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_RWLOCK(xfrm_state_afinfo_lock);
 162 static struct xfrm_state_afinfo *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 pid);
 170
 171 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 172 {
 173         struct xfrm_state_afinfo *afinfo;
 174         if (unlikely(family >= NPROTO))
 175                 return NULL;
 176         write_lock_bh(&xfrm_state_afinfo_lock);
 177         afinfo = xfrm_state_afinfo[family];
 178         if (unlikely(!afinfo))
 179                 write_unlock_bh(&xfrm_state_afinfo_lock);
 180         return afinfo;
 181 }
 182
 183 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 184         __releases(xfrm_state_afinfo_lock)
 185 {
 186         write_unlock_bh(&xfrm_state_afinfo_lock);
 187 }
 188
 189 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 190 {
 191         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 192         const struct xfrm_type **typemap;
 193         int err = 0;
 194
 195         if (unlikely(afinfo == NULL))
 196                 return -EAFNOSUPPORT;
 197         typemap = afinfo->type_map;
 198
 199         if (likely(typemap[type->proto] == NULL))
 200                 typemap[type->proto] = type;
 201         else
 202                 err = -EEXIST;
 203         xfrm_state_unlock_afinfo(afinfo);
 204         return err;
 205 }
 206 EXPORT_SYMBOL(xfrm_register_type);
 207
 208 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 209 {
 210         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 211         const struct xfrm_type **typemap;
 212         int err = 0;
 213
 214         if (unlikely(afinfo == NULL))
 215                 return -EAFNOSUPPORT;
 216         typemap = afinfo->type_map;
 217
 218         if (unlikely(typemap[type->proto] != type))
 219                 err = -ENOENT;
 220         else
 221                 typemap[type->proto] = NULL;
 222         xfrm_state_unlock_afinfo(afinfo);
 223         return err;
 224 }
 225 EXPORT_SYMBOL(xfrm_unregister_type);
 226
 227 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 228 {
 229         struct xfrm_state_afinfo *afinfo;
 230         const struct xfrm_type **typemap;
 231         const struct xfrm_type *type;
 232         int modload_attempted = 0;
 233
 234 retry:
 235         afinfo = xfrm_state_get_afinfo(family);
 236         if (unlikely(afinfo == NULL))
 237                 return NULL;
 238         typemap = afinfo->type_map;
 239
 240         type = typemap[proto];
 241         if (unlikely(type && !try_module_get(type->owner)))
 242                 type = NULL;
 243         if (!type && !modload_attempted) {
 244                 xfrm_state_put_afinfo(afinfo);
 245                 request_module("xfrm-type-%d-%d", family, proto);
 246                 modload_attempted = 1;
 247                 goto retry;
 248         }
 249
 250         xfrm_state_put_afinfo(afinfo);
 251         return type;
 252 }
 253
 254 static void xfrm_put_type(const struct xfrm_type *type)
 255 {
 256         module_put(type->owner);
 257 }
 258
 259 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 260 {
 261         struct xfrm_state_afinfo *afinfo;
 262         struct xfrm_mode **modemap;
 263         int err;
 264
 265         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 266                 return -EINVAL;
 267
 268         afinfo = xfrm_state_lock_afinfo(family);
 269         if (unlikely(afinfo == NULL))
 270                 return -EAFNOSUPPORT;
 271
 272         err = -EEXIST;
 273         modemap = afinfo->mode_map;
 274         if (modemap[mode->encap])
 275                 goto out;
 276
 277         err = -ENOENT;
 278         if (!try_module_get(afinfo->owner))
 279                 goto out;
 280
 281         mode->afinfo = afinfo;
 282         modemap[mode->encap] = mode;
 283         err = 0;
 284
 285 out:
 286         xfrm_state_unlock_afinfo(afinfo);
 287         return err;
 288 }
 289 EXPORT_SYMBOL(xfrm_register_mode);
 290
 291 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 292 {
 293         struct xfrm_state_afinfo *afinfo;
 294         struct xfrm_mode **modemap;
 295         int err;
 296
 297         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 298                 return -EINVAL;
 299
 300         afinfo = xfrm_state_lock_afinfo(family);
 301         if (unlikely(afinfo == NULL))
 302                 return -EAFNOSUPPORT;
 303
 304         err = -ENOENT;
 305         modemap = afinfo->mode_map;
 306         if (likely(modemap[mode->encap] == mode)) {
 307                 modemap[mode->encap] = NULL;
 308                 module_put(mode->afinfo->owner);
 309                 err = 0;
 310         }
 311
 312         xfrm_state_unlock_afinfo(afinfo);
 313         return err;
 314 }
 315 EXPORT_SYMBOL(xfrm_unregister_mode);
 316
 317 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 318 {
 319         struct xfrm_state_afinfo *afinfo;
 320         struct xfrm_mode *mode;
 321         int modload_attempted = 0;
 322
 323         if (unlikely(encap >= XFRM_MODE_MAX))
 324                 return NULL;
 325
 326 retry:
 327         afinfo = xfrm_state_get_afinfo(family);
 328         if (unlikely(afinfo == NULL))
 329                 return NULL;
 330
 331         mode = afinfo->mode_map[encap];
 332         if (unlikely(mode && !try_module_get(mode->owner)))
 333                 mode = NULL;
 334         if (!mode && !modload_attempted) {
 335                 xfrm_state_put_afinfo(afinfo);
 336                 request_module("xfrm-mode-%d-%d", family, encap);
 337                 modload_attempted = 1;
 338                 goto retry;
 339         }
 340
 341         xfrm_state_put_afinfo(afinfo);
 342         return mode;
 343 }
 344
 345 static void xfrm_put_mode(struct xfrm_mode *mode)
 346 {
 347         module_put(mode->owner);
 348 }
 349
 350 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 351 {
 352         tasklet_hrtimer_cancel(&x->mtimer);
 353         del_timer_sync(&x->rtimer);
 354         kfree(x->aead);
 355         kfree(x->aalg);
 356         kfree(x->ealg);
 357         kfree(x->calg);
 358         kfree(x->encap);
 359         kfree(x->coaddr);
 360         kfree(x->replay_esn);
 361         kfree(x->preplay_esn);
