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 inline unsigned int xfrm_dst_hash(struct net *net,
  43                                          const xfrm_address_t *daddr,
  44                                          const xfrm_address_t *saddr,
  45                                          u32 reqid,
  46                                          unsigned short family)
  47 {
  48         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  49 }
  50
  51 static inline unsigned int xfrm_src_hash(struct net *net,
  52                                          const xfrm_address_t *daddr,
  53                                          const xfrm_address_t *saddr,
  54                                          unsigned short family)
  55 {
  56         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  57 }
  58
  59 static inline unsigned int
  60 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  61               __be32 spi, u8 proto, unsigned short family)
  62 {
  63         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  64 }
  65
  66 static void xfrm_hash_transfer(struct hlist_head *list,
  67                                struct hlist_head *ndsttable,
  68                                struct hlist_head *nsrctable,
  69                                struct hlist_head *nspitable,
  70                                unsigned int nhashmask)
  71 {
  72         struct hlist_node *tmp;
  73         struct xfrm_state *x;
  74
  75         hlist_for_each_entry_safe(x, tmp, list, bydst) {
  76                 unsigned int h;
  77
  78                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  79                                     x->props.reqid, x->props.family,
  80                                     nhashmask);
  81                 hlist_add_head(&x->bydst, ndsttable+h);
  82
  83                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  84                                     x->props.family,
  85                                     nhashmask);
  86                 hlist_add_head(&x->bysrc, nsrctable+h);
  87
  88                 if (x->id.spi) {
  89                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  90                                             x->id.proto, x->props.family,
  91                                             nhashmask);
  92                         hlist_add_head(&x->byspi, nspitable+h);
  93                 }
  94         }
  95 }
  96
  97 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
  98 {
  99         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 100 }
 101
 102 static DEFINE_MUTEX(hash_resize_mutex);
 103
 104 static void xfrm_hash_resize(struct work_struct *work)
 105 {
 106         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 107         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 108         unsigned long nsize, osize;
 109         unsigned int nhashmask, ohashmask;
 110         int i;
 111
 112         mutex_lock(&hash_resize_mutex);
 113
 114         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 115         ndst = xfrm_hash_alloc(nsize);
 116         if (!ndst)
 117                 goto out_unlock;
 118         nsrc = xfrm_hash_alloc(nsize);
 119         if (!nsrc) {
 120                 xfrm_hash_free(ndst, nsize);
 121                 goto out_unlock;
 122         }
 123         nspi = xfrm_hash_alloc(nsize);
 124         if (!nspi) {
 125                 xfrm_hash_free(ndst, nsize);
 126                 xfrm_hash_free(nsrc, nsize);
 127                 goto out_unlock;
 128         }
 129
 130         spin_lock_bh(&xfrm_state_lock);
 131
 132         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 133         for (i = net->xfrm.state_hmask; i >= 0; i--)
 134                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 135                                    nhashmask);
 136
 137         odst = net->xfrm.state_bydst;
 138         osrc = net->xfrm.state_bysrc;
 139         ospi = net->xfrm.state_byspi;
 140         ohashmask = net->xfrm.state_hmask;
 141
 142         net->xfrm.state_bydst = ndst;
 143         net->xfrm.state_bysrc = nsrc;
 144         net->xfrm.state_byspi = nspi;
 145         net->xfrm.state_hmask = nhashmask;
 146
 147         spin_unlock_bh(&xfrm_state_lock);
 148
 149         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 150         xfrm_hash_free(odst, osize);
 151         xfrm_hash_free(osrc, osize);
 152         xfrm_hash_free(ospi, osize);
 153
 154 out_unlock:
 155         mutex_unlock(&hash_resize_mutex);
 156 }
 157
 158 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 159 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
 160
 161 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 162
 163 int __xfrm_state_delete(struct xfrm_state *x);
 164
 165 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 166 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 167
 168 static DEFINE_SPINLOCK(xfrm_type_lock);
 169 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 170 {
 171         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 172         const struct xfrm_type **typemap;
 173         int err = 0;
 174
 175         if (unlikely(afinfo == NULL))
 176                 return -EAFNOSUPPORT;
 177         typemap = afinfo->type_map;
 178         spin_lock_bh(&xfrm_type_lock);
 179
 180         if (likely(typemap[type->proto] == NULL))
 181                 typemap[type->proto] = type;
 182         else
 183                 err = -EEXIST;
 184         spin_unlock_bh(&xfrm_type_lock);
 185         xfrm_state_put_afinfo(afinfo);
 186         return err;
 187 }
 188 EXPORT_SYMBOL(xfrm_register_type);
 189
 190 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 191 {
 192         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 193         const struct xfrm_type **typemap;
 194         int err = 0;
 195
 196         if (unlikely(afinfo == NULL))
 197                 return -EAFNOSUPPORT;
 198         typemap = afinfo->type_map;
 199         spin_lock_bh(&xfrm_type_lock);
 200
 201         if (unlikely(typemap[type->proto] != type))
 202                 err = -ENOENT;
 203         else
 204                 typemap[type->proto] = NULL;
 205         spin_unlock_bh(&xfrm_type_lock);
 206         xfrm_state_put_afinfo(afinfo);
 207         return err;
 208 }
 209 EXPORT_SYMBOL(xfrm_unregister_type);
 210
 211 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 212 {
 213         struct xfrm_state_afinfo *afinfo;
 214         const struct xfrm_type **typemap;
 215         const struct xfrm_type *type;
 216         int modload_attempted = 0;
 217
 218 retry:
 219         afinfo = xfrm_state_get_afinfo(family);
 220         if (unlikely(afinfo == NULL))
 221                 return NULL;
 222         typemap = afinfo->type_map;
 223
 224         type = typemap[proto];
 225         if (unlikely(type && !try_module_get(type->owner)))
 226                 type = NULL;
 227         if (!type && !modload_attempted) {
 228                 xfrm_state_put_afinfo(afinfo);
 229                 request_module("xfrm-type-%d-%d", family, proto);
 230                 modload_attempted = 1;
 231                 goto retry;
 232         }
 233
 234         xfrm_state_put_afinfo(afinfo);
 235         return type;
 236 }
 237
 238 static void xfrm_put_type(const struct xfrm_type *type)
 239 {
 240         module_put(type->owner);
 241 }
 242
 243 static DEFINE_SPINLOCK(xfrm_mode_lock);
 244 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 245 {
 246         struct xfrm_state_afinfo *afinfo;
 247         struct xfrm_mode **modemap;
 248         int err;
 249
 250         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 251                 return -EINVAL;
 252
 253         afinfo = xfrm_state_get_afinfo(family);
 254         if (unlikely(afinfo == NULL))
 255                 return -EAFNOSUPPORT;
 256
 257         err = -EEXIST;
 258         modemap = afinfo->mode_map;
 259         spin_lock_bh(&xfrm_mode_lock);
 260         if (modemap[mode->encap])
 261                 goto out;
 262
 263         err = -ENOENT;
 264         if (!try_module_get(afinfo->owner))
 265                 goto out;
 266
 267         mode->afinfo = afinfo;
 268         modemap[mode->encap] = mode;
 269         err = 0;
 270
 271 out:
 272         spin_unlock_bh(&xfrm_mode_lock);
 273         xfrm_state_put_afinfo(afinfo);
 274         return err;
 275 }
 276 EXPORT_SYMBOL(xfrm_register_mode);
 277
 278 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 279 {
 280         struct xfrm_state_afinfo *afinfo;
 281         struct xfrm_mode **modemap;
 282         int err;
 283
 284         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 285                 return -EINVAL;
 286
 287         afinfo = xfrm_state_get_afinfo(family);
 288         if (unlikely(afinfo == NULL))
 289                 return -EAFNOSUPPORT;
 290
 291         err = -ENOENT;
 292         modemap = afinfo->mode_map;
 293         spin_lock_bh(&xfrm_mode_lock);
 294         if (likely(modemap[mode->encap] == mode)) {
 295                 modemap[mode->encap] = NULL;
 296                 module_put(mode->afinfo->owner);
 297                 err = 0;
 298         }
