SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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 #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, cnt = 0;
 607
 608         spin_lock_bh(&xfrm_state_lock);
 609         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 610         if (err)
 611                 goto out;
 612
 613         err = -ESRCH;
 614         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 615                 struct hlist_node *entry;
 616                 struct xfrm_state *x;
 617 restart:
 618                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 619                         if (!xfrm_state_kern(x) &&
 620                             xfrm_id_proto_match(x->id.proto, proto)) {
 621                                 xfrm_state_hold(x);
 622                                 spin_unlock_bh(&xfrm_state_lock);
 623
 624                                 err = xfrm_state_delete(x);
 625                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 626                                                         audit_info->loginuid,
 627                                                         audit_info->sessionid,
 628                                                         audit_info->secid);
 629                                 xfrm_state_put(x);
 630                                 if (!err)
 631                                         cnt++;
 632
 633                                 spin_lock_bh(&xfrm_state_lock);
 634                                 goto restart;
 635                         }
 636                 }
 637         }
 638         if (cnt)
 639                 err = 0;
 640
 641 out:
 642         spin_unlock_bh(&xfrm_state_lock);
 643         wake_up(&net->xfrm.km_waitq);
 644         return err;
 645 }
 646 EXPORT_SYMBOL(xfrm_state_flush);
 647
 648 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 649 {
 650         spin_lock_bh(&xfrm_state_lock);
 651         si->sadcnt = net->xfrm.state_num;
 652         si->sadhcnt = net->xfrm.state_hmask;
 653         si->sadhmcnt = xfrm_state_hashmax;
 654         spin_unlock_bh(&xfrm_state_lock);
 655 }
 656 EXPORT_SYMBOL(xfrm_sad_getinfo);
 657
 658 static int
 659 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
 660                   struct xfrm_tmpl *tmpl,
 661                   xfrm_address_t *daddr, xfrm_address_t *saddr,
 662                   unsigned short family)
 663 {
 664         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 665         if (!afinfo)
 666                 return -1;
 667         afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
 668         xfrm_state_put_afinfo(afinfo);
 669         return 0;
 670 }
 671
 672 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
 673 {
 674         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 675         struct xfrm_state *x;
 676         struct hlist_node *entry;
 677
 678         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 679                 if (x->props.family != family ||
 680                     x->id.spi       != spi ||
 681                     x->id.proto     != proto ||
 682                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 683                         continue;
 684
 685                 if ((mark & x->mark.m) != x->mark.v)
 686                         continue;
 687                 xfrm_state_hold(x);
 688                 return x;
 689         }
 690
 691         return NULL;
 692 }
 693
 694 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark, xfrm_address_t *daddr, xfrm_address_t *saddr, u8 proto, unsigned short family)
 695 {
 696         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 697         struct xfrm_state *x;
 698         struct hlist_node *entry;
 699
 700         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 701                 if (x->props.family != family ||
 702                     x->id.proto     != proto ||
 703                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 704                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 705                         continue;
 706
 707                 if ((mark & x->mark.m) != x->mark.v)
 708                         continue;
 709                 xfrm_state_hold(x);
 710                 return x;
 711         }
 712
 713         return NULL;
 714 }
 715
 716 static inline struct xfrm_state *
 717 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 718 {
 719         struct net *net = xs_net(x);
 720         u32 mark = x->mark.v & x->mark.m;
 721
 722         if (use_spi)
 723                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 724                                            x->id.spi, x->id.proto, family);
 725         else
 726                 return __xfrm_state_lookup_byaddr(net, mark,
 727                                                   &x->id.daddr,
 728                                                   &x->props.saddr,
 729                                                   x->id.proto, family);
 730 }
 731
 732 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 733 {
 734         if (have_hash_collision &&
 735             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 736             net->xfrm.state_num > net->xfrm.state_hmask)
 737                 schedule_work(&net->xfrm.state_hash_work);
 738 }
 739
 740 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 741                                struct flowi *fl, unsigned short family,
 742                                xfrm_address_t *daddr, xfrm_address_t *saddr,
 743                                struct xfrm_state **best, int *acq_in_progress,
 744                                int *error)
 745 {
 746         /* Resolution logic:
 747          * 1. There is a valid state with matching selector. Done.
 748          * 2. Valid state with inappropriate selector. Skip.
 749          *
 750          * Entering area of "sysdeps".
 751          *
 752          * 3. If state is not valid, selector is temporary, it selects
 753          *    only session which triggered previous resolution. Key
 754          *    manager will do something to install a state with proper
 755          *    selector.
