net/xfrm/xfrm_state.c

   1 /*
   2  * xfrm_state.c
   3  *
   4  * Changes:
   5  *      Mitsuru KANDA @USAGI
   6  *      Kazunori MIYAZAWA @USAGI
   7  *      Kunihiro Ishiguro <kunihiro@ipinfusion.com>
   8  *              IPv6 support
   9  *      YOSHIFUJI Hideaki @USAGI
  10  *              Split up af-specific functions
  11  *      Derek Atkins <derek@ihtfp.com>
  12  *              Add UDP Encapsulation
  13  *
  14  */
  15
  16 #include <linux/workqueue.h>
  17 #include <net/xfrm.h>
  18 #include <linux/pfkeyv2.h>
  19 #include <linux/ipsec.h>
  20 #include <linux/module.h>
  21 #include <linux/cache.h>
  22 #include <linux/audit.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/ktime.h>
  25 #include <linux/slab.h>
  26 #include <linux/interrupt.h>
  27 #include <linux/kernel.h>
  28
  29 #include "xfrm_hash.h"
  30
  31 /* Each xfrm_state may be linked to two tables:
  32
  33    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  34    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  35       destination/tunnel endpoint. (output)
  36  */
  37
  38 static DEFINE_SPINLOCK(xfrm_state_lock);
  39
  40 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  41
  42 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
  43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
  44
  45 #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                                          const xfrm_address_t *daddr,
  54                                          const 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                                          const xfrm_address_t *daddr,
  63                                          const 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, const xfrm_address_t *daddr,
  71               __be32 spi, u8 proto, unsigned short family)
  72 {
  73         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  74 }
  75
  76 static void xfrm_hash_transfer(struct hlist_head *list,
  77                                struct hlist_head *ndsttable,
  78                                struct hlist_head *nsrctable,
  79                                struct hlist_head *nspitable,
  80                                unsigned int nhashmask)
  81 {
  82         struct hlist_node *entry, *tmp;
  83         struct xfrm_state *x;
  84
  85         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
  86                 unsigned int h;
  87
  88                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  89                                     x->props.reqid, x->props.family,
  90                                     nhashmask);
  91                 hlist_add_head(&x->bydst, ndsttable+h);
  92
  93                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  94                                     x->props.family,
  95                                     nhashmask);
  96                 hlist_add_head(&x->bysrc, nsrctable+h);
  97
  98                 if (x->id.spi) {
  99                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
 100                                             x->id.proto, x->props.family,
 101                                             nhashmask);
 102                         hlist_add_head(&x->byspi, nspitable+h);
 103                 }
 104         }
 105 }
 106
 107 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 108 {
 109         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 110 }
 111
 112 static DEFINE_MUTEX(hash_resize_mutex);
 113
 114 static void xfrm_hash_resize(struct work_struct *work)
 115 {
 116         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 117         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 118         unsigned long nsize, osize;
 119         unsigned int nhashmask, ohashmask;
 120         int i;
 121
 122         mutex_lock(&hash_resize_mutex);
 123
 124         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 125         ndst = xfrm_hash_alloc(nsize);
 126         if (!ndst)
 127                 goto out_unlock;
 128         nsrc = xfrm_hash_alloc(nsize);
 129         if (!nsrc) {
 130                 xfrm_hash_free(ndst, nsize);
 131                 goto out_unlock;
 132         }
 133         nspi = xfrm_hash_alloc(nsize);
 134         if (!nspi) {
 135                 xfrm_hash_free(ndst, nsize);
 136                 xfrm_hash_free(nsrc, nsize);
 137                 goto out_unlock;
 138         }
 139
 140         spin_lock_bh(&xfrm_state_lock);
 141
 142         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 143         for (i = net->xfrm.state_hmask; i >= 0; i--)
 144                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 145                                    nhashmask);
 146
 147         odst = net->xfrm.state_bydst;
 148         osrc = net->xfrm.state_bysrc;
 149         ospi = net->xfrm.state_byspi;
 150         ohashmask = net->xfrm.state_hmask;
 151
 152         net->xfrm.state_bydst = ndst;
 153         net->xfrm.state_bysrc = nsrc;
 154         net->xfrm.state_byspi = nspi;
 155         net->xfrm.state_hmask = nhashmask;
 156
 157         spin_unlock_bh(&xfrm_state_lock);
 158
 159         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 160         xfrm_hash_free(odst, osize);
 161         xfrm_hash_free(osrc, osize);
 162         xfrm_hash_free(ospi, osize);
 163
 164 out_unlock:
 165         mutex_unlock(&hash_resize_mutex);
 166 }
 167
 168 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
 169 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
 170
 171 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 172
 173 int __xfrm_state_delete(struct xfrm_state *x);
 174
 175 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 176 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
 177
 178 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 179 {
 180         struct xfrm_state_afinfo *afinfo;
 181         if (unlikely(family >= NPROTO))
 182                 return NULL;
 183         write_lock_bh(&xfrm_state_afinfo_lock);
 184         afinfo = xfrm_state_afinfo[family];
 185         if (unlikely(!afinfo))
 186                 write_unlock_bh(&xfrm_state_afinfo_lock);
 187         return afinfo;
 188 }
 189
 190 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 191         __releases(xfrm_state_afinfo_lock)
 192 {
 193         write_unlock_bh(&xfrm_state_afinfo_lock);
 194 }
 195
 196 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 197 {
 198         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 199         const struct xfrm_type **typemap;
 200         int err = 0;
 201
 202         if (unlikely(afinfo == NULL))
 203                 return -EAFNOSUPPORT;
 204         typemap = afinfo->type_map;
 205
 206         if (likely(typemap[type->proto] == NULL))
 207                 typemap[type->proto] = type;
 208         else
 209                 err = -EEXIST;
 210         xfrm_state_unlock_afinfo(afinfo);
 211         return err;
 212 }
 213 EXPORT_SYMBOL(xfrm_register_type);
 214
 215 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 216 {
 217         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 218         const struct xfrm_type **typemap;
 219         int err = 0;
 220
 221         if (unlikely(afinfo == NULL))
 222                 return -EAFNOSUPPORT;
 223         typemap = afinfo->type_map;
 224
 225         if (unlikely(typemap[type->proto] != type))
 226                 err = -ENOENT;
 227         else
 228                 typemap[type->proto] = NULL;
 229         xfrm_state_unlock_afinfo(afinfo);
 230         return err;
 231 }
 232 EXPORT_SYMBOL(xfrm_unregister_type);
 233
 234 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 235 {
 236         struct xfrm_state_afinfo *afinfo;
 237         const struct xfrm_type **typemap;
 238         const struct xfrm_type *type;
 239         int modload_attempted = 0;
 240
 241 retry:
 242         afinfo = xfrm_state_get_afinfo(family);
 243         if (unlikely(afinfo == NULL))
 244                 return NULL;
 245         typemap = afinfo->type_map;
 246
 247         type = typemap[proto];
 248         if (unlikely(type && !try_module_get(type->owner)))
 249                 type = NULL;
 250         if (!type && !modload_attempted) {
 251                 xfrm_state_put_afinfo(afinfo);
 252                 request_module("xfrm-type-%d-%d", family, proto);
 253                 modload_attempted = 1;
 254                 goto retry;
 255         }
 256
 257         xfrm_state_put_afinfo(afinfo);
 258         return type;
 259 }
 260
 261 static void xfrm_put_type(const struct xfrm_type *type)
 262 {
 263         module_put(type->owner);
 264 }
 265
 266 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 267 {
 268         struct xfrm_state_afinfo *afinfo;
 269         struct xfrm_mode **modemap;
 270         int err;
 271
 272         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 273                 return -EINVAL;
 274
 275         afinfo = xfrm_state_lock_afinfo(family);
 276         if (unlikely(afinfo == NULL))
 277                 return -EAFNOSUPPORT;
 278
 279         err = -EEXIST;
 280         modemap = afinfo->mode_map;
 281         if (modemap[mode->encap])
 282                 goto out;
 283
 284         err = -ENOENT;
 285         if (!try_module_get(afinfo->owner))
