SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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/interrupt.h>
  26 #include <linux/kernel.h>
  27
  28 #include "xfrm_hash.h"
  29
  30 /* Each xfrm_state may be linked to two tables:
  31
  32    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  33    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  34       destination/tunnel endpoint. (output)
  35  */
  36
  37 static DEFINE_SPINLOCK(xfrm_state_lock);
  38
  39 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  40 static unsigned int xfrm_state_genid;
  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 #ifdef CONFIG_AUDITSYSCALL
  46 static void xfrm_audit_state_replay(struct xfrm_state *x,
  47                                     struct sk_buff *skb, __be32 net_seq);
  48 #else
  49 #define xfrm_audit_state_replay(x, s, sq)       do { ; } while (0)
  50 #endif /* CONFIG_AUDITSYSCALL */
  51
  52 static inline unsigned int xfrm_dst_hash(struct net *net,
  53                                          xfrm_address_t *daddr,
  54                                          xfrm_address_t *saddr,
  55                                          u32 reqid,
  56                                          unsigned short family)
  57 {
  58         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  59 }
  60
  61 static inline unsigned int xfrm_src_hash(struct net *net,
  62                                          xfrm_address_t *daddr,
  63                                          xfrm_address_t *saddr,
  64                                          unsigned short family)
  65 {
  66         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  67 }
  68
  69 static inline unsigned int
  70 xfrm_spi_hash(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
  71 {
  72         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  73 }
  74
  75 static void xfrm_hash_transfer(struct hlist_head *list,
  76                                struct hlist_head *ndsttable,
  77                                struct hlist_head *nsrctable,
  78                                struct hlist_head *nspitable,
  79                                unsigned int nhashmask)
  80 {
  81         struct hlist_node *entry, *tmp;
  82         struct xfrm_state *x;
  83
  84         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
  85                 unsigned int h;
  86
  87                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  88                                     x->props.reqid, x->props.family,
  89                                     nhashmask);
  90                 hlist_add_head(&x->bydst, ndsttable+h);
  91
  92                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  93                                     x->props.family,
  94                                     nhashmask);
  95                 hlist_add_head(&x->bysrc, nsrctable+h);
  96
  97                 if (x->id.spi) {
  98                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  99                                             x->id.proto, x->props.family,
 100                                             nhashmask);
 101                         hlist_add_head(&x->byspi, nspitable+h);
 102                 }
 103         }
 104 }
 105
 106 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 107 {
 108         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 109 }
 110
 111 static DEFINE_MUTEX(hash_resize_mutex);
 112
 113 static void xfrm_hash_resize(struct work_struct *work)
 114 {
 115         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 116         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 117         unsigned long nsize, osize;
 118         unsigned int nhashmask, ohashmask;
 119         int i;
 120
 121         mutex_lock(&hash_resize_mutex);
 122
 123         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 124         ndst = xfrm_hash_alloc(nsize);
 125         if (!ndst)
 126                 goto out_unlock;
 127         nsrc = xfrm_hash_alloc(nsize);
 128         if (!nsrc) {
 129                 xfrm_hash_free(ndst, nsize);
 130                 goto out_unlock;
 131         }
 132         nspi = xfrm_hash_alloc(nsize);
 133         if (!nspi) {
 134                 xfrm_hash_free(ndst, nsize);
 135                 xfrm_hash_free(nsrc, nsize);
 136                 goto out_unlock;
 137         }
 138
 139         spin_lock_bh(&xfrm_state_lock);
 140
 141         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 142         for (i = net->xfrm.state_hmask; i >= 0; i--)
 143                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 144                                    nhashmask);
 145
 146         odst = net->xfrm.state_bydst;
 147         osrc = net->xfrm.state_bysrc;
 148         ospi = net->xfrm.state_byspi;
 149         ohashmask = net->xfrm.state_hmask;
 150
 151         net->xfrm.state_bydst = ndst;
 152         net->xfrm.state_bysrc = nsrc;
 153         net->xfrm.state_byspi = nspi;
 154         net->xfrm.state_hmask = nhashmask;
 155
 156         spin_unlock_bh(&xfrm_state_lock);
 157
 158         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 159         xfrm_hash_free(odst, osize);
 160         xfrm_hash_free(osrc, osize);
 161         xfrm_hash_free(ospi, osize);
 162
 163 out_unlock:
 164         mutex_unlock(&hash_resize_mutex);
 165 }
 166
 167 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
 168 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
 169
 170 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 171
 172 int __xfrm_state_delete(struct xfrm_state *x);
 173
 174 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 175 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
 176
 177 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 178 {
 179         struct xfrm_state_afinfo *afinfo;
 180         if (unlikely(family >= NPROTO))
 181                 return NULL;
 182         write_lock_bh(&xfrm_state_afinfo_lock);
 183         afinfo = xfrm_state_afinfo[family];
 184         if (unlikely(!afinfo))
 185                 write_unlock_bh(&xfrm_state_afinfo_lock);
 186         return afinfo;
 187 }
 188
 189 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 190         __releases(xfrm_state_afinfo_lock)
 191 {
 192         write_unlock_bh(&xfrm_state_afinfo_lock);
 193 }
 194
 195 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 196 {
 197         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 198         const struct xfrm_type **typemap;
 199         int err = 0;
 200
 201         if (unlikely(afinfo == NULL))
 202                 return -EAFNOSUPPORT;
 203         typemap = afinfo->type_map;
 204
 205         if (likely(typemap[type->proto] == NULL))
 206                 typemap[type->proto] = type;
 207         else
 208                 err = -EEXIST;
 209         xfrm_state_unlock_afinfo(afinfo);
 210         return err;
 211 }
 212 EXPORT_SYMBOL(xfrm_register_type);
 213
 214 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 215 {
 216         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 217         const struct xfrm_type **typemap;
 218         int err = 0;
 219
 220         if (unlikely(afinfo == NULL))
 221                 return -EAFNOSUPPORT;
 222         typemap = afinfo->type_map;
 223
 224         if (unlikely(typemap[type->proto] != type))
 225                 err = -ENOENT;
 226         else
 227                 typemap[type->proto] = NULL;
 228         xfrm_state_unlock_afinfo(afinfo);
 229         return err;
 230 }
 231 EXPORT_SYMBOL(xfrm_unregister_type);
 232
 233 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 234 {
 235         struct xfrm_state_afinfo *afinfo;
 236         const struct xfrm_type **typemap;
 237         const struct xfrm_type *type;
 238         int modload_attempted = 0;
 239
 240 retry:
 241         afinfo = xfrm_state_get_afinfo(family);
 242         if (unlikely(afinfo == NULL))
 243                 return NULL;
 244         typemap = afinfo->type_map;
 245
 246         type = typemap[proto];
 247         if (unlikely(type && !try_module_get(type->owner)))
 248                 type = NULL;
 249         if (!type && !modload_attempted) {
 250                 xfrm_state_put_afinfo(afinfo);
 251                 request_module("xfrm-type-%d-%d", family, proto);
 252                 modload_attempted = 1;
 253                 goto retry;
 254         }
 255
 256         xfrm_state_put_afinfo(afinfo);
 257         return type;
 258 }
 259
 260 static void xfrm_put_type(const struct xfrm_type *type)
 261 {
 262         module_put(type->owner);
 263 }
 264
 265 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 266 {
 267         struct xfrm_state_afinfo *afinfo;
 268         struct xfrm_mode **modemap;
 269         int err;
 270
 271         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 272                 return -EINVAL;
 273
 274         afinfo = xfrm_state_lock_afinfo(family);
 275         if (unlikely(afinfo == NULL))
 276                 return -EAFNOSUPPORT;
 277
 278         err = -EEXIST;
 279         modemap = afinfo->mode_map;
 280         if (modemap[mode->encap])
 281                 goto out;
 282
 283         err = -ENOENT;
 284         if (!try_module_get(afinfo->owner))
