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/interrupt.h>
  26 #include <linux/kernel.h>
  27
  28 #include "xfrm_hash.h"
  29
  30 /* Each xfrm_state may be linked to two tables:
  31
  32    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  33    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  34       destination/tunnel endpoint. (output)
  35  */
  36
  37 static DEFINE_SPINLOCK(xfrm_state_lock);
  38
  39 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  40 static unsigned int xfrm_state_genid;
  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 #ifdef CONFIG_AUDITSYSCALL
  46 static void xfrm_audit_state_replay(struct xfrm_state *x,
  47                                     struct sk_buff *skb, __be32 net_seq);
  48 #else
  49 #define xfrm_audit_state_replay(x, s, sq)       do { ; } while (0)
  50 #endif /* CONFIG_AUDITSYSCALL */
  51
  52 static inline unsigned int xfrm_dst_hash(struct net *net,
  53                                          xfrm_address_t *daddr,
  54                                          xfrm_address_t *saddr,
  55                                          u32 reqid,
  56                                          unsigned short family)
  57 {
  58         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  59 }
  60
  61 static inline unsigned int xfrm_src_hash(struct net *net,
  62                                          xfrm_address_t *daddr,
  63                                          xfrm_address_t *saddr,
  64                                          unsigned short family)
  65 {
  66         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  67 }
  68
  69 static inline unsigned int
  70 xfrm_spi_hash(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
  71 {
  72         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  73 }
  74
  75 static void xfrm_hash_transfer(struct hlist_head *list,
  76                                struct hlist_head *ndsttable,
  77                                struct hlist_head *nsrctable,
  78                                struct hlist_head *nspitable,
  79                                unsigned int nhashmask)
  80 {
  81         struct hlist_node *entry, *tmp;
  82         struct xfrm_state *x;
  83
  84         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
  85                 unsigned int h;
  86
  87                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  88                                     x->props.reqid, x->props.family,
  89                                     nhashmask);
  90                 hlist_add_head(&x->bydst, ndsttable+h);
  91
  92                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  93                                     x->props.family,
  94                                     nhashmask);
  95                 hlist_add_head(&x->bysrc, nsrctable+h);
  96
  97                 if (x->id.spi) {
  98                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  99                                             x->id.proto, x->props.family,
 100                                             nhashmask);
 101                         hlist_add_head(&x->byspi, nspitable+h);
 102                 }
 103         }
 104 }
 105
 106 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 107 {
 108         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 109 }
 110
 111 static DEFINE_MUTEX(hash_resize_mutex);
 112
 113 static void xfrm_hash_resize(struct work_struct *work)
 114 {
 115         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 116         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 117         unsigned long nsize, osize;
 118         unsigned int nhashmask, ohashmask;
 119         int i;
 120
 121         mutex_lock(&hash_resize_mutex);
 122
 123         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 124         ndst = xfrm_hash_alloc(nsize);
 125         if (!ndst)
 126                 goto out_unlock;
 127         nsrc = xfrm_hash_alloc(nsize);
 128         if (!nsrc) {
 129                 xfrm_hash_free(ndst, nsize);
 130                 goto out_unlock;
 131         }
 132         nspi = xfrm_hash_alloc(nsize);
 133         if (!nspi) {
 134                 xfrm_hash_free(ndst, nsize);
 135                 xfrm_hash_free(nsrc, nsize);
 136                 goto out_unlock;
 137         }
 138
 139         spin_lock_bh(&xfrm_state_lock);
 140
 141         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 142         for (i = net->xfrm.state_hmask; i >= 0; i--)
 143                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 144                                    nhashmask);
 145
 146         odst = net->xfrm.state_bydst;
 147         osrc = net->xfrm.state_bysrc;
 148         ospi = net->xfrm.state_byspi;
 149         ohashmask = net->xfrm.state_hmask;
 150
 151         net->xfrm.state_bydst = ndst;
 152         net->xfrm.state_bysrc = nsrc;
 153         net->xfrm.state_byspi = nspi;
 154         net->xfrm.state_hmask = nhashmask;
 155
 156         spin_unlock_bh(&xfrm_state_lock);
 157
 158         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 159         xfrm_hash_free(odst, osize);
 160         xfrm_hash_free(osrc, osize);
 161         xfrm_hash_free(ospi, osize);
 162
 163 out_unlock:
 164         mutex_unlock(&hash_resize_mutex);
 165 }
 166
 167 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
 168 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
 169
 170 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 171
 172 int __xfrm_state_delete(struct xfrm_state *x);
 173
 174 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 175 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
 176
 177 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 178 {
 179         struct xfrm_state_afinfo *afinfo;
 180         if (unlikely(family >= NPROTO))
 181                 return NULL;
 182         write_lock_bh(&xfrm_state_afinfo_lock);
 183         afinfo = xfrm_state_afinfo[family];
 184         if (unlikely(!afinfo))
 185                 write_unlock_bh(&xfrm_state_afinfo_lock);
 186         return afinfo;
 187 }
 188
 189 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 190         __releases(xfrm_state_afinfo_lock)
 191 {
 192         write_unlock_bh(&xfrm_state_afinfo_lock);
 193 }
 194
 195 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 196 {
 197         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 198         const struct xfrm_type **typemap;
 199         int err = 0;
 200
 201         if (unlikely(afinfo == NULL))
 202                 return -EAFNOSUPPORT;
 203         typemap = afinfo->type_map;
 204
 205         if (likely(typemap[type->proto] == NULL))
 206                 typemap[type->proto] = type;
 207         else
 208                 err = -EEXIST;
 209         xfrm_state_unlock_afinfo(afinfo);
 210         return err;
 211 }
 212 EXPORT_SYMBOL(xfrm_register_type);
 213
 214 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 215 {
 216         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 217         const struct xfrm_type **typemap;
 218         int err = 0;
 219
 220         if (unlikely(afinfo == NULL))
 221                 return -EAFNOSUPPORT;
 222         typemap = afinfo->type_map;
 223
 224         if (unlikely(typemap[type->proto] != type))
 225                 err = -ENOENT;
 226         else
 227                 typemap[type->proto] = NULL;
 228         xfrm_state_unlock_afinfo(afinfo);
 229         return err;
 230 }
 231 EXPORT_SYMBOL(xfrm_unregister_type);
 232
 233 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 234 {
 235         struct xfrm_state_afinfo *afinfo;
 236         const struct xfrm_type **typemap;
 237         const struct xfrm_type *type;
 238         int modload_attempted = 0;
 239
 240 retry:
 241         afinfo = xfrm_state_get_afinfo(family);
 242         if (unlikely(afinfo == NULL))
 243                 return NULL;
 244         typemap = afinfo->type_map;
 245
 246         type = typemap[proto];
 247         if (unlikely(type && !try_module_get(type->owner)))
 248                 type = NULL;
 249         if (!type && !modload_attempted) {
 250                 xfrm_state_put_afinfo(afinfo);
 251                 request_module("xfrm-type-%d-%d", family, proto);
 252                 modload_attempted = 1;
 253                 goto retry;
 254         }
 255
 256         xfrm_state_put_afinfo(afinfo);
 257         return type;
 258 }
 259
 260 static void xfrm_put_type(const struct xfrm_type *type)
 261 {
 262         module_put(type->owner);
 263 }
 264
 265 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 266 {
 267         struct xfrm_state_afinfo *afinfo;
 268         struct xfrm_mode **modemap;
 269         int err;
 270
 271         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 272                 return -EINVAL;
 273
 274         afinfo = xfrm_state_lock_afinfo(family);
 275         if (unlikely(afinfo == NULL))
 276                 return -EAFNOSUPPORT;
 277
 278         err = -EEXIST;
 279         modemap = afinfo->mode_map;
 280         if (modemap[mode->encap])
 281                 goto out;
 282
 283         err = -ENOENT;
 284         if (!try_module_get(afinfo->owner))
 285                 goto out;
