net/xfrm/xfrm_state.c

   1 /*
   2  * xfrm_state.c
   3  *
   4  * Changes:
   5  *      Mitsuru KANDA @USAGI
   6  *      Kazunori MIYAZAWA @USAGI
   7  *      Kunihiro Ishiguro <[email protected]>
   8  *              IPv6 support
   9  *      YOSHIFUJI Hideaki @USAGI
  10  *              Split up af-specific functions
  11  *      Derek Atkins <[email protected]>
  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                     xfrm_address_t *daddr, 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                                xfrm_address_t *daddr, xfrm_address_t *saddr,
 758                                struct xfrm_state **best, int *acq_in_progress,
 759                                int *error)
 760 {
 761         /* Resolution logic:
 762          * 1. There is a valid state with matching selector. Done.
 763          * 2. Valid state with inappropriate selector. Skip.
 764          *
 765          * Entering area of "sysdeps".
 766          *
 767          * 3. If state is not valid, selector is temporary, it selects
 768          *    only session which triggered previous resolution. Key
 769          *    manager will do something to install a state with proper
 770          *    selector.
 771          */
 772         if (x->km.state == XFRM_STATE_VALID) {
 773                 if ((x->sel.family &&
 774                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 775                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 776                         return;
 777
 778                 if (!*best ||
 779                     (*best)->km.dying > x->km.dying ||
 780                     ((*best)->km.dying == x->km.dying &&
 781                      (*best)->curlft.add_time < x->curlft.add_time))
 782                         *best = x;
 783         } else if (x->km.state == XFRM_STATE_ACQ) {
 784                 *acq_in_progress = 1;
 785         } else if (x->km.state == XFRM_STATE_ERROR ||
 786                    x->km.state == XFRM_STATE_EXPIRED) {
 787                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 788                     security_xfrm_state_pol_flow_match(x, pol, fl))
 789                         *error = -ESRCH;
 790         }
 791 }
 792
 793 struct xfrm_state *
 794 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
 795                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 796                 struct xfrm_policy *pol, int *err,
 797                 unsigned short family)
 798 {
 799         static xfrm_address_t saddr_wildcard = { };
 800         struct net *net = xp_net(pol);
 801         unsigned int h, h_wildcard;
 802         struct hlist_node *entry;
 803         struct xfrm_state *x, *x0, *to_put;
 804         int acquire_in_progress = 0;
 805         int error = 0;
 806         struct xfrm_state *best = NULL;
 807         u32 mark = pol->mark.v & pol->mark.m;
 808         unsigned short encap_family = tmpl->encap_family;
 809
 810         to_put = NULL;
 811
 812         spin_lock_bh(&xfrm_state_lock);
 813         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 814         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 815                 if (x->props.family == encap_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, encap_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, encap_family, daddr, saddr,
 824                                            &best, &acquire_in_progress, &error);
 825         }
 826         if (best)
 827                 goto found;
 828
 829         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 830         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 831                 if (x->props.family == encap_family &&
 832                     x->props.reqid == tmpl->reqid &&
 833                     (mark & x->mark.m) == x->mark.v &&
 834                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 835                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 836                     tmpl->mode == x->props.mode &&
 837                     tmpl->id.proto == x->id.proto &&
 838                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 839                         xfrm_state_look_at(pol, x, fl, encap_family, daddr, saddr,
 840                                            &best, &acquire_in_progress, &error);
 841         }
 842
 843 found:
 844         x = best;
 845         if (!x && !error && !acquire_in_progress) {
 846                 if (tmpl->id.spi &&
 847                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 848                                               tmpl->id.proto, encap_family)) != NULL) {
 849                         to_put = x0;
 850                         error = -EEXIST;
 851                         goto out;
 852                 }
 853                 x = xfrm_state_alloc(net);
 854                 if (x == NULL) {
 855                         error = -ENOMEM;
 856                         goto out;
 857                 }
 858                 /* Initialize temporary state matching only
 859                  * to current session. */
 860                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 861                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 862
 863                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
 864                 if (error) {
 865                         x->km.state = XFRM_STATE_DEAD;
 866                         to_put = x;
 867                         x = NULL;
 868                         goto out;
 869                 }
 870
 871                 if (km_query(x, tmpl, pol) == 0) {
 872                         x->km.state = XFRM_STATE_ACQ;
 873                         list_add(&x->km.all, &net->xfrm.state_all);
 874                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 875                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 876                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 877                         if (x->id.spi) {
 878                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 879                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 880                         }
 881                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 882                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 883                         net->xfrm.state_num++;
 884                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 885                 } else {
 886                         x->km.state = XFRM_STATE_DEAD;
 887                         to_put = x;
 888                         x = NULL;
 889                         error = -ESRCH;
 890                 }
 891         }
 892 out:
 893         if (x)
 894                 xfrm_state_hold(x);
 895         else
 896                 *err = acquire_in_progress ? -EAGAIN : error;
 897         spin_unlock_bh(&xfrm_state_lock);
 898         if (to_put)
 899                 xfrm_state_put(to_put);
 900         return x;
 901 }
 902