 362         if (x->inner_mode)
 363                 xfrm_put_mode(x->inner_mode);
 364         if (x->inner_mode_iaf)
 365                 xfrm_put_mode(x->inner_mode_iaf);
 366         if (x->outer_mode)
 367                 xfrm_put_mode(x->outer_mode);
 368         if (x->type) {
 369                 x->type->destructor(x);
 370                 xfrm_put_type(x->type);
 371         }
 372         security_xfrm_state_free(x);
 373         kfree(x);
 374 }
 375
 376 static void xfrm_state_gc_task(struct work_struct *work)
 377 {
 378         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 379         struct xfrm_state *x;
 380         struct hlist_node *entry, *tmp;
 381         struct hlist_head gc_list;
 382
 383         spin_lock_bh(&xfrm_state_gc_lock);
 384         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 385         spin_unlock_bh(&xfrm_state_gc_lock);
 386
 387         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 388                 xfrm_state_gc_destroy(x);
 389
 390         wake_up(&net->xfrm.km_waitq);
 391 }
 392
 393 static inline unsigned long make_jiffies(long secs)
 394 {
 395         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 396                 return MAX_SCHEDULE_TIMEOUT-1;
 397         else
 398                 return secs*HZ;
 399 }
 400
 401 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 402 {
 403         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 404         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 405         struct net *net = xs_net(x);
 406         unsigned long now = get_seconds();
 407         long next = LONG_MAX;
 408         int warn = 0;
 409         int err = 0;
 410
 411         spin_lock(&x->lock);
 412         if (x->km.state == XFRM_STATE_DEAD)
 413                 goto out;
 414         if (x->km.state == XFRM_STATE_EXPIRED)
 415                 goto expired;
 416         if (x->lft.hard_add_expires_seconds) {
 417                 long tmo = x->lft.hard_add_expires_seconds +
 418                         x->curlft.add_time - now;
 419                 if (tmo <= 0)
 420                         goto expired;
 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                 else if (tmo < next)
 440                         next = tmo;
 441         }
 442         if (x->lft.soft_use_expires_seconds) {
 443                 long tmo = x->lft.soft_use_expires_seconds +
 444                         (x->curlft.use_time ? : now) - now;
 445                 if (tmo <= 0)
 446                         warn = 1;
 447                 else if (tmo < next)
 448                         next = tmo;
 449         }
 450
 451         x->km.dying = warn;
 452         if (warn)
 453                 km_state_expired(x, 0, 0);
 454 resched:
 455         if (next != LONG_MAX){
 456                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 457         }
 458
 459         goto out;
 460
 461 expired:
 462         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 463                 x->km.state = XFRM_STATE_EXPIRED;
 464                 wake_up(&net->xfrm.km_waitq);
 465                 next = 2;
 466                 goto resched;
 467         }
 468
 469         err = __xfrm_state_delete(x);
 470         if (!err && x->id.spi)
 471                 km_state_expired(x, 1, 0);
 472
 473         xfrm_audit_state_delete(x, err ? 0 : 1,
 474                                 audit_get_loginuid(current),
 475                                 audit_get_sessionid(current), 0);
 476
 477 out:
 478         spin_unlock(&x->lock);
 479         return HRTIMER_NORESTART;
 480 }
 481
 482 static void xfrm_replay_timer_handler(unsigned long data);
 483
 484 struct xfrm_state *xfrm_state_alloc(struct net *net)
 485 {
 486         struct xfrm_state *x;
 487
 488         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 489
 490         if (x) {
 491                 write_pnet(&x->xs_net, net);
 492                 atomic_set(&x->refcnt, 1);
 493                 atomic_set(&x->tunnel_users, 0);
 494                 INIT_LIST_HEAD(&x->km.all);
 495                 INIT_HLIST_NODE(&x->bydst);
 496                 INIT_HLIST_NODE(&x->bysrc);
 497                 INIT_HLIST_NODE(&x->byspi);
 498                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 499                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 500                                 (unsigned long)x);
 501                 x->curlft.add_time = get_seconds();
 502                 x->lft.soft_byte_limit = XFRM_INF;
 503                 x->lft.soft_packet_limit = XFRM_INF;
 504                 x->lft.hard_byte_limit = XFRM_INF;
 505                 x->lft.hard_packet_limit = XFRM_INF;
 506                 x->replay_maxage = 0;
 507                 x->replay_maxdiff = 0;
 508                 x->inner_mode = NULL;
 509                 x->inner_mode_iaf = NULL;
 510                 spin_lock_init(&x->lock);
 511         }
 512         return x;
 513 }
 514 EXPORT_SYMBOL(xfrm_state_alloc);
 515
 516 void __xfrm_state_destroy(struct xfrm_state *x)
 517 {
 518         struct net *net = xs_net(x);
 519
 520         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 521
 522         spin_lock_bh(&xfrm_state_gc_lock);
 523         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 524         spin_unlock_bh(&xfrm_state_gc_lock);
 525         schedule_work(&net->xfrm.state_gc_work);
 526 }
 527 EXPORT_SYMBOL(__xfrm_state_destroy);
 528
 529 int __xfrm_state_delete(struct xfrm_state *x)
 530 {
 531         struct net *net = xs_net(x);
 532         int err = -ESRCH;
 533
 534         if (x->km.state != XFRM_STATE_DEAD) {
 535                 x->km.state = XFRM_STATE_DEAD;
 536                 spin_lock(&xfrm_state_lock);
 537                 list_del(&x->km.all);
 538                 hlist_del(&x->bydst);
 539                 hlist_del(&x->bysrc);
 540                 if (x->id.spi)
 541                         hlist_del(&x->byspi);
 542                 net->xfrm.state_num--;
 543                 spin_unlock(&xfrm_state_lock);
 544
 545                 /* All xfrm_state objects are created by xfrm_state_alloc.
 546                  * The xfrm_state_alloc call gives a reference, and that
 547                  * is what we are dropping here.