 299
 300         spin_unlock_bh(&xfrm_mode_lock);
 301         xfrm_state_put_afinfo(afinfo);
 302         return err;
 303 }
 304 EXPORT_SYMBOL(xfrm_unregister_mode);
 305
 306 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 307 {
 308         struct xfrm_state_afinfo *afinfo;
 309         struct xfrm_mode *mode;
 310         int modload_attempted = 0;
 311
 312         if (unlikely(encap >= XFRM_MODE_MAX))
 313                 return NULL;
 314
 315 retry:
 316         afinfo = xfrm_state_get_afinfo(family);
 317         if (unlikely(afinfo == NULL))
 318                 return NULL;
 319
 320         mode = afinfo->mode_map[encap];
 321         if (unlikely(mode && !try_module_get(mode->owner)))
 322                 mode = NULL;
 323         if (!mode && !modload_attempted) {
 324                 xfrm_state_put_afinfo(afinfo);
 325                 request_module("xfrm-mode-%d-%d", family, encap);
 326                 modload_attempted = 1;
 327                 goto retry;
 328         }
 329
 330         xfrm_state_put_afinfo(afinfo);
 331         return mode;
 332 }
 333
 334 static void xfrm_put_mode(struct xfrm_mode *mode)
 335 {
 336         module_put(mode->owner);
 337 }
 338
 339 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 340 {
 341         tasklet_hrtimer_cancel(&x->mtimer);
 342         del_timer_sync(&x->rtimer);
 343         kfree(x->aalg);
 344         kfree(x->ealg);
 345         kfree(x->calg);
 346         kfree(x->encap);
 347         kfree(x->coaddr);
 348         kfree(x->replay_esn);
 349         kfree(x->preplay_esn);
 350         if (x->inner_mode)
 351                 xfrm_put_mode(x->inner_mode);
 352         if (x->inner_mode_iaf)
 353                 xfrm_put_mode(x->inner_mode_iaf);
 354         if (x->outer_mode)
 355                 xfrm_put_mode(x->outer_mode);
 356         if (x->type) {
 357                 x->type->destructor(x);
 358                 xfrm_put_type(x->type);
 359         }
 360         security_xfrm_state_free(x);
 361         kfree(x);
 362 }
 363
 364 static void xfrm_state_gc_task(struct work_struct *work)
 365 {
 366         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 367         struct xfrm_state *x;
 368         struct hlist_node *tmp;
 369         struct hlist_head gc_list;
 370
 371         spin_lock_bh(&xfrm_state_gc_lock);
 372         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 373         spin_unlock_bh(&xfrm_state_gc_lock);
 374
 375         hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 376                 xfrm_state_gc_destroy(x);
 377
 378         wake_up(&net->xfrm.km_waitq);
 379 }
 380
 381 static inline unsigned long make_jiffies(long secs)
 382 {
 383         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 384                 return MAX_SCHEDULE_TIMEOUT-1;
 385         else
 386                 return secs*HZ;
 387 }
 388
 389 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 390 {
 391         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 392         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 393         struct net *net = xs_net(x);
 394         unsigned long now = get_seconds();
 395         long next = LONG_MAX;
 396         int warn = 0;
 397         int err = 0;
 398
 399         spin_lock(&x->lock);
 400         if (x->km.state == XFRM_STATE_DEAD)
 401                 goto out;
 402         if (x->km.state == XFRM_STATE_EXPIRED)
 403                 goto expired;
 404         if (x->lft.hard_add_expires_seconds) {
 405                 long tmo = x->lft.hard_add_expires_seconds +
 406                         x->curlft.add_time - now;
 407                 if (tmo <= 0) {
 408                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 409                                 /* enter hard expire without soft expire first?!
 410                                  * setting a new date could trigger this.
 411                                  * workarbound: fix x->curflt.add_time by below:
 412                                  */
 413                                 x->curlft.add_time = now - x->saved_tmo - 1;
 414                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 415                         } else
 416                                 goto expired;
 417                 }
 418                 if (tmo < next)
 419                         next = tmo;
 420         }
 421         if (x->lft.hard_use_expires_seconds) {
 422                 long tmo = x->lft.hard_use_expires_seconds +
 423                         (x->curlft.use_time ? : now) - now;
 424                 if (tmo <= 0)
 425                         goto expired;
 426                 if (tmo < next)
 427                         next = tmo;
 428         }
 429         if (x->km.dying)
 430                 goto resched;
 431         if (x->lft.soft_add_expires_seconds) {
 432                 long tmo = x->lft.soft_add_expires_seconds +
 433                         x->curlft.add_time - now;
 434                 if (tmo <= 0) {
 435                         warn = 1;
 436                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 437                 } else if (tmo < next) {
 438                         next = tmo;
 439                         x->xflags |= XFRM_SOFT_EXPIRE;
 440                         x->saved_tmo = tmo;
 441                 }
 442         }
 443         if (x->lft.soft_use_expires_seconds) {
 444                 long tmo = x->lft.soft_use_expires_seconds +
 445                         (x->curlft.use_time ? : now) - now;
 446                 if (tmo <= 0)
 447                         warn = 1;
 448                 else if (tmo < next)
 449                         next = tmo;
 450         }
 451
 452         x->km.dying = warn;
 453         if (warn)
 454                 km_state_expired(x, 0, 0);
 455 resched:
 456         if (next != LONG_MAX){
 457                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 458         }
 459
 460         goto out;
 461
 462 expired:
 463         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 464                 x->km.state = XFRM_STATE_EXPIRED;
 465                 wake_up(&net->xfrm.km_waitq);
 466                 next = 2;
 467                 goto resched;
 468         }
 469
 470         err = __xfrm_state_delete(x);
 471         if (!err && x->id.spi)
 472                 km_state_expired(x, 1, 0);
 473
 474         xfrm_audit_state_delete(x, err ? 0 : 1,
 475                                 audit_get_loginuid(current),
 476                                 audit_get_sessionid(current), 0);
 477
 478 out:
 479         spin_unlock(&x->lock);
 480         return HRTIMER_NORESTART;
 481 }
 482
 483 static void xfrm_replay_timer_handler(unsigned long data);
 484
 485 struct xfrm_state *xfrm_state_alloc(struct net *net)
 486 {
 487         struct xfrm_state *x;
 488
 489         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 490
 491         if (x) {
 492                 write_pnet(&x->xs_net, net);
 493                 atomic_set(&x->refcnt, 1);
 494                 atomic_set(&x->tunnel_users, 0);
 495                 INIT_LIST_HEAD(&x->km.all);
 496                 INIT_HLIST_NODE(&x->bydst);
 497                 INIT_HLIST_NODE(&x->bysrc);
 498                 INIT_HLIST_NODE(&x->byspi);
 499                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 500                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 501                                 (unsigned long)x);
 502                 x->curlft.add_time = get_seconds();
 503                 x->lft.soft_byte_limit = XFRM_INF;
 504                 x->lft.soft_packet_limit = XFRM_INF;
 505                 x->lft.hard_byte_limit = XFRM_INF;
 506                 x->lft.hard_packet_limit = XFRM_INF;
 507                 x->replay_maxage = 0;
 508                 x->replay_maxdiff = 0;
 509                 x->inner_mode = NULL;
 510                 x->inner_mode_iaf = NULL;
 511                 spin_lock_init(&x->lock);
 512         }
 513         return x;
 514 }
 515 EXPORT_SYMBOL(xfrm_state_alloc);
 516
 517 void __xfrm_state_destroy(struct xfrm_state *x)
 518 {
 519         struct net *net = xs_net(x);
 520
 521         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 522
 523         spin_lock_bh(&xfrm_state_gc_lock);
 524         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 525         spin_unlock_bh(&xfrm_state_gc_lock);
 526         schedule_work(&net->xfrm.state_gc_work);
 527 }
 528 EXPORT_SYMBOL(__xfrm_state_destroy);
 529
 530 int __xfrm_state_delete(struct xfrm_state *x)
 531 {
 532         struct net *net = xs_net(x);
 533         int err = -ESRCH;
 534
 535         if (x->km.state != XFRM_STATE_DEAD) {
 536                 x->km.state = XFRM_STATE_DEAD;
 537                 spin_lock(&xfrm_state_lock);
 538                 list_del(&x->km.all);
 539                 hlist_del(&x->bydst);
 540                 hlist_del(&x->bysrc);
 541                 if (x->id.spi)
 542                         hlist_del(&x->byspi);
 543                 net->xfrm.state_num--;
 544                 spin_unlock(&xfrm_state_lock);
 545
 546                 /* All xfrm_state objects are created by xfrm_state_alloc.