 903 struct xfrm_state *
 904 xfrm_stateonly_find(struct net *net, u32 mark,
 905                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 906                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 907 {
 908         unsigned int h;
 909         struct xfrm_state *rx = NULL, *x = NULL;
 910         struct hlist_node *entry;
 911
 912         spin_lock(&xfrm_state_lock);
 913         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 914         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 915                 if (x->props.family == family &&
 916                     x->props.reqid == reqid &&
 917                     (mark & x->mark.m) == x->mark.v &&
 918                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 919                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 920                     mode == x->props.mode &&
 921                     proto == x->id.proto &&
 922                     x->km.state == XFRM_STATE_VALID) {
 923                         rx = x;
 924                         break;
 925                 }
 926         }
 927
 928         if (rx)
 929                 xfrm_state_hold(rx);
 930         spin_unlock(&xfrm_state_lock);
 931
 932
 933         return rx;
 934 }
 935 EXPORT_SYMBOL(xfrm_stateonly_find);
 936
 937 static void __xfrm_state_insert(struct xfrm_state *x)
 938 {
 939         struct net *net = xs_net(x);
 940         unsigned int h;
 941
 942         list_add(&x->km.all, &net->xfrm.state_all);
 943
 944         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 945                           x->props.reqid, x->props.family);
 946         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 947
 948         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 949         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 950
 951         if (x->id.spi) {
 952                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 953                                   x->props.family);
 954
 955                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 956         }
 957
 958         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 959         if (x->replay_maxage)
 960                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 961
 962         wake_up(&net->xfrm.km_waitq);
 963
 964         net->xfrm.state_num++;
 965
 966         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 967 }
 968
 969 /* xfrm_state_lock is held */
 970 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 971 {
 972         struct net *net = xs_net(xnew);
 973         unsigned short family = xnew->props.family;
 974         u32 reqid = xnew->props.reqid;
 975         struct xfrm_state *x;
 976         struct hlist_node *entry;
 977         unsigned int h;
 978         u32 mark = xnew->mark.v & xnew->mark.m;
 979
 980         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 981         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 982                 if (x->props.family     == family &&
 983                     x->props.reqid      == reqid &&
 984                     (mark & x->mark.m) == x->mark.v &&
 985                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 986                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 987                         x->genid++;
 988         }
 989 }
 990
 991 void xfrm_state_insert(struct xfrm_state *x)
 992 {
 993         spin_lock_bh(&xfrm_state_lock);
 994         __xfrm_state_bump_genids(x);
 995         __xfrm_state_insert(x);
 996         spin_unlock_bh(&xfrm_state_lock);
 997 }
 998 EXPORT_SYMBOL(xfrm_state_insert);
 999
1000 /* xfrm_state_lock is held */
1001 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)
1002 {
1003         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1004         struct hlist_node *entry;
1005         struct xfrm_state *x;
1006         u32 mark = m->v & m->m;
1007
1008         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1009                 if (x->props.reqid  != reqid ||
1010                     x->props.mode   != mode ||
1011                     x->props.family != family ||
1012                     x->km.state     != XFRM_STATE_ACQ ||
1013                     x->id.spi       != 0 ||
1014                     x->id.proto     != proto ||
1015                     (mark & x->mark.m) != x->mark.v ||
1016                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1017                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1018                         continue;
1019
1020                 xfrm_state_hold(x);
1021                 return x;
1022         }
1023
1024         if (!create)
1025                 return NULL;
1026
1027         x = xfrm_state_alloc(net);
1028         if (likely(x)) {
1029                 switch (family) {
1030                 case AF_INET:
1031                         x->sel.daddr.a4 = daddr->a4;
1032                         x->sel.saddr.a4 = saddr->a4;
1033                         x->sel.prefixlen_d = 32;
1034                         x->sel.prefixlen_s = 32;
1035                         x->props.saddr.a4 = saddr->a4;
1036                         x->id.daddr.a4 = daddr->a4;
1037                         break;
1038
1039                 case AF_INET6:
1040                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1041                                        (struct in6_addr *)daddr);
1042                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1043                                        (struct in6_addr *)saddr);
1044                         x->sel.prefixlen_d = 128;
1045                         x->sel.prefixlen_s = 128;
1046                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1047                                        (struct in6_addr *)saddr);
1048                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1049                                        (struct in6_addr *)daddr);
1050                         break;
1051                 }
1052
1053                 x->km.state = XFRM_STATE_ACQ;
1054                 x->id.proto = proto;
1055                 x->props.family = family;
1056                 x->props.mode = mode;
1057                 x->props.reqid = reqid;
1058                 x->mark.v = m->v;
1059                 x->mark.m = m->m;
1060                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1061                 xfrm_state_hold(x);
1062                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1063                 list_add(&x->km.all, &net->xfrm.state_all);
1064                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1065                 h = xfrm_src_hash(net, daddr, saddr, family);
1066                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1067
1068                 net->xfrm.state_num++;
1069
1070                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1071         }
1072
1073         return x;
1074 }
1075
1076 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1077