 286                 goto out;
 287
 288         mode->afinfo = afinfo;
 289         modemap[mode->encap] = mode;
 290         err = 0;
 291
 292 out:
 293         xfrm_state_unlock_afinfo(afinfo);
 294         return err;
 295 }
 296 EXPORT_SYMBOL(xfrm_register_mode);
 297
 298 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 299 {
 300         struct xfrm_state_afinfo *afinfo;
 301         struct xfrm_mode **modemap;
 302         int err;
 303
 304         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 305                 return -EINVAL;
 306
 307         afinfo = xfrm_state_lock_afinfo(family);
 308         if (unlikely(afinfo == NULL))
 309                 return -EAFNOSUPPORT;
 310
 311         err = -ENOENT;
 312         modemap = afinfo->mode_map;
 313         if (likely(modemap[mode->encap] == mode)) {
 314                 modemap[mode->encap] = NULL;
 315                 module_put(mode->afinfo->owner);
 316                 err = 0;
 317         }
 318
 319         xfrm_state_unlock_afinfo(afinfo);
 320         return err;
 321 }
 322 EXPORT_SYMBOL(xfrm_unregister_mode);
 323
 324 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 325 {
 326         struct xfrm_state_afinfo *afinfo;
 327         struct xfrm_mode *mode;
 328         int modload_attempted = 0;
 329
 330         if (unlikely(encap >= XFRM_MODE_MAX))
 331                 return NULL;
 332
 333 retry:
 334         afinfo = xfrm_state_get_afinfo(family);
 335         if (unlikely(afinfo == NULL))
 336                 return NULL;
 337
 338         mode = afinfo->mode_map[encap];
 339         if (unlikely(mode && !try_module_get(mode->owner)))
 340                 mode = NULL;
 341         if (!mode && !modload_attempted) {
 342                 xfrm_state_put_afinfo(afinfo);
 343                 request_module("xfrm-mode-%d-%d", family, encap);
 344                 modload_attempted = 1;
 345                 goto retry;
 346         }
 347
 348         xfrm_state_put_afinfo(afinfo);
 349         return mode;
 350 }
 351
 352 static void xfrm_put_mode(struct xfrm_mode *mode)
 353 {
 354         module_put(mode->owner);
 355 }
 356
 357 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 358 {
 359         tasklet_hrtimer_cancel(&x->mtimer);
 360         del_timer_sync(&x->rtimer);
 361         kfree(x->aalg);
 362         kfree(x->ealg);
 363         kfree(x->calg);
 364         kfree(x->encap);
 365         kfree(x->coaddr);
 366         if (x->inner_mode)
 367                 xfrm_put_mode(x->inner_mode);
 368         if (x->inner_mode_iaf)
 369                 xfrm_put_mode(x->inner_mode_iaf);
 370         if (x->outer_mode)
 371                 xfrm_put_mode(x->outer_mode);
 372         if (x->type) {
 373                 x->type->destructor(x);
 374                 xfrm_put_type(x->type);
 375         }
 376         security_xfrm_state_free(x);
 377         kfree(x);
 378 }
 379
 380 static void xfrm_state_gc_task(struct work_struct *work)
 381 {
 382         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 383         struct xfrm_state *x;
 384         struct hlist_node *entry, *tmp;
 385         struct hlist_head gc_list;
 386
 387         spin_lock_bh(&xfrm_state_gc_lock);
 388         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 389         spin_unlock_bh(&xfrm_state_gc_lock);
 390
 391         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 392                 xfrm_state_gc_destroy(x);
 393
 394         wake_up(&net->xfrm.km_waitq);
 395 }
 396
 397 static inline unsigned long make_jiffies(long secs)
 398 {
 399         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 400                 return MAX_SCHEDULE_TIMEOUT-1;
 401         else
 402                 return secs*HZ;
 403 }
 404
 405 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 406 {
 407         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 408         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 409         struct net *net = xs_net(x);
 410         unsigned long now = get_seconds();
 411         long next = LONG_MAX;
 412         int warn = 0;
 413         int err = 0;
 414
 415         spin_lock(&x->lock);
 416         if (x->km.state == XFRM_STATE_DEAD)
 417                 goto out;
 418         if (x->km.state == XFRM_STATE_EXPIRED)
 419                 goto expired;
 420         if (x->lft.hard_add_expires_seconds) {
 421                 long tmo = x->lft.hard_add_expires_seconds +
 422                         x->curlft.add_time - now;
 423                 if (tmo <= 0)
 424                         goto expired;
 425                 if (tmo < next)
 426                         next = tmo;
 427         }
 428         if (x->lft.hard_use_expires_seconds) {
 429                 long tmo = x->lft.hard_use_expires_seconds +
 430                         (x->curlft.use_time ? : now) - now;
 431                 if (tmo <= 0)
 432                         goto expired;
 433                 if (tmo < next)
 434                         next = tmo;
 435         }
 436         if (x->km.dying)
 437                 goto resched;
 438         if (x->lft.soft_add_expires_seconds) {
 439                 long tmo = x->lft.soft_add_expires_seconds +
 440                         x->curlft.add_time - now;
 441                 if (tmo <= 0)
 442                         warn = 1;
 443                 else if (tmo < next)
 444                         next = tmo;
 445         }
 446         if (x->lft.soft_use_expires_seconds) {
 447                 long tmo = x->lft.soft_use_expires_seconds +
 448                         (x->curlft.use_time ? : now) - now;
 449                 if (tmo <= 0)
 450                         warn = 1;
 451                 else if (tmo < next)
 452                         next = tmo;
 453         }
 454
 455         x->km.dying = warn;
 456         if (warn)
 457                 km_state_expired(x, 0, 0);
 458 resched:
 459         if (next != LONG_MAX){
 460                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 461         }
 462
 463         goto out;
 464
 465 expired:
 466         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 467                 x->km.state = XFRM_STATE_EXPIRED;
 468                 wake_up(&net->xfrm.km_waitq);
 469                 next = 2;
 470                 goto resched;
 471         }
 472
 473         err = __xfrm_state_delete(x);
 474         if (!err && x->id.spi)
 475                 km_state_expired(x, 1, 0);
 476
 477         xfrm_audit_state_delete(x, err ? 0 : 1,
 478                                 audit_get_loginuid(current),
 479                                 audit_get_sessionid(current), 0);
 480
 481 out:
 482         spin_unlock(&x->lock);
 483         return HRTIMER_NORESTART;
 484 }
 485
 486 static void xfrm_replay_timer_handler(unsigned long data);
 487
 488 struct xfrm_state *xfrm_state_alloc(struct net *net)
 489 {
 490         struct xfrm_state *x;
 491
 492         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 493
 494         if (x) {
 495                 write_pnet(&x->xs_net, net);
 496                 atomic_set(&x->refcnt, 1);
 497                 atomic_set(&x->tunnel_users, 0);
 498                 INIT_LIST_HEAD(&x->km.all);
 499                 INIT_HLIST_NODE(&x->bydst);
 500                 INIT_HLIST_NODE(&x->bysrc);
 501                 INIT_HLIST_NODE(&x->byspi);
 502                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 503                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 504                                 (unsigned long)x);
 505                 x->curlft.add_time = get_seconds();
 506                 x->lft.soft_byte_limit = XFRM_INF;
 507                 x->lft.soft_packet_limit = XFRM_INF;
 508                 x->lft.hard_byte_limit = XFRM_INF;
 509                 x->lft.hard_packet_limit = XFRM_INF;
 510                 x->replay_maxage = 0;
 511                 x->replay_maxdiff = 0;
 512                 x->inner_mode = NULL;
 513                 x->inner_mode_iaf = NULL;
 514                 spin_lock_init(&x->lock);
 515         }
 516         return x;
 517 }
 518 EXPORT_SYMBOL(xfrm_state_alloc);
 519
 520 void __xfrm_state_destroy(struct xfrm_state *x)
 521 {
 522         struct net *net = xs_net(x);
 523
 524         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 525
 526         spin_lock_bh(&xfrm_state_gc_lock);
 527         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 528         spin_unlock_bh(&xfrm_state_gc_lock);
 529         schedule_work(&net->xfrm.state_gc_work);
 530 }
 531 EXPORT_SYMBOL(__xfrm_state_destroy);
 532
 533 int __xfrm_state_delete(struct xfrm_state *x)
 534 {
 535         struct net *net = xs_net(x);
 536         int err = -ESRCH;
 537
 538         if (x->km.state != XFRM_STATE_DEAD) {
 539                 x->km.state = XFRM_STATE_DEAD;
 540                 spin_lock(&xfrm_state_lock);
 541                 list_del(&x->km.all);
 542                 hlist_del(&x->bydst);
 543                 hlist_del(&x->bysrc);
 544                 if (x->id.spi)
 545                         hlist_del(&x->byspi);
 546                 net->xfrm.state_num--;
 547                 spin_unlock(&xfrm_state_lock);
 548
 549                 /* All xfrm_state objects are created by xfrm_state_alloc.