 547                  * The xfrm_state_alloc call gives a reference, and that
 548                  * is what we are dropping here.
 549                  */
 550                 xfrm_state_put(x);
 551                 err = 0;
 552         }
 553
 554         return err;
 555 }
 556 EXPORT_SYMBOL(__xfrm_state_delete);
 557
 558 int xfrm_state_delete(struct xfrm_state *x)
 559 {
 560         int err;
 561
 562         spin_lock_bh(&x->lock);
 563         err = __xfrm_state_delete(x);
 564         spin_unlock_bh(&x->lock);
 565
 566         return err;
 567 }
 568 EXPORT_SYMBOL(xfrm_state_delete);
 569
 570 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 571 static inline int
 572 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 573 {
 574         int i, err = 0;
 575
 576         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 577                 struct xfrm_state *x;
 578
 579                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 580                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 581                            (err = security_xfrm_state_delete(x)) != 0) {
 582                                 xfrm_audit_state_delete(x, 0,
 583                                                         audit_info->loginuid,
 584                                                         audit_info->sessionid,
 585                                                         audit_info->secid);
 586                                 return err;
 587                         }
 588                 }
 589         }
 590
 591         return err;
 592 }
 593 #else
 594 static inline int
 595 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 596 {
 597         return 0;
 598 }
 599 #endif
 600
 601 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 602 {
 603         int i, err = 0, cnt = 0;
 604
 605         spin_lock_bh(&xfrm_state_lock);
 606         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 607         if (err)
 608                 goto out;
 609
 610         err = -ESRCH;
 611         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 612                 struct xfrm_state *x;
 613 restart:
 614                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 615                         if (!xfrm_state_kern(x) &&
 616                             xfrm_id_proto_match(x->id.proto, proto)) {
 617                                 xfrm_state_hold(x);
 618                                 spin_unlock_bh(&xfrm_state_lock);
 619
 620                                 err = xfrm_state_delete(x);
 621                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 622                                                         audit_info->loginuid,
 623                                                         audit_info->sessionid,
 624                                                         audit_info->secid);
 625                                 xfrm_state_put(x);
 626                                 if (!err)
 627                                         cnt++;
 628
 629                                 spin_lock_bh(&xfrm_state_lock);
 630                                 goto restart;
 631                         }
 632                 }
 633         }
 634         if (cnt)
 635                 err = 0;
 636
 637 out:
 638         spin_unlock_bh(&xfrm_state_lock);
 639         wake_up(&net->xfrm.km_waitq);
 640         return err;
 641 }
 642 EXPORT_SYMBOL(xfrm_state_flush);
 643
 644 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 645 {
 646         spin_lock_bh(&xfrm_state_lock);
 647         si->sadcnt = net->xfrm.state_num;
 648         si->sadhcnt = net->xfrm.state_hmask;
 649         si->sadhmcnt = xfrm_state_hashmax;
 650         spin_unlock_bh(&xfrm_state_lock);
 651 }
 652 EXPORT_SYMBOL(xfrm_sad_getinfo);
 653
 654 static int
 655 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 656                     const struct xfrm_tmpl *tmpl,
 657                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 658                     unsigned short family)
 659 {
 660         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 661         if (!afinfo)
 662                 return -1;
 663         afinfo->init_tempsel(&x->sel, fl);
 664
 665         if (family != tmpl->encap_family) {
 666                 xfrm_state_put_afinfo(afinfo);
 667                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 668                 if (!afinfo)
 669                         return -1;
 670         }
 671         afinfo->init_temprop(x, tmpl, daddr, saddr);
 672         xfrm_state_put_afinfo(afinfo);
 673         return 0;
 674 }
 675
 676 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 677                                               const xfrm_address_t *daddr,
 678                                               __be32 spi, u8 proto,
 679                                               unsigned short family)
 680 {
 681         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 682         struct xfrm_state *x;
 683
 684         hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
 685                 if (x->props.family != family ||
 686                     x->id.spi       != spi ||
 687                     x->id.proto     != proto ||
 688                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 689                         continue;
 690
 691                 if ((mark & x->mark.m) != x->mark.v)
 692                         continue;
 693                 xfrm_state_hold(x);
 694                 return x;
 695         }
 696
 697         return NULL;
 698 }
 699
 700 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 701                                                      const xfrm_address_t *daddr,
 702                                                      const xfrm_address_t *saddr,
 703                                                      u8 proto, unsigned short family)
 704 {
 705         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 706         struct xfrm_state *x;
 707
 708         hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
 709                 if (x->props.family != family ||
 710                     x->id.proto     != proto ||
 711                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 712                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 713                         continue;
 714
 715                 if ((mark & x->mark.m) != x->mark.v)
 716                         continue;
 717                 xfrm_state_hold(x);
 718                 return x;
 719         }
 720
 721         return NULL;
 722 }
 723
 724 static inline struct xfrm_state *
 725 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 726 {
 727         struct net *net = xs_net(x);
 728         u32 mark = x->mark.v & x->mark.m;
 729
 730         if (use_spi)
 731                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 732                                            x->id.spi, x->id.proto, family);
 733         else
 734                 return __xfrm_state_lookup_byaddr(net, mark,
 735                                                   &x->id.daddr,
 736                                                   &x->props.saddr,
 737                                                   x->id.proto, family);
 738 }
 739
 740 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 741 {
 742         if (have_hash_collision &&
 743             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 744             net->xfrm.state_num > net->xfrm.state_hmask)
 745                 schedule_work(&net->xfrm.state_hash_work);
 746 }
 747
 748 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 749                                const struct flowi *fl, unsigned short family,
 750                                struct xfrm_state **best, int *acq_in_progress,
 751                                int *error)
 752 {
 753         /* Resolution logic:
 754          * 1. There is a valid state with matching selector. Done.
 755          * 2. Valid state with inappropriate selector. Skip.
 756          *
 757          * Entering area of "sysdeps".
 758          *
 759          * 3. If state is not valid, selector is temporary, it selects
 760          *    only session which triggered previous resolution. Key
 761          *    manager will do something to install a state with proper
 762          *    selector.
