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 *entry, *tmp;
  76         struct xfrm_state *x;
  77
  78         hlist_for_each_entry_safe(x, entry, 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 *entry, *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, entry, 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 hlist_node *entry;
 581                 struct xfrm_state *x;
 582
 583                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 584                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 585                            (err = security_xfrm_state_delete(x)) != 0) {
 586                                 xfrm_audit_state_delete(x, 0,
 587                                                         audit_info->loginuid,
 588                                                         audit_info->sessionid,
 589                                                         audit_info->secid);
 590                                 return err;
 591                         }
 592                 }
 593         }
 594
 595         return err;
 596 }
 597 #else
 598 static inline int
 599 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 600 {
 601         return 0;
 602 }
 603 #endif
 604
 605 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 606 {
 607         int i, err = 0, cnt = 0;
 608
 609         spin_lock_bh(&xfrm_state_lock);
 610         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 611         if (err)
 612                 goto out;
 613
 614         err = -ESRCH;
 615         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 616                 struct hlist_node *entry;
 617                 struct xfrm_state *x;
 618 restart:
 619                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 620                         if (!xfrm_state_kern(x) &&
 621                             xfrm_id_proto_match(x->id.proto, proto)) {
 622                                 xfrm_state_hold(x);
 623                                 spin_unlock_bh(&xfrm_state_lock);
 624
 625                                 err = xfrm_state_delete(x);
 626                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 627                                                         audit_info->loginuid,
 628                                                         audit_info->sessionid,
 629                                                         audit_info->secid);
 630                                 xfrm_state_put(x);
 631                                 if (!err)
 632                                         cnt++;
 633
 634                                 spin_lock_bh(&xfrm_state_lock);
 635                                 goto restart;
 636                         }
 637                 }
 638         }
 639         if (cnt)
 640                 err = 0;
 641
 642 out:
 643         spin_unlock_bh(&xfrm_state_lock);
 644         wake_up(&net->xfrm.km_waitq);
 645         return err;
 646 }
 647 EXPORT_SYMBOL(xfrm_state_flush);
 648
 649 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 650 {
 651         spin_lock_bh(&xfrm_state_lock);
 652         si->sadcnt = net->xfrm.state_num;
 653         si->sadhcnt = net->xfrm.state_hmask;
 654         si->sadhmcnt = xfrm_state_hashmax;
 655         spin_unlock_bh(&xfrm_state_lock);
 656 }
 657 EXPORT_SYMBOL(xfrm_sad_getinfo);
 658
 659 static int
 660 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 661                     const struct xfrm_tmpl *tmpl,
 662                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 663                     unsigned short family)
 664 {
 665         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 666         if (!afinfo)
 667                 return -1;
 668         afinfo->init_tempsel(&x->sel, fl);
 669
 670         if (family != tmpl->encap_family) {
 671                 xfrm_state_put_afinfo(afinfo);
 672                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 673                 if (!afinfo)
 674                         return -1;
 675         }
 676         afinfo->init_temprop(x, tmpl, daddr, saddr);
 677         xfrm_state_put_afinfo(afinfo);
 678         return 0;
 679 }
 680
 681 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 682                                               const xfrm_address_t *daddr,
 683                                               __be32 spi, u8 proto,
 684                                               unsigned short family)
 685 {
 686         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 687         struct xfrm_state *x;
 688         struct hlist_node *entry;
 689
 690         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 691                 if (x->props.family != family ||
 692                     x->id.spi       != spi ||
 693                     x->id.proto     != proto ||
 694                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 695                         continue;
 696
 697                 if ((mark & x->mark.m) != x->mark.v)
 698                         continue;
 699                 xfrm_state_hold(x);
 700                 return x;
 701         }
 702
 703         return NULL;
 704 }
 705
 706 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 707                                                      const xfrm_address_t *daddr,
 708                                                      const xfrm_address_t *saddr,
 709                                                      u8 proto, unsigned short family)
 710 {
 711         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 712         struct xfrm_state *x;
 713         struct hlist_node *entry;
 714
 715         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 716                 if (x->props.family != family ||
 717                     x->id.proto     != proto ||
 718                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 719                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 720                         continue;
 721
 722                 if ((mark & x->mark.m) != x->mark.v)
 723                         continue;
 724                 xfrm_state_hold(x);
 725                 return x;
 726         }
 727
 728         return NULL;
 729 }
 730
 731 static inline struct xfrm_state *
 732 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 733 {
 734         struct net *net = xs_net(x);
 735         u32 mark = x->mark.v & x->mark.m;
 736
 737         if (use_spi)
 738                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 739                                            x->id.spi, x->id.proto, family);
 740         else
 741                 return __xfrm_state_lookup_byaddr(net, mark,
 742                                                   &x->id.daddr,
 743                                                   &x->props.saddr,
 744                                                   x->id.proto, family);
 745 }
 746
 747 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 748 {
 749         if (have_hash_collision &&
 750             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 751             net->xfrm.state_num > net->xfrm.state_hmask)
 752                 schedule_work(&net->xfrm.state_hash_work);
 753 }
 754
 755 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 756                                const struct flowi *fl, unsigned short family,
 757                                struct xfrm_state **best, int *acq_in_progress,
 758                                int *error)
 759 {
 760         /* Resolution logic:
 761          * 1. There is a valid state with matching selector. Done.
 762          * 2. Valid state with inappropriate selector. Skip.
 763          *
 764          * Entering area of "sysdeps".
 765          *
 766          * 3. If state is not valid, selector is temporary, it selects
 767          *    only session which triggered previous resolution. Key
 768          *    manager will do something to install a state with proper
 769          *    selector.