 550                  * The xfrm_state_alloc call gives a reference, and that
 551                  * is what we are dropping here.
 552                  */
 553                 xfrm_state_put(x);
 554                 err = 0;
 555         }
 556
 557         return err;
 558 }
 559 EXPORT_SYMBOL(__xfrm_state_delete);
 560
 561 int xfrm_state_delete(struct xfrm_state *x)
 562 {
 563         int err;
 564
 565         spin_lock_bh(&x->lock);
 566         err = __xfrm_state_delete(x);
 567         spin_unlock_bh(&x->lock);
 568
 569         return err;
 570 }
 571 EXPORT_SYMBOL(xfrm_state_delete);
 572
 573 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 574 static inline int
 575 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 576 {
 577         int i, err = 0;
 578
 579         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 580                 struct hlist_node *entry;
 581                 struct xfrm_state *x;
 582
 583                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 584                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 585                            (err = security_xfrm_state_delete(x)) != 0) {
 586                                 xfrm_audit_state_delete(x, 0,
 587                                                         audit_info->loginuid,
 588                                                         audit_info->sessionid,
 589                                                         audit_info->secid);
 590                                 return err;
 591                         }
 592                 }
 593         }
 594
 595         return err;
 596 }
 597 #else
 598 static inline int
 599 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 600 {
 601         return 0;
 602 }
 603 #endif
 604
 605 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 606 {
 607         int i, err = 0, cnt = 0;
 608
 609         spin_lock_bh(&xfrm_state_lock);
 610         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 611         if (err)
 612                 goto out;
 613
 614         err = -ESRCH;
 615         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 616                 struct hlist_node *entry;
 617                 struct xfrm_state *x;
 618 restart:
 619                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 620                         if (!xfrm_state_kern(x) &&
 621                             xfrm_id_proto_match(x->id.proto, proto)) {
 622                                 xfrm_state_hold(x);
 623                                 spin_unlock_bh(&xfrm_state_lock);
 624
 625                                 err = xfrm_state_delete(x);
 626                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 627                                                         audit_info->loginuid,
 628                                                         audit_info->sessionid,
 629                                                         audit_info->secid);
 630                                 xfrm_state_put(x);
 631                                 if (!err)
 632                                         cnt++;
 633
 634                                 spin_lock_bh(&xfrm_state_lock);
 635                                 goto restart;
 636                         }
 637                 }
 638         }
 639         if (cnt)
 640                 err = 0;
 641
 642 out:
 643         spin_unlock_bh(&xfrm_state_lock);
 644         wake_up(&net->xfrm.km_waitq);
 645         return err;
 646 }
 647 EXPORT_SYMBOL(xfrm_state_flush);
 648
 649 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 650 {
 651         spin_lock_bh(&xfrm_state_lock);
 652         si->sadcnt = net->xfrm.state_num;
 653         si->sadhcnt = net->xfrm.state_hmask;
 654         si->sadhmcnt = xfrm_state_hashmax;
 655         spin_unlock_bh(&xfrm_state_lock);
 656 }
 657 EXPORT_SYMBOL(xfrm_sad_getinfo);
 658
 659 static int
 660 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 661                     const struct xfrm_tmpl *tmpl,
 662                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 663                     unsigned short family)
 664 {
 665         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 666         if (!afinfo)
 667                 return -1;
 668         afinfo->init_tempsel(&x->sel, fl);
 669
 670         if (family != tmpl->encap_family) {
 671                 xfrm_state_put_afinfo(afinfo);
 672                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 673                 if (!afinfo)
 674                         return -1;
 675         }
 676         afinfo->init_temprop(x, tmpl, daddr, saddr);
 677         xfrm_state_put_afinfo(afinfo);
 678         return 0;
 679 }
 680
 681 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 682                                               const xfrm_address_t *daddr,
 683                                               __be32 spi, u8 proto,
 684                                               unsigned short family)
 685 {
 686         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 687         struct xfrm_state *x;
 688         struct hlist_node *entry;
 689
 690         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 691                 if (x->props.family != family ||
 692                     x->id.spi       != spi ||
 693                     x->id.proto     != proto ||
 694                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 695                         continue;
 696
 697                 if ((mark & x->mark.m) != x->mark.v)
 698                         continue;
 699                 xfrm_state_hold(x);
 700                 return x;
 701         }
 702
 703         return NULL;
 704 }
 705
 706 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 707                                                      const xfrm_address_t *daddr,
 708                                                      const xfrm_address_t *saddr,
 709                                                      u8 proto, unsigned short family)
 710 {
 711         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 712         struct xfrm_state *x;
 713         struct hlist_node *entry;
 714
 715         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 716                 if (x->props.family != family ||
 717                     x->id.proto     != proto ||
 718                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 719                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 720                         continue;
 721
 722                 if ((mark & x->mark.m) != x->mark.v)
 723                         continue;
 724                 xfrm_state_hold(x);
 725                 return x;
 726         }
 727
 728         return NULL;
 729 }
 730
 731 static inline struct xfrm_state *
 732 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 733 {
 734         struct net *net = xs_net(x);
 735         u32 mark = x->mark.v & x->mark.m;
 736
 737         if (use_spi)
 738                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 739                                            x->id.spi, x->id.proto, family);
 740         else
 741                 return __xfrm_state_lookup_byaddr(net, mark,
 742                                                   &x->id.daddr,
 743                                                   &x->props.saddr,
 744                                                   x->id.proto, family);
 745 }
 746
 747 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 748 {
 749         if (have_hash_collision &&
 750             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 751             net->xfrm.state_num > net->xfrm.state_hmask)
 752                 schedule_work(&net->xfrm.state_hash_work);
 753 }
 754
 755 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 756                                const struct flowi *fl, unsigned short family,
 757                                struct xfrm_state **best, int *acq_in_progress,
 758                                int *error)
 759 {
 760         /* Resolution logic:
 761          * 1. There is a valid state with matching selector. Done.
 762          * 2. Valid state with inappropriate selector. Skip.
 763          *
 764          * Entering area of "sysdeps".
 765          *
 766          * 3. If state is not valid, selector is temporary, it selects
 767          *    only session which triggered previous resolution. Key
 768          *    manager will do something to install a state with proper
 769          *    selector.
