net/xfrm/xfrm_state.c

   1 /*
   2  * xfrm_state.c
   3  *
   4  * Changes:
   5  *      Mitsuru KANDA @USAGI
   6  *      Kazunori MIYAZAWA @USAGI
   7  *      Kunihiro Ishiguro <kunihiro@ipinfusion.com>
   8  *              IPv6 support
   9  *      YOSHIFUJI Hideaki @USAGI
  10  *              Split up af-specific functions
  11  *      Derek Atkins <derek@ihtfp.com>
  12  *              Add UDP Encapsulation
  13  *
  14  */
  15
  16 #include <linux/workqueue.h>
  17 #include <net/xfrm.h>
  18 #include <linux/pfkeyv2.h>
  19 #include <linux/ipsec.h>
  20 #include <linux/module.h>
  21 #include <linux/cache.h>
  22 #include <linux/audit.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/ktime.h>
  25 #include <linux/slab.h>
  26 #include <linux/interrupt.h>
  27 #include <linux/kernel.h>
  28
  29 #include "xfrm_hash.h"
  30
  31 /* Each xfrm_state may be linked to two tables:
  32
  33    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  34    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  35       destination/tunnel endpoint. (output)
  36  */
  37
  38 static DEFINE_SPINLOCK(xfrm_state_lock);
  39
  40 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  41
  42 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
  43 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
  44
  45 static inline unsigned int xfrm_dst_hash(struct net *net,
  46                                          const xfrm_address_t *daddr,
  47                                          const xfrm_address_t *saddr,
  48                                          u32 reqid,
  49                                          unsigned short family)
  50 {
  51         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  52 }
  53
  54 static inline unsigned int xfrm_src_hash(struct net *net,
  55                                          const xfrm_address_t *daddr,
  56                                          const xfrm_address_t *saddr,
  57                                          unsigned short family)
  58 {
  59         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  60 }
  61
  62 static inline unsigned int
  63 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  64               __be32 spi, u8 proto, unsigned short family)
  65 {
  66         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  67 }
  68
  69 static void xfrm_hash_transfer(struct hlist_head *list,
  70                                struct hlist_head *ndsttable,
  71                                struct hlist_head *nsrctable,
  72                                struct hlist_head *nspitable,
  73                                unsigned int nhashmask)
  74 {
  75         struct hlist_node *tmp;
  76         struct xfrm_state *x;
  77
  78         hlist_for_each_entry_safe(x, tmp, list, bydst) {
  79                 unsigned int h;
  80
  81                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  82                                     x->props.reqid, x->props.family,
  83                                     nhashmask);
  84                 hlist_add_head(&x->bydst, ndsttable+h);
  85
  86                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  87                                     x->props.family,
  88                                     nhashmask);
  89                 hlist_add_head(&x->bysrc, nsrctable+h);
  90
  91                 if (x->id.spi) {
  92                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  93                                             x->id.proto, x->props.family,
  94                                             nhashmask);
  95                         hlist_add_head(&x->byspi, nspitable+h);
  96                 }
  97         }
  98 }
  99
 100 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 101 {
 102         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 103 }
 104
 105 static DEFINE_MUTEX(hash_resize_mutex);
 106
 107 static void xfrm_hash_resize(struct work_struct *work)
 108 {
 109         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 110         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 111         unsigned long nsize, osize;
 112         unsigned int nhashmask, ohashmask;
 113         int i;
 114
 115         mutex_lock(&hash_resize_mutex);
 116
 117         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 118         ndst = xfrm_hash_alloc(nsize);
 119         if (!ndst)
 120                 goto out_unlock;
 121         nsrc = xfrm_hash_alloc(nsize);
 122         if (!nsrc) {
 123                 xfrm_hash_free(ndst, nsize);
 124                 goto out_unlock;
 125         }
 126         nspi = xfrm_hash_alloc(nsize);
 127         if (!nspi) {
 128                 xfrm_hash_free(ndst, nsize);
 129                 xfrm_hash_free(nsrc, nsize);
 130                 goto out_unlock;
 131         }
 132
 133         spin_lock_bh(&xfrm_state_lock);
 134
 135         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 136         for (i = net->xfrm.state_hmask; i >= 0; i--)
 137                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 138                                    nhashmask);
 139
 140         odst = net->xfrm.state_bydst;
 141         osrc = net->xfrm.state_bysrc;
 142         ospi = net->xfrm.state_byspi;
 143         ohashmask = net->xfrm.state_hmask;
 144
 145         net->xfrm.state_bydst = ndst;
 146         net->xfrm.state_bysrc = nsrc;
 147         net->xfrm.state_byspi = nspi;
 148         net->xfrm.state_hmask = nhashmask;
 149
 150         spin_unlock_bh(&xfrm_state_lock);
 151
 152         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 153         xfrm_hash_free(odst, osize);
 154         xfrm_hash_free(osrc, osize);
 155         xfrm_hash_free(ospi, osize);
 156
 157 out_unlock:
 158         mutex_unlock(&hash_resize_mutex);
 159 }
 160
 161 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 162 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
 163
 164 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 165
 166 int __xfrm_state_delete(struct xfrm_state *x);
 167
 168 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 169 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 170
 171 static DEFINE_SPINLOCK(xfrm_type_lock);
 172 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 173 {
 174         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 175         const struct xfrm_type **typemap;
 176         int err = 0;
 177
 178         if (unlikely(afinfo == NULL))
 179                 return -EAFNOSUPPORT;
 180         typemap = afinfo->type_map;
 181         spin_lock_bh(&xfrm_type_lock);
 182
 183         if (likely(typemap[type->proto] == NULL))
 184                 typemap[type->proto] = type;
 185         else
 186                 err = -EEXIST;
 187         spin_unlock_bh(&xfrm_type_lock);
 188         xfrm_state_put_afinfo(afinfo);
 189         return err;
 190 }
 191 EXPORT_SYMBOL(xfrm_register_type);
 192
 193 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 194 {
 195         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 196         const struct xfrm_type **typemap;
 197         int err = 0;
 198
 199         if (unlikely(afinfo == NULL))
 200                 return -EAFNOSUPPORT;
 201         typemap = afinfo->type_map;
 202         spin_lock_bh(&xfrm_type_lock);
 203
 204         if (unlikely(typemap[type->proto] != type))
 205                 err = -ENOENT;
 206         else
 207                 typemap[type->proto] = NULL;
 208         spin_unlock_bh(&xfrm_type_lock);
 209         xfrm_state_put_afinfo(afinfo);
 210         return err;
 211 }
 212 EXPORT_SYMBOL(xfrm_unregister_type);
 213
 214 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 215 {
 216         struct xfrm_state_afinfo *afinfo;
 217         const struct xfrm_type **typemap;
 218         const struct xfrm_type *type;
 219         int modload_attempted = 0;
 220
 221 retry:
 222         afinfo = xfrm_state_get_afinfo(family);
 223         if (unlikely(afinfo == NULL))
 224                 return NULL;
 225         typemap = afinfo->type_map;
 226
 227         type = typemap[proto];
 228         if (unlikely(type && !try_module_get(type->owner)))
 229                 type = NULL;
 230         if (!type && !modload_attempted) {
 231                 xfrm_state_put_afinfo(afinfo);
 232                 request_module("xfrm-type-%d-%d", family, proto);
 233                 modload_attempted = 1;
 234                 goto retry;
 235         }
 236
 237         xfrm_state_put_afinfo(afinfo);
 238         return type;
 239 }
 240
 241 static void xfrm_put_type(const struct xfrm_type *type)
 242 {
 243         module_put(type->owner);
 244 }
 245
 246 static DEFINE_SPINLOCK(xfrm_mode_lock);
 247 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 248 {
 249         struct xfrm_state_afinfo *afinfo;
 250         struct xfrm_mode **modemap;
 251         int err;
 252