 756          */
 757         if (x->km.state == XFRM_STATE_VALID) {
 758                 if ((x->sel.family &&
 759                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 760                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 761                         return;
 762
 763                 if (!*best ||
 764                     (*best)->km.dying > x->km.dying ||
 765                     ((*best)->km.dying == x->km.dying &&
 766                      (*best)->curlft.add_time < x->curlft.add_time))
 767                         *best = x;
 768         } else if (x->km.state == XFRM_STATE_ACQ) {
 769                 *acq_in_progress = 1;
 770         } else if (x->km.state == XFRM_STATE_ERROR ||
 771                    x->km.state == XFRM_STATE_EXPIRED) {
 772                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 773                     security_xfrm_state_pol_flow_match(x, pol, fl))
 774                         *error = -ESRCH;
 775         }
 776 }
 777
 778 struct xfrm_state *
 779 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
 780                 struct flowi *fl, struct xfrm_tmpl *tmpl,
 781                 struct xfrm_policy *pol, int *err,
 782                 unsigned short family)
 783 {
 784         static xfrm_address_t saddr_wildcard = { };
 785         struct net *net = xp_net(pol);
 786         unsigned int h, h_wildcard;
 787         struct hlist_node *entry;
 788         struct xfrm_state *x, *x0, *to_put;
 789         int acquire_in_progress = 0;
 790         int error = 0;
 791         struct xfrm_state *best = NULL;
 792         u32 mark = pol->mark.v & pol->mark.m;
 793
 794         to_put = NULL;
 795
 796         spin_lock_bh(&xfrm_state_lock);
 797         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
 798         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 799                 if (x->props.family == family &&
 800                     x->props.reqid == tmpl->reqid &&
 801                     (mark & x->mark.m) == x->mark.v &&
 802                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 803                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 804                     tmpl->mode == x->props.mode &&
 805                     tmpl->id.proto == x->id.proto &&
 806                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 807                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 808                                            &best, &acquire_in_progress, &error);
 809         }
 810         if (best)
 811                 goto found;
 812
 813         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
 814         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 815                 if (x->props.family == family &&
 816                     x->props.reqid == tmpl->reqid &&
 817                     (mark & x->mark.m) == x->mark.v &&
 818                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 819                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 820                     tmpl->mode == x->props.mode &&
 821                     tmpl->id.proto == x->id.proto &&
 822                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 823                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 824                                            &best, &acquire_in_progress, &error);
 825         }
 826
 827 found:
 828         x = best;
 829         if (!x && !error && !acquire_in_progress) {
 830                 if (tmpl->id.spi &&
 831                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 832                                               tmpl->id.proto, family)) != NULL) {
 833                         to_put = x0;
 834                         error = -EEXIST;
 835                         goto out;
 836                 }
 837                 x = xfrm_state_alloc(net);
 838                 if (x == NULL) {
 839                         error = -ENOMEM;
 840                         goto out;
 841                 }
 842                 /* Initialize temporary selector matching only
 843                  * to current session. */
 844                 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
 845                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 846
 847                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
 848                 if (error) {
 849                         x->km.state = XFRM_STATE_DEAD;
 850                         to_put = x;
 851                         x = NULL;
 852                         goto out;
 853                 }
 854
 855                 if (km_query(x, tmpl, pol) == 0) {
 856                         x->km.state = XFRM_STATE_ACQ;
 857                         list_add(&x->km.all, &net->xfrm.state_all);
 858                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 859                         h = xfrm_src_hash(net, daddr, saddr, family);
 860                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 861                         if (x->id.spi) {
 862                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
 863                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 864                         }
 865                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 866                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 867                         net->xfrm.state_num++;
 868                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 869                 } else {
 870                         x->km.state = XFRM_STATE_DEAD;
 871                         to_put = x;
 872                         x = NULL;
 873                         error = -ESRCH;
 874                 }
 875         }
 876 out:
 877         if (x)
 878                 xfrm_state_hold(x);
 879         else
 880                 *err = acquire_in_progress ? -EAGAIN : error;
 881         spin_unlock_bh(&xfrm_state_lock);
 882         if (to_put)
 883                 xfrm_state_put(to_put);
 884         return x;
 885 }
 886
 887 struct xfrm_state *
 888 xfrm_stateonly_find(struct net *net, u32 mark,
 889                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 890                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 891 {
 892         unsigned int h;
 893         struct xfrm_state *rx = NULL, *x = NULL;
 894         struct hlist_node *entry;
 895
 896         spin_lock(&xfrm_state_lock);
 897         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 898         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 899                 if (x->props.family == family &&
 900                     x->props.reqid == reqid &&
 901                     (mark & x->mark.m) == x->mark.v &&
 902                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 903                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 904                     mode == x->props.mode &&
 905                     proto == x->id.proto &&
 906                     x->km.state == XFRM_STATE_VALID) {
 907                         rx = x;
 908                         break;
 909                 }
 910         }
 911
 912         if (rx)
 913                 xfrm_state_hold(rx);
 914         spin_unlock(&xfrm_state_lock);
 915
 916
 917         return rx;
 918 }
 919 EXPORT_SYMBOL(xfrm_stateonly_find);
 920
 921 static void __xfrm_state_insert(struct xfrm_state *x)
 922 {
 923         struct net *net = xs_net(x);
 924         unsigned int h;
 925
 926         list_add(&x->km.all, &net->xfrm.state_all);
 927
 928         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 929                           x->props.reqid, x->props.family);
 930         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 931
 932         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 933         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 934
 935         if (x->id.spi) {
 936                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 937                                   x->props.family);