 763          */
 764         if (x->km.state == XFRM_STATE_VALID) {
 765                 if ((x->sel.family &&
 766                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 767                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 768                         return;
 769
 770                 if (!*best ||
 771                     (*best)->km.dying > x->km.dying ||
 772                     ((*best)->km.dying == x->km.dying &&
 773                      (*best)->curlft.add_time < x->curlft.add_time))
 774                         *best = x;
 775         } else if (x->km.state == XFRM_STATE_ACQ) {
 776                 *acq_in_progress = 1;
 777         } else if (x->km.state == XFRM_STATE_ERROR ||
 778                    x->km.state == XFRM_STATE_EXPIRED) {
 779                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 780                     security_xfrm_state_pol_flow_match(x, pol, fl))
 781                         *error = -ESRCH;
 782         }
 783 }
 784
 785 struct xfrm_state *
 786 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 787                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 788                 struct xfrm_policy *pol, int *err,
 789                 unsigned short family)
 790 {
 791         static xfrm_address_t saddr_wildcard = { };
 792         struct net *net = xp_net(pol);
 793         unsigned int h, h_wildcard;
 794         struct xfrm_state *x, *x0, *to_put;
 795         int acquire_in_progress = 0;
 796         int error = 0;
 797         struct xfrm_state *best = NULL;
 798         u32 mark = pol->mark.v & pol->mark.m;
 799         unsigned short encap_family = tmpl->encap_family;
 800
 801         to_put = NULL;
 802
 803         spin_lock_bh(&xfrm_state_lock);
 804         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 805         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 806                 if (x->props.family == encap_family &&
 807                     x->props.reqid == tmpl->reqid &&
 808                     (mark & x->mark.m) == x->mark.v &&
 809                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 810                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 811                     tmpl->mode == x->props.mode &&
 812                     tmpl->id.proto == x->id.proto &&
 813                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 814                         xfrm_state_look_at(pol, x, fl, encap_family,
 815                                            &best, &acquire_in_progress, &error);
 816         }
 817         if (best)
 818                 goto found;
 819
 820         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 821         hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
 822                 if (x->props.family == encap_family &&
 823                     x->props.reqid == tmpl->reqid &&
 824                     (mark & x->mark.m) == x->mark.v &&
 825                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 826                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 827                     tmpl->mode == x->props.mode &&
 828                     tmpl->id.proto == x->id.proto &&
 829                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 830                         xfrm_state_look_at(pol, x, fl, encap_family,
 831                                            &best, &acquire_in_progress, &error);
 832         }
 833
 834 found:
 835         x = best;
 836         if (!x && !error && !acquire_in_progress) {
 837                 if (tmpl->id.spi &&
 838                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 839                                               tmpl->id.proto, encap_family)) != NULL) {
 840                         to_put = x0;
 841                         error = -EEXIST;
 842                         goto out;
 843                 }
 844                 x = xfrm_state_alloc(net);
 845                 if (x == NULL) {
 846                         error = -ENOMEM;
 847                         goto out;
 848                 }
 849                 /* Initialize temporary state matching only
 850                  * to current session. */
 851                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 852                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 853
 854                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 855                 if (error) {
 856                         x->km.state = XFRM_STATE_DEAD;
 857                         to_put = x;
 858                         x = NULL;
 859                         goto out;
 860                 }
 861
 862                 if (km_query(x, tmpl, pol) == 0) {
 863                         x->km.state = XFRM_STATE_ACQ;
 864                         list_add(&x->km.all, &net->xfrm.state_all);
 865                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 866                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 867                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 868                         if (x->id.spi) {
 869                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 870                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 871                         }
 872                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 873                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 874                         net->xfrm.state_num++;
 875                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 876                 } else {
 877                         x->km.state = XFRM_STATE_DEAD;
 878                         to_put = x;
 879                         x = NULL;
 880                         error = -ESRCH;
 881                 }
 882         }
 883 out:
 884         if (x)
 885                 xfrm_state_hold(x);
 886         else
 887                 *err = acquire_in_progress ? -EAGAIN : error;
 888         spin_unlock_bh(&xfrm_state_lock);
 889         if (to_put)
 890                 xfrm_state_put(to_put);
 891         return x;
 892 }
 893
 894 struct xfrm_state *
 895 xfrm_stateonly_find(struct net *net, u32 mark,
 896                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 897                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 898 {
 899         unsigned int h;
 900         struct xfrm_state *rx = NULL, *x = NULL;
 901
 902         spin_lock(&xfrm_state_lock);
 903         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 904         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 905                 if (x->props.family == family &&
 906                     x->props.reqid == reqid &&
 907                     (mark & x->mark.m) == x->mark.v &&
 908                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 909                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 910                     mode == x->props.mode &&
 911                     proto == x->id.proto &&
 912                     x->km.state == XFRM_STATE_VALID) {
 913                         rx = x;
 914                         break;
 915                 }
 916         }
 917
 918         if (rx)
 919                 xfrm_state_hold(rx);
 920         spin_unlock(&xfrm_state_lock);
 921
 922
 923         return rx;
 924 }
 925 EXPORT_SYMBOL(xfrm_stateonly_find);
 926
 927 static void __xfrm_state_insert(struct xfrm_state *x)
 928 {
 929         struct net *net = xs_net(x);
 930         unsigned int h;
 931
 932         list_add(&x->km.all, &net->xfrm.state_all);
 933
 934         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 935                           x->props.reqid, x->props.family);
 936         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 937
 938         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 939         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 940
 941         if (x->id.spi) {
 942                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 943                                   x->props.family);
 944
 945                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 946         }
 947
 948         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 949         if (x->replay_maxage)
 950                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 951
 952         wake_up(&net->xfrm.km_waitq);
 953
 954         net->xfrm.state_num++;
 955
 956         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 957 }
 958
 959 /* xfrm_state_lock is held */
 960 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 961 {
 962         struct net *net = xs_net(xnew);
 963         unsigned short family = xnew->props.family;
 964         u32 reqid = xnew->props.reqid;
 965         struct xfrm_state *x;
 966         unsigned int h;
 967         u32 mark = xnew->mark.v & xnew->mark.m;
 968
 969         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 970         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 971                 if (x->props.family     == family &&
 972                     x->props.reqid      == reqid &&
 973                     (mark & x->mark.m) == x->mark.v &&
 974                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 975                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
 976                         x->genid++;
 977         }
 978 }
 979
 980 void xfrm_state_insert(struct xfrm_state *x)
 981 {
 982         spin_lock_bh(&xfrm_state_lock);
 983         __xfrm_state_bump_genids(x);
 984         __xfrm_state_insert(x);
 985         spin_unlock_bh(&xfrm_state_lock);
 986 }
 987 EXPORT_SYMBOL(xfrm_state_insert);
 988
 989 /* xfrm_state_lock is held */
 990 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
 991                                           unsigned short family, u8 mode,
 992                                           u32 reqid, u8 proto,
 993                                           const xfrm_address_t *daddr,
 994                                           const xfrm_address_t *saddr, int create)
 995 {
 996         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 997         struct xfrm_state *x;
 998         u32 mark = m->v & m->m;
 999
1000         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1001                 if (x->props.reqid  != reqid ||
1002                     x->props.mode   != mode ||
1003                     x->props.family != family ||
1004                     x->km.state     != XFRM_STATE_ACQ ||