 253         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 254                 return -EINVAL;
 255
 256         afinfo = xfrm_state_get_afinfo(family);
 257         if (unlikely(afinfo == NULL))
 258                 return -EAFNOSUPPORT;
 259
 260         err = -EEXIST;
 261         modemap = afinfo->mode_map;
 262         spin_lock_bh(&xfrm_mode_lock);
 263         if (modemap[mode->encap])
 264                 goto out;
 265
 266         err = -ENOENT;
 267         if (!try_module_get(afinfo->owner))
 268                 goto out;
 269
 270         mode->afinfo = afinfo;
 271         modemap[mode->encap] = mode;
 272         err = 0;
 273
 274 out:
 275         spin_unlock_bh(&xfrm_mode_lock);
 276         xfrm_state_put_afinfo(afinfo);
 277         return err;
 278 }
 279 EXPORT_SYMBOL(xfrm_register_mode);
 280
 281 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 282 {
 283         struct xfrm_state_afinfo *afinfo;
 284         struct xfrm_mode **modemap;
 285         int err;
 286
 287         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 288                 return -EINVAL;
 289
 290         afinfo = xfrm_state_get_afinfo(family);
 291         if (unlikely(afinfo == NULL))
 292                 return -EAFNOSUPPORT;
 293
 294         err = -ENOENT;
 295         modemap = afinfo->mode_map;
 296         spin_lock_bh(&xfrm_mode_lock);
 297         if (likely(modemap[mode->encap] == mode)) {
 298                 modemap[mode->encap] = NULL;
 299                 module_put(mode->afinfo->owner);
 300                 err = 0;
 301         }
 302
 303         spin_unlock_bh(&xfrm_mode_lock);
 304         xfrm_state_put_afinfo(afinfo);
 305         return err;
 306 }
 307 EXPORT_SYMBOL(xfrm_unregister_mode);
 308
 309 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 310 {
 311         struct xfrm_state_afinfo *afinfo;
 312         struct xfrm_mode *mode;
 313         int modload_attempted = 0;
 314
 315         if (unlikely(encap >= XFRM_MODE_MAX))
 316                 return NULL;
 317
 318 retry:
 319         afinfo = xfrm_state_get_afinfo(family);
 320         if (unlikely(afinfo == NULL))
 321                 return NULL;
 322
 323         mode = afinfo->mode_map[encap];
 324         if (unlikely(mode && !try_module_get(mode->owner)))
 325                 mode = NULL;
 326         if (!mode && !modload_attempted) {
 327                 xfrm_state_put_afinfo(afinfo);
 328                 request_module("xfrm-mode-%d-%d", family, encap);
 329                 modload_attempted = 1;
 330                 goto retry;
 331         }
 332
 333         xfrm_state_put_afinfo(afinfo);
 334         return mode;
 335 }
 336
 337 static void xfrm_put_mode(struct xfrm_mode *mode)
 338 {
 339         module_put(mode->owner);
 340 }
 341
 342 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 343 {
 344         tasklet_hrtimer_cancel(&x->mtimer);
 345         del_timer_sync(&x->rtimer);
 346         kfree(x->aalg);
 347         kfree(x->ealg);
 348         kfree(x->calg);
 349         kfree(x->encap);
 350         kfree(x->coaddr);
 351         kfree(x->replay_esn);
 352         kfree(x->preplay_esn);
 353         if (x->inner_mode)
 354                 xfrm_put_mode(x->inner_mode);
 355         if (x->inner_mode_iaf)
 356                 xfrm_put_mode(x->inner_mode_iaf);
 357         if (x->outer_mode)
 358                 xfrm_put_mode(x->outer_mode);
 359         if (x->type) {
 360                 x->type->destructor(x);
 361                 xfrm_put_type(x->type);
 362         }
 363         security_xfrm_state_free(x);
 364         kfree(x);
 365 }
 366
 367 static void xfrm_state_gc_task(struct work_struct *work)
 368 {
 369         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 370         struct xfrm_state *x;
 371         struct hlist_node *tmp;
 372         struct hlist_head gc_list;
 373
 374         spin_lock_bh(&xfrm_state_gc_lock);
 375         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 376         spin_unlock_bh(&xfrm_state_gc_lock);
 377
 378         hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 379                 xfrm_state_gc_destroy(x);
 380
 381         wake_up(&net->xfrm.km_waitq);
 382 }
 383
 384 static inline unsigned long make_jiffies(long secs)
 385 {
 386         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 387                 return MAX_SCHEDULE_TIMEOUT-1;
 388         else
 389                 return secs*HZ;
 390 }
 391
 392 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 393 {
 394         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 395         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 396         struct net *net = xs_net(x);
 397         unsigned long now = get_seconds();
 398         long next = LONG_MAX;
 399         int warn = 0;
 400         int err = 0;
 401
 402         spin_lock(&x->lock);
 403         if (x->km.state == XFRM_STATE_DEAD)
 404                 goto out;
 405         if (x->km.state == XFRM_STATE_EXPIRED)
 406                 goto expired;
 407         if (x->lft.hard_add_expires_seconds) {
 408                 long tmo = x->lft.hard_add_expires_seconds +
 409                         x->curlft.add_time - now;
 410                 if (tmo <= 0) {
 411                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 412                                 /* enter hard expire without soft expire first?!
 413                                  * setting a new date could trigger this.
 414                                  * workarbound: fix x->curflt.add_time by below:
 415                                  */
 416                                 x->curlft.add_time = now - x->saved_tmo - 1;
 417                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 418                         } else
 419                                 goto expired;
 420                 }
 421                 if (tmo < next)
 422                         next = tmo;
 423         }
 424         if (x->lft.hard_use_expires_seconds) {
 425                 long tmo = x->lft.hard_use_expires_seconds +
 426                         (x->curlft.use_time ? : now) - now;
 427                 if (tmo <= 0)
 428                         goto expired;
 429                 if (tmo < next)
 430                         next = tmo;
 431         }
 432         if (x->km.dying)
 433                 goto resched;
 434         if (x->lft.soft_add_expires_seconds) {
 435                 long tmo = x->lft.soft_add_expires_seconds +
 436                         x->curlft.add_time - now;
 437                 if (tmo <= 0) {
 438                         warn = 1;
 439                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 440                 } else if (tmo < next) {
 441                         next = tmo;
 442                         x->xflags |= XFRM_SOFT_EXPIRE;
 443                         x->saved_tmo = tmo;
 444                 }
 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,
 503                                         CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
 504                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 505                                 (unsigned long)x);
 506                 x->curlft.add_time = get_seconds();
 507                 x->lft.soft_byte_limit = XFRM_INF;
 508                 x->lft.soft_packet_limit = XFRM_INF;
 509                 x->lft.hard_byte_limit = XFRM_INF;
 510                 x->lft.hard_packet_limit = XFRM_INF;
 511                 x->replay_maxage = 0;
 512                 x->replay_maxdiff = 0;
 513                 x->inner_mode = NULL;
 514                 x->inner_mode_iaf = NULL;
 515                 spin_lock_init(&x->lock);
 516         }
 517         return x;
 518 }
 519 EXPORT_SYMBOL(xfrm_state_alloc);
 520
 521 void __xfrm_state_destroy(struct xfrm_state *x)
 522 {
 523         struct net *net = xs_net(x);
 524
 525         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 526
 527         spin_lock_bh(&xfrm_state_gc_lock);
 528         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 529         spin_unlock_bh(&xfrm_state_gc_lock);
 530         schedule_work(&net->xfrm.state_gc_work);
 531 }
 532 EXPORT_SYMBOL(__xfrm_state_destroy);
 533
 534 int __xfrm_state_delete(struct xfrm_state *x)
 535 {
 536         struct net *net = xs_net(x);
 537         int err = -ESRCH;
 538
 539         if (x->km.state != XFRM_STATE_DEAD) {
 540                 x->km.state = XFRM_STATE_DEAD;
 541                 spin_lock(&xfrm_state_lock);
 542                 list_del(&x->km.all);
 543                 hlist_del(&x->bydst);
 544                 hlist_del(&x->bysrc);
 545                 if (x->id.spi)
 546                         hlist_del(&x->byspi);
 547                 net->xfrm.state_num--;
 548                 spin_unlock(&xfrm_state_lock);
 549
 550                 /* All xfrm_state objects are created by xfrm_state_alloc.