 938
 939                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 940         }
 941
 942         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 943         if (x->replay_maxage)
 944                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 945
 946         wake_up(&net->xfrm.km_waitq);
 947
 948         net->xfrm.state_num++;
 949
 950         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 951 }
 952
 953 /* xfrm_state_lock is held */
 954 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 955 {
 956         struct net *net = xs_net(xnew);
 957         unsigned short family = xnew->props.family;
 958         u32 reqid = xnew->props.reqid;
 959         struct xfrm_state *x;
 960         struct hlist_node *entry;
 961         unsigned int h;
 962         u32 mark = xnew->mark.v & xnew->mark.m;
 963
 964         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 965         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 966                 if (x->props.family     == family &&
 967                     x->props.reqid      == reqid &&
 968                     (mark & x->mark.m) == x->mark.v &&
 969                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 970                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 971                         x->genid++;
 972         }
 973 }
 974
 975 void xfrm_state_insert(struct xfrm_state *x)
 976 {
 977         spin_lock_bh(&xfrm_state_lock);
 978         __xfrm_state_bump_genids(x);
 979         __xfrm_state_insert(x);
 980         spin_unlock_bh(&xfrm_state_lock);
 981 }
 982 EXPORT_SYMBOL(xfrm_state_insert);
 983
 984 /* xfrm_state_lock is held */
 985 static struct xfrm_state *__find_acq_core(struct net *net, struct xfrm_mark *m, unsigned short family, u8 mode, u32 reqid, u8 proto, xfrm_address_t *daddr, xfrm_address_t *saddr, int create)
 986 {
 987         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 988         struct hlist_node *entry;
 989         struct xfrm_state *x;
 990         u32 mark = m->v & m->m;
 991
 992         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 993                 if (x->props.reqid  != reqid ||
 994                     x->props.mode   != mode ||
 995                     x->props.family != family ||
 996                     x->km.state     != XFRM_STATE_ACQ ||
 997                     x->id.spi       != 0 ||
 998                     x->id.proto     != proto ||
 999                     (mark & x->mark.m) != x->mark.v ||
1000                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1001                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1002                         continue;
1003
1004                 xfrm_state_hold(x);
1005                 return x;
1006         }
1007
1008         if (!create)
1009                 return NULL;
1010
1011         x = xfrm_state_alloc(net);
1012         if (likely(x)) {
1013                 switch (family) {
1014                 case AF_INET:
1015                         x->sel.daddr.a4 = daddr->a4;
1016                         x->sel.saddr.a4 = saddr->a4;
1017                         x->sel.prefixlen_d = 32;
1018                         x->sel.prefixlen_s = 32;
1019                         x->props.saddr.a4 = saddr->a4;
1020                         x->id.daddr.a4 = daddr->a4;
1021                         break;
1022
1023                 case AF_INET6:
1024                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1025                                        (struct in6_addr *)daddr);
1026                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1027                                        (struct in6_addr *)saddr);
1028                         x->sel.prefixlen_d = 128;
1029                         x->sel.prefixlen_s = 128;
1030                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1031                                        (struct in6_addr *)saddr);
1032                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1033                                        (struct in6_addr *)daddr);
1034                         break;
1035                 }
1036
1037                 x->km.state = XFRM_STATE_ACQ;
1038                 x->id.proto = proto;
1039                 x->props.family = family;
1040                 x->props.mode = mode;
1041                 x->props.reqid = reqid;
1042                 x->mark.v = m->v;
1043                 x->mark.m = m->m;
1044                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1045                 xfrm_state_hold(x);
1046                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1047                 list_add(&x->km.all, &net->xfrm.state_all);
1048                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1049                 h = xfrm_src_hash(net, daddr, saddr, family);
1050                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1051
1052                 net->xfrm.state_num++;
1053
1054                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1055         }
1056
1057         return x;
1058 }
1059
1060 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1061
1062 int xfrm_state_add(struct xfrm_state *x)
1063 {
1064         struct net *net = xs_net(x);
1065         struct xfrm_state *x1, *to_put;
1066         int family;
1067         int err;
1068         u32 mark = x->mark.v & x->mark.m;
1069         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1070
1071         family = x->props.family;
1072
1073         to_put = NULL;
1074
1075         spin_lock_bh(&xfrm_state_lock);
1076
1077         x1 = __xfrm_state_locate(x, use_spi, family);
1078         if (x1) {
1079                 to_put = x1;
1080                 x1 = NULL;
1081                 err = -EEXIST;
1082                 goto out;
1083         }
1084
1085         if (use_spi && x->km.seq) {
1086                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1087                 if (x1 && ((x1->id.proto != x->id.proto) ||
1088                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1089                         to_put = x1;
1090                         x1 = NULL;
1091                 }
1092         }
1093
1094         if (use_spi && !x1)
1095                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1096                                      x->props.reqid, x->id.proto,
1097                                      &x->id.daddr, &x->props.saddr, 0);
1098
1099         __xfrm_state_bump_genids(x);
1100         __xfrm_state_insert(x);
1101         err = 0;
1102
1103 out:
1104         spin_unlock_bh(&xfrm_state_lock);
1105
1106         if (x1) {
1107                 xfrm_state_delete(x1);
1108                 xfrm_state_put(x1);
1109         }
1110
1111         if (to_put)
1112                 xfrm_state_put(to_put);
1113
1114         return err;
1115 }
1116 EXPORT_SYMBOL(xfrm_state_add);
1117
1118 #ifdef CONFIG_XFRM_MIGRATE
1119 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1120 {
1121         struct net *net = xs_net(orig);
1122         int err = -ENOMEM;
1123         struct xfrm_state *x = xfrm_state_alloc(net);
1124         if (!x)
1125                 goto out;
1126
1127         memcpy(&x->id, &orig->id, sizeof(x->id));
1128         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1129         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1130         x->props.mode = orig->props.mode;
1131         x->props.replay_window = orig->props.replay_window;
1132         x->props.reqid = orig->props.reqid;
1133         x->props.family = orig->props.family;
1134         x->props.saddr = orig->props.saddr;
1135
1136         if (orig->aalg) {
1137                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1138                 if (!x->aalg)
1139                         goto error;
1140         }
1141         x->props.aalgo = orig->props.aalgo;
1142
1143         if (orig->ealg) {
1144                 x->ealg = xfrm_algo_clone(orig->ealg);