 286
 287         mode->afinfo = afinfo;
 288         modemap[mode->encap] = mode;
 289         err = 0;
 290
 291 out:
 292         xfrm_state_unlock_afinfo(afinfo);
 293         return err;
 294 }
 295 EXPORT_SYMBOL(xfrm_register_mode);
 296
 297 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 298 {
 299         struct xfrm_state_afinfo *afinfo;
 300         struct xfrm_mode **modemap;
 301         int err;
 302
 303         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 304                 return -EINVAL;
 305
 306         afinfo = xfrm_state_lock_afinfo(family);
 307         if (unlikely(afinfo == NULL))
 308                 return -EAFNOSUPPORT;
 309
 310         err = -ENOENT;
 311         modemap = afinfo->mode_map;
 312         if (likely(modemap[mode->encap] == mode)) {
 313                 modemap[mode->encap] = NULL;
 314                 module_put(mode->afinfo->owner);
 315                 err = 0;
 316         }
 317
 318         xfrm_state_unlock_afinfo(afinfo);
 319         return err;
 320 }
 321 EXPORT_SYMBOL(xfrm_unregister_mode);
 322
 323 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 324 {
 325         struct xfrm_state_afinfo *afinfo;
 326         struct xfrm_mode *mode;
 327         int modload_attempted = 0;
 328
 329         if (unlikely(encap >= XFRM_MODE_MAX))
 330                 return NULL;
 331
 332 retry:
 333         afinfo = xfrm_state_get_afinfo(family);
 334         if (unlikely(afinfo == NULL))
 335                 return NULL;
 336
 337         mode = afinfo->mode_map[encap];
 338         if (unlikely(mode && !try_module_get(mode->owner)))
 339                 mode = NULL;
 340         if (!mode && !modload_attempted) {
 341                 xfrm_state_put_afinfo(afinfo);
 342                 request_module("xfrm-mode-%d-%d", family, encap);
 343                 modload_attempted = 1;
 344                 goto retry;
 345         }
 346
 347         xfrm_state_put_afinfo(afinfo);
 348         return mode;
 349 }
 350
 351 static void xfrm_put_mode(struct xfrm_mode *mode)
 352 {
 353         module_put(mode->owner);
 354 }
 355
 356 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 357 {
 358         tasklet_hrtimer_cancel(&x->mtimer);
 359         del_timer_sync(&x->rtimer);
 360         kfree(x->aalg);
 361         kfree(x->ealg);
 362         kfree(x->calg);
 363         kfree(x->encap);
 364         kfree(x->coaddr);
 365         if (x->inner_mode)
 366                 xfrm_put_mode(x->inner_mode);
 367         if (x->inner_mode_iaf)
 368                 xfrm_put_mode(x->inner_mode_iaf);
 369         if (x->outer_mode)
 370                 xfrm_put_mode(x->outer_mode);
 371         if (x->type) {
 372                 x->type->destructor(x);
 373                 xfrm_put_type(x->type);
 374         }
 375         security_xfrm_state_free(x);
 376         kfree(x);
 377 }
 378
 379 static void xfrm_state_gc_task(struct work_struct *work)
 380 {
 381         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 382         struct xfrm_state *x;
 383         struct hlist_node *entry, *tmp;
 384         struct hlist_head gc_list;
 385
 386         spin_lock_bh(&xfrm_state_gc_lock);
 387         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 388         spin_unlock_bh(&xfrm_state_gc_lock);
 389
 390         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 391                 xfrm_state_gc_destroy(x);
 392
 393         wake_up(&net->xfrm.km_waitq);
 394 }
 395
 396 static inline unsigned long make_jiffies(long secs)
 397 {
 398         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 399                 return MAX_SCHEDULE_TIMEOUT-1;
 400         else
 401                 return secs*HZ;
 402 }
 403
 404 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 405 {
 406         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 407         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 408         struct net *net = xs_net(x);
 409         unsigned long now = get_seconds();
 410         long next = LONG_MAX;
 411         int warn = 0;
 412         int err = 0;
 413
 414         spin_lock(&x->lock);
 415         if (x->km.state == XFRM_STATE_DEAD)
 416                 goto out;
 417         if (x->km.state == XFRM_STATE_EXPIRED)
 418                 goto expired;
 419         if (x->lft.hard_add_expires_seconds) {
 420                 long tmo = x->lft.hard_add_expires_seconds +
 421                         x->curlft.add_time - now;
 422                 if (tmo <= 0)
 423                         goto expired;
 424                 if (tmo < next)
 425                         next = tmo;
 426         }
 427         if (x->lft.hard_use_expires_seconds) {
 428                 long tmo = x->lft.hard_use_expires_seconds +
 429                         (x->curlft.use_time ? : now) - now;
 430                 if (tmo <= 0)
 431                         goto expired;
 432                 if (tmo < next)
 433                         next = tmo;
 434         }
 435         if (x->km.dying)
 436                 goto resched;
 437         if (x->lft.soft_add_expires_seconds) {
 438                 long tmo = x->lft.soft_add_expires_seconds +
 439                         x->curlft.add_time - now;
 440                 if (tmo <= 0)
 441                         warn = 1;
 442                 else if (tmo < next)
 443                         next = tmo;
 444         }
 445         if (x->lft.soft_use_expires_seconds) {
 446                 long tmo = x->lft.soft_use_expires_seconds +
 447                         (x->curlft.use_time ? : now) - now;
 448                 if (tmo <= 0)
 449                         warn = 1;
 450                 else if (tmo < next)
 451                         next = tmo;
 452         }
 453
 454         x->km.dying = warn;
 455         if (warn)
 456                 km_state_expired(x, 0, 0);
 457 resched:
 458         if (next != LONG_MAX){
 459                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 460         }
 461
 462         goto out;
 463
 464 expired:
 465         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 466                 x->km.state = XFRM_STATE_EXPIRED;
 467                 wake_up(&net->xfrm.km_waitq);
 468                 next = 2;
 469                 goto resched;
 470         }
 471
 472         err = __xfrm_state_delete(x);
 473         if (!err && x->id.spi)
 474                 km_state_expired(x, 1, 0);
 475
 476         xfrm_audit_state_delete(x, err ? 0 : 1,
 477                                 audit_get_loginuid(current),
 478                                 audit_get_sessionid(current), 0);
 479
 480 out:
 481         spin_unlock(&x->lock);
 482         return HRTIMER_NORESTART;
 483 }
 484
 485 static void xfrm_replay_timer_handler(unsigned long data);
 486
 487 struct xfrm_state *xfrm_state_alloc(struct net *net)
 488 {
 489         struct xfrm_state *x;
 490
 491         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 492
 493         if (x) {
 494                 write_pnet(&x->xs_net, net);
 495                 atomic_set(&x->refcnt, 1);
 496                 atomic_set(&x->tunnel_users, 0);
 497                 INIT_LIST_HEAD(&x->km.all);
 498                 INIT_HLIST_NODE(&x->bydst);
 499                 INIT_HLIST_NODE(&x->bysrc);
 500                 INIT_HLIST_NODE(&x->byspi);
 501                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 502                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 503                                 (unsigned long)x);
 504                 x->curlft.add_time = get_seconds();
 505                 x->lft.soft_byte_limit = XFRM_INF;
 506                 x->lft.soft_packet_limit = XFRM_INF;
 507                 x->lft.hard_byte_limit = XFRM_INF;
 508                 x->lft.hard_packet_limit = XFRM_INF;
 509                 x->replay_maxage = 0;
 510                 x->replay_maxdiff = 0;
 511                 x->inner_mode = NULL;
 512                 x->inner_mode_iaf = NULL;
 513                 spin_lock_init(&x->lock);
 514         }
 515         return x;
 516 }
 517 EXPORT_SYMBOL(xfrm_state_alloc);
 518
 519 void __xfrm_state_destroy(struct xfrm_state *x)
 520 {
 521         struct net *net = xs_net(x);
 522
 523         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 524
 525         spin_lock_bh(&xfrm_state_gc_lock);
 526         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 527         spin_unlock_bh(&xfrm_state_gc_lock);
 528         schedule_work(&net->xfrm.state_gc_work);
 529 }
 530 EXPORT_SYMBOL(__xfrm_state_destroy);
 531
 532 int __xfrm_state_delete(struct xfrm_state *x)
 533 {
 534         struct net *net = xs_net(x);
 535         int err = -ESRCH;
 536
 537         if (x->km.state != XFRM_STATE_DEAD) {
 538                 x->km.state = XFRM_STATE_DEAD;
 539                 spin_lock(&xfrm_state_lock);
 540                 list_del(&x->km.all);
 541                 hlist_del(&x->bydst);
 542                 hlist_del(&x->bysrc);
 543                 if (x->id.spi)
 544                         hlist_del(&x->byspi);
 545                 net->xfrm.state_num--;
 546                 spin_unlock(&xfrm_state_lock);
 547
 548                 /* All xfrm_state objects are created by xfrm_state_alloc.