 551                  * The xfrm_state_alloc call gives a reference, and that
 552                  * is what we are dropping here.
 553                  */
 554                 xfrm_state_put(x);
 555                 err = 0;
 556         }
 557
 558         return err;
 559 }
 560 EXPORT_SYMBOL(__xfrm_state_delete);
 561
 562 int xfrm_state_delete(struct xfrm_state *x)
 563 {
 564         int err;
 565
 566         spin_lock_bh(&x->lock);
 567         err = __xfrm_state_delete(x);
 568         spin_unlock_bh(&x->lock);
 569
 570         return err;
 571 }
 572 EXPORT_SYMBOL(xfrm_state_delete);
 573
 574 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 575 static inline int
 576 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 577 {
 578         int i, err = 0;
 579
 580         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 581                 struct xfrm_state *x;
 582
 583                 hlist_for_each_entry(x, 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 xfrm_state *x;
 617 restart:
 618                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 619                         if (!xfrm_state_kern(x) &&
 620                             xfrm_id_proto_match(x->id.proto, proto)) {
 621                                 xfrm_state_hold(x);
 622                                 spin_unlock_bh(&xfrm_state_lock);
 623
 624                                 err = xfrm_state_delete(x);
 625                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 626                                                         audit_info->loginuid,
 627                                                         audit_info->sessionid,
 628                                                         audit_info->secid);
 629                                 xfrm_state_put(x);
 630                                 if (!err)
 631                                         cnt++;
 632
 633                                 spin_lock_bh(&xfrm_state_lock);
 634                                 goto restart;
 635                         }
 636                 }
 637         }
 638         if (cnt)
 639                 err = 0;
 640
 641 out:
 642         spin_unlock_bh(&xfrm_state_lock);
 643         wake_up(&net->xfrm.km_waitq);
 644         return err;
 645 }
 646 EXPORT_SYMBOL(xfrm_state_flush);
 647
 648 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 649 {
 650         spin_lock_bh(&xfrm_state_lock);
 651         si->sadcnt = net->xfrm.state_num;
 652         si->sadhcnt = net->xfrm.state_hmask;
 653         si->sadhmcnt = xfrm_state_hashmax;
 654         spin_unlock_bh(&xfrm_state_lock);
 655 }
 656 EXPORT_SYMBOL(xfrm_sad_getinfo);
 657
 658 static int
 659 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 660                     const struct xfrm_tmpl *tmpl,
 661                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 662                     unsigned short family)
 663 {
 664         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 665         if (!afinfo)
 666                 return -1;
 667         afinfo->init_tempsel(&x->sel, fl);
 668
 669         if (family != tmpl->encap_family) {
 670                 xfrm_state_put_afinfo(afinfo);
 671                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 672                 if (!afinfo)
 673                         return -1;
 674         }
 675         afinfo->init_temprop(x, tmpl, daddr, saddr);
 676         xfrm_state_put_afinfo(afinfo);
 677         return 0;
 678 }
 679
 680 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 681                                               const xfrm_address_t *daddr,
 682                                               __be32 spi, u8 proto,
 683                                               unsigned short family)
 684 {
 685         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 686         struct xfrm_state *x;
 687
 688         hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
 689                 if (x->props.family != family ||
 690                     x->id.spi       != spi ||
 691                     x->id.proto     != proto ||
 692                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 693                         continue;
 694
 695                 if ((mark & x->mark.m) != x->mark.v)
 696                         continue;
 697                 xfrm_state_hold(x);
 698                 return x;
 699         }
 700
 701         return NULL;
 702 }
 703
 704 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 705                                                      const xfrm_address_t *daddr,
 706                                                      const xfrm_address_t *saddr,
 707                                                      u8 proto, unsigned short family)
 708 {
 709         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 710         struct xfrm_state *x;
 711
 712         hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
 713                 if (x->props.family != family ||
 714                     x->id.proto     != proto ||
 715                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 716                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 717                         continue;
 718
 719                 if ((mark & x->mark.m) != x->mark.v)
 720                         continue;
 721                 xfrm_state_hold(x);
 722                 return x;
 723         }
 724
 725         return NULL;
 726 }
 727
 728 static inline struct xfrm_state *
 729 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 730 {
 731         struct net *net = xs_net(x);
 732         u32 mark = x->mark.v & x->mark.m;
 733
 734         if (use_spi)
 735                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 736                                            x->id.spi, x->id.proto, family);
 737         else
 738                 return __xfrm_state_lookup_byaddr(net, mark,
 739                                                   &x->id.daddr,
 740                                                   &x->props.saddr,
 741                                                   x->id.proto, family);
 742 }
 743
 744 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 745 {
 746         if (have_hash_collision &&
 747             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 748             net->xfrm.state_num > net->xfrm.state_hmask)
 749                 schedule_work(&net->xfrm.state_hash_work);
 750 }
 751
 752 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 753                                const struct flowi *fl, unsigned short family,
 754                                struct xfrm_state **best, int *acq_in_progress,
 755                                int *error)
 756 {
 757         /* Resolution logic:
 758          * 1. There is a valid state with matching selector. Done.
 759          * 2. Valid state with inappropriate selector. Skip.
 760          *
 761          * Entering area of "sysdeps".
 762          *
 763          * 3. If state is not valid, selector is temporary, it selects
 764          *    only session which triggered previous resolution. Key
 765          *    manager will do something to install a state with proper
 766          *    selector.