 548                  */
 549                 xfrm_state_put(x);
 550                 err = 0;
 551         }
 552
 553         return err;
 554 }
 555 EXPORT_SYMBOL(__xfrm_state_delete);
 556
 557 int xfrm_state_delete(struct xfrm_state *x)
 558 {
 559         int err;
 560
 561         spin_lock_bh(&x->lock);
 562         err = __xfrm_state_delete(x);
 563         spin_unlock_bh(&x->lock);
 564
 565         return err;
 566 }
 567 EXPORT_SYMBOL(xfrm_state_delete);
 568
 569 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 570 static inline int
 571 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 572 {
 573         int i, err = 0;
 574
 575         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 576                 struct hlist_node *entry;
 577                 struct xfrm_state *x;
 578
 579                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 580                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 581                            (err = security_xfrm_state_delete(x)) != 0) {
 582                                 xfrm_audit_state_delete(x, 0,
 583                                                         audit_info->loginuid,
 584                                                         audit_info->sessionid,
 585                                                         audit_info->secid);
 586                                 return err;
 587                         }
 588                 }
 589         }
 590
 591         return err;
 592 }
 593 #else
 594 static inline int
 595 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 596 {
 597         return 0;
 598 }
 599 #endif
 600
 601 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 602 {
 603         int i, err = 0, cnt = 0;
 604
 605         spin_lock_bh(&xfrm_state_lock);
 606         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 607         if (err)
 608                 goto out;
 609
 610         err = -ESRCH;
 611         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 612                 struct hlist_node *entry;
 613                 struct xfrm_state *x;
 614 restart:
 615                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 616                         if (!xfrm_state_kern(x) &&
 617                             xfrm_id_proto_match(x->id.proto, proto)) {
 618                                 xfrm_state_hold(x);
 619                                 spin_unlock_bh(&xfrm_state_lock);
 620
 621                                 err = xfrm_state_delete(x);
 622                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 623                                                         audit_info->loginuid,
 624                                                         audit_info->sessionid,
 625                                                         audit_info->secid);
 626                                 xfrm_state_put(x);
 627                                 if (!err)
 628                                         cnt++;
 629
 630                                 spin_lock_bh(&xfrm_state_lock);
 631                                 goto restart;
 632                         }
 633                 }
 634         }
 635         if (cnt)
 636                 err = 0;
 637
 638 out:
 639         spin_unlock_bh(&xfrm_state_lock);
 640         wake_up(&net->xfrm.km_waitq);
 641         return err;
 642 }
 643 EXPORT_SYMBOL(xfrm_state_flush);
 644
 645 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 646 {
 647         spin_lock_bh(&xfrm_state_lock);
 648         si->sadcnt = net->xfrm.state_num;
 649         si->sadhcnt = net->xfrm.state_hmask;
 650         si->sadhmcnt = xfrm_state_hashmax;
 651         spin_unlock_bh(&xfrm_state_lock);
 652 }
 653 EXPORT_SYMBOL(xfrm_sad_getinfo);
 654
 655 static int
 656 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 657                     const struct xfrm_tmpl *tmpl,
 658                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 659                     unsigned short family)
 660 {
 661         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 662         if (!afinfo)
 663                 return -1;
 664         afinfo->init_tempsel(&x->sel, fl);
 665
 666         if (family != tmpl->encap_family) {
 667                 xfrm_state_put_afinfo(afinfo);
 668                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 669                 if (!afinfo)
 670                         return -1;
 671         }
 672         afinfo->init_temprop(x, tmpl, daddr, saddr);
 673         xfrm_state_put_afinfo(afinfo);
 674         return 0;
 675 }
 676
 677 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 678                                               const xfrm_address_t *daddr,
 679                                               __be32 spi, u8 proto,
 680                                               unsigned short family)
 681 {
 682         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 683         struct xfrm_state *x;
 684         struct hlist_node *entry;
 685
 686         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 687                 if (x->props.family != family ||
 688                     x->id.spi       != spi ||
 689                     x->id.proto     != proto ||
 690                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 691                         continue;
 692
 693                 if ((mark & x->mark.m) != x->mark.v)
 694                         continue;
 695                 xfrm_state_hold(x);
 696                 return x;
 697         }
 698
 699         return NULL;
 700 }
 701
 702 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 703                                                      const xfrm_address_t *daddr,
 704                                                      const xfrm_address_t *saddr,
 705                                                      u8 proto, unsigned short family)
 706 {
 707         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 708         struct xfrm_state *x;
 709         struct hlist_node *entry;
 710
 711         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 712                 if (x->props.family != family ||
 713                     x->id.proto     != proto ||
 714                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 715                     xfrm_addr_cmp(&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 hlist_node *entry;
 798         struct xfrm_state *x, *x0, *to_put;
 799         int acquire_in_progress = 0;
 800         int error = 0;
 801         struct xfrm_state *best = NULL;
 802         u32 mark = pol->mark.v & pol->mark.m;
 803         unsigned short encap_family = tmpl->encap_family;
 804
 805         to_put = NULL;
 806
 807         spin_lock_bh(&xfrm_state_lock);
 808         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 809         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 810                 if (x->props.family == encap_family &&
 811                     x->props.reqid == tmpl->reqid &&
 812                     (mark & x->mark.m) == x->mark.v &&
 813                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 814                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 815                     tmpl->mode == x->props.mode &&
 816                     tmpl->id.proto == x->id.proto &&
 817                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 818                         xfrm_state_look_at(pol, x, fl, encap_family,
 819                                            &best, &acquire_in_progress, &error);
 820         }
 821         if (best)
 822                 goto found;
 823
 824         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 825         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 826                 if (x->props.family == encap_family &&
