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, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 503                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 504                                 (unsigned long)x);
 505                 x->curlft.add_time = get_seconds();
 506                 x->lft.soft_byte_limit = XFRM_INF;
 507                 x->lft.soft_packet_limit = XFRM_INF;
 508                 x->lft.hard_byte_limit = XFRM_INF;
 509                 x->lft.hard_packet_limit = XFRM_INF;
 510                 x->replay_maxage = 0;
 511                 x->replay_maxdiff = 0;
 512                 x->inner_mode = NULL;
 513                 x->inner_mode_iaf = NULL;
 514                 spin_lock_init(&x->lock);
 515         }
 516         return x;
 517 }
 518 EXPORT_SYMBOL(xfrm_state_alloc);
 519
 520 void __xfrm_state_destroy(struct xfrm_state *x)
 521 {
 522         struct net *net = xs_net(x);
 523
 524         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 525
 526         spin_lock_bh(&xfrm_state_gc_lock);
 527         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 528         spin_unlock_bh(&xfrm_state_gc_lock);
 529         schedule_work(&net->xfrm.state_gc_work);
 530 }
 531 EXPORT_SYMBOL(__xfrm_state_destroy);
 532
 533 int __xfrm_state_delete(struct xfrm_state *x)
 534 {
 535         struct net *net = xs_net(x);
 536         int err = -ESRCH;
 537
 538         if (x->km.state != XFRM_STATE_DEAD) {
 539                 x->km.state = XFRM_STATE_DEAD;
 540                 spin_lock(&xfrm_state_lock);
 541                 list_del(&x->km.all);
 542                 hlist_del(&x->bydst);
 543                 hlist_del(&x->bysrc);
 544                 if (x->id.spi)
 545                         hlist_del(&x->byspi);
 546                 net->xfrm.state_num--;
 547                 spin_unlock(&xfrm_state_lock);
 548
 549                 /* All xfrm_state objects are created by xfrm_state_alloc.
 550                  * The xfrm_state_alloc call gives a reference, and that
 551                  * is what we are dropping here.
 552                  */
 553                 xfrm_state_put(x);
 554                 err = 0;
 555         }
 556
 557         return err;
 558 }
 559 EXPORT_SYMBOL(__xfrm_state_delete);
 560
 561 int xfrm_state_delete(struct xfrm_state *x)
 562 {
 563         int err;
 564
 565         spin_lock_bh(&x->lock);
 566         err = __xfrm_state_delete(x);
 567         spin_unlock_bh(&x->lock);
 568
 569         return err;
 570 }
 571 EXPORT_SYMBOL(xfrm_state_delete);
 572
 573 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 574 static inline int
 575 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 576 {
 577         int i, err = 0;
 578
 579         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 580                 struct xfrm_state *x;
 581
 582                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 583                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 584                            (err = security_xfrm_state_delete(x)) != 0) {
 585                                 xfrm_audit_state_delete(x, 0,
 586                                                         audit_info->loginuid,
 587                                                         audit_info->sessionid,
 588                                                         audit_info->secid);
 589                                 return err;
 590                         }
 591                 }
 592         }
 593
 594         return err;
 595 }
 596 #else
 597 static inline int
 598 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 599 {
 600         return 0;
 601 }
 602 #endif
 603
 604 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 605 {
 606         int i, err = 0, cnt = 0;
 607
 608         spin_lock_bh(&xfrm_state_lock);
 609         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 610         if (err)
 611                 goto out;
 612
 613         err = -ESRCH;
 614         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 615                 struct xfrm_state *x;
 616 restart:
 617                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 618                         if (!xfrm_state_kern(x) &&
 619                             xfrm_id_proto_match(x->id.proto, proto)) {
 620                                 xfrm_state_hold(x);
 621                                 spin_unlock_bh(&xfrm_state_lock);
 622
 623                                 err = xfrm_state_delete(x);
 624                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 625                                                         audit_info->loginuid,
 626                                                         audit_info->sessionid,
 627                                                         audit_info->secid);
 628                                 xfrm_state_put(x);
 629                                 if (!err)
 630                                         cnt++;
 631
 632                                 spin_lock_bh(&xfrm_state_lock);
 633                                 goto restart;
 634                         }
 635                 }
 636         }
 637         if (cnt)
 638                 err = 0;
 639
 640 out:
 641         spin_unlock_bh(&xfrm_state_lock);
 642         wake_up(&net->xfrm.km_waitq);
 643         return err;
 644 }
 645 EXPORT_SYMBOL(xfrm_state_flush);
 646
 647 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 648 {
 649         spin_lock_bh(&xfrm_state_lock);
 650         si->sadcnt = net->xfrm.state_num;
 651         si->sadhcnt = net->xfrm.state_hmask;
 652         si->sadhmcnt = xfrm_state_hashmax;
 653         spin_unlock_bh(&xfrm_state_lock);
 654 }
 655 EXPORT_SYMBOL(xfrm_sad_getinfo);
 656
 657 static int
 658 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 659                     const struct xfrm_tmpl *tmpl,
 660                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 661                     unsigned short family)
 662 {
 663         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 664         if (!afinfo)
 665                 return -1;
 666         afinfo->init_tempsel(&x->sel, fl);
 667
 668         if (family != tmpl->encap_family) {
 669                 xfrm_state_put_afinfo(afinfo);
 670                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 671                 if (!afinfo)
 672                         return -1;
 673         }
 674         afinfo->init_temprop(x, tmpl, daddr, saddr);
 675         xfrm_state_put_afinfo(afinfo);
 676         return 0;
 677 }
 678
 679 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 680                                               const xfrm_address_t *daddr,
 681                                               __be32 spi, u8 proto,
 682                                               unsigned short family)
 683 {
 684         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 685         struct xfrm_state *x;
 686
 687         hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
 688                 if (x->props.family != family ||
 689                     x->id.spi       != spi ||
 690                     x->id.proto     != proto ||
 691                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 692                         continue;
 693
 694                 if ((mark & x->mark.m) != x->mark.v)
 695                         continue;
 696                 xfrm_state_hold(x);
 697                 return x;
 698         }
 699
 700         return NULL;
 701 }
 702
 703 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 704                                                      const xfrm_address_t *daddr,
 705                                                      const xfrm_address_t *saddr,
 706                                                      u8 proto, unsigned short family)
 707 {
 708         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 709         struct xfrm_state *x;
 710
 711         hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
 712                 if (x->props.family != family ||
 713                     x->id.proto     != proto ||
 714                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 715                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 716                         continue;
 717
 718                 if ((mark & x->mark.m) != x->mark.v)
 719                         continue;
 720                 xfrm_state_hold(x);
 721                 return x;
 722         }
 723
 724         return NULL;
 725 }
 726
 727 static inline struct xfrm_state *
 728 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 729 {
 730         struct net *net = xs_net(x);
 731         u32 mark = x->mark.v & x->mark.m;
 732
 733         if (use_spi)
 734                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 735                                            x->id.spi, x->id.proto, family);
 736         else
 737                 return __xfrm_state_lookup_byaddr(net, mark,
 738                                                   &x->id.daddr,
 739                                                   &x->props.saddr,
 740                                                   x->id.proto, family);
 741 }
 742
 743 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 744 {
 745         if (have_hash_collision &&
 746             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 747             net->xfrm.state_num > net->xfrm.state_hmask)
 748                 schedule_work(&net->xfrm.state_hash_work);
 749 }
 750
 751 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 752                                const struct flowi *fl, unsigned short family,
 753                                struct xfrm_state **best, int *acq_in_progress,
 754                                int *error)
 755 {
 756         /* Resolution logic:
 757          * 1. There is a valid state with matching selector. Done.