 767          */
 768         if (x->km.state == XFRM_STATE_VALID) {
 769                 if ((x->sel.family &&
 770                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 771                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 772                         return;
 773
 774                 if (!*best ||
 775                     (*best)->km.dying > x->km.dying ||
 776                     ((*best)->km.dying == x->km.dying &&
 777                      (*best)->curlft.add_time < x->curlft.add_time))
 778                         *best = x;
 779         } else if (x->km.state == XFRM_STATE_ACQ) {
 780                 *acq_in_progress = 1;
 781         } else if (x->km.state == XFRM_STATE_ERROR ||
 782                    x->km.state == XFRM_STATE_EXPIRED) {
 783                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 784                     security_xfrm_state_pol_flow_match(x, pol, fl))
 785                         *error = -ESRCH;
 786         }
 787 }
 788
 789 struct xfrm_state *
 790 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 791                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 792                 struct xfrm_policy *pol, int *err,
 793                 unsigned short family)
 794 {
 795         static xfrm_address_t saddr_wildcard = { };
 796         struct net *net = xp_net(pol);
 797         unsigned int h, h_wildcard;
 798         struct xfrm_state *x, *x0, *to_put;
 799         int acquire_in_progress = 0;
 800         int error = 0;
 801         struct xfrm_state *best = NULL;
 802         u32 mark = pol->mark.v & pol->mark.m;
 803         unsigned short encap_family = tmpl->encap_family;
 804
 805         to_put = NULL;
 806
 807         spin_lock_bh(&xfrm_state_lock);
 808         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 809         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 810                 if (x->props.family == encap_family &&
 811                     x->props.reqid == tmpl->reqid &&
 812                     (mark & x->mark.m) == x->mark.v &&
 813                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 814                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 815                     tmpl->mode == x->props.mode &&
 816                     tmpl->id.proto == x->id.proto &&
 817                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 818                         xfrm_state_look_at(pol, x, fl, encap_family,
 819                                            &best, &acquire_in_progress, &error);
 820         }
 821         if (best)
 822                 goto found;
 823
 824         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 825         hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
 826                 if (x->props.family == encap_family &&
 827                     x->props.reqid == tmpl->reqid &&
 828                     (mark & x->mark.m) == x->mark.v &&
 829                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 830                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 831                     tmpl->mode == x->props.mode &&
 832                     tmpl->id.proto == x->id.proto &&
 833                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 834                         xfrm_state_look_at(pol, x, fl, encap_family,
 835                                            &best, &acquire_in_progress, &error);
 836         }
 837
 838 found:
 839         x = best;
 840         if (!x && !error && !acquire_in_progress) {
 841                 if (tmpl->id.spi &&
 842                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 843                                               tmpl->id.proto, encap_family)) != NULL) {
 844                         to_put = x0;
 845                         error = -EEXIST;
 846                         goto out;
 847                 }
 848                 x = xfrm_state_alloc(net);
 849                 if (x == NULL) {
 850                         error = -ENOMEM;
 851                         goto out;
 852                 }
 853                 /* Initialize temporary state matching only
 854                  * to current session. */
 855                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 856                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 857
 858                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 859                 if (error) {
 860                         x->km.state = XFRM_STATE_DEAD;
 861                         to_put = x;
 862                         x = NULL;
 863                         goto out;
 864                 }
 865
 866                 if (km_query(x, tmpl, pol) == 0) {
 867                         x->km.state = XFRM_STATE_ACQ;
 868                         list_add(&x->km.all, &net->xfrm.state_all);
 869                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 870                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 871                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 872                         if (x->id.spi) {
 873                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 874                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 875                         }
 876                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 877                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 878                         net->xfrm.state_num++;
 879                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 880                 } else {
 881                         x->km.state = XFRM_STATE_DEAD;
 882                         to_put = x;
 883                         x = NULL;
 884                         error = -ESRCH;
 885                 }
 886         }
 887 out:
 888         if (x)
 889                 xfrm_state_hold(x);
 890         else
 891                 *err = acquire_in_progress ? -EAGAIN : error;
 892         spin_unlock_bh(&xfrm_state_lock);
 893         if (to_put)
 894                 xfrm_state_put(to_put);
 895         return x;
 896 }
 897
 898 struct xfrm_state *
 899 xfrm_stateonly_find(struct net *net, u32 mark,
 900                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 901                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 902 {
 903         unsigned int h;
 904         struct xfrm_state *rx = NULL, *x = NULL;
 905
 906         spin_lock(&xfrm_state_lock);
 907         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 908         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 909                 if (x->props.family == family &&
 910                     x->props.reqid == reqid &&
 911                     (mark & x->mark.m) == x->mark.v &&
 912                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 913                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 914                     mode == x->props.mode &&
 915                     proto == x->id.proto &&
 916                     x->km.state == XFRM_STATE_VALID) {
 917                         rx = x;
 918                         break;
 919                 }
 920         }
 921
 922         if (rx)
 923                 xfrm_state_hold(rx);
 924         spin_unlock(&xfrm_state_lock);
 925
 926
 927         return rx;
 928 }
 929 EXPORT_SYMBOL(xfrm_stateonly_find);
 930
 931 static void __xfrm_state_insert(struct xfrm_state *x)
 932 {
 933         struct net *net = xs_net(x);
 934         unsigned int h;
 935
 936         list_add(&x->km.all, &net->xfrm.state_all);
 937
 938         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 939                           x->props.reqid, x->props.family);
 940         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 941
 942         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 943         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 944
 945         if (x->id.spi) {
 946                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 947                                   x->props.family);
 948
 949                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 950         }
 951
 952         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 953         if (x->replay_maxage)
 954                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 955
 956         wake_up(&net->xfrm.km_waitq);
 957
 958         net->xfrm.state_num++;
 959
 960         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 961 }
 962
 963 /* xfrm_state_lock is held */
 964 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 965 {
 966         struct net *net = xs_net(xnew);
 967         unsigned short family = xnew->props.family;
 968         u32 reqid = xnew->props.reqid;
 969         struct xfrm_state *x;
 970         unsigned int h;
 971         u32 mark = xnew->mark.v & xnew->mark.m;
 972
 973         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 974         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 975                 if (x->props.family     == family &&
 976                     x->props.reqid      == reqid &&
 977                     (mark & x->mark.m) == x->mark.v &&
 978                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 979                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
 980                         x->genid++;
 981         }
 982 }
 983
 984 void xfrm_state_insert(struct xfrm_state *x)
 985 {
 986         spin_lock_bh(&xfrm_state_lock);
 987         __xfrm_state_bump_genids(x);
 988         __xfrm_state_insert(x);
 989         spin_unlock_bh(&xfrm_state_lock);
 990 }