 827                     x->props.reqid == tmpl->reqid &&
 828                     (mark & x->mark.m) == x->mark.v &&
 829                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 830                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 831                     tmpl->mode == x->props.mode &&
 832                     tmpl->id.proto == x->id.proto &&
 833                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 834                         xfrm_state_look_at(pol, x, fl, encap_family,
 835                                            &best, &acquire_in_progress, &error);
 836         }
 837
 838 found:
 839         x = best;
 840         if (!x && !error && !acquire_in_progress) {
 841                 if (tmpl->id.spi &&
 842                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 843                                               tmpl->id.proto, encap_family)) != NULL) {
 844                         to_put = x0;
 845                         error = -EEXIST;
 846                         goto out;
 847                 }
 848                 x = xfrm_state_alloc(net);
 849                 if (x == NULL) {
 850                         error = -ENOMEM;
 851                         goto out;
 852                 }
 853                 /* Initialize temporary state matching only
 854                  * to current session. */
 855                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 856                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 857
 858                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 859                 if (error) {
 860                         x->km.state = XFRM_STATE_DEAD;
 861                         to_put = x;
 862                         x = NULL;
 863                         goto out;
 864                 }
 865
 866                 if (km_query(x, tmpl, pol) == 0) {
 867                         x->km.state = XFRM_STATE_ACQ;
 868                         list_add(&x->km.all, &net->xfrm.state_all);
 869                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 870                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 871                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 872                         if (x->id.spi) {
 873                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 874                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 875                         }
 876                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 877                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 878                         net->xfrm.state_num++;
 879                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 880                 } else {
 881                         x->km.state = XFRM_STATE_DEAD;
 882                         to_put = x;
 883                         x = NULL;
 884                         error = -ESRCH;
 885                 }
 886         }
 887 out:
 888         if (x)
 889                 xfrm_state_hold(x);
 890         else
 891                 *err = acquire_in_progress ? -EAGAIN : error;
 892         spin_unlock_bh(&xfrm_state_lock);
 893         if (to_put)
 894                 xfrm_state_put(to_put);
 895         return x;
 896 }
 897
 898 struct xfrm_state *
 899 xfrm_stateonly_find(struct net *net, u32 mark,
 900                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 901                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 902 {
 903         unsigned int h;
 904         struct xfrm_state *rx = NULL, *x = NULL;
 905         struct hlist_node *entry;
 906
 907         spin_lock(&xfrm_state_lock);
 908         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 909         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 910                 if (x->props.family == family &&
 911                     x->props.reqid == reqid &&
 912                     (mark & x->mark.m) == x->mark.v &&
 913                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 914                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 915                     mode == x->props.mode &&
 916                     proto == x->id.proto &&
 917                     x->km.state == XFRM_STATE_VALID) {
 918                         rx = x;
 919                         break;
 920                 }
 921         }
 922
 923         if (rx)
 924                 xfrm_state_hold(rx);
 925         spin_unlock(&xfrm_state_lock);
 926
 927
 928         return rx;
 929 }
 930 EXPORT_SYMBOL(xfrm_stateonly_find);
 931
 932 static void __xfrm_state_insert(struct xfrm_state *x)
 933 {
 934         struct net *net = xs_net(x);
 935         unsigned int h;
 936
 937         list_add(&x->km.all, &net->xfrm.state_all);
 938
 939         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 940                           x->props.reqid, x->props.family);
 941         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 942
 943         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 944         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 945
 946         if (x->id.spi) {
 947                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 948                                   x->props.family);
 949
 950                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 951         }
 952
 953         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 954         if (x->replay_maxage)
 955                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 956
 957         wake_up(&net->xfrm.km_waitq);
 958
 959         net->xfrm.state_num++;
 960
 961         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 962 }
 963
 964 /* xfrm_state_lock is held */
 965 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 966 {
 967         struct net *net = xs_net(xnew);
 968         unsigned short family = xnew->props.family;
 969         u32 reqid = xnew->props.reqid;
 970         struct xfrm_state *x;
 971         struct hlist_node *entry;
 972         unsigned int h;
 973         u32 mark = xnew->mark.v & xnew->mark.m;
 974
 975         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 976         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 977                 if (x->props.family     == family &&
 978                     x->props.reqid      == reqid &&
 979                     (mark & x->mark.m) == x->mark.v &&
 980                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 981                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 982                         x->genid++;
 983         }
 984 }
 985
 986 void xfrm_state_insert(struct xfrm_state *x)
 987 {
 988         spin_lock_bh(&xfrm_state_lock);
 989         __xfrm_state_bump_genids(x);
 990         __xfrm_state_insert(x);
 991         spin_unlock_bh(&xfrm_state_lock);
 992 }
 993 EXPORT_SYMBOL(xfrm_state_insert);
 994
 995 /* xfrm_state_lock is held */
 996 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
 997                                           unsigned short family, u8 mode,
 998                                           u32 reqid, u8 proto,
 999                                           const xfrm_address_t *daddr,
1000                                           const xfrm_address_t *saddr, int create)
1001 {
1002         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1003         struct hlist_node *entry;
1004         struct xfrm_state *x;
1005         u32 mark = m->v & m->m;
1006
1007         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1008                 if (x->props.reqid  != reqid ||
1009                     x->props.mode   != mode ||
1010                     x->props.family != family ||
1011                     x->km.state     != XFRM_STATE_ACQ ||