 758          * 2. Valid state with inappropriate selector. Skip.
 759          *
 760          * Entering area of "sysdeps".
 761          *
 762          * 3. If state is not valid, selector is temporary, it selects
 763          *    only session which triggered previous resolution. Key
 764          *    manager will do something to install a state with proper
 765          *    selector.
 766          */
 767         if (x->km.state == XFRM_STATE_VALID) {
 768                 if ((x->sel.family &&
 769                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 770                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 771                         return;
 772
 773                 if (!*best ||
 774                     (*best)->km.dying > x->km.dying ||
 775                     ((*best)->km.dying == x->km.dying &&
 776                      (*best)->curlft.add_time < x->curlft.add_time))
 777                         *best = x;
 778         } else if (x->km.state == XFRM_STATE_ACQ) {
 779                 *acq_in_progress = 1;
 780         } else if (x->km.state == XFRM_STATE_ERROR ||
 781                    x->km.state == XFRM_STATE_EXPIRED) {
 782                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 783                     security_xfrm_state_pol_flow_match(x, pol, fl))
 784                         *error = -ESRCH;
 785         }
 786 }
 787
 788 struct xfrm_state *
 789 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 790                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 791                 struct xfrm_policy *pol, int *err,
 792                 unsigned short family)
 793 {
 794         static xfrm_address_t saddr_wildcard = { };
 795         struct net *net = xp_net(pol);
 796         unsigned int h, h_wildcard;
 797         struct xfrm_state *x, *x0, *to_put;
 798         int acquire_in_progress = 0;
 799         int error = 0;
 800         struct xfrm_state *best = NULL;
 801         u32 mark = pol->mark.v & pol->mark.m;
 802         unsigned short encap_family = tmpl->encap_family;
 803
 804         to_put = NULL;
 805
 806         spin_lock_bh(&xfrm_state_lock);
 807         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 808         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 809                 if (x->props.family == encap_family &&
 810                     x->props.reqid == tmpl->reqid &&
 811                     (mark & x->mark.m) == x->mark.v &&
 812                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 813                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 814                     tmpl->mode == x->props.mode &&
 815                     tmpl->id.proto == x->id.proto &&
 816                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 817                         xfrm_state_look_at(pol, x, fl, encap_family,
 818                                            &best, &acquire_in_progress, &error);
 819         }
 820         if (best)
 821                 goto found;
 822
 823         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 824         hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
 825                 if (x->props.family == encap_family &&
 826                     x->props.reqid == tmpl->reqid &&
 827                     (mark & x->mark.m) == x->mark.v &&
 828                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 829                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 830                     tmpl->mode == x->props.mode &&
 831                     tmpl->id.proto == x->id.proto &&
 832                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 833                         xfrm_state_look_at(pol, x, fl, encap_family,
 834                                            &best, &acquire_in_progress, &error);
 835         }
 836
 837 found:
 838         x = best;
 839         if (!x && !error && !acquire_in_progress) {
 840                 if (tmpl->id.spi &&
 841                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 842                                               tmpl->id.proto, encap_family)) != NULL) {
 843                         to_put = x0;
 844                         error = -EEXIST;
 845                         goto out;
 846                 }
 847                 x = xfrm_state_alloc(net);
 848                 if (x == NULL) {
 849                         error = -ENOMEM;
 850                         goto out;
 851                 }
 852                 /* Initialize temporary state matching only
 853                  * to current session. */
 854                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 855                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 856
 857                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 858                 if (error) {
 859                         x->km.state = XFRM_STATE_DEAD;
 860                         to_put = x;
 861                         x = NULL;
 862                         goto out;
 863                 }
 864
 865                 if (km_query(x, tmpl, pol) == 0) {
 866                         x->km.state = XFRM_STATE_ACQ;
 867                         list_add(&x->km.all, &net->xfrm.state_all);
 868                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 869                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 870                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 871                         if (x->id.spi) {
 872                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 873                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 874                         }
 875                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 876                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 877                         net->xfrm.state_num++;
 878                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 879                 } else {
 880                         x->km.state = XFRM_STATE_DEAD;
 881                         to_put = x;
 882                         x = NULL;
 883                         error = -ESRCH;
 884                 }
 885         }
 886 out:
 887         if (x)
 888                 xfrm_state_hold(x);
 889         else
 890                 *err = acquire_in_progress ? -EAGAIN : error;
 891         spin_unlock_bh(&xfrm_state_lock);
 892         if (to_put)
 893                 xfrm_state_put(to_put);
 894         return x;
 895 }
 896
 897 struct xfrm_state *
 898 xfrm_stateonly_find(struct net *net, u32 mark,
 899                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 900                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 901 {
 902         unsigned int h;
 903         struct xfrm_state *rx = NULL, *x = NULL;
 904
 905         spin_lock(&xfrm_state_lock);
 906         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 907         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 908                 if (x->props.family == family &&
 909                     x->props.reqid == reqid &&
 910                     (mark & x->mark.m) == x->mark.v &&
 911                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 912                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 913                     mode == x->props.mode &&
 914                     proto == x->id.proto &&
 915                     x->km.state == XFRM_STATE_VALID) {
 916                         rx = x;
 917                         break;
 918                 }