1005                     x->id.spi       != 0 ||
1006                     x->id.proto     != proto ||
1007                     (mark & x->mark.m) != x->mark.v ||
1008                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1009                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1010                         continue;
1011
1012                 xfrm_state_hold(x);
1013                 return x;
1014         }
1015
1016         if (!create)
1017                 return NULL;
1018
1019         x = xfrm_state_alloc(net);
1020         if (likely(x)) {
1021                 switch (family) {
1022                 case AF_INET:
1023                         x->sel.daddr.a4 = daddr->a4;
1024                         x->sel.saddr.a4 = saddr->a4;
1025                         x->sel.prefixlen_d = 32;
1026                         x->sel.prefixlen_s = 32;
1027                         x->props.saddr.a4 = saddr->a4;
1028                         x->id.daddr.a4 = daddr->a4;
1029                         break;
1030
1031                 case AF_INET6:
1032                         *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1033                         *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1034                         x->sel.prefixlen_d = 128;
1035                         x->sel.prefixlen_s = 128;
1036                         *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1037                         *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1038                         break;
1039                 }
1040
1041                 x->km.state = XFRM_STATE_ACQ;
1042                 x->id.proto = proto;
1043                 x->props.family = family;
1044                 x->props.mode = mode;
1045                 x->props.reqid = reqid;
1046                 x->mark.v = m->v;
1047                 x->mark.m = m->m;
1048                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1049                 xfrm_state_hold(x);
1050                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1051                 list_add(&x->km.all, &net->xfrm.state_all);
1052                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1053                 h = xfrm_src_hash(net, daddr, saddr, family);
1054                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1055
1056                 net->xfrm.state_num++;
1057
1058                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1059         }
1060
1061         return x;
1062 }
1063
1064 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1065
1066 int xfrm_state_add(struct xfrm_state *x)
1067 {
1068         struct net *net = xs_net(x);
1069         struct xfrm_state *x1, *to_put;
1070         int family;
1071         int err;
1072         u32 mark = x->mark.v & x->mark.m;
1073         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1074
1075         family = x->props.family;
1076
1077         to_put = NULL;
1078
1079         spin_lock_bh(&xfrm_state_lock);
1080
1081         x1 = __xfrm_state_locate(x, use_spi, family);
1082         if (x1) {
1083                 to_put = x1;
1084                 x1 = NULL;
1085                 err = -EEXIST;
1086                 goto out;
1087         }
1088
1089         if (use_spi && x->km.seq) {
1090                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1091                 if (x1 && ((x1->id.proto != x->id.proto) ||
1092                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1093                         to_put = x1;
1094                         x1 = NULL;
1095                 }
1096         }
1097
1098         if (use_spi && !x1)
1099                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1100                                      x->props.reqid, x->id.proto,
1101                                      &x->id.daddr, &x->props.saddr, 0);
1102
1103         __xfrm_state_bump_genids(x);
1104         __xfrm_state_insert(x);
1105         err = 0;
1106
1107 out:
1108         spin_unlock_bh(&xfrm_state_lock);
1109
1110         if (x1) {
1111                 xfrm_state_delete(x1);
1112                 xfrm_state_put(x1);
1113         }
1114
1115         if (to_put)
1116                 xfrm_state_put(to_put);
1117
1118         return err;
1119 }
1120 EXPORT_SYMBOL(xfrm_state_add);
1121
1122 #ifdef CONFIG_XFRM_MIGRATE
1123 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1124 {
1125         struct net *net = xs_net(orig);
1126         int err = -ENOMEM;
1127         struct xfrm_state *x = xfrm_state_alloc(net);
1128         if (!x)
1129                 goto out;
1130
1131         memcpy(&x->id, &orig->id, sizeof(x->id));
1132         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1133         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1134         x->props.mode = orig->props.mode;
1135         x->props.replay_window = orig->props.replay_window;
1136         x->props.reqid = orig->props.reqid;
1137         x->props.family = orig->props.family;
1138         x->props.saddr = orig->props.saddr;
1139
1140         if (orig->aalg) {
1141                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1142                 if (!x->aalg)
1143                         goto error;
1144         }
1145         x->props.aalgo = orig->props.aalgo;
1146
1147         if (orig->ealg) {
1148                 x->ealg = xfrm_algo_clone(orig->ealg);
1149                 if (!x->ealg)
1150                         goto error;
1151         }
1152         x->props.ealgo = orig->props.ealgo;
1153
1154         if (orig->calg) {
1155                 x->calg = xfrm_algo_clone(orig->calg);
1156                 if (!x->calg)
1157                         goto error;
1158         }
1159         x->props.calgo = orig->props.calgo;
1160
1161         if (orig->encap) {
1162                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1163                 if (!x->encap)
1164                         goto error;
1165         }
1166
1167         if (orig->coaddr) {
1168                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1169                                     GFP_KERNEL);
1170                 if (!x->coaddr)
1171                         goto error;
1172         }
1173
1174         if (orig->replay_esn) {
1175                 err = xfrm_replay_clone(x, orig);
1176                 if (err)
1177                         goto error;
1178         }
1179
1180         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1181
1182         err = xfrm_init_state(x);
1183         if (err)
1184                 goto error;
1185
1186         x->props.flags = orig->props.flags;
1187         x->props.extra_flags = orig->props.extra_flags;
1188
1189         x->curlft.add_time = orig->curlft.add_time;
1190         x->km.state = orig->km.state;
1191         x->km.seq = orig->km.seq;
1192
1193         return x;
1194
1195  error:
1196         xfrm_state_put(x);
1197 out:
1198         if (errp)
1199                 *errp = err;
1200         return NULL;
1201 }
1202
1203 /* xfrm_state_lock is held */
1204 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1205 {
1206         unsigned int h;
1207         struct xfrm_state *x;
1208
1209         if (m->reqid) {
1210                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1211                                   m->reqid, m->old_family);
1212                 hlist_for_each_entry(x, init_net.xfrm.state_bydst+h, bydst) {
1213                         if (x->props.mode != m->mode ||
1214                             x->id.proto != m->proto)
1215                                 continue;
1216                         if (m->reqid && x->props.reqid != m->reqid)
1217                                 continue;
1218                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1219                                              m->old_family) ||
1220                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1221                                              m->old_family))
1222                                 continue;
1223                         xfrm_state_hold(x);
1224                         return x;
1225                 }
1226         } else {
1227                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1228                                   m->old_family);
1229                 hlist_for_each_entry(x, init_net.xfrm.state_bysrc+h, bysrc) {
1230                         if (x->props.mode != m->mode ||
1231                             x->id.proto != m->proto)
1232                                 continue;
1233                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1234                                              m->old_family) ||
1235                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1236                                              m->old_family))
1237                                 continue;
1238                         xfrm_state_hold(x);
1239                         return x;
1240                 }
1241         }
1242
1243         return NULL;
1244 }
1245 EXPORT_SYMBOL(xfrm_migrate_state_find);
1246
1247 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1248                                        struct xfrm_migrate *m)
1249 {
1250         struct xfrm_state *xc;
1251         int err;
1252
1253         xc = xfrm_state_clone(x, &err);
1254         if (!xc)
1255                 return NULL;
1256
1257         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1258         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1259
1260         /* add state */
1261         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1262                 /* a care is needed when the destination address of the
1263                    state is to be updated as it is a part of triplet */
1264                 xfrm_state_insert(xc);
1265         } else {
1266                 if ((err = xfrm_state_add(xc)) < 0)
1267                         goto error;
1268         }
1269
1270         return xc;
1271 error:
1272         xfrm_state_put(xc);
1273         return NULL;
1274 }
1275 EXPORT_SYMBOL(xfrm_state_migrate);
1276 #endif
1277
1278 int xfrm_state_update(struct xfrm_state *x)
1279 {
1280         struct xfrm_state *x1, *to_put;
1281         int err;
1282         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1283
1284         to_put = NULL;
1285
1286         spin_lock_bh(&xfrm_state_lock);
1287         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1288
1289         err = -ESRCH;
1290         if (!x1)
1291                 goto out;
1292