 770          */
 771         if (x->km.state == XFRM_STATE_VALID) {
 772                 if ((x->sel.family &&
 773                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 774                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 775                         return;
 776
 777                 if (!*best ||
 778                     (*best)->km.dying > x->km.dying ||
 779                     ((*best)->km.dying == x->km.dying &&
 780                      (*best)->curlft.add_time < x->curlft.add_time))
 781                         *best = x;
 782         } else if (x->km.state == XFRM_STATE_ACQ) {
 783                 *acq_in_progress = 1;
 784         } else if (x->km.state == XFRM_STATE_ERROR ||
 785                    x->km.state == XFRM_STATE_EXPIRED) {
 786                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 787                     security_xfrm_state_pol_flow_match(x, pol, fl))
 788                         *error = -ESRCH;
 789         }
 790 }
 791
 792 struct xfrm_state *
 793 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 794                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 795                 struct xfrm_policy *pol, int *err,
 796                 unsigned short family)
 797 {
 798         static xfrm_address_t saddr_wildcard = { };
 799         struct net *net = xp_net(pol);
 800         unsigned int h, h_wildcard;
 801         struct hlist_node *entry;
 802         struct xfrm_state *x, *x0, *to_put;
 803         int acquire_in_progress = 0;
 804         int error = 0;
 805         struct xfrm_state *best = NULL;
 806         u32 mark = pol->mark.v & pol->mark.m;
 807         unsigned short encap_family = tmpl->encap_family;
 808
 809         to_put = NULL;
 810
 811         spin_lock_bh(&xfrm_state_lock);
 812         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 813         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 814                 if (x->props.family == encap_family &&
 815                     x->props.reqid == tmpl->reqid &&
 816                     (mark & x->mark.m) == x->mark.v &&
 817                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 818                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 819                     tmpl->mode == x->props.mode &&
 820                     tmpl->id.proto == x->id.proto &&
 821                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 822                         xfrm_state_look_at(pol, x, fl, encap_family,
 823                                            &best, &acquire_in_progress, &error);
 824         }
 825         if (best)
 826                 goto found;
 827
 828         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 829         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 830                 if (x->props.family == encap_family &&
 831                     x->props.reqid == tmpl->reqid &&
 832                     (mark & x->mark.m) == x->mark.v &&
 833                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 834                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 835                     tmpl->mode == x->props.mode &&
 836                     tmpl->id.proto == x->id.proto &&
 837                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 838                         xfrm_state_look_at(pol, x, fl, encap_family,
 839                                            &best, &acquire_in_progress, &error);
 840         }
 841
 842 found:
 843         x = best;
 844         if (!x && !error && !acquire_in_progress) {
 845                 if (tmpl->id.spi &&
 846                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 847                                               tmpl->id.proto, encap_family)) != NULL) {
 848                         to_put = x0;
 849                         error = -EEXIST;
 850                         goto out;
 851                 }
 852                 x = xfrm_state_alloc(net);
 853                 if (x == NULL) {
 854                         error = -ENOMEM;
 855                         goto out;
 856                 }
 857                 /* Initialize temporary state matching only
 858                  * to current session. */
 859                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 860                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 861
 862                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 863                 if (error) {
 864                         x->km.state = XFRM_STATE_DEAD;
 865                         to_put = x;
 866                         x = NULL;
 867                         goto out;
 868                 }
 869
 870                 if (km_query(x, tmpl, pol) == 0) {
 871                         x->km.state = XFRM_STATE_ACQ;
 872                         list_add(&x->km.all, &net->xfrm.state_all);
 873                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 874                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 875                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 876                         if (x->id.spi) {
 877                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 878                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 879                         }
 880                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 881                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 882                         net->xfrm.state_num++;
 883                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 884                 } else {
 885                         x->km.state = XFRM_STATE_DEAD;
 886                         to_put = x;
 887                         x = NULL;
 888                         error = -ESRCH;
 889                 }
 890         }
 891 out:
 892         if (x)
 893                 xfrm_state_hold(x);
 894         else
 895                 *err = acquire_in_progress ? -EAGAIN : error;
 896         spin_unlock_bh(&xfrm_state_lock);
 897         if (to_put)
 898                 xfrm_state_put(to_put);
 899         return x;
 900 }
 901
 902 struct xfrm_state *
 903 xfrm_stateonly_find(struct net *net, u32 mark,
 904                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 905                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 906 {
 907         unsigned int h;
 908         struct xfrm_state *rx = NULL, *x = NULL;
 909         struct hlist_node *entry;
 910
 911         spin_lock(&xfrm_state_lock);
 912         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 913         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 914                 if (x->props.family == family &&
 915                     x->props.reqid == reqid &&
 916                     (mark & x->mark.m) == x->mark.v &&
 917                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 918                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 919                     mode == x->props.mode &&
 920                     proto == x->id.proto &&
 921                     x->km.state == XFRM_STATE_VALID) {
 922                         rx = x;
 923                         break;
 924                 }
 925         }
 926
 927         if (rx)
 928                 xfrm_state_hold(rx);
 929         spin_unlock(&xfrm_state_lock);
 930
 931
 932         return rx;
 933 }
 934 EXPORT_SYMBOL(xfrm_stateonly_find);
 935
 936 static void __xfrm_state_insert(struct xfrm_state *x)
 937 {