1012                     x->id.spi       != 0 ||
1013                     x->id.proto     != proto ||
1014                     (mark & x->mark.m) != x->mark.v ||
1015                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1016                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1017                         continue;
1018
1019                 xfrm_state_hold(x);
1020                 return x;
1021         }
1022
1023         if (!create)
1024                 return NULL;
1025
1026         x = xfrm_state_alloc(net);
1027         if (likely(x)) {
1028                 switch (family) {
1029                 case AF_INET:
1030                         x->sel.daddr.a4 = daddr->a4;
1031                         x->sel.saddr.a4 = saddr->a4;
1032                         x->sel.prefixlen_d = 32;
1033                         x->sel.prefixlen_s = 32;
1034                         x->props.saddr.a4 = saddr->a4;
1035                         x->id.daddr.a4 = daddr->a4;
1036                         break;
1037
1038                 case AF_INET6:
1039                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1040                                        (const struct in6_addr *)daddr);
1041                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1042                                        (const struct in6_addr *)saddr);
1043                         x->sel.prefixlen_d = 128;
1044                         x->sel.prefixlen_s = 128;
1045                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1046                                        (const struct in6_addr *)saddr);
1047                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1048                                        (const 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->km.state != XFRM_STATE_VALID)
1366                 return -EINVAL;
1367
1368         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1369             x->curlft.packets >= x->lft.hard_packet_limit) {
1370                 x->km.state = XFRM_STATE_EXPIRED;
1371                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1372                 return -EINVAL;
1373         }
1374
1375         if (!x->km.dying &&
1376             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1377              x->curlft.packets >= x->lft.soft_packet_limit)) {
1378                 x->km.dying = 1;
1379                 km_state_expired(x, 0, 0);
1380         }
1381         return 0;
1382 }
1383 EXPORT_SYMBOL(xfrm_state_check_expire);
1384
1385 struct xfrm_state *
1386 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1387                   u8 proto, unsigned short family)
1388 {
1389         struct xfrm_state *x;
1390
1391         spin_lock_bh(&xfrm_state_lock);
1392         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1393         spin_unlock_bh(&xfrm_state_lock);
1394         return x;
1395 }
1396 EXPORT_SYMBOL(xfrm_state_lookup);
1397
1398 struct xfrm_state *
1399 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1400                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1401                          u8 proto, unsigned short family)
1402 {
1403         struct xfrm_state *x;
1404
1405         spin_lock_bh(&xfrm_state_lock);
1406         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1407         spin_unlock_bh(&xfrm_state_lock);
1408         return x;
1409 }
1410 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1411
1412 struct xfrm_state *
1413 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1414               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1415               int create, unsigned short family)
1416 {
1417         struct xfrm_state *x;
1418
1419         spin_lock_bh(&xfrm_state_lock);
1420         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1421         spin_unlock_bh(&xfrm_state_lock);
1422
1423         return x;
1424 }
1425 EXPORT_SYMBOL(xfrm_find_acq);
1426
1427 #ifdef CONFIG_XFRM_SUB_POLICY
1428 int
1429 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1430                unsigned short family)
1431 {
1432         int err = 0;
1433         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1434         if (!afinfo)
1435                 return -EAFNOSUPPORT;
1436
1437         spin_lock_bh(&xfrm_state_lock);
1438         if (afinfo->tmpl_sort)
1439                 err = afinfo->tmpl_sort(dst, src, n);
1440         spin_unlock_bh(&xfrm_state_lock);
1441         xfrm_state_put_afinfo(afinfo);
1442         return err;
1443 }
1444 EXPORT_SYMBOL(xfrm_tmpl_sort);
1445
1446 int
1447 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1448                 unsigned short family)
1449 {
1450         int err = 0;
1451         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1452         if (!afinfo)
1453                 return -EAFNOSUPPORT;
1454
1455         spin_lock_bh(&xfrm_state_lock);
1456         if (afinfo->state_sort)
1457                 err = afinfo->state_sort(dst, src, n);
1458         spin_unlock_bh(&xfrm_state_lock);
1459         xfrm_state_put_afinfo(afinfo);
1460         return err;
1461 }
1462 EXPORT_SYMBOL(xfrm_state_sort);
1463 #endif
1464
1465 /* Silly enough, but I'm lazy to build resolution list */
1466
1467 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1468 {
1469         int i;
1470
1471         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1472                 struct hlist_node *entry;
1473                 struct xfrm_state *x;
1474
1475                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1476                         if (x->km.seq == seq &&
1477                             (mark & x->mark.m) == x->mark.v &&
1478                             x->km.state == XFRM_STATE_ACQ) {
1479                                 xfrm_state_hold(x);
1480                                 return x;
1481                         }
1482                 }
1483         }
1484         return NULL;
1485 }
1486
1487 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1488 {
1489         struct xfrm_state *x;
1490
1491         spin_lock_bh(&xfrm_state_lock);
1492         x = __xfrm_find_acq_byseq(net, mark, seq);
1493         spin_unlock_bh(&xfrm_state_lock);
1494         return x;
1495 }
1496 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1497
1498 u32 xfrm_get_acqseq(void)
1499 {
1500         u32 res;
1501         static atomic_t acqseq;
1502
1503         do {
1504                 res = atomic_inc_return(&acqseq);
1505         } while (!res);
1506
1507         return res;
1508 }
1509 EXPORT_SYMBOL(xfrm_get_acqseq);
1510
1511 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1512 {
1513         struct net *net = xs_net(x);
1514         unsigned int h;
1515         struct xfrm_state *x0;
1516         int err = -ENOENT;
1517         __be32 minspi = htonl(low);
1518         __be32 maxspi = htonl(high);
1519         u32 mark = x->mark.v & x->mark.m;
1520
1521         spin_lock_bh(&x->lock);
1522         if (x->km.state == XFRM_STATE_DEAD)
1523                 goto unlock;
1524
1525         err = 0;
1526         if (x->id.spi)
1527                 goto unlock;
1528
1529         err = -ENOENT;
1530
1531         if (minspi == maxspi) {
1532                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1533                 if (x0) {
1534                         xfrm_state_put(x0);
1535                         goto unlock;
1536                 }
1537                 x->id.spi = minspi;
1538         } else {
1539                 u32 spi = 0;
1540                 for (h=0; h<high-low+1; h++) {
1541                         spi = low + net_random()%(high-low+1);
1542                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1543                         if (x0 == NULL) {
1544                                 x->id.spi = htonl(spi);