 991 EXPORT_SYMBOL(xfrm_state_insert);
 992
 993 /* xfrm_state_lock is held */
 994 static struct xfrm_state *__find_acq_core(struct net *net,
 995                                           const struct xfrm_mark *m,
 996                                           unsigned short family, u8 mode,
 997                                           u32 reqid, u8 proto,
 998                                           const xfrm_address_t *daddr,
 999                                           const xfrm_address_t *saddr,
1000                                           int create)
1001 {
1002         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1003         struct xfrm_state *x;
1004         u32 mark = m->v & m->m;
1005
1006         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1007                 if (x->props.reqid  != reqid ||
1008                     x->props.mode   != mode ||
1009                     x->props.family != family ||
1010                     x->km.state     != XFRM_STATE_ACQ ||
1011                     x->id.spi       != 0 ||
1012                     x->id.proto     != proto ||
1013                     (mark & x->mark.m) != x->mark.v ||
1014                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1015                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1016                         continue;
1017
1018                 xfrm_state_hold(x);
1019                 return x;
1020         }
1021
1022         if (!create)
1023                 return NULL;
1024
1025         x = xfrm_state_alloc(net);
1026         if (likely(x)) {
1027                 switch (family) {
1028                 case AF_INET:
1029                         x->sel.daddr.a4 = daddr->a4;
1030                         x->sel.saddr.a4 = saddr->a4;
1031                         x->sel.prefixlen_d = 32;
1032                         x->sel.prefixlen_s = 32;
1033                         x->props.saddr.a4 = saddr->a4;
1034                         x->id.daddr.a4 = daddr->a4;
1035                         break;
1036
1037                 case AF_INET6:
1038                         *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1039                         *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1040                         x->sel.prefixlen_d = 128;
1041                         x->sel.prefixlen_s = 128;
1042                         *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1043                         *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1044                         break;
1045                 }
1046
1047                 x->km.state = XFRM_STATE_ACQ;
1048                 x->id.proto = proto;
1049                 x->props.family = family;
1050                 x->props.mode = mode;
1051                 x->props.reqid = reqid;
1052                 x->mark.v = m->v;
1053                 x->mark.m = m->m;
1054                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1055                 xfrm_state_hold(x);
1056                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1057                 list_add(&x->km.all, &net->xfrm.state_all);
1058                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1059                 h = xfrm_src_hash(net, daddr, saddr, family);
1060                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1061
1062                 net->xfrm.state_num++;
1063
1064                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1065         }
1066
1067         return x;
1068 }
1069
1070 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1071
1072 int xfrm_state_add(struct xfrm_state *x)
1073 {
1074         struct net *net = xs_net(x);
1075         struct xfrm_state *x1, *to_put;
1076         int family;
1077         int err;
1078         u32 mark = x->mark.v & x->mark.m;
1079         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1080
1081         family = x->props.family;
1082
1083         to_put = NULL;
1084
1085         spin_lock_bh(&xfrm_state_lock);
1086
1087         x1 = __xfrm_state_locate(x, use_spi, family);
1088         if (x1) {
1089                 to_put = x1;
1090                 x1 = NULL;
1091                 err = -EEXIST;
1092                 goto out;
1093         }
1094
1095         if (use_spi && x->km.seq) {
1096                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1097                 if (x1 && ((x1->id.proto != x->id.proto) ||
1098                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1099                         to_put = x1;
1100                         x1 = NULL;
1101                 }
1102         }
1103
1104         if (use_spi && !x1)
1105                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1106                                      x->props.reqid, x->id.proto,
1107                                      &x->id.daddr, &x->props.saddr, 0);
1108
1109         __xfrm_state_bump_genids(x);
1110         __xfrm_state_insert(x);
1111         err = 0;
1112
1113 out:
1114         spin_unlock_bh(&xfrm_state_lock);
1115
1116         if (x1) {
1117                 xfrm_state_delete(x1);
1118                 xfrm_state_put(x1);
1119         }
1120
1121         if (to_put)
1122                 xfrm_state_put(to_put);
1123
1124         return err;
1125 }
1126 EXPORT_SYMBOL(xfrm_state_add);
1127
1128 #ifdef CONFIG_XFRM_MIGRATE
1129 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1130 {
1131         struct net *net = xs_net(orig);
1132         int err = -ENOMEM;
1133         struct xfrm_state *x = xfrm_state_alloc(net);
1134         if (!x)
1135                 goto out;
1136
1137         memcpy(&x->id, &orig->id, sizeof(x->id));
1138         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1139         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1140         x->props.mode = orig->props.mode;
1141         x->props.replay_window = orig->props.replay_window;
1142         x->props.reqid = orig->props.reqid;
1143         x->props.family = orig->props.family;
1144         x->props.saddr = orig->props.saddr;
1145
1146         if (orig->aalg) {
1147                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1148                 if (!x->aalg)
1149                         goto error;
1150         }
1151         x->props.aalgo = orig->props.aalgo;
1152
1153         if (orig->ealg) {
1154                 x->ealg = xfrm_algo_clone(orig->ealg);
1155                 if (!x->ealg)
1156                         goto error;
1157         }
1158         x->props.ealgo = orig->props.ealgo;
1159
1160         if (orig->calg) {
1161                 x->calg = xfrm_algo_clone(orig->calg);
1162                 if (!x->calg)
1163                         goto error;
1164         }
1165         x->props.calgo = orig->props.calgo;
1166
1167         if (orig->encap) {
1168                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1169                 if (!x->encap)
1170                         goto error;
1171         }
1172
1173         if (orig->coaddr) {
1174                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1175                                     GFP_KERNEL);
1176                 if (!x->coaddr)
1177                         goto error;
1178         }
1179
1180         if (orig->replay_esn) {
1181                 err = xfrm_replay_clone(x, orig);
1182                 if (err)
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         x->props.extra_flags = orig->props.extra_flags;
1194
1195         x->curlft.add_time = orig->curlft.add_time;
1196         x->km.state = orig->km.state;
1197         x->km.seq = orig->km.seq;
1198
1199         return x;
1200
1201  error:
1202         xfrm_state_put(x);
1203 out:
1204         if (errp)
1205                 *errp = err;
1206         return NULL;
1207 }
1208
1209 /* xfrm_state_lock is held */
1210 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1211 {
1212         unsigned int h;
1213         struct xfrm_state *x;
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, 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_equal(&x->id.daddr, &m->old_daddr,
1225                                              m->old_family) ||
1226                             !xfrm_addr_equal(&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, 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_equal(&x->id.daddr, &m->old_daddr,
1240                                              m->old_family) ||
1241                             !xfrm_addr_equal(&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_equal(&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                 x->km.state = XFRM_STATE_DEAD;
1345                 __xfrm_state_put(x);
1346         }
1347         spin_unlock_bh(&x1->lock);
1348
1349         xfrm_state_put(x1);
1350
1351         return err;
1352 }
1353 EXPORT_SYMBOL(xfrm_state_update);
1354
1355 int xfrm_state_check_expire(struct xfrm_state *x)
1356 {
1357         if (!x->curlft.use_time)
1358                 x->curlft.use_time = get_seconds();
1359
1360         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1361             x->curlft.packets >= x->lft.hard_packet_limit) {
1362                 x->km.state = XFRM_STATE_EXPIRED;
1363                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1364                 return -EINVAL;
1365         }
1366
1367         if (!x->km.dying &&
1368             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1369              x->curlft.packets >= x->lft.soft_packet_limit)) {
1370                 x->km.dying = 1;
1371                 km_state_expired(x, 0, 0);
1372         }
1373         return 0;
1374 }
1375 EXPORT_SYMBOL(xfrm_state_check_expire);
