock;
1499                 }
1500                 x->id.spi = minspi;
1501         } else {
1502                 u32 spi = 0;
1503                 for (h=0; h<high-low+1; h++) {
1504                         spi = low + net_random()%(high-low+1);
1505                         x0 = xfrm_state_lookup(&x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1506                         if (x0 == NULL) {
1507                                 x->id.spi = htonl(spi);
1508                                 break;
1509                         }
1510                         xfrm_state_put(x0);
1511                 }
1512         }
1513         if (x->id.spi) {
1514                 spin_lock_bh(&xfrm_state_lock);
1515                 h = xfrm_spi_hash(&x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1516                 hlist_add_head(&x->byspi, xfrm_state_byspi+h);
1517                 spin_unlock_bh(&xfrm_state_lock);
1518
1519                 err = 0;
1520         }
1521
1522 unlock:
1523         spin_unlock_bh(&x->lock);
1524
1525         return err;
1526 }
1527 EXPORT_SYMBOL(xfrm_alloc_spi);
1528
1529 int xfrm_state_walk(struct xfrm_state_walk *walk,
1530                     int (*func)(struct xfrm_state *, int, void*),
1531                     void *data)
1532 {
1533         struct xfrm_state *old, *x, *last = NULL;
1534         int err = 0;
1535
1536         if (walk->state == NULL && walk->count != 0)
1537                 return 0;
1538
1539         old = x = walk->state;
1540         walk->state = NULL;
1541         spin_lock_bh(&xfrm_state_lock);
1542         if (x == NULL)
1543                 x = list_first_entry(&xfrm_state_all, struct xfrm_state, all);
1544         list_for_each_entry_from(x, &xfrm_state_all, all) {
1545                 if (x->km.state == XFRM_STATE_DEAD)
1546                         continue;
1547                 if (!xfrm_id_proto_match(x->id.proto, walk->proto))
1548                         continue;
1549                 if (last) {
1550                         err = func(last, walk->count, data);
1551                         if (err) {
1552                                 xfrm_state_hold(last);
1553                                 walk->state = last;
1554                                 goto out;
1555                         }
1556                 }
1557                 last = x;
1558                 walk->count++;
1559         }
1560         if (walk->count == 0) {
1561                 err = -ENOENT;
1562                 goto out;
1563         }
1564         if (last)
1565                 err = func(last, 0, data);
1566 out:
1567         spin_unlock_bh(&xfrm_state_lock);
1568         if (old != NULL)
1569                 xfrm_state_put(old);
1570         return err;
1571 }
1572 EXPORT_SYMBOL(xfrm_state_walk);
1573
1574
1575 void xfrm_replay_notify(struct xfrm_state *x, int event)
1576 {
1577         struct km_event c;
1578         /* we send notify messages in case
1579          *  1. we updated on of the sequence numbers, and the seqno difference
1580          *     is at least x->replay_maxdiff, in this case we also update the
1581          *     timeout of our timer function
1582          *  2. if x->replay_maxage has elapsed since last update,
1583          *     and there were changes
1584          *
1585          *  The state structure must be locked!
1586          */
1587
1588         switch (event) {
1589         case XFRM_REPLAY_UPDATE:
1590                 if (x->replay_maxdiff &&
1591                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1592                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1593                         if (x->xflags & XFRM_TIME_DEFER)
1594                                 event = XFRM_REPLAY_TIMEOUT;
1595                         else
1596                                 return;
1597                 }
1598
1599                 break;
1600
1601         case XFRM_REPLAY_TIMEOUT:
1602                 if ((x->replay.seq == x->preplay.seq) &&
1603                     (x->replay.bitmap == x->preplay.bitmap) &&
1604                     (x->replay.oseq == x->preplay.oseq)) {
1605                         x->xflags |= XFRM_TIME_DEFER;
1606                         return;
1607                 }
1608
1609                 break;
1610         }
1611
1612         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1613         c.event = XFRM_MSG_NEWAE;
1614         c.data.aevent = event;
1615         km_state_notify(x, &c);
1616
1617         if (x->replay_maxage &&
1618             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1619                 x->xflags &= ~XFRM_TIME_DEFER;
1620 }
1621
1622 static void xfrm_replay_timer_handler(unsigned long data)
1623 {
1624         struct xfrm_state *x = (struct xfrm_state*)data;
1625
1626         spin_lock(&x->lock);
1627
1628         if (x->km.state == XFRM_STATE_VALID) {