1078 int xfrm_state_add(struct xfrm_state *x)
1079 {
1080         struct net *net = xs_net(x);
1081         struct xfrm_state *x1, *to_put;
1082         int family;
1083         int err;
1084         u32 mark = x->mark.v & x->mark.m;
1085         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1086
1087         family = x->props.family;
1088
1089         to_put = NULL;
1090
1091         spin_lock_bh(&xfrm_state_lock);
1092
1093         x1 = __xfrm_state_locate(x, use_spi, family);
1094         if (x1) {
1095                 to_put = x1;
1096                 x1 = NULL;
1097                 err = -EEXIST;
1098                 goto out;
1099         }
1100
1101         if (use_spi && x->km.seq) {
1102                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1103                 if (x1 && ((x1->id.proto != x->id.proto) ||
1104                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1105                         to_put = x1;
1106                         x1 = NULL;
1107                 }
1108         }
1109
1110         if (use_spi && !x1)
1111                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1112                                      x->props.reqid, x->id.proto,
1113                                      &x->id.daddr, &x->props.saddr, 0);
1114
1115         __xfrm_state_bump_genids(x);
1116         __xfrm_state_insert(x);
1117         err = 0;
1118
1119 out:
1120         spin_unlock_bh(&xfrm_state_lock);
1121
1122         if (x1) {
1123                 xfrm_state_delete(x1);
1124                 xfrm_state_put(x1);
1125         }
1126
1127         if (to_put)
1128                 xfrm_state_put(to_put);
1129
1130         return err;
1131 }
1132 EXPORT_SYMBOL(xfrm_state_add);
1133
1134 #ifdef CONFIG_XFRM_MIGRATE
1135 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1136 {
1137         struct net *net = xs_net(orig);
1138         int err = -ENOMEM;
1139         struct xfrm_state *x = xfrm_state_alloc(net);
1140         if (!x)
1141                 goto out;
1142
1143         memcpy(&x->id, &orig->id, sizeof(x->id));
1144         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1145         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1146         x->props.mode = orig->props.mode;
1147         x->props.replay_window = orig->props.replay_window;
1148         x->props.reqid = orig->props.reqid;
1149         x->props.family = orig->props.family;
1150         x->props.saddr = orig->props.saddr;
1151
1152         if (orig->aalg) {
1153                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1154                 if (!x->aalg)
1155                         goto error;
1156         }
1157         x->props.aalgo = orig->props.aalgo;
1158
1159         if (orig->ealg) {
1160                 x->ealg = xfrm_algo_clone(orig->ealg);
1161                 if (!x->ealg)
1162                         goto error;
1163         }
1164         x->props.ealgo = orig->props.ealgo;
1165
1166         if (orig->calg) {
1167                 x->calg = xfrm_algo_clone(orig->calg);
1168                 if (!x->calg)
1169                         goto error;
1170         }
1171         x->props.calgo = orig->props.calgo;
1172
1173         if (orig->encap) {
1174                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1175                 if (!x->encap)
1176                         goto error;
1177         }
1178
1179         if (orig->coaddr) {
1180                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1181                                     GFP_KERNEL);
1182                 if (!x->coaddr)
1183                         goto error;
1184         }
1185
1186         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1187
1188         err = xfrm_init_state(x);
1189         if (err)
1190                 goto error;
1191
1192         x->props.flags = orig->props.flags;
1193
1194         x->curlft.add_time = orig->curlft.add_time;
1195         x->km.state = orig->km.state;
1196         x->km.seq = orig->km.seq;
1197
1198         return x;
1199
1200  error:
1201         xfrm_state_put(x);
1202 out:
1203         if (errp)
1204                 *errp = err;
1205         return NULL;
1206 }
1207
1208 /* xfrm_state_lock is held */
1209 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1210 {
1211         unsigned int h;
1212         struct xfrm_state *x;
1213         struct hlist_node *entry;
1214
1215         if (m->reqid) {
1216                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1217                                   m->reqid, m->old_family);
1218                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1219                         if (x->props.mode != m->mode ||
1220                             x->id.proto != m->proto)
1221                                 continue;
1222                         if (m->reqid && x->props.reqid != m->reqid)
1223                                 continue;
1224                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1225                                           m->old_family) ||
1226                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1227                                           m->old_family))
1228                                 continue;
1229                         xfrm_state_hold(x);
1230                         return x;
1231                 }
1232         } else {
1233                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1234                                   m->old_family);
1235                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1236                         if (x->props.mode != m->mode ||
1237                             x->id.proto != m->proto)
1238                                 continue;
1239                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1240                                           m->old_family) ||
1241                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1242                                           m->old_family))
1243                                 continue;
1244                         xfrm_state_hold(x);
1245                         return x;
1246                 }
1247         }
1248
1249         return NULL;
1250 }
1251 EXPORT_SYMBOL(xfrm_migrate_state_find);
1252
1253 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1254                                        struct xfrm_migrate *m)
1255 {
1256         struct xfrm_state *xc;
1257         int err;
1258
1259         xc = xfrm_state_clone(x, &err);
1260         if (!xc)
1261                 return NULL;
1262
1263         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1264         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1265
1266         /* add state */
1267         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1268                 /* a care is needed when the destination address of the
1269                    state is to be updated as it is a part of triplet */
1270                 xfrm_state_insert(xc);
1271         } else {