1376
1377 struct xfrm_state *
1378 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1379                   u8 proto, unsigned short family)
1380 {
1381         struct xfrm_state *x;
1382
1383         spin_lock_bh(&xfrm_state_lock);
1384         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1385         spin_unlock_bh(&xfrm_state_lock);
1386         return x;
1387 }
1388 EXPORT_SYMBOL(xfrm_state_lookup);
1389
1390 struct xfrm_state *
1391 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1392                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1393                          u8 proto, unsigned short family)
1394 {
1395         struct xfrm_state *x;
1396
1397         spin_lock_bh(&xfrm_state_lock);
1398         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1399         spin_unlock_bh(&xfrm_state_lock);
1400         return x;
1401 }
1402 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1403
1404 struct xfrm_state *
1405 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1406               u8 proto, const xfrm_address_t *daddr,
1407               const xfrm_address_t *saddr, int create, unsigned short family)
1408 {
1409         struct xfrm_state *x;
1410
1411         spin_lock_bh(&xfrm_state_lock);
1412         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1413         spin_unlock_bh(&xfrm_state_lock);
1414
1415         return x;
1416 }
1417 EXPORT_SYMBOL(xfrm_find_acq);
1418
1419 #ifdef CONFIG_XFRM_SUB_POLICY
1420 int
1421 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1422                unsigned short family)
1423 {
1424         int err = 0;
1425         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1426         if (!afinfo)
1427                 return -EAFNOSUPPORT;
1428
1429         spin_lock_bh(&xfrm_state_lock);
1430         if (afinfo->tmpl_sort)
1431                 err = afinfo->tmpl_sort(dst, src, n);
1432         spin_unlock_bh(&xfrm_state_lock);
1433         xfrm_state_put_afinfo(afinfo);
1434         return err;
1435 }
1436 EXPORT_SYMBOL(xfrm_tmpl_sort);
1437
1438 int
1439 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1440                 unsigned short family)
1441 {
1442         int err = 0;
1443         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1444         if (!afinfo)
1445                 return -EAFNOSUPPORT;
1446
1447         spin_lock_bh(&xfrm_state_lock);
1448         if (afinfo->state_sort)
1449                 err = afinfo->state_sort(dst, src, n);
1450         spin_unlock_bh(&xfrm_state_lock);
1451         xfrm_state_put_afinfo(afinfo);
1452         return err;
1453 }
1454 EXPORT_SYMBOL(xfrm_state_sort);
1455 #endif
1456
1457 /* Silly enough, but I'm lazy to build resolution list */
1458
1459 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1460 {
1461         int i;
1462
1463         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1464                 struct xfrm_state *x;
1465
1466                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1467                         if (x->km.seq == seq &&
1468                             (mark & x->mark.m) == x->mark.v &&
1469                             x->km.state == XFRM_STATE_ACQ) {
1470                                 xfrm_state_hold(x);
1471                                 return x;
1472                         }
1473                 }
1474         }
1475         return NULL;
1476 }
1477
1478 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1479 {
1480         struct xfrm_state *x;
1481
1482         spin_lock_bh(&xfrm_state_lock);
1483         x = __xfrm_find_acq_byseq(net, mark, seq);
1484         spin_unlock_bh(&xfrm_state_lock);
1485         return x;
1486 }
1487 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1488
1489 u32 xfrm_get_acqseq(void)
1490 {
1491         u32 res;
1492         static atomic_t acqseq;
1493
1494         do {
1495                 res = atomic_inc_return(&acqseq);
1496         } while (!res);
1497
1498         return res;
1499 }
1500 EXPORT_SYMBOL(xfrm_get_acqseq);
1501
1502 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1503 {
1504         struct net *net = xs_net(x);
1505         unsigned int h;
1506         struct xfrm_state *x0;
1507         int err = -ENOENT;
1508         __be32 minspi = htonl(low);
1509         __be32 maxspi = htonl(high);
1510         u32 mark = x->mark.v & x->mark.m;
1511
1512         spin_lock_bh(&x->lock);
1513         if (x->km.state == XFRM_STATE_DEAD)
1514                 goto unlock;
1515
1516         err = 0;
1517         if (x->id.spi)
1518                 goto unlock;
1519
1520         err = -ENOENT;
1521
1522         if (minspi == maxspi) {
1523                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1524                 if (x0) {
1525                         xfrm_state_put(x0);
1526                         goto unlock;
1527                 }
1528                 x->id.spi = minspi;
1529         } else {
1530                 u32 spi = 0;
1531                 for (h=0; h<high-low+1; h++) {
1532                         spi = low + net_random()%(high-low+1);
1533                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1534                         if (x0 == NULL) {
1535                                 x->id.spi = htonl(spi);
1536                                 break;
1537                         }
1538                         xfrm_state_put(x0);
1539                 }
1540         }
1541         if (x->id.spi) {
1542                 spin_lock_bh(&xfrm_state_lock);
1543                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1544                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1545                 spin_unlock_bh(&xfrm_state_lock);
1546
1547                 err = 0;
1548         }
1549
1550 unlock:
1551         spin_unlock_bh(&x->lock);
1552
1553         return err;
1554 }
1555 EXPORT_SYMBOL(xfrm_alloc_spi);
1556
1557 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1558                     int (*func)(struct xfrm_state *, int, void*),
1559                     void *data)
1560 {
1561         struct xfrm_state *state;
1562         struct xfrm_state_walk *x;
1563         int err = 0;
1564
1565         if (walk->seq != 0 && list_empty(&walk->all))
1566                 return 0;
1567
1568         spin_lock_bh(&xfrm_state_lock);
1569         if (list_empty(&walk->all))
1570                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1571         else
1572                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1573         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1574                 if (x->state == XFRM_STATE_DEAD)
1575                         continue;
1576                 state = container_of(x, struct xfrm_state, km);
1577                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1578                         continue;
1579                 err = func(state, walk->seq, data);
1580                 if (err) {
1581                         list_move_tail(&walk->all, &x->all);
1582                         goto out;
1583                 }
1584                 walk->seq++;
1585         }
1586         if (walk->seq == 0) {
1587                 err = -ENOENT;
1588                 goto out;
1589         }
1590         list_del_init(&walk->all);
1591 out:
1592         spin_unlock_bh(&xfrm_state_lock);
1593         return err;
1594 }
1595 EXPORT_SYMBOL(xfrm_state_walk);
1596
1597 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1598 {
1599         INIT_LIST_HEAD(&walk->all);
1600         walk->proto = proto;
1601         walk->state = XFRM_STATE_DEAD;
1602         walk->seq = 0;
1603 }
1604 EXPORT_SYMBOL(xfrm_state_walk_init);
1605
1606 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1607 {
1608         if (list_empty(&walk->all))
1609                 return;
1610
1611         spin_lock_bh(&xfrm_state_lock);
1612         list_del(&walk->all);
1613         spin_unlock_bh(&xfrm_state_lock);
1614 }
1615 EXPORT_SYMBOL(xfrm_state_walk_done);
1616
1617 static void xfrm_replay_timer_handler(unsigned long data)
1618 {
1619         struct xfrm_state *x = (struct xfrm_state*)data;
1620
1621         spin_lock(&x->lock);
1622
1623         if (x->km.state == XFRM_STATE_VALID) {
1624                 if (xfrm_aevent_is_on(xs_net(x)))
1625                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1626                 else
1627                         x->xflags |= XFRM_TIME_DEFER;
1628         }
1629
1630         spin_unlock(&x->lock);
1631 }
1632
1633 static LIST_HEAD(xfrm_km_list);
1634
1635 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1636 {
1637         struct xfrm_mgr *km;
1638
1639         rcu_read_lock();
1640         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1641                 if (km->notify_policy)
1642                         km->notify_policy(xp, dir, c);
1643         rcu_read_unlock();
1644 }
1645
1646 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1647 {
1648         struct xfrm_mgr *km;
1649         rcu_read_lock();
1650         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1651                 if (km->notify)
1652                         km->notify(x, c);
1653         rcu_read_unlock();
1654 }
1655
1656 EXPORT_SYMBOL(km_policy_notify);
1657 EXPORT_SYMBOL(km_state_notify);
1658
1659 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1660 {
1661         struct net *net = xs_net(x);
1662         struct km_event c;
1663
1664         c.data.hard = hard;
1665         c.portid = portid;
1666         c.event = XFRM_MSG_EXPIRE;
1667         km_state_notify(x, &c);
1668
1669         if (hard)
1670                 wake_up(&net->xfrm.km_waitq);
1671 }
1672