 919         }
 920
 921         if (rx)
 922                 xfrm_state_hold(rx);
 923         spin_unlock(&xfrm_state_lock);
 924
 925
 926         return rx;
 927 }
 928 EXPORT_SYMBOL(xfrm_stateonly_find);
 929
 930 static void __xfrm_state_insert(struct xfrm_state *x)
 931 {
 932         struct net *net = xs_net(x);
 933         unsigned int h;
 934
 935         list_add(&x->km.all, &net->xfrm.state_all);
 936
 937         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 938                           x->props.reqid, x->props.family);
 939         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 940
 941         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 942         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 943
 944         if (x->id.spi) {
 945                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 946                                   x->props.family);
 947
 948                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 949         }
 950
 951         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 952         if (x->replay_maxage)
 953                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 954
 955         wake_up(&net->xfrm.km_waitq);
 956
 957         net->xfrm.state_num++;
 958
 959         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 960 }
 961
 962 /* xfrm_state_lock is held */
 963 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 964 {
 965         struct net *net = xs_net(xnew);
 966         unsigned short family = xnew->props.family;
 967         u32 reqid = xnew->props.reqid;
 968         struct xfrm_state *x;
 969         unsigned int h;
 970         u32 mark = xnew->mark.v & xnew->mark.m;
 971
 972         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 973         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 974                 if (x->props.family     == family &&
 975                     x->props.reqid      == reqid &&
 976                     (mark & x->mark.m) == x->mark.v &&
 977                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 978                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
 979                         x->genid++;
 980         }
 981 }
 982
 983 void xfrm_state_insert(struct xfrm_state *x)
 984 {
 985         spin_lock_bh(&xfrm_state_lock);
 986         __xfrm_state_bump_genids(x);
 987         __xfrm_state_insert(x);
 988         spin_unlock_bh(&xfrm_state_lock);
 989 }
 990 EXPORT_SYMBOL(xfrm_state_insert);
 991
 992 /* xfrm_state_lock is held */
 993 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
 994                                           unsigned short family, u8 mode,
 995                                           u32 reqid, u8 proto,
 996                                           const xfrm_address_t *daddr,
 997                                           const xfrm_address_t *saddr, int create)
 998 {
 999         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1000         struct xfrm_state *x;
1001         u32 mark = m->v & m->m;
1002
1003         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1004                 if (x->props.reqid  != reqid ||
1005                     x->props.mode   != mode ||
1006                     x->props.family != family ||
1007                     x->km.state     != XFRM_STATE_ACQ ||
1008                     x->id.spi       != 0 ||
1009                     x->id.proto     != proto ||
1010                     (mark & x->mark.m) != x->mark.v ||
1011                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1012                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1013                         continue;
1014
1015                 xfrm_state_hold(x);
1016                 return x;
1017         }
1018
1019         if (!create)
1020                 return NULL;
1021
1022         x = xfrm_state_alloc(net);
1023         if (likely(x)) {
1024                 switch (family) {
1025                 case AF_INET:
1026                         x->sel.daddr.a4 = daddr->a4;
1027                         x->sel.saddr.a4 = saddr->a4;
1028                         x->sel.prefixlen_d = 32;
1029                         x->sel.prefixlen_s = 32;
1030                         x->props.saddr.a4 = saddr->a4;
1031                         x->id.daddr.a4 = daddr->a4;
1032                         break;
1033
1034                 case AF_INET6:
1035                         *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1036                         *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1037                         x->sel.prefixlen_d = 128;
1038                         x->sel.prefixlen_s = 128;
1039                         *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1040                         *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1041                         break;
1042                 }
1043
1044                 x->km.state = XFRM_STATE_ACQ;
1045                 x->id.proto = proto;
1046                 x->props.family = family;
1047                 x->props.mode = mode;
1048                 x->props.reqid = reqid;
1049                 x->mark.v = m->v;
1050                 x->mark.m = m->m;
1051                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1052                 xfrm_state_hold(x);
1053                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1054                 list_add(&x->km.all, &net->xfrm.state_all);
1055                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1056                 h = xfrm_src_hash(net, daddr, saddr, family);
1057                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1058
1059                 net->xfrm.state_num++;
1060
1061                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1062         }
1063
1064         return x;
1065 }
1066
1067 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1068
1069 int xfrm_state_add(struct xfrm_state *x)
1070 {
1071         struct net *net = xs_net(x);
1072         struct xfrm_state *x1, *to_put;
1073         int family;
1074         int err;
1075         u32 mark = x->mark.v & x->mark.m;
1076         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1077
1078         family = x->props.family;
1079
1080         to_put = NULL;
1081
1082         spin_lock_bh(&xfrm_state_lock);
1083
1084         x1 = __xfrm_state_locate(x, use_spi, family);
1085         if (x1) {
1086                 to_put = x1;
1087                 x1 = NULL;
1088                 err = -EEXIST;
1089                 goto out;
1090         }
1091
1092         if (use_spi && x->km.seq) {
1093                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1094                 if (x1 && ((x1->id.proto != x->id.proto) ||
1095                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1096                         to_put = x1;
1097                         x1 = NULL;
1098                 }
1099         }
1100
1101         if (use_spi && !x1)
1102                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1103                                      x->props.reqid, x->id.proto,
1104                                      &x->id.daddr, &x->props.saddr, 0);
1105
1106         __xfrm_state_bump_genids(x);
1107         __xfrm_state_insert(x);
1108         err = 0;
1109
1110 out:
1111         spin_unlock_bh(&xfrm_state_lock);
1112
1113         if (x1) {
1114                 xfrm_state_delete(x1);
1115                 xfrm_state_put(x1);
1116         }
1117
1118         if (to_put)
1119                 xfrm_state_put(to_put);