 770          */
 771         if (x->km.state == XFRM_STATE_VALID) {
 772                 if ((x->sel.family &&
 773                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 774                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 775                         return;
 776
 777                 if (!*best ||
 778                     (*best)->km.dying > x->km.dying ||
 779                     ((*best)->km.dying == x->km.dying &&
 780                      (*best)->curlft.add_time < x->curlft.add_time))
 781                         *best = x;
 782         } else if (x->km.state == XFRM_STATE_ACQ) {
 783                 *acq_in_progress = 1;
 784         } else if (x->km.state == XFRM_STATE_ERROR ||
 785                    x->km.state == XFRM_STATE_EXPIRED) {
 786                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 787                     security_xfrm_state_pol_flow_match(x, pol, fl))
 788                         *error = -ESRCH;
 789         }
 790 }
 791
 792 struct xfrm_state *
 793 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 794                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 795                 struct xfrm_policy *pol, int *err,
 796                 unsigned short family)
 797 {
 798         static xfrm_address_t saddr_wildcard = { };
 799         struct net *net = xp_net(pol);
 800         unsigned int h, h_wildcard;
 801         struct hlist_node *entry;
 802         struct xfrm_state *x, *x0, *to_put;
 803         int acquire_in_progress = 0;
 804         int error = 0;
 805         struct xfrm_state *best = NULL;
 806         u32 mark = pol->mark.v & pol->mark.m;
 807         unsigned short encap_family = tmpl->encap_family;
 808
 809         to_put = NULL;
 810
 811         spin_lock_bh(&xfrm_state_lock);
 812         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 813         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 814                 if (x->props.family == encap_family &&
 815                     x->props.reqid == tmpl->reqid &&
 816                     (mark & x->mark.m) == x->mark.v &&
 817                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 818                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 819                     tmpl->mode == x->props.mode &&
 820                     tmpl->id.proto == x->id.proto &&
 821                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 822                         xfrm_state_look_at(pol, x, fl, encap_family,
 823                                            &best, &acquire_in_progress, &error);
 824         }
 825         if (best)
 826                 goto found;
 827
 828         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 829         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 830                 if (x->props.family == encap_family &&
 831                     x->props.reqid == tmpl->reqid &&
 832                     (mark & x->mark.m) == x->mark.v &&
 833                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 834                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 835                     tmpl->mode == x->props.mode &&
 836                     tmpl->id.proto == x->id.proto &&
 837                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 838                         xfrm_state_look_at(pol, x, fl, encap_family,
 839                                            &best, &acquire_in_progress, &error);
 840         }
 841
 842 found:
 843         x = best;
 844         if (!x && !error && !acquire_in_progress) {
 845                 if (tmpl->id.spi &&
 846                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 847                                               tmpl->id.proto, encap_family)) != NULL) {
 848                         to_put = x0;
 849                         error = -EEXIST;
 850                         goto out;
 851                 }
 852                 x = xfrm_state_alloc(net);
 853                 if (x == NULL) {
 854                         error = -ENOMEM;
 855                         goto out;
 856                 }
 857                 /* Initialize temporary state matching only
 858                  * to current session. */
 859                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 860                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 861
 862                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 863                 if (error) {
 864                         x->km.state = XFRM_STATE_DEAD;
 865                         to_put = x;
 866                         x = NULL;
 867                         goto out;
 868                 }
 869
 870                 if (km_query(x, tmpl, pol) == 0) {
 871                         x->km.state = XFRM_STATE_ACQ;
 872                         list_add(&x->km.all, &net->xfrm.state_all);
 873                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 874                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 875                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 876                         if (x->id.spi) {
 877                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 878                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 879                         }
 880                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 881                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 882                         net->xfrm.state_num++;
 883                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 884                 } else {
 885                         x->km.state = XFRM_STATE_DEAD;
 886                         to_put = x;
 887                         x = NULL;
 888                         error = -ESRCH;
 889                 }
 890         }
 891 out:
 892         if (x)
 893                 xfrm_state_hold(x);
 894         else
 895                 *err = acquire_in_progress ? -EAGAIN : error;
 896         spin_unlock_bh(&xfrm_state_lock);
 897         if (to_put)
 898                 xfrm_state_put(to_put);
 899         return x;
 900 }
 901
 902 struct xfrm_state *
 903 xfrm_stateonly_find(struct net *net, u32 mark,
 904                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 905                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 906 {
 907         unsigned int h;
 908         struct xfrm_state *rx = NULL, *x = NULL;
 909         struct hlist_node *entry;
 910
 911         spin_lock(&xfrm_state_lock);
 912         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 913         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 914                 if (x->props.family == family &&
 915                     x->props.reqid == reqid &&
 916                     (mark & x->mark.m) == x->mark.v &&
 917                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 918                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 919                     mode == x->props.mode &&
 920                     proto == x->id.proto &&
 921                     x->km.state == XFRM_STATE_VALID) {
 922                         rx = x;
 923                         break;
 924                 }
 925         }
 926
 927         if (rx)
 928                 xfrm_state_hold(rx);
 929         spin_unlock(&xfrm_state_lock);
 930
 931
 932         return rx;
 933 }
 934 EXPORT_SYMBOL(xfrm_stateonly_find);
 935
 936 static void __xfrm_state_insert(struct xfrm_state *x)
 937 {
 938         struct net *net = xs_net(x);
 939         unsigned int h;
 940
 941         list_add(&x->km.all, &net->xfrm.state_all);
 942
 943         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 944                           x->props.reqid, x->props.family);
 945         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 946
 947         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 948         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 949
 950         if (x->id.spi) {
 951                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 952                                   x->props.family);
 953
 954                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 955         }
 956
 957         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 958         if (x->replay_maxage)
 959                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 960
 961         wake_up(&net->xfrm.km_waitq);
 962
 963         net->xfrm.state_num++;
 964
 965         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 966 }
 967
 968 /* xfrm_state_lock is held */
 969 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 970 {
 971         struct net *net = xs_net(xnew);
 972         unsigned short family = xnew->props.family;
 973         u32 reqid = xnew->props.reqid;
 974         struct xfrm_state *x;
 975         struct hlist_node *entry;
 976         unsigned int h;
 977         u32 mark = xnew->mark.v & xnew->mark.m;
 978
 979         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 980         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 981                 if (x->props.family     == family &&
 982                     x->props.reqid      == reqid &&
 983                     (mark & x->mark.m) == x->mark.v &&
 984                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 985                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 986                         x->genid++;
 987         }
 988 }
 989
 990 void xfrm_state_insert(struct xfrm_state *x)
 991 {
 992         spin_lock_bh(&xfrm_state_lock);
 993         __xfrm_state_bump_genids(x);
 994         __xfrm_state_insert(x);
 995         spin_unlock_bh(&xfrm_state_lock);
 996 }
 997 EXPORT_SYMBOL(xfrm_state_insert);
 998
 999 /* xfrm_state_lock is held */
1000 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
1001                                           unsigned short family, u8 mode,
1002                                           u32 reqid, u8 proto,
1003                                           const xfrm_address_t *daddr,
1004                                           const xfrm_address_t *saddr, int create)
1005 {
1006         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1007         struct hlist_node *entry;
1008         struct xfrm_state *x;
1009         u32 mark = m->v & m->m;
1010
1011         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1012                 if (x->props.reqid  != reqid ||
1013                     x->props.mode   != mode ||
1014                     x->props.family != family ||
1015                     x->km.state     != XFRM_STATE_ACQ ||
1016                     x->id.spi       != 0 ||
1017                     x->id.proto     != proto ||
1018                     (mark & x->mark.m) != x->mark.v ||
1019                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1020                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1021                         continue;
1022
1023                 xfrm_state_hold(x);
1024                 return x;
1025         }
1026
1027         if (!create)
