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 portid);
 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->aalg);
 355         kfree(x->ealg);
 356         kfree(x->calg);
 357         kfree(x->encap);
 358         kfree(x->coaddr);
 359         kfree(x->replay_esn);
 360         kfree(x->preplay_esn);
 361         if (x->inner_mode)
 362                 xfrm_put_mode(x->inner_mode);
 363         if (x->inner_mode_iaf)
 364                 xfrm_put_mode(x->inner_mode_iaf);
 365         if (x->outer_mode)
 366                 xfrm_put_mode(x->outer_mode);
 367         if (x->type) {
 368                 x->type->destructor(x);
 369                 xfrm_put_type(x->type);
 370         }
 371         security_xfrm_state_free(x);
 372         kfree(x);
 373 }
 374
 375 static void xfrm_state_gc_task(struct work_struct *work)
 376 {
 377         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 378         struct xfrm_state *x;
 379         struct hlist_node *entry, *tmp;
 380         struct hlist_head gc_list;
 381
 382         spin_lock_bh(&xfrm_state_gc_lock);
 383         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 384         spin_unlock_bh(&xfrm_state_gc_lock);
 385
 386         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 387                 xfrm_state_gc_destroy(x);
 388
 389         wake_up(&net->xfrm.km_waitq);
 390 }
 391
 392 static inline unsigned long make_jiffies(long secs)
 393 {
 394         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 395                 return MAX_SCHEDULE_TIMEOUT-1;
 396         else
 397                 return secs*HZ;
 398 }
 399
 400 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 401 {
 402         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 403         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 404         struct net *net = xs_net(x);
 405         unsigned long now = get_seconds();
 406         long next = LONG_MAX;
 407         int warn = 0;
 408         int err = 0;
 409
 410         spin_lock(&x->lock);
 411         if (x->km.state == XFRM_STATE_DEAD)
 412                 goto out;
 413         if (x->km.state == XFRM_STATE_EXPIRED)
 414                 goto expired;
 415         if (x->lft.hard_add_expires_seconds) {
 416                 long tmo = x->lft.hard_add_expires_seconds +
 417                         x->curlft.add_time - now;
 418                 if (tmo <= 0) {
 419                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 420                                 /* enter hard expire without soft expire first?!
 421                                  * setting a new date could trigger this.
 422                                  * workarbound: fix x->curflt.add_time by below:
 423                                  */
 424                                 x->curlft.add_time = now - x->saved_tmo - 1;
 425                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 426                         } else
 427                                 goto expired;
 428                 }
 429                 if (tmo < next)
 430                         next = tmo;
 431         }
 432         if (x->lft.hard_use_expires_seconds) {
 433                 long tmo = x->lft.hard_use_expires_seconds +
 434                         (x->curlft.use_time ? : now) - now;
 435                 if (tmo <= 0)
 436                         goto expired;
 437                 if (tmo < next)
 438                         next = tmo;
 439         }
 440         if (x->km.dying)
 441                 goto resched;
 442         if (x->lft.soft_add_expires_seconds) {
 443                 long tmo = x->lft.soft_add_expires_seconds +
 444                         x->curlft.add_time - now;
 445                 if (tmo <= 0) {
 446                         warn = 1;
 447                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 448                 } else if (tmo < next) {
 449                         next = tmo;
 450                         x->xflags |= XFRM_SOFT_EXPIRE;
 451                         x->saved_tmo = tmo;
 452                 }
 453         }
 454         if (x->lft.soft_use_expires_seconds) {
 455                 long tmo = x->lft.soft_use_expires_seconds +
 456                         (x->curlft.use_time ? : now) - now;
 457                 if (tmo <= 0)
 458                         warn = 1;
 459                 else if (tmo < next)
 460                         next = tmo;
 461         }
 462
 463         x->km.dying = warn;
 464         if (warn)
 465                 km_state_expired(x, 0, 0);
 466 resched:
 467         if (next != LONG_MAX){
 468                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 469         }
 470
 471         goto out;
 472
 473 expired:
 474         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 475                 x->km.state = XFRM_STATE_EXPIRED;
 476                 wake_up(&net->xfrm.km_waitq);
 477                 next = 2;
 478                 goto resched;
 479         }
 480
 481         err = __xfrm_state_delete(x);
 482         if (!err && x->id.spi)
 483                 km_state_expired(x, 1, 0);
 484
 485         xfrm_audit_state_delete(x, err ? 0 : 1,
 486                                 audit_get_loginuid(current),
 487                                 audit_get_sessionid(current), 0);
 488
 489 out:
 490         spin_unlock(&x->lock);
 491         return HRTIMER_NORESTART;
 492 }
 493
 494 static void xfrm_replay_timer_handler(unsigned long data);
 495
 496 struct xfrm_state *xfrm_state_alloc(struct net *net)
 497 {
 498         struct xfrm_state *x;
 499
 500         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 501
 502         if (x) {
 503                 write_pnet(&x->xs_net, net);
 504                 atomic_set(&x->refcnt, 1);
 505                 atomic_set(&x->tunnel_users, 0);
 506                 INIT_LIST_HEAD(&x->km.all);
 507                 INIT_HLIST_NODE(&x->bydst);
 508                 INIT_HLIST_NODE(&x->bysrc);
 509                 INIT_HLIST_NODE(&x->byspi);
 510                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 511                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 512                                 (unsigned long)x);
 513                 x->curlft.add_time = get_seconds();
 514                 x->lft.soft_byte_limit = XFRM_INF;
 515                 x->lft.soft_packet_limit = XFRM_INF;
 516                 x->lft.hard_byte_limit = XFRM_INF;
 517                 x->lft.hard_packet_limit = XFRM_INF;
 518                 x->replay_maxage = 0;
 519                 x->replay_maxdiff = 0;
 520                 x->inner_mode = NULL;
 521                 x->inner_mode_iaf = NULL;
 522                 spin_lock_init(&x->lock);
 523         }
 524         return x;
 525 }
 526 EXPORT_SYMBOL(xfrm_state_alloc);
 527
 528 void __xfrm_state_destroy(struct xfrm_state *x)
 529 {
 530         struct net *net = xs_net(x);
 531
 532         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 533
 534         spin_lock_bh(&xfrm_state_gc_lock);
 535         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 536         spin_unlock_bh(&xfrm_state_gc_lock);
 537         schedule_work(&net->xfrm.state_gc_work);
 538 }
 539 EXPORT_SYMBOL(__xfrm_state_destroy);
 540
 541 int __xfrm_state_delete(struct xfrm_state *x)
 542 {
 543         struct net *net = xs_net(x);
 544         int err = -ESRCH;
 545
 546         if (x->km.state != XFRM_STATE_DEAD) {
 547                 x->km.state = XFRM_STATE_DEAD;
 548                 spin_lock(&xfrm_state_lock);
 549                 list_del(&x->km.all);
 550                 hlist_del(&x->bydst);
 551                 hlist_del(&x->bysrc);
 552                 if (x->id.spi)
 553                         hlist_del(&x->byspi);
 554                 net->xfrm.state_num--;
 555                 spin_unlock(&xfrm_state_lock);
 556
 557                 /* All xfrm_state objects are created by xfrm_state_alloc.