1120
1121         return err;
1122 }
1123 EXPORT_SYMBOL(xfrm_state_add);
1124
1125 #ifdef CONFIG_XFRM_MIGRATE
1126 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1127 {
1128         struct net *net = xs_net(orig);
1129         int err = -ENOMEM;
1130         struct xfrm_state *x = xfrm_state_alloc(net);
1131         if (!x)
1132                 goto out;
1133
1134         memcpy(&x->id, &orig->id, sizeof(x->id));
1135         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1136         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1137         x->props.mode = orig->props.mode;
1138         x->props.replay_window = orig->props.replay_window;
1139         x->props.reqid = orig->props.reqid;
1140         x->props.family = orig->props.family;
1141         x->props.saddr = orig->props.saddr;
1142
1143         if (orig->aalg) {
1144                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1145                 if (!x->aalg)
1146                         goto error;
1147         }
1148         x->props.aalgo = orig->props.aalgo;
1149
1150         if (orig->ealg) {
1151                 x->ealg = xfrm_algo_clone(orig->ealg);
1152                 if (!x->ealg)
1153                         goto error;
1154         }
1155         x->props.ealgo = orig->props.ealgo;
1156
1157         if (orig->calg) {
1158                 x->calg = xfrm_algo_clone(orig->calg);
1159                 if (!x->calg)
1160                         goto error;
1161         }
1162         x->props.calgo = orig->props.calgo;
1163
1164         if (orig->encap) {
1165                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1166                 if (!x->encap)
1167                         goto error;
1168         }
1169
1170         if (orig->coaddr) {
1171                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1172                                     GFP_KERNEL);
1173                 if (!x->coaddr)
1174                         goto error;
1175         }
1176
1177         if (orig->replay_esn) {
1178                 err = xfrm_replay_clone(x, orig);
1179                 if (err)
1180                         goto error;
1181         }
1182
1183         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1184
1185         err = xfrm_init_state(x);
1186         if (err)
1187                 goto error;
1188
1189         x->props.flags = orig->props.flags;
1190
1191         x->curlft.add_time = orig->curlft.add_time;
1192         x->km.state = orig->km.state;
1193         x->km.seq = orig->km.seq;
1194
1195         return x;
1196
1197  error:
1198         xfrm_state_put(x);
1199 out:
1200         if (errp)
1201                 *errp = err;
1202         return NULL;
1203 }
1204
1205 /* xfrm_state_lock is held */
1206 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1207 {
1208         unsigned int h;
1209         struct xfrm_state *x;
1210
1211         if (m->reqid) {
1212                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1213                                   m->reqid, m->old_family);
1214                 hlist_for_each_entry(x, init_net.xfrm.state_bydst+h, bydst) {
1215                         if (x->props.mode != m->mode ||
1216                             x->id.proto != m->proto)
1217                                 continue;
1218                         if (m->reqid && x->props.reqid != m->reqid)
1219                                 continue;
1220                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1221                                              m->old_family) ||
1222                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1223                                              m->old_family))
1224                                 continue;
1225                         xfrm_state_hold(x);
1226                         return x;
1227                 }
1228         } else {
1229                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1230                                   m->old_family);
1231                 hlist_for_each_entry(x, init_net.xfrm.state_bysrc+h, bysrc) {
1232                         if (x->props.mode != m->mode ||
1233                             x->id.proto != m->proto)
1234                                 continue;
1235                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1236                                              m->old_family) ||
1237                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1238                                              m->old_family))
1239                                 continue;
1240                         xfrm_state_hold(x);
1241                         return x;
1242                 }
1243         }
1244
1245         return NULL;
1246 }
1247 EXPORT_SYMBOL(xfrm_migrate_state_find);
1248
1249 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1250                                        struct xfrm_migrate *m)
1251 {
1252         struct xfrm_state *xc;
1253         int err;
1254
1255         xc = xfrm_state_clone(x, &err);
1256         if (!xc)
1257                 return NULL;
1258
1259         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1260         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1261
1262         /* add state */
1263         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1264                 /* a care is needed when the destination address of the
1265                    state is to be updated as it is a part of triplet */
1266                 xfrm_state_insert(xc);
1267         } else {
1268                 if ((err = xfrm_state_add(xc)) < 0)
1269                         goto error;
1270         }
1271
1272         return xc;
1273 error:
1274         xfrm_state_put(xc);
1275         return NULL;
1276 }
1277 EXPORT_SYMBOL(xfrm_state_migrate);
1278 #endif
1279
1280 int xfrm_state_update(struct xfrm_state *x)
1281 {
1282         struct xfrm_state *x1, *to_put;
1283         int err;
1284         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1285
1286         to_put = NULL;
1287
1288         spin_lock_bh(&xfrm_state_lock);
1289         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1290
1291         err = -ESRCH;
1292         if (!x1)
1293                 goto out;
1294
1295         if (xfrm_state_kern(x1)) {
1296                 to_put = x1;
1297                 err = -EEXIST;
1298                 goto out;
1299         }
1300
1301         if (x1->km.state == XFRM_STATE_ACQ) {
1302                 __xfrm_state_insert(x);
1303                 x = NULL;
1304         }
1305         err = 0;
1306
1307 out:
1308         spin_unlock_bh(&xfrm_state_lock);
1309
1310         if (to_put)
1311                 xfrm_state_put(to_put);
1312
1313         if (err)
1314                 return err;
1315
1316         if (!x) {
1317                 xfrm_state_delete(x1);
1318                 xfrm_state_put(x1);
1319                 return 0;
1320         }
1321
1322         err = -EINVAL;
1323         spin_lock_bh(&x1->lock);
1324         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1325                 if (x->encap && x1->encap)
1326                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1327                 if (x->coaddr && x1->coaddr) {
1328                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1329                 }
1330                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1331                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1332                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1333                 x1->km.dying = 0;