 549                  * The xfrm_state_alloc call gives a reference, and that
 550                  * is what we are dropping here.
 551                  */
 552                 xfrm_state_put(x);
 553                 err = 0;
 554         }
 555
 556         return err;
 557 }
 558 EXPORT_SYMBOL(__xfrm_state_delete);
 559
 560 int xfrm_state_delete(struct xfrm_state *x)
 561 {
 562         int err;
 563
 564         spin_lock_bh(&x->lock);
 565         err = __xfrm_state_delete(x);
 566         spin_unlock_bh(&x->lock);
 567
 568         return err;
 569 }
 570 EXPORT_SYMBOL(xfrm_state_delete);
 571
 572 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 573 static inline int
 574 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 575 {
 576         int i, err = 0;
 577
 578         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 579                 struct hlist_node *entry;
 580                 struct xfrm_state *x;
 581
 582                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 583                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 584                            (err = security_xfrm_state_delete(x)) != 0) {
 585                                 xfrm_audit_state_delete(x, 0,
 586                                                         audit_info->loginuid,
 587                                                         audit_info->sessionid,
 588                                                         audit_info->secid);
 589                                 return err;
 590                         }
 591                 }
 592         }
 593
 594         return err;
 595 }
 596 #else
 597 static inline int
 598 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 599 {
 600         return 0;
 601 }
 602 #endif
 603
 604 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 605 {
 606         int i, err = 0;
 607
 608         spin_lock_bh(&xfrm_state_lock);
 609         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 610         if (err)
 611                 goto out;
 612
 613         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 614                 struct hlist_node *entry;
 615                 struct xfrm_state *x;
 616 restart:
 617                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 618                         if (!xfrm_state_kern(x) &&
 619                             xfrm_id_proto_match(x->id.proto, proto)) {
 620                                 xfrm_state_hold(x);
 621                                 spin_unlock_bh(&xfrm_state_lock);
 622
 623                                 err = xfrm_state_delete(x);
 624                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 625                                                         audit_info->loginuid,
 626                                                         audit_info->sessionid,
 627                                                         audit_info->secid);
 628                                 xfrm_state_put(x);
 629
 630                                 spin_lock_bh(&xfrm_state_lock);
 631                                 goto restart;
 632                         }
 633                 }
 634         }
 635         err = 0;
 636
 637 out:
 638         spin_unlock_bh(&xfrm_state_lock);
 639         wake_up(&net->xfrm.km_waitq);
 640         return err;
 641 }
 642 EXPORT_SYMBOL(xfrm_state_flush);
 643
 644 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 645 {
 646         spin_lock_bh(&xfrm_state_lock);
 647         si->sadcnt = net->xfrm.state_num;
 648         si->sadhcnt = net->xfrm.state_hmask;
 649         si->sadhmcnt = xfrm_state_hashmax;
 650         spin_unlock_bh(&xfrm_state_lock);
 651 }
 652 EXPORT_SYMBOL(xfrm_sad_getinfo);
 653
 654 static int
 655 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
 656                   struct xfrm_tmpl *tmpl,
 657                   xfrm_address_t *daddr, xfrm_address_t *saddr,
 658                   unsigned short family)
 659 {
 660         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 661         if (!afinfo)
 662                 return -1;
 663         afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
 664         xfrm_state_put_afinfo(afinfo);
 665         return 0;
 666 }
 667
 668 static struct xfrm_state *__xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
 669 {
 670         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 671         struct xfrm_state *x;
 672         struct hlist_node *entry;
 673
 674         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 675                 if (x->props.family != family ||
 676                     x->id.spi       != spi ||
 677                     x->id.proto     != proto ||
 678                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 679                         continue;
 680
 681                 xfrm_state_hold(x);
 682                 return x;
 683         }
 684
 685         return NULL;
 686 }
 687
 688 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
 689 {
 690         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 691         struct xfrm_state *x;
 692         struct hlist_node *entry;
 693
 694         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 695                 if (x->props.family != family ||
 696                     x->id.proto     != proto ||
 697                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 698                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 699                         continue;
 700
 701                 xfrm_state_hold(x);
 702                 return x;
 703         }
 704
 705         return NULL;
 706 }
 707
 708 static inline struct xfrm_state *
 709 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 710 {
 711         struct net *net = xs_net(x);
 712
 713         if (use_spi)
 714                 return __xfrm_state_lookup(net, &x->id.daddr, x->id.spi,
 715                                            x->id.proto, family);
 716         else
 717                 return __xfrm_state_lookup_byaddr(net, &x->id.daddr,
 718                                                   &x->props.saddr,
 719                                                   x->id.proto, family);
 720 }
 721
 722 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 723 {
 724         if (have_hash_collision &&
 725             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 726             net->xfrm.state_num > net->xfrm.state_hmask)
 727                 schedule_work(&net->xfrm.state_hash_work);
 728 }
 729
 730 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 731                                struct flowi *fl, unsigned short family,
 732                                xfrm_address_t *daddr, xfrm_address_t *saddr,
 733                                struct xfrm_state **best, int *acq_in_progress,
 734                                int *error)
 735 {
 736         /* Resolution logic:
 737          * 1. There is a valid state with matching selector. Done.
 738          * 2. Valid state with inappropriate selector. Skip.
 739          *
 740          * Entering area of "sysdeps".
 741          *
 742          * 3. If state is not valid, selector is temporary, it selects
 743          *    only session which triggered previous resolution. Key
 744          *    manager will do something to install a state with proper
 745          *    selector.
 746          */
 747         if (x->km.state == XFRM_STATE_VALID) {
 748                 if ((x->sel.family &&
 749                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 750                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 751                         return;
 752
 753                 if (!*best ||
 754                     (*best)->km.dying > x->km.dying ||
 755                     ((*best)->km.dying == x->km.dying &&
 756                      (*best)->curlft.add_time < x->curlft.add_time))
 757                         *best = x;
 758         } else if (x->km.state == XFRM_STATE_ACQ) {
 759                 *acq_in_progress = 1;
 760         } else if (x->km.state == XFRM_STATE_ERROR ||
 761                    x->km.state == XFRM_STATE_EXPIRED) {
 762                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 763                     security_xfrm_state_pol_flow_match(x, pol, fl))
 764                         *error = -ESRCH;
 765         }
 766 }
 767
 768 struct xfrm_state *
 769 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
 770                 struct flowi *fl, struct xfrm_tmpl *tmpl,
 771                 struct xfrm_policy *pol, int *err,
 772                 unsigned short family)
 773 {
 774         static xfrm_address_t saddr_wildcard = { };
 775         struct net *net = xp_net(pol);
 776         unsigned int h, h_wildcard;
 777         struct hlist_node *entry;
 778         struct xfrm_state *x, *x0, *to_put;
 779         int acquire_in_progress = 0;
 780         int error = 0;
 781         struct xfrm_state *best = NULL;
 782
 783         to_put = NULL;
 784
 785         spin_lock_bh(&xfrm_state_lock);
 786         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
 787         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 788                 if (x->props.family == family &&
 789                     x->props.reqid == tmpl->reqid &&
 790                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 791                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 792                     tmpl->mode == x->props.mode &&
 793                     tmpl->id.proto == x->id.proto &&
 794                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 795                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 796                                            &best, &acquire_in_progress, &error);
 797         }
 798         if (best)
 799                 goto found;
 800
 801         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
 802         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 803                 if (x->props.family == family &&
 804                     x->props.reqid == tmpl->reqid &&
 805                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 806                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 807                     tmpl->mode == x->props.mode &&
 808                     tmpl->id.proto == x->id.proto &&