1145                 if (!x->ealg)
1146                         goto error;
1147         }
1148         x->props.ealgo = orig->props.ealgo;
1149
1150         if (orig->calg) {
1151                 x->calg = xfrm_algo_clone(orig->calg);
1152                 if (!x->calg)
1153                         goto error;
1154         }
1155         x->props.calgo = orig->props.calgo;
1156
1157         if (orig->encap) {
1158                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1159                 if (!x->encap)
1160                         goto error;
1161         }
1162
1163         if (orig->coaddr) {
1164                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1165                                     GFP_KERNEL);
1166                 if (!x->coaddr)
1167                         goto error;
1168         }
1169
1170         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1171
1172         err = xfrm_init_state(x);
1173         if (err)
1174                 goto error;
1175
1176         x->props.flags = orig->props.flags;
1177
1178         x->curlft.add_time = orig->curlft.add_time;
1179         x->km.state = orig->km.state;
1180         x->km.seq = orig->km.seq;
1181
1182         return x;
1183
1184  error:
1185         xfrm_state_put(x);
1186 out:
1187         if (errp)
1188                 *errp = err;
1189         return NULL;
1190 }
1191
1192 /* xfrm_state_lock is held */
1193 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1194 {
1195         unsigned int h;
1196         struct xfrm_state *x;
1197         struct hlist_node *entry;
1198
1199         if (m->reqid) {
1200                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1201                                   m->reqid, m->old_family);
1202                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1203                         if (x->props.mode != m->mode ||
1204                             x->id.proto != m->proto)
1205                                 continue;
1206                         if (m->reqid && x->props.reqid != m->reqid)
1207                                 continue;
1208                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1209                                           m->old_family) ||
1210                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1211                                           m->old_family))
1212                                 continue;
1213                         xfrm_state_hold(x);
1214                         return x;
1215                 }
1216         } else {
1217                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1218                                   m->old_family);
1219                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1220                         if (x->props.mode != m->mode ||
1221                             x->id.proto != m->proto)
1222                                 continue;
1223                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1224                                           m->old_family) ||
1225                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1226                                           m->old_family))
1227                                 continue;
1228                         xfrm_state_hold(x);
1229                         return x;
1230                 }
1231         }
1232
1233         return NULL;
1234 }
1235 EXPORT_SYMBOL(xfrm_migrate_state_find);
1236
1237 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1238                                        struct xfrm_migrate *m)
1239 {
1240         struct xfrm_state *xc;
1241         int err;
1242
1243         xc = xfrm_state_clone(x, &err);
1244         if (!xc)
1245                 return NULL;
1246
1247         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1248         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1249
1250         /* add state */
1251         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1252                 /* a care is needed when the destination address of the
1253                    state is to be updated as it is a part of triplet */
1254                 xfrm_state_insert(xc);
1255         } else {
1256                 if ((err = xfrm_state_add(xc)) < 0)
1257                         goto error;
1258         }
1259
1260         return xc;
1261 error:
1262         kfree(xc);
1263         return NULL;
1264 }
1265 EXPORT_SYMBOL(xfrm_state_migrate);
1266 #endif
1267
1268 int xfrm_state_update(struct xfrm_state *x)
1269 {
1270         struct xfrm_state *x1, *to_put;
1271         int err;
1272         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1273
1274         to_put = NULL;
1275
1276         spin_lock_bh(&xfrm_state_lock);
1277         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1278
1279         err = -ESRCH;
1280         if (!x1)
1281                 goto out;
1282
1283         if (xfrm_state_kern(x1)) {
1284                 to_put = x1;
1285                 err = -EEXIST;
1286                 goto out;
1287         }
1288
1289         if (x1->km.state == XFRM_STATE_ACQ) {
1290                 __xfrm_state_insert(x);
1291                 x = NULL;
1292         }
1293         err = 0;
1294
1295 out:
1296         spin_unlock_bh(&xfrm_state_lock);
1297
1298         if (to_put)
1299                 xfrm_state_put(to_put);
1300
1301         if (err)
1302                 return err;
1303
1304         if (!x) {
1305                 xfrm_state_delete(x1);
1306                 xfrm_state_put(x1);
1307                 return 0;
1308         }
1309
1310         err = -EINVAL;
1311         spin_lock_bh(&x1->lock);
1312         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1313                 if (x->encap && x1->encap)
1314                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1315                 if (x->coaddr && x1->coaddr) {
1316                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1317                 }
1318                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1319                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1320                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1321                 x1->km.dying = 0;
1322
1323                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1324                 if (x1->curlft.use_time)
1325                         xfrm_state_check_expire(x1);
1326
1327                 err = 0;
1328         }
1329         spin_unlock_bh(&x1->lock);
1330
1331         xfrm_state_put(x1);
1332
1333         return err;
1334 }
1335 EXPORT_SYMBOL(xfrm_state_update);
1336
1337 int xfrm_state_check_expire(struct xfrm_state *x)
1338 {
1339         if (!x->curlft.use_time)
1340                 x->curlft.use_time = get_seconds();
1341
1342         if (x->km.state != XFRM_STATE_VALID)
1343                 return -EINVAL;
1344
1345         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1346             x->curlft.packets >= x->lft.hard_packet_limit) {
1347                 x->km.state = XFRM_STATE_EXPIRED;
1348                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1349                 return -EINVAL;
1350         }
1351
1352         if (!x->km.dying &&
1353             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1354              x->curlft.packets >= x->lft.soft_packet_limit)) {
1355                 x->km.dying = 1;
1356                 km_state_expired(x, 0, 0);
1357         }
1358         return 0;
1359 }
1360 EXPORT_SYMBOL(xfrm_state_check_expire);
1361
1362 struct xfrm_state *
1363 xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi,
1364                   u8 proto, unsigned short family)
1365 {
1366         struct xfrm_state *x;
1367
1368         spin_lock_bh(&xfrm_state_lock);
1369         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1370         spin_unlock_bh(&xfrm_state_lock);
1371         return x;
1372 }
1373 EXPORT_SYMBOL(xfrm_state_lookup);
1374
1375 struct xfrm_state *
1376 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1377                          xfrm_address_t *daddr, xfrm_address_t *saddr,