 285                 goto out;
 286
 287         mode->afinfo = afinfo;
 288         modemap[mode->encap] = mode;
 289         err = 0;
 290
 291 out:
 292         xfrm_state_unlock_afinfo(afinfo);
 293         return err;
 294 }
 295 EXPORT_SYMBOL(xfrm_register_mode);
 296
 297 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 298 {
 299         struct xfrm_state_afinfo *afinfo;
 300         struct xfrm_mode **modemap;
 301         int err;
 302
 303         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 304                 return -EINVAL;
 305
 306         afinfo = xfrm_state_lock_afinfo(family);
 307         if (unlikely(afinfo == NULL))
 308                 return -EAFNOSUPPORT;
 309
 310         err = -ENOENT;
 311         modemap = afinfo->mode_map;
 312         if (likely(modemap[mode->encap] == mode)) {
 313                 modemap[mode->encap] = NULL;
 314                 module_put(mode->afinfo->owner);
 315                 err = 0;
 316         }
 317
 318         xfrm_state_unlock_afinfo(afinfo);
 319         return err;
 320 }
 321 EXPORT_SYMBOL(xfrm_unregister_mode);
 322
 323 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 324 {
 325         struct xfrm_state_afinfo *afinfo;
 326         struct xfrm_mode *mode;
 327         int modload_attempted = 0;
 328
 329         if (unlikely(encap >= XFRM_MODE_MAX))
 330                 return NULL;
 331
 332 retry:
 333         afinfo = xfrm_state_get_afinfo(family);
 334         if (unlikely(afinfo == NULL))
 335                 return NULL;
 336
 337         mode = afinfo->mode_map[encap];
 338         if (unlikely(mode && !try_module_get(mode->owner)))
 339                 mode = NULL;
 340         if (!mode && !modload_attempted) {
 341                 xfrm_state_put_afinfo(afinfo);
 342                 request_module("xfrm-mode-%d-%d", family, encap);
 343                 modload_attempted = 1;
 344                 goto retry;
 345         }
 346
 347         xfrm_state_put_afinfo(afinfo);
 348         return mode;
 349 }
 350
 351 static void xfrm_put_mode(struct xfrm_mode *mode)
 352 {
 353         module_put(mode->owner);
 354 }
 355
 356 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 357 {
 358         tasklet_hrtimer_cancel(&x->mtimer);
 359         del_timer_sync(&x->rtimer);
 360         kfree(x->aalg);
 361         kfree(x->ealg);
 362         kfree(x->calg);
 363         kfree(x->encap);
 364         kfree(x->coaddr);
 365         if (x->inner_mode)
 366                 xfrm_put_mode(x->inner_mode);
 367         if (x->inner_mode_iaf)
 368                 xfrm_put_mode(x->inner_mode_iaf);
 369         if (x->outer_mode)
 370                 xfrm_put_mode(x->outer_mode);
 371         if (x->type) {
 372                 x->type->destructor(x);
 373                 xfrm_put_type(x->type);
 374         }
 375         security_xfrm_state_free(x);
 376         kfree(x);
 377 }
 378
 379 static void xfrm_state_gc_task(struct work_struct *work)
 380 {
 381         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 382         struct xfrm_state *x;
 383         struct hlist_node *entry, *tmp;
 384         struct hlist_head gc_list;
 385
 386         spin_lock_bh(&xfrm_state_gc_lock);
 387         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 388         spin_unlock_bh(&xfrm_state_gc_lock);
 389
 390         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 391                 xfrm_state_gc_destroy(x);
 392
 393         wake_up(&net->xfrm.km_waitq);
 394 }
 395
 396 static inline unsigned long make_jiffies(long secs)
 397 {
 398         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 399                 return MAX_SCHEDULE_TIMEOUT-1;
 400         else
 401                 return secs*HZ;
 402 }
 403
 404 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 405 {
 406         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 407         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 408         struct net *net = xs_net(x);
 409         unsigned long now = get_seconds();
 410         long next = LONG_MAX;
 411         int warn = 0;
 412         int err = 0;
 413
 414         spin_lock(&x->lock);
 415         if (x->km.state == XFRM_STATE_DEAD)
 416                 goto out;
 417         if (x->km.state == XFRM_STATE_EXPIRED)
 418                 goto expired;
 419         if (x->lft.hard_add_expires_seconds) {
 420                 long tmo = x->lft.hard_add_expires_seconds +
 421                         x->curlft.add_time - now;
 422                 if (tmo <= 0)
 423                         goto expired;
 424                 if (tmo < next)
 425                         next = tmo;
 426         }
 427         if (x->lft.hard_use_expires_seconds) {
 428                 long tmo = x->lft.hard_use_expires_seconds +
 429                         (x->curlft.use_time ? : now) - now;
 430                 if (tmo <= 0)
 431                         goto expired;
 432                 if (tmo < next)
 433                         next = tmo;
 434         }
 435         if (x->km.dying)
 436                 goto resched;
 437         if (x->lft.soft_add_expires_seconds) {
 438                 long tmo = x->lft.soft_add_expires_seconds +
 439                         x->curlft.add_time - now;
 440                 if (tmo <= 0)
 441                         warn = 1;
 442                 else if (tmo < next)
 443                         next = tmo;
 444         }
 445         if (x->lft.soft_use_expires_seconds) {
 446                 long tmo = x->lft.soft_use_expires_seconds +
 447                         (x->curlft.use_time ? : now) - now;
 448                 if (tmo <= 0)
 449                         warn = 1;
 450                 else if (tmo < next)
 451                         next = tmo;
 452         }
 453
 454         x->km.dying = warn;
 455         if (warn)
 456                 km_state_expired(x, 0, 0);
 457 resched:
 458         if (next != LONG_MAX){
 459                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 460         }
 461
 462         goto out;
 463
 464 expired:
 465         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 466                 x->km.state = XFRM_STATE_EXPIRED;
 467                 wake_up(&net->xfrm.km_waitq);
 468                 next = 2;
 469                 goto resched;
 470         }
 471
 472         err = __xfrm_state_delete(x);
 473         if (!err && x->id.spi)
 474                 km_state_expired(x, 1, 0);
 475
 476         xfrm_audit_state_delete(x, err ? 0 : 1,
 477                                 audit_get_loginuid(current),
 478                                 audit_get_sessionid(current), 0);
 479
 480 out:
 481         spin_unlock(&x->lock);
 482         return HRTIMER_NORESTART;
 483 }
 484
 485 static void xfrm_replay_timer_handler(unsigned long data);
 486
 487 struct xfrm_state *xfrm_state_alloc(struct net *net)
 488 {
 489         struct xfrm_state *x;
 490
 491         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 492
 493         if (x) {
 494                 write_pnet(&x->xs_net, net);
 495                 atomic_set(&x->refcnt, 1);
 496                 atomic_set(&x->tunnel_users, 0);
 497                 INIT_LIST_HEAD(&x->km.all);
 498                 INIT_HLIST_NODE(&x->bydst);
 499                 INIT_HLIST_NODE(&x->bysrc);
 500                 INIT_HLIST_NODE(&x->byspi);
 501                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 502                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 503                                 (unsigned long)x);
 504                 x->curlft.add_time = get_seconds();
 505                 x->lft.soft_byte_limit = XFRM_INF;
 506                 x->lft.soft_packet_limit = XFRM_INF;
 507                 x->lft.hard_byte_limit = XFRM_INF;
 508                 x->lft.hard_packet_limit = XFRM_INF;
 509                 x->replay_maxage = 0;
 510                 x->replay_maxdiff = 0;
 511                 x->inner_mode = NULL;
 512                 x->inner_mode_iaf = NULL;
 513                 spin_lock_init(&x->lock);
 514         }
 515         return x;
 516 }
 517 EXPORT_SYMBOL(xfrm_state_alloc);
 518
 519 void __xfrm_state_destroy(struct xfrm_state *x)
 520 {
 521         struct net *net = xs_net(x);
 522
 523         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 524
 525         spin_lock_bh(&xfrm_state_gc_lock);
 526         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 527         spin_unlock_bh(&xfrm_state_gc_lock);
 528         schedule_work(&net->xfrm.state_gc_work);
 529 }
 530 EXPORT_SYMBOL(__xfrm_state_destroy);
 531
 532 int __xfrm_state_delete(struct xfrm_state *x)
 533 {
 534         struct net *net = xs_net(x);
 535         int err = -ESRCH;
 536
 537         if (x->km.state != XFRM_STATE_DEAD) {
 538                 x->km.state = XFRM_STATE_DEAD;
 539                 spin_lock(&xfrm_state_lock);
 540                 list_del(&x->km.all);
 541                 hlist_del(&x->bydst);
 542                 hlist_del(&x->bysrc);
 543                 if (x->id.spi)
 544                         hlist_del(&x->byspi);
 545                 net->xfrm.state_num--;
 546                 spin_unlock(&xfrm_state_lock);
 547
 548                 /* All xfrm_state objects are created by xfrm_state_alloc.