 558                  * The xfrm_state_alloc call gives a reference, and that
 559                  * is what we are dropping here.
 560                  */
 561                 xfrm_state_put(x);
 562                 err = 0;
 563         }
 564
 565         return err;
 566 }
 567 EXPORT_SYMBOL(__xfrm_state_delete);
 568
 569 int xfrm_state_delete(struct xfrm_state *x)
 570 {
 571         int err;
 572
 573         spin_lock_bh(&x->lock);
 574         err = __xfrm_state_delete(x);
 575         spin_unlock_bh(&x->lock);
 576
 577         return err;
 578 }
 579 EXPORT_SYMBOL(xfrm_state_delete);
 580
 581 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 582 static inline int
 583 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 584 {
 585         int i, err = 0;
 586
 587         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 588                 struct hlist_node *entry;
 589                 struct xfrm_state *x;
 590
 591                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 592                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 593                            (err = security_xfrm_state_delete(x)) != 0) {
 594                                 xfrm_audit_state_delete(x, 0,
 595                                                         audit_info->loginuid,
 596                                                         audit_info->sessionid,
 597                                                         audit_info->secid);
 598                                 return err;
 599                         }
 600                 }
 601         }
 602
 603         return err;
 604 }
 605 #else
 606 static inline int
 607 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 608 {
 609         return 0;
 610 }
 611 #endif
 612
 613 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 614 {
 615         int i, err = 0, cnt = 0;
 616
 617         spin_lock_bh(&xfrm_state_lock);
 618         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 619         if (err)
 620                 goto out;
 621
 622         err = -ESRCH;
 623         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 624                 struct hlist_node *entry;
 625                 struct xfrm_state *x;
 626 restart:
 627                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 628                         if (!xfrm_state_kern(x) &&
 629                             xfrm_id_proto_match(x->id.proto, proto)) {
 630                                 xfrm_state_hold(x);
 631                                 spin_unlock_bh(&xfrm_state_lock);
 632
 633                                 err = xfrm_state_delete(x);
 634                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 635                                                         audit_info->loginuid,
 636                                                         audit_info->sessionid,
 637                                                         audit_info->secid);
 638                                 xfrm_state_put(x);
 639                                 if (!err)
 640                                         cnt++;
 641
 642                                 spin_lock_bh(&xfrm_state_lock);
 643                                 goto restart;
 644                         }
 645                 }
 646         }
 647         if (cnt)
 648                 err = 0;
 649
 650 out:
 651         spin_unlock_bh(&xfrm_state_lock);
 652         wake_up(&net->xfrm.km_waitq);
 653         return err;
 654 }
 655 EXPORT_SYMBOL(xfrm_state_flush);
 656
 657 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 658 {
 659         spin_lock_bh(&xfrm_state_lock);
 660         si->sadcnt = net->xfrm.state_num;
 661         si->sadhcnt = net->xfrm.state_hmask;
 662         si->sadhmcnt = xfrm_state_hashmax;
 663         spin_unlock_bh(&xfrm_state_lock);
 664 }
 665 EXPORT_SYMBOL(xfrm_sad_getinfo);
 666
 667 static int
 668 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 669                     const struct xfrm_tmpl *tmpl,
 670                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 671                     unsigned short family)
 672 {
 673         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 674         if (!afinfo)
 675                 return -1;
 676         afinfo->init_tempsel(&x->sel, fl);
 677
 678         if (family != tmpl->encap_family) {
 679                 xfrm_state_put_afinfo(afinfo);
 680                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 681                 if (!afinfo)
 682                         return -1;
 683         }
 684         afinfo->init_temprop(x, tmpl, daddr, saddr);
 685         xfrm_state_put_afinfo(afinfo);
 686         return 0;
 687 }
 688
 689 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 690                                               const xfrm_address_t *daddr,
 691                                               __be32 spi, u8 proto,
 692                                               unsigned short family)
 693 {
 694         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 695         struct xfrm_state *x;
 696         struct hlist_node *entry;
 697
 698         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 699                 if (x->props.family != family ||
 700                     x->id.spi       != spi ||
 701                     x->id.proto     != proto ||
 702                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 703                         continue;
 704
 705                 if ((mark & x->mark.m) != x->mark.v)
 706                         continue;
 707                 xfrm_state_hold(x);
 708                 return x;
 709         }
 710
 711         return NULL;
 712 }
 713
 714 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 715                                                      const xfrm_address_t *daddr,
 716                                                      const xfrm_address_t *saddr,
 717                                                      u8 proto, unsigned short family)
 718 {
 719         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 720         struct xfrm_state *x;
 721         struct hlist_node *entry;
 722
 723         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 724                 if (x->props.family != family ||
 725                     x->id.proto     != proto ||
 726                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 727                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 728                         continue;
 729
 730                 if ((mark & x->mark.m) != x->mark.v)
 731                         continue;
 732                 xfrm_state_hold(x);
 733                 return x;
 734         }
 735
 736         return NULL;
 737 }
 738
 739 static inline struct xfrm_state *
 740 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 741 {
 742         struct net *net = xs_net(x);
 743         u32 mark = x->mark.v & x->mark.m;
 744
 745         if (use_spi)
 746                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 747                                            x->id.spi, x->id.proto, family);
 748         else
 749                 return __xfrm_state_lookup_byaddr(net, mark,
 750                                                   &x->id.daddr,
 751                                                   &x->props.saddr,
 752                                                   x->id.proto, family);
 753 }
 754
 755 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 756 {
 757         if (have_hash_collision &&
 758             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 759             net->xfrm.state_num > net->xfrm.state_hmask)
 760                 schedule_work(&net->xfrm.state_hash_work);
 761 }
 762
 763 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 764                                const struct flowi *fl, unsigned short family,
 765                                struct xfrm_state **best, int *acq_in_progress,
 766                                int *error)
 767 {
 768         /* Resolution logic:
 769          * 1. There is a valid state with matching selector. Done.
 770          * 2. Valid state with inappropriate selector. Skip.
 771          *
 772          * Entering area of "sysdeps".
 773          *
 774          * 3. If state is not valid, selector is temporary, it selects
 775          *    only session which triggered previous resolution. Key
 776          *    manager will do something to install a state with proper
 777          *    selector.