1334
1335                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1336                 if (x1->curlft.use_time)
1337                         xfrm_state_check_expire(x1);
1338
1339                 err = 0;
1340                 x->km.state = XFRM_STATE_DEAD;
1341                 __xfrm_state_put(x);
1342         }
1343         spin_unlock_bh(&x1->lock);
1344
1345         xfrm_state_put(x1);
1346
1347         return err;
1348 }
1349 EXPORT_SYMBOL(xfrm_state_update);
1350
1351 int xfrm_state_check_expire(struct xfrm_state *x)
1352 {
1353         if (!x->curlft.use_time)
1354                 x->curlft.use_time = get_seconds();
1355
1356         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1357             x->curlft.packets >= x->lft.hard_packet_limit) {
1358                 x->km.state = XFRM_STATE_EXPIRED;
1359                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1360                 return -EINVAL;
1361         }
1362
1363         if (!x->km.dying &&
1364             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1365              x->curlft.packets >= x->lft.soft_packet_limit)) {
1366                 x->km.dying = 1;
1367                 km_state_expired(x, 0, 0);
1368         }
1369         return 0;
1370 }
1371 EXPORT_SYMBOL(xfrm_state_check_expire);
1372
1373 struct xfrm_state *
1374 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1375                   u8 proto, unsigned short family)
1376 {
1377         struct xfrm_state *x;
1378
1379         spin_lock_bh(&xfrm_state_lock);
1380         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1381         spin_unlock_bh(&xfrm_state_lock);
1382         return x;
1383 }
1384 EXPORT_SYMBOL(xfrm_state_lookup);
1385
1386 struct xfrm_state *
1387 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1388                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1389                          u8 proto, unsigned short family)
1390 {
1391         struct xfrm_state *x;
1392
1393         spin_lock_bh(&xfrm_state_lock);
1394         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1395         spin_unlock_bh(&xfrm_state_lock);
1396         return x;
1397 }
1398 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1399
1400 struct xfrm_state *
1401 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1402               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1403               int create, unsigned short family)
1404 {
1405         struct xfrm_state *x;
1406
1407         spin_lock_bh(&xfrm_state_lock);
1408         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1409         spin_unlock_bh(&xfrm_state_lock);
1410
1411         return x;
1412 }
1413 EXPORT_SYMBOL(xfrm_find_acq);
1414
1415 #ifdef CONFIG_XFRM_SUB_POLICY
1416 int
1417 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1418                unsigned short family)
1419 {
1420         int err = 0;
1421         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1422         if (!afinfo)
1423                 return -EAFNOSUPPORT;
1424
1425         spin_lock_bh(&xfrm_state_lock);
1426         if (afinfo->tmpl_sort)
1427                 err = afinfo->tmpl_sort(dst, src, n);
1428         spin_unlock_bh(&xfrm_state_lock);
1429         xfrm_state_put_afinfo(afinfo);
1430         return err;
1431 }
1432 EXPORT_SYMBOL(xfrm_tmpl_sort);
1433
1434 int
1435 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1436                 unsigned short family)
1437 {
1438         int err = 0;
1439         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1440         if (!afinfo)
1441                 return -EAFNOSUPPORT;
1442
1443         spin_lock_bh(&xfrm_state_lock);
1444         if (afinfo->state_sort)
1445                 err = afinfo->state_sort(dst, src, n);
1446         spin_unlock_bh(&xfrm_state_lock);
1447         xfrm_state_put_afinfo(afinfo);
1448         return err;
1449 }
1450 EXPORT_SYMBOL(xfrm_state_sort);
1451 #endif
1452
1453 /* Silly enough, but I'm lazy to build resolution list */
1454
1455 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1456 {
1457         int i;
1458
1459         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1460                 struct xfrm_state *x;
1461
1462                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1463                         if (x->km.seq == seq &&
1464                             (mark & x->mark.m) == x->mark.v &&
1465                             x->km.state == XFRM_STATE_ACQ) {
1466                                 xfrm_state_hold(x);
1467                                 return x;
1468                         }
1469                 }
1470         }
1471         return NULL;
1472 }
1473
1474 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1475 {
1476         struct xfrm_state *x;
1477
1478         spin_lock_bh(&xfrm_state_lock);
1479         x = __xfrm_find_acq_byseq(net, mark, seq);
1480         spin_unlock_bh(&xfrm_state_lock);
1481         return x;
1482 }
1483 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1484
1485 u32 xfrm_get_acqseq(void)
1486 {
1487         u32 res;
1488         static atomic_t acqseq;
1489
1490         do {
1491                 res = atomic_inc_return(&acqseq);
1492         } while (!res);
1493
1494         return res;
1495 }
1496 EXPORT_SYMBOL(xfrm_get_acqseq);
1497
1498 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1499 {
1500         struct net *net = xs_net(x);
1501         unsigned int h;
1502         struct xfrm_state *x0;
1503         int err = -ENOENT;
1504         __be32 minspi = htonl(low);
1505         __be32 maxspi = htonl(high);
1506         u32 mark = x->mark.v & x->mark.m;
1507
1508         spin_lock_bh(&x->lock);
1509         if (x->km.state == XFRM_STATE_DEAD)
1510                 goto unlock;
1511
1512         err = 0;
1513         if (x->id.spi)
1514                 goto unlock;
1515
1516         err = -ENOENT;
1517
1518         if (minspi == maxspi) {
1519                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1520                 if (x0) {
1521                         xfrm_state_put(x0);
1522                         goto unlock;
1523                 }
1524                 x->id.spi = minspi;
1525         } else {
1526                 u32 spi = 0;
1527                 for (h=0; h<high-low+1; h++) {
1528                         spi = low + net_random()%(high-low+1);
1529                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1530                         if (x0 == NULL) {
1531                                 x->id.spi = htonl(spi);
1532                                 break;
1533                         }
1534                         xfrm_state_put(x0);
1535                 }
1536         }
1537         if (x->id.spi) {
1538                 spin_lock_bh(&xfrm_state_lock);
1539                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1540                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1541                 spin_unlock_bh(&xfrm_state_lock);
1542
1543                 err = 0;
1544         }
1545
1546 unlock:
1547         spin_unlock_bh(&x->lock);
1548
1549         return err;
1550 }
1551 EXPORT_SYMBOL(xfrm_alloc_spi);
1552
1553 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1554                     int (*func)(struct xfrm_state *, int, void*),
1555                     void *data)
1556 {
1557         struct xfrm_state *state;
1558         struct xfrm_state_walk *x;
1559         int err = 0;
1560
1561         if (walk->seq != 0 && list_empty(&walk->all))
1562                 return 0;
1563