1629                 if (xfrm_aevent_is_on())
1630                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1631                 else
1632                         x->xflags |= XFRM_TIME_DEFER;
1633         }
1634
1635         spin_unlock(&x->lock);
1636 }
1637
1638 int xfrm_replay_check(struct xfrm_state *x,
1639                       struct sk_buff *skb, __be32 net_seq)
1640 {
1641         u32 diff;
1642         u32 seq = ntohl(net_seq);
1643
1644         if (unlikely(seq == 0))
1645                 goto err;
1646
1647         if (likely(seq > x->replay.seq))
1648                 return 0;
1649
1650         diff = x->replay.seq - seq;
1651         if (diff >= min_t(unsigned int, x->props.replay_window,
1652                           sizeof(x->replay.bitmap) * 8)) {
1653                 x->stats.replay_window++;
1654                 goto err;
1655         }
1656
1657         if (x->replay.bitmap & (1U << diff)) {
1658                 x->stats.replay++;
1659                 goto err;
1660         }
1661         return 0;
1662
1663 err:
1664         xfrm_audit_state_replay(x, skb, net_seq);
1665         return -EINVAL;
1666 }
1667
1668 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1669 {
1670         u32 diff;
1671         u32 seq = ntohl(net_seq);
1672
1673         if (seq > x->replay.seq) {
1674                 diff = seq - x->replay.seq;
1675                 if (diff < x->props.replay_window)
1676                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1677                 else
1678                         x->replay.bitmap = 1;
1679                 x->replay.seq = seq;
1680         } else {
1681                 diff = x->replay.seq - seq;
1682                 x->replay.bitmap |= (1U << diff);
1683         }
1684
1685         if (xfrm_aevent_is_on())
1686                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1687 }
1688
1689 static LIST_HEAD(xfrm_km_list);
1690 static DEFINE_RWLOCK(xfrm_km_lock);
1691
1692 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1693 {
1694         struct xfrm_mgr *km;
1695
1696         read_lock(&xfrm_km_lock);
1697         list_for_each_entry(km, &xfrm_km_list, list)
1698                 if (km->notify_policy)
1699                         km->notify_policy(xp, dir, c);
1700         read_unlock(&xfrm_km_lock);
1701 }
1702
1703 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1704 {
1705         struct xfrm_mgr *km;
1706         read_lock(&xfrm_km_lock);
1707         list_for_each_entry(km, &xfrm_km_list, list)
1708                 if (km->notify)
1709                         km->notify(x, c);
1710         read_unlock(&xfrm_km_lock);
1711 }
1712
1713 EXPORT_SYMBOL(km_policy_notify);
1714 EXPORT_SYMBOL(km_state_notify);
1715
1716 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1717 {
1718         struct km_event c;
1719
1720         c.data.hard = hard;
1721         c.pid = pid;
1722         c.event = XFRM_MSG_EXPIRE;
1723         km_state_notify(x, &c);
1724
1725         if (hard)
1726                 wake_up(&km_waitq);
1727 }
1728
1729 EXPORT_SYMBOL(km_state_expired);
1730 /*
1731  * We send to all registered managers regardless of failure
1732  * We are happy with one success
1733 */
1734 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1735 {
1736         int err = -EINVAL, acqret;
1737         struct xfrm_mgr *km;
1738
1739         read_lock(&xfrm_km_lock);
1740         list_for_each_entry(km, &xfrm_km_list, list) {
1741                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1742                 if (!acqret)
1743                         err = acqret;
1744         }
1745         read_unlock(&xfrm_km_lock);
1746         return err;
1747 }
1748 EXPORT_SYMBOL(km_query);
1749
1750 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1751 {
1752         int err = -EINVAL;
1753         struct xfrm_mgr *km;
1754
1755         read_lock(&xfrm_km_lock);
1756         list_for_each_entry(km, &xfrm_km_list, list) {
1757                 if (km->new_mapping)
1758                         err = km->new_mapping(x, ipaddr, sport);
1759                 if (!err)
1760                         break;
1761         }
1762         read_unlock(&xfrm_km_lock);
1763         return err;
1764 }
1765 EXPORT_SYMBOL(km_new_mapping);
1766
1767 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1768 {
1769         struct km_event c;
1770
1771         c.data.hard = hard;
1772         c.pid = pid;
1773         c.event = XFRM_MSG_POLEXPIRE;
1774         km_policy_notify(pol, dir, &c);
1775
1776         if (hard)
1777                 wake_up(&km_waitq);
1778 }
1779 EXPORT_SYMBOL(km_policy_expired);
1780
1781 #ifdef CONFIG_XFRM_MIGRATE
1782 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1783                struct xfrm_migrate *m, int num_migrate)
1784 {
1785         int err = -EINVAL;
1786         int ret;
1787         struct xfrm_mgr *km;