 809                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 810                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 811                                            &best, &acquire_in_progress, &error);
 812         }
 813
 814 found:
 815         x = best;
 816         if (!x && !error && !acquire_in_progress) {
 817                 if (tmpl->id.spi &&
 818                     (x0 = __xfrm_state_lookup(net, daddr, tmpl->id.spi,
 819                                               tmpl->id.proto, family)) != NULL) {
 820                         to_put = x0;
 821                         error = -EEXIST;
 822                         goto out;
 823                 }
 824                 x = xfrm_state_alloc(net);
 825                 if (x == NULL) {
 826                         error = -ENOMEM;
 827                         goto out;
 828                 }
 829                 /* Initialize temporary selector matching only
 830                  * to current session. */
 831                 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
 832
 833                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
 834                 if (error) {
 835                         x->km.state = XFRM_STATE_DEAD;
 836                         to_put = x;
 837                         x = NULL;
 838                         goto out;
 839                 }
 840
 841                 if (km_query(x, tmpl, pol) == 0) {
 842                         x->km.state = XFRM_STATE_ACQ;
 843                         list_add(&x->km.all, &net->xfrm.state_all);
 844                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 845                         h = xfrm_src_hash(net, daddr, saddr, family);
 846                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 847                         if (x->id.spi) {
 848                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
 849                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 850                         }
 851                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 852                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 853                         net->xfrm.state_num++;
 854                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 855                 } else {
 856                         x->km.state = XFRM_STATE_DEAD;
 857                         to_put = x;
 858                         x = NULL;
 859                         error = -ESRCH;
 860                 }
 861         }
 862 out:
 863         if (x)
 864                 xfrm_state_hold(x);
 865         else
 866                 *err = acquire_in_progress ? -EAGAIN : error;
 867         spin_unlock_bh(&xfrm_state_lock);
 868         if (to_put)
 869                 xfrm_state_put(to_put);
 870         return x;
 871 }
 872
 873 struct xfrm_state *
 874 xfrm_stateonly_find(struct net *net,
 875                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 876                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 877 {
 878         unsigned int h;
 879         struct xfrm_state *rx = NULL, *x = NULL;
 880         struct hlist_node *entry;
 881
 882         spin_lock(&xfrm_state_lock);
 883         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 884         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 885                 if (x->props.family == family &&
 886                     x->props.reqid == reqid &&
 887                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 888                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 889                     mode == x->props.mode &&
 890                     proto == x->id.proto &&
 891                     x->km.state == XFRM_STATE_VALID) {
 892                         rx = x;
 893                         break;
 894                 }
 895         }
 896
 897         if (rx)
 898                 xfrm_state_hold(rx);
 899         spin_unlock(&xfrm_state_lock);
 900
 901
 902         return rx;
 903 }
 904 EXPORT_SYMBOL(xfrm_stateonly_find);
 905
 906 static void __xfrm_state_insert(struct xfrm_state *x)
 907 {
 908         struct net *net = xs_net(x);
 909         unsigned int h;
 910
 911         x->genid = ++xfrm_state_genid;
 912
 913         list_add(&x->km.all, &net->xfrm.state_all);
 914
 915         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 916                           x->props.reqid, x->props.family);
 917         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 918
 919         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 920         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 921
 922         if (x->id.spi) {
 923                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 924                                   x->props.family);
 925
 926                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 927         }
 928
 929         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 930         if (x->replay_maxage)
 931                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 932
 933         wake_up(&net->xfrm.km_waitq);
 934
 935         net->xfrm.state_num++;
 936
 937         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 938 }
 939
 940 /* xfrm_state_lock is held */
 941 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 942 {
 943         struct net *net = xs_net(xnew);
 944         unsigned short family = xnew->props.family;
 945         u32 reqid = xnew->props.reqid;
 946         struct xfrm_state *x;
 947         struct hlist_node *entry;
 948         unsigned int h;
 949
 950         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 951         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 952                 if (x->props.family     == family &&
 953                     x->props.reqid      == reqid &&
 954                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 955                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 956                         x->genid = xfrm_state_genid;
 957         }
 958 }
 959
 960 void xfrm_state_insert(struct xfrm_state *x)
 961 {
 962         spin_lock_bh(&xfrm_state_lock);
 963         __xfrm_state_bump_genids(x);
 964         __xfrm_state_insert(x);
 965         spin_unlock_bh(&xfrm_state_lock);
 966 }
 967 EXPORT_SYMBOL(xfrm_state_insert);
 968
 969 /* xfrm_state_lock is held */
 970 static struct xfrm_state *__find_acq_core(struct net *net, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
 971 {
 972         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 973         struct hlist_node *entry;
 974         struct xfrm_state *x;
 975
 976         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 977                 if (x->props.reqid  != reqid ||
 978                     x->props.mode   != mode ||
 979                     x->props.family != family ||
 980                     x->km.state     != XFRM_STATE_ACQ ||
 981                     x->id.spi       != 0 ||
 982                     x->id.proto     != proto ||
 983                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 984                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 985                         continue;
 986
 987                 xfrm_state_hold(x);
 988                 return x;
 989         }
 990
 991         if (!create)
 992                 return NULL;
 993
 994         x = xfrm_state_alloc(net);
 995         if (likely(x)) {
 996                 switch (family) {
 997                 case AF_INET:
 998                         x->sel.daddr.a4 = daddr->a4;
 999                         x->sel.saddr.a4 = saddr->a4;
1000                         x->sel.prefixlen_d = 32;
1001                         x->sel.prefixlen_s = 32;
1002                         x->props.saddr.a4 = saddr->a4;
1003                         x->id.daddr.a4 = daddr->a4;
1004                         break;
1005
1006                 case AF_INET6:
1007                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1008                                        (struct in6_addr *)daddr);
1009                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1010                                        (struct in6_addr *)saddr);
1011                         x->sel.prefixlen_d = 128;
1012                         x->sel.prefixlen_s = 128;
1013                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1014                                        (struct in6_addr *)saddr);
1015                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1016                                        (struct in6_addr *)daddr);
1017                         break;
1018                 }
1019
1020                 x->km.state = XFRM_STATE_ACQ;
1021                 x->id.proto = proto;
1022                 x->props.family = family;
1023                 x->props.mode = mode;
1024                 x->props.reqid = reqid;
1025                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1026                 xfrm_state_hold(x);
1027                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1028                 list_add(&x->km.all, &net->xfrm.state_all);
1029                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1030                 h = xfrm_src_hash(net, daddr, saddr, family);
1031                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1032
1033                 net->xfrm.state_num++;
1034
1035                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1036         }
1037
1038         return x;
1039 }
1040
1041 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq);
1042
1043 int xfrm_state_add(struct xfrm_state *x)
1044 {
1045         struct net *net = xs_net(x);
1046         struct xfrm_state *x1, *to_put;
1047         int family;
1048         int err;
1049         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1050
1051         family = x->props.family;
1052
1053         to_put = NULL;
1054
1055         spin_lock_bh(&xfrm_state_lock);
1056
1057         x1 = __xfrm_state_locate(x, use_spi, family);
1058         if (x1) {
1059                 to_put = x1;
1060                 x1 = NULL;
1061                 err = -EEXIST;
1062                 goto out;
1063         }
1064
1065         if (use_spi && x->km.seq) {
1066                 x1 = __xfrm_find_acq_byseq(net, x->km.seq);