 938         struct net *net = xs_net(x);
 939         unsigned int h;
 940
 941         list_add(&x->km.all, &net->xfrm.state_all);
 942
 943         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 944                           x->props.reqid, x->props.family);
 945         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 946
 947         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 948         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 949
 950         if (x->id.spi) {
 951                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 952                                   x->props.family);
 953
 954                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 955         }
 956
 957         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 958         if (x->replay_maxage)
 959                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 960
 961         wake_up(&net->xfrm.km_waitq);
 962
 963         net->xfrm.state_num++;
 964
 965         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 966 }
 967
 968 /* xfrm_state_lock is held */
 969 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 970 {
 971         struct net *net = xs_net(xnew);
 972         unsigned short family = xnew->props.family;
 973         u32 reqid = xnew->props.reqid;
 974         struct xfrm_state *x;
 975         struct hlist_node *entry;
 976         unsigned int h;
 977         u32 mark = xnew->mark.v & xnew->mark.m;
 978
 979         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 980         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 981                 if (x->props.family     == family &&
 982                     x->props.reqid      == reqid &&
 983                     (mark & x->mark.m) == x->mark.v &&
 984                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 985                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 986                         x->genid++;
 987         }
 988 }
 989
 990 void xfrm_state_insert(struct xfrm_state *x)
 991 {
 992         spin_lock_bh(&xfrm_state_lock);
 993         __xfrm_state_bump_genids(x);
 994         __xfrm_state_insert(x);
 995         spin_unlock_bh(&xfrm_state_lock);
 996 }
 997 EXPORT_SYMBOL(xfrm_state_insert);
 998
 999 /* xfrm_state_lock is held */
1000 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
1001                                           unsigned short family, u8 mode,
1002                                           u32 reqid, u8 proto,
1003                                           const xfrm_address_t *daddr,
1004                                           const xfrm_address_t *saddr, int create)
1005 {
1006         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1007         struct hlist_node *entry;
1008         struct xfrm_state *x;
1009         u32 mark = m->v & m->m;
1010
1011         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1012                 if (x->props.reqid  != reqid ||
1013                     x->props.mode   != mode ||
1014                     x->props.family != family ||
1015                     x->km.state     != XFRM_STATE_ACQ ||
1016                     x->id.spi       != 0 ||
1017                     x->id.proto     != proto ||
1018                     (mark & x->mark.m) != x->mark.v ||
1019                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1020                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1021                         continue;
1022
1023                 xfrm_state_hold(x);
1024                 return x;
1025         }
1026
1027         if (!create)
1028                 return NULL;
1029
1030         x = xfrm_state_alloc(net);
1031         if (likely(x)) {
1032                 switch (family) {
1033                 case AF_INET:
1034                         x->sel.daddr.a4 = daddr->a4;
1035                         x->sel.saddr.a4 = saddr->a4;
1036                         x->sel.prefixlen_d = 32;
1037                         x->sel.prefixlen_s = 32;
1038                         x->props.saddr.a4 = saddr->a4;
1039                         x->id.daddr.a4 = daddr->a4;
1040                         break;
1041
1042                 case AF_INET6:
1043                         *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1044                         *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1045                         x->sel.prefixlen_d = 128;
1046                         x->sel.prefixlen_s = 128;
1047                         *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1048                         *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1049                         break;
1050                 }
1051
1052                 x->km.state = XFRM_STATE_ACQ;
1053                 x->id.proto = proto;
1054                 x->props.family = family;
1055                 x->props.mode = mode;
1056                 x->props.reqid = reqid;
1057                 x->mark.v = m->v;
1058                 x->mark.m = m->m;
1059                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1060                 xfrm_state_hold(x);
1061                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1062                 list_add(&x->km.all, &net->xfrm.state_all);
1063                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1064                 h = xfrm_src_hash(net, daddr, saddr, family);
1065                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1066
1067                 net->xfrm.state_num++;
1068
1069                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1070         }
1071
1072         return x;
1073 }
1074
1075 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1076
1077 int xfrm_state_add(struct xfrm_state *x)
1078 {
1079         struct net *net = xs_net(x);
1080         struct xfrm_state *x1, *to_put;
1081         int family;
1082         int err;
1083         u32 mark = x->mark.v & x->mark.m;
1084         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1085
1086         family = x->props.family;
1087
1088         to_put = NULL;
1089
1090         spin_lock_bh(&xfrm_state_lock);
1091
1092         x1 = __xfrm_state_locate(x, use_spi, family);
1093         if (x1) {
1094                 to_put = x1;
1095                 x1 = NULL;
1096                 err = -EEXIST;
1097                 goto out;
1098         }
1099
1100         if (use_spi && x->km.seq) {
1101                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1102                 if (x1 && ((x1->id.proto != x->id.proto) ||
1103                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1104                         to_put = x1;
1105                         x1 = NULL;
1106                 }
1107         }
1108
1109         if (use_spi && !x1)
1110                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1111                                      x->props.reqid, x->id.proto,
1112                                      &x->id.daddr, &x->props.saddr, 0);
1113
1114         __xfrm_state_bump_genids(x);
1115         __xfrm_state_insert(x);
1116         err = 0;
1117
1118 out:
1119         spin_unlock_bh(&xfrm_state_lock);
1120
1121         if (x1) {
1122                 xfrm_state_delete(x1);
1123                 xfrm_state_put(x1);
1124         }
1125
1126         if (to_put)
1127                 xfrm_state_put(to_put);
1128
1129         return err;
1130 }
1131 EXPORT_SYMBOL(xfrm_state_add);
1132
1133 #ifdef CONFIG_XFRM_MIGRATE
1134 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1135 {