1378                          u8 proto, unsigned short family)
1379 {
1380         struct xfrm_state *x;
1381
1382         spin_lock_bh(&xfrm_state_lock);
1383         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1384         spin_unlock_bh(&xfrm_state_lock);
1385         return x;
1386 }
1387 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1388
1389 struct xfrm_state *
1390 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1391               xfrm_address_t *daddr, xfrm_address_t *saddr,
1392               int create, unsigned short family)
1393 {
1394         struct xfrm_state *x;
1395
1396         spin_lock_bh(&xfrm_state_lock);
1397         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1398         spin_unlock_bh(&xfrm_state_lock);
1399
1400         return x;
1401 }
1402 EXPORT_SYMBOL(xfrm_find_acq);
1403
1404 #ifdef CONFIG_XFRM_SUB_POLICY
1405 int
1406 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1407                unsigned short family)
1408 {
1409         int err = 0;
1410         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1411         if (!afinfo)
1412                 return -EAFNOSUPPORT;
1413
1414         spin_lock_bh(&xfrm_state_lock);
1415         if (afinfo->tmpl_sort)
1416                 err = afinfo->tmpl_sort(dst, src, n);
1417         spin_unlock_bh(&xfrm_state_lock);
1418         xfrm_state_put_afinfo(afinfo);
1419         return err;
1420 }
1421 EXPORT_SYMBOL(xfrm_tmpl_sort);
1422
1423 int
1424 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1425                 unsigned short family)
1426 {
1427         int err = 0;
1428         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1429         if (!afinfo)
1430                 return -EAFNOSUPPORT;
1431
1432         spin_lock_bh(&xfrm_state_lock);
1433         if (afinfo->state_sort)
1434                 err = afinfo->state_sort(dst, src, n);
1435         spin_unlock_bh(&xfrm_state_lock);
1436         xfrm_state_put_afinfo(afinfo);
1437         return err;
1438 }
1439 EXPORT_SYMBOL(xfrm_state_sort);
1440 #endif
1441
1442 /* Silly enough, but I'm lazy to build resolution list */
1443
1444 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1445 {
1446         int i;
1447
1448         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1449                 struct hlist_node *entry;
1450                 struct xfrm_state *x;
1451
1452                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1453                         if (x->km.seq == seq &&
1454                             (mark & x->mark.m) == x->mark.v &&
1455                             x->km.state == XFRM_STATE_ACQ) {
1456                                 xfrm_state_hold(x);
1457                                 return x;
1458                         }
1459                 }
1460         }
1461         return NULL;
1462 }
1463
1464 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1465 {
1466         struct xfrm_state *x;
1467
1468         spin_lock_bh(&xfrm_state_lock);
1469         x = __xfrm_find_acq_byseq(net, mark, seq);
1470         spin_unlock_bh(&xfrm_state_lock);
1471         return x;
1472 }
1473 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1474
1475 u32 xfrm_get_acqseq(void)
1476 {
1477         u32 res;
1478         static atomic_t acqseq;
1479
1480         do {
1481                 res = atomic_inc_return(&acqseq);
1482         } while (!res);
1483
1484         return res;
1485 }
1486 EXPORT_SYMBOL(xfrm_get_acqseq);
1487
1488 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1489 {
1490         struct net *net = xs_net(x);
1491         unsigned int h;
1492         struct xfrm_state *x0;
1493         int err = -ENOENT;
1494         __be32 minspi = htonl(low);
1495         __be32 maxspi = htonl(high);
1496         u32 mark = x->mark.v & x->mark.m;
1497
1498         spin_lock_bh(&x->lock);
1499         if (x->km.state == XFRM_STATE_DEAD)
1500                 goto unlock;
1501
1502         err = 0;
1503         if (x->id.spi)
1504                 goto unlock;
1505
1506         err = -ENOENT;
1507
1508         if (minspi == maxspi) {
1509                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1510                 if (x0) {
1511                         xfrm_state_put(x0);
1512                         goto unlock;
1513                 }
1514                 x->id.spi = minspi;
1515         } else {
1516                 u32 spi = 0;
1517                 for (h=0; h<high-low+1; h++) {
1518                         spi = low + net_random()%(high-low+1);
1519                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1520                         if (x0 == NULL) {
1521                                 x->id.spi = htonl(spi);
1522                                 break;
1523                         }
1524                         xfrm_state_put(x0);
1525                 }
1526         }
1527         if (x->id.spi) {
1528                 spin_lock_bh(&xfrm_state_lock);
1529                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1530                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1531                 spin_unlock_bh(&xfrm_state_lock);
1532
1533                 err = 0;
1534         }
1535
1536 unlock:
1537         spin_unlock_bh(&x->lock);
1538
1539         return err;
1540 }
1541 EXPORT_SYMBOL(xfrm_alloc_spi);
1542
1543 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1544                     int (*func)(struct xfrm_state *, int, void*),
1545                     void *data)
1546 {
1547         struct xfrm_state *state;
1548         struct xfrm_state_walk *x;
1549         int err = 0;
1550
1551         if (walk->seq != 0 && list_empty(&walk->all))
1552                 return 0;
1553
1554         spin_lock_bh(&xfrm_state_lock);
1555         if (list_empty(&walk->all))
1556                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1557         else
1558                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1559         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1560                 if (x->state == XFRM_STATE_DEAD)
1561                         continue;
1562                 state = container_of(x, struct xfrm_state, km);
1563                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1564                         continue;
1565                 err = func(state, walk->seq, data);
1566                 if (err) {
1567                         list_move_tail(&walk->all, &x->all);
1568                         goto out;
1569                 }
1570                 walk->seq++;
1571         }
1572         if (walk->seq == 0) {
1573                 err = -ENOENT;
1574                 goto out;
1575         }
1576         list_del_init(&walk->all);
1577 out:
1578         spin_unlock_bh(&xfrm_state_lock);
1579         return err;
1580 }
1581 EXPORT_SYMBOL(xfrm_state_walk);
1582
1583 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1584 {
1585         INIT_LIST_HEAD(&walk->all);
1586         walk->proto = proto;
1587         walk->state = XFRM_STATE_DEAD;
1588         walk->seq = 0;
1589 }
1590 EXPORT_SYMBOL(xfrm_state_walk_init);
1591
1592 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1593 {
1594         if (list_empty(&walk->all))
1595                 return;
1596
1597         spin_lock_bh(&xfrm_state_lock);
1598         list_del(&walk->all);
1599         spin_unlock_bh(&xfrm_state_lock);
1600 }
1601 EXPORT_SYMBOL(xfrm_state_walk_done);
1602
1603
1604 void xfrm_replay_notify(struct xfrm_state *x, int event)
1605 {
1606         struct km_event c;
1607         /* we send notify messages in case
1608          *  1. we updated on of the sequence numbers, and the seqno difference
1609          *     is at least x->replay_maxdiff, in this case we also update the
1610          *     timeout of our timer function
1611          *  2. if x->replay_maxage has elapsed since last update,
1612          *     and there were changes
1613          *
1614          *  The state structure must be locked!