1028                 return NULL;
1029
1030         x = xfrm_state_alloc(net);
1031         if (likely(x)) {
1032                 switch (family) {
1033                 case AF_INET:
1034                         x->sel.daddr.a4 = daddr->a4;
1035                         x->sel.saddr.a4 = saddr->a4;
1036                         x->sel.prefixlen_d = 32;
1037                         x->sel.prefixlen_s = 32;
1038                         x->props.saddr.a4 = saddr->a4;
1039                         x->id.daddr.a4 = daddr->a4;
1040                         break;
1041
1042                 case AF_INET6:
1043                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1044                                        (struct in6_addr *)daddr);
1045                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1046                                        (struct in6_addr *)saddr);
1047                         x->sel.prefixlen_d = 128;
1048                         x->sel.prefixlen_s = 128;
1049                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1050                                        (struct in6_addr *)saddr);
1051                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1052                                        (struct in6_addr *)daddr);
1053                         break;
1054                 }
1055
1056                 x->km.state = XFRM_STATE_ACQ;
1057                 x->id.proto = proto;
1058                 x->props.family = family;
1059                 x->props.mode = mode;
1060                 x->props.reqid = reqid;
1061                 x->mark.v = m->v;
1062                 x->mark.m = m->m;
1063                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1064                 xfrm_state_hold(x);
1065                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1066                 list_add(&x->km.all, &net->xfrm.state_all);
1067                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1068                 h = xfrm_src_hash(net, daddr, saddr, family);
1069                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1070
1071                 net->xfrm.state_num++;
1072
1073                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1074         }
1075
1076         return x;
1077 }
1078
1079 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1080
1081 int xfrm_state_add(struct xfrm_state *x)
1082 {
1083         struct net *net = xs_net(x);
1084         struct xfrm_state *x1, *to_put;
1085         int family;
1086         int err;
1087         u32 mark = x->mark.v & x->mark.m;
1088         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1089
1090         family = x->props.family;
1091
1092         to_put = NULL;
1093
1094         spin_lock_bh(&xfrm_state_lock);
1095
1096         x1 = __xfrm_state_locate(x, use_spi, family);
1097         if (x1) {
1098                 to_put = x1;
1099                 x1 = NULL;
1100                 err = -EEXIST;
1101                 goto out;
1102         }
1103
1104         if (use_spi && x->km.seq) {
1105                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1106                 if (x1 && ((x1->id.proto != x->id.proto) ||
1107                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1108                         to_put = x1;
1109                         x1 = NULL;
1110                 }
1111         }
1112
1113         if (use_spi && !x1)
1114                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1115                                      x->props.reqid, x->id.proto,
1116                                      &x->id.daddr, &x->props.saddr, 0);
1117
1118         __xfrm_state_bump_genids(x);
1119         __xfrm_state_insert(x);
1120         err = 0;
1121
1122 out:
1123         spin_unlock_bh(&xfrm_state_lock);
1124
1125         if (x1) {
1126                 xfrm_state_delete(x1);
1127                 xfrm_state_put(x1);
1128         }
1129
1130         if (to_put)
1131                 xfrm_state_put(to_put);
1132
1133         return err;
1134 }
1135 EXPORT_SYMBOL(xfrm_state_add);
1136
1137 #ifdef CONFIG_XFRM_MIGRATE
1138 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1139 {
1140         struct net *net = xs_net(orig);
1141         int err = -ENOMEM;
1142         struct xfrm_state *x = xfrm_state_alloc(net);
1143         if (!x)
1144                 goto out;
1145
1146         memcpy(&x->id, &orig->id, sizeof(x->id));
1147         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1148         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1149         x->props.mode = orig->props.mode;
1150         x->props.replay_window = orig->props.replay_window;
1151         x->props.reqid = orig->props.reqid;
1152         x->props.family = orig->props.family;
1153         x->props.saddr = orig->props.saddr;
1154
1155         if (orig->aalg) {
1156                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1157                 if (!x->aalg)
1158                         goto error;
1159         }
1160         x->props.aalgo = orig->props.aalgo;
1161
1162         if (orig->ealg) {
1163                 x->ealg = xfrm_algo_clone(orig->ealg);
1164                 if (!x->ealg)
1165                         goto error;
1166         }
1167         x->props.ealgo = orig->props.ealgo;
1168
1169         if (orig->calg) {
1170                 x->calg = xfrm_algo_clone(orig->calg);
1171                 if (!x->calg)
1172                         goto error;
1173         }
1174         x->props.calgo = orig->props.calgo;
1175
1176         if (orig->encap) {
1177                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1178                 if (!x->encap)
1179                         goto error;
1180         }
1181
1182         if (orig->coaddr) {
1183                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1184                                     GFP_KERNEL);
1185                 if (!x->coaddr)
1186                         goto error;
1187         }
1188
1189         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1190
1191         err = xfrm_init_state(x);
1192         if (err)
1193                 goto error;
1194
1195         x->props.flags = orig->props.flags;
1196
1197         x->curlft.add_time = orig->curlft.add_time;
1198         x->km.state = orig->km.state;
1199         x->km.seq = orig->km.seq;
1200
1201         return x;
1202
1203  error:
1204         xfrm_state_put(x);
1205 out:
1206         if (errp)
1207                 *errp = err;
1208         return NULL;
1209 }
1210
1211 /* xfrm_state_lock is held */
1212 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1213 {
1214         unsigned int h;
1215         struct xfrm_state *x;
1216         struct hlist_node *entry;
1217
1218         if (m->reqid) {
1219                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1220                                   m->reqid, m->old_family);
1221                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1222                         if (x->props.mode != m->mode ||
1223                             x->id.proto != m->proto)
1224                                 continue;
1225                         if (m->reqid && x->props.reqid != m->reqid)
1226                                 continue;
1227                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1228                                           m->old_family) ||
1229                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1230                                           m->old_family))
1231                                 continue;
1232                         xfrm_state_hold(x);
1233                         return x;
1234                 }
1235         } else {
1236                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1237                                   m->old_family);
1238                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1239                         if (x->props.mode != m->mode ||
1240                             x->id.proto != m->proto)
1241                                 continue;
1242                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1243                                           m->old_family) ||
1244                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1245                                           m->old_family))
1246                                 continue;
1247                         xfrm_state_hold(x);
1248                         return x;
1249                 }
1250         }
1251
1252         return NULL;
1253 }
1254 EXPORT_SYMBOL(xfrm_migrate_state_find);
1255
1256 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1257                                        struct xfrm_migrate *m)
1258 {
1259         struct xfrm_state *xc;
1260         int err;
1261
1262         xc = xfrm_state_clone(x, &err);
1263         if (!xc)
1264                 return NULL;
1265
1266         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1267         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1268
1269         /* add state */
1270         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1271                 /* a care is needed when the destination address of the
1272                    state is to be updated as it is a part of triplet */
1273                 xfrm_state_insert(xc);
1274         } else {
1275                 if ((err = xfrm_state_add(xc)) < 0)
1276                         goto error;
1277         }
1278
1279         return xc;
1280 error:
1281         xfrm_state_put(xc);
1282         return NULL;
1283 }
1284 EXPORT_SYMBOL(xfrm_state_migrate);
1285 #endif
1286
1287 int xfrm_state_update(struct xfrm_state *x)
1288 {
1289         struct xfrm_state *x1, *to_put;
1290         int err;
1291         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1292
1293         to_put = NULL;
1294
1295         spin_lock_bh(&xfrm_state_lock);
1296         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1297
1298         err = -ESRCH;
1299         if (!x1)
1300                 goto out;
1301
1302         if (xfrm_state_kern(x1)) {
1303                 to_put = x1;
1304                 err = -EEXIST;
1305                 goto out;
1306         }
1307
1308         if (x1->km.state == XFRM_STATE_ACQ) {
1309                 __xfrm_state_insert(x);
1310                 x = NULL;
1311         }
1312         err = 0;
1313
1314 out:
1315         spin_unlock_bh(&xfrm_state_lock);
1316
1317         if (to_put)
1318                 xfrm_state_put(to_put);
1319
1320         if (err)
1321                 return err;
1322
1323         if (!x) {
1324                 xfrm_state_delete(x1);
1325                 xfrm_state_put(x1);
1326                 return 0;
1327         }
1328
1329         err = -EINVAL;
1330         spin_lock_bh(&x1->lock);
1331         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1332                 if (x->encap && x1->encap)
1333                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1334                 if (x->coaddr && x1->coaddr) {
1335                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1336                 }