1136         struct net *net = xs_net(orig);
1137         int err = -ENOMEM;
1138         struct xfrm_state *x = xfrm_state_alloc(net);
1139         if (!x)
1140                 goto out;
1141
1142         memcpy(&x->id, &orig->id, sizeof(x->id));
1143         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1144         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1145         x->props.mode = orig->props.mode;
1146         x->props.replay_window = orig->props.replay_window;
1147         x->props.reqid = orig->props.reqid;
1148         x->props.family = orig->props.family;
1149         x->props.saddr = orig->props.saddr;
1150
1151         if (orig->aalg) {
1152                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1153                 if (!x->aalg)
1154                         goto error;
1155         }
1156         x->props.aalgo = orig->props.aalgo;
1157
1158         if (orig->ealg) {
1159                 x->ealg = xfrm_algo_clone(orig->ealg);
1160                 if (!x->ealg)
1161                         goto error;
1162         }
1163         x->props.ealgo = orig->props.ealgo;
1164
1165         if (orig->calg) {
1166                 x->calg = xfrm_algo_clone(orig->calg);
1167                 if (!x->calg)
1168                         goto error;
1169         }
1170         x->props.calgo = orig->props.calgo;
1171
1172         if (orig->encap) {
1173                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1174                 if (!x->encap)
1175                         goto error;
1176         }
1177
1178         if (orig->coaddr) {
1179                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1180                                     GFP_KERNEL);
1181                 if (!x->coaddr)
1182                         goto error;
1183         }
1184
1185         if (orig->replay_esn) {
1186                 err = xfrm_replay_clone(x, orig);
1187                 if (err)
1188                         goto error;
1189         }
1190
1191         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1192
1193         err = xfrm_init_state(x);
1194         if (err)
1195                 goto error;
1196
1197         x->props.flags = orig->props.flags;
1198
1199         x->curlft.add_time = orig->curlft.add_time;
1200         x->km.state = orig->km.state;
1201         x->km.seq = orig->km.seq;
1202
1203         return x;
1204
1205  error:
1206         xfrm_state_put(x);
1207 out:
1208         if (errp)
1209                 *errp = err;
1210         return NULL;
1211 }
1212
1213 /* xfrm_state_lock is held */
1214 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1215 {
1216         unsigned int h;
1217         struct xfrm_state *x;
1218         struct hlist_node *entry;
1219
1220         if (m->reqid) {
1221                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1222                                   m->reqid, m->old_family);
1223                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1224                         if (x->props.mode != m->mode ||
1225                             x->id.proto != m->proto)
1226                                 continue;
1227                         if (m->reqid && x->props.reqid != m->reqid)
1228                                 continue;
1229                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1230                                           m->old_family) ||
1231                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1232                                           m->old_family))
1233                                 continue;
1234                         xfrm_state_hold(x);
1235                         return x;
1236                 }
1237         } else {
1238                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1239                                   m->old_family);
1240                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1241                         if (x->props.mode != m->mode ||
1242                             x->id.proto != m->proto)
1243                                 continue;
1244                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1245                                           m->old_family) ||
1246                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1247                                           m->old_family))
1248                                 continue;
1249                         xfrm_state_hold(x);
1250                         return x;
1251                 }
1252         }
1253
1254         return NULL;
1255 }
1256 EXPORT_SYMBOL(xfrm_migrate_state_find);
1257
1258 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1259                                        struct xfrm_migrate *m)
1260 {
1261         struct xfrm_state *xc;
1262         int err;
1263
1264         xc = xfrm_state_clone(x, &err);
1265         if (!xc)
1266                 return NULL;
1267
1268         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1269         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1270
1271         /* add state */
1272         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1273                 /* a care is needed when the destination address of the
1274                    state is to be updated as it is a part of triplet */
1275                 xfrm_state_insert(xc);
1276         } else {
1277                 if ((err = xfrm_state_add(xc)) < 0)
1278                         goto error;
1279         }
1280
1281         return xc;
1282 error:
1283         xfrm_state_put(xc);
1284         return NULL;
1285 }
1286 EXPORT_SYMBOL(xfrm_state_migrate);
1287 #endif
1288
1289 int xfrm_state_update(struct xfrm_state *x)
1290 {
1291         struct xfrm_state *x1, *to_put;
1292         int err;
1293         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1294
1295         to_put = NULL;
1296
1297         spin_lock_bh(&xfrm_state_lock);
1298         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1299
1300         err = -ESRCH;
1301         if (!x1)
1302                 goto out;
1303
1304         if (xfrm_state_kern(x1)) {
1305                 to_put = x1;
1306                 err = -EEXIST;
1307                 goto out;
1308         }
1309
1310         if (x1->km.state == XFRM_STATE_ACQ) {
1311                 __xfrm_state_insert(x);
1312                 x = NULL;
1313         }
1314         err = 0;
1315
1316 out:
1317         spin_unlock_bh(&xfrm_state_lock);
1318
1319         if (to_put)
1320                 xfrm_state_put(to_put);
1321
1322         if (err)
1323                 return err;
1324
1325         if (!x) {
1326                 xfrm_state_delete(x1);
1327                 xfrm_state_put(x1);
1328                 return 0;
1329         }
1330
1331         err = -EINVAL;
1332         spin_lock_bh(&x1->lock);
1333         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1334                 if (x->encap && x1->encap)
1335                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1336                 if (x->coaddr && x1->coaddr) {
1337                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1338                 }
1339                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1340                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1341                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1342                 x1->km.dying = 0;
1343
1344                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1345                 if (x1->curlft.use_time)