 549                  * The xfrm_state_alloc call gives a reference, and that
 550                  * is what we are dropping here.
 551                  */
 552                 xfrm_state_put(x);
 553                 err = 0;
 554         }
 555
 556         return err;
 557 }
 558 EXPORT_SYMBOL(__xfrm_state_delete);
 559
 560 int xfrm_state_delete(struct xfrm_state *x)
 561 {
 562         int err;
 563
 564         spin_lock_bh(&x->lock);
 565         err = __xfrm_state_delete(x);
 566         spin_unlock_bh(&x->lock);
 567
 568         return err;
 569 }
 570 EXPORT_SYMBOL(xfrm_state_delete);
 571
 572 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 573 static inline int
 574 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 575 {
 576         int i, err = 0;
 577
 578         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 579                 struct hlist_node *entry;
 580                 struct xfrm_state *x;
 581
 582                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 583                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 584                            (err = security_xfrm_state_delete(x)) != 0) {
 585                                 xfrm_audit_state_delete(x, 0,
 586                                                         audit_info->loginuid,
 587                                                         audit_info->sessionid,
 588                                                         audit_info->secid);
 589                                 return err;
 590                         }
 591                 }
 592         }
 593
 594         return err;
 595 }
 596 #else
 597 static inline int
 598 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 599 {
 600         return 0;
 601 }
 602 #endif
 603
 604 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 605 {
 606         int i, err = 0, cnt = 0;
 607
 608         spin_lock_bh(&xfrm_state_lock);
 609         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 610         if (err)
 611                 goto out;
 612
 613         err = -ESRCH;
 614         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 615                 struct hlist_node *entry;
 616                 struct xfrm_state *x;
 617 restart:
 618                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 619                         if (!xfrm_state_kern(x) &&
 620                             xfrm_id_proto_match(x->id.proto, proto)) {
 621                                 xfrm_state_hold(x);
 622                                 spin_unlock_bh(&xfrm_state_lock);
 623
 624                                 err = xfrm_state_delete(x);
 625                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 626                                                         audit_info->loginuid,
 627                                                         audit_info->sessionid,
 628                                                         audit_info->secid);
 629                                 xfrm_state_put(x);
 630                                 if (!err)
 631                                         cnt++;
 632
 633                                 spin_lock_bh(&xfrm_state_lock);
 634                                 goto restart;
 635                         }
 636                 }
 637         }
 638         if (cnt)
 639                 err = 0;
 640
 641 out:
 642         spin_unlock_bh(&xfrm_state_lock);
 643         wake_up(&net->xfrm.km_waitq);
 644         return err;
 645 }
 646 EXPORT_SYMBOL(xfrm_state_flush);
 647
 648 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 649 {
 650         spin_lock_bh(&xfrm_state_lock);
 651         si->sadcnt = net->xfrm.state_num;
 652         si->sadhcnt = net->xfrm.state_hmask;
 653         si->sadhmcnt = xfrm_state_hashmax;
 654         spin_unlock_bh(&xfrm_state_lock);
 655 }
 656 EXPORT_SYMBOL(xfrm_sad_getinfo);
 657
 658 static int
 659 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
 660                   struct xfrm_tmpl *tmpl,
 661                   xfrm_address_t *daddr, xfrm_address_t *saddr,
 662                   unsigned short family)
 663 {
 664         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 665         if (!afinfo)
 666                 return -1;
 667         afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
 668         xfrm_state_put_afinfo(afinfo);
 669         return 0;
 670 }
 671
 672 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
 673 {
 674         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 675         struct xfrm_state *x;
 676         struct hlist_node *entry;
 677
 678         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 679                 if (x->props.family != family ||
 680                     x->id.spi       != spi ||
 681                     x->id.proto     != proto ||
 682                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 683                         continue;
 684
 685                 if ((mark & x->mark.m) != x->mark.v)
 686                         continue;
 687                 xfrm_state_hold(x);
 688                 return x;
 689         }
 690
 691         return NULL;
 692 }
 693
 694 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
 695 {
 696         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 697         struct xfrm_state *x;
 698         struct hlist_node *entry;
 699
 700         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 701                 if (x->props.family != family ||
 702                     x->id.proto     != proto ||
 703                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 704                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 705                         continue;
 706
 707                 if ((mark & x->mark.m) != x->mark.v)
 708                         continue;
 709                 xfrm_state_hold(x);
 710                 return x;
 711         }
 712
 713         return NULL;
 714 }
 715
 716 static inline struct xfrm_state *
 717 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 718 {
 719         struct net *net = xs_net(x);
 720         u32 mark = x->mark.v & x->mark.m;
 721
 722         if (use_spi)
 723                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 724                                            x->id.spi, x->id.proto, family);
 725         else
 726                 return __xfrm_state_lookup_byaddr(net, mark,
 727                                                   &x->id.daddr,
 728                                                   &x->props.saddr,
 729                                                   x->id.proto, family);
 730 }
 731
 732 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 733 {
 734         if (have_hash_collision &&
 735             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 736             net->xfrm.state_num > net->xfrm.state_hmask)
 737                 schedule_work(&net->xfrm.state_hash_work);
 738 }
 739
 740 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 741                                struct flowi *fl, unsigned short family,
 742                                xfrm_address_t *daddr, xfrm_address_t *saddr,
 743                                struct xfrm_state **best, int *acq_in_progress,
 744                                int *error)
 745 {
 746         /* Resolution logic:
 747          * 1. There is a valid state with matching selector. Done.
 748          * 2. Valid state with inappropriate selector. Skip.
 749          *
 750          * Entering area of "sysdeps".
 751          *
 752          * 3. If state is not valid, selector is temporary, it selects
 753          *    only session which triggered previous resolution. Key
 754          *    manager will do something to install a state with proper
 755          *    selector.
 756          */
 757         if (x->km.state == XFRM_STATE_VALID) {
 758                 if ((x->sel.family &&
 759                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 760                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 761                         return;
 762
 763                 if (!*best ||
 764                     (*best)->km.dying > x->km.dying ||
 765                     ((*best)->km.dying == x->km.dying &&
 766                      (*best)->curlft.add_time < x->curlft.add_time))
 767                         *best = x;
 768         } else if (x->km.state == XFRM_STATE_ACQ) {
 769                 *acq_in_progress = 1;
 770         } else if (x->km.state == XFRM_STATE_ERROR ||
 771                    x->km.state == XFRM_STATE_EXPIRED) {
 772                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 773                     security_xfrm_state_pol_flow_match(x, pol, fl))
 774                         *error = -ESRCH;
 775         }
 776 }
 777
 778 struct xfrm_state *
 779 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
 780                 struct flowi *fl, struct xfrm_tmpl *tmpl,
 781                 struct xfrm_policy *pol, int *err,
 782                 unsigned short family)
 783 {
 784         static xfrm_address_t saddr_wildcard = { };
 785         struct net *net = xp_net(pol);
 786         unsigned int h, h_wildcard;
 787         struct hlist_node *entry;
 788         struct xfrm_state *x, *x0, *to_put;
 789         int acquire_in_progress = 0;
 790         int error = 0;
 791         struct xfrm_state *best = NULL;
 792         u32 mark = pol->mark.v & pol->mark.m;
 793
 794         to_put = NULL;
 795
 796         spin_lock_bh(&xfrm_state_lock);
 797         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
 798         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 799                 if (x->props.family == family &&
 800                     x->props.reqid == tmpl->reqid &&
 801                     (mark & x->mark.m) == x->mark.v &&
 802                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 803                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 804                     tmpl->mode == x->props.mode &&
 805                     tmpl->id.proto == x->id.proto &&
 806                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 807                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 808                                            &best, &acquire_in_progress, &error);
 809         }
 810         if (best)
 811                 goto found;
 812
 813         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