1067                 if (x1 && ((x1->id.proto != x->id.proto) ||
1068                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1069                         to_put = x1;
1070                         x1 = NULL;
1071                 }
1072         }
1073
1074         if (use_spi && !x1)
1075                 x1 = __find_acq_core(net, family, x->props.mode, x->props.reqid,
1076                                      x->id.proto,
1077                                      &x->id.daddr, &x->props.saddr, 0);
1078
1079         __xfrm_state_bump_genids(x);
1080         __xfrm_state_insert(x);
1081         err = 0;
1082
1083 out:
1084         spin_unlock_bh(&xfrm_state_lock);
1085
1086         if (x1) {
1087                 xfrm_state_delete(x1);
1088                 xfrm_state_put(x1);
1089         }
1090
1091         if (to_put)
1092                 xfrm_state_put(to_put);
1093
1094         return err;
1095 }
1096 EXPORT_SYMBOL(xfrm_state_add);
1097
1098 #ifdef CONFIG_XFRM_MIGRATE
1099 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1100 {
1101         struct net *net = xs_net(orig);
1102         int err = -ENOMEM;
1103         struct xfrm_state *x = xfrm_state_alloc(net);
1104         if (!x)
1105                 goto out;
1106
1107         memcpy(&x->id, &orig->id, sizeof(x->id));
1108         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1109         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1110         x->props.mode = orig->props.mode;
1111         x->props.replay_window = orig->props.replay_window;
1112         x->props.reqid = orig->props.reqid;
1113         x->props.family = orig->props.family;
1114         x->props.saddr = orig->props.saddr;
1115
1116         if (orig->aalg) {
1117                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1118                 if (!x->aalg)
1119                         goto error;
1120         }
1121         x->props.aalgo = orig->props.aalgo;
1122
1123         if (orig->ealg) {
1124                 x->ealg = xfrm_algo_clone(orig->ealg);
1125                 if (!x->ealg)
1126                         goto error;
1127         }
1128         x->props.ealgo = orig->props.ealgo;
1129
1130         if (orig->calg) {
1131                 x->calg = xfrm_algo_clone(orig->calg);
1132                 if (!x->calg)
1133                         goto error;
1134         }
1135         x->props.calgo = orig->props.calgo;
1136
1137         if (orig->encap) {
1138                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1139                 if (!x->encap)
1140                         goto error;
1141         }
1142
1143         if (orig->coaddr) {
1144                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1145                                     GFP_KERNEL);
1146                 if (!x->coaddr)
1147                         goto error;
1148         }
1149
1150         err = xfrm_init_state(x);
1151         if (err)
1152                 goto error;
1153
1154         x->props.flags = orig->props.flags;
1155
1156         x->curlft.add_time = orig->curlft.add_time;
1157         x->km.state = orig->km.state;
1158         x->km.seq = orig->km.seq;
1159
1160         return x;
1161
1162  error:
1163         xfrm_state_put(x);
1164 out:
1165         if (errp)
1166                 *errp = err;
1167         return NULL;
1168 }
1169
1170 /* xfrm_state_lock is held */
1171 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1172 {
1173         unsigned int h;
1174         struct xfrm_state *x;
1175         struct hlist_node *entry;
1176
1177         if (m->reqid) {
1178                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1179                                   m->reqid, m->old_family);
1180                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1181                         if (x->props.mode != m->mode ||
1182                             x->id.proto != m->proto)
1183                                 continue;
1184                         if (m->reqid && x->props.reqid != m->reqid)
1185                                 continue;
1186                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1187                                           m->old_family) ||
1188                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1189                                           m->old_family))
1190                                 continue;
1191                         xfrm_state_hold(x);
1192                         return x;
1193                 }
1194         } else {
1195                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1196                                   m->old_family);
1197                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1198                         if (x->props.mode != m->mode ||
1199                             x->id.proto != m->proto)
1200                                 continue;
1201                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1202                                           m->old_family) ||
1203                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1204                                           m->old_family))
1205                                 continue;
1206                         xfrm_state_hold(x);
1207                         return x;
1208                 }
1209         }
1210
1211         return NULL;
1212 }
1213 EXPORT_SYMBOL(xfrm_migrate_state_find);
1214
1215 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1216                                        struct xfrm_migrate *m)
1217 {
1218         struct xfrm_state *xc;
1219         int err;
1220
1221         xc = xfrm_state_clone(x, &err);
1222         if (!xc)
1223                 return NULL;
1224
1225         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1226         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1227
1228         /* add state */
1229         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1230                 /* a care is needed when the destination address of the
1231                    state is to be updated as it is a part of triplet */
1232                 xfrm_state_insert(xc);
1233         } else {
1234                 if ((err = xfrm_state_add(xc)) < 0)
1235                         goto error;
1236         }
1237
1238         return xc;
1239 error:
1240         kfree(xc);
1241         return NULL;
1242 }
1243 EXPORT_SYMBOL(xfrm_state_migrate);
1244 #endif
1245
1246 int xfrm_state_update(struct xfrm_state *x)
1247 {
1248         struct xfrm_state *x1, *to_put;
1249         int err;
1250         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1251
1252         to_put = NULL;
1253
1254         spin_lock_bh(&xfrm_state_lock);
1255         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1256
1257         err = -ESRCH;
1258         if (!x1)
1259                 goto out;
1260
1261         if (xfrm_state_kern(x1)) {
1262                 to_put = x1;
1263                 err = -EEXIST;
1264                 goto out;
1265         }
1266
1267         if (x1->km.state == XFRM_STATE_ACQ) {
1268                 __xfrm_state_insert(x);
1269                 x = NULL;
1270         }
1271         err = 0;
1272
1273 out:
1274         spin_unlock_bh(&xfrm_state_lock);
1275
1276         if (to_put)
1277                 xfrm_state_put(to_put);
1278
1279         if (err)
1280                 return err;
1281
1282         if (!x) {
1283                 xfrm_state_delete(x1);
1284                 xfrm_state_put(x1);
1285                 return 0;
1286         }
1287
1288         err = -EINVAL;
1289         spin_lock_bh(&x1->lock);
1290         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1291                 if (x->encap && x1->encap)
1292                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1293                 if (x->coaddr && x1->coaddr) {
1294                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1295                 }
1296                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1297                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1298                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1299                 x1->km.dying = 0;
1300
1301                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1302                 if (x1->curlft.use_time)
1303                         xfrm_state_check_expire(x1);
1304
1305                 err = 0;
1306         }
1307         spin_unlock_bh(&x1->lock);
1308
1309         xfrm_state_put(x1);
1310
1311         return err;
1312 }
1313 EXPORT_SYMBOL(xfrm_state_update);
1314
1315 int xfrm_state_check_expire(struct xfrm_state *x)
1316 {
1317         if (!x->curlft.use_time)
1318                 x->curlft.use_time = get_seconds();
1319
1320         if (x->km.state != XFRM_STATE_VALID)
1321                 return -EINVAL;
1322
1323         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1324             x->curlft.packets >= x->lft.hard_packet_limit) {
1325                 x->km.state = XFRM_STATE_EXPIRED;
1326                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1327                 return -EINVAL;
1328         }
1329
1330         if (!x->km.dying &&
1331             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1332              x->curlft.packets >= x->lft.soft_packet_limit)) {
1333                 x->km.dying = 1;
1334                 km_state_expired(x, 0, 0);
1335         }
1336         return 0;
1337 }
1338 EXPORT_SYMBOL(xfrm_state_check_expire);
1339
1340 struct xfrm_state *
1341 xfrm_state_lookup(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto,
1342                   unsigned short family)
1343 {
1344         struct xfrm_state *x;
1345
1346         spin_lock_bh(&xfrm_state_lock);
1347         x = __xfrm_state_lookup(net, daddr, spi, proto, family);
1348         spin_unlock_bh(&xfrm_state_lock);
1349         return x;
1350 }
1351 EXPORT_SYMBOL(xfrm_state_lookup);
1352
1353 struct xfrm_state *
1354 xfrm_state_lookup_byaddr(struct net *net,
1355                          xfrm_address_t *daddr, xfrm_address_t *saddr,
1356                          u8 proto, unsigned short family)
1357 {
1358         struct xfrm_state *x;
1359
1360         spin_lock_bh(&xfrm_state_lock);
1361         x = __xfrm_state_lookup_byaddr(net, daddr, saddr, proto, family);
1362         spin_unlock_bh(&xfrm_state_lock);