1673 EXPORT_SYMBOL(km_state_expired);
1674 /*
1675  * We send to all registered managers regardless of failure
1676  * We are happy with one success
1677 */
1678 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1679 {
1680         int err = -EINVAL, acqret;
1681         struct xfrm_mgr *km;
1682
1683         rcu_read_lock();
1684         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1685                 acqret = km->acquire(x, t, pol);
1686                 if (!acqret)
1687                         err = acqret;
1688         }
1689         rcu_read_unlock();
1690         return err;
1691 }
1692 EXPORT_SYMBOL(km_query);
1693
1694 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1695 {
1696         int err = -EINVAL;
1697         struct xfrm_mgr *km;
1698
1699         rcu_read_lock();
1700         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1701                 if (km->new_mapping)
1702                         err = km->new_mapping(x, ipaddr, sport);
1703                 if (!err)
1704                         break;
1705         }
1706         rcu_read_unlock();
1707         return err;
1708 }
1709 EXPORT_SYMBOL(km_new_mapping);
1710
1711 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1712 {
1713         struct net *net = xp_net(pol);
1714         struct km_event c;
1715
1716         c.data.hard = hard;
1717         c.portid = portid;
1718         c.event = XFRM_MSG_POLEXPIRE;
1719         km_policy_notify(pol, dir, &c);
1720
1721         if (hard)
1722                 wake_up(&net->xfrm.km_waitq);
1723 }
1724 EXPORT_SYMBOL(km_policy_expired);
1725
1726 #ifdef CONFIG_XFRM_MIGRATE
1727 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1728                const struct xfrm_migrate *m, int num_migrate,
1729                const struct xfrm_kmaddress *k)
1730 {
1731         int err = -EINVAL;
1732         int ret;
1733         struct xfrm_mgr *km;
1734
1735         rcu_read_lock();
1736         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1737                 if (km->migrate) {
1738                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1739                         if (!ret)
1740                                 err = ret;
1741                 }
1742         }
1743         rcu_read_unlock();
1744         return err;
1745 }
1746 EXPORT_SYMBOL(km_migrate);
1747 #endif
1748
1749 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1750 {
1751         int err = -EINVAL;
1752         int ret;
1753         struct xfrm_mgr *km;
1754
1755         rcu_read_lock();
1756         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1757                 if (km->report) {
1758                         ret = km->report(net, proto, sel, addr);
1759                         if (!ret)
1760                                 err = ret;
1761                 }
1762         }
1763         rcu_read_unlock();
1764         return err;
1765 }
1766 EXPORT_SYMBOL(km_report);
1767
1768 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1769 {
1770         int err;
1771         u8 *data;
1772         struct xfrm_mgr *km;
1773         struct xfrm_policy *pol = NULL;
1774
1775         if (optlen <= 0 || optlen > PAGE_SIZE)
1776                 return -EMSGSIZE;
1777
1778         data = kmalloc(optlen, GFP_KERNEL);
1779         if (!data)
1780                 return -ENOMEM;
1781
1782         err = -EFAULT;
1783         if (copy_from_user(data, optval, optlen))
1784                 goto out;
1785
1786         err = -EINVAL;
1787         rcu_read_lock();
1788         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1789                 pol = km->compile_policy(sk, optname, data,
1790                                          optlen, &err);
1791                 if (err >= 0)
1792                         break;
1793         }
1794         rcu_read_unlock();
1795
1796         if (err >= 0) {
1797                 xfrm_sk_policy_insert(sk, err, pol);
1798                 xfrm_pol_put(pol);
1799                 err = 0;
1800         }
1801
1802 out:
1803         kfree(data);
1804         return err;
1805 }
1806 EXPORT_SYMBOL(xfrm_user_policy);
1807
1808 static DEFINE_SPINLOCK(xfrm_km_lock);
1809
1810 int xfrm_register_km(struct xfrm_mgr *km)
1811 {
1812         spin_lock_bh(&xfrm_km_lock);
1813         list_add_tail_rcu(&km->list, &xfrm_km_list);
1814         spin_unlock_bh(&xfrm_km_lock);
1815         return 0;
1816 }
1817 EXPORT_SYMBOL(xfrm_register_km);
1818
1819 int xfrm_unregister_km(struct xfrm_mgr *km)
1820 {
1821         spin_lock_bh(&xfrm_km_lock);
1822         list_del_rcu(&km->list);
1823         spin_unlock_bh(&xfrm_km_lock);
1824         synchronize_rcu();
1825         return 0;
1826 }
1827 EXPORT_SYMBOL(xfrm_unregister_km);
1828
1829 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1830 {
1831         int err = 0;
1832         if (unlikely(afinfo == NULL))
1833                 return -EINVAL;
1834         if (unlikely(afinfo->family >= NPROTO))
1835                 return -EAFNOSUPPORT;
1836         spin_lock_bh(&xfrm_state_afinfo_lock);
1837         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1838                 err = -ENOBUFS;
1839         else
1840                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1841         spin_unlock_bh(&xfrm_state_afinfo_lock);
1842         return err;
1843 }
1844 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1845
1846 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1847 {
1848         int err = 0;
1849         if (unlikely(afinfo == NULL))
1850                 return -EINVAL;
1851         if (unlikely(afinfo->family >= NPROTO))
1852                 return -EAFNOSUPPORT;
1853         spin_lock_bh(&xfrm_state_afinfo_lock);
1854         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1855                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1856                         err = -EINVAL;
1857                 else
1858                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1859         }
1860         spin_unlock_bh(&xfrm_state_afinfo_lock);
1861         synchronize_rcu();
1862         return err;
1863 }
1864 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1865
1866 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1867 {
1868         struct xfrm_state_afinfo *afinfo;
1869         if (unlikely(family >= NPROTO))
1870                 return NULL;
1871         rcu_read_lock();
1872         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1873         if (unlikely(!afinfo))
1874                 rcu_read_unlock();
1875         return afinfo;
1876 }
1877
1878 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1879 {
1880         rcu_read_unlock();
1881 }
1882
1883 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1884 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1885 {
1886         if (x->tunnel) {
1887                 struct xfrm_state *t = x->tunnel;
1888
1889                 if (atomic_read(&t->tunnel_users) == 2)
1890                         xfrm_state_delete(t);
1891                 atomic_dec(&t->tunnel_users);
1892                 xfrm_state_put(t);
1893                 x->tunnel = NULL;
1894         }
1895 }
1896 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1897
1898 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1899 {
1900         int res;
1901
1902         spin_lock_bh(&x->lock);
1903         if (x->km.state == XFRM_STATE_VALID &&
1904             x->type && x->type->get_mtu)
1905                 res = x->type->get_mtu(x, mtu);
1906         else
1907                 res = mtu - x->props.header_len;
1908         spin_unlock_bh(&x->lock);
1909         return res;
1910 }
1911
1912 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1913 {
1914         struct xfrm_state_afinfo *afinfo;
1915         struct xfrm_mode *inner_mode;
1916         int family = x->props.family;
1917         int err;
1918
1919         err = -EAFNOSUPPORT;
1920         afinfo = xfrm_state_get_afinfo(family);
1921         if (!afinfo)
1922                 goto error;
1923
1924         err = 0;
1925         if (afinfo->init_flags)
1926                 err = afinfo->init_flags(x);
1927
1928         xfrm_state_put_afinfo(afinfo);
1929
1930         if (err)
1931                 goto error;
1932
1933         err = -EPROTONOSUPPORT;
1934
1935         if (x->sel.family != AF_UNSPEC) {
1936                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1937                 if (inner_mode == NULL)
1938                         goto error;
1939
1940                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1941                     family != x->sel.family) {
1942                         xfrm_put_mode(inner_mode);
1943                         goto error;
1944                 }
1945
1946                 x->inner_mode = inner_mode;
1947         } else {
1948                 struct xfrm_mode *inner_mode_iaf;
1949                 int iafamily = AF_INET;
1950
1951                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1952                 if (inner_mode == NULL)
1953                         goto error;
1954
1955                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1956                         xfrm_put_mode(inner_mode);
1957                         goto error;
1958                 }
1959                 x->inner_mode = inner_mode;
1960
1961                 if (x->props.family == AF_INET)
1962                         iafamily = AF_INET6;
1963
1964                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1965                 if (inner_mode_iaf) {
1966                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1967                                 x->inner_mode_iaf = inner_mode_iaf;