1337                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1338                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1339                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1340                 x1->km.dying = 0;
1341
1342                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1343                 if (x1->curlft.use_time)
1344                         xfrm_state_check_expire(x1);
1345
1346                 err = 0;
1347         }
1348         spin_unlock_bh(&x1->lock);
1349
1350         xfrm_state_put(x1);
1351
1352         return err;
1353 }
1354 EXPORT_SYMBOL(xfrm_state_update);
1355
1356 int xfrm_state_check_expire(struct xfrm_state *x)
1357 {
1358         if (!x->curlft.use_time)
1359                 x->curlft.use_time = get_seconds();
1360
1361         if (x->km.state != XFRM_STATE_VALID)
1362                 return -EINVAL;
1363
1364         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1365             x->curlft.packets >= x->lft.hard_packet_limit) {
1366                 x->km.state = XFRM_STATE_EXPIRED;
1367                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1368                 return -EINVAL;
1369         }
1370
1371         if (!x->km.dying &&
1372             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1373              x->curlft.packets >= x->lft.soft_packet_limit)) {
1374                 x->km.dying = 1;
1375                 km_state_expired(x, 0, 0);
1376         }
1377         return 0;
1378 }
1379 EXPORT_SYMBOL(xfrm_state_check_expire);
1380
1381 struct xfrm_state *
1382 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1383                   u8 proto, unsigned short family)
1384 {
1385         struct xfrm_state *x;
1386
1387         spin_lock_bh(&xfrm_state_lock);
1388         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1389         spin_unlock_bh(&xfrm_state_lock);
1390         return x;
1391 }
1392 EXPORT_SYMBOL(xfrm_state_lookup);
1393
1394 struct xfrm_state *
1395 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1396                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1397                          u8 proto, unsigned short family)
1398 {
1399         struct xfrm_state *x;
1400
1401         spin_lock_bh(&xfrm_state_lock);
1402         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1403         spin_unlock_bh(&xfrm_state_lock);
1404         return x;
1405 }
1406 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1407
1408 struct xfrm_state *
1409 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1410               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1411               int create, unsigned short family)
1412 {
1413         struct xfrm_state *x;
1414
1415         spin_lock_bh(&xfrm_state_lock);
1416         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1417         spin_unlock_bh(&xfrm_state_lock);
1418
1419         return x;
1420 }
1421 EXPORT_SYMBOL(xfrm_find_acq);
1422
1423 #ifdef CONFIG_XFRM_SUB_POLICY
1424 int
1425 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1426                unsigned short family)
1427 {
1428         int err = 0;
1429         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1430         if (!afinfo)
1431                 return -EAFNOSUPPORT;
1432
1433         spin_lock_bh(&xfrm_state_lock);
1434         if (afinfo->tmpl_sort)
1435                 err = afinfo->tmpl_sort(dst, src, n);
1436         spin_unlock_bh(&xfrm_state_lock);
1437         xfrm_state_put_afinfo(afinfo);
1438         return err;
1439 }
1440 EXPORT_SYMBOL(xfrm_tmpl_sort);
1441
1442 int
1443 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1444                 unsigned short family)
1445 {
1446         int err = 0;
1447         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1448         if (!afinfo)
1449                 return -EAFNOSUPPORT;
1450
1451         spin_lock_bh(&xfrm_state_lock);
1452         if (afinfo->state_sort)
1453                 err = afinfo->state_sort(dst, src, n);
1454         spin_unlock_bh(&xfrm_state_lock);
1455         xfrm_state_put_afinfo(afinfo);
1456         return err;
1457 }
1458 EXPORT_SYMBOL(xfrm_state_sort);
1459 #endif
1460
1461 /* Silly enough, but I'm lazy to build resolution list */
1462
1463 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1464 {
1465         int i;
1466
1467         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1468                 struct hlist_node *entry;
1469                 struct xfrm_state *x;
1470
1471                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1472                         if (x->km.seq == seq &&
1473                             (mark & x->mark.m) == x->mark.v &&
1474                             x->km.state == XFRM_STATE_ACQ) {
1475                                 xfrm_state_hold(x);
1476                                 return x;
1477                         }
1478                 }
1479         }
1480         return NULL;
1481 }
1482
1483 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1484 {
1485         struct xfrm_state *x;
1486
1487         spin_lock_bh(&xfrm_state_lock);
1488         x = __xfrm_find_acq_byseq(net, mark, seq);
1489         spin_unlock_bh(&xfrm_state_lock);
1490         return x;
1491 }
1492 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1493
1494 u32 xfrm_get_acqseq(void)
1495 {
1496         u32 res;
1497         static atomic_t acqseq;
1498
1499         do {
1500                 res = atomic_inc_return(&acqseq);
1501         } while (!res);
1502
1503         return res;
1504 }
1505 EXPORT_SYMBOL(xfrm_get_acqseq);
1506
1507 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1508 {
1509         struct net *net = xs_net(x);
1510         unsigned int h;
1511         struct xfrm_state *x0;
1512         int err = -ENOENT;
1513         __be32 minspi = htonl(low);
1514         __be32 maxspi = htonl(high);
1515         u32 mark = x->mark.v & x->mark.m;
1516
1517         spin_lock_bh(&x->lock);
1518         if (x->km.state == XFRM_STATE_DEAD)
1519                 goto unlock;
1520
1521         err = 0;
1522         if (x->id.spi)
1523                 goto unlock;
1524
1525         err = -ENOENT;
1526
1527         if (minspi == maxspi) {
1528                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1529                 if (x0) {
1530                         xfrm_state_put(x0);
1531                         goto unlock;
1532                 }
1533                 x->id.spi = minspi;
1534         } else {
1535                 u32 spi = 0;
1536                 for (h=0; h<high-low+1; h++) {
1537                         spi = low + net_random()%(high-low+1);
1538                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1539                         if (x0 == NULL) {
1540                                 x->id.spi = htonl(spi);
1541                                 break;
1542                         }
1543                         xfrm_state_put(x0);
1544                 }
1545         }
1546         if (x->id.spi) {
1547                 spin_lock_bh(&xfrm_state_lock);
1548                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1549                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1550                 spin_unlock_bh(&xfrm_state_lock);
1551
1552                 err = 0;
1553         }
1554
1555 unlock:
1556         spin_unlock_bh(&x->lock);
1557
1558         return err;
1559 }
1560 EXPORT_SYMBOL(xfrm_alloc_spi);
1561
1562 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1563                     int (*func)(struct xfrm_state *, int, void*),
1564                     void *data)
1565 {
1566         struct xfrm_state *state;
1567         struct xfrm_state_walk *x;
1568         int err = 0;
1569
1570         if (walk->seq != 0 && list_empty(&walk->all))
1571                 return 0;
1572
1573         spin_lock_bh(&xfrm_state_lock);
1574         if (list_empty(&walk->all))
1575                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1576         else
1577                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1578         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1579                 if (x->state == XFRM_STATE_DEAD)
1580                         continue;
1581                 state = container_of(x, struct xfrm_state, km);
1582                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1583                         continue;
1584                 err = func(state, walk->seq, data);
1585                 if (err) {
1586                         list_move_tail(&walk->all, &x->all);
1587                         goto out;
1588                 }
1589                 walk->seq++;
1590         }
1591         if (walk->seq == 0) {
1592                 err = -ENOENT;
1593                 goto out;
1594         }
1595         list_del_init(&walk->all);
1596 out:
1597         spin_unlock_bh(&xfrm_state_lock);
1598         return err;
1599 }
1600 EXPORT_SYMBOL(xfrm_state_walk);
1601
1602 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1603 {
1604         INIT_LIST_HEAD(&walk->all);
1605         walk->proto = proto;
1606         walk->state = XFRM_STATE_DEAD;
1607         walk->seq = 0;
1608 }
1609 EXPORT_SYMBOL(xfrm_state_walk_init);
1610
1611 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1612 {
1613         if (list_empty(&walk->all))
1614                 return;
1615
1616         spin_lock_bh(&xfrm_state_lock);
1617         list_del(&walk->all);
1618         spin_unlock_bh(&xfrm_state_lock);
1619 }
1620 EXPORT_SYMBOL(xfrm_state_walk_done);
1621
1622
1623 void xfrm_replay_notify(struct xfrm_state *x, int event)
1624 {
1625         struct km_event c;
1626         /* we send notify messages in case
1627          *  1. we updated on of the sequence numbers, and the seqno difference
1628          *     is at least x->replay_maxdiff, in this case we also update the
1629          *     timeout of our timer function
1630          *  2. if x->replay_maxage has elapsed since last update,
1631          *     and there were changes
1632          *
1633          *  The state structure must be locked!
1634          */
1635
1636         switch (event) {
1637         case XFRM_REPLAY_UPDATE:
1638                 if (x->replay_maxdiff &&
1639                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1640                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1641                         if (x->xflags & XFRM_TIME_DEFER)
1642                                 event = XFRM_REPLAY_TIMEOUT;
1643                         else
1644                                 return;
1645                 }
1646
1647                 break;
1648
1649         case XFRM_REPLAY_TIMEOUT:
1650                 if ((x->replay.seq == x->preplay.seq) &&
1651                     (x->replay.bitmap == x->preplay.bitmap) &&
1652                     (x->replay.oseq == x->preplay.oseq)) {
1653                         x->xflags |= XFRM_TIME_DEFER;
1654                         return;
1655                 }
1656
1657                 break;
1658         }
1659
1660         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1661         c.event = XFRM_MSG_NEWAE;