1564         spin_lock_bh(&xfrm_state_lock);
1565         if (list_empty(&walk->all))
1566                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1567         else
1568                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1569         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1570                 if (x->state == XFRM_STATE_DEAD)
1571                         continue;
1572                 state = container_of(x, struct xfrm_state, km);
1573                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1574                         continue;
1575                 err = func(state, walk->seq, data);
1576                 if (err) {
1577                         list_move_tail(&walk->all, &x->all);
1578                         goto out;
1579                 }
1580                 walk->seq++;
1581         }
1582         if (walk->seq == 0) {
1583                 err = -ENOENT;
1584                 goto out;
1585         }
1586         list_del_init(&walk->all);
1587 out:
1588         spin_unlock_bh(&xfrm_state_lock);
1589         return err;
1590 }
1591 EXPORT_SYMBOL(xfrm_state_walk);
1592
1593 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1594 {
1595         INIT_LIST_HEAD(&walk->all);
1596         walk->proto = proto;
1597         walk->state = XFRM_STATE_DEAD;
1598         walk->seq = 0;
1599 }
1600 EXPORT_SYMBOL(xfrm_state_walk_init);
1601
1602 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1603 {
1604         if (list_empty(&walk->all))
1605                 return;
1606
1607         spin_lock_bh(&xfrm_state_lock);
1608         list_del(&walk->all);
1609         spin_unlock_bh(&xfrm_state_lock);
1610 }
1611 EXPORT_SYMBOL(xfrm_state_walk_done);
1612
1613 static void xfrm_replay_timer_handler(unsigned long data)
1614 {
1615         struct xfrm_state *x = (struct xfrm_state*)data;
1616
1617         spin_lock(&x->lock);
1618
1619         if (x->km.state == XFRM_STATE_VALID) {
1620                 if (xfrm_aevent_is_on(xs_net(x)))
1621                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1622                 else
1623                         x->xflags |= XFRM_TIME_DEFER;
1624         }
1625
1626         spin_unlock(&x->lock);
1627 }
1628
1629 static LIST_HEAD(xfrm_km_list);
1630
1631 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1632 {
1633         struct xfrm_mgr *km;
1634
1635         rcu_read_lock();
1636         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1637                 if (km->notify_policy)
1638                         km->notify_policy(xp, dir, c);
1639         rcu_read_unlock();
1640 }
1641
1642 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1643 {
1644         struct xfrm_mgr *km;
1645         rcu_read_lock();
1646         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1647                 if (km->notify)
1648                         km->notify(x, c);
1649         rcu_read_unlock();
1650 }
1651
1652 EXPORT_SYMBOL(km_policy_notify);
1653 EXPORT_SYMBOL(km_state_notify);
1654
1655 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1656 {
1657         struct net *net = xs_net(x);
1658         struct km_event c;
1659
1660         c.data.hard = hard;
1661         c.portid = portid;
1662         c.event = XFRM_MSG_EXPIRE;
1663         km_state_notify(x, &c);
1664
1665         if (hard)
1666                 wake_up(&net->xfrm.km_waitq);
1667 }
1668
1669 EXPORT_SYMBOL(km_state_expired);
1670 /*
1671  * We send to all registered managers regardless of failure
1672  * We are happy with one success
1673 */
1674 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1675 {
1676         int err = -EINVAL, acqret;
1677         struct xfrm_mgr *km;
1678
1679         rcu_read_lock();
1680         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1681                 acqret = km->acquire(x, t, pol);
1682                 if (!acqret)
1683                         err = acqret;
1684         }
1685         rcu_read_unlock();
1686         return err;
1687 }
1688 EXPORT_SYMBOL(km_query);
1689
1690 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1691 {
1692         int err = -EINVAL;
1693         struct xfrm_mgr *km;
1694
1695         rcu_read_lock();
1696         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1697                 if (km->new_mapping)
1698                         err = km->new_mapping(x, ipaddr, sport);
1699                 if (!err)
1700                         break;
1701         }
1702         rcu_read_unlock();
1703         return err;
1704 }
1705 EXPORT_SYMBOL(km_new_mapping);
1706
1707 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1708 {
1709         struct net *net = xp_net(pol);
1710         struct km_event c;
1711
1712         c.data.hard = hard;
1713         c.portid = portid;
1714         c.event = XFRM_MSG_POLEXPIRE;
1715         km_policy_notify(pol, dir, &c);
1716
1717         if (hard)
1718                 wake_up(&net->xfrm.km_waitq);
1719 }
1720 EXPORT_SYMBOL(km_policy_expired);
1721
1722 #ifdef CONFIG_XFRM_MIGRATE
1723 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1724                const struct xfrm_migrate *m, int num_migrate,
1725                const struct xfrm_kmaddress *k)
1726 {
1727         int err = -EINVAL;
1728         int ret;
1729         struct xfrm_mgr *km;
1730
1731         rcu_read_lock();
1732         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1733                 if (km->migrate) {
1734                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1735                         if (!ret)
1736                                 err = ret;
1737                 }
1738         }
1739         rcu_read_unlock();
1740         return err;
1741 }
1742 EXPORT_SYMBOL(km_migrate);
1743 #endif
1744
1745 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1746 {
1747         int err = -EINVAL;
1748         int ret;
1749         struct xfrm_mgr *km;
1750
1751         rcu_read_lock();
1752         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1753                 if (km->report) {
1754                         ret = km->report(net, proto, sel, addr);
1755                         if (!ret)
1756                                 err = ret;
1757                 }
1758         }
1759         rcu_read_unlock();
1760         return err;
1761 }
1762 EXPORT_SYMBOL(km_report);
1763
1764 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1765 {
1766         int err;
1767         u8 *data;
1768         struct xfrm_mgr *km;
1769         struct xfrm_policy *pol = NULL;
1770
1771         if (optlen <= 0 || optlen > PAGE_SIZE)
1772                 return -EMSGSIZE;
1773
1774         data = kmalloc(optlen, GFP_KERNEL);
1775         if (!data)
1776                 return -ENOMEM;
1777
1778         err = -EFAULT;
1779         if (copy_from_user(data, optval, optlen))
1780                 goto out;
1781
1782         err = -EINVAL;
1783         rcu_read_lock();
1784         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1785                 pol = km->compile_policy(sk, optname, data,
1786                                          optlen, &err);
1787                 if (err >= 0)
1788                         break;
1789         }
1790         rcu_read_unlock();
1791
1792         if (err >= 0) {
1793                 xfrm_sk_policy_insert(sk, err, pol);
1794                 xfrm_pol_put(pol);
1795                 err = 0;
1796         }
1797
1798 out:
1799         kfree(data);
1800         return err;
1801 }
1802 EXPORT_SYMBOL(xfrm_user_policy);
1803
1804 static DEFINE_SPINLOCK(xfrm_km_lock);
1805
1806 int xfrm_register_km(struct xfrm_mgr *km)