1272                 if ((err = xfrm_state_add(xc)) < 0)
1273                         goto error;
1274         }
1275
1276         return xc;
1277 error:
1278         xfrm_state_put(xc);
1279         return NULL;
1280 }
1281 EXPORT_SYMBOL(xfrm_state_migrate);
1282 #endif
1283
1284 int xfrm_state_update(struct xfrm_state *x)
1285 {
1286         struct xfrm_state *x1, *to_put;
1287         int err;
1288         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1289
1290         to_put = NULL;
1291
1292         spin_lock_bh(&xfrm_state_lock);
1293         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1294
1295         err = -ESRCH;
1296         if (!x1)
1297                 goto out;
1298
1299         if (xfrm_state_kern(x1)) {
1300                 to_put = x1;
1301                 err = -EEXIST;
1302                 goto out;
1303         }
1304
1305         if (x1->km.state == XFRM_STATE_ACQ) {
1306                 __xfrm_state_insert(x);
1307                 x = NULL;
1308         }
1309         err = 0;
1310
1311 out:
1312         spin_unlock_bh(&xfrm_state_lock);
1313
1314         if (to_put)
1315                 xfrm_state_put(to_put);
1316
1317         if (err)
1318                 return err;
1319
1320         if (!x) {
1321                 xfrm_state_delete(x1);
1322                 xfrm_state_put(x1);
1323                 return 0;
1324         }
1325
1326         err = -EINVAL;
1327         spin_lock_bh(&x1->lock);
1328         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1329                 if (x->encap && x1->encap)
1330                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1331                 if (x->coaddr && x1->coaddr) {
1332                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1333                 }
1334                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1335                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1336                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1337                 x1->km.dying = 0;
1338
1339                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1340                 if (x1->curlft.use_time)
1341                         xfrm_state_check_expire(x1);
1342
1343                 err = 0;
1344         }
1345         spin_unlock_bh(&x1->lock);
1346
1347         xfrm_state_put(x1);
1348
1349         return err;
1350 }
1351 EXPORT_SYMBOL(xfrm_state_update);
1352
1353 int xfrm_state_check_expire(struct xfrm_state *x)
1354 {
1355         if (!x->curlft.use_time)
1356                 x->curlft.use_time = get_seconds();
1357
1358         if (x->km.state != XFRM_STATE_VALID)
1359                 return -EINVAL;
1360
1361         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1362             x->curlft.packets >= x->lft.hard_packet_limit) {
1363                 x->km.state = XFRM_STATE_EXPIRED;
1364                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1365                 return -EINVAL;
1366         }
1367
1368         if (!x->km.dying &&
1369             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1370              x->curlft.packets >= x->lft.soft_packet_limit)) {
1371                 x->km.dying = 1;
1372                 km_state_expired(x, 0, 0);
1373         }
1374         return 0;
1375 }
1376 EXPORT_SYMBOL(xfrm_state_check_expire);
1377
1378 struct xfrm_state *
1379 xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi,
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(net, mark, daddr, spi, proto, family);
1386         spin_unlock_bh(&xfrm_state_lock);
1387         return x;
1388 }
1389 EXPORT_SYMBOL(xfrm_state_lookup);
1390
1391 struct xfrm_state *
1392 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1393                          xfrm_address_t *daddr, xfrm_address_t *saddr,
1394                          u8 proto, unsigned short family)
1395 {
1396         struct xfrm_state *x;
1397
1398         spin_lock_bh(&xfrm_state_lock);
1399         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1400         spin_unlock_bh(&xfrm_state_lock);
1401         return x;
1402 }
1403 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1404
1405 struct xfrm_state *
1406 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1407               xfrm_address_t *daddr, xfrm_address_t *saddr,
1408               int create, unsigned short family)
1409 {
1410         struct xfrm_state *x;
1411
1412         spin_lock_bh(&xfrm_state_lock);
1413         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1414         spin_unlock_bh(&xfrm_state_lock);
1415
1416         return x;
1417 }
1418 EXPORT_SYMBOL(xfrm_find_acq);
1419
1420 #ifdef CONFIG_XFRM_SUB_POLICY
1421 int
1422 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1423                unsigned short family)
1424 {
1425         int err = 0;
1426         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1427         if (!afinfo)
1428                 return -EAFNOSUPPORT;
1429
1430         spin_lock_bh(&xfrm_state_lock);
1431         if (afinfo->tmpl_sort)
1432                 err = afinfo->tmpl_sort(dst, src, n);
1433         spin_unlock_bh(&xfrm_state_lock);
1434         xfrm_state_put_afinfo(afinfo);
1435         return err;
1436 }
1437 EXPORT_SYMBOL(xfrm_tmpl_sort);
1438
1439 int
1440 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1441                 unsigned short family)
1442 {
1443         int err = 0;
1444         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1445         if (!afinfo)
1446                 return -EAFNOSUPPORT;
1447
1448         spin_lock_bh(&xfrm_state_lock);
1449         if (afinfo->state_sort)
1450                 err = afinfo->state_sort(dst, src, n);
1451         spin_unlock_bh(&xfrm_state_lock);
1452         xfrm_state_put_afinfo(afinfo);
1453         return err;
1454 }
1455 EXPORT_SYMBOL(xfrm_state_sort);
1456 #endif
1457
1458 /* Silly enough, but I'm lazy to build resolution list */
1459
1460 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1461 {
1462         int i;
1463
1464         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1465                 struct hlist_node *entry;
1466                 struct xfrm_state *x;
1467
1468                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1469                         if (x->km.seq == seq &&
1470                             (mark & x->mark.m) == x->mark.v &&
1471                             x->km.state == XFRM_STATE_ACQ) {
1472                                 xfrm_state_hold(x);
1473                                 return x;
1474                         }
1475                 }
1476         }
1477         return NULL;
1478 }
1479
1480 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1481 {
1482         struct xfrm_state *x;
1483
1484         spin_lock_bh(&xfrm_state_lock);
1485         x = __xfrm_find_acq_byseq(net, mark, seq);
1486         spin_unlock_bh(&xfrm_state_lock);
1487         return x;
1488 }
1489 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1490
1491 u32 xfrm_get_acqseq(void)
1492 {
1493         u32 res;