1662         c.data.aevent = event;
1663         km_state_notify(x, &c);
1664
1665         if (x->replay_maxage &&
1666             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1667                 x->xflags &= ~XFRM_TIME_DEFER;
1668 }
1669
1670 static void xfrm_replay_timer_handler(unsigned long data)
1671 {
1672         struct xfrm_state *x = (struct xfrm_state*)data;
1673
1674         spin_lock(&x->lock);
1675
1676         if (x->km.state == XFRM_STATE_VALID) {
1677                 if (xfrm_aevent_is_on(xs_net(x)))
1678                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1679                 else
1680                         x->xflags |= XFRM_TIME_DEFER;
1681         }
1682
1683         spin_unlock(&x->lock);
1684 }
1685
1686 int xfrm_replay_check(struct xfrm_state *x,
1687                       struct sk_buff *skb, __be32 net_seq)
1688 {
1689         u32 diff;
1690         u32 seq = ntohl(net_seq);
1691
1692         if (unlikely(seq == 0))
1693                 goto err;
1694
1695         if (likely(seq > x->replay.seq))
1696                 return 0;
1697
1698         diff = x->replay.seq - seq;
1699         if (diff >= min_t(unsigned int, x->props.replay_window,
1700                           sizeof(x->replay.bitmap) * 8)) {
1701                 x->stats.replay_window++;
1702                 goto err;
1703         }
1704
1705         if (x->replay.bitmap & (1U << diff)) {
1706                 x->stats.replay++;
1707                 goto err;
1708         }
1709         return 0;
1710
1711 err:
1712         xfrm_audit_state_replay(x, skb, net_seq);
1713         return -EINVAL;
1714 }
1715
1716 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1717 {
1718         u32 diff;
1719         u32 seq = ntohl(net_seq);
1720
1721         if (seq > x->replay.seq) {
1722                 diff = seq - x->replay.seq;
1723                 if (diff < x->props.replay_window)
1724                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1725                 else
1726                         x->replay.bitmap = 1;
1727                 x->replay.seq = seq;
1728         } else {
1729                 diff = x->replay.seq - seq;
1730                 x->replay.bitmap |= (1U << diff);
1731         }
1732
1733         if (xfrm_aevent_is_on(xs_net(x)))
1734                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1735 }
1736
1737 static LIST_HEAD(xfrm_km_list);
1738 static DEFINE_RWLOCK(xfrm_km_lock);
1739
1740 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1741 {
1742         struct xfrm_mgr *km;
1743
1744         read_lock(&xfrm_km_lock);
1745         list_for_each_entry(km, &xfrm_km_list, list)
1746                 if (km->notify_policy)
1747                         km->notify_policy(xp, dir, c);
1748         read_unlock(&xfrm_km_lock);
1749 }
1750
1751 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1752 {
1753         struct xfrm_mgr *km;
1754         read_lock(&xfrm_km_lock);
1755         list_for_each_entry(km, &xfrm_km_list, list)
1756                 if (km->notify)
1757                         km->notify(x, c);
1758         read_unlock(&xfrm_km_lock);
1759 }
1760
1761 EXPORT_SYMBOL(km_policy_notify);
1762 EXPORT_SYMBOL(km_state_notify);
1763
1764 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1765 {
1766         struct net *net = xs_net(x);
1767         struct km_event c;
1768
1769         c.data.hard = hard;
1770         c.pid = pid;
1771         c.event = XFRM_MSG_EXPIRE;
1772         km_state_notify(x, &c);
1773
1774         if (hard)
1775                 wake_up(&net->xfrm.km_waitq);
1776 }
1777
1778 EXPORT_SYMBOL(km_state_expired);
1779 /*
1780  * We send to all registered managers regardless of failure
1781  * We are happy with one success
1782 */
1783 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1784 {
1785         int err = -EINVAL, acqret;
1786         struct xfrm_mgr *km;
1787
1788         read_lock(&xfrm_km_lock);
1789         list_for_each_entry(km, &xfrm_km_list, list) {
1790                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1791                 if (!acqret)
1792                         err = acqret;
1793         }
1794         read_unlock(&xfrm_km_lock);
1795         return err;
1796 }
1797 EXPORT_SYMBOL(km_query);
1798
1799 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1800 {
1801         int err = -EINVAL;
1802         struct xfrm_mgr *km;
1803
1804         read_lock(&xfrm_km_lock);
1805         list_for_each_entry(km, &xfrm_km_list, list) {
1806                 if (km->new_mapping)
1807                         err = km->new_mapping(x, ipaddr, sport);
1808                 if (!err)
1809                         break;
1810         }
1811         read_unlock(&xfrm_km_lock);
1812         return err;
1813 }
1814 EXPORT_SYMBOL(km_new_mapping);
1815
1816 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1817 {
1818         struct net *net = xp_net(pol);
1819         struct km_event c;
1820
1821         c.data.hard = hard;
1822         c.pid = pid;
1823         c.event = XFRM_MSG_POLEXPIRE;
1824         km_policy_notify(pol, dir, &c);
1825
1826         if (hard)
1827                 wake_up(&net->xfrm.km_waitq);
1828 }
1829 EXPORT_SYMBOL(km_policy_expired);
1830
1831 #ifdef CONFIG_XFRM_MIGRATE
1832 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1833                const struct xfrm_migrate *m, int num_migrate,
1834                const struct xfrm_kmaddress *k)
1835 {
1836         int err = -EINVAL;
1837         int ret;
1838         struct xfrm_mgr *km;
1839
1840         read_lock(&xfrm_km_lock);
1841         list_for_each_entry(km, &xfrm_km_list, list) {
1842                 if (km->migrate) {
1843                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1844                         if (!ret)
1845                                 err = ret;
1846                 }
1847         }
1848         read_unlock(&xfrm_km_lock);
1849         return err;
1850 }
1851 EXPORT_SYMBOL(km_migrate);
1852 #endif
1853
1854 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1855 {
1856         int err = -EINVAL;
1857         int ret;
1858         struct xfrm_mgr *km;
1859
1860         read_lock(&xfrm_km_lock);
1861         list_for_each_entry(km, &xfrm_km_list, list) {
1862                 if (km->report) {
1863                         ret = km->report(net, proto, sel, addr);
1864                         if (!ret)
1865                                 err = ret;
1866                 }
1867         }
1868         read_unlock(&xfrm_km_lock);
1869         return err;
1870 }
1871 EXPORT_SYMBOL(km_report);
1872
1873 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1874 {
1875         int err;
1876         u8 *data;
1877         struct xfrm_mgr *km;
1878         struct xfrm_policy *pol = NULL;
1879
1880         if (optlen <= 0 || optlen > PAGE_SIZE)
1881                 return -EMSGSIZE;
1882
1883         data = kmalloc(optlen, GFP_KERNEL);
1884         if (!data)
1885                 return -ENOMEM;
1886
1887         err = -EFAULT;
1888         if (copy_from_user(data, optval, optlen))
1889                 goto out;
1890
1891         err = -EINVAL;
1892         read_lock(&xfrm_km_lock);
1893         list_for_each_entry(km, &xfrm_km_list, list) {
1894                 pol = km->compile_policy(sk, optname, data,
1895                                          optlen, &err);
1896                 if (err >= 0)
1897                         break;
1898         }
1899         read_unlock(&xfrm_km_lock);
1900
1901         if (err >= 0) {
1902                 xfrm_sk_policy_insert(sk, err, pol);
1903                 xfrm_pol_put(pol);
1904                 err = 0;
1905         }
1906
1907 out:
1908         kfree(data);
1909         return err;
1910 }
1911 EXPORT_SYMBOL(xfrm_user_policy);
1912
1913 int xfrm_register_km(struct xfrm_mgr *km)
1914 {
1915         write_lock_bh(&xfrm_km_lock);
1916         list_add_tail(&km->list, &xfrm_km_list);
1917         write_unlock_bh(&xfrm_km_lock);
1918         return 0;
1919 }
1920 EXPORT_SYMBOL(xfrm_register_km);
1921
1922 int xfrm_unregister_km(struct xfrm_mgr *km)
1923 {
1924         write_lock_bh(&xfrm_km_lock);
1925         list_del(&km->list);
1926         write_unlock_bh(&xfrm_km_lock);
1927         return 0;
1928 }
1929 EXPORT_SYMBOL(xfrm_unregister_km);
1930
1931 int xfrm_state_register_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 (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1940                 err = -ENOBUFS;
1941         else
1942                 xfrm_state_afinfo[afinfo->family] = afinfo;
1943         write_unlock_bh(&xfrm_state_afinfo_lock);
1944         return err;
1945 }
1946 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1947
1948 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1949 {
1950         int err = 0;
1951         if (unlikely(afinfo == NULL))
1952                 return -EINVAL;
1953         if (unlikely(afinfo->family >= NPROTO))
1954                 return -EAFNOSUPPORT;
1955         write_lock_bh(&xfrm_state_afinfo_lock);
1956         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1957                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1958                         err = -EINVAL;
1959                 else
1960                         xfrm_state_afinfo[afinfo->family] = NULL;
1961         }
1962         write_unlock_bh(&xfrm_state_afinfo_lock);
1963         return err;
1964 }
1965 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1966
1967 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1968 {
1969         struct xfrm_state_afinfo *afinfo;
1970         if (unlikely(family >= NPROTO))
1971                 return NULL;
1972         read_lock(&xfrm_state_afinfo_lock);
1973         afinfo = xfrm_state_afinfo[family];
1974         if (unlikely(!afinfo))
1975                 read_unlock(&xfrm_state_afinfo_lock);
1976         return afinfo;
1977 }
1978
1979 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1980         __releases(xfrm_state_afinfo_lock)
1981 {
1982         read_unlock(&xfrm_state_afinfo_lock);
1983 }
1984
1985 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1986 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1987 {
1988         if (x->tunnel) {
1989                 struct xfrm_state *t = x->tunnel;
1990
1991                 if (atomic_read(&t->tunnel_users) == 2)
1992                         xfrm_state_delete(t);
1993                 atomic_dec(&t->tunnel_users);
1994                 xfrm_state_put(t);
1995                 x->tunnel = NULL;
1996         }
1997 }
1998 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1999
2000 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2001 {
2002         int res;
2003
2004         spin_lock_bh(&x->lock);
2005         if (x->km.state == XFRM_STATE_VALID &&
2006             x->type && x->type->get_mtu)
2007                 res = x->type->get_mtu(x, mtu);
2008         else
2009                 res = mtu - x->props.header_len;
2010         spin_unlock_bh(&x->lock);
2011         return res;
2012 }
2013
2014 int xfrm_init_state(struct xfrm_state *x)
2015 {
2016         struct xfrm_state_afinfo *afinfo;