 778          */
 779         if (x->km.state == XFRM_STATE_VALID) {
 780                 if ((x->sel.family &&
 781                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 782                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 783                         return;
 784
 785                 if (!*best ||
 786                     (*best)->km.dying > x->km.dying ||
 787                     ((*best)->km.dying == x->km.dying &&
 788                      (*best)->curlft.add_time < x->curlft.add_time))
 789                         *best = x;
 790         } else if (x->km.state == XFRM_STATE_ACQ) {
 791                 *acq_in_progress = 1;
 792         } else if (x->km.state == XFRM_STATE_ERROR ||
 793                    x->km.state == XFRM_STATE_EXPIRED) {
 794                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 795                     security_xfrm_state_pol_flow_match(x, pol, fl))
 796                         *error = -ESRCH;
 797         }
 798 }
 799
 800 struct xfrm_state *
 801 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 802                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 803                 struct xfrm_policy *pol, int *err,
 804                 unsigned short family)
 805 {
 806         static xfrm_address_t saddr_wildcard = { };
 807         struct net *net = xp_net(pol);
 808         unsigned int h, h_wildcard;
 809         struct hlist_node *entry;
 810         struct xfrm_state *x, *x0, *to_put;
 811         int acquire_in_progress = 0;
 812         int error = 0;
 813         struct xfrm_state *best = NULL;
 814         u32 mark = pol->mark.v & pol->mark.m;
 815         unsigned short encap_family = tmpl->encap_family;
 816
 817         to_put = NULL;
 818
 819         spin_lock_bh(&xfrm_state_lock);
 820         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 821         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 822                 if (x->props.family == encap_family &&
 823                     x->props.reqid == tmpl->reqid &&
 824                     (mark & x->mark.m) == x->mark.v &&
 825                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 826                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 827                     tmpl->mode == x->props.mode &&
 828                     tmpl->id.proto == x->id.proto &&
 829                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 830                         xfrm_state_look_at(pol, x, fl, encap_family,
 831                                            &best, &acquire_in_progress, &error);
 832         }
 833         if (best)
 834                 goto found;
 835
 836         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 837         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 838                 if (x->props.family == encap_family &&
 839                     x->props.reqid == tmpl->reqid &&
 840                     (mark & x->mark.m) == x->mark.v &&
 841                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 842                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 843                     tmpl->mode == x->props.mode &&
 844                     tmpl->id.proto == x->id.proto &&
 845                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 846                         xfrm_state_look_at(pol, x, fl, encap_family,
 847                                            &best, &acquire_in_progress, &error);
 848         }
 849
 850 found:
 851         x = best;
 852         if (!x && !error && !acquire_in_progress) {
 853                 if (tmpl->id.spi &&
 854                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 855                                               tmpl->id.proto, encap_family)) != NULL) {
 856                         to_put = x0;
 857                         error = -EEXIST;
 858                         goto out;
 859                 }
 860                 x = xfrm_state_alloc(net);
 861                 if (x == NULL) {
 862                         error = -ENOMEM;
 863                         goto out;
 864                 }
 865                 /* Initialize temporary state matching only
 866                  * to current session. */
 867                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 868                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 869
 870                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 871                 if (error) {
 872                         x->km.state = XFRM_STATE_DEAD;
 873                         to_put = x;
 874                         x = NULL;
 875                         goto out;
 876                 }
 877
 878                 if (km_query(x, tmpl, pol) == 0) {
 879                         x->km.state = XFRM_STATE_ACQ;
 880                         list_add(&x->km.all, &net->xfrm.state_all);
 881                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 882                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 883                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 884                         if (x->id.spi) {
 885                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 886                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 887                         }
 888                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 889                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 890                         net->xfrm.state_num++;
 891                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 892                 } else {
 893                         x->km.state = XFRM_STATE_DEAD;
 894                         to_put = x;
 895                         x = NULL;
 896                         error = -ESRCH;
 897                 }
 898         }
 899 out:
 900         if (x)
 901                 xfrm_state_hold(x);
 902         else
 903                 *err = acquire_in_progress ? -EAGAIN : error;
 904         spin_unlock_bh(&xfrm_state_lock);
 905         if (to_put)
 906                 xfrm_state_put(to_put);
 907         return x;
 908 }
 909
 910 struct xfrm_state *
 911 xfrm_stateonly_find(struct net *net, u32 mark,
 912                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 913                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 914 {
 915         unsigned int h;
 916         struct xfrm_state *rx = NULL, *x = NULL;
 917         struct hlist_node *entry;
 918
 919         spin_lock(&xfrm_state_lock);
 920         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 921         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 922                 if (x->props.family == family &&
 923                     x->props.reqid == reqid &&
 924                     (mark & x->mark.m) == x->mark.v &&
 925                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 926                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 927                     mode == x->props.mode &&
 928                     proto == x->id.proto &&
 929                     x->km.state == XFRM_STATE_VALID) {
 930                         rx = x;
 931                         break;
 932                 }
 933         }
 934
 935         if (rx)
 936                 xfrm_state_hold(rx);
 937         spin_unlock(&xfrm_state_lock);
 938
 939
 940         return rx;
 941 }
 942 EXPORT_SYMBOL(xfrm_stateonly_find);
 943
 944 static void __xfrm_state_insert(struct xfrm_state *x)
 945 {
 946         struct net *net = xs_net(x);
 947         unsigned int h;
 948
 949         list_add(&x->km.all, &net->xfrm.state_all);
 950
 951         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 952                           x->props.reqid, x->props.family);
 953         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 954
 955         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 956         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 957
 958         if (x->id.spi) {
 959                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 960                                   x->props.family);
 961
 962                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 963         }
 964
 965         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 966         if (x->replay_maxage)
 967                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 968
 969         wake_up(&net->xfrm.km_waitq);
 970
 971         net->xfrm.state_num++;
 972
 973         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 974 }
 975
 976 /* xfrm_state_lock is held */
 977 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 978 {
 979         struct net *net = xs_net(xnew);
 980         unsigned short family = xnew->props.family;
 981         u32 reqid = xnew->props.reqid;