1494         static atomic_t acqseq;
1495
1496         do {
1497                 res = atomic_inc_return(&acqseq);
1498         } while (!res);
1499
1500         return res;
1501 }
1502 EXPORT_SYMBOL(xfrm_get_acqseq);
1503
1504 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1505 {
1506         struct net *net = xs_net(x);
1507         unsigned int h;
1508         struct xfrm_state *x0;
1509         int err = -ENOENT;
1510         __be32 minspi = htonl(low);
1511         __be32 maxspi = htonl(high);
1512         u32 mark = x->mark.v & x->mark.m;
1513
1514         spin_lock_bh(&x->lock);
1515         if (x->km.state == XFRM_STATE_DEAD)
1516                 goto unlock;
1517
1518         err = 0;
1519         if (x->id.spi)
1520                 goto unlock;
1521
1522         err = -ENOENT;
1523
1524         if (minspi == maxspi) {
1525                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1526                 if (x0) {
1527                         xfrm_state_put(x0);
1528                         goto unlock;
1529                 }
1530                 x->id.spi = minspi;
1531         } else {
1532                 u32 spi = 0;
1533                 for (h=0; h<high-low+1; h++) {
1534                         spi = low + net_random()%(high-low+1);
1535                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1536                         if (x0 == NULL) {
1537                                 x->id.spi = htonl(spi);
1538                                 break;
1539                         }
1540                         xfrm_state_put(x0);
1541                 }
1542         }
1543         if (x->id.spi) {
1544                 spin_lock_bh(&xfrm_state_lock);
1545                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1546                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1547                 spin_unlock_bh(&xfrm_state_lock);
1548
1549                 err = 0;
1550         }
1551
1552 unlock:
1553         spin_unlock_bh(&x->lock);
1554
1555         return err;
1556 }
1557 EXPORT_SYMBOL(xfrm_alloc_spi);
1558
1559 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1560                     int (*func)(struct xfrm_state *, int, void*),
1561                     void *data)
1562 {
1563         struct xfrm_state *state;
1564         struct xfrm_state_walk *x;
1565         int err = 0;
1566
1567         if (walk->seq != 0 && list_empty(&walk->all))
1568                 return 0;
1569
1570         spin_lock_bh(&xfrm_state_lock);
1571         if (list_empty(&walk->all))
1572                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1573         else
1574                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1575         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1576                 if (x->state == XFRM_STATE_DEAD)
1577                         continue;
1578                 state = container_of(x, struct xfrm_state, km);
1579                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1580                         continue;
1581                 err = func(state, walk->seq, data);
1582                 if (err) {
1583                         list_move_tail(&walk->all, &x->all);
1584                         goto out;
1585                 }
1586                 walk->seq++;
1587         }
1588         if (walk->seq == 0) {
1589                 err = -ENOENT;
1590                 goto out;
1591         }
1592         list_del_init(&walk->all);
1593 out:
1594         spin_unlock_bh(&xfrm_state_lock);
1595         return err;
1596 }
1597 EXPORT_SYMBOL(xfrm_state_walk);
1598
1599 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1600 {
1601         INIT_LIST_HEAD(&walk->all);
1602         walk->proto = proto;
1603         walk->state = XFRM_STATE_DEAD;
1604         walk->seq = 0;
1605 }
1606 EXPORT_SYMBOL(xfrm_state_walk_init);
1607
1608 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1609 {
1610         if (list_empty(&walk->all))
1611                 return;
1612
1613         spin_lock_bh(&xfrm_state_lock);
1614         list_del(&walk->all);
1615         spin_unlock_bh(&xfrm_state_lock);
1616 }
1617 EXPORT_SYMBOL(xfrm_state_walk_done);
1618
1619
1620 void xfrm_replay_notify(struct xfrm_state *x, int event)
1621 {
1622         struct km_event c;
1623         /* we send notify messages in case
1624          *  1. we updated on of the sequence numbers, and the seqno difference
1625          *     is at least x->replay_maxdiff, in this case we also update the
1626          *     timeout of our timer function
1627          *  2. if x->replay_maxage has elapsed since last update,
1628          *     and there were changes
1629          *
1630          *  The state structure must be locked!
1631          */
1632
1633         switch (event) {
1634         case XFRM_REPLAY_UPDATE:
1635                 if (x->replay_maxdiff &&
1636                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1637                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1638                         if (x->xflags & XFRM_TIME_DEFER)
1639                                 event = XFRM_REPLAY_TIMEOUT;
1640                         else
1641                                 return;
1642                 }
1643
1644                 break;
1645
1646         case XFRM_REPLAY_TIMEOUT:
1647                 if ((x->replay.seq == x->preplay.seq) &&
1648                     (x->replay.bitmap == x->preplay.bitmap) &&
1649                     (x->replay.oseq == x->preplay.oseq)) {
1650                         x->xflags |= XFRM_TIME_DEFER;
1651                         return;
1652                 }
1653
1654                 break;
1655         }
1656
1657         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1658         c.event = XFRM_MSG_NEWAE;
1659         c.data.aevent = event;
1660         km_state_notify(x, &c);
1661
1662         if (x->replay_maxage &&
1663             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1664                 x->xflags &= ~XFRM_TIME_DEFER;
1665 }
1666
1667 static void xfrm_replay_timer_handler(unsigned long data)
1668 {
1669         struct xfrm_state *x = (struct xfrm_state*)data;
1670
1671         spin_lock(&x->lock);
1672
1673         if (x->km.state == XFRM_STATE_VALID) {
1674                 if (xfrm_aevent_is_on(xs_net(x)))
1675                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1676                 else
1677                         x->xflags |= XFRM_TIME_DEFER;
1678         }
1679
1680         spin_unlock(&x->lock);
1681 }
1682
1683 int xfrm_replay_check(struct xfrm_state *x,
1684                       struct sk_buff *skb, __be32 net_seq)
1685 {
1686         u32 diff;
1687         u32 seq = ntohl(net_seq);
1688
1689         if (unlikely(seq == 0))
1690                 goto err;
1691
1692         if (likely(seq > x->replay.seq))
1693                 return 0;
1694
1695         diff = x->replay.seq - seq;
1696         if (diff >= min_t(unsigned int, x->props.replay_window,
1697                           sizeof(x->replay.bitmap) * 8)) {