1545                                 break;
1546                         }
1547                         xfrm_state_put(x0);
1548                 }
1549         }
1550         if (x->id.spi) {
1551                 spin_lock_bh(&xfrm_state_lock);
1552                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1553                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1554                 spin_unlock_bh(&xfrm_state_lock);
1555
1556                 err = 0;
1557         }
1558
1559 unlock:
1560         spin_unlock_bh(&x->lock);
1561
1562         return err;
1563 }
1564 EXPORT_SYMBOL(xfrm_alloc_spi);
1565
1566 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1567                     int (*func)(struct xfrm_state *, int, void*),
1568                     void *data)
1569 {
1570         struct xfrm_state *state;
1571         struct xfrm_state_walk *x;
1572         int err = 0;
1573
1574         if (walk->seq != 0 && list_empty(&walk->all))
1575                 return 0;
1576
1577         spin_lock_bh(&xfrm_state_lock);
1578         if (list_empty(&walk->all))
1579                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1580         else
1581                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1582         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1583                 if (x->state == XFRM_STATE_DEAD)
1584                         continue;
1585                 state = container_of(x, struct xfrm_state, km);
1586                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1587                         continue;
1588                 err = func(state, walk->seq, data);
1589                 if (err) {
1590                         list_move_tail(&walk->all, &x->all);
1591                         goto out;
1592                 }
1593                 walk->seq++;
1594         }
1595         if (walk->seq == 0) {
1596                 err = -ENOENT;
1597                 goto out;
1598         }
1599         list_del_init(&walk->all);
1600 out:
1601         spin_unlock_bh(&xfrm_state_lock);
1602         return err;
1603 }
1604 EXPORT_SYMBOL(xfrm_state_walk);
1605
1606 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1607 {
1608         INIT_LIST_HEAD(&walk->all);
1609         walk->proto = proto;
1610         walk->state = XFRM_STATE_DEAD;
1611         walk->seq = 0;
1612 }
1613 EXPORT_SYMBOL(xfrm_state_walk_init);
1614
1615 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1616 {
1617         if (list_empty(&walk->all))
1618                 return;
1619
1620         spin_lock_bh(&xfrm_state_lock);
1621         list_del(&walk->all);
1622         spin_unlock_bh(&xfrm_state_lock);
1623 }
1624 EXPORT_SYMBOL(xfrm_state_walk_done);
1625
1626 static void xfrm_replay_timer_handler(unsigned long data)
1627 {
1628         struct xfrm_state *x = (struct xfrm_state*)data;
1629
1630         spin_lock(&x->lock);
1631
1632         if (x->km.state == XFRM_STATE_VALID) {
1633                 if (xfrm_aevent_is_on(xs_net(x)))
1634                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1635                 else
1636                         x->xflags |= XFRM_TIME_DEFER;
1637         }
1638
1639         spin_unlock(&x->lock);
1640 }
1641
1642 static LIST_HEAD(xfrm_km_list);
1643 static DEFINE_RWLOCK(xfrm_km_lock);
1644
1645 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1646 {
1647         struct xfrm_mgr *km;
1648
1649         read_lock(&xfrm_km_lock);
1650         list_for_each_entry(km, &xfrm_km_list, list)
1651                 if (km->notify_policy)
1652                         km->notify_policy(xp, dir, c);
1653         read_unlock(&xfrm_km_lock);
1654 }
1655
1656 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1657 {
1658         struct xfrm_mgr *km;
1659         read_lock(&xfrm_km_lock);
1660         list_for_each_entry(km, &xfrm_km_list, list)
1661                 if (km->notify)
1662                         km->notify(x, c);
1663         read_unlock(&xfrm_km_lock);
1664 }
1665
1666 EXPORT_SYMBOL(km_policy_notify);
1667 EXPORT_SYMBOL(km_state_notify);
1668
1669 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1670 {
1671         struct net *net = xs_net(x);
1672         struct km_event c;
1673
1674         c.data.hard = hard;
1675         c.pid = pid;
1676         c.event = XFRM_MSG_EXPIRE;
1677         km_state_notify(x, &c);
1678
1679         if (hard)
1680                 wake_up(&net->xfrm.km_waitq);
1681 }
1682
1683 EXPORT_SYMBOL(km_state_expired);
1684 /*
1685  * We send to all registered managers regardless of failure
1686  * We are happy with one success
1687 */
1688 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1689 {
1690         int err = -EINVAL, acqret;
1691         struct xfrm_mgr *km;
1692
1693         read_lock(&xfrm_km_lock);
1694         list_for_each_entry(km, &xfrm_km_list, list) {
1695                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1696                 if (!acqret)
1697                         err = acqret;
1698         }
1699         read_unlock(&xfrm_km_lock);
1700         return err;
1701 }
1702 EXPORT_SYMBOL(km_query);
1703
1704 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1705 {
1706         int err = -EINVAL;
1707         struct xfrm_mgr *km;
1708
1709         read_lock(&xfrm_km_lock);
1710         list_for_each_entry(km, &xfrm_km_list, list) {
1711                 if (km->new_mapping)
1712                         err = km->new_mapping(x, ipaddr, sport);
1713                 if (!err)
1714                         break;
1715         }
1716         read_unlock(&xfrm_km_lock);
1717         return err;
1718 }
1719 EXPORT_SYMBOL(km_new_mapping);
1720
1721 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1722 {
1723         struct net *net = xp_net(pol);
1724         struct km_event c;
1725
1726         c.data.hard = hard;
1727         c.pid = pid;
1728         c.event = XFRM_MSG_POLEXPIRE;
1729         km_policy_notify(pol, dir, &c);
1730
1731         if (hard)
1732                 wake_up(&net->xfrm.km_waitq);
1733 }
1734 EXPORT_SYMBOL(km_policy_expired);
1735
1736 #ifdef CONFIG_XFRM_MIGRATE
1737 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1738                const struct xfrm_migrate *m, int num_migrate,
1739                const struct xfrm_kmaddress *k)
1740 {
1741         int err = -EINVAL;
1742         int ret;
1743         struct xfrm_mgr *km;
1744
1745         read_lock(&xfrm_km_lock);
1746         list_for_each_entry(km, &xfrm_km_list, list) {
1747                 if (km->migrate) {
1748                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1749                         if (!ret)
1750                                 err = ret;
1751                 }
1752         }
1753         read_unlock(&xfrm_km_lock);
1754         return err;
1755 }
1756 EXPORT_SYMBOL(km_migrate);
1757 #endif
1758
1759 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1760 {
1761         int err = -EINVAL;
1762         int ret;
1763         struct xfrm_mgr *km;
1764
1765         read_lock(&xfrm_km_lock);
1766         list_for_each_entry(km, &xfrm_km_list, list) {
1767                 if (km->report) {
1768                         ret = km->report(net, proto, sel, addr);
1769                         if (!ret)
1770                                 err = ret;
1771                 }
1772         }
1773         read_unlock(&xfrm_km_lock);
1774         return err;
1775 }
1776 EXPORT_SYMBOL(km_report);
1777
1778 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1779 {
1780         int err;
1781         u8 *data;
1782         struct xfrm_mgr *km;
1783         struct xfrm_policy *pol = NULL;
1784
1785         if (optlen <= 0 || optlen > PAGE_SIZE)
1786                 return -EMSGSIZE;
1787
1788         data = kmalloc(optlen, GFP_KERNEL);
1789         if (!data)
1790                 return -ENOMEM;
1791