1807 {
1808         spin_lock_bh(&xfrm_km_lock);
1809         list_add_tail_rcu(&km->list, &xfrm_km_list);
1810         spin_unlock_bh(&xfrm_km_lock);
1811         return 0;
1812 }
1813 EXPORT_SYMBOL(xfrm_register_km);
1814
1815 int xfrm_unregister_km(struct xfrm_mgr *km)
1816 {
1817         spin_lock_bh(&xfrm_km_lock);
1818         list_del_rcu(&km->list);
1819         spin_unlock_bh(&xfrm_km_lock);
1820         synchronize_rcu();
1821         return 0;
1822 }
1823 EXPORT_SYMBOL(xfrm_unregister_km);
1824
1825 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1826 {
1827         int err = 0;
1828         if (unlikely(afinfo == NULL))
1829                 return -EINVAL;
1830         if (unlikely(afinfo->family >= NPROTO))
1831                 return -EAFNOSUPPORT;
1832         spin_lock_bh(&xfrm_state_afinfo_lock);
1833         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1834                 err = -ENOBUFS;
1835         else
1836                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1837         spin_unlock_bh(&xfrm_state_afinfo_lock);
1838         return err;
1839 }
1840 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1841
1842 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1843 {
1844         int err = 0;
1845         if (unlikely(afinfo == NULL))
1846                 return -EINVAL;
1847         if (unlikely(afinfo->family >= NPROTO))
1848                 return -EAFNOSUPPORT;
1849         spin_lock_bh(&xfrm_state_afinfo_lock);
1850         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1851                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1852                         err = -EINVAL;
1853                 else
1854                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1855         }
1856         spin_unlock_bh(&xfrm_state_afinfo_lock);
1857         synchronize_rcu();
1858         return err;
1859 }
1860 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1861
1862 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1863 {
1864         struct xfrm_state_afinfo *afinfo;
1865         if (unlikely(family >= NPROTO))
1866                 return NULL;
1867         rcu_read_lock();
1868         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1869         if (unlikely(!afinfo))
1870                 rcu_read_unlock();
1871         return afinfo;
1872 }
1873
1874 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1875 {
1876         rcu_read_unlock();
1877 }
1878
1879 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1880 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1881 {
1882         if (x->tunnel) {
1883                 struct xfrm_state *t = x->tunnel;
1884
1885                 if (atomic_read(&t->tunnel_users) == 2)
1886                         xfrm_state_delete(t);
1887                 atomic_dec(&t->tunnel_users);
1888                 xfrm_state_put(t);
1889                 x->tunnel = NULL;
1890         }
1891 }
1892 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1893
1894 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1895 {
1896         int res;
1897
1898         spin_lock_bh(&x->lock);
1899         if (x->km.state == XFRM_STATE_VALID &&
1900             x->type && x->type->get_mtu)
1901                 res = x->type->get_mtu(x, mtu);
1902         else
1903                 res = mtu - x->props.header_len;
1904         spin_unlock_bh(&x->lock);
1905         return res;
1906 }
1907
1908 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1909 {
1910         struct xfrm_state_afinfo *afinfo;
1911         struct xfrm_mode *inner_mode;
1912         int family = x->props.family;
1913         int err;
1914
1915         err = -EAFNOSUPPORT;
1916         afinfo = xfrm_state_get_afinfo(family);
1917         if (!afinfo)
1918                 goto error;
1919
1920         err = 0;
1921         if (afinfo->init_flags)
1922                 err = afinfo->init_flags(x);
1923
1924         xfrm_state_put_afinfo(afinfo);
1925
1926         if (err)
1927                 goto error;
1928
1929         err = -EPROTONOSUPPORT;
1930
1931         if (x->sel.family != AF_UNSPEC) {
1932                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1933                 if (inner_mode == NULL)
1934                         goto error;
1935
1936                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1937                     family != x->sel.family) {
1938                         xfrm_put_mode(inner_mode);
1939                         goto error;
1940                 }
1941
1942                 x->inner_mode = inner_mode;
1943         } else {
1944                 struct xfrm_mode *inner_mode_iaf;
1945                 int iafamily = AF_INET;
1946
1947                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1948                 if (inner_mode == NULL)
1949                         goto error;
1950
1951                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1952                         xfrm_put_mode(inner_mode);
1953                         goto error;
1954                 }
1955                 x->inner_mode = inner_mode;
1956
1957                 if (x->props.family == AF_INET)
1958                         iafamily = AF_INET6;
1959
1960                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1961                 if (inner_mode_iaf) {
1962                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1963                                 x->inner_mode_iaf = inner_mode_iaf;
1964                         else
1965                                 xfrm_put_mode(inner_mode_iaf);
1966                 }
1967         }
1968
1969         x->type = xfrm_get_type(x->id.proto, family);
1970         if (x->type == NULL)
1971                 goto error;
1972
1973         err = x->type->init_state(x);
1974         if (err)
1975                 goto error;
1976
1977         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1978         if (x->outer_mode == NULL) {
1979                 err = -EPROTONOSUPPORT;
1980                 goto error;
1981         }
1982
1983         if (init_replay) {
1984                 err = xfrm_init_replay(x);
1985                 if (err)
1986                         goto error;
1987         }
1988
1989         x->km.state = XFRM_STATE_VALID;
1990
1991 error:
1992         return err;
1993 }
1994
1995 EXPORT_SYMBOL(__xfrm_init_state);
1996
1997 int xfrm_init_state(struct xfrm_state *x)
1998 {
1999         return __xfrm_init_state(x, true);
2000 }
2001
2002 EXPORT_SYMBOL(xfrm_init_state);
2003
2004 int __net_init xfrm_state_init(struct net *net)
2005 {
2006         unsigned int sz;
2007
2008         INIT_LIST_HEAD(&net->xfrm.state_all);
2009
2010         sz = sizeof(struct hlist_head) * 8;
2011
2012         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2013         if (!net->xfrm.state_bydst)
2014                 goto out_bydst;
2015         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2016         if (!net->xfrm.state_bysrc)
2017                 goto out_bysrc;
2018         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2019         if (!net->xfrm.state_byspi)
2020                 goto out_byspi;
2021         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2022
2023         net->xfrm.state_num = 0;
2024         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2025         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2026         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2027         init_waitqueue_head(&net->xfrm.km_waitq);
2028         return 0;
2029
2030 out_byspi:
2031         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2032 out_bysrc:
2033         xfrm_hash_free(net->xfrm.state_bydst, sz);
2034 out_bydst:
2035         return -ENOMEM;
2036 }
2037
2038 void xfrm_state_fini(struct net *net)
2039 {
2040         struct xfrm_audit audit_info;
2041         unsigned int sz;
2042
2043         flush_work(&net->xfrm.state_hash_work);
2044         audit_info.loginuid = INVALID_UID;
2045         audit_info.sessionid = -1;
2046         audit_info.secid = 0;
2047         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2048         flush_work(&net->xfrm.state_gc_work);
2049
2050         WARN_ON(!list_empty(&net->xfrm.state_all));
2051
2052         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2053         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2054         xfrm_hash_free(net->xfrm.state_byspi, sz);
2055         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2056         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2057         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2058         xfrm_hash_free(net->xfrm.state_bydst, sz);
2059 }
2060
2061 #ifdef CONFIG_AUDITSYSCALL
2062 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2063                                      struct audit_buffer *audit_buf)
2064 {
2065         struct xfrm_sec_ctx *ctx = x->security;
2066         u32 spi = ntohl(x->id.spi);
2067
2068         if (ctx)
2069                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2070                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2071
2072         switch(x->props.family) {
2073         case AF_INET:
2074                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2075                                  &x->props.saddr.a4, &x->id.daddr.a4);
2076                 break;
2077         case AF_INET6:
2078                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2079                                  x->props.saddr.a6, x->id.daddr.a6);
2080                 break;
2081         }
2082
2083         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2084 }
2085
2086 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2087                                       struct audit_buffer *audit_buf)
2088 {
2089         const struct iphdr *iph4;
2090         const struct ipv6hdr *iph6;
2091
2092         switch (family) {
2093         case AF_INET:
2094                 iph4 = ip_hdr(skb);
2095                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2096                                  &iph4->saddr, &iph4->daddr);
2097                 break;
2098         case AF_INET6:
2099                 iph6 = ipv6_hdr(skb);
2100                 audit_log_format(audit_buf,
2101                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2102                                  &iph6->saddr,&iph6->daddr,
2103                                  iph6->flow_lbl[0] & 0x0f,
2104                                  iph6->flow_lbl[1],
2105                                  iph6->flow_lbl[2]);
2106                 break;
2107         }
2108 }
2109
2110 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2111                           kuid_t auid, u32 sessionid, u32 secid)
2112 {
2113         struct audit_buffer *audit_buf;
2114
2115         audit_buf = xfrm_audit_start("SAD-add");
2116         if (audit_buf == NULL)
2117                 return;
2118         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2119         xfrm_audit_helper_sainfo(x, audit_buf);
2120         audit_log_format(audit_buf, " res=%u", result);
2121         audit_log_end(audit_buf);
2122 }
2123 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2124
2125 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2126                              kuid_t auid, u32 sessionid, u32 secid)
2127 {
2128         struct audit_buffer *audit_buf;
2129
2130         audit_buf = xfrm_audit_start("SAD-delete");
2131         if (audit_buf == NULL)
2132                 return;
2133         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2134         xfrm_audit_helper_sainfo(x, audit_buf);
2135         audit_log_format(audit_buf, " res=%u", result);
2136         audit_log_end(audit_buf);
2137 }
2138 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2139
2140 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2141                                       struct sk_buff *skb)
2142 {
2143         struct audit_buffer *audit_buf;
2144         u32 spi;
2145
2146         audit_buf = xfrm_audit_start("SA-replay-overflow");
2147         if (audit_buf == NULL)
2148                 return;
2149         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2150         /* don't record the sequence number because it's inherent in this kind
2151          * of audit message */
2152         spi = ntohl(x->id.spi);
2153         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2154         audit_log_end(audit_buf);
2155 }
2156 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2157
2158 void xfrm_audit_state_replay(struct xfrm_state *x,
2159                              struct sk_buff *skb, __be32 net_seq)
2160 {
2161         struct audit_buffer *audit_buf;
2162         u32 spi;
2163
2164         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2165         if (audit_buf == NULL)
2166                 return;
2167         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2168         spi = ntohl(x->id.spi);
2169         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2170                          spi, spi, ntohl(net_seq));
2171         audit_log_end(audit_buf);
2172 }
2173 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2174
2175 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2176 {
2177         struct audit_buffer *audit_buf;
2178
2179         audit_buf = xfrm_audit_start("SA-notfound");
2180         if (audit_buf == NULL)
2181                 return;
2182         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2183         audit_log_end(audit_buf);
2184 }
2185 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2186
2187 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2188                                __be32 net_spi, __be32 net_seq)
2189 {
2190         struct audit_buffer *audit_buf;
2191         u32 spi;
2192
2193         audit_buf = xfrm_audit_start("SA-notfound");
2194         if (audit_buf == NULL)
2195                 return;
2196         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2197         spi = ntohl(net_spi);
2198         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2199                          spi, spi, ntohl(net_seq));
2200         audit_log_end(audit_buf);
2201 }
2202 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2203
2204 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2205                               struct sk_buff *skb, u8 proto)
2206 {
2207         struct audit_buffer *audit_buf;
2208         __be32 net_spi;
2209         __be32 net_seq;
2210
2211         audit_buf = xfrm_audit_start("SA-icv-failure");
2212         if (audit_buf == NULL)
2213                 return;
2214         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2215         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2216                 u32 spi = ntohl(net_spi);
2217                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2218                                  spi, spi, ntohl(net_seq));
2219         }
2220         audit_log_end(audit_buf);
2221 }
2222 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2223 #endif /* CONFIG_AUDITSYSCALL */