 814         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 815                 if (x->props.family == family &&
 816                     x->props.reqid == tmpl->reqid &&
 817                     (mark & x->mark.m) == x->mark.v &&
 818                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 819                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 820                     tmpl->mode == x->props.mode &&
 821                     tmpl->id.proto == x->id.proto &&
 822                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 823                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 824                                            &best, &acquire_in_progress, &error);
 825         }
 826
 827 found:
 828         x = best;
 829         if (!x && !error && !acquire_in_progress) {
 830                 if (tmpl->id.spi &&
 831                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 832                                               tmpl->id.proto, family)) != NULL) {
 833                         to_put = x0;
 834                         error = -EEXIST;
 835                         goto out;
 836                 }
 837                 x = xfrm_state_alloc(net);
 838                 if (x == NULL) {
 839                         error = -ENOMEM;
 840                         goto out;
 841                 }
 842                 /* Initialize temporary selector matching only
 843                  * to current session. */
 844                 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
 845                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 846
 847                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
 848                 if (error) {
 849                         x->km.state = XFRM_STATE_DEAD;
 850                         to_put = x;
 851                         x = NULL;
 852                         goto out;
 853                 }
 854
 855                 if (km_query(x, tmpl, pol) == 0) {
 856                         x->km.state = XFRM_STATE_ACQ;
 857                         list_add(&x->km.all, &net->xfrm.state_all);
 858                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 859                         h = xfrm_src_hash(net, daddr, saddr, family);
 860                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 861                         if (x->id.spi) {
 862                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
 863                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 864                         }
 865                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 866                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 867                         net->xfrm.state_num++;
 868                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 869                 } else {
 870                         x->km.state = XFRM_STATE_DEAD;
 871                         to_put = x;
 872                         x = NULL;
 873                         error = -ESRCH;
 874                 }
 875         }
 876 out:
 877         if (x)
 878                 xfrm_state_hold(x);
 879         else
 880                 *err = acquire_in_progress ? -EAGAIN : error;
 881         spin_unlock_bh(&xfrm_state_lock);
 882         if (to_put)
 883                 xfrm_state_put(to_put);
 884         return x;
 885 }
 886
 887 struct xfrm_state *
 888 xfrm_stateonly_find(struct net *net, u32 mark,
 889                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 890                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 891 {
 892         unsigned int h;
 893         struct xfrm_state *rx = NULL, *x = NULL;
 894         struct hlist_node *entry;
 895
 896         spin_lock(&xfrm_state_lock);
 897         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 898         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 899                 if (x->props.family == family &&
 900                     x->props.reqid == reqid &&
 901                     (mark & x->mark.m) == x->mark.v &&
 902                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 903                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 904                     mode == x->props.mode &&
 905                     proto == x->id.proto &&
 906                     x->km.state == XFRM_STATE_VALID) {
 907                         rx = x;
 908                         break;
 909                 }
 910         }
 911
 912         if (rx)
 913                 xfrm_state_hold(rx);
 914         spin_unlock(&xfrm_state_lock);
 915
 916
 917         return rx;
 918 }
 919 EXPORT_SYMBOL(xfrm_stateonly_find);
 920
 921 static void __xfrm_state_insert(struct xfrm_state *x)
 922 {
 923         struct net *net = xs_net(x);
 924         unsigned int h;
 925
 926         x->genid = ++xfrm_state_genid;
 927
 928         list_add(&x->km.all, &net->xfrm.state_all);
 929
 930         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 931                           x->props.reqid, x->props.family);
 932         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 933
 934         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 935         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 936
 937         if (x->id.spi) {
 938                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 939                                   x->props.family);
 940
 941                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 942         }
 943
 944         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 945         if (x->replay_maxage)
 946                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 947
 948         wake_up(&net->xfrm.km_waitq);
 949
 950         net->xfrm.state_num++;
 951
 952         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 953 }
 954
 955 /* xfrm_state_lock is held */
 956 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 957 {
 958         struct net *net = xs_net(xnew);
 959         unsigned short family = xnew->props.family;
 960         u32 reqid = xnew->props.reqid;
 961         struct xfrm_state *x;
 962         struct hlist_node *entry;
 963         unsigned int h;
 964         u32 mark = xnew->mark.v & xnew->mark.m;
 965
 966         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 967         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 968                 if (x->props.family     == family &&
 969                     x->props.reqid      == reqid &&
 970                     (mark & x->mark.m) == x->mark.v &&
 971                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 972                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 973                         x->genid = xfrm_state_genid;
 974         }
 975 }
 976
 977 void xfrm_state_insert(struct xfrm_state *x)
 978 {
 979         spin_lock_bh(&xfrm_state_lock);
 980         __xfrm_state_bump_genids(x);
 981         __xfrm_state_insert(x);
 982         spin_unlock_bh(&xfrm_state_lock);
 983 }
 984 EXPORT_SYMBOL(xfrm_state_insert);
 985
 986 /* xfrm_state_lock is held */
 987 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
 988 {
 989         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 990         struct hlist_node *entry;
 991         struct xfrm_state *x;
 992         u32 mark = m->v & m->m;
 993
 994         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 995                 if (x->props.reqid  != reqid ||
 996                     x->props.mode   != mode ||
 997                     x->props.family != family ||
 998                     x->km.state     != XFRM_STATE_ACQ ||
 999                     x->id.spi       != 0 ||
1000                     x->id.proto     != proto ||
1001                     (mark & x->mark.m) != x->mark.v ||
1002                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1003                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1004                         continue;
1005
1006                 xfrm_state_hold(x);
1007                 return x;
1008         }
1009
1010         if (!create)
1011                 return NULL;
1012
1013         x = xfrm_state_alloc(net);
1014         if (likely(x)) {
1015                 switch (family) {
1016                 case AF_INET:
1017                         x->sel.daddr.a4 = daddr->a4;
1018                         x->sel.saddr.a4 = saddr->a4;
1019                         x->sel.prefixlen_d = 32;
1020                         x->sel.prefixlen_s = 32;
1021                         x->props.saddr.a4 = saddr->a4;
1022                         x->id.daddr.a4 = daddr->a4;
1023                         break;
1024
1025                 case AF_INET6:
1026                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1027                                        (struct in6_addr *)daddr);
1028                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1029                                        (struct in6_addr *)saddr);
1030                         x->sel.prefixlen_d = 128;
1031                         x->sel.prefixlen_s = 128;
1032                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1033                                        (struct in6_addr *)saddr);
1034                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1035                                        (struct in6_addr *)daddr);
1036                         break;
1037                 }
1038
1039                 x->km.state = XFRM_STATE_ACQ;
1040                 x->id.proto = proto;
1041                 x->props.family = family;
1042                 x->props.mode = mode;
1043                 x->props.reqid = reqid;
1044                 x->mark.v = m->v;
1045                 x->mark.m = m->m;
1046                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1047                 xfrm_state_hold(x);
1048                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1049                 list_add(&x->km.all, &net->xfrm.state_all);
1050                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1051                 h = xfrm_src_hash(net, daddr, saddr, family);
1052                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1053
1054                 net->xfrm.state_num++;
1055
1056                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1057         }
1058
1059         return x;
1060 }
1061
1062 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1063
1064 int xfrm_state_add(struct xfrm_state *x)