 982         struct xfrm_state *x;
 983         struct hlist_node *entry;
 984         unsigned int h;
 985         u32 mark = xnew->mark.v & xnew->mark.m;
 986
 987         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 988         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 989                 if (x->props.family     == family &&
 990                     x->props.reqid      == reqid &&
 991                     (mark & x->mark.m) == x->mark.v &&
 992                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 993                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 994                         x->genid++;
 995         }
 996 }
 997
 998 void xfrm_state_insert(struct xfrm_state *x)
 999 {
1000         spin_lock_bh(&xfrm_state_lock);
1001         __xfrm_state_bump_genids(x);
1002         __xfrm_state_insert(x);
1003         spin_unlock_bh(&xfrm_state_lock);
1004 }
1005 EXPORT_SYMBOL(xfrm_state_insert);
1006
1007 /* xfrm_state_lock is held */
1008 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m,
1009                                           unsigned short family, u8 mode,
1010                                           u32 reqid, u8 proto,
1011                                           const xfrm_address_t *daddr,
1012                                           const xfrm_address_t *saddr, int create)
1013 {
1014         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1015         struct hlist_node *entry;
1016         struct xfrm_state *x;
1017         u32 mark = m->v & m->m;
1018
1019         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
1020                 if (x->props.reqid  != reqid ||
1021                     x->props.mode   != mode ||
1022                     x->props.family != family ||
1023                     x->km.state     != XFRM_STATE_ACQ ||
1024                     x->id.spi       != 0 ||
1025                     x->id.proto     != proto ||
1026                     (mark & x->mark.m) != x->mark.v ||
1027                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1028                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1029                         continue;
1030
1031                 xfrm_state_hold(x);
1032                 return x;
1033         }
1034
1035         if (!create)
1036                 return NULL;
1037
1038         x = xfrm_state_alloc(net);
1039         if (likely(x)) {
1040                 switch (family) {
1041                 case AF_INET:
1042                         x->sel.daddr.a4 = daddr->a4;
1043                         x->sel.saddr.a4 = saddr->a4;
1044                         x->sel.prefixlen_d = 32;
1045                         x->sel.prefixlen_s = 32;
1046                         x->props.saddr.a4 = saddr->a4;
1047                         x->id.daddr.a4 = daddr->a4;
1048                         break;
1049
1050                 case AF_INET6:
1051                         *(struct in6_addr *)x->sel.daddr.a6 = *(struct in6_addr *)daddr;
1052                         *(struct in6_addr *)x->sel.saddr.a6 = *(struct in6_addr *)saddr;
1053                         x->sel.prefixlen_d = 128;
1054                         x->sel.prefixlen_s = 128;
1055                         *(struct in6_addr *)x->props.saddr.a6 = *(struct in6_addr *)saddr;
1056                         *(struct in6_addr *)x->id.daddr.a6 = *(struct in6_addr *)daddr;
1057                         break;
1058                 }
1059
1060                 x->km.state = XFRM_STATE_ACQ;
1061                 x->id.proto = proto;
1062                 x->props.family = family;
1063                 x->props.mode = mode;
1064                 x->props.reqid = reqid;
1065                 x->mark.v = m->v;
1066                 x->mark.m = m->m;
1067                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1068                 xfrm_state_hold(x);
1069                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1070                 list_add(&x->km.all, &net->xfrm.state_all);
1071                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1072                 h = xfrm_src_hash(net, daddr, saddr, family);
1073                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1074
1075                 net->xfrm.state_num++;
1076
1077                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1078         }
1079
1080         return x;
1081 }
1082
1083 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1084
1085 int xfrm_state_add(struct xfrm_state *x)
1086 {
1087         struct net *net = xs_net(x);
1088         struct xfrm_state *x1, *to_put;
1089         int family;
1090         int err;
1091         u32 mark = x->mark.v & x->mark.m;
1092         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1093
1094         family = x->props.family;
1095
1096         to_put = NULL;
1097
1098         spin_lock_bh(&xfrm_state_lock);
1099
1100         x1 = __xfrm_state_locate(x, use_spi, family);
1101         if (x1) {
1102                 to_put = x1;
1103                 x1 = NULL;
1104                 err = -EEXIST;
1105                 goto out;
1106         }
1107
1108         if (use_spi && x->km.seq) {
1109                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1110                 if (x1 && ((x1->id.proto != x->id.proto) ||
1111                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1112                         to_put = x1;
1113                         x1 = NULL;
1114                 }
1115         }
1116
1117         if (use_spi && !x1)
1118                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1119                                      x->props.reqid, x->id.proto,
1120                                      &x->id.daddr, &x->props.saddr, 0);
1121
1122         __xfrm_state_bump_genids(x);
1123         __xfrm_state_insert(x);
1124         err = 0;
1125
1126 out:
1127         spin_unlock_bh(&xfrm_state_lock);
1128
1129         if (x1) {
1130                 xfrm_state_delete(x1);
1131                 xfrm_state_put(x1);
1132         }
1133
1134         if (to_put)
1135                 xfrm_state_put(to_put);
1136
1137         return err;
1138 }
1139 EXPORT_SYMBOL(xfrm_state_add);
1140
1141 #ifdef CONFIG_XFRM_MIGRATE
1142 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1143 {
1144         struct net *net = xs_net(orig);
1145         int err = -ENOMEM;
1146         struct xfrm_state *x = xfrm_state_alloc(net);
1147         if (!x)
1148                 goto out;
1149
1150         memcpy(&x->id, &orig->id, sizeof(x->id));
1151         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1152         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1153         x->props.mode = orig->props.mode;
1154         x->props.replay_window = orig->props.replay_window;
1155         x->props.reqid = orig->props.reqid;
1156         x->props.family = orig->props.family;
1157         x->props.saddr = orig->props.saddr;
1158
1159         if (orig->aalg) {
1160                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1161                 if (!x->aalg)
1162                         goto error;
1163         }
1164         x->props.aalgo = orig->props.aalgo;
1165
1166         if (orig->ealg) {
1167                 x->ealg = xfrm_algo_clone(orig->ealg);
1168                 if (!x->ealg)
1169                         goto error;
1170         }
1171         x->props.ealgo = orig->props.ealgo;
1172
1173         if (orig->calg) {
1174                 x->calg = xfrm_algo_clone(orig->calg);
1175                 if (!x->calg)
1176                         goto error;
1177         }
1178         x->props.calgo = orig->props.calgo;
1179
1180         if (orig->encap) {
1181                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1182                 if (!x->encap)
1183                         goto error;
1184         }
1185
1186         if (orig->coaddr) {
1187                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1188                                     GFP_KERNEL);
1189                 if (!x->coaddr)
1190                         goto error;
1191         }
1192
1193         if (orig->replay_esn) {
1194                 err = xfrm_replay_clone(x, orig);
1195                 if (err)
1196                         goto error;
1197         }
1198
1199         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1200
1201         err = xfrm_init_state(x);
1202         if (err)
1203                 goto error;
1204
1205         x->props.flags = orig->props.flags;
1206
1207         x->curlft.add_time = orig->curlft.add_time;
1208         x->km.state = orig->km.state;
1209         x->km.seq = orig->km.seq;
1210
1211         return x;
1212
1213  error:
1214         xfrm_state_put(x);
1215 out:
1216         if (errp)
1217                 *errp = err;
1218         return NULL;
1219 }
1220