1698                 x->stats.replay_window++;
1699                 goto err;
1700         }
1701
1702         if (x->replay.bitmap & (1U << diff)) {
1703                 x->stats.replay++;
1704                 goto err;
1705         }
1706         return 0;
1707
1708 err:
1709         xfrm_audit_state_replay(x, skb, net_seq);
1710         return -EINVAL;
1711 }
1712
1713 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1714 {
1715         u32 diff;
1716         u32 seq = ntohl(net_seq);
1717
1718         if (seq > x->replay.seq) {
1719                 diff = seq - x->replay.seq;
1720                 if (diff < x->props.replay_window)
1721                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1722                 else
1723                         x->replay.bitmap = 1;
1724                 x->replay.seq = seq;
1725         } else {
1726                 diff = x->replay.seq - seq;
1727                 x->replay.bitmap |= (1U << diff);
1728         }
1729
1730         if (xfrm_aevent_is_on(xs_net(x)))
1731                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1732 }
1733
1734 static LIST_HEAD(xfrm_km_list);
1735 static DEFINE_RWLOCK(xfrm_km_lock);
1736
1737 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1738 {
1739         struct xfrm_mgr *km;
1740
1741         read_lock(&xfrm_km_lock);
1742         list_for_each_entry(km, &xfrm_km_list, list)
1743                 if (km->notify_policy)
1744                         km->notify_policy(xp, dir, c);
1745         read_unlock(&xfrm_km_lock);
1746 }
1747
1748 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1749 {
1750         struct xfrm_mgr *km;
1751         read_lock(&xfrm_km_lock);
1752         list_for_each_entry(km, &xfrm_km_list, list)
1753                 if (km->notify)
1754                         km->notify(x, c);
1755         read_unlock(&xfrm_km_lock);
1756 }
1757
1758 EXPORT_SYMBOL(km_policy_notify);
1759 EXPORT_SYMBOL(km_state_notify);
1760
1761 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1762 {
1763         struct net *net = xs_net(x);
1764         struct km_event c;
1765
1766         c.data.hard = hard;
1767         c.pid = pid;
1768         c.event = XFRM_MSG_EXPIRE;
1769         km_state_notify(x, &c);
1770
1771         if (hard)
1772                 wake_up(&net->xfrm.km_waitq);
1773 }
1774
1775 EXPORT_SYMBOL(km_state_expired);
1776 /*
1777  * We send to all registered managers regardless of failure
1778  * We are happy with one success
1779 */
1780 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1781 {
1782         int err = -EINVAL, acqret;
1783         struct xfrm_mgr *km;
1784
1785         read_lock(&xfrm_km_lock);
1786         list_for_each_entry(km, &xfrm_km_list, list) {
1787                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1788                 if (!acqret)
1789                         err = acqret;
1790         }
1791         read_unlock(&xfrm_km_lock);
1792         return err;
1793 }
1794 EXPORT_SYMBOL(km_query);
1795
1796 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1797 {
1798         int err = -EINVAL;
1799         struct xfrm_mgr *km;
1800
1801         read_lock(&xfrm_km_lock);
1802         list_for_each_entry(km, &xfrm_km_list, list) {
1803                 if (km->new_mapping)
1804                         err = km->new_mapping(x, ipaddr, sport);
1805                 if (!err)
1806                         break;
1807         }
1808         read_unlock(&xfrm_km_lock);
1809         return err;
1810 }
1811 EXPORT_SYMBOL(km_new_mapping);
1812
1813 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1814 {
1815         struct net *net = xp_net(pol);
1816         struct km_event c;
1817
1818         c.data.hard = hard;
1819         c.pid = pid;
1820         c.event = XFRM_MSG_POLEXPIRE;
1821         km_policy_notify(pol, dir, &c);
1822
1823         if (hard)
1824                 wake_up(&net->xfrm.km_waitq);
1825 }
1826 EXPORT_SYMBOL(km_policy_expired);
1827
1828 #ifdef CONFIG_XFRM_MIGRATE
1829 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1830                const struct xfrm_migrate *m, int num_migrate,
1831                const struct xfrm_kmaddress *k)
1832 {
1833         int err = -EINVAL;
1834         int ret;
1835         struct xfrm_mgr *km;
1836
1837         read_lock(&xfrm_km_lock);
1838         list_for_each_entry(km, &xfrm_km_list, list) {
1839                 if (km->migrate) {
1840                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1841                         if (!ret)
1842                                 err = ret;
1843                 }
1844         }
1845         read_unlock(&xfrm_km_lock);
1846         return err;
1847 }
1848 EXPORT_SYMBOL(km_migrate);
1849 #endif
1850
1851 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1852 {
1853         int err = -EINVAL;
1854         int ret;
1855         struct xfrm_mgr *km;
1856
1857         read_lock(&xfrm_km_lock);
1858         list_for_each_entry(km, &xfrm_km_list, list) {
1859                 if (km->report) {
1860                         ret = km->report(net, proto, sel, addr);
1861                         if (!ret)
1862                                 err = ret;
1863                 }
1864         }
1865         read_unlock(&xfrm_km_lock);
1866         return err;
1867 }
1868 EXPORT_SYMBOL(km_report);
1869
1870 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1871 {
1872         int err;
1873         u8 *data;
1874         struct xfrm_mgr *km;
1875         struct xfrm_policy *pol = NULL;
1876
1877         if (optlen <= 0 || optlen > PAGE_SIZE)
1878                 return -EMSGSIZE;
1879
1880         data = kmalloc(optlen, GFP_KERNEL);
1881         if (!data)
1882                 return -ENOMEM;
1883
1884         err = -EFAULT;
1885         if (copy_from_user(data, optval, optlen))
1886                 goto out;
1887
1888         err = -EINVAL;
1889         read_lock(&xfrm_km_lock);
1890         list_for_each_entry(km, &xfrm_km_list, list) {
1891                 pol = km->compile_policy(sk, optname, data,
1892                                          optlen, &err);
1893                 if (err >= 0)
1894                         break;
1895         }
1896         read_unlock(&xfrm_km_lock);
1897
1898         if (err >= 0) {
1899                 xfrm_sk_policy_insert(sk, err, pol);
1900                 xfrm_pol_put(pol);
1901                 err = 0;
1902         }
1903
1904 out:
1905         kfree(data);
1906         return err;
1907 }
1908 EXPORT_SYMBOL(xfrm_user_policy);
1909
1910 int xfrm_register_km(struct xfrm_mgr *km)
1911 {
1912         write_lock_bh(&xfrm_km_lock);
1913         list_add_tail(&km->list, &xfrm_km_list);
1914         write_unlock_bh(&xfrm_km_lock);
1915         return 0;
1916 }
1917 EXPORT_SYMBOL(xfrm_register_km);
1918
1919 int xfrm_unregister_km(struct xfrm_mgr *km)
1920 {
1921         write_lock_bh(&xfrm_km_lock);
1922         list_del(&km->list);
1923         write_unlock_bh(&xfrm_km_lock);
1924         return 0;
1925 }
1926 EXPORT_SYMBOL(xfrm_unregister_km);