1788
1789         read_lock(&xfrm_km_lock);
1790         list_for_each_entry(km, &xfrm_km_list, list) {
1791                 if (km->migrate) {
1792                         ret = km->migrate(sel, dir, type, m, num_migrate);
1793                         if (!ret)
1794                                 err = ret;
1795                 }
1796         }
1797         read_unlock(&xfrm_km_lock);
1798         return err;
1799 }
1800 EXPORT_SYMBOL(km_migrate);
1801 #endif
1802
1803 int km_report(u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1804 {
1805         int err = -EINVAL;
1806         int ret;
1807         struct xfrm_mgr *km;
1808
1809         read_lock(&xfrm_km_lock);
1810         list_for_each_entry(km, &xfrm_km_list, list) {
1811                 if (km->report) {
1812                         ret = km->report(proto, sel, addr);
1813                         if (!ret)
1814                                 err = ret;
1815                 }
1816         }
1817         read_unlock(&xfrm_km_lock);
1818         return err;
1819 }
1820 EXPORT_SYMBOL(km_report);
1821
1822 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1823 {
1824         int err;
1825         u8 *data;
1826         struct xfrm_mgr *km;
1827         struct xfrm_policy *pol = NULL;
1828
1829         if (optlen <= 0 || optlen > PAGE_SIZE)
1830                 return -EMSGSIZE;
1831
1832         data = kmalloc(optlen, GFP_KERNEL);
1833         if (!data)
1834                 return -ENOMEM;
1835
1836         err = -EFAULT;
1837         if (copy_from_user(data, optval, optlen))
1838                 goto out;
1839
1840         err = -EINVAL;
1841         read_lock(&xfrm_km_lock);
1842         list_for_each_entry(km, &xfrm_km_list, list) {
1843                 pol = km->compile_policy(sk, optname, data,
1844                                          optlen, &err);
1845                 if (err >= 0)
1846                         break;
1847         }
1848         read_unlock(&xfrm_km_lock);
1849
1850         if (err >= 0) {
1851                 xfrm_sk_policy_insert(sk, err, pol);
1852                 xfrm_pol_put(pol);
1853                 err = 0;
1854         }
1855
1856 out:
1857         kfree(data);
1858         return err;
1859 }
1860 EXPORT_SYMBOL(xfrm_user_policy);
1861
1862 int xfrm_register_km(struct xfrm_mgr *km)
1863 {
1864         write_lock_bh(&xfrm_km_lock);
1865         list_add_tail(&km->list, &xfrm_km_list);
1866         write_unlock_bh(&xfrm_km_lock);
1867         return 0;
1868 }
1869 EXPORT_SYMBOL(xfrm_register_km);
1870
1871 int xfrm_unregister_km(struct xfrm_mgr *km)
1872 {
1873         write_lock_bh(&xfrm_km_lock);
1874         list_del(&km->list);
1875         write_unlock_bh(&xfrm_km_lock);
1876         return 0;
1877 }
1878 EXPORT_SYMBOL(xfrm_unregister_km);
1879
1880 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1881 {
1882         int err = 0;
1883         if (unlikely(afinfo == NULL))
1884                 return -EINVAL;
1885         if (unlikely(afinfo->family >= NPROTO))
1886                 return -EAFNOSUPPORT;
1887         write_lock_bh(&xfrm_state_afinfo_lock);
1888         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1889                 err = -ENOBUFS;
1890         else
1891                 xfrm_state_afinfo[afinfo->family] = afinfo;
1892         write_unlock_bh(&xfrm_state_afinfo_lock);
1893         return err;
1894 }
1895 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1896
1897 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1898 {
1899         int err = 0;
1900         if (unlikely(afinfo == NULL))
1901                 return -EINVAL;
1902         if (unlikely(afinfo->family >= NPROTO))
1903                 return -EAFNOSUPPORT;
1904         write_lock_bh(&xfrm_state_afinfo_lock);
1905         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1906                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1907                         err = -EINVAL;
1908                 else
1909                         xfrm_state_afinfo[afinfo->family] = NULL;
1910         }
1911         write_unlock_bh(&xfrm_state_afinfo_lock);
1912         return err;
1913 }
1914 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1915
1916 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1917 {
1918         struct xfrm_state_afinfo *afinfo;
1919         if (unlikely(family >= NPROTO))
1920                 return NULL;
1921         read_lock(&xfrm_state_afinfo_lock);
1922         afinfo = xfrm_state_afinfo[family];
1923         if (unlikely(!afinfo))
1924                 read_unlock(&xfrm_state_afinfo_lock);
1925         return afinfo;
1926 }
1927
1928 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1929         __releases(xfrm_state_afinfo_lock)
1930 {
1931         read_unlock(&xfrm_state_afinfo_lock);
1932 }
1933