1615          */
1616
1617         switch (event) {
1618         case XFRM_REPLAY_UPDATE:
1619                 if (x->replay_maxdiff &&
1620                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1621                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1622                         if (x->xflags & XFRM_TIME_DEFER)
1623                                 event = XFRM_REPLAY_TIMEOUT;
1624                         else
1625                                 return;
1626                 }
1627
1628                 break;
1629
1630         case XFRM_REPLAY_TIMEOUT:
1631                 if ((x->replay.seq == x->preplay.seq) &&
1632                     (x->replay.bitmap == x->preplay.bitmap) &&
1633                     (x->replay.oseq == x->preplay.oseq)) {
1634                         x->xflags |= XFRM_TIME_DEFER;
1635                         return;
1636                 }
1637
1638                 break;
1639         }
1640
1641         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1642         c.event = XFRM_MSG_NEWAE;
1643         c.data.aevent = event;
1644         km_state_notify(x, &c);
1645
1646         if (x->replay_maxage &&
1647             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1648                 x->xflags &= ~XFRM_TIME_DEFER;
1649 }
1650
1651 static void xfrm_replay_timer_handler(unsigned long data)
1652 {
1653         struct xfrm_state *x = (struct xfrm_state*)data;
1654
1655         spin_lock(&x->lock);
1656
1657         if (x->km.state == XFRM_STATE_VALID) {
1658                 if (xfrm_aevent_is_on(xs_net(x)))
1659                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1660                 else
1661                         x->xflags |= XFRM_TIME_DEFER;
1662         }
1663
1664         spin_unlock(&x->lock);
1665 }
1666
1667 int xfrm_replay_check(struct xfrm_state *x,
1668                       struct sk_buff *skb, __be32 net_seq)
1669 {
1670         u32 diff;
1671         u32 seq = ntohl(net_seq);
1672
1673         if (unlikely(seq == 0))
1674                 goto err;
1675
1676         if (likely(seq > x->replay.seq))
1677                 return 0;
1678
1679         diff = x->replay.seq - seq;
1680         if (diff >= min_t(unsigned int, x->props.replay_window,
1681                           sizeof(x->replay.bitmap) * 8)) {
1682                 x->stats.replay_window++;
1683                 goto err;
1684         }
1685
1686         if (x->replay.bitmap & (1U << diff)) {
1687                 x->stats.replay++;
1688                 goto err;
1689         }
1690         return 0;
1691
1692 err:
1693         xfrm_audit_state_replay(x, skb, net_seq);
1694         return -EINVAL;
1695 }
1696
1697 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1698 {
1699         u32 diff;
1700         u32 seq = ntohl(net_seq);
1701
1702         if (seq > x->replay.seq) {
1703                 diff = seq - x->replay.seq;
1704                 if (diff < x->props.replay_window)
1705                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1706                 else
1707                         x->replay.bitmap = 1;
1708                 x->replay.seq = seq;
1709         } else {
1710                 diff = x->replay.seq - seq;
1711                 x->replay.bitmap |= (1U << diff);
1712         }
1713
1714         if (xfrm_aevent_is_on(xs_net(x)))
1715                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1716 }
1717
1718 static LIST_HEAD(xfrm_km_list);
1719 static DEFINE_RWLOCK(xfrm_km_lock);
1720
1721 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1722 {
1723         struct xfrm_mgr *km;
1724
1725         read_lock(&xfrm_km_lock);
1726         list_for_each_entry(km, &xfrm_km_list, list)
1727                 if (km->notify_policy)
1728                         km->notify_policy(xp, dir, c);
1729         read_unlock(&xfrm_km_lock);
1730 }
1731
1732 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1733 {
1734         struct xfrm_mgr *km;
1735         read_lock(&xfrm_km_lock);
1736         list_for_each_entry(km, &xfrm_km_list, list)
1737                 if (km->notify)
1738                         km->notify(x, c);
1739         read_unlock(&xfrm_km_lock);
1740 }
1741
1742 EXPORT_SYMBOL(km_policy_notify);
1743 EXPORT_SYMBOL(km_state_notify);
1744
1745 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1746 {
1747         struct net *net = xs_net(x);
1748         struct km_event c;
1749
1750         c.data.hard = hard;
1751         c.pid = pid;
1752         c.event = XFRM_MSG_EXPIRE;
1753         km_state_notify(x, &c);
1754
1755         if (hard)
1756                 wake_up(&net->xfrm.km_waitq);
1757 }
1758
1759 EXPORT_SYMBOL(km_state_expired);
1760 /*
1761  * We send to all registered managers regardless of failure
1762  * We are happy with one success
1763 */
1764 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1765 {
1766         int err = -EINVAL, acqret;
1767         struct xfrm_mgr *km;
1768
1769         read_lock(&xfrm_km_lock);
1770         list_for_each_entry(km, &xfrm_km_list, list) {
1771                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1772                 if (!acqret)
1773                         err = acqret;
1774         }
1775         read_unlock(&xfrm_km_lock);
1776         return err;
1777 }
1778 EXPORT_SYMBOL(km_query);
1779
1780 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1781 {
1782         int err = -EINVAL;
1783         struct xfrm_mgr *km;
1784
1785         read_lock(&xfrm_km_lock);
1786         list_for_each_entry(km, &xfrm_km_list, list) {
1787                 if (km->new_mapping)
1788                         err = km->new_mapping(x, ipaddr, sport);
1789                 if (!err)
1790                         break;
1791         }
1792         read_unlock(&xfrm_km_lock);
1793         return err;
1794 }
1795 EXPORT_SYMBOL(km_new_mapping);
1796
1797 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1798 {
1799         struct net *net = xp_net(pol);
1800         struct km_event c;
1801
1802         c.data.hard = hard;
1803         c.pid = pid;
1804         c.event = XFRM_MSG_POLEXPIRE;
1805         km_policy_notify(pol, dir, &c);
1806
1807         if (hard)
1808                 wake_up(&net->xfrm.km_waitq);
1809 }
1810 EXPORT_SYMBOL(km_policy_expired);
1811
1812 #ifdef CONFIG_XFRM_MIGRATE
1813 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1814                struct xfrm_migrate *m, int num_migrate,
1815                struct xfrm_kmaddress *k)
1816 {
1817         int err = -EINVAL;
1818         int ret;
1819         struct xfrm_mgr *km;
1820
1821         read_lock(&xfrm_km_lock);
1822         list_for_each_entry(km, &xfrm_km_list, list) {
1823                 if (km->migrate) {
1824                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1825                         if (!ret)
1826                                 err = ret;
1827                 }
1828         }
1829         read_unlock(&xfrm_km_lock);
1830         return err;
1831 }
1832 EXPORT_SYMBOL(km_migrate);
1833 #endif
1834
1835 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1836 {
1837         int err = -EINVAL;
1838         int ret;
1839         struct xfrm_mgr *km;
1840
1841         read_lock(&xfrm_km_lock);
1842         list_for_each_entry(km, &xfrm_km_list, list) {
1843                 if (km->report) {
1844                         ret = km->report(net, proto, sel, addr);
1845                         if (!ret)
1846                                 err = ret;
1847                 }
1848         }
1849         read_unlock(&xfrm_km_lock);
1850         return err;
1851 }
1852 EXPORT_SYMBOL(km_report);
1853
1854 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1855 {
1856         int err;
1857         u8 *data;
1858         struct xfrm_mgr *km;
1859         struct xfrm_policy *pol = NULL;
1860