1792         err = -EFAULT;
1793         if (copy_from_user(data, optval, optlen))
1794                 goto out;
1795
1796         err = -EINVAL;
1797         read_lock(&xfrm_km_lock);
1798         list_for_each_entry(km, &xfrm_km_list, list) {
1799                 pol = km->compile_policy(sk, optname, data,
1800                                          optlen, &err);
1801                 if (err >= 0)
1802                         break;
1803         }
1804         read_unlock(&xfrm_km_lock);
1805
1806         if (err >= 0) {
1807                 xfrm_sk_policy_insert(sk, err, pol);
1808                 xfrm_pol_put(pol);
1809                 err = 0;
1810         }
1811
1812 out:
1813         kfree(data);
1814         return err;
1815 }
1816 EXPORT_SYMBOL(xfrm_user_policy);
1817
1818 int xfrm_register_km(struct xfrm_mgr *km)
1819 {
1820         write_lock_bh(&xfrm_km_lock);
1821         list_add_tail(&km->list, &xfrm_km_list);
1822         write_unlock_bh(&xfrm_km_lock);
1823         return 0;
1824 }
1825 EXPORT_SYMBOL(xfrm_register_km);
1826
1827 int xfrm_unregister_km(struct xfrm_mgr *km)
1828 {
1829         write_lock_bh(&xfrm_km_lock);
1830         list_del(&km->list);
1831         write_unlock_bh(&xfrm_km_lock);
1832         return 0;
1833 }
1834 EXPORT_SYMBOL(xfrm_unregister_km);
1835
1836 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1837 {
1838         int err = 0;
1839         if (unlikely(afinfo == NULL))
1840                 return -EINVAL;
1841         if (unlikely(afinfo->family >= NPROTO))
1842                 return -EAFNOSUPPORT;
1843         write_lock_bh(&xfrm_state_afinfo_lock);
1844         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1845                 err = -ENOBUFS;
1846         else
1847                 xfrm_state_afinfo[afinfo->family] = afinfo;
1848         write_unlock_bh(&xfrm_state_afinfo_lock);
1849         return err;
1850 }
1851 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1852
1853 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1854 {
1855         int err = 0;
1856         if (unlikely(afinfo == NULL))
1857                 return -EINVAL;
1858         if (unlikely(afinfo->family >= NPROTO))
1859                 return -EAFNOSUPPORT;
1860         write_lock_bh(&xfrm_state_afinfo_lock);
1861         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1862                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1863                         err = -EINVAL;
1864                 else
1865                         xfrm_state_afinfo[afinfo->family] = NULL;
1866         }
1867         write_unlock_bh(&xfrm_state_afinfo_lock);
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         read_lock(&xfrm_state_afinfo_lock);
1878         afinfo = xfrm_state_afinfo[family];
1879         if (unlikely(!afinfo))
1880                 read_unlock(&xfrm_state_afinfo_lock);
1881         return afinfo;
1882 }
1883
1884 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1885         __releases(xfrm_state_afinfo_lock)
1886 {
1887         read_unlock(&xfrm_state_afinfo_lock);
1888 }
1889
1890 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1891 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1892 {
1893         if (x->tunnel) {
1894                 struct xfrm_state *t = x->tunnel;
1895
1896                 if (atomic_read(&t->tunnel_users) == 2)
1897                         xfrm_state_delete(t);
1898                 atomic_dec(&t->tunnel_users);
1899                 xfrm_state_put(t);
1900                 x->tunnel = NULL;
1901         }
1902 }
1903 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1904
1905 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1906 {
1907         int res;
1908
1909         spin_lock_bh(&x->lock);
1910         if (x->km.state == XFRM_STATE_VALID &&
1911             x->type && x->type->get_mtu)
1912                 res = x->type->get_mtu(x, mtu);
1913         else
1914                 res = mtu - x->props.header_len;
1915         spin_unlock_bh(&x->lock);
1916         return res;
1917 }
1918
1919 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1920 {
1921         struct xfrm_state_afinfo *afinfo;
1922         struct xfrm_mode *inner_mode;
1923         int family = x->props.family;
1924         int err;
1925
1926         err = -EAFNOSUPPORT;
1927         afinfo = xfrm_state_get_afinfo(family);
1928         if (!afinfo)
1929                 goto error;
1930
1931         err = 0;
1932         if (afinfo->init_flags)
1933                 err = afinfo->init_flags(x);
1934
1935         xfrm_state_put_afinfo(afinfo);
1936
1937         if (err)
1938                 goto error;
1939
1940         err = -EPROTONOSUPPORT;
1941
1942         if (x->sel.family != AF_UNSPEC) {
1943                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1944                 if (inner_mode == NULL)
1945                         goto error;
1946
1947                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1948                     family != x->sel.family) {
1949                         xfrm_put_mode(inner_mode);
1950                         goto error;
1951                 }
1952
1953                 x->inner_mode = inner_mode;
1954         } else {
1955                 struct xfrm_mode *inner_mode_iaf;
1956                 int iafamily = AF_INET;
1957
1958                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1959                 if (inner_mode == NULL)
1960                         goto error;
1961
1962                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1963                         xfrm_put_mode(inner_mode);
1964                         goto error;
1965                 }
1966                 x->inner_mode = inner_mode;
1967
1968                 if (x->props.family == AF_INET)
1969                         iafamily = AF_INET6;
1970
1971                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1972                 if (inner_mode_iaf) {
1973                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1974                                 x->inner_mode_iaf = inner_mode_iaf;
1975                         else
1976                                 xfrm_put_mode(inner_mode_iaf);
1977                 }
1978         }
1979
1980         x->type = xfrm_get_type(x->id.proto, family);
1981         if (x->type == NULL)
1982                 goto error;
1983
1984         err = x->type->init_state(x);
1985         if (err)
1986                 goto error;
1987
1988         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1989         if (x->outer_mode == NULL)
1990                 goto error;
1991
1992         if (init_replay) {
1993                 err = xfrm_init_replay(x);
1994                 if (err)
1995                         goto error;
1996         }
1997
1998         x->km.state = XFRM_STATE_VALID;
1999
2000 error:
2001         return err;
2002 }
2003
2004 EXPORT_SYMBOL(__xfrm_init_state);
2005
2006 int xfrm_init_state(struct xfrm_state *x)
2007 {
2008         return __xfrm_init_state(x, true);
2009 }
2010
2011 EXPORT_SYMBOL(xfrm_init_state);
2012
2013 int __net_init xfrm_state_init(struct net *net)
2014 {
2015         unsigned int sz;
2016
2017         INIT_LIST_HEAD(&net->xfrm.state_all);
2018
2019         sz = sizeof(struct hlist_head) * 8;
2020
2021         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2022         if (!net->xfrm.state_bydst)
2023                 goto out_bydst;
2024         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2025         if (!net->xfrm.state_bysrc)
2026                 goto out_bysrc;
2027         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2028         if (!net->xfrm.state_byspi)
2029                 goto out_byspi;
2030         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2031
2032         net->xfrm.state_num = 0;