1346                         xfrm_state_check_expire(x1);
1347
1348                 err = 0;
1349                 x->km.state = XFRM_STATE_DEAD;
1350                 __xfrm_state_put(x);
1351         }
1352         spin_unlock_bh(&x1->lock);
1353
1354         xfrm_state_put(x1);
1355
1356         return err;
1357 }
1358 EXPORT_SYMBOL(xfrm_state_update);
1359
1360 int xfrm_state_check_expire(struct xfrm_state *x)
1361 {
1362         if (!x->curlft.use_time)
1363                 x->curlft.use_time = get_seconds();
1364
1365         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1366             x->curlft.packets >= x->lft.hard_packet_limit) {
1367                 x->km.state = XFRM_STATE_EXPIRED;
1368                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1369                 return -EINVAL;
1370         }
1371
1372         if (!x->km.dying &&
1373             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1374              x->curlft.packets >= x->lft.soft_packet_limit)) {
1375                 x->km.dying = 1;
1376                 km_state_expired(x, 0, 0);
1377         }
1378         return 0;
1379 }
1380 EXPORT_SYMBOL(xfrm_state_check_expire);
1381
1382 struct xfrm_state *
1383 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1384                   u8 proto, unsigned short family)
1385 {
1386         struct xfrm_state *x;
1387
1388         spin_lock_bh(&xfrm_state_lock);
1389         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1390         spin_unlock_bh(&xfrm_state_lock);
1391         return x;
1392 }
1393 EXPORT_SYMBOL(xfrm_state_lookup);
1394
1395 struct xfrm_state *
1396 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1397                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1398                          u8 proto, unsigned short family)
1399 {
1400         struct xfrm_state *x;
1401
1402         spin_lock_bh(&xfrm_state_lock);
1403         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1404         spin_unlock_bh(&xfrm_state_lock);
1405         return x;
1406 }
1407 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1408
1409 struct xfrm_state *
1410 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1411               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1412               int create, unsigned short family)
1413 {
1414         struct xfrm_state *x;
1415
1416         spin_lock_bh(&xfrm_state_lock);
1417         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1418         spin_unlock_bh(&xfrm_state_lock);
1419
1420         return x;
1421 }
1422 EXPORT_SYMBOL(xfrm_find_acq);
1423
1424 #ifdef CONFIG_XFRM_SUB_POLICY
1425 int
1426 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1427                unsigned short family)
1428 {
1429         int err = 0;
1430         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1431         if (!afinfo)
1432                 return -EAFNOSUPPORT;
1433
1434         spin_lock_bh(&xfrm_state_lock);
1435         if (afinfo->tmpl_sort)
1436                 err = afinfo->tmpl_sort(dst, src, n);
1437         spin_unlock_bh(&xfrm_state_lock);
1438         xfrm_state_put_afinfo(afinfo);
1439         return err;
1440 }
1441 EXPORT_SYMBOL(xfrm_tmpl_sort);
1442
1443 int
1444 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1445                 unsigned short family)
1446 {
1447         int err = 0;
1448         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1449         if (!afinfo)
1450                 return -EAFNOSUPPORT;
1451
1452         spin_lock_bh(&xfrm_state_lock);
1453         if (afinfo->state_sort)
1454                 err = afinfo->state_sort(dst, src, n);
1455         spin_unlock_bh(&xfrm_state_lock);
1456         xfrm_state_put_afinfo(afinfo);
1457         return err;
1458 }
1459 EXPORT_SYMBOL(xfrm_state_sort);
1460 #endif
1461
1462 /* Silly enough, but I'm lazy to build resolution list */
1463
1464 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1465 {
1466         int i;
1467
1468         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1469                 struct hlist_node *entry;
1470                 struct xfrm_state *x;
1471
1472                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1473                         if (x->km.seq == seq &&
1474                             (mark & x->mark.m) == x->mark.v &&
1475                             x->km.state == XFRM_STATE_ACQ) {
1476                                 xfrm_state_hold(x);
1477                                 return x;
1478                         }
1479                 }
1480         }
1481         return NULL;
1482 }
1483
1484 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1485 {
1486         struct xfrm_state *x;
1487
1488         spin_lock_bh(&xfrm_state_lock);
1489         x = __xfrm_find_acq_byseq(net, mark, seq);
1490         spin_unlock_bh(&xfrm_state_lock);
1491         return x;
1492 }
1493 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1494
1495 u32 xfrm_get_acqseq(void)
1496 {
1497         u32 res;
1498         static atomic_t acqseq;
1499
1500         do {
1501                 res = atomic_inc_return(&acqseq);
1502         } while (!res);
1503
1504         return res;
1505 }
1506 EXPORT_SYMBOL(xfrm_get_acqseq);
1507
1508 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1509 {
1510         struct net *net = xs_net(x);
1511         unsigned int h;
1512         struct xfrm_state *x0;
1513         int err = -ENOENT;
1514         __be32 minspi = htonl(low);
1515         __be32 maxspi = htonl(high);
1516         u32 mark = x->mark.v & x->mark.m;
1517
1518         spin_lock_bh(&x->lock);
1519         if (x->km.state == XFRM_STATE_DEAD)
1520                 goto unlock;
1521
1522         err = 0;
1523         if (x->id.spi)
1524                 goto unlock;
1525
1526         err = -ENOENT;
1527
1528         if (minspi == maxspi) {
1529                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1530                 if (x0) {
1531                         xfrm_state_put(x0);
1532                         goto unlock;
1533                 }
1534                 x->id.spi = minspi;
1535         } else {
1536                 u32 spi = 0;
1537                 for (h=0; h<high-low+1; h++) {
1538                         spi = low + net_random()%(high-low+1);
1539                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1540                         if (x0 == NULL) {
1541                                 x->id.spi = htonl(spi);
1542                                 break;
1543                         }
1544                         xfrm_state_put(x0);
1545                 }
1546         }
1547         if (x->id.spi) {
1548                 spin_lock_bh(&xfrm_state_lock);
1549                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1550                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1551                 spin_unlock_bh(&xfrm_state_lock);
1552
1553                 err = 0;
1554         }
1555
1556 unlock:
1557         spin_unlock_bh(&x->lock);
1558
1559         return err;
1560 }
1561 EXPORT_SYMBOL(xfrm_alloc_spi);
1562
1563 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1564                     int (*func)(struct xfrm_state *, int, void*),
1565                     void *data)
1566 {
1567         struct xfrm_state *state;
1568         struct xfrm_state_walk *x;
1569         int err = 0;
1570
1571         if (walk->seq != 0 && list_empty(&walk->all))
1572                 return 0;
1573
1574         spin_lock_bh(&xfrm_state_lock);
1575         if (list_empty(&walk->all))