1293         if (xfrm_state_kern(x1)) {
1294                 to_put = x1;
1295                 err = -EEXIST;
1296                 goto out;
1297         }
1298
1299         if (x1->km.state == XFRM_STATE_ACQ) {
1300                 __xfrm_state_insert(x);
1301                 x = NULL;
1302         }
1303         err = 0;
1304
1305 out:
1306         spin_unlock_bh(&xfrm_state_lock);
1307
1308         if (to_put)
1309                 xfrm_state_put(to_put);
1310
1311         if (err)
1312                 return err;
1313
1314         if (!x) {
1315                 xfrm_state_delete(x1);
1316                 xfrm_state_put(x1);
1317                 return 0;
1318         }
1319
1320         err = -EINVAL;
1321         spin_lock_bh(&x1->lock);
1322         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1323                 if (x->encap && x1->encap)
1324                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1325                 if (x->coaddr && x1->coaddr) {
1326                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1327                 }
1328                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1329                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1330                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1331                 x1->km.dying = 0;
1332
1333                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1334                 if (x1->curlft.use_time)
1335                         xfrm_state_check_expire(x1);
1336
1337                 err = 0;
1338                 x->km.state = XFRM_STATE_DEAD;
1339                 __xfrm_state_put(x);
1340         }
1341         spin_unlock_bh(&x1->lock);
1342
1343         xfrm_state_put(x1);
1344
1345         return err;
1346 }
1347 EXPORT_SYMBOL(xfrm_state_update);
1348
1349 int xfrm_state_check_expire(struct xfrm_state *x)
1350 {
1351         if (!x->curlft.use_time)
1352                 x->curlft.use_time = get_seconds();
1353
1354         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1355             x->curlft.packets >= x->lft.hard_packet_limit) {
1356                 x->km.state = XFRM_STATE_EXPIRED;
1357                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1358                 return -EINVAL;
1359         }
1360
1361         if (!x->km.dying &&
1362             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1363              x->curlft.packets >= x->lft.soft_packet_limit)) {
1364                 x->km.dying = 1;
1365                 km_state_expired(x, 0, 0);
1366         }
1367         return 0;
1368 }
1369 EXPORT_SYMBOL(xfrm_state_check_expire);
1370
1371 struct xfrm_state *
1372 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1373                   u8 proto, unsigned short family)
1374 {
1375         struct xfrm_state *x;
1376
1377         spin_lock_bh(&xfrm_state_lock);
1378         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1379         spin_unlock_bh(&xfrm_state_lock);
1380         return x;
1381 }
1382 EXPORT_SYMBOL(xfrm_state_lookup);
1383
1384 struct xfrm_state *
1385 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1386                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1387                          u8 proto, unsigned short family)
1388 {
1389         struct xfrm_state *x;
1390
1391         spin_lock_bh(&xfrm_state_lock);
1392         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1393         spin_unlock_bh(&xfrm_state_lock);
1394         return x;
1395 }
1396 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1397
1398 struct xfrm_state *
1399 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1400               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1401               int create, unsigned short family)
1402 {
1403         struct xfrm_state *x;
1404
1405         spin_lock_bh(&xfrm_state_lock);
1406         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1407         spin_unlock_bh(&xfrm_state_lock);
1408
1409         return x;
1410 }
1411 EXPORT_SYMBOL(xfrm_find_acq);
1412
1413 #ifdef CONFIG_XFRM_SUB_POLICY
1414 int
1415 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1416                unsigned short family)
1417 {
1418         int err = 0;
1419         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1420         if (!afinfo)
1421                 return -EAFNOSUPPORT;
1422
1423         spin_lock_bh(&xfrm_state_lock);
1424         if (afinfo->tmpl_sort)
1425                 err = afinfo->tmpl_sort(dst, src, n);
1426         spin_unlock_bh(&xfrm_state_lock);
1427         xfrm_state_put_afinfo(afinfo);
1428         return err;
1429 }
1430 EXPORT_SYMBOL(xfrm_tmpl_sort);
1431
1432 int
1433 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1434                 unsigned short family)
1435 {
1436         int err = 0;
1437         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1438         if (!afinfo)
1439                 return -EAFNOSUPPORT;
1440
1441         spin_lock_bh(&xfrm_state_lock);
1442         if (afinfo->state_sort)
1443                 err = afinfo->state_sort(dst, src, n);
1444         spin_unlock_bh(&xfrm_state_lock);
1445         xfrm_state_put_afinfo(afinfo);
1446         return err;
1447 }
1448 EXPORT_SYMBOL(xfrm_state_sort);
1449 #endif
1450
1451 /* Silly enough, but I'm lazy to build resolution list */
1452
1453 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1454 {
1455         int i;
1456
1457         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1458                 struct xfrm_state *x;
1459
1460                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1461                         if (x->km.seq == seq &&
1462                             (mark & x->mark.m) == x->mark.v &&
1463                             x->km.state == XFRM_STATE_ACQ) {
1464                                 xfrm_state_hold(x);
1465                                 return x;
1466                         }
1467                 }
1468         }
1469         return NULL;
1470 }
1471
1472 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1473 {
1474         struct xfrm_state *x;
1475
1476         spin_lock_bh(&xfrm_state_lock);
1477         x = __xfrm_find_acq_byseq(net, mark, seq);
1478         spin_unlock_bh(&xfrm_state_lock);
1479         return x;
1480 }
1481 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1482
1483 u32 xfrm_get_acqseq(void)
1484 {
1485         u32 res;
1486         static atomic_t acqseq;
1487
1488         do {
1489                 res = atomic_inc_return(&acqseq);
1490         } while (!res);
1491
1492         return res;
1493 }
1494 EXPORT_SYMBOL(xfrm_get_acqseq);
1495
1496 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1497 {
1498         struct net *net = xs_net(x);
1499         unsigned int h;
1500         struct xfrm_state *x0;
1501         int err = -ENOENT;
1502         __be32 minspi = htonl(low);
1503         __be32 maxspi = htonl(high);
1504         u32 mark = x->mark.v & x->mark.m;
1505
1506         spin_lock_bh(&x->lock);
1507         if (x->km.state == XFRM_STATE_DEAD)
1508                 goto unlock;
1509
1510         err = 0;
1511         if (x->id.spi)
1512                 goto unlock;
1513
1514         err = -ENOENT;
1515
1516         if (minspi == maxspi) {
1517                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1518                 if (x0) {
1519                         xfrm_state_put(x0);
1520                         goto unlock;
1521                 }
1522                 x->id.spi = minspi;
1523         } else {
1524                 u32 spi = 0;
1525                 for (h=0; h<high-low+1; h++) {
1526                         spi = low + net_random()%(high-low+1);
1527                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1528                         if (x0 == NULL) {
1529                                 x->id.spi = htonl(spi);
1530                                 break;
1531                         }
1532                         xfrm_state_put(x0);
1533                 }
1534         }
1535         if (x->id.spi) {
1536                 spin_lock_bh(&xfrm_state_lock);
1537                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1538                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1539                 spin_unlock_bh(&xfrm_state_lock);
1540
1541                 err = 0;
1542         }
1543
1544 unlock:
1545         spin_unlock_bh(&x->lock);
1546
1547         return err;
1548 }
1549 EXPORT_SYMBOL(xfrm_alloc_spi);
1550
1551 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1552                     int (*func)(struct xfrm_state *, int, void*),
1553                     void *data)
1554 {
1555         struct xfrm_state *state;
1556         struct xfrm_state_walk *x;
1557         int err = 0;
1558
1559         if (walk->seq != 0 && list_empty(&walk->all))
1560                 return 0;
1561
1562         spin_lock_bh(&xfrm_state_lock);
1563         if (list_empty(&walk->all))
1564                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1565         else
1566                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1567         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1568                 if (x->state == XFRM_STATE_DEAD)
1569                         continue;
1570                 state = container_of(x, struct xfrm_state, km);
1571                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1572                         continue;
1573                 err = func(state, walk->seq, data);
1574                 if (err) {
1575                         list_move_tail(&walk->all, &x->all);
1576                         goto out;
1577                 }
1578                 walk->seq++;
1579         }
1580         if (walk->seq == 0) {
1581                 err = -ENOENT;
1582                 goto out;
1583         }
1584         list_del_init(&walk->all);
1585 out:
1586         spin_unlock_bh(&xfrm_state_lock);
1587         return err;
1588 }
1589 EXPORT_SYMBOL(xfrm_state_walk);
1590
1591 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1592 {
1593         INIT_LIST_HEAD(&walk->all);
1594         walk->proto = proto;
1595         walk->state = XFRM_STATE_DEAD;
1596         walk->seq = 0;
1597 }
1598 EXPORT_SYMBOL(xfrm_state_walk_init);
1599
1600 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1601 {
1602         if (list_empty(&walk->all))
1603                 return;
1604
1605         spin_lock_bh(&xfrm_state_lock);
1606         list_del(&walk->all);
1607         spin_unlock_bh(&xfrm_state_lock);