1968                         else
1969                                 xfrm_put_mode(inner_mode_iaf);
1970                 }
1971         }
1972
1973         x->type = xfrm_get_type(x->id.proto, family);
1974         if (x->type == NULL)
1975                 goto error;
1976
1977         err = x->type->init_state(x);
1978         if (err)
1979                 goto error;
1980
1981         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1982         if (x->outer_mode == NULL) {
1983                 err = -EPROTONOSUPPORT;
1984                 goto error;
1985         }
1986
1987         if (init_replay) {
1988                 err = xfrm_init_replay(x);
1989                 if (err)
1990                         goto error;
1991         }
1992
1993         x->km.state = XFRM_STATE_VALID;
1994
1995 error:
1996         return err;
1997 }
1998
1999 EXPORT_SYMBOL(__xfrm_init_state);
2000
2001 int xfrm_init_state(struct xfrm_state *x)
2002 {
2003         return __xfrm_init_state(x, true);
2004 }
2005
2006 EXPORT_SYMBOL(xfrm_init_state);
2007
2008 int __net_init xfrm_state_init(struct net *net)
2009 {
2010         unsigned int sz;
2011
2012         INIT_LIST_HEAD(&net->xfrm.state_all);
2013
2014         sz = sizeof(struct hlist_head) * 8;
2015
2016         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2017         if (!net->xfrm.state_bydst)
2018                 goto out_bydst;
2019         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2020         if (!net->xfrm.state_bysrc)
2021                 goto out_bysrc;
2022         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2023         if (!net->xfrm.state_byspi)
2024                 goto out_byspi;
2025         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2026
2027         net->xfrm.state_num = 0;
2028         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2029         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2030         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2031         init_waitqueue_head(&net->xfrm.km_waitq);
2032         return 0;
2033
2034 out_byspi:
2035         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2036 out_bysrc:
2037         xfrm_hash_free(net->xfrm.state_bydst, sz);
2038 out_bydst:
2039         return -ENOMEM;
2040 }
2041
2042 void xfrm_state_fini(struct net *net)
2043 {
2044         struct xfrm_audit audit_info;
2045         unsigned int sz;
2046
2047         flush_work(&net->xfrm.state_hash_work);
2048         audit_info.loginuid = INVALID_UID;
2049         audit_info.sessionid = -1;
2050         audit_info.secid = 0;
2051         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2052         flush_work(&net->xfrm.state_gc_work);
2053
2054         WARN_ON(!list_empty(&net->xfrm.state_all));
2055
2056         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2057         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2058         xfrm_hash_free(net->xfrm.state_byspi, sz);
2059         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2060         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2061         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2062         xfrm_hash_free(net->xfrm.state_bydst, sz);
2063 }
2064
2065 #ifdef CONFIG_AUDITSYSCALL
2066 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2067                                      struct audit_buffer *audit_buf)
2068 {
2069         struct xfrm_sec_ctx *ctx = x->security;
2070         u32 spi = ntohl(x->id.spi);
2071
2072         if (ctx)
2073                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2074                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2075
2076         switch(x->props.family) {
2077         case AF_INET:
2078                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2079                                  &x->props.saddr.a4, &x->id.daddr.a4);
2080                 break;
2081         case AF_INET6:
2082                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2083                                  x->props.saddr.a6, x->id.daddr.a6);
2084                 break;
2085         }
2086
2087         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2088 }
2089
2090 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2091                                       struct audit_buffer *audit_buf)
2092 {
2093         const struct iphdr *iph4;
2094         const struct ipv6hdr *iph6;
2095
2096         switch (family) {
2097         case AF_INET:
2098                 iph4 = ip_hdr(skb);
2099                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2100                                  &iph4->saddr, &iph4->daddr);
2101                 break;
2102         case AF_INET6:
2103                 iph6 = ipv6_hdr(skb);
2104                 audit_log_format(audit_buf,
2105                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2106                                  &iph6->saddr,&iph6->daddr,
2107                                  iph6->flow_lbl[0] & 0x0f,
2108                                  iph6->flow_lbl[1],
2109                                  iph6->flow_lbl[2]);
2110                 break;
2111         }
2112 }
2113
2114 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2115                           kuid_t auid, u32 sessionid, u32 secid)
2116 {
2117         struct audit_buffer *audit_buf;
2118
2119         audit_buf = xfrm_audit_start("SAD-add");
2120         if (audit_buf == NULL)
2121                 return;
2122         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2123         xfrm_audit_helper_sainfo(x, audit_buf);
2124         audit_log_format(audit_buf, " res=%u", result);
2125         audit_log_end(audit_buf);
2126 }
2127 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2128
2129 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2130                              kuid_t auid, u32 sessionid, u32 secid)
2131 {
2132         struct audit_buffer *audit_buf;
2133
2134         audit_buf = xfrm_audit_start("SAD-delete");
2135         if (audit_buf == NULL)
2136                 return;
2137         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2138         xfrm_audit_helper_sainfo(x, audit_buf);
2139         audit_log_format(audit_buf, " res=%u", result);
2140         audit_log_end(audit_buf);
2141 }
2142 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2143
2144 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2145                                       struct sk_buff *skb)
2146 {
2147         struct audit_buffer *audit_buf;
2148         u32 spi;
2149
2150         audit_buf = xfrm_audit_start("SA-replay-overflow");
2151         if (audit_buf == NULL)
2152                 return;
2153         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2154         /* don't record the sequence number because it's inherent in this kind
2155          * of audit message */
2156         spi = ntohl(x->id.spi);
2157         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158         audit_log_end(audit_buf);
2159 }
2160 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2161
2162 void xfrm_audit_state_replay(struct xfrm_state *x,
2163                              struct sk_buff *skb, __be32 net_seq)
2164 {
2165         struct audit_buffer *audit_buf;
2166         u32 spi;
2167
2168         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2169         if (audit_buf == NULL)
2170                 return;
2171         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2172         spi = ntohl(x->id.spi);
2173         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2174                          spi, spi, ntohl(net_seq));
2175         audit_log_end(audit_buf);
2176 }
2177 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2178
2179 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2180 {
2181         struct audit_buffer *audit_buf;
2182
2183         audit_buf = xfrm_audit_start("SA-notfound");
2184         if (audit_buf == NULL)
2185                 return;
2186         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2187         audit_log_end(audit_buf);
2188 }
2189 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2190
2191 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2192                                __be32 net_spi, __be32 net_seq)
2193 {
2194         struct audit_buffer *audit_buf;
2195         u32 spi;
2196
2197         audit_buf = xfrm_audit_start("SA-notfound");
2198         if (audit_buf == NULL)
2199                 return;
2200         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2201         spi = ntohl(net_spi);
2202         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2203                          spi, spi, ntohl(net_seq));
2204         audit_log_end(audit_buf);
2205 }
2206 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2207
2208 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2209                               struct sk_buff *skb, u8 proto)
2210 {
2211         struct audit_buffer *audit_buf;
2212         __be32 net_spi;
2213         __be32 net_seq;
2214
2215         audit_buf = xfrm_audit_start("SA-icv-failure");
2216         if (audit_buf == NULL)
2217                 return;
2218         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2219         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2220                 u32 spi = ntohl(net_spi);
2221                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2222                                  spi, spi, ntohl(net_seq));
2223         }
2224         audit_log_end(audit_buf);
2225 }
2226 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2227 #endif /* CONFIG_AUDITSYSCALL */