1576                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1577         else
1578                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1579         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1580                 if (x->state == XFRM_STATE_DEAD)
1581                         continue;
1582                 state = container_of(x, struct xfrm_state, km);
1583                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1584                         continue;
1585                 err = func(state, walk->seq, data);
1586                 if (err) {
1587                         list_move_tail(&walk->all, &x->all);
1588                         goto out;
1589                 }
1590                 walk->seq++;
1591         }
1592         if (walk->seq == 0) {
1593                 err = -ENOENT;
1594                 goto out;
1595         }
1596         list_del_init(&walk->all);
1597 out:
1598         spin_unlock_bh(&xfrm_state_lock);
1599         return err;
1600 }
1601 EXPORT_SYMBOL(xfrm_state_walk);
1602
1603 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1604 {
1605         INIT_LIST_HEAD(&walk->all);
1606         walk->proto = proto;
1607         walk->state = XFRM_STATE_DEAD;
1608         walk->seq = 0;
1609 }
1610 EXPORT_SYMBOL(xfrm_state_walk_init);
1611
1612 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1613 {
1614         if (list_empty(&walk->all))
1615                 return;
1616
1617         spin_lock_bh(&xfrm_state_lock);
1618         list_del(&walk->all);
1619         spin_unlock_bh(&xfrm_state_lock);
1620 }
1621 EXPORT_SYMBOL(xfrm_state_walk_done);
1622
1623 static void xfrm_replay_timer_handler(unsigned long data)
1624 {
1625         struct xfrm_state *x = (struct xfrm_state*)data;
1626
1627         spin_lock(&x->lock);
1628
1629         if (x->km.state == XFRM_STATE_VALID) {
1630                 if (xfrm_aevent_is_on(xs_net(x)))
1631                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1632                 else
1633                         x->xflags |= XFRM_TIME_DEFER;
1634         }
1635
1636         spin_unlock(&x->lock);
1637 }
1638
1639 static LIST_HEAD(xfrm_km_list);
1640
1641 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1642 {
1643         struct xfrm_mgr *km;
1644
1645         rcu_read_lock();
1646         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1647                 if (km->notify_policy)
1648                         km->notify_policy(xp, dir, c);
1649         rcu_read_unlock();
1650 }
1651
1652 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1653 {
1654         struct xfrm_mgr *km;
1655         rcu_read_lock();
1656         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1657                 if (km->notify)
1658                         km->notify(x, c);
1659         rcu_read_unlock();
1660 }
1661
1662 EXPORT_SYMBOL(km_policy_notify);
1663 EXPORT_SYMBOL(km_state_notify);
1664
1665 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1666 {
1667         struct net *net = xs_net(x);
1668         struct km_event c;
1669
1670         c.data.hard = hard;
1671         c.portid = portid;
1672         c.event = XFRM_MSG_EXPIRE;
1673         km_state_notify(x, &c);
1674
1675         if (hard)
1676                 wake_up(&net->xfrm.km_waitq);
1677 }
1678
1679 EXPORT_SYMBOL(km_state_expired);
1680 /*
1681  * We send to all registered managers regardless of failure
1682  * We are happy with one success
1683 */
1684 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1685 {
1686         int err = -EINVAL, acqret;
1687         struct xfrm_mgr *km;
1688
1689         rcu_read_lock();
1690         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1691                 acqret = km->acquire(x, t, pol);
1692                 if (!acqret)
1693                         err = acqret;
1694         }
1695         rcu_read_unlock();
1696         return err;
1697 }
1698 EXPORT_SYMBOL(km_query);
1699
1700 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1701 {
1702         int err = -EINVAL;
1703         struct xfrm_mgr *km;
1704
1705         rcu_read_lock();
1706         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1707                 if (km->new_mapping)
1708                         err = km->new_mapping(x, ipaddr, sport);
1709                 if (!err)
1710                         break;
1711         }
1712         rcu_read_unlock();
1713         return err;
1714 }
1715 EXPORT_SYMBOL(km_new_mapping);
1716
1717 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1718 {
1719         struct net *net = xp_net(pol);
1720         struct km_event c;
1721
1722         c.data.hard = hard;
1723         c.portid = portid;
1724         c.event = XFRM_MSG_POLEXPIRE;
1725         km_policy_notify(pol, dir, &c);
1726
1727         if (hard)
1728                 wake_up(&net->xfrm.km_waitq);
1729 }
1730 EXPORT_SYMBOL(km_policy_expired);
1731
1732 #ifdef CONFIG_XFRM_MIGRATE
1733 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1734                const struct xfrm_migrate *m, int num_migrate,
1735                const struct xfrm_kmaddress *k)
1736 {
1737         int err = -EINVAL;
1738         int ret;
1739         struct xfrm_mgr *km;
1740
1741         rcu_read_lock();
1742         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1743                 if (km->migrate) {
1744                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1745                         if (!ret)
1746                                 err = ret;
1747                 }
1748         }
1749         rcu_read_unlock();
1750         return err;
1751 }
1752 EXPORT_SYMBOL(km_migrate);
1753 #endif
1754
1755 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1756 {
1757         int err = -EINVAL;
1758         int ret;
1759         struct xfrm_mgr *km;
1760
1761         rcu_read_lock();
1762         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1763                 if (km->report) {
1764                         ret = km->report(net, proto, sel, addr);
1765                         if (!ret)
1766                                 err = ret;
1767                 }
1768         }
1769         rcu_read_unlock();
1770         return err;
1771 }
1772 EXPORT_SYMBOL(km_report);
1773
1774 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1775 {
1776         int err;
1777         u8 *data;
1778         struct xfrm_mgr *km;
1779         struct xfrm_policy *pol = NULL;
1780
1781         if (optlen <= 0 || optlen > PAGE_SIZE)
1782                 return -EMSGSIZE;
1783
1784         data = kmalloc(optlen, GFP_KERNEL);
1785         if (!data)
1786                 return -ENOMEM;
1787
1788         err = -EFAULT;
1789         if (copy_from_user(data, optval, optlen))
1790                 goto out;
1791
1792         err = -EINVAL;
1793         rcu_read_lock();
1794         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1795                 pol = km->compile_policy(sk, optname, data,
1796                                          optlen, &err);
1797                 if (err >= 0)
1798                         break;
1799         }
1800         rcu_read_unlock();
1801
1802         if (err >= 0) {
1803                 xfrm_sk_policy_insert(sk, err, pol);
1804                 xfrm_pol_put(pol);
1805                 err = 0;
1806         }
1807
1808 out:
1809         kfree(data);
1810         return err;
1811 }
1812 EXPORT_SYMBOL(xfrm_user_policy);
1813
1814 static DEFINE_SPINLOCK(xfrm_km_lock);
1815
1816 int xfrm_register_km(struct xfrm_mgr *km)
1817 {
1818         spin_lock_bh(&xfrm_km_lock);