1065 {
1066         struct net *net = xs_net(x);
1067         struct xfrm_state *x1, *to_put;
1068         int family;
1069         int err;
1070         u32 mark = x->mark.v & x->mark.m;
1071         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1072
1073         family = x->props.family;
1074
1075         to_put = NULL;
1076
1077         spin_lock_bh(&xfrm_state_lock);
1078
1079         x1 = __xfrm_state_locate(x, use_spi, family);
1080         if (x1) {
1081                 to_put = x1;
1082                 x1 = NULL;
1083                 err = -EEXIST;
1084                 goto out;
1085         }
1086
1087         if (use_spi && x->km.seq) {
1088                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1089                 if (x1 && ((x1->id.proto != x->id.proto) ||
1090                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1091                         to_put = x1;
1092                         x1 = NULL;
1093                 }
1094         }
1095
1096         if (use_spi && !x1)
1097                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1098                                      x->props.reqid, x->id.proto,
1099                                      &x->id.daddr, &x->props.saddr, 0);
1100
1101         __xfrm_state_bump_genids(x);
1102         __xfrm_state_insert(x);
1103         err = 0;
1104
1105 out:
1106         spin_unlock_bh(&xfrm_state_lock);
1107
1108         if (x1) {
1109                 xfrm_state_delete(x1);
1110                 xfrm_state_put(x1);
1111         }
1112
1113         if (to_put)
1114                 xfrm_state_put(to_put);
1115
1116         return err;
1117 }
1118 EXPORT_SYMBOL(xfrm_state_add);
1119
1120 #ifdef CONFIG_XFRM_MIGRATE
1121 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1122 {
1123         struct net *net = xs_net(orig);
1124         int err = -ENOMEM;
1125         struct xfrm_state *x = xfrm_state_alloc(net);
1126         if (!x)
1127                 goto out;
1128
1129         memcpy(&x->id, &orig->id, sizeof(x->id));
1130         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1131         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1132         x->props.mode = orig->props.mode;
1133         x->props.replay_window = orig->props.replay_window;
1134         x->props.reqid = orig->props.reqid;
1135         x->props.family = orig->props.family;
1136         x->props.saddr = orig->props.saddr;
1137
1138         if (orig->aalg) {
1139                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1140                 if (!x->aalg)
1141                         goto error;
1142         }
1143         x->props.aalgo = orig->props.aalgo;
1144
1145         if (orig->ealg) {
1146                 x->ealg = xfrm_algo_clone(orig->ealg);
1147                 if (!x->ealg)
1148                         goto error;
1149         }
1150         x->props.ealgo = orig->props.ealgo;
1151
1152         if (orig->calg) {
1153                 x->calg = xfrm_algo_clone(orig->calg);
1154                 if (!x->calg)
1155                         goto error;
1156         }
1157         x->props.calgo = orig->props.calgo;
1158
1159         if (orig->encap) {
1160                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1161                 if (!x->encap)
1162                         goto error;
1163         }
1164
1165         if (orig->coaddr) {
1166                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1167                                     GFP_KERNEL);
1168                 if (!x->coaddr)
1169                         goto error;
1170         }
1171
1172         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1173
1174         err = xfrm_init_state(x);
1175         if (err)
1176                 goto error;
1177
1178         x->props.flags = orig->props.flags;
1179
1180         x->curlft.add_time = orig->curlft.add_time;
1181         x->km.state = orig->km.state;
1182         x->km.seq = orig->km.seq;
1183
1184         return x;
1185
1186  error:
1187         xfrm_state_put(x);
1188 out:
1189         if (errp)
1190                 *errp = err;
1191         return NULL;
1192 }
1193
1194 /* xfrm_state_lock is held */
1195 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1196 {
1197         unsigned int h;
1198         struct xfrm_state *x;
1199         struct hlist_node *entry;
1200
1201         if (m->reqid) {
1202                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1203                                   m->reqid, m->old_family);
1204                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1205                         if (x->props.mode != m->mode ||
1206                             x->id.proto != m->proto)
1207                                 continue;
1208                         if (m->reqid && x->props.reqid != m->reqid)
1209                                 continue;
1210                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1211                                           m->old_family) ||
1212                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1213                                           m->old_family))
1214                                 continue;
1215                         xfrm_state_hold(x);
1216                         return x;
1217                 }
1218         } else {
1219                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1220                                   m->old_family);
1221                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1222                         if (x->props.mode != m->mode ||
1223                             x->id.proto != m->proto)
1224                                 continue;
1225                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1226                                           m->old_family) ||
1227                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1228                                           m->old_family))
1229                                 continue;
1230                         xfrm_state_hold(x);
1231                         return x;
1232                 }
1233         }
1234
1235         return NULL;
1236 }
1237 EXPORT_SYMBOL(xfrm_migrate_state_find);
1238
1239 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1240                                        struct xfrm_migrate *m)
1241 {
1242         struct xfrm_state *xc;
1243         int err;
1244
1245         xc = xfrm_state_clone(x, &err);
1246         if (!xc)
1247                 return NULL;
1248
1249         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1250         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1251
1252         /* add state */
1253         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1254                 /* a care is needed when the destination address of the
1255                    state is to be updated as it is a part of triplet */
1256                 xfrm_state_insert(xc);
1257         } else {
1258                 if ((err = xfrm_state_add(xc)) < 0)
1259                         goto error;
1260         }
1261
1262         return xc;
1263 error:
1264         kfree(xc);
1265         return NULL;
1266 }
1267 EXPORT_SYMBOL(xfrm_state_migrate);
1268 #endif
1269
1270 int xfrm_state_update(struct xfrm_state *x)
1271 {
1272         struct xfrm_state *x1, *to_put;
1273         int err;
1274         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1275
1276         to_put = NULL;
1277
1278         spin_lock_bh(&xfrm_state_lock);
1279         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1280
1281         err = -ESRCH;
1282         if (!x1)
1283                 goto out;
1284
1285         if (xfrm_state_kern(x1)) {
1286                 to_put = x1;
1287                 err = -EEXIST;
1288                 goto out;
1289         }
1290
1291         if (x1->km.state == XFRM_STATE_ACQ) {
1292                 __xfrm_state_insert(x);
1293                 x = NULL;
1294         }
1295         err = 0;
1296
1297 out:
1298         spin_unlock_bh(&xfrm_state_lock);
1299
1300         if (to_put)
1301                 xfrm_state_put(to_put);
1302
1303         if (err)
1304                 return err;
1305
1306         if (!x) {
1307                 xfrm_state_delete(x1);
1308                 xfrm_state_put(x1);
1309                 return 0;
1310         }
1311
1312         err = -EINVAL;
1313         spin_lock_bh(&x1->lock);
1314         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1315                 if (x->encap && x1->encap)
1316                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1317                 if (x->coaddr && x1->coaddr) {
1318                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1319                 }
1320                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1321                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1322                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1323                 x1->km.dying = 0;
1324
1325                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1326                 if (x1->curlft.use_time)
1327                         xfrm_state_check_expire(x1);
1328
1329                 err = 0;
1330         }
1331         spin_unlock_bh(&x1->lock);
1332
1333         xfrm_state_put(x1);
1334
1335         return err;
1336 }
1337 EXPORT_SYMBOL(xfrm_state_update);
1338
1339 int xfrm_state_check_expire(struct xfrm_state *x)
1340 {
1341         if (!x->curlft.use_time)
1342                 x->curlft.use_time = get_seconds();
1343
1344         if (x->km.state != XFRM_STATE_VALID)
1345                 return -EINVAL;
1346
1347         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1348             x->curlft.packets >= x->lft.hard_packet_limit) {
1349                 x->km.state = XFRM_STATE_EXPIRED;
1350                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1351                 return -EINVAL;
1352         }
1353
1354         if (!x->km.dying &&
1355             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1356              x->curlft.packets >= x->lft.soft_packet_limit)) {
1357                 x->km.dying = 1;
1358                 km_state_expired(x, 0, 0);
1359         }
1360         return 0;
1361 }
1362 EXPORT_SYMBOL(xfrm_state_check_expire);
1363
1364 struct xfrm_state *
1365 xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi,
1366                   u8 proto, unsigned short family)
1367 {
1368         struct xfrm_state *x;
1369