1363         return x;
1364 }
1365 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1366
1367 struct xfrm_state *
1368 xfrm_find_acq(struct net *net, u8 mode, u32 reqid, u8 proto,
1369               xfrm_address_t *daddr, xfrm_address_t *saddr,
1370               int create, unsigned short family)
1371 {
1372         struct xfrm_state *x;
1373
1374         spin_lock_bh(&xfrm_state_lock);
1375         x = __find_acq_core(net, family, mode, reqid, proto, daddr, saddr, create);
1376         spin_unlock_bh(&xfrm_state_lock);
1377
1378         return x;
1379 }
1380 EXPORT_SYMBOL(xfrm_find_acq);
1381
1382 #ifdef CONFIG_XFRM_SUB_POLICY
1383 int
1384 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1385                unsigned short family)
1386 {
1387         int err = 0;
1388         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1389         if (!afinfo)
1390                 return -EAFNOSUPPORT;
1391
1392         spin_lock_bh(&xfrm_state_lock);
1393         if (afinfo->tmpl_sort)
1394                 err = afinfo->tmpl_sort(dst, src, n);
1395         spin_unlock_bh(&xfrm_state_lock);
1396         xfrm_state_put_afinfo(afinfo);
1397         return err;
1398 }
1399 EXPORT_SYMBOL(xfrm_tmpl_sort);
1400
1401 int
1402 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1403                 unsigned short family)
1404 {
1405         int err = 0;
1406         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1407         if (!afinfo)
1408                 return -EAFNOSUPPORT;
1409
1410         spin_lock_bh(&xfrm_state_lock);
1411         if (afinfo->state_sort)
1412                 err = afinfo->state_sort(dst, src, n);
1413         spin_unlock_bh(&xfrm_state_lock);
1414         xfrm_state_put_afinfo(afinfo);
1415         return err;
1416 }
1417 EXPORT_SYMBOL(xfrm_state_sort);
1418 #endif
1419
1420 /* Silly enough, but I'm lazy to build resolution list */
1421
1422 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 seq)
1423 {
1424         int i;
1425
1426         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1427                 struct hlist_node *entry;
1428                 struct xfrm_state *x;
1429
1430                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1431                         if (x->km.seq == seq &&
1432                             x->km.state == XFRM_STATE_ACQ) {
1433                                 xfrm_state_hold(x);
1434                                 return x;
1435                         }
1436                 }
1437         }
1438         return NULL;
1439 }
1440
1441 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 seq)
1442 {
1443         struct xfrm_state *x;
1444
1445         spin_lock_bh(&xfrm_state_lock);
1446         x = __xfrm_find_acq_byseq(net, seq);
1447         spin_unlock_bh(&xfrm_state_lock);
1448         return x;
1449 }
1450 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1451
1452 u32 xfrm_get_acqseq(void)
1453 {
1454         u32 res;
1455         static atomic_t acqseq;
1456
1457         do {
1458                 res = atomic_inc_return(&acqseq);
1459         } while (!res);
1460
1461         return res;
1462 }
1463 EXPORT_SYMBOL(xfrm_get_acqseq);
1464
1465 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1466 {
1467         struct net *net = xs_net(x);
1468         unsigned int h;
1469         struct xfrm_state *x0;
1470         int err = -ENOENT;
1471         __be32 minspi = htonl(low);
1472         __be32 maxspi = htonl(high);
1473
1474         spin_lock_bh(&x->lock);
1475         if (x->km.state == XFRM_STATE_DEAD)
1476                 goto unlock;
1477
1478         err = 0;
1479         if (x->id.spi)
1480                 goto unlock;
1481
1482         err = -ENOENT;
1483
1484         if (minspi == maxspi) {
1485                 x0 = xfrm_state_lookup(net, &x->id.daddr, minspi, x->id.proto, x->props.family);
1486                 if (x0) {
1487                         xfrm_state_put(x0);
1488                         goto unlock;
1489                 }
1490                 x->id.spi = minspi;
1491         } else {
1492                 u32 spi = 0;
1493                 for (h=0; h<high-low+1; h++) {
1494                         spi = low + net_random()%(high-low+1);
1495                         x0 = xfrm_state_lookup(net, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1496                         if (x0 == NULL) {
1497                                 x->id.spi = htonl(spi);
1498                                 break;
1499                         }
1500                         xfrm_state_put(x0);
1501                 }
1502         }
1503         if (x->id.spi) {
1504                 spin_lock_bh(&xfrm_state_lock);
1505                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1506                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1507                 spin_unlock_bh(&xfrm_state_lock);
1508
1509                 err = 0;
1510         }
1511
1512 unlock:
1513         spin_unlock_bh(&x->lock);
1514
1515         return err;
1516 }
1517 EXPORT_SYMBOL(xfrm_alloc_spi);
1518
1519 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1520                     int (*func)(struct xfrm_state *, int, void*),
1521                     void *data)
1522 {
1523         struct xfrm_state *state;
1524         struct xfrm_state_walk *x;
1525         int err = 0;
1526
1527         if (walk->seq != 0 && list_empty(&walk->all))
1528                 return 0;
1529
1530         spin_lock_bh(&xfrm_state_lock);
1531         if (list_empty(&walk->all))
1532                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1533         else
1534                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1535         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1536                 if (x->state == XFRM_STATE_DEAD)
1537                         continue;
1538                 state = container_of(x, struct xfrm_state, km);
1539                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1540                         continue;
1541                 err = func(state, walk->seq, data);
1542                 if (err) {
1543                         list_move_tail(&walk->all, &x->all);
1544                         goto out;
1545                 }
1546                 walk->seq++;
1547         }
1548         if (walk->seq == 0) {
1549                 err = -ENOENT;
1550                 goto out;
1551         }
1552         list_del_init(&walk->all);
1553 out:
1554         spin_unlock_bh(&xfrm_state_lock);
1555         return err;
1556 }
1557 EXPORT_SYMBOL(xfrm_state_walk);
1558
1559 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1560 {
1561         INIT_LIST_HEAD(&walk->all);
1562         walk->proto = proto;
1563         walk->state = XFRM_STATE_DEAD;
1564         walk->seq = 0;
1565 }
1566 EXPORT_SYMBOL(xfrm_state_walk_init);
1567
1568 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1569 {
1570         if (list_empty(&walk->all))
1571                 return;
1572
1573         spin_lock_bh(&xfrm_state_lock);
1574         list_del(&walk->all);
1575         spin_unlock_bh(&xfrm_state_lock);
1576 }
1577 EXPORT_SYMBOL(xfrm_state_walk_done);
1578
1579
1580 void xfrm_replay_notify(struct xfrm_state *x, int event)
1581 {
1582         struct km_event c;
1583         /* we send notify messages in case
1584          *  1. we updated on of the sequence numbers, and the seqno difference
1585          *     is at least x->replay_maxdiff, in this case we also update the
1586          *     timeout of our timer function
1587          *  2. if x->replay_maxage has elapsed since last update,
1588          *     and there were changes
1589          *
1590          *  The state structure must be locked!
1591          */
1592
1593         switch (event) {
1594         case XFRM_REPLAY_UPDATE:
1595                 if (x->replay_maxdiff &&
1596                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1597                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1598                         if (x->xflags & XFRM_TIME_DEFER)
1599                                 event = XFRM_REPLAY_TIMEOUT;
1600                         else
1601                                 return;
1602                 }
1603
1604                 break;
1605
1606         case XFRM_REPLAY_TIMEOUT:
1607                 if ((x->replay.seq == x->preplay.seq) &&
1608                     (x->replay.bitmap == x->preplay.bitmap) &&
1609                     (x->replay.oseq == x->preplay.oseq)) {
1610                         x->xflags |= XFRM_TIME_DEFER;
1611                         return;
1612                 }
1613
1614                 break;
1615         }
1616
1617         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1618         c.event = XFRM_MSG_NEWAE;
1619         c.data.aevent = event;
1620         km_state_notify(x, &c);
1621
1622         if (x->replay_maxage &&
1623             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1624                 x->xflags &= ~XFRM_TIME_DEFER;
1625 }
1626
1627 static void xfrm_replay_timer_handler(unsigned long data)
1628 {
1629         struct xfrm_state *x = (struct xfrm_state*)data;
1630
1631         spin_lock(&x->lock);
1632
1633         if (x->km.state == XFRM_STATE_VALID) {
1634                 if (xfrm_aevent_is_on(xs_net(x)))
1635                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1636                 else
1637                         x->xflags |= XFRM_TIME_DEFER;
1638         }
1639
1640         spin_unlock(&x->lock);
1641 }
1642
1643 int xfrm_replay_check(struct xfrm_state *x,
1644                       struct sk_buff *skb, __be32 net_seq)
1645 {
1646         u32 diff;
1647         u32 seq = ntohl(net_seq);
1648
1649         if (unlikely(seq == 0))
1650                 goto err;
1651
1652         if (likely(seq > x->replay.seq))
1653                 return 0;
1654
1655         diff = x->replay.seq - seq;
1656         if (diff >= min_t(unsigned int, x->props.replay_window,
1657                           sizeof(x->replay.bitmap) * 8)) {
1658                 x->stats.replay_window++;
1659                 goto err;
1660         }
1661
1662         if (x->replay.bitmap & (1U << diff)) {
1663                 x->stats.replay++;
1664                 goto err;
1665         }
1666         return 0;
1667
1668 err:
1669         xfrm_audit_state_replay(x, skb, net_seq);
1670         return -EINVAL;