2017         struct xfrm_mode *inner_mode;
2018         int family = x->props.family;
2019         int err;
2020
2021         err = -EAFNOSUPPORT;
2022         afinfo = xfrm_state_get_afinfo(family);
2023         if (!afinfo)
2024                 goto error;
2025
2026         err = 0;
2027         if (afinfo->init_flags)
2028                 err = afinfo->init_flags(x);
2029
2030         xfrm_state_put_afinfo(afinfo);
2031
2032         if (err)
2033                 goto error;
2034
2035         err = -EPROTONOSUPPORT;
2036
2037         if (x->sel.family != AF_UNSPEC) {
2038                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2039                 if (inner_mode == NULL)
2040                         goto error;
2041
2042                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2043                     family != x->sel.family) {
2044                         xfrm_put_mode(inner_mode);
2045                         goto error;
2046                 }
2047
2048                 x->inner_mode = inner_mode;
2049         } else {
2050                 struct xfrm_mode *inner_mode_iaf;
2051                 int iafamily = AF_INET;
2052
2053                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2054                 if (inner_mode == NULL)
2055                         goto error;
2056
2057                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2058                         xfrm_put_mode(inner_mode);
2059                         goto error;
2060                 }
2061                 x->inner_mode = inner_mode;
2062
2063                 if (x->props.family == AF_INET)
2064                         iafamily = AF_INET6;
2065
2066                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2067                 if (inner_mode_iaf) {
2068                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2069                                 x->inner_mode_iaf = inner_mode_iaf;
2070                         else
2071                                 xfrm_put_mode(inner_mode_iaf);
2072                 }
2073         }
2074
2075         x->type = xfrm_get_type(x->id.proto, family);
2076         if (x->type == NULL)
2077                 goto error;
2078
2079         err = x->type->init_state(x);
2080         if (err)
2081                 goto error;
2082
2083         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2084         if (x->outer_mode == NULL)
2085                 goto error;
2086
2087         x->km.state = XFRM_STATE_VALID;
2088
2089 error:
2090         return err;
2091 }
2092
2093 EXPORT_SYMBOL(xfrm_init_state);
2094
2095 int __net_init xfrm_state_init(struct net *net)
2096 {
2097         unsigned int sz;
2098
2099         INIT_LIST_HEAD(&net->xfrm.state_all);
2100
2101         sz = sizeof(struct hlist_head) * 8;
2102
2103         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2104         if (!net->xfrm.state_bydst)
2105                 goto out_bydst;
2106         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2107         if (!net->xfrm.state_bysrc)
2108                 goto out_bysrc;
2109         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2110         if (!net->xfrm.state_byspi)
2111                 goto out_byspi;
2112         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2113
2114         net->xfrm.state_num = 0;
2115         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2116         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2117         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2118         init_waitqueue_head(&net->xfrm.km_waitq);
2119         return 0;
2120
2121 out_byspi:
2122         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2123 out_bysrc:
2124         xfrm_hash_free(net->xfrm.state_bydst, sz);
2125 out_bydst:
2126         return -ENOMEM;
2127 }
2128
2129 void xfrm_state_fini(struct net *net)
2130 {
2131         struct xfrm_audit audit_info;
2132         unsigned int sz;
2133
2134         flush_work(&net->xfrm.state_hash_work);
2135         audit_info.loginuid = -1;
2136         audit_info.sessionid = -1;
2137         audit_info.secid = 0;
2138         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2139         flush_work(&net->xfrm.state_gc_work);
2140
2141         WARN_ON(!list_empty(&net->xfrm.state_all));
2142
2143         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2144         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2145         xfrm_hash_free(net->xfrm.state_byspi, sz);
2146         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2147         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2148         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2149         xfrm_hash_free(net->xfrm.state_bydst, sz);
2150 }
2151
2152 #ifdef CONFIG_AUDITSYSCALL
2153 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2154                                      struct audit_buffer *audit_buf)
2155 {
2156         struct xfrm_sec_ctx *ctx = x->security;
2157         u32 spi = ntohl(x->id.spi);
2158
2159         if (ctx)
2160                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2161                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2162
2163         switch(x->props.family) {
2164         case AF_INET:
2165                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2166                                  &x->props.saddr.a4, &x->id.daddr.a4);
2167                 break;
2168         case AF_INET6:
2169                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2170                                  x->props.saddr.a6, x->id.daddr.a6);
2171                 break;
2172         }
2173
2174         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2175 }
2176
2177 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2178                                       struct audit_buffer *audit_buf)
2179 {
2180         struct iphdr *iph4;
2181         struct ipv6hdr *iph6;
2182
2183         switch (family) {
2184         case AF_INET:
2185                 iph4 = ip_hdr(skb);
2186                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2187                                  &iph4->saddr, &iph4->daddr);
2188                 break;
2189         case AF_INET6:
2190                 iph6 = ipv6_hdr(skb);
2191                 audit_log_format(audit_buf,
2192                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2193                                  &iph6->saddr,&iph6->daddr,
2194                                  iph6->flow_lbl[0] & 0x0f,
2195                                  iph6->flow_lbl[1],
2196                                  iph6->flow_lbl[2]);
2197                 break;
2198         }
2199 }
2200
2201 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2202                           uid_t auid, u32 sessionid, u32 secid)
2203 {
2204         struct audit_buffer *audit_buf;
2205
2206         audit_buf = xfrm_audit_start("SAD-add");
2207         if (audit_buf == NULL)
2208                 return;
2209         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2210         xfrm_audit_helper_sainfo(x, audit_buf);
2211         audit_log_format(audit_buf, " res=%u", result);
2212         audit_log_end(audit_buf);
2213 }
2214 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2215
2216 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2217                              uid_t auid, u32 sessionid, u32 secid)
2218 {
2219         struct audit_buffer *audit_buf;
2220
2221         audit_buf = xfrm_audit_start("SAD-delete");
2222         if (audit_buf == NULL)
2223                 return;
2224         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2225         xfrm_audit_helper_sainfo(x, audit_buf);
2226         audit_log_format(audit_buf, " res=%u", result);
2227         audit_log_end(audit_buf);
2228 }
2229 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2230
2231 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2232                                       struct sk_buff *skb)
2233 {
2234         struct audit_buffer *audit_buf;
2235         u32 spi;
2236
2237         audit_buf = xfrm_audit_start("SA-replay-overflow");
2238         if (audit_buf == NULL)
2239                 return;
2240         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2241         /* don't record the sequence number because it's inherent in this kind
2242          * of audit message */
2243         spi = ntohl(x->id.spi);
2244         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2245         audit_log_end(audit_buf);
2246 }
2247 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2248
2249 static void xfrm_audit_state_replay(struct xfrm_state *x,
2250                              struct sk_buff *skb, __be32 net_seq)
2251 {
2252         struct audit_buffer *audit_buf;
2253         u32 spi;
2254
2255         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2256         if (audit_buf == NULL)
2257                 return;
2258         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2259         spi = ntohl(x->id.spi);
2260         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2261                          spi, spi, ntohl(net_seq));
2262         audit_log_end(audit_buf);
2263 }
2264
2265 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2266 {
2267         struct audit_buffer *audit_buf;
2268
2269         audit_buf = xfrm_audit_start("SA-notfound");
2270         if (audit_buf == NULL)
2271                 return;
2272         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2273         audit_log_end(audit_buf);
2274 }
2275 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2276
2277 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2278                                __be32 net_spi, __be32 net_seq)
2279 {
2280         struct audit_buffer *audit_buf;
2281         u32 spi;
2282
2283         audit_buf = xfrm_audit_start("SA-notfound");
2284         if (audit_buf == NULL)
2285                 return;
2286         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2287         spi = ntohl(net_spi);
2288         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2289                          spi, spi, ntohl(net_seq));
2290         audit_log_end(audit_buf);
2291 }
2292 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2293
2294 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2295                               struct sk_buff *skb, u8 proto)
2296 {
2297         struct audit_buffer *audit_buf;
2298         __be32 net_spi;
2299         __be32 net_seq;
2300
2301         audit_buf = xfrm_audit_start("SA-icv-failure");
2302         if (audit_buf == NULL)
2303                 return;
2304         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2305         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2306                 u32 spi = ntohl(net_spi);
2307                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2308                                  spi, spi, ntohl(net_seq));
2309         }
2310         audit_log_end(audit_buf);
2311 }
2312 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2313 #endif /* CONFIG_AUDITSYSCALL */