1370         spin_lock_bh(&xfrm_state_lock);
1371         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1372         spin_unlock_bh(&xfrm_state_lock);
1373         return x;
1374 }
1375 EXPORT_SYMBOL(xfrm_state_lookup);
1376
1377 struct xfrm_state *
1378 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1379                          xfrm_address_t *daddr, xfrm_address_t *saddr,
1380                          u8 proto, unsigned short family)
1381 {
1382         struct xfrm_state *x;
1383
1384         spin_lock_bh(&xfrm_state_lock);
1385         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1386         spin_unlock_bh(&xfrm_state_lock);
1387         return x;
1388 }
1389 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1390
1391 struct xfrm_state *
1392 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1393               xfrm_address_t *daddr, xfrm_address_t *saddr,
1394               int create, unsigned short family)
1395 {
1396         struct xfrm_state *x;
1397
1398         spin_lock_bh(&xfrm_state_lock);
1399         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1400         spin_unlock_bh(&xfrm_state_lock);
1401
1402         return x;
1403 }
1404 EXPORT_SYMBOL(xfrm_find_acq);
1405
1406 #ifdef CONFIG_XFRM_SUB_POLICY
1407 int
1408 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1409                unsigned short family)
1410 {
1411         int err = 0;
1412         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1413         if (!afinfo)
1414                 return -EAFNOSUPPORT;
1415
1416         spin_lock_bh(&xfrm_state_lock);
1417         if (afinfo->tmpl_sort)
1418                 err = afinfo->tmpl_sort(dst, src, n);
1419         spin_unlock_bh(&xfrm_state_lock);
1420         xfrm_state_put_afinfo(afinfo);
1421         return err;
1422 }
1423 EXPORT_SYMBOL(xfrm_tmpl_sort);
1424
1425 int
1426 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **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->state_sort)
1436                 err = afinfo->state_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_state_sort);
1442 #endif
1443
1444 /* Silly enough, but I'm lazy to build resolution list */
1445
1446 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1447 {
1448         int i;
1449
1450         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1451                 struct hlist_node *entry;
1452                 struct xfrm_state *x;
1453
1454                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1455                         if (x->km.seq == seq &&
1456                             (mark & x->mark.m) == x->mark.v &&
1457                             x->km.state == XFRM_STATE_ACQ) {
1458                                 xfrm_state_hold(x);
1459                                 return x;
1460                         }
1461                 }
1462         }
1463         return NULL;
1464 }
1465
1466 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1467 {
1468         struct xfrm_state *x;
1469
1470         spin_lock_bh(&xfrm_state_lock);
1471         x = __xfrm_find_acq_byseq(net, mark, seq);
1472         spin_unlock_bh(&xfrm_state_lock);
1473         return x;
1474 }
1475 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1476
1477 u32 xfrm_get_acqseq(void)
1478 {
1479         u32 res;
1480         static atomic_t acqseq;
1481
1482         do {
1483                 res = atomic_inc_return(&acqseq);
1484         } while (!res);
1485
1486         return res;
1487 }
1488 EXPORT_SYMBOL(xfrm_get_acqseq);
1489
1490 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1491 {
1492         struct net *net = xs_net(x);
1493         unsigned int h;
1494         struct xfrm_state *x0;
1495         int err = -ENOENT;
1496         __be32 minspi = htonl(low);
1497         __be32 maxspi = htonl(high);
1498         u32 mark = x->mark.v & x->mark.m;
1499
1500         spin_lock_bh(&x->lock);
1501         if (x->km.state == XFRM_STATE_DEAD)
1502                 goto unlock;
1503
1504         err = 0;
1505         if (x->id.spi)
1506                 goto unlock;
1507
1508         err = -ENOENT;
1509
1510         if (minspi == maxspi) {
1511                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1512                 if (x0) {
1513                         xfrm_state_put(x0);
1514                         goto unlock;
1515                 }
1516                 x->id.spi = minspi;
1517         } else {
1518                 u32 spi = 0;
1519                 for (h=0; h<high-low+1; h++) {
1520                         spi = low + net_random()%(high-low+1);
1521                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1522                         if (x0 == NULL) {
1523                                 x->id.spi = htonl(spi);
1524                                 break;
1525                         }
1526                         xfrm_state_put(x0);
1527                 }
1528         }
1529         if (x->id.spi) {
1530                 spin_lock_bh(&xfrm_state_lock);
1531                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1532                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1533                 spin_unlock_bh(&xfrm_state_lock);
1534
1535                 err = 0;
1536         }
1537
1538 unlock:
1539         spin_unlock_bh(&x->lock);
1540
1541         return err;
1542 }
1543 EXPORT_SYMBOL(xfrm_alloc_spi);
1544
1545 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1546                     int (*func)(struct xfrm_state *, int, void*),
1547                     void *data)
1548 {
1549         struct xfrm_state *state;
1550         struct xfrm_state_walk *x;
1551         int err = 0;
1552
1553         if (walk->seq != 0 && list_empty(&walk->all))
1554                 return 0;
1555
1556         spin_lock_bh(&xfrm_state_lock);
1557         if (list_empty(&walk->all))
1558                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1559         else
1560                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1561         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1562                 if (x->state == XFRM_STATE_DEAD)
1563                         continue;
1564                 state = container_of(x, struct xfrm_state, km);
1565                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1566                         continue;
1567                 err = func(state, walk->seq, data);
1568                 if (err) {
1569                         list_move_tail(&walk->all, &x->all);
1570                         goto out;
1571                 }
1572                 walk->seq++;
1573         }
1574         if (walk->seq == 0) {
1575                 err = -ENOENT;
1576                 goto out;
1577         }
1578         list_del_init(&walk->all);
1579 out:
1580         spin_unlock_bh(&xfrm_state_lock);
1581         return err;
1582 }
1583 EXPORT_SYMBOL(xfrm_state_walk);
1584
1585 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1586 {
1587         INIT_LIST_HEAD(&walk->all);
1588         walk->proto = proto;
1589         walk->state = XFRM_STATE_DEAD;
1590         walk->seq = 0;
1591 }
1592 EXPORT_SYMBOL(xfrm_state_walk_init);
1593
1594 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1595 {
1596         if (list_empty(&walk->all))
1597                 return;
1598
1599         spin_lock_bh(&xfrm_state_lock);
1600         list_del(&walk->all);
1601         spin_unlock_bh(&xfrm_state_lock);
1602 }
1603 EXPORT_SYMBOL(xfrm_state_walk_done);
1604
1605
1606 void xfrm_replay_notify(struct xfrm_state *x, int event)
1607 {
1608         struct km_event c;
1609         /* we send notify messages in case
1610          *  1. we updated on of the sequence numbers, and the seqno difference
1611          *     is at least x->replay_maxdiff, in this case we also update the
1612          *     timeout of our timer function
1613          *  2. if x->replay_maxage has elapsed since last update,
1614          *     and there were changes
1615          *
1616          *  The state structure must be locked!
1617          */
1618
1619         switch (event) {
1620         case XFRM_REPLAY_UPDATE:
1621                 if (x->replay_maxdiff &&
1622                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1623                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1624                         if (x->xflags & XFRM_TIME_DEFER)
1625                                 event = XFRM_REPLAY_TIMEOUT;
1626                         else
1627                                 return;
1628                 }
1629
1630                 break;
1631
1632         case XFRM_REPLAY_TIMEOUT:
1633                 if ((x->replay.seq == x->preplay.seq) &&
1634                     (x->replay.bitmap == x->preplay.bitmap) &&
1635                     (x->replay.oseq == x->preplay.oseq)) {
1636                         x->xflags |= XFRM_TIME_DEFER;
1637                         return;
1638                 }
1639
1640                 break;
1641         }
1642
1643         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1644         c.event = XFRM_MSG_NEWAE;
1645         c.data.aevent = event;
1646         km_state_notify(x, &c);
1647
1648         if (x->replay_maxage &&
1649             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1650                 x->xflags &= ~XFRM_TIME_DEFER;
1651 }
1652
1653 static void xfrm_replay_timer_handler(unsigned long data)
1654 {
1655         struct xfrm_state *x = (struct xfrm_state*)data;
1656
1657         spin_lock(&x->lock);
1658
1659         if (x->km.state == XFRM_STATE_VALID) {
1660                 if (xfrm_aevent_is_on(xs_net(x)))
1661                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1662                 else
1663                         x->xflags |= XFRM_TIME_DEFER;
1664         }
1665
1666         spin_unlock(&x->lock);
1667 }
1668
1669 int xfrm_replay_check(struct xfrm_state *x,
1670                       struct sk_buff *skb, __be32 net_seq)
1671 {
1672         u32 diff;
1673         u32 seq = ntohl(net_seq);
1674
1675         if (unlikely(seq == 0))
1676                 goto err;
1677
1678         if (likely(seq > x->replay.seq))
1679                 return 0;
1680
1681         diff = x->replay.seq - seq;
1682         if (diff >= min_t(unsigned int, x->props.replay_window,
1683                           sizeof(x->replay.bitmap) * 8)) {
1684                 x->stats.replay_window++;
1685                 goto err;
1686         }
1687
1688         if (x->replay.bitmap & (1U << diff)) {
1689                 x->stats.replay++;