2033         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2034         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2035         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2036         init_waitqueue_head(&net->xfrm.km_waitq);
2037         return 0;
2038
2039 out_byspi:
2040         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2041 out_bysrc:
2042         xfrm_hash_free(net->xfrm.state_bydst, sz);
2043 out_bydst:
2044         return -ENOMEM;
2045 }
2046
2047 void xfrm_state_fini(struct net *net)
2048 {
2049         struct xfrm_audit audit_info;
2050         unsigned int sz;
2051
2052         flush_work(&net->xfrm.state_hash_work);
2053         audit_info.loginuid = -1;
2054         audit_info.sessionid = -1;
2055         audit_info.secid = 0;
2056         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2057         flush_work(&net->xfrm.state_gc_work);
2058
2059         WARN_ON(!list_empty(&net->xfrm.state_all));
2060
2061         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2062         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2063         xfrm_hash_free(net->xfrm.state_byspi, sz);
2064         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2065         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2066         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2067         xfrm_hash_free(net->xfrm.state_bydst, sz);
2068 }
2069
2070 #ifdef CONFIG_AUDITSYSCALL
2071 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2072                                      struct audit_buffer *audit_buf)
2073 {
2074         struct xfrm_sec_ctx *ctx = x->security;
2075         u32 spi = ntohl(x->id.spi);
2076
2077         if (ctx)
2078                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2079                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2080
2081         switch(x->props.family) {
2082         case AF_INET:
2083                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2084                                  &x->props.saddr.a4, &x->id.daddr.a4);
2085                 break;
2086         case AF_INET6:
2087                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2088                                  x->props.saddr.a6, x->id.daddr.a6);
2089                 break;
2090         }
2091
2092         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2093 }
2094
2095 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2096                                       struct audit_buffer *audit_buf)
2097 {
2098         const struct iphdr *iph4;
2099         const struct ipv6hdr *iph6;
2100
2101         switch (family) {
2102         case AF_INET:
2103                 iph4 = ip_hdr(skb);
2104                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2105                                  &iph4->saddr, &iph4->daddr);
2106                 break;
2107         case AF_INET6:
2108                 iph6 = ipv6_hdr(skb);
2109                 audit_log_format(audit_buf,
2110                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2111                                  &iph6->saddr,&iph6->daddr,
2112                                  iph6->flow_lbl[0] & 0x0f,
2113                                  iph6->flow_lbl[1],
2114                                  iph6->flow_lbl[2]);
2115                 break;
2116         }
2117 }
2118
2119 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2120                           uid_t auid, u32 sessionid, u32 secid)
2121 {
2122         struct audit_buffer *audit_buf;
2123
2124         audit_buf = xfrm_audit_start("SAD-add");
2125         if (audit_buf == NULL)
2126                 return;
2127         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2128         xfrm_audit_helper_sainfo(x, audit_buf);
2129         audit_log_format(audit_buf, " res=%u", result);
2130         audit_log_end(audit_buf);
2131 }
2132 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2133
2134 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2135                              uid_t auid, u32 sessionid, u32 secid)
2136 {
2137         struct audit_buffer *audit_buf;
2138
2139         audit_buf = xfrm_audit_start("SAD-delete");
2140         if (audit_buf == NULL)
2141                 return;
2142         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2143         xfrm_audit_helper_sainfo(x, audit_buf);
2144         audit_log_format(audit_buf, " res=%u", result);
2145         audit_log_end(audit_buf);
2146 }
2147 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2148
2149 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2150                                       struct sk_buff *skb)
2151 {
2152         struct audit_buffer *audit_buf;
2153         u32 spi;
2154
2155         audit_buf = xfrm_audit_start("SA-replay-overflow");
2156         if (audit_buf == NULL)
2157                 return;
2158         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2159         /* don't record the sequence number because it's inherent in this kind
2160          * of audit message */
2161         spi = ntohl(x->id.spi);
2162         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2163         audit_log_end(audit_buf);
2164 }
2165 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2166
2167 void xfrm_audit_state_replay(struct xfrm_state *x,
2168                              struct sk_buff *skb, __be32 net_seq)
2169 {
2170         struct audit_buffer *audit_buf;
2171         u32 spi;
2172
2173         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2174         if (audit_buf == NULL)
2175                 return;
2176         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2177         spi = ntohl(x->id.spi);
2178         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2179                          spi, spi, ntohl(net_seq));
2180         audit_log_end(audit_buf);
2181 }
2182 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2183
2184 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2185 {
2186         struct audit_buffer *audit_buf;
2187
2188         audit_buf = xfrm_audit_start("SA-notfound");
2189         if (audit_buf == NULL)
2190                 return;
2191         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2192         audit_log_end(audit_buf);
2193 }
2194 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2195
2196 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2197                                __be32 net_spi, __be32 net_seq)
2198 {
2199         struct audit_buffer *audit_buf;
2200         u32 spi;
2201
2202         audit_buf = xfrm_audit_start("SA-notfound");
2203         if (audit_buf == NULL)
2204                 return;
2205         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2206         spi = ntohl(net_spi);
2207         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2208                          spi, spi, ntohl(net_seq));
2209         audit_log_end(audit_buf);
2210 }
2211 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2212
2213 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2214                               struct sk_buff *skb, u8 proto)
2215 {
2216         struct audit_buffer *audit_buf;
2217         __be32 net_spi;
2218         __be32 net_seq;
2219
2220         audit_buf = xfrm_audit_start("SA-icv-failure");
2221         if (audit_buf == NULL)
2222                 return;
2223         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2224         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2225                 u32 spi = ntohl(net_spi);
2226                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2227                                  spi, spi, ntohl(net_seq));
2228         }
2229         audit_log_end(audit_buf);
2230 }
2231 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2232 #endif /* CONFIG_AUDITSYSCALL */