1221 /* xfrm_state_lock is held */
1222 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1223 {
1224         unsigned int h;
1225         struct xfrm_state *x;
1226         struct hlist_node *entry;
1227
1228         if (m->reqid) {
1229                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1230                                   m->reqid, m->old_family);
1231                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1232                         if (x->props.mode != m->mode ||
1233                             x->id.proto != m->proto)
1234                                 continue;
1235                         if (m->reqid && x->props.reqid != m->reqid)
1236                                 continue;
1237                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1238                                           m->old_family) ||
1239                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1240                                           m->old_family))
1241                                 continue;
1242                         xfrm_state_hold(x);
1243                         return x;
1244                 }
1245         } else {
1246                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1247                                   m->old_family);
1248                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1249                         if (x->props.mode != m->mode ||
1250                             x->id.proto != m->proto)
1251                                 continue;
1252                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1253                                           m->old_family) ||
1254                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1255                                           m->old_family))
1256                                 continue;
1257                         xfrm_state_hold(x);
1258                         return x;
1259                 }
1260         }
1261
1262         return NULL;
1263 }
1264 EXPORT_SYMBOL(xfrm_migrate_state_find);
1265
1266 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1267                                        struct xfrm_migrate *m)
1268 {
1269         struct xfrm_state *xc;
1270         int err;
1271
1272         xc = xfrm_state_clone(x, &err);
1273         if (!xc)
1274                 return NULL;
1275
1276         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1277         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1278
1279         /* add state */
1280         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1281                 /* a care is needed when the destination address of the
1282                    state is to be updated as it is a part of triplet */
1283                 xfrm_state_insert(xc);
1284         } else {
1285                 if ((err = xfrm_state_add(xc)) < 0)
1286                         goto error;
1287         }
1288
1289         return xc;
1290 error:
1291         xfrm_state_put(xc);
1292         return NULL;
1293 }
1294 EXPORT_SYMBOL(xfrm_state_migrate);
1295 #endif
1296
1297 int xfrm_state_update(struct xfrm_state *x)
1298 {
1299         struct xfrm_state *x1, *to_put;
1300         int err;
1301         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1302
1303         to_put = NULL;
1304
1305         spin_lock_bh(&xfrm_state_lock);
1306         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1307
1308         err = -ESRCH;
1309         if (!x1)
1310                 goto out;
1311
1312         if (xfrm_state_kern(x1)) {
1313                 to_put = x1;
1314                 err = -EEXIST;
1315                 goto out;
1316         }
1317
1318         if (x1->km.state == XFRM_STATE_ACQ) {
1319                 __xfrm_state_insert(x);
1320                 x = NULL;
1321         }
1322         err = 0;
1323
1324 out:
1325         spin_unlock_bh(&xfrm_state_lock);
1326
1327         if (to_put)
1328                 xfrm_state_put(to_put);
1329
1330         if (err)
1331                 return err;
1332
1333         if (!x) {
1334                 xfrm_state_delete(x1);
1335                 xfrm_state_put(x1);
1336                 return 0;
1337         }
1338
1339         err = -EINVAL;
1340         spin_lock_bh(&x1->lock);
1341         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1342                 if (x->encap && x1->encap)
1343                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1344                 if (x->coaddr && x1->coaddr) {
1345                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1346                 }
1347                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1348                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1349                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1350                 x1->km.dying = 0;
1351
1352                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1353                 if (x1->curlft.use_time)
1354                         xfrm_state_check_expire(x1);
1355
1356                 err = 0;
1357                 x->km.state = XFRM_STATE_DEAD;
1358                 __xfrm_state_put(x);
1359         }
1360         spin_unlock_bh(&x1->lock);
1361
1362         xfrm_state_put(x1);
1363
1364         return err;
1365 }
1366 EXPORT_SYMBOL(xfrm_state_update);
1367
1368 int xfrm_state_check_expire(struct xfrm_state *x)
1369 {
1370         if (!x->curlft.use_time)
1371                 x->curlft.use_time = get_seconds();
1372
1373         if (x->km.state != XFRM_STATE_VALID)
1374                 return -EINVAL;
1375
1376         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1377             x->curlft.packets >= x->lft.hard_packet_limit) {
1378                 x->km.state = XFRM_STATE_EXPIRED;
1379                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1380                 return -EINVAL;
1381         }
1382
1383         if (!x->km.dying &&
1384             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1385              x->curlft.packets >= x->lft.soft_packet_limit)) {
1386                 x->km.dying = 1;
1387                 km_state_expired(x, 0, 0);
1388         }
1389         return 0;
1390 }
1391 EXPORT_SYMBOL(xfrm_state_check_expire);
1392
1393 struct xfrm_state *
1394 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1395                   u8 proto, unsigned short family)
1396 {
1397         struct xfrm_state *x;
1398
1399         spin_lock_bh(&xfrm_state_lock);
1400         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1401         spin_unlock_bh(&xfrm_state_lock);
1402         return x;
1403 }
1404 EXPORT_SYMBOL(xfrm_state_lookup);
1405
1406 struct xfrm_state *
1407 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1408                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1409                          u8 proto, unsigned short family)
1410 {
1411         struct xfrm_state *x;
1412
1413         spin_lock_bh(&xfrm_state_lock);
1414         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1415         spin_unlock_bh(&xfrm_state_lock);
1416         return x;
1417 }
1418 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1419
1420 struct xfrm_state *
1421 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1422               const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1423               int create, unsigned short family)
1424 {
1425         struct xfrm_state *x;
1426
1427         spin_lock_bh(&xfrm_state_lock);
1428         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1429         spin_unlock_bh(&xfrm_state_lock);
1430
1431         return x;
1432 }
1433 EXPORT_SYMBOL(xfrm_find_acq);
1434
1435 #ifdef CONFIG_XFRM_SUB_POLICY
1436 int
1437 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1438                unsigned short family)
1439 {
1440         int err = 0;
1441         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1442         if (!afinfo)
1443                 return -EAFNOSUPPORT;
1444
1445         spin_lock_bh(&xfrm_state_lock);
1446         if (afinfo->tmpl_sort)
1447                 err = afinfo->tmpl_sort(dst, src, n);
1448         spin_unlock_bh(&xfrm_state_lock);
1449         xfrm_state_put_afinfo(afinfo);
1450         return err;
1451 }
1452 EXPORT_SYMBOL(xfrm_tmpl_sort);
1453
1454 int
1455 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1456                 unsigned short family)
1457 {
1458         int err = 0;
1459         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1460         if (!afinfo)
1461                 return -EAFNOSUPPORT;
1462
1463         spin_lock_bh(&xfrm_state_lock);
1464         if (afinfo->state_sort)
1465                 err = afinfo->state_sort(dst, src, n);
1466         spin_unlock_bh(&xfrm_state_lock);
1467         xfrm_state_put_afinfo(afinfo);
1468         return err;