1608 }
1609 EXPORT_SYMBOL(xfrm_state_walk_done);
1610
1611 static void xfrm_replay_timer_handler(unsigned long data)
1612 {
1613         struct xfrm_state *x = (struct xfrm_state*)data;
1614
1615         spin_lock(&x->lock);
1616
1617         if (x->km.state == XFRM_STATE_VALID) {
1618                 if (xfrm_aevent_is_on(xs_net(x)))
1619                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1620                 else
1621                         x->xflags |= XFRM_TIME_DEFER;
1622         }
1623
1624         spin_unlock(&x->lock);
1625 }
1626
1627 static LIST_HEAD(xfrm_km_list);
1628
1629 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1630 {
1631         struct xfrm_mgr *km;
1632
1633         rcu_read_lock();
1634         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1635                 if (km->notify_policy)
1636                         km->notify_policy(xp, dir, c);
1637         rcu_read_unlock();
1638 }
1639
1640 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1641 {
1642         struct xfrm_mgr *km;
1643         rcu_read_lock();
1644         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1645                 if (km->notify)
1646                         km->notify(x, c);
1647         rcu_read_unlock();
1648 }
1649
1650 EXPORT_SYMBOL(km_policy_notify);
1651 EXPORT_SYMBOL(km_state_notify);
1652
1653 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1654 {
1655         struct net *net = xs_net(x);
1656         struct km_event c;
1657
1658         c.data.hard = hard;
1659         c.portid = portid;
1660         c.event = XFRM_MSG_EXPIRE;
1661         km_state_notify(x, &c);
1662
1663         if (hard)
1664                 wake_up(&net->xfrm.km_waitq);
1665 }
1666
1667 EXPORT_SYMBOL(km_state_expired);
1668 /*
1669  * We send to all registered managers regardless of failure
1670  * We are happy with one success
1671 */
1672 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1673 {
1674         int err = -EINVAL, acqret;
1675         struct xfrm_mgr *km;
1676
1677         rcu_read_lock();
1678         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1679                 acqret = km->acquire(x, t, pol);
1680                 if (!acqret)
1681                         err = acqret;
1682         }
1683         rcu_read_unlock();
1684         return err;
1685 }
1686 EXPORT_SYMBOL(km_query);
1687
1688 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1689 {
1690         int err = -EINVAL;
1691         struct xfrm_mgr *km;
1692
1693         rcu_read_lock();
1694         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1695                 if (km->new_mapping)
1696                         err = km->new_mapping(x, ipaddr, sport);
1697                 if (!err)
1698                         break;
1699         }
1700         rcu_read_unlock();
1701         return err;
1702 }
1703 EXPORT_SYMBOL(km_new_mapping);
1704
1705 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1706 {
1707         struct net *net = xp_net(pol);
1708         struct km_event c;
1709
1710         c.data.hard = hard;
1711         c.portid = portid;
1712         c.event = XFRM_MSG_POLEXPIRE;
1713         km_policy_notify(pol, dir, &c);
1714
1715         if (hard)
1716                 wake_up(&net->xfrm.km_waitq);
1717 }
1718 EXPORT_SYMBOL(km_policy_expired);
1719
1720 #ifdef CONFIG_XFRM_MIGRATE
1721 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1722                const struct xfrm_migrate *m, int num_migrate,
1723                const struct xfrm_kmaddress *k)
1724 {
1725         int err = -EINVAL;
1726         int ret;
1727         struct xfrm_mgr *km;
1728
1729         rcu_read_lock();
1730         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1731                 if (km->migrate) {
1732                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1733                         if (!ret)
1734                                 err = ret;
1735                 }
1736         }
1737         rcu_read_unlock();
1738         return err;
1739 }
1740 EXPORT_SYMBOL(km_migrate);
1741 #endif
1742
1743 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1744 {
1745         int err = -EINVAL;
1746         int ret;
1747         struct xfrm_mgr *km;
1748
1749         rcu_read_lock();
1750         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1751                 if (km->report) {
1752                         ret = km->report(net, proto, sel, addr);
1753                         if (!ret)
1754                                 err = ret;
1755                 }
1756         }
1757         rcu_read_unlock();
1758         return err;
1759 }
1760 EXPORT_SYMBOL(km_report);
1761
1762 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1763 {
1764         int err;
1765         u8 *data;
1766         struct xfrm_mgr *km;
1767         struct xfrm_policy *pol = NULL;
1768
1769         if (optlen <= 0 || optlen > PAGE_SIZE)
1770                 return -EMSGSIZE;
1771
1772         data = kmalloc(optlen, GFP_KERNEL);
1773         if (!data)
1774                 return -ENOMEM;
1775
1776         err = -EFAULT;
1777         if (copy_from_user(data, optval, optlen))
1778                 goto out;
1779
1780         err = -EINVAL;
1781         rcu_read_lock();
1782         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1783                 pol = km->compile_policy(sk, optname, data,
1784                                          optlen, &err);
1785                 if (err >= 0)
1786                         break;
1787         }
1788         rcu_read_unlock();
1789
1790         if (err >= 0) {
1791                 xfrm_sk_policy_insert(sk, err, pol);
1792                 xfrm_pol_put(pol);
1793                 err = 0;
1794         }
1795
1796 out:
1797         kfree(data);
1798         return err;
1799 }
1800 EXPORT_SYMBOL(xfrm_user_policy);
1801
1802 static DEFINE_SPINLOCK(xfrm_km_lock);
1803
1804 int xfrm_register_km(struct xfrm_mgr *km)
1805 {
1806         spin_lock_bh(&xfrm_km_lock);
1807         list_add_tail_rcu(&km->list, &xfrm_km_list);
1808         spin_unlock_bh(&xfrm_km_lock);
1809         return 0;
1810 }
1811 EXPORT_SYMBOL(xfrm_register_km);
1812
1813 int xfrm_unregister_km(struct xfrm_mgr *km)
1814 {
1815         spin_lock_bh(&xfrm_km_lock);
1816         list_del_rcu(&km->list);
1817         spin_unlock_bh(&xfrm_km_lock);
1818         synchronize_rcu();
1819         return 0;
1820 }
1821 EXPORT_SYMBOL(xfrm_unregister_km);
1822
1823 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1824 {
1825         int err = 0;
1826         if (unlikely(afinfo == NULL))
1827                 return -EINVAL;
1828         if (unlikely(afinfo->family >= NPROTO))
1829                 return -EAFNOSUPPORT;
1830         spin_lock_bh(&xfrm_state_afinfo_lock);
1831         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1832                 err = -ENOBUFS;
1833         else
1834                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1835         spin_unlock_bh(&xfrm_state_afinfo_lock);
1836         return err;
1837 }
1838 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1839
1840 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1841 {
1842         int err = 0;
1843         if (unlikely(afinfo == NULL))
1844                 return -EINVAL;
1845         if (unlikely(afinfo->family >= NPROTO))
1846                 return -EAFNOSUPPORT;
1847         spin_lock_bh(&xfrm_state_afinfo_lock);
1848         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1849                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1850                         err = -EINVAL;
1851                 else
1852                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1853         }
1854         spin_unlock_bh(&xfrm_state_afinfo_lock);
1855         synchronize_rcu();
1856         return err;
1857 }
1858 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1859
1860 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1861 {
1862         struct xfrm_state_afinfo *afinfo;
1863         if (unlikely(family >= NPROTO))
1864                 return NULL;
1865         rcu_read_lock();
1866         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1867         if (unlikely(!afinfo))
1868                 rcu_read_unlock();
1869         return afinfo;
1870 }
1871
1872 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1873 {
1874         rcu_read_unlock();
1875 }
1876
1877 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1878 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1879 {
1880         if (x->tunnel) {
1881                 struct xfrm_state *t = x->tunnel;
1882
1883                 if (atomic_read(&t->tunnel_users) == 2)
1884                         xfrm_state_delete(t);
1885                 atomic_dec(&t->tunnel_users);
1886                 xfrm_state_put(t);
1887                 x->tunnel = NULL;
1888         }
1889 }
1890 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1891
1892 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1893 {
1894         int res;
1895
1896         spin_lock_bh(&x->lock);
1897         if (x->km.state == XFRM_STATE_VALID &&
1898             x->type && x->type->get_mtu)
1899                 res = x->type->get_mtu(x, mtu);
1900         else
1901                 res = mtu - x->props.header_len;
1902         spin_unlock_bh(&x->lock);
1903         return res;
1904 }
1905
1906 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1907 {
1908         struct xfrm_state_afinfo *afinfo;
1909         struct xfrm_mode *inner_mode;
1910         int family = x->props.family;
1911         int err;
1912
1913         err = -EAFNOSUPPORT;
1914         afinfo = xfrm_state_get_afinfo(family);
1915         if (!afinfo)
1916                 goto error;
1917
1918         err = 0;
1919         if (afinfo->init_flags)
1920                 err = afinfo->init_flags(x);
1921
1922         xfrm_state_put_afinfo(afinfo);
1923
1924         if (err)
1925                 goto error;
1926
1927         err = -EPROTONOSUPPORT;
1928
1929         if (x->sel.family != AF_UNSPEC) {
1930                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1931                 if (inner_mode == NULL)
1932                         goto error;
1933
1934                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1935                     family != x->sel.family) {
1936                         xfrm_put_mode(inner_mode);
1937                         goto error;
1938                 }
1939
1940                 x->inner_mode = inner_mode;
1941         } else {