1671 }
1672
1673 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1674 {
1675         u32 diff;
1676         u32 seq = ntohl(net_seq);
1677
1678         if (seq > x->replay.seq) {
1679                 diff = seq - x->replay.seq;
1680                 if (diff < x->props.replay_window)
1681                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1682                 else
1683                         x->replay.bitmap = 1;
1684                 x->replay.seq = seq;
1685         } else {
1686                 diff = x->replay.seq - seq;
1687                 x->replay.bitmap |= (1U << diff);
1688         }
1689
1690         if (xfrm_aevent_is_on(xs_net(x)))
1691                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1692 }
1693
1694 static LIST_HEAD(xfrm_km_list);
1695 static DEFINE_RWLOCK(xfrm_km_lock);
1696
1697 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1698 {
1699         struct xfrm_mgr *km;
1700
1701         read_lock(&xfrm_km_lock);
1702         list_for_each_entry(km, &xfrm_km_list, list)
1703                 if (km->notify_policy)
1704                         km->notify_policy(xp, dir, c);
1705         read_unlock(&xfrm_km_lock);
1706 }
1707
1708 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1709 {
1710         struct xfrm_mgr *km;
1711         read_lock(&xfrm_km_lock);
1712         list_for_each_entry(km, &xfrm_km_list, list)
1713                 if (km->notify)
1714                         km->notify(x, c);
1715         read_unlock(&xfrm_km_lock);
1716 }
1717
1718 EXPORT_SYMBOL(km_policy_notify);
1719 EXPORT_SYMBOL(km_state_notify);
1720
1721 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1722 {
1723         struct net *net = xs_net(x);
1724         struct km_event c;
1725
1726         c.data.hard = hard;
1727         c.pid = pid;
1728         c.event = XFRM_MSG_EXPIRE;
1729         km_state_notify(x, &c);
1730
1731         if (hard)
1732                 wake_up(&net->xfrm.km_waitq);
1733 }
1734
1735 EXPORT_SYMBOL(km_state_expired);
1736 /*
1737  * We send to all registered managers regardless of failure
1738  * We are happy with one success
1739 */
1740 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1741 {
1742         int err = -EINVAL, acqret;
1743         struct xfrm_mgr *km;
1744
1745         read_lock(&xfrm_km_lock);
1746         list_for_each_entry(km, &xfrm_km_list, list) {
1747                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1748                 if (!acqret)
1749                         err = acqret;
1750         }
1751         read_unlock(&xfrm_km_lock);
1752         return err;
1753 }
1754 EXPORT_SYMBOL(km_query);
1755
1756 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1757 {
1758         int err = -EINVAL;
1759         struct xfrm_mgr *km;
1760
1761         read_lock(&xfrm_km_lock);
1762         list_for_each_entry(km, &xfrm_km_list, list) {
1763                 if (km->new_mapping)
1764                         err = km->new_mapping(x, ipaddr, sport);
1765                 if (!err)
1766                         break;
1767         }
1768         read_unlock(&xfrm_km_lock);
1769         return err;
1770 }
1771 EXPORT_SYMBOL(km_new_mapping);
1772
1773 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1774 {
1775         struct net *net = xp_net(pol);
1776         struct km_event c;
1777
1778         c.data.hard = hard;
1779         c.pid = pid;
1780         c.event = XFRM_MSG_POLEXPIRE;
1781         km_policy_notify(pol, dir, &c);
1782
1783         if (hard)
1784                 wake_up(&net->xfrm.km_waitq);
1785 }
1786 EXPORT_SYMBOL(km_policy_expired);
1787
1788 #ifdef CONFIG_XFRM_MIGRATE
1789 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1790                struct xfrm_migrate *m, int num_migrate,
1791                struct xfrm_kmaddress *k)
1792 {
1793         int err = -EINVAL;
1794         int ret;
1795         struct xfrm_mgr *km;
1796
1797         read_lock(&xfrm_km_lock);
1798         list_for_each_entry(km, &xfrm_km_list, list) {
1799                 if (km->migrate) {
1800                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1801                         if (!ret)
1802                                 err = ret;
1803                 }
1804         }
1805         read_unlock(&xfrm_km_lock);
1806         return err;
1807 }
1808 EXPORT_SYMBOL(km_migrate);
1809 #endif
1810
1811 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1812 {
1813         int err = -EINVAL;
1814         int ret;
1815         struct xfrm_mgr *km;
1816
1817         read_lock(&xfrm_km_lock);
1818         list_for_each_entry(km, &xfrm_km_list, list) {
1819                 if (km->report) {
1820                         ret = km->report(net, proto, sel, addr);
1821                         if (!ret)
1822                                 err = ret;
1823                 }
1824         }
1825         read_unlock(&xfrm_km_lock);
1826         return err;
1827 }
1828 EXPORT_SYMBOL(km_report);
1829
1830 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1831 {
1832         int err;
1833         u8 *data;
1834         struct xfrm_mgr *km;
1835         struct xfrm_policy *pol = NULL;
1836
1837         if (optlen <= 0 || optlen > PAGE_SIZE)
1838                 return -EMSGSIZE;
1839
1840         data = kmalloc(optlen, GFP_KERNEL);
1841         if (!data)
1842                 return -ENOMEM;
1843
1844         err = -EFAULT;
1845         if (copy_from_user(data, optval, optlen))
1846                 goto out;
1847
1848         err = -EINVAL;
1849         read_lock(&xfrm_km_lock);
1850         list_for_each_entry(km, &xfrm_km_list, list) {
1851                 pol = km->compile_policy(sk, optname, data,
1852                                          optlen, &err);
1853                 if (err >= 0)
1854                         break;
1855         }
1856         read_unlock(&xfrm_km_lock);
1857
1858         if (err >= 0) {
1859                 xfrm_sk_policy_insert(sk, err, pol);
1860                 xfrm_pol_put(pol);
1861                 err = 0;
1862         }
1863
1864 out:
1865         kfree(data);
1866         return err;
1867 }
1868 EXPORT_SYMBOL(xfrm_user_policy);
1869
1870 int xfrm_register_km(struct xfrm_mgr *km)
1871 {
1872         write_lock_bh(&xfrm_km_lock);
1873         list_add_tail(&km->list, &xfrm_km_list);
1874         write_unlock_bh(&xfrm_km_lock);
1875         return 0;
1876 }
1877 EXPORT_SYMBOL(xfrm_register_km);
1878
1879 int xfrm_unregister_km(struct xfrm_mgr *km)
1880 {
1881         write_lock_bh(&xfrm_km_lock);
1882         list_del(&km->list);
1883         write_unlock_bh(&xfrm_km_lock);
1884         return 0;
1885 }
1886 EXPORT_SYMBOL(xfrm_unregister_km);
1887
1888 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1889 {
1890         int err = 0;
1891         if (unlikely(afinfo == NULL))
1892                 return -EINVAL;
1893         if (unlikely(afinfo->family >= NPROTO))
1894                 return -EAFNOSUPPORT;
1895         write_lock_bh(&xfrm_state_afinfo_lock);
1896         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1897                 err = -ENOBUFS;
1898         else
1899                 xfrm_state_afinfo[afinfo->family] = afinfo;
1900         write_unlock_bh(&xfrm_state_afinfo_lock);
1901         return err;
1902 }
1903 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1904
1905 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1906 {
1907         int err = 0;
1908         if (unlikely(afinfo == NULL))
1909                 return -EINVAL;
1910         if (unlikely(afinfo->family >= NPROTO))
1911                 return -EAFNOSUPPORT;
1912         write_lock_bh(&xfrm_state_afinfo_lock);
1913         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1914                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1915                         err = -EINVAL;
1916                 else
1917                         xfrm_state_afinfo[afinfo->family] = NULL;
1918         }
1919         write_unlock_bh(&xfrm_state_afinfo_lock);
1920         return err;
1921 }
1922 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1923
1924 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1925 {
1926         struct xfrm_state_afinfo *afinfo;
1927         if (unlikely(family >= NPROTO))
1928                 return NULL;
1929         read_lock(&xfrm_state_afinfo_lock);
1930         afinfo = xfrm_state_afinfo[family];
1931         if (unlikely(!afinfo))
1932                 read_unlock(&xfrm_state_afinfo_lock);
1933         return afinfo;
1934 }
1935
1936 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1937         __releases(xfrm_state_afinfo_lock)
1938 {
1939         read_unlock(&xfrm_state_afinfo_lock);
1940 }
1941
1942 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1943 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1944 {
1945         if (x->tunnel) {
1946                 struct xfrm_state *t = x->tunnel;
1947
1948                 if (atomic_read(&t->tunnel_users) == 2)
1949                         xfrm_state_delete(t);
1950                 atomic_dec(&t->tunnel_users);
1951                 xfrm_state_put(t);
1952                 x->tunnel = NULL;
1953         }
1954 }
1955 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1956
1957 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1958 {
1959         int res;
1960
1961         spin_lock_bh(&x->lock);
1962         if (x->km.state == XFRM_STATE_VALID &&
1963             x->type && x->type->get_mtu)
1964                 res = x->type->get_mtu(x, mtu);
1965         else
1966                 res = mtu - x->props.header_len;
1967         spin_unlock_bh(&x->lock);
1968         return res;
1969 }
1970
1971 int xfrm_init_state(struct xfrm_state *x)
1972 {
1973         struct xfrm_state_afinfo *afinfo;
1974         struct xfrm_mode *inner_mode;
1975         int family = x->props.family;
1976         int err;
1977
1978         err = -EAFNOSUPPORT;
1979         afinfo = xfrm_state_get_afinfo(family);
1980         if (!afinfo)
1981                 goto error;
1982
1983         err = 0;
1984         if (afinfo->init_flags)
1985                 err = afinfo->init_flags(x);
1986
1987         xfrm_state_put_afinfo(afinfo);
1988
1989         if (err)
1990                 goto error;
1991
1992         err = -EPROTONOSUPPORT;
1993
1994         if (x->sel.family != AF_UNSPEC) {
1995                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
