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/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
  41 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family);
  42 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo);
  43
  44 #ifdef CONFIG_AUDITSYSCALL
  45 static void xfrm_audit_state_replay(struct xfrm_state *x,
  46                                     struct sk_buff *skb, __be32 net_seq);
  47 #else
  48 #define xfrm_audit_state_replay(x, s, sq)       do { ; } while (0)
  49 #endif /* CONFIG_AUDITSYSCALL */
  50
  51 static inline unsigned int xfrm_dst_hash(struct net *net,
  52                                          xfrm_address_t *daddr,
  53                                          xfrm_address_t *saddr,
  54                                          u32 reqid,
  55                                          unsigned short family)
  56 {
  57         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  58 }
  59
  60 static inline unsigned int xfrm_src_hash(struct net *net,
  61                                          xfrm_address_t *daddr,
  62                                          xfrm_address_t *saddr,
  63                                          unsigned short family)
  64 {
  65         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  66 }
  67
  68 static inline unsigned int
  69 xfrm_spi_hash(struct net *net, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
  70 {
  71         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  72 }
  73
  74 static void xfrm_hash_transfer(struct hlist_head *list,
  75                                struct hlist_head *ndsttable,
  76                                struct hlist_head *nsrctable,
  77                                struct hlist_head *nspitable,
  78                                unsigned int nhashmask)
  79 {
  80         struct hlist_node *entry, *tmp;
  81         struct xfrm_state *x;
  82
  83         hlist_for_each_entry_safe(x, entry, tmp, list, bydst) {
  84                 unsigned int