1469 }
1470 EXPORT_SYMBOL(xfrm_state_sort);
1471 #endif
1472
1473 /* Silly enough, but I'm lazy to build resolution list */
1474
1475 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1476 {
1477         int i;
1478
1479         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1480                 struct hlist_node *entry;
1481                 struct xfrm_state *x;
1482
1483                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1484                         if (x->km.seq == seq &&
1485                             (mark & x->mark.m) == x->mark.v &&
1486                             x->km.state == XFRM_STATE_ACQ) {
1487                                 xfrm_state_hold(x);
1488                                 return x;
1489                         }
1490                 }
1491         }
1492         return NULL;
1493 }
1494
1495 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1496 {
1497         struct xfrm_state *x;
1498
1499         spin_lock_bh(&xfrm_state_lock);
1500         x = __xfrm_find_acq_byseq(net, mark, seq);
1501         spin_unlock_bh(&xfrm_state_lock);
1502         return x;
1503 }
1504 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1505
1506 u32 xfrm_get_acqseq(void)
1507 {
1508         u32 res;
1509         static atomic_t acqseq;
1510
1511         do {
1512                 res = atomic_inc_return(&acqseq);
1513         } while (!res);
1514
1515         return res;
1516 }
1517 EXPORT_SYMBOL(xfrm_get_acqseq);
1518
1519 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1520 {
1521         struct net *net = xs_net(x);
1522         unsigned int h;
1523         struct xfrm_state *x0;
1524         int err = -ENOENT;
1525         __be32 minspi = htonl(low);
1526         __be32 maxspi = htonl(high);
1527         u32 mark = x->mark.v & x->mark.m;
1528
1529         spin_lock_bh(&x->lock);
1530         if (x->km.state == XFRM_STATE_DEAD)
1531                 goto unlock;
1532
1533         err = 0;
1534         if (x->id.spi)
1535                 goto unlock;
1536
1537         err = -ENOENT;
1538
1539         if (minspi == maxspi) {
1540                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1541                 if (x0) {
1542                         xfrm_state_put(x0);
1543                         goto unlock;
1544                 }
1545                 x->id.spi = minspi;
1546         } else {
1547                 u32 spi = 0;
1548                 for (h=0; h<high-low+1; h++) {
1549                         spi = low + net_random()%(high-low+1);
1550                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1551                         if (x0 == NULL) {
1552                                 x->id.spi = htonl(spi);
1553                                 break;
1554                         }
1555                         xfrm_state_put(x0);
1556                 }
1557         }
1558         if (x->id.spi) {
1559                 spin_lock_bh(&xfrm_state_lock);
1560                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1561                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1562                 spin_unlock_bh(&xfrm_state_lock);
1563
1564                 err = 0;
1565         }
1566
1567 unlock:
1568         spin_unlock_bh(&x->lock);
1569
1570         return err;
1571 }
1572 EXPORT_SYMBOL(xfrm_alloc_spi);
1573
1574 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1575                     int (*func)(struct xfrm_state *, int, void*),
1576                     void *data)
1577 {
1578         struct xfrm_state *state;
1579         struct xfrm_state_walk *x;
1580         int err = 0;
1581
1582         if (walk->seq != 0 && list_empty(&walk->all))
1583                 return 0;
1584
1585         spin_lock_bh(&xfrm_state_lock);
1586         if (list_empty(&walk->all))
1587                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1588         else
1589                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1590         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1591                 if (x->state == XFRM_STATE_DEAD)
1592                         continue;
1593                 state = container_of(x, struct xfrm_state, km);
1594                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1595                         continue;
1596                 err = func(state, walk->seq, data);
1597                 if (err) {
1598                         list_move_tail(&walk->all, &x->all);
1599                         goto out;
1600                 }
1601                 walk->seq++;
1602         }
1603         if (walk->seq == 0) {
1604                 err = -ENOENT;
1605                 goto out;
1606         }
1607         list_del_init(&walk->all);
1608 out:
1609         spin_unlock_bh(&xfrm_state_lock);
1610         return err;
1611 }
1612 EXPORT_SYMBOL(xfrm_state_walk);
1613
1614 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1615 {
1616         INIT_LIST_HEAD(&walk->all);
1617         walk->proto = proto;
1618         walk->state = XFRM_STATE_DEAD;
1619         walk->seq = 0;
1620 }
1621 EXPORT_SYMBOL(xfrm_state_walk_init);
1622
1623 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1624 {
1625         if (list_empty(&walk->all))
1626                 return;
1627
1628         spin_lock_bh(&xfrm_state_lock);
1629         list_del(&walk->all);
1630         spin_unlock_bh(&xfrm_state_lock);
1631 }
1632 EXPORT_SYMBOL(xfrm_state_walk_done);
1633
1634 static void xfrm_replay_timer_handler(unsigned long data)
1635 {
1636         struct xfrm_state *x = (struct xfrm_state*)data;
1637
1638         spin_lock(&x->lock);
1639
1640         if (x->km.state == XFRM_STATE_VALID) {
1641                 if (xfrm_aevent_is_on(xs_net(x)))
1642                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1643                 else
1644                         x->xflags |= XFRM_TIME_DEFER;
1645         }
1646
1647         spin_unlock(&x->lock);
1648 }
1649
1650 static LIST_HEAD(xfrm_km_list);
1651 static DEFINE_RWLOCK(xfrm_km_lock);
1652
1653 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1654 {
1655         struct xfrm_mgr *km;
1656
1657         read_lock(&xfrm_km_lock);
1658         list_for_each_entry(km, &xfrm_km_list, list)
1659                 if (km->notify_policy)
1660                         km->notify_policy(xp, dir, c);
1661         read_unlock(&xfrm_km_lock);
1662 }
1663
1664 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1665 {
1666         struct xfrm_mgr *km;
1667         read_lock(&xfrm_km_lock);
1668         list_for_each_entry(km, &xfrm_km_list, list)
1669                 if (km->notify)
1670                         km->notify(x, c);
1671         read_unlock(&xfrm_km_lock);
1672 }
1673
1674 EXPORT_SYMBOL(km_policy_notify);
1675 EXPORT_SYMBOL(km_state_notify);
1676
1677 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1678 {
1679         struct net *net = xs_net(x);
1680         struct km_event c;
1681
1682         c.data.hard = hard;
1683         c.portid = portid;
1684         c.event = XFRM_MSG_EXPIRE;
1685         km_state_notify(x, &c);
1686
1687         if (hard)
1688                 wake_up(&net->xfrm.km_waitq);
1689 }
1690
1691 EXPORT_SYMBOL(km_state_expired);
1692 /*
1693  * We send to all registered managers regardless of failure
1694  * We are happy with one success
1695 */
1696 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1697 {
1698         int err = -EINVAL, acqret;
1699         struct xfrm_mgr *km;
1700
1701         read_lock(&xfrm_km_lock);
1702         list_for_each_entry(km, &xfrm_km_list, list) {
1703                 acqret = km->acquire(x, t, pol);
1704                 if (!acqret)
1705                         err = acqret;
1706         }
1707         read_unlock(&xfrm_km_lock);
1708         return err;
1709 }
1710 EXPORT_SYMBOL(km_query);
1711
1712 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1713 {
1714         int err = -EINVAL;
1715         struct xfrm_mgr *km;
1716
1717         read_lock(&xfrm_km_lock);
1718         list_for_each_entry(km, &xfrm_km_list, list) {
1719                 if (km->new_mapping)
1720                         err = km->new_mapping(x, ipaddr, sport);
1721                 if (!err)
1722                         break;
1723         }
1724         read_unlock(&xfrm_km_lock);
1725         return err;
1726 }
1727 EXPORT_SYMBOL(km_new_mapping);
1728
1729 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1730 {
1731         struct net *net = xp_net(pol);
1732         struct km_event c;
1733
1734         c.data.hard = hard;
1735         c.portid = portid;
1736         c.event = XFRM_MSG_POLEXPIRE;
1737         km_policy_notify(pol, dir, &c);
1738
1739         if (hard)
1740                 wake_up(&net->xfrm.km_waitq);
1741 }
1742 EXPORT_SYMBOL(km_policy_expired);
1743
1744 #ifdef CONFIG_XFRM_MIGRATE