1927
1928 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1929 {
1930         int err = 0;
1931         if (unlikely(afinfo == NULL))
1932                 return -EINVAL;
1933         if (unlikely(afinfo->family >= NPROTO))
1934                 return -EAFNOSUPPORT;
1935         write_lock_bh(&xfrm_state_afinfo_lock);
1936         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1937                 err = -ENOBUFS;
1938         else
1939                 xfrm_state_afinfo[afinfo->family] = afinfo;
1940         write_unlock_bh(&xfrm_state_afinfo_lock);
1941         return err;
1942 }
1943 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1944
1945 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1946 {
1947         int err = 0;
1948         if (unlikely(afinfo == NULL))
1949                 return -EINVAL;
1950         if (unlikely(afinfo->family >= NPROTO))
1951                 return -EAFNOSUPPORT;
1952         write_lock_bh(&xfrm_state_afinfo_lock);
1953         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1954                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1955                         err = -EINVAL;
1956                 else
1957                         xfrm_state_afinfo[afinfo->family] = NULL;
1958         }
1959         write_unlock_bh(&xfrm_state_afinfo_lock);
1960         return err;
1961 }
1962 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1963
1964 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1965 {
1966         struct xfrm_state_afinfo *afinfo;
1967         if (unlikely(family >= NPROTO))
1968                 return NULL;
1969         read_lock(&xfrm_state_afinfo_lock);
1970         afinfo = xfrm_state_afinfo[family];
1971         if (unlikely(!afinfo))
1972                 read_unlock(&xfrm_state_afinfo_lock);
1973         return afinfo;
1974 }
1975
1976 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1977         __releases(xfrm_state_afinfo_lock)
1978 {
1979         read_unlock(&xfrm_state_afinfo_lock);
1980 }
1981
1982 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1983 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1984 {
1985         if (x->tunnel) {
1986                 struct xfrm_state *t = x->tunnel;
1987
1988                 if (atomic_read(&t->tunnel_users) == 2)
1989                         xfrm_state_delete(t);
1990                 atomic_dec(&t->tunnel_users);
1991                 xfrm_state_put(t);
1992                 x->tunnel = NULL;
1993         }
1994 }
1995 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1996
1997 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1998 {
1999         int res;
2000
2001         spin_lock_bh(&x->lock);
2002         if (x->km.state == XFRM_STATE_VALID &&
2003             x->type && x->type->get_mtu)
2004                 res = x->type->get_mtu(x, mtu);
2005         else
2006                 res = mtu - x->props.header_len;
2007         spin_unlock_bh(&x->lock);
2008         return res;
2009 }
2010
2011 int xfrm_init_state(struct xfrm_state *x)
2012 {
2013         struct xfrm_state_afinfo *afinfo;
2014         struct xfrm_mode *inner_mode;
2015         int family = x->props.family;
2016         int err;
2017
2018         err = -EAFNOSUPPORT;
2019         afinfo = xfrm_state_get_afinfo(family);
2020         if (!afinfo)
2021                 goto error;
2022
2023         err = 0;
2024         if (afinfo->init_flags)
2025                 err = afinfo->init_flags(x);
2026
2027         xfrm_state_put_afinfo(afinfo);
2028
2029         if (err)
2030                 goto error;
2031
2032         err = -EPROTONOSUPPORT;
2033
2034         if (x->sel.family != AF_UNSPEC) {
2035                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2036                 if (inner_mode == NULL)
2037                         goto error;
2038
2039                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2040                     family != x->sel.family) {
2041                         xfrm_put_mode(inner_mode);
2042                         goto error;
2043                 }
2044
2045                 x->inner_mode = inner_mode;
2046         } else {
2047                 struct xfrm_mode *inner_mode_iaf;
2048                 int iafamily = AF_INET;
2049
2050                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2051                 if (inner_mode == NULL)
2052                         goto error;
2053
2054                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2055                         xfrm_put_mode(inner_mode);
2056                         goto error;
2057                 }
2058                 x->inner_mode = inner_mode;
2059
2060                 if (x->props.family == AF_INET)
2061                         iafamily = AF_INET6;
2062
2063                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2064                 if (inner_mode_iaf) {
2065                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2066                                 x->inner_mode_iaf = inner_mode_iaf;
2067                         else
2068                                 xfrm_put_mode(inner_mode_iaf);
2069                 }
2070         }
2071
2072         x->type = xfrm_get_type(x->id.proto, family);
2073         if (x->type == NULL)
2074                 goto error;
2075
2076         err = x->type->init_state(x);
2077         if (err)
2078                 goto error;
2079
2080         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2081         if (x->outer_mode == NULL)
2082                 goto error;
2083
2084         x->km.state = XFRM_STATE_VALID;
2085
2086 error:
2087         return err;
2088 }
2089
2090 EXPORT_SYMBOL(xfrm_init_state);
2091
2092 int __net_init xfrm_state_init(struct net *net)
2093 {
2094         unsigned int sz;
2095
2096         INIT_LIST_HEAD(&net->xfrm.state_all);
2097
2098         sz = sizeof(struct hlist_head) * 8;
2099
2100         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2101         if (!net->xfrm.state_bydst)
2102                 goto out_bydst;
2103         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2104         if (!net->xfrm.state_bysrc)
2105                 goto out_bysrc;
2106         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2107         if (!net->xfrm.state_byspi)
2108                 goto out_byspi;
2109         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2110
2111         net->xfrm.state_num = 0;
2112         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2113         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2114         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2115         init_waitqueue_head(&net->xfrm.km_waitq);
2116         return 0;
2117
2118 out_byspi:
2119         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2120 out_bysrc:
2121         xfrm_hash_free(net->xfrm.state_bydst, sz);
2122 out_bydst:
2123         return -ENOMEM;
2124 }
2125
2126 void xfrm_state_fini(struct net *net)
2127 {
2128         struct xfrm_audit audit_info;
2129         unsigned int sz;
2130