1861         if (optlen <= 0 || optlen > PAGE_SIZE)
1862                 return -EMSGSIZE;
1863
1864         data = kmalloc(optlen, GFP_KERNEL);
1865         if (!data)
1866                 return -ENOMEM;
1867
1868         err = -EFAULT;
1869         if (copy_from_user(data, optval, optlen))
1870                 goto out;
1871
1872         err = -EINVAL;
1873         read_lock(&xfrm_km_lock);
1874         list_for_each_entry(km, &xfrm_km_list, list) {
1875                 pol = km->compile_policy(sk, optname, data,
1876                                          optlen, &err);
1877                 if (err >= 0)
1878                         break;
1879         }
1880         read_unlock(&xfrm_km_lock);
1881
1882         if (err >= 0) {
1883                 xfrm_sk_policy_insert(sk, err, pol);
1884                 xfrm_pol_put(pol);
1885                 err = 0;
1886         }
1887
1888 out:
1889         kfree(data);
1890         return err;
1891 }
1892 EXPORT_SYMBOL(xfrm_user_policy);
1893
1894 int xfrm_register_km(struct xfrm_mgr *km)
1895 {
1896         write_lock_bh(&xfrm_km_lock);
1897         list_add_tail(&km->list, &xfrm_km_list);
1898         write_unlock_bh(&xfrm_km_lock);
1899         return 0;
1900 }
1901 EXPORT_SYMBOL(xfrm_register_km);
1902
1903 int xfrm_unregister_km(struct xfrm_mgr *km)
1904 {
1905         write_lock_bh(&xfrm_km_lock);
1906         list_del(&km->list);
1907         write_unlock_bh(&xfrm_km_lock);
1908         return 0;
1909 }
1910 EXPORT_SYMBOL(xfrm_unregister_km);
1911
1912 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1913 {
1914         int err = 0;
1915         if (unlikely(afinfo == NULL))
1916                 return -EINVAL;
1917         if (unlikely(afinfo->family >= NPROTO))
1918                 return -EAFNOSUPPORT;
1919         write_lock_bh(&xfrm_state_afinfo_lock);
1920         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1921                 err = -ENOBUFS;
1922         else
1923                 xfrm_state_afinfo[afinfo->family] = afinfo;
1924         write_unlock_bh(&xfrm_state_afinfo_lock);
1925         return err;
1926 }
1927 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1928
1929 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1930 {
1931         int err = 0;
1932         if (unlikely(afinfo == NULL))
1933                 return -EINVAL;
1934         if (unlikely(afinfo->family >= NPROTO))
1935                 return -EAFNOSUPPORT;
1936         write_lock_bh(&xfrm_state_afinfo_lock);
1937         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1938                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1939                         err = -EINVAL;
1940                 else
1941                         xfrm_state_afinfo[afinfo->family] = NULL;
1942         }
1943         write_unlock_bh(&xfrm_state_afinfo_lock);
1944         return err;
1945 }
1946 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1947
1948 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1949 {
1950         struct xfrm_state_afinfo *afinfo;
1951         if (unlikely(family >= NPROTO))
1952                 return NULL;
1953         read_lock(&xfrm_state_afinfo_lock);
1954         afinfo = xfrm_state_afinfo[family];
1955         if (unlikely(!afinfo))
1956                 read_unlock(&xfrm_state_afinfo_lock);
1957         return afinfo;
1958 }
1959
1960 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1961         __releases(xfrm_state_afinfo_lock)
1962 {
1963         read_unlock(&xfrm_state_afinfo_lock);
1964 }
1965
1966 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1967 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1968 {
1969         if (x->tunnel) {
1970                 struct xfrm_state *t = x->tunnel;
1971
1972                 if (atomic_read(&t->tunnel_users) == 2)
1973                         xfrm_state_delete(t);
1974                 atomic_dec(&t->tunnel_users);
1975                 xfrm_state_put(t);
1976                 x->tunnel = NULL;
1977         }
1978 }
1979 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1980
1981 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1982 {
1983         int res;
1984
1985         spin_lock_bh(&x->lock);
1986         if (x->km.state == XFRM_STATE_VALID &&
1987             x->type && x->type->get_mtu)
1988                 res = x->type->get_mtu(x, mtu);
1989         else
1990                 res = mtu - x->props.header_len;
1991         spin_unlock_bh(&x->lock);
1992         return res;
1993 }
1994
1995 int xfrm_init_state(struct xfrm_state *x)
1996 {
1997         struct xfrm_state_afinfo *afinfo;
1998         struct xfrm_mode *inner_mode;
1999         int family = x->props.family;
2000         int err;
2001
2002         err = -EAFNOSUPPORT;
2003         afinfo = xfrm_state_get_afinfo(family);
2004         if (!afinfo)
2005                 goto error;
2006
2007         err = 0;
2008         if (afinfo->init_flags)
2009                 err = afinfo->init_flags(x);
2010
2011         xfrm_state_put_afinfo(afinfo);
2012
2013         if (err)
2014                 goto error;
2015
2016         err = -EPROTONOSUPPORT;
2017
2018         if (x->sel.family != AF_UNSPEC) {
2019                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2020                 if (inner_mode == NULL)
2021                         goto error;
2022
2023                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2024                     family != x->sel.family) {
2025                         xfrm_put_mode(inner_mode);
2026                         goto error;
2027                 }
2028
2029                 x->inner_mode = inner_mode;
2030         } else {
2031                 struct xfrm_mode *inner_mode_iaf;
2032                 int iafamily = AF_INET;
2033
2034                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2035                 if (inner_mode == NULL)
2036                         goto error;
2037
2038                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2039                         xfrm_put_mode(inner_mode);
2040                         goto error;
2041                 }
2042                 x->inner_mode = inner_mode;
2043
2044                 if (x->props.family == AF_INET)
2045                         iafamily = AF_INET6;
2046
2047                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2048                 if (inner_mode_iaf) {
2049                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2050                                 x->inner_mode_iaf = inner_mode_iaf;
2051                         else
2052                                 xfrm_put_mode(inner_mode_iaf);
2053                 }
2054         }
2055
2056         x->type = xfrm_get_type(x->id.proto, family);
2057         if (x->type == NULL)
2058                 goto error;
2059
2060         err = x->type->init_state(x);
2061         if (err)
2062                 goto error;
2063
2064         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2065         if (x->outer_mode == NULL)
2066                 goto error;
2067
2068         x->km.state = XFRM_STATE_VALID;
2069
2070 error:
2071         return err;
2072 }
2073
2074 EXPORT_SYMBOL(xfrm_init_state);
2075
2076 int __net_init xfrm_state_init(struct net *net)
2077 {
2078         unsigned int sz;
2079
2080         INIT_LIST_HEAD(&net->xfrm.state_all);
2081
2082         sz = sizeof(struct hlist_head) * 8;
2083
2084         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2085         if (!net->xfrm.state_bydst)
2086                 goto out_bydst;
2087         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2088         if (!net->xfrm.state_bysrc)
2089                 goto out_bysrc;
2090         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2091         if (!net->xfrm.state_byspi)
2092                 goto out_byspi;
2093         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2094