1942                 struct xfrm_mode *inner_mode_iaf;
1943                 int iafamily = AF_INET;
1944
1945                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1946                 if (inner_mode == NULL)
1947                         goto error;
1948
1949                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1950                         xfrm_put_mode(inner_mode);
1951                         goto error;
1952                 }
1953                 x->inner_mode = inner_mode;
1954
1955                 if (x->props.family == AF_INET)
1956                         iafamily = AF_INET6;
1957
1958                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1959                 if (inner_mode_iaf) {
1960                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1961                                 x->inner_mode_iaf = inner_mode_iaf;
1962                         else
1963                                 xfrm_put_mode(inner_mode_iaf);
1964                 }
1965         }
1966
1967         x->type = xfrm_get_type(x->id.proto, family);
1968         if (x->type == NULL)
1969                 goto error;
1970
1971         err = x->type->init_state(x);
1972         if (err)
1973                 goto error;
1974
1975         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1976         if (x->outer_mode == NULL) {
1977                 err = -EPROTONOSUPPORT;
1978                 goto error;
1979         }
1980
1981         if (init_replay) {
1982                 err = xfrm_init_replay(x);
1983                 if (err)
1984                         goto error;
1985         }
1986
1987         x->km.state = XFRM_STATE_VALID;
1988
1989 error:
1990         return err;
1991 }
1992
1993 EXPORT_SYMBOL(__xfrm_init_state);
1994
1995 int xfrm_init_state(struct xfrm_state *x)
1996 {
1997         return __xfrm_init_state(x, true);
1998 }
1999
2000 EXPORT_SYMBOL(xfrm_init_state);
2001
2002 int __net_init xfrm_state_init(struct net *net)
2003 {
2004         unsigned int sz;
2005
2006         INIT_LIST_HEAD(&net->xfrm.state_all);
2007
2008         sz = sizeof(struct hlist_head) * 8;
2009
2010         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2011         if (!net->xfrm.state_bydst)
2012                 goto out_bydst;
2013         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2014         if (!net->xfrm.state_bysrc)
2015                 goto out_bysrc;
2016         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2017         if (!net->xfrm.state_byspi)
2018                 goto out_byspi;
2019         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2020
2021         net->xfrm.state_num = 0;
2022         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2023         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2024         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2025         init_waitqueue_head(&net->xfrm.km_waitq);
2026         return 0;
2027
2028 out_byspi:
2029         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2030 out_bysrc:
2031         xfrm_hash_free(net->xfrm.state_bydst, sz);
2032 out_bydst:
2033         return -ENOMEM;
2034 }
2035
2036 void xfrm_state_fini(struct net *net)
2037 {
2038         struct xfrm_audit audit_info;
2039         unsigned int sz;
2040
2041         flush_work(&net->xfrm.state_hash_work);
2042         audit_info.loginuid = INVALID_UID;
2043         audit_info.sessionid = -1;
2044         audit_info.secid = 0;
2045         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2046         flush_work(&net->xfrm.state_gc_work);
2047
2048         WARN_ON(!list_empty(&net->xfrm.state_all));
2049
2050         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2051         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2052         xfrm_hash_free(net->xfrm.state_byspi, sz);
2053         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2054         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2055         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2056         xfrm_hash_free(net->xfrm.state_bydst, sz);
2057 }
2058
2059 #ifdef CONFIG_AUDITSYSCALL
2060 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2061                                      struct audit_buffer *audit_buf)
2062 {
2063         struct xfrm_sec_ctx *ctx = x->security;
2064         u32 spi = ntohl(x->id.spi);
2065
2066         if (ctx)
2067                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2068                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2069
2070         switch(x->props.family) {
2071         case AF_INET:
2072                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2073                                  &x->props.saddr.a4, &x->id.daddr.a4);
2074                 break;
2075         case AF_INET6:
2076                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2077                                  x->props.saddr.a6, x->id.daddr.a6);
2078                 break;
2079         }
2080
2081         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2082 }
2083
2084 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2085                                       struct audit_buffer *audit_buf)
2086 {
2087         const struct iphdr *iph4;
2088         const struct ipv6hdr *iph6;
2089
2090         switch (family) {
2091         case AF_INET:
2092                 iph4 = ip_hdr(skb);
2093                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2094                                  &iph4->saddr, &iph4->daddr);
2095                 break;
2096         case AF_INET6:
2097                 iph6 = ipv6_hdr(skb);
2098                 audit_log_format(audit_buf,
2099                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2100                                  &iph6->saddr,&iph6->daddr,
2101                                  iph6->flow_lbl[0] & 0x0f,
2102                                  iph6->flow_lbl[1],
2103                                  iph6->flow_lbl[2]);
2104                 break;
2105         }
2106 }
2107
2108 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2109                           kuid_t auid, u32 sessionid, u32 secid)
2110 {
2111         struct audit_buffer *audit_buf;
2112
2113         audit_buf = xfrm_audit_start("SAD-add");
2114         if (audit_buf == NULL)
2115                 return;
2116         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2117         xfrm_audit_helper_sainfo(x, audit_buf);
2118         audit_log_format(audit_buf, " res=%u", result);
2119         audit_log_end(audit_buf);
2120 }
2121 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2122
2123 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2124                              kuid_t auid, u32 sessionid, u32 secid)
2125 {
2126         struct audit_buffer *audit_buf;
2127
2128         audit_buf = xfrm_audit_start("SAD-delete");
2129         if (audit_buf == NULL)
2130                 return;
2131         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2132         xfrm_audit_helper_sainfo(x, audit_buf);
2133         audit_log_format(audit_buf, " res=%u", result);
2134         audit_log_end(audit_buf);
2135 }
2136 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2137
2138 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2139                                       struct sk_buff *skb)
2140 {
2141         struct audit_buffer *audit_buf;
2142         u32 spi;
2143
2144         audit_buf = xfrm_audit_start("SA-replay-overflow");
2145         if (audit_buf == NULL)
2146                 return;
2147         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2148         /* don't record the sequence number because it's inherent in this kind
2149          * of audit message */
2150         spi = ntohl(x->id.spi);
2151         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2152         audit_log_end(audit_buf);
2153 }
2154 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2155
2156 void xfrm_audit_state_replay(struct xfrm_state *x,
2157                              struct sk_buff *skb, __be32 net_seq)
2158 {
2159         struct audit_buffer *audit_buf;
2160         u32 spi;
2161
2162         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2163         if (audit_buf == NULL)
2164                 return;
2165         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2166         spi = ntohl(x->id.spi);
2167         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2168                          spi, spi, ntohl(net_seq));
2169         audit_log_end(audit_buf);
2170 }
2171 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2172
2173 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2174 {
2175         struct audit_buffer *audit_buf;
2176
2177         audit_buf = xfrm_audit_start("SA-notfound");
2178         if (audit_buf == NULL)
2179                 return;
2180         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2181         audit_log_end(audit_buf);
2182 }
2183 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2184
2185 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2186                                __be32 net_spi, __be32 net_seq)
2187 {
2188         struct audit_buffer *audit_buf;
2189         u32 spi;
2190
2191         audit_buf = xfrm_audit_start("SA-notfound");
2192         if (audit_buf == NULL)
2193                 return;
2194         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2195         spi = ntohl(net_spi);
2196         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2197                          spi, spi, ntohl(net_seq));
2198         audit_log_end(audit_buf);
2199 }
2200 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2201
2202 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2203                               struct sk_buff *skb, u8 proto)
2204 {
2205         struct audit_buffer *audit_buf;
2206         __be32 net_spi;
2207         __be32 net_seq;
2208
2209         audit_buf = xfrm_audit_start("SA-icv-failure");
2210         if (audit_buf == NULL)
2211                 return;
2212         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2213         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2214                 u32 spi = ntohl(net_spi);
2215                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2216                                  spi, spi, ntohl(net_seq));
2217         }
2218         audit_log_end(audit_buf);
2219 }
2220 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2221 #endif /* CONFIG_AUDITSYSCALL */