1934 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1935 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1936 {
1937         if (x->tunnel) {
1938                 struct xfrm_state *t = x->tunnel;
1939
1940                 if (atomic_read(&t->tunnel_users) == 2)
1941                         xfrm_state_delete(t);
1942                 atomic_dec(&t->tunnel_users);
1943                 xfrm_state_put(t);
1944                 x->tunnel = NULL;
1945         }
1946 }
1947 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1948
1949 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1950 {
1951         int res;
1952
1953         spin_lock_bh(&x->lock);
1954         if (x->km.state == XFRM_STATE_VALID &&
1955             x->type && x->type->get_mtu)
1956                 res = x->type->get_mtu(x, mtu);
1957         else
1958                 res = mtu - x->props.header_len;
1959         spin_unlock_bh(&x->lock);
1960         return res;
1961 }
1962
1963 int xfrm_init_state(struct xfrm_state *x)
1964 {
1965         struct xfrm_state_afinfo *afinfo;
1966         int family = x->props.family;
1967         int err;
1968
1969         err = -EAFNOSUPPORT;
1970         afinfo = xfrm_state_get_afinfo(family);
1971         if (!afinfo)
1972                 goto error;
1973
1974         err = 0;
1975         if (afinfo->init_flags)
1976                 err = afinfo->init_flags(x);
1977
1978         xfrm_state_put_afinfo(afinfo);
1979
1980         if (err)
1981                 goto error;
1982
1983         err = -EPROTONOSUPPORT;
1984         x->inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1985         if (x->inner_mode == NULL)
1986                 goto error;
1987
1988         if (!(x->inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1989             family != x->sel.family)
1990                 goto error;
1991
1992         x->type = xfrm_get_type(x->id.proto, family);
1993         if (x->type == NULL)
1994                 goto error;
1995
1996         err = x->type->init_state(x);
1997         if (err)
1998                 goto error;
1999
2000         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2001         if (x->outer_mode == NULL)
2002                 goto error;
2003
2004         x->km.state = XFRM_STATE_VALID;
2005
2006 error:
2007         return err;
2008 }
2009
2010 EXPORT_SYMBOL(xfrm_init_state);
2011
2012 void __init xfrm_state_init(void)
2013 {
2014         unsigned int sz;
2015
2016         sz = sizeof(struct hlist_head) * 8;
2017
2018         xfrm_state_bydst = xfrm_hash_alloc(sz);
2019         xfrm_state_bysrc = xfrm_hash_alloc(sz);
2020         xfrm_state_byspi = xfrm_hash_alloc(sz);
2021         if (!xfrm_state_bydst || !xfrm_state_bysrc || !xfrm_state_byspi)
2022                 panic("XFRM: Cannot allocate bydst/bysrc/byspi hashes.");
2023         xfrm_state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2024
2025         INIT_WORK(&xfrm_state_gc_work, xfrm_state_gc_task);
2026 }
2027
2028 #ifdef CONFIG_AUDITSYSCALL
2029 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2030                                      struct audit_buffer *audit_buf)
2031 {
2032         struct xfrm_sec_ctx *ctx = x->security;
2033         u32 spi = ntohl(x->id.spi);
2034
2035         if (ctx)
2036                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2037                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2038
2039         switch(x->props.family) {
2040         case AF_INET:
2041                 audit_log_format(audit_buf,
2042                                  " src=" NIPQUAD_FMT " dst=" NIPQUAD_FMT,
2043                                  NIPQUAD(x->props.saddr.a4),
2044                                  NIPQUAD(x->id.daddr.a4));
2045                 break;
2046         case AF_INET6:
2047                 audit_log_format(audit_buf,
2048                                  " src=" NIP6_FMT " dst=" NIP6_FMT,
2049                                  NIP6(*(struct in6_addr *)x->props.saddr.a6),
2050                                  NIP6(*(struct in6_addr *)x->id.daddr.a6));
2051                 break;
2052         }
2053
2054         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2055 }
2056
2057 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2058                                       struct audit_buffer *audit_buf)
2059 {
2060         struct iphdr *iph4;
2061         struct ipv6hdr *iph6;
2062
2063         switch (family) {
2064         case AF_INET:
2065                 iph4 = ip_hdr(skb);
2066                 audit_log_format(audit_buf,
2067                                  " src=" NIPQUAD_FMT " dst=" NIPQUAD_FMT,
2068                                  NIPQUAD(iph4->saddr),
2069                                  NIPQUAD(iph4->daddr));
2070                 break;
2071         case AF_INET6:
2072                 iph6 = ipv6_hdr(skb);
2073                 audit_log_format(audit_buf,