1745 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1746                const struct xfrm_migrate *m, int num_migrate,
1747                const struct xfrm_kmaddress *k)
1748 {
1749         int err = -EINVAL;
1750         int ret;
1751         struct xfrm_mgr *km;
1752
1753         read_lock(&xfrm_km_lock);
1754         list_for_each_entry(km, &xfrm_km_list, list) {
1755                 if (km->migrate) {
1756                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1757                         if (!ret)
1758                                 err = ret;
1759                 }
1760         }
1761         read_unlock(&xfrm_km_lock);
1762         return err;
1763 }
1764 EXPORT_SYMBOL(km_migrate);
1765 #endif
1766
1767 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1768 {
1769         int err = -EINVAL;
1770         int ret;
1771         struct xfrm_mgr *km;
1772
1773         read_lock(&xfrm_km_lock);
1774         list_for_each_entry(km, &xfrm_km_list, list) {
1775                 if (km->report) {
1776                         ret = km->report(net, proto, sel, addr);
1777                         if (!ret)
1778                                 err = ret;
1779                 }
1780         }
1781         read_unlock(&xfrm_km_lock);
1782         return err;
1783 }
1784 EXPORT_SYMBOL(km_report);
1785
1786 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1787 {
1788         int err;
1789         u8 *data;
1790         struct xfrm_mgr *km;
1791         struct xfrm_policy *pol = NULL;
1792
1793         if (optlen <= 0 || optlen > PAGE_SIZE)
1794                 return -EMSGSIZE;
1795
1796         data = kmalloc(optlen, GFP_KERNEL);
1797         if (!data)
1798                 return -ENOMEM;
1799
1800         err = -EFAULT;
1801         if (copy_from_user(data, optval, optlen))
1802                 goto out;
1803
1804         err = -EINVAL;
1805         read_lock(&xfrm_km_lock);
1806         list_for_each_entry(km, &xfrm_km_list, list) {
1807                 pol = km->compile_policy(sk, optname, data,
1808                                          optlen, &err);
1809                 if (err >= 0)
1810                         break;
1811         }
1812         read_unlock(&xfrm_km_lock);
1813
1814         if (err >= 0) {
1815                 xfrm_sk_policy_insert(sk, err, pol);
1816                 xfrm_pol_put(pol);
1817                 err = 0;
1818         }
1819
1820 out:
1821         kfree(data);
1822         return err;
1823 }
1824 EXPORT_SYMBOL(xfrm_user_policy);
1825
1826 int xfrm_register_km(struct xfrm_mgr *km)
1827 {
1828         write_lock_bh(&xfrm_km_lock);
1829         list_add_tail(&km->list, &xfrm_km_list);
1830         write_unlock_bh(&xfrm_km_lock);
1831         return 0;
1832 }
1833 EXPORT_SYMBOL(xfrm_register_km);
1834
1835 int xfrm_unregister_km(struct xfrm_mgr *km)
1836 {
1837         write_lock_bh(&xfrm_km_lock);
1838         list_del(&km->list);
1839         write_unlock_bh(&xfrm_km_lock);
1840         return 0;
1841 }
1842 EXPORT_SYMBOL(xfrm_unregister_km);
1843
1844 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1845 {
1846         int err = 0;
1847         if (unlikely(afinfo == NULL))
1848                 return -EINVAL;
1849         if (unlikely(afinfo->family >= NPROTO))
1850                 return -EAFNOSUPPORT;
1851         write_lock_bh(&xfrm_state_afinfo_lock);
1852         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1853                 err = -ENOBUFS;
1854         else
1855                 xfrm_state_afinfo[afinfo->family] = afinfo;
1856         write_unlock_bh(&xfrm_state_afinfo_lock);
1857         return err;
1858 }
1859 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1860
1861 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1862 {
1863         int err = 0;
1864         if (unlikely(afinfo == NULL))
1865                 return -EINVAL;
1866         if (unlikely(afinfo->family >= NPROTO))
1867                 return -EAFNOSUPPORT;
1868         write_lock_bh(&xfrm_state_afinfo_lock);
1869         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1870                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1871                         err = -EINVAL;
1872                 else
1873                         xfrm_state_afinfo[afinfo->family] = NULL;
1874         }
1875         write_unlock_bh(&xfrm_state_afinfo_lock);
1876         return err;
1877 }
1878 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1879
1880 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1881 {
1882         struct xfrm_state_afinfo *afinfo;
1883         if (unlikely(family >= NPROTO))
1884                 return NULL;
1885         read_lock(&xfrm_state_afinfo_lock);
1886         afinfo = xfrm_state_afinfo[family];
1887         if (unlikely(!afinfo))
1888                 read_unlock(&xfrm_state_afinfo_lock);
1889         return afinfo;
1890 }
1891
1892 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1893         __releases(xfrm_state_afinfo_lock)
1894 {
1895         read_unlock(&xfrm_state_afinfo_lock);
1896 }
1897
1898 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1899 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1900 {
1901         if (x->tunnel) {
1902                 struct xfrm_state *t = x->tunnel;
1903
1904                 if (atomic_read(&t->tunnel_users) == 2)
1905                         xfrm_state_delete(t);
1906                 atomic_dec(&t->tunnel_users);
1907                 xfrm_state_put(t);
1908                 x->tunnel = NULL;
1909         }
1910 }
1911 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1912
1913 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1914 {
1915         int res;
1916
1917         spin_lock_bh(&x->lock);
1918         if (x->km.state == XFRM_STATE_VALID &&
1919             x->type && x->type->get_mtu)
1920                 res = x->type->get_mtu(x, mtu);
1921         else
1922                 res = mtu - x->props.header_len;
1923         spin_unlock_bh(&x->lock);
1924         return res;
1925 }
1926
1927 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1928 {
1929         struct xfrm_state_afinfo *afinfo;
1930         struct xfrm_mode *inner_mode;
1931         int family = x->props.family;
1932         int err;
1933
1934         err = -EAFNOSUPPORT;
1935         afinfo = xfrm_state_get_afinfo(family);
1936         if (!afinfo)
1937                 goto error;
1938
1939         err = 0;
1940         if (afinfo->init_flags)
1941                 err = afinfo->init_flags(x);
1942
1943         xfrm_state_put_afinfo(afinfo);
1944
1945         if (err)
1946                 goto error;
1947
1948         err = -EPROTONOSUPPORT;
1949
1950         if (x->sel.family != AF_UNSPEC) {
1951                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1952                 if (inner_mode == NULL)
1953                         goto error;
1954
1955                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
1956                     family != x->sel.family) {
1957                         xfrm_put_mode(inner_mode);
1958                         goto error;
1959                 }
1960
1961                 x->inner_mode = inner_mode;
1962         } else {
1963                 struct xfrm_mode *inner_mode_iaf;
1964                 int iafamily = AF_INET;
1965
1966                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
1967                 if (inner_mode == NULL)
1968                         goto error;
1969
1970                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
1971                         xfrm_put_mode(inner_mode);
1972                         goto error;
1973                 }
1974                 x->inner_mode = inner_mode;
1975
1976                 if (x->props.family == AF_INET)
1977                         iafamily = AF_INET6;
1978
1979                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
1980                 if (inner_mode_iaf) {
1981                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
1982                                 x->inner_mode_iaf = inner_mode_iaf;
1983                         else
1984                                 xfrm_put_mode(inner_mode_iaf);
1985                 }
1986         }
1987
1988         x->type = xfrm_get_type(x->id.proto, family);
1989         if (x->type == NULL)
1990                 goto error;
1991
1992         err = x->type->init_state(x);
1993         if (err)
1994                 goto error;
1995
1996         x->outer_mode = xfrm_get_mode(x->props.mode, family);
1997         if (x->outer_mode == NULL) {
1998                 err = -EPROTONOSUPPORT;
1999                 goto error;
2000         }
2001
2002         if (init_replay) {
2003                 err = xfrm_init_replay(x);
2004                 if (err)
2005                         goto error;
2006         }
2007
2008         x->km.state = XFRM_STATE_VALID;
2009
2010 error:
2011         return err;
2012 }
2013
2014 EXPORT_SYMBOL(__xfrm_init_state);
2015
2016 int xfrm_init_state(struct xfrm_state *x)
2017 {
2018         return __xfrm_init_state(x, true);
2019 }
2020
2021 EXPORT_SYMBOL(xfrm_init_state);
2022
2023 int __net_init xfrm_state_init(struct net *net)
2024 {
2025         unsigned int sz;
2026
2027         INIT_LIST_HEAD(&net->xfrm.state_all);
2028
2029         sz = sizeof(struct hlist_head) * 8;
2030
2031         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2032         if (!net->xfrm.state_bydst)
2033                 goto out_bydst;
2034         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2035         if (!net->xfrm.state_bysrc)
2036                 goto out_bysrc;
2037         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2038         if (!net->xfrm.state_byspi)
2039                 goto out_byspi;
2040         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2041
2042         net->xfrm.state_num = 0;
2043         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2044         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2045         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2046         init_waitqueue_head(&net->xfrm.km_waitq);
2047         return 0;
2048
2049 out_byspi:
2050         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2051 out_bysrc:
2052         xfrm_hash_free(net->xfrm.state_bydst, sz);
2053 out_bydst:
2054         return -ENOMEM;
2055 }
2056
2057 void xfrm_state_fini(struct net *net)
2058 {
2059         struct xfrm_audit audit_info;
2060         unsigned int sz;
2061
2062         flush_work(&net->xfrm.state_hash_work);
2063         audit_info.loginuid = INVALID_UID;
2064         audit_info.sessionid = -1;
2065         audit_info.secid = 0;
2066         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2067         flush_work(&net->xfrm.state_gc_work);
2068
2069         WARN_ON(!list_empty(&net->xfrm.state_all));
2070
2071         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2072         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2073         xfrm_hash_free(net->xfrm.state_byspi, sz);
2074         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2075         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2076         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2077         xfrm_hash_free(net->xfrm.state_bydst, sz);
2078 }
2079
2080 #ifdef CONFIG_AUDITSYSCALL
2081 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2082                                      struct audit_buffer *audit_buf)
2083 {
2084         struct xfrm_sec_ctx *ctx = x->security;
2085         u32 spi = ntohl(x->id.spi);
2086
2087         if (ctx)
2088                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2089                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2090
2091         switch(x->props.family) {
2092         case AF_INET:
2093                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2094                                  &x->props.saddr.a4, &x->id.daddr.a4);
2095                 break;
2096         case AF_INET6:
2097                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2098                                  x->props.saddr.a6, x->id.daddr.a6);
2099                 break;
2100         }
2101
2102         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2103 }
2104
2105 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2106                                       struct audit_buffer *audit_buf)
2107 {
2108         const struct iphdr *iph4;
2109         const struct ipv6hdr *iph6;
2110
2111         switch (family) {
2112         case AF_INET:
2113                 iph4 = ip_hdr(skb);
2114                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2115                                  &iph4->saddr, &iph4->daddr);
2116                 break;
2117         case AF_INET6:
2118                 iph6 = ipv6_hdr(skb);
2119                 audit_log_format(audit_buf,
2120                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2121                                  &iph6->saddr,&iph6->daddr,
2122                                  iph6->flow_lbl[0] & 0x0f,
2123                                  iph6->flow_lbl[1],
2124                                  iph6->flow_lbl[2]);
2125                 break;
2126         }
2127 }
2128
2129 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2130                           kuid_t auid, u32 sessionid, u32 secid)
2131 {
2132         struct audit_buffer *audit_buf;
2133
2134         audit_buf = xfrm_audit_start("SAD-add");
2135         if (audit_buf == NULL)
2136                 return;
2137         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2138         xfrm_audit_helper_sainfo(x, audit_buf);
2139         audit_log_format(audit_buf, " res=%u", result);
2140         audit_log_end(audit_buf);
2141 }
2142 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2143
2144 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2145                              kuid_t auid, u32 sessionid, u32 secid)
2146 {
2147         struct audit_buffer *audit_buf;
2148
2149         audit_buf = xfrm_audit_start("SAD-delete");
2150         if (audit_buf == NULL)
2151                 return;
2152         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2153         xfrm_audit_helper_sainfo(x, audit_buf);
2154         audit_log_format(audit_buf, " res=%u", result);
2155         audit_log_end(audit_buf);
2156 }
2157 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2158
2159 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2160                                       struct sk_buff *skb)
2161 {
2162         struct audit_buffer *audit_buf;
2163         u32 spi;
2164
2165         audit_buf = xfrm_audit_start("SA-replay-overflow");
2166         if (audit_buf == NULL)
2167                 return;
2168         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2169         /* don't record the sequence number because it's inherent in this kind
2170          * of audit message */
2171         spi = ntohl(x->id.spi);
2172         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2173         audit_log_end(audit_buf);
2174 }
2175 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2176
2177 void xfrm_audit_state_replay(struct xfrm_state *x,
2178                              struct sk_buff *skb, __be32 net_seq)
2179 {
2180         struct audit_buffer *audit_buf;
2181         u32 spi;
2182
2183         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2184         if (audit_buf == NULL)
2185                 return;
2186         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2187         spi = ntohl(x->id.spi);
2188         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2189                          spi, spi, ntohl(net_seq));
2190         audit_log_end(audit_buf);
2191 }
2192 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2193
2194 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2195 {
2196         struct audit_buffer *audit_buf;
2197
2198         audit_buf = xfrm_audit_start("SA-notfound");
2199         if (audit_buf == NULL)
2200                 return;
2201         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2202         audit_log_end(audit_buf);
2203 }
2204 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2205
2206 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2207                                __be32 net_spi, __be32 net_seq)
2208 {
2209         struct audit_buffer *audit_buf;
2210         u32 spi;
2211
2212         audit_buf = xfrm_audit_start("SA-notfound");
2213         if (audit_buf == NULL)
2214                 return;
2215         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2216         spi = ntohl(net_spi);
2217         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2218                          spi, spi, ntohl(net_seq));
2219         audit_log_end(audit_buf);
2220 }
2221 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2222
2223 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2224                               struct sk_buff *skb, u8 proto)
2225 {
2226         struct audit_buffer *audit_buf;
2227         __be32 net_spi;
2228         __be32 net_seq;
2229
2230         audit_buf = xfrm_audit_start("SA-icv-failure");
2231         if (audit_buf == NULL)
2232                 return;
2233         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2234         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2235                 u32 spi = ntohl(net_spi);
2236                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2237                                  spi, spi, ntohl(net_seq));
2238         }
2239         audit_log_end(audit_buf);
2240 }
2241 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2242 #endif /* CONFIG_AUDITSYSCALL */