1819         list_add_tail_rcu(&km->list, &xfrm_km_list);
1820         spin_unlock_bh(&xfrm_km_lock);
1821         return 0;
1822 }
1823 EXPORT_SYMBOL(xfrm_register_km);
1824
1825 int xfrm_unregister_km(struct xfrm_mgr *km)
1826 {
1827         spin_lock_bh(&xfrm_km_lock);
1828         list_del_rcu(&km->list);
1829         spin_unlock_bh(&xfrm_km_lock);
1830         synchronize_rcu();
1831         return 0;
1832 }
1833 EXPORT_SYMBOL(xfrm_unregister_km);
1834
1835 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1836 {
1837         int err = 0;
1838         if (unlikely(afinfo == NULL))
1839                 return -EINVAL;
1840         if (unlikely(afinfo->family >= NPROTO))
1841                 return -EAFNOSUPPORT;
1842         spin_lock_bh(&xfrm_state_afinfo_lock);
1843         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1844                 err = -ENOBUFS;
1845         else
1846                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1847         spin_unlock_bh(&xfrm_state_afinfo_lock);
1848         return err;
1849 }
1850 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1851
1852 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1853 {
1854         int err = 0;
1855         if (unlikely(afinfo == NULL))
1856                 return -EINVAL;
1857         if (unlikely(afinfo->family >= NPROTO))
1858                 return -EAFNOSUPPORT;
1859         spin_lock_bh(&xfrm_state_afinfo_lock);
1860         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1861                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1862                         err = -EINVAL;
1863                 else
1864                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1865         }
1866         spin_unlock_bh(&xfrm_state_afinfo_lock);
1867         synchronize_rcu();
1868         return err;
1869 }
1870 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1871
1872 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1873 {
1874         struct xfrm_state_afinfo *afinfo;
1875         if (unlikely(family >= NPROTO))
1876                 return NULL;
1877         rcu_read_lock();
1878         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1879         if (unlikely(!afinfo))
1880                 rcu_read_unlock();
1881         return afinfo;
1882 }
1883
1884 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1885 {
1886         rcu_read_unlock();
1887 }
1888
1889 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1890 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1891 {
1892         if (x->tunnel) {
1893                 struct xfrm_state *t = x->tunnel;
1894
1895                 if (atomic_read(&t->tunnel_users) == 2)
1896                         xfrm_state_delete(t);
1897                 atomic_dec(&t->tunnel_users);
1898                 xfrm_state_put(t);
1899                 x->tunnel = NULL;
1900         }
1901 }
1902 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1903
1904 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1905 {
1906         int res;
1907
1908         spin_lock_bh(&x->lock);
1909         if (x->km.state == XFRM_STATE_VALID &&
1910             x->type && x->type->get_mtu)
1911                 res = x->type->get_mtu(x, mtu);
1912         else
1913                 res = mtu - x->props.header_len;
1914         spin_unlock_bh(&x->lock);
1915         return res;
1916 }
1917
1918 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1919 {
1920         struct xfrm_state_afinfo *afinfo;
1921         struct xfrm_mode *inner_mode;
1922         int family = x->props.family;
1923         int err;
1924
1925         err = -EAFNOSUPPORT;
1926         afinfo = xfrm_state_get_afinfo(family);
1927         if (!afinfo)
1928                 goto error;
1929
1930         err = 0;
1931         if (afinfo->init_flags)
1932                 err = afinfo->init_flags(x);
1933
1934         xfrm_state_put_afinfo(afinfo);
1935
1936         if (err)
1937                 goto error;
1938
1939         err = -EPROTONOSUPPORT;
1940
1941         if (x->sel.family != AF_UNSPEC) {
1942                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1943                 if (inner_mode == NULL)
1944                         goto error;
1945
1946                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1947                     family != x->sel.family) {
1948                         xfrm_put_mode(inner_mode);
1949                         goto error;
1950                 }
1951
1952                 x->inner_mode = inner_mode;
1953         } else {
1954                 struct xfrm_mode *inner_mode_iaf;
1955                 int iafamily = AF_INET;
1956
1957                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1958                 if (inner_mode == NULL)
1959                         goto error;
1960
1961                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1962                         xfrm_put_mode(inner_mode);
1963                         goto error;
1964                 }
1965                 x->inner_mode = inner_mode;
1966
1967                 if (x->props.family == AF_INET)
1968                         iafamily = AF_INET6;
1969
1970                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1971                 if (inner_mode_iaf) {
1972                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1973                                 x->inner_mode_iaf = inner_mode_iaf;
1974                         else
1975                                 xfrm_put_mode(inner_mode_iaf);
1976                 }
1977         }
1978
1979         x->type = xfrm_get_type(x->id.proto, family);
1980         if (x->type == NULL)
1981                 goto error;
1982
1983         err = x->type->init_state(x);
1984         if (err)
1985                 goto error;
1986
1987         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1988         if (x->outer_mode == NULL) {
1989                 err = -EPROTONOSUPPORT;
1990                 goto error;
1991         }
1992
1993         if (init_replay) {
1994                 err = xfrm_init_replay(x);
1995                 if (err)
1996                         goto error;
1997         }
1998
1999         x->km.state = XFRM_STATE_VALID;
2000
2001 error:
2002         return err;
2003 }
2004
2005 EXPORT_SYMBOL(__xfrm_init_state);
2006
2007 int xfrm_init_state(struct xfrm_state *x)
2008 {
2009         return __xfrm_init_state(x, true);
2010 }
2011
2012 EXPORT_SYMBOL(xfrm_init_state);
2013
2014 int __net_init xfrm_state_init(struct net *net)
2015 {
2016         unsigned int sz;
2017
2018         INIT_LIST_HEAD(&net->xfrm.state_all);
2019
2020         sz = sizeof(struct hlist_head) * 8;
2021
2022         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2023         if (!net->xfrm.state_bydst)
2024                 goto out_bydst;
2025         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2026         if (!net->xfrm.state_bysrc)
2027                 goto out_bysrc;
2028         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2029         if (!net->xfrm.state_byspi)
2030                 goto out_byspi;
2031         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2032
2033         net->xfrm.state_num = 0;
2034         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2035         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2036         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2037         init_waitqueue_head(&net->xfrm.km_waitq);
2038         return 0;
2039
2040 out_byspi:
2041         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2042 out_bysrc:
2043         xfrm_hash_free(net->xfrm.state_bydst, sz);
2044 out_bydst:
2045         return -ENOMEM;
2046 }
2047
2048 void xfrm_state_fini(struct net *net)
2049 {
2050         struct xfrm_audit audit_info;
2051         unsigned int sz;
2052
2053         flush_work(&net->xfrm.state_hash_work);
2054         audit_info.loginuid = INVALID_UID;
2055         audit_info.sessionid = -1;
2056         audit_info.secid = 0;
2057         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2058         flush_work(&net->xfrm.state_gc_work);
2059
2060         WARN_ON(!list_empty(&net->xfrm.state_all));
2061
2062         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2063         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2064         xfrm_hash_free(net->xfrm.state_byspi, sz);
2065         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2066         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2067         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2068         xfrm_hash_free(net->xfrm.state_bydst, sz);
2069 }
2070
2071 #ifdef CONFIG_AUDITSYSCALL
2072 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2073                                      struct audit_buffer *audit_buf)
2074 {
2075         struct xfrm_sec_ctx *ctx = x->security;
2076         u32 spi = ntohl(x->id.spi);
2077
2078         if (ctx)
2079                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2080                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2081
2082         switch(x->props.family) {
2083         case AF_INET:
2084                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2085                                  &x->props.saddr.a4, &x->id.daddr.a4);
2086                 break;
2087         case AF_INET6:
2088                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2089                                  x->props.saddr.a6, x->id.daddr.a6);
2090                 break;
2091         }
2092
2093         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2094 }
2095
2096 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2097                                       struct audit_buffer *audit_buf)
2098 {
2099         const struct iphdr *iph4;
2100         const struct ipv6hdr *iph6;
2101
2102         switch (family) {
2103         case AF_INET:
2104                 iph4 = ip_hdr(skb);
2105                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2106                                  &iph4->saddr, &iph4->daddr);
2107                 break;
2108         case AF_INET6:
2109                 iph6 = ipv6_hdr(skb);
2110                 audit_log_format(audit_buf,
2111                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2112                                  &iph6->saddr,&iph6->daddr,
2113                                  iph6->flow_lbl[0] & 0x0f,
2114                                  iph6->flow_lbl[1],
2115                                  iph6->flow_lbl[2]);
2116                 break;
2117         }
2118 }
2119
2120 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2121                           kuid_t auid, u32 sessionid, u32 secid)
2122 {
2123         struct audit_buffer *audit_buf;
2124
2125         audit_buf = xfrm_audit_start("SAD-add");
2126         if (audit_buf == NULL)
2127                 return;
2128         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2129         xfrm_audit_helper_sainfo(x, audit_buf);
2130         audit_log_format(audit_buf, " res=%u", result);
2131         audit_log_end(audit_buf);
2132 }
2133 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2134
2135 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2136                              kuid_t auid, u32 sessionid, u32 secid)
2137 {
2138         struct audit_buffer *audit_buf;
2139
2140         audit_buf = xfrm_audit_start("SAD-delete");
2141         if (audit_buf == NULL)
2142                 return;
2143         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2144         xfrm_audit_helper_sainfo(x, audit_buf);
2145         audit_log_format(audit_buf, " res=%u", result);
2146         audit_log_end(audit_buf);
2147 }
2148 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2149
2150 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2151                                       struct sk_buff *skb)
2152 {
2153         struct audit_buffer *audit_buf;
2154         u32 spi;
2155
2156         audit_buf = xfrm_audit_start("SA-replay-overflow");
2157         if (audit_buf == NULL)
2158                 return;
2159         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2160         /* don't record the sequence number because it's inherent in this kind
2161          * of audit message */
2162         spi = ntohl(x->id.spi);
2163         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2164         audit_log_end(audit_buf);
2165 }
2166 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2167
2168 void xfrm_audit_state_replay(struct xfrm_state *x,
2169                              struct sk_buff *skb, __be32 net_seq)
2170 {
2171         struct audit_buffer *audit_buf;
2172         u32 spi;
2173
2174         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2175         if (audit_buf == NULL)
2176                 return;
2177         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2178         spi = ntohl(x->id.spi);
2179         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2180                          spi, spi, ntohl(net_seq));
2181         audit_log_end(audit_buf);
2182 }
2183 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2184
2185 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2186 {
2187         struct audit_buffer *audit_buf;
2188
2189         audit_buf = xfrm_audit_start("SA-notfound");
2190         if (audit_buf == NULL)
2191                 return;
2192         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2193         audit_log_end(audit_buf);
2194 }
2195 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2196
2197 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2198                                __be32 net_spi, __be32 net_seq)
2199 {
2200         struct audit_buffer *audit_buf;
2201         u32 spi;
2202
2203         audit_buf = xfrm_audit_start("SA-notfound");
2204         if (audit_buf == NULL)
2205                 return;
2206         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2207         spi = ntohl(net_spi);
2208         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2209                          spi, spi, ntohl(net_seq));
2210         audit_log_end(audit_buf);
2211 }
2212 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2213
2214 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2215                               struct sk_buff *skb, u8 proto)
2216 {
2217         struct audit_buffer *audit_buf;
2218         __be32 net_spi;
2219         __be32 net_seq;
2220
2221         audit_buf = xfrm_audit_start("SA-icv-failure");
2222         if (audit_buf == NULL)
2223                 return;
2224         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2225         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2226                 u32 spi = ntohl(net_spi);
2227                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2228                                  spi, spi, ntohl(net_seq));
2229         }
2230         audit_log_end(audit_buf);
2231 }
2232 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2233 #endif /* CONFIG_AUDITSYSCALL */