2131         flush_work(&net->xfrm.state_hash_work);
2132         audit_info.loginuid = -1;
2133         audit_info.sessionid = -1;
2134         audit_info.secid = 0;
2135         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2136         flush_work(&net->xfrm.state_gc_work);
2137
2138         WARN_ON(!list_empty(&net->xfrm.state_all));
2139
2140         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2141         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2142         xfrm_hash_free(net->xfrm.state_byspi, sz);
2143         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2144         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2145         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2146         xfrm_hash_free(net->xfrm.state_bydst, sz);
2147 }
2148
2149 #ifdef CONFIG_AUDITSYSCALL
2150 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2151                                      struct audit_buffer *audit_buf)
2152 {
2153         struct xfrm_sec_ctx *ctx = x->security;
2154         u32 spi = ntohl(x->id.spi);
2155
2156         if (ctx)
2157                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2158                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2159
2160         switch(x->props.family) {
2161         case AF_INET:
2162                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2163                                  &x->props.saddr.a4, &x->id.daddr.a4);
2164                 break;
2165         case AF_INET6:
2166                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2167                                  x->props.saddr.a6, x->id.daddr.a6);
2168                 break;
2169         }
2170
2171         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2172 }
2173
2174 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2175                                       struct audit_buffer *audit_buf)
2176 {
2177         struct iphdr *iph4;
2178         struct ipv6hdr *iph6;
2179
2180         switch (family) {
2181         case AF_INET:
2182                 iph4 = ip_hdr(skb);
2183                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2184                                  &iph4->saddr, &iph4->daddr);
2185                 break;
2186         case AF_INET6:
2187                 iph6 = ipv6_hdr(skb);
2188                 audit_log_format(audit_buf,
2189                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2190                                  &iph6->saddr,&iph6->daddr,
2191                                  iph6->flow_lbl[0] & 0x0f,
2192                                  iph6->flow_lbl[1],
2193                                  iph6->flow_lbl[2]);
2194                 break;
2195         }
2196 }
2197
2198 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2199                           uid_t auid, u32 sessionid, u32 secid)
2200 {
2201         struct audit_buffer *audit_buf;
2202
2203         audit_buf = xfrm_audit_start("SAD-add");
2204         if (audit_buf == NULL)
2205                 return;
2206         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2207         xfrm_audit_helper_sainfo(x, audit_buf);
2208         audit_log_format(audit_buf, " res=%u", result);
2209         audit_log_end(audit_buf);
2210 }
2211 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2212
2213 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2214                              uid_t auid, u32 sessionid, u32 secid)
2215 {
2216         struct audit_buffer *audit_buf;
2217
2218         audit_buf = xfrm_audit_start("SAD-delete");
2219         if (audit_buf == NULL)
2220                 return;
2221         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2222         xfrm_audit_helper_sainfo(x, audit_buf);
2223         audit_log_format(audit_buf, " res=%u", result);
2224         audit_log_end(audit_buf);
2225 }
2226 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2227
2228 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2229                                       struct sk_buff *skb)
2230 {
2231         struct audit_buffer *audit_buf;
2232         u32 spi;
2233
2234         audit_buf = xfrm_audit_start("SA-replay-overflow");
2235         if (audit_buf == NULL)
2236                 return;
2237         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2238         /* don't record the sequence number because it's inherent in this kind
2239          * of audit message */
2240         spi = ntohl(x->id.spi);
2241         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2242         audit_log_end(audit_buf);
2243 }
2244 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2245
2246 static void xfrm_audit_state_replay(struct xfrm_state *x,
2247                              struct sk_buff *skb, __be32 net_seq)
2248 {
2249         struct audit_buffer *audit_buf;
2250         u32 spi;
2251
2252         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2253         if (audit_buf == NULL)
2254                 return;
2255         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2256         spi = ntohl(x->id.spi);
2257         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2258                          spi, spi, ntohl(net_seq));
2259         audit_log_end(audit_buf);
2260 }
2261
2262 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2263 {
2264         struct audit_buffer *audit_buf;
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         audit_log_end(audit_buf);
2271 }
2272 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2273
2274 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2275                                __be32 net_spi, __be32 net_seq)
2276 {
2277         struct audit_buffer *audit_buf;
2278         u32 spi;
2279
2280         audit_buf = xfrm_audit_start("SA-notfound");
2281         if (audit_buf == NULL)
2282                 return;
2283         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2284         spi = ntohl(net_spi);
2285         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2286                          spi, spi, ntohl(net_seq));
2287         audit_log_end(audit_buf);
2288 }
2289 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2290
2291 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2292                               struct sk_buff *skb, u8 proto)
2293 {
2294         struct audit_buffer *audit_buf;
2295         __be32 net_spi;
2296         __be32 net_seq;
2297
2298         audit_buf = xfrm_audit_start("SA-icv-failure");
2299         if (audit_buf == NULL)
2300                 return;
2301         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2302         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2303                 u32 spi = ntohl(net_spi);
2304                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2305                                  spi, spi, ntohl(net_seq));
2306         }
2307         audit_log_end(audit_buf);
2308 }
2309 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2310 #endif /* CONFIG_AUDITSYSCALL */