2095         net->xfrm.state_num = 0;
2096         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2097         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2098         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2099         init_waitqueue_head(&net->xfrm.km_waitq);
2100         return 0;
2101
2102 out_byspi:
2103         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2104 out_bysrc:
2105         xfrm_hash_free(net->xfrm.state_bydst, sz);
2106 out_bydst:
2107         return -ENOMEM;
2108 }
2109
2110 void xfrm_state_fini(struct net *net)
2111 {
2112         struct xfrm_audit audit_info;
2113         unsigned int sz;
2114
2115         flush_work(&net->xfrm.state_hash_work);
2116         audit_info.loginuid = -1;
2117         audit_info.sessionid = -1;
2118         audit_info.secid = 0;
2119         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2120         flush_work(&net->xfrm.state_gc_work);
2121
2122         WARN_ON(!list_empty(&net->xfrm.state_all));
2123
2124         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2125         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2126         xfrm_hash_free(net->xfrm.state_byspi, sz);
2127         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2128         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2129         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2130         xfrm_hash_free(net->xfrm.state_bydst, sz);
2131 }
2132
2133 #ifdef CONFIG_AUDITSYSCALL
2134 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2135                                      struct audit_buffer *audit_buf)
2136 {
2137         struct xfrm_sec_ctx *ctx = x->security;
2138         u32 spi = ntohl(x->id.spi);
2139
2140         if (ctx)
2141                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2142                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2143
2144         switch(x->props.family) {
2145         case AF_INET:
2146                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2147                                  &x->props.saddr.a4, &x->id.daddr.a4);
2148                 break;
2149         case AF_INET6:
2150                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2151                                  x->props.saddr.a6, x->id.daddr.a6);
2152                 break;
2153         }
2154
2155         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2156 }
2157
2158 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2159                                       struct audit_buffer *audit_buf)
2160 {
2161         struct iphdr *iph4;
2162         struct ipv6hdr *iph6;
2163
2164         switch (family) {
2165         case AF_INET:
2166                 iph4 = ip_hdr(skb);
2167                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2168                                  &iph4->saddr, &iph4->daddr);
2169                 break;
2170         case AF_INET6:
2171                 iph6 = ipv6_hdr(skb);
2172                 audit_log_format(audit_buf,
2173                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2174                                  &iph6->saddr,&iph6->daddr,
2175                                  iph6->flow_lbl[0] & 0x0f,
2176                                  iph6->flow_lbl[1],
2177                                  iph6->flow_lbl[2]);
2178                 break;
2179         }
2180 }
2181
2182 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2183                           uid_t auid, u32 sessionid, u32 secid)
2184 {
2185         struct audit_buffer *audit_buf;
2186
2187         audit_buf = xfrm_audit_start("SAD-add");
2188         if (audit_buf == NULL)
2189                 return;
2190         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2191         xfrm_audit_helper_sainfo(x, audit_buf);
2192         audit_log_format(audit_buf, " res=%u", result);
2193         audit_log_end(audit_buf);
2194 }
2195 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2196
2197 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2198                              uid_t auid, u32 sessionid, u32 secid)
2199 {
2200         struct audit_buffer *audit_buf;
2201
2202         audit_buf = xfrm_audit_start("SAD-delete");
2203         if (audit_buf == NULL)
2204                 return;
2205         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2206         xfrm_audit_helper_sainfo(x, audit_buf);
2207         audit_log_format(audit_buf, " res=%u", result);
2208         audit_log_end(audit_buf);
2209 }
2210 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2211
2212 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2213                                       struct sk_buff *skb)
2214 {
2215         struct audit_buffer *audit_buf;
2216         u32 spi;
2217
2218         audit_buf = xfrm_audit_start("SA-replay-overflow");
2219         if (audit_buf == NULL)
2220                 return;
2221         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2222         /* don't record the sequence number because it's inherent in this kind
2223          * of audit message */
2224         spi = ntohl(x->id.spi);
2225         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2226         audit_log_end(audit_buf);
2227 }
2228 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2229
2230 static void xfrm_audit_state_replay(struct xfrm_state *x,
2231                              struct sk_buff *skb, __be32 net_seq)
2232 {
2233         struct audit_buffer *audit_buf;
2234         u32 spi;
2235
2236         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2237         if (audit_buf == NULL)
2238                 return;
2239         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2240         spi = ntohl(x->id.spi);
2241         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2242                          spi, spi, ntohl(net_seq));
2243         audit_log_end(audit_buf);
2244 }
2245
2246 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2247 {
2248         struct audit_buffer *audit_buf;
2249
2250         audit_buf = xfrm_audit_start("SA-notfound");
2251         if (audit_buf == NULL)
2252                 return;
2253         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2254         audit_log_end(audit_buf);
2255 }
2256 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2257
2258 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2259                                __be32 net_spi, __be32 net_seq)
2260 {
2261         struct audit_buffer *audit_buf;
2262         u32 spi;
2263
2264         audit_buf = xfrm_audit_start("SA-notfound");
2265         if (audit_buf == NULL)
2266                 return;
2267         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2268         spi = ntohl(net_spi);
2269         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2270                          spi, spi, ntohl(net_seq));
2271         audit_log_end(audit_buf);
2272 }
2273 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2274
2275 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2276                               struct sk_buff *skb, u8 proto)
2277 {
2278         struct audit_buffer *audit_buf;
2279         __be32 net_spi;
2280         __be32 net_seq;
2281
2282         audit_buf = xfrm_audit_start("SA-icv-failure");
2283         if (audit_buf == NULL)
2284                 return;
2285         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2286         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2287                 u32 spi = ntohl(net_spi);
2288                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2289                                  spi, spi, ntohl(net_seq));
2290         }
2291         audit_log_end(audit_buf);
2292 }
2293 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2294 #endif /* CONFIG_AUDITSYSCALL */