1690                 goto err;
1691         }
1692         return 0;
1693
1694 err:
1695         xfrm_audit_state_replay(x, skb, net_seq);
1696         return -EINVAL;
1697 }
1698
1699 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1700 {
1701         u32 diff;
1702         u32 seq = ntohl(net_seq);
1703
1704         if (seq > x->replay.seq) {
1705                 diff = seq - x->replay.seq;
1706                 if (diff < x->props.replay_window)
1707                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1708                 else
1709                         x->replay.bitmap = 1;
1710                 x->replay.seq = seq;
1711         } else {
1712                 diff = x->replay.seq - seq;
1713                 x->replay.bitmap |= (1U << diff);
1714         }
1715
1716         if (xfrm_aevent_is_on(xs_net(x)))
1717                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1718 }
1719
1720 static LIST_HEAD(xfrm_km_list);
1721 static DEFINE_RWLOCK(xfrm_km_lock);
1722
1723 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1724 {
1725         struct xfrm_mgr *km;
1726
1727         read_lock(&xfrm_km_lock);
1728         list_for_each_entry(km, &xfrm_km_list, list)
1729                 if (km->notify_policy)
1730                         km->notify_policy(xp, dir, c);
1731         read_unlock(&xfrm_km_lock);
1732 }
1733
1734 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1735 {
1736         struct xfrm_mgr *km;
1737         read_lock(&xfrm_km_lock);
1738         list_for_each_entry(km, &xfrm_km_list, list)
1739                 if (km->notify)
1740                         km->notify(x, c);
1741         read_unlock(&xfrm_km_lock);
1742 }
1743
1744 EXPORT_SYMBOL(km_policy_notify);
1745 EXPORT_SYMBOL(km_state_notify);
1746
1747 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1748 {
1749         struct net *net = xs_net(x);
1750         struct km_event c;
1751
1752         c.data.hard = hard;
1753         c.pid = pid;
1754         c.event = XFRM_MSG_EXPIRE;
1755         km_state_notify(x, &c);
1756
1757         if (hard)
1758                 wake_up(&net->xfrm.km_waitq);
1759 }
1760
1761 EXPORT_SYMBOL(km_state_expired);
1762 /*
1763  * We send to all registered managers regardless of failure
1764  * We are happy with one success
1765 */
1766 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1767 {
1768         int err = -EINVAL, acqret;
1769         struct xfrm_mgr *km;
1770
1771         read_lock(&xfrm_km_lock);
1772         list_for_each_entry(km, &xfrm_km_list, list) {
1773                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1774                 if (!acqret)
1775                         err = acqret;
1776         }
1777         read_unlock(&xfrm_km_lock);
1778         return err;
1779 }
1780 EXPORT_SYMBOL(km_query);
1781
1782 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1783 {
1784         int err = -EINVAL;
1785         struct xfrm_mgr *km;
1786
1787         read_lock(&xfrm_km_lock);
1788         list_for_each_entry(km, &xfrm_km_list, list) {
1789                 if (km->new_mapping)
1790                         err = km->new_mapping(x, ipaddr, sport);
1791                 if (!err)
1792                         break;
1793         }
1794         read_unlock(&xfrm_km_lock);
1795         return err;
1796 }
1797 EXPORT_SYMBOL(km_new_mapping);
1798
1799 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1800 {
1801         struct net *net = xp_net(pol);
1802         struct km_event c;
1803
1804         c.data.hard = hard;
1805         c.pid = pid;
1806         c.event = XFRM_MSG_POLEXPIRE;
1807         km_policy_notify(pol, dir, &c);
1808
1809         if (hard)
1810                 wake_up(&net->xfrm.km_waitq);
1811 }
1812 EXPORT_SYMBOL(km_policy_expired);
1813
1814 #ifdef CONFIG_XFRM_MIGRATE
1815 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1816                struct xfrm_migrate *m, int num_migrate,
1817                struct xfrm_kmaddress *k)
1818 {
1819         int err = -EINVAL;
1820         int ret;
1821         struct xfrm_mgr *km;
1822
1823         read_lock(&xfrm_km_lock);
1824         list_for_each_entry(km, &xfrm_km_list, list) {
1825                 if (km->migrate) {
1826                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1827                         if (!ret)
1828                                 err = ret;
1829                 }
1830         }
1831         read_unlock(&xfrm_km_lock);
1832         return err;
1833 }
1834 EXPORT_SYMBOL(km_migrate);
1835 #endif
1836
1837 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1838 {
1839         int err = -EINVAL;
1840         int ret;
1841         struct xfrm_mgr *km;
1842
1843         read_lock(&xfrm_km_lock);
1844         list_for_each_entry(km, &xfrm_km_list, list) {
1845                 if (km->report) {
1846                         ret = km->report(net, proto, sel, addr);
1847                         if (!ret)
1848                                 err = ret;
1849                 }
1850         }
1851         read_unlock(&xfrm_km_lock);
1852         return err;
1853 }
1854 EXPORT_SYMBOL(km_report);
1855
1856 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1857 {
1858         int err;
1859         u8 *data;
1860         struct xfrm_mgr *km;
1861         struct xfrm_policy *pol = NULL;
1862
1863         if (optlen <= 0 || optlen > PAGE_SIZE)
1864                 return -EMSGSIZE;
1865
1866         data = kmalloc(optlen, GFP_KERNEL);
1867         if (!data)
1868                 return -ENOMEM;
1869
1870         err = -EFAULT;
1871         if (copy_from_user(data, optval, optlen))
1872                 goto out;
1873
1874         err = -EINVAL;
1875         read_lock(&xfrm_km_lock);
1876         list_for_each_entry(km, &xfrm_km_list, list) {
1877                 pol = km->compile_policy(sk, optname, data,
1878                                          optlen, &err);
1879                 if (err >= 0)
1880                         break;
1881         }
1882         read_unlock(&xfrm_km_lock);
1883
1884         if (err >= 0) {
1885                 xfrm_sk_policy_insert(sk, err, pol);
1886                 xfrm_pol_put(pol);
1887                 err = 0;
1888         }
1889
1890 out:
1891         kfree(data);
1892         return err;
1893 }
1894 EXPORT_SYMBOL(xfrm_user_policy);
1895
1896 int xfrm_register_km(struct xfrm_mgr *km)
1897 {
1898         write_lock_bh(&xfrm_km_lock);
1899         list_add_tail(&km->list, &xfrm_km_list);
1900         write_unlock_bh(&xfrm_km_lock);
1901         return 0;
1902 }
1903 EXPORT_SYMBOL(xfrm_register_km);
1904
1905 int xfrm_unregister_km(struct xfrm_mgr *km)
1906 {
1907         write_lock_bh(&xfrm_km_lock);
1908         list_del(&km->list);
1909         write_unlock_bh(&xfrm_km_lock);
1910         return 0;
1911 }
1912 EXPORT_SYMBOL(xfrm_unregister_km);
1913
1914 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1915 {
1916         int err = 0;
1917         if (unlikely(afinfo == NULL))
1918                 return -EINVAL;
1919         if (unlikely(afinfo->family >= NPROTO))
1920                 return -EAFNOSUPPORT;
1921         write_lock_bh(&xfrm_state_afinfo_lock);
1922         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1923                 err = -ENOBUFS;
1924         else
1925                 xfrm_state_afinfo[afinfo->family] = afinfo;
1926         write_unlock_bh(&xfrm_state_afinfo_lock);
1927         return err;
1928 }
1929 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1930
1931 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1932 {
1933         int err = 0;
1934         if (unlikely(afinfo == NULL))
1935                 return -EINVAL;
1936         if (unlikely(afinfo->family >= NPROTO))
1937                 return -EAFNOSUPPORT;
1938         write_lock_bh(&xfrm_state_afinfo_lock);
1939         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1940                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1941                         err = -EINVAL;
1942                 else
1943                         xfrm_state_afinfo[afinfo->family] = NULL;
1944         }
1945         write_unlock_bh(&xfrm_state_afinfo_lock);
1946         return err;
1947 }
1948 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1949
1950 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1951 {
1952         struct xfrm_state_afinfo *afinfo;
1953         if (unlikely(family >= NPROTO))
1954                 return NULL;
1955         read_lock(&xfrm_state_afinfo_lock);
1956         afinfo = xfrm_state_afinfo[family];
1957         if (unlikely(!afinfo))
1958                 read_unlock(&xfrm_state_afinfo_lock);
1959         return afinfo;
1960 }
1961
1962 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1963         __releases(xfrm_state_afinfo_lock)
1964 {
1965         read_unlock(&xfrm_state_afinfo_lock);
1966 }
1967
1968 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1969 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1970 {
1971         if (x->tunnel) {
1972                 struct xfrm_state *t = x->tunnel;
1973
1974                 if (atomic_read(&t->tunnel_users) == 2)
1975                         xfrm_state_delete(t);
1976                 atomic_dec(&t->tunnel_users);
1977                 xfrm_state_put(t);
1978                 x->tunnel = NULL;
1979         }
1980 }
1981 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1982
1983 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1984 {
1985         int res;
1986
1987         spin_lock_bh(&x->lock);
1988         if (x->km.state == XFRM_STATE_VALID &&
1989             x->type && x->type->get_mtu)
1990                 res = x->type->get_mtu(x, mtu);
1991         else
1992                 res = mtu - x->props.header_len;
1993         spin_unlock_bh(&x->lock);
1994         return res;
1995 }
1996
1997 int xfrm_init_state(struct xfrm_state *x)
1998 {
1999         struct xfrm_state_afinfo *afinfo;
2000         struct xfrm_mode *inner_mode;
2001         int family = x->props.family;
2002         int err;
2003
2004         err = -EAFNOSUPPORT;
2005         afinfo = xfrm_state_get_afinfo(family);
2006         if (!afinfo)
2007                 goto error;
2008
2009         err = 0;
2010         if (afinfo->init_flags)
2011                 err = afinfo->init_flags(x);
2012
2013         xfrm_state_put_afinfo(afinfo);
2014
2015         if (err)
2016                 goto error;
2017
2018         err = -EPROTONOSUPPORT;
2019
2020         if (x->sel.family != AF_UNSPEC) {