2074                                  " src=" NIP6_FMT " dst=" NIP6_FMT
2075                                  " flowlbl=0x%x%x%x",
2076                                  NIP6(iph6->saddr),
2077                                  NIP6(iph6->daddr),
2078                                  iph6->flow_lbl[0] & 0x0f,
2079                                  iph6->flow_lbl[1],
2080                                  iph6->flow_lbl[2]);
2081                 break;
2082         }
2083 }
2084
2085 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2086                           u32 auid, u32 secid)
2087 {
2088         struct audit_buffer *audit_buf;
2089
2090         audit_buf = xfrm_audit_start("SAD-add");
2091         if (audit_buf == NULL)
2092                 return;
2093         xfrm_audit_helper_usrinfo(auid, secid, audit_buf);
2094         xfrm_audit_helper_sainfo(x, audit_buf);
2095         audit_log_format(audit_buf, " res=%u", result);
2096         audit_log_end(audit_buf);
2097 }
2098 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2099
2100 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2101                              u32 auid, u32 secid)
2102 {
2103         struct audit_buffer *audit_buf;
2104
2105         audit_buf = xfrm_audit_start("SAD-delete");
2106         if (audit_buf == NULL)
2107                 return;
2108         xfrm_audit_helper_usrinfo(auid, secid, audit_buf);
2109         xfrm_audit_helper_sainfo(x, audit_buf);
2110         audit_log_format(audit_buf, " res=%u", result);
2111         audit_log_end(audit_buf);
2112 }
2113 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2114
2115 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2116                                       struct sk_buff *skb)
2117 {
2118         struct audit_buffer *audit_buf;
2119         u32 spi;
2120
2121         audit_buf = xfrm_audit_start("SA-replay-overflow");
2122         if (audit_buf == NULL)
2123                 return;
2124         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2125         /* don't record the sequence number because it's inherent in this kind
2126          * of audit message */
2127         spi = ntohl(x->id.spi);
2128         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2129         audit_log_end(audit_buf);
2130 }
2131 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2132
2133 static void xfrm_audit_state_replay(struct xfrm_state *x,
2134                              struct sk_buff *skb, __be32 net_seq)
2135 {
2136         struct audit_buffer *audit_buf;
2137         u32 spi;
2138
2139         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2140         if (audit_buf == NULL)
2141                 return;
2142         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2143         spi = ntohl(x->id.spi);
2144         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2145                          spi, spi, ntohl(net_seq));
2146         audit_log_end(audit_buf);
2147 }
2148
2149 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2150 {
2151         struct audit_buffer *audit_buf;
2152
2153         audit_buf = xfrm_audit_start("SA-notfound");
2154         if (audit_buf == NULL)
2155                 return;
2156         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2157         audit_log_end(audit_buf);
2158 }
2159 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2160
2161 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2162                                __be32 net_spi, __be32 net_seq)
2163 {
2164         struct audit_buffer *audit_buf;
2165         u32 spi;
2166
2167         audit_buf = xfrm_audit_start("SA-notfound");
2168         if (audit_buf == NULL)
2169                 return;
2170         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2171         spi = ntohl(net_spi);
2172         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2173                          spi, spi, ntohl(net_seq));
2174         audit_log_end(audit_buf);
2175 }
2176 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2177
2178 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2179                               struct sk_buff *skb, u8 proto)
2180 {
2181         struct audit_buffer *audit_buf;
2182         __be32 net_spi;
2183         __be32 net_seq;
2184
2185         audit_buf = xfrm_audit_start("SA-icv-failure");
2186         if (audit_buf == NULL)
2187                 return;
2188         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2189         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2190                 u32 spi = ntohl(net_spi);
2191                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2192                                  spi, spi, ntohl(net_seq));
2193         }
2194         audit_log_end(audit_buf);
2195 }
2196 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2197 #endif /* CONFIG_AUDITSYSCALL */