1996                 if (inner_mode == NULL)
1997                         goto error;
1998
1999                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2000                     family != x->sel.family) {
2001                         xfrm_put_mode(inner_mode);
2002                         goto error;
2003                 }
2004
2005                 x->inner_mode = inner_mode;
2006         } else {
2007                 struct xfrm_mode *inner_mode_iaf;
2008                 int iafamily = AF_INET;
2009
2010                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2011                 if (inner_mode == NULL)
2012                         goto error;
2013
2014                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2015                         xfrm_put_mode(inner_mode);
2016                         goto error;
2017                 }
2018                 x->inner_mode = inner_mode;
2019
2020                 if (x->props.family == AF_INET)
2021                         iafamily = AF_INET6;
2022
2023                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2024                 if (inner_mode_iaf) {
2025                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2026                                 x->inner_mode_iaf = inner_mode_iaf;
2027                         else
2028                                 xfrm_put_mode(inner_mode_iaf);
2029                 }
2030         }
2031
2032         x->type = xfrm_get_type(x->id.proto, family);
2033         if (x->type == NULL)
2034                 goto error;
2035
2036         err = x->type->init_state(x);
2037         if (err)
2038                 goto error;
2039
2040         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2041         if (x->outer_mode == NULL)
2042                 goto error;
2043
2044         x->km.state = XFRM_STATE_VALID;
2045
2046 error:
2047         return err;
2048 }
2049
2050 EXPORT_SYMBOL(xfrm_init_state);
2051
2052 int __net_init xfrm_state_init(struct net *net)
2053 {
2054         unsigned int sz;
2055
2056         INIT_LIST_HEAD(&net->xfrm.state_all);
2057
2058         sz = sizeof(struct hlist_head) * 8;
2059
2060         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2061         if (!net->xfrm.state_bydst)
2062                 goto out_bydst;
2063         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2064         if (!net->xfrm.state_bysrc)
2065                 goto out_bysrc;
2066         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2067         if (!net->xfrm.state_byspi)
2068                 goto out_byspi;
2069         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2070
2071         net->xfrm.state_num = 0;
2072         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2073         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2074         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2075         init_waitqueue_head(&net->xfrm.km_waitq);
2076         return 0;
2077
2078 out_byspi:
2079         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2080 out_bysrc:
2081         xfrm_hash_free(net->xfrm.state_bydst, sz);
2082 out_bydst:
2083         return -ENOMEM;
2084 }
2085
2086 void xfrm_state_fini(struct net *net)
2087 {
2088         struct xfrm_audit audit_info;
2089         unsigned int sz;
2090
2091         flush_work(&net->xfrm.state_hash_work);
2092         audit_info.loginuid = -1;
2093         audit_info.sessionid = -1;
2094         audit_info.secid = 0;
2095         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2096         flush_work(&net->xfrm.state_gc_work);
2097
2098         WARN_ON(!list_empty(&net->xfrm.state_all));
2099
2100         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2101         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2102         xfrm_hash_free(net->xfrm.state_byspi, sz);
2103         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2104         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2105         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2106         xfrm_hash_free(net->xfrm.state_bydst, sz);
2107 }
2108
2109 #ifdef CONFIG_AUDITSYSCALL
2110 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2111                                      struct audit_buffer *audit_buf)
2112 {
2113         struct xfrm_sec_ctx *ctx = x->security;
2114         u32 spi = ntohl(x->id.spi);
2115
2116         if (ctx)
2117                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2118                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2119
2120         switch(x->props.family) {
2121         case AF_INET:
2122                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2123                                  &x->props.saddr.a4, &x->id.daddr.a4);
2124                 break;
2125         case AF_INET6:
2126                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2127                                  x->props.saddr.a6, x->id.daddr.a6);
2128                 break;
2129         }
2130
2131         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2132 }
2133
2134 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2135                                       struct audit_buffer *audit_buf)
2136 {
2137         struct iphdr *iph4;
2138         struct ipv6hdr *iph6;
2139
2140         switch (family) {
2141         case AF_INET:
2142                 iph4 = ip_hdr(skb);
2143                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2144                                  &iph4->saddr, &iph4->daddr);
2145                 break;
2146         case AF_INET6:
2147                 iph6 = ipv6_hdr(skb);
2148                 audit_log_format(audit_buf,
2149                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2150                                  &iph6->saddr,&iph6->daddr,
2151                                  iph6->flow_lbl[0] & 0x0f,
2152                                  iph6->flow_lbl[1],
2153                                  iph6->flow_lbl[2]);
2154                 break;
2155         }
2156 }
2157
2158 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2159                           uid_t auid, u32 sessionid, u32 secid)
2160 {
2161         struct audit_buffer *audit_buf;
2162
2163         audit_buf = xfrm_audit_start("SAD-add");
2164         if (audit_buf == NULL)
2165                 return;
2166         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2167         xfrm_audit_helper_sainfo(x, audit_buf);
2168         audit_log_format(audit_buf, " res=%u", result);
2169         audit_log_end(audit_buf);
2170 }
2171 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2172
2173 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2174                              uid_t auid, u32 sessionid, u32 secid)
2175 {
2176         struct audit_buffer *audit_buf;
2177
2178         audit_buf = xfrm_audit_start("SAD-delete");
2179         if (audit_buf == NULL)
2180                 return;
2181         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2182         xfrm_audit_helper_sainfo(x, audit_buf);
2183         audit_log_format(audit_buf, " res=%u", result);
2184         audit_log_end(audit_buf);
2185 }
2186 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2187
2188 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2189                                       struct sk_buff *skb)
2190 {
2191         struct audit_buffer *audit_buf;
2192         u32 spi;
2193
2194         audit_buf = xfrm_audit_start("SA-replay-overflow");
2195         if (audit_buf == NULL)
2196                 return;
2197         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2198         /* don't record the sequence number because it's inherent in this kind
2199          * of audit message */
2200         spi = ntohl(x->id.spi);
2201         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2202         audit_log_end(audit_buf);
2203 }
2204 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2205
2206 static void xfrm_audit_state_replay(struct xfrm_state *x,
2207                              struct sk_buff *skb, __be32 net_seq)
2208 {
2209         struct audit_buffer *audit_buf;
2210         u32 spi;
2211
2212         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2213         if (audit_buf == NULL)
2214                 return;
2215         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2216         spi = ntohl(x->id.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
2222 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2223 {
2224         struct audit_buffer *audit_buf;
2225
2226         audit_buf = xfrm_audit_start("SA-notfound");
2227         if (audit_buf == NULL)
2228                 return;
2229         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2230         audit_log_end(audit_buf);
2231 }
2232 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2233
2234 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2235                                __be32 net_spi, __be32 net_seq)
2236 {
2237         struct audit_buffer *audit_buf;
2238         u32 spi;
2239
2240         audit_buf = xfrm_audit_start("SA-notfound");
2241         if (audit_buf == NULL)
2242                 return;
2243         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2244         spi = ntohl(net_spi);
2245         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2246                          spi, spi, ntohl(net_seq));
2247         audit_log_end(audit_buf);
2248 }
2249 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2250
2251 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2252                               struct sk_buff *skb, u8 proto)
2253 {
2254         struct audit_buffer *audit_buf;
2255         __be32 net_spi;
2256         __be32 net_seq;
2257
2258         audit_buf = xfrm_audit_start("SA-icv-failure");
2259         if (audit_buf == NULL)
2260                 return;
2261         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2262         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2263                 u32 spi = ntohl(net_spi);
2264                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2265                                  spi, spi, ntohl(net_seq));
2266         }
2267         audit_log_end(audit_buf);
2268 }
2269 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2270 #endif /* CONFIG_AUDITSYSCALL */