2021                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2022                 if (inner_mode == NULL)
2023                         goto error;
2024
2025                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2026                     family != x->sel.family) {
2027                         xfrm_put_mode(inner_mode);
2028                         goto error;
2029                 }
2030
2031                 x->inner_mode = inner_mode;
2032         } else {
2033                 struct xfrm_mode *inner_mode_iaf;
2034                 int iafamily = AF_INET;
2035
2036                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2037                 if (inner_mode == NULL)
2038                         goto error;
2039
2040                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2041                         xfrm_put_mode(inner_mode);
2042                         goto error;
2043                 }
2044                 x->inner_mode = inner_mode;
2045
2046                 if (x->props.family == AF_INET)
2047                         iafamily = AF_INET6;
2048
2049                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2050                 if (inner_mode_iaf) {
2051                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2052                                 x->inner_mode_iaf = inner_mode_iaf;
2053                         else
2054                                 xfrm_put_mode(inner_mode_iaf);
2055                 }
2056         }
2057
2058         x->type = xfrm_get_type(x->id.proto, family);
2059         if (x->type == NULL)
2060                 goto error;
2061
2062         err = x->type->init_state(x);
2063         if (err)
2064                 goto error;
2065
2066         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2067         if (x->outer_mode == NULL)
2068                 goto error;
2069
2070         x->km.state = XFRM_STATE_VALID;
2071
2072 error:
2073         return err;
2074 }
2075
2076 EXPORT_SYMBOL(xfrm_init_state);
2077
2078 int __net_init xfrm_state_init(struct net *net)
2079 {
2080         unsigned int sz;
2081
2082         INIT_LIST_HEAD(&net->xfrm.state_all);
2083
2084         sz = sizeof(struct hlist_head) * 8;
2085
2086         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2087         if (!net->xfrm.state_bydst)
2088                 goto out_bydst;
2089         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2090         if (!net->xfrm.state_bysrc)
2091                 goto out_bysrc;
2092         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2093         if (!net->xfrm.state_byspi)
2094                 goto out_byspi;
2095         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2096
2097         net->xfrm.state_num = 0;
2098         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2099         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2100         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2101         init_waitqueue_head(&net->xfrm.km_waitq);
2102         return 0;
2103
2104 out_byspi:
2105         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2106 out_bysrc:
2107         xfrm_hash_free(net->xfrm.state_bydst, sz);
2108 out_bydst:
2109         return -ENOMEM;
2110 }
2111
2112 void xfrm_state_fini(struct net *net)
2113 {
2114         struct xfrm_audit audit_info;
2115         unsigned int sz;
2116
2117         flush_work(&net->xfrm.state_hash_work);
2118         audit_info.loginuid = -1;
2119         audit_info.sessionid = -1;
2120         audit_info.secid = 0;
2121         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2122         flush_work(&net->xfrm.state_gc_work);
2123
2124         WARN_ON(!list_empty(&net->xfrm.state_all));
2125
2126         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2127         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2128         xfrm_hash_free(net->xfrm.state_byspi, sz);
2129         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2130         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2131         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2132         xfrm_hash_free(net->xfrm.state_bydst, sz);
2133 }
2134
2135 #ifdef CONFIG_AUDITSYSCALL
2136 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2137                                      struct audit_buffer *audit_buf)
2138 {
2139         struct xfrm_sec_ctx *ctx = x->security;
2140         u32 spi = ntohl(x->id.spi);
2141
2142         if (ctx)
2143                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2144                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2145
2146         switch(x->props.family) {
2147         case AF_INET:
2148                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2149                                  &x->props.saddr.a4, &x->id.daddr.a4);
2150                 break;
2151         case AF_INET6:
2152                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2153                                  x->props.saddr.a6, x->id.daddr.a6);
2154                 break;
2155         }
2156
2157         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158 }
2159
2160 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2161                                       struct audit_buffer *audit_buf)
2162 {
2163         struct iphdr *iph4;
2164         struct ipv6hdr *iph6;
2165
2166         switch (family) {
2167         case AF_INET:
2168                 iph4 = ip_hdr(skb);
2169                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2170                                  &iph4->saddr, &iph4->daddr);
2171                 break;
2172         case AF_INET6:
2173                 iph6 = ipv6_hdr(skb);
2174                 audit_log_format(audit_buf,
2175                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2176                                  &iph6->saddr,&iph6->daddr,
2177                                  iph6->flow_lbl[0] & 0x0f,
2178                                  iph6->flow_lbl[1],
2179                                  iph6->flow_lbl[2]);
2180                 break;
2181         }
2182 }
2183
2184 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2185                           uid_t auid, u32 sessionid, u32 secid)
2186 {
2187         struct audit_buffer *audit_buf;
2188
2189         audit_buf = xfrm_audit_start("SAD-add");
2190         if (audit_buf == NULL)
2191                 return;
2192         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2193         xfrm_audit_helper_sainfo(x, audit_buf);
2194         audit_log_format(audit_buf, " res=%u", result);
2195         audit_log_end(audit_buf);
2196 }
2197 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2198
2199 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2200                              uid_t auid, u32 sessionid, u32 secid)
2201 {
2202         struct audit_buffer *audit_buf;
2203
2204         audit_buf = xfrm_audit_start("SAD-delete");
2205         if (audit_buf == NULL)
2206                 return;
2207         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2208         xfrm_audit_helper_sainfo(x, audit_buf);
2209         audit_log_format(audit_buf, " res=%u", result);
2210         audit_log_end(audit_buf);
2211 }
2212 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2213
2214 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2215                                       struct sk_buff *skb)
2216 {
2217         struct audit_buffer *audit_buf;
2218         u32 spi;
2219
2220         audit_buf = xfrm_audit_start("SA-replay-overflow");
2221         if (audit_buf == NULL)
2222                 return;
2223         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2224         /* don't record the sequence number because it's inherent in this kind
2225          * of audit message */
2226         spi = ntohl(x->id.spi);
2227         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2228         audit_log_end(audit_buf);
2229 }
2230 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2231
2232 static void xfrm_audit_state_replay(struct xfrm_state *x,
2233                              struct sk_buff *skb, __be32 net_seq)
2234 {
2235         struct audit_buffer *audit_buf;
2236         u32 spi;
2237
2238         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2239         if (audit_buf == NULL)
2240                 return;
2241         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2242         spi = ntohl(x->id.spi);
2243         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2244                          spi, spi, ntohl(net_seq));
2245         audit_log_end(audit_buf);
2246 }
2247
2248 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2249 {
2250         struct audit_buffer *audit_buf;
2251
2252         audit_buf = xfrm_audit_start("SA-notfound");
2253         if (audit_buf == NULL)
2254                 return;
2255         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2256         audit_log_end(audit_buf);
2257 }
2258 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2259
2260 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2261                                __be32 net_spi, __be32 net_seq)
2262 {
2263         struct audit_buffer *audit_buf;
2264         u32 spi;
2265
2266         audit_buf = xfrm_audit_start("SA-notfound");
2267         if (audit_buf == NULL)
2268                 return;
2269         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2270         spi = ntohl(net_spi);
2271         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2272                          spi, spi, ntohl(net_seq));
2273         audit_log_end(audit_buf);
2274 }
2275 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2276
2277 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2278                               struct sk_buff *skb, u8 proto)
2279 {
2280         struct audit_buffer *audit_buf;
2281         __be32 net_spi;
2282         __be32 net_seq;
2283
2284         audit_buf = xfrm_audit_start("SA-icv-failure");
2285         if (audit_buf == NULL)
2286                 return;
2287         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2288         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2289                 u32 spi = ntohl(net_spi);
2290                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2291                                  spi, spi, ntohl(net_seq));
2292         }
2293         audit_log_end(audit_buf);
2294 }
2295 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2296 #endif /* CONFIG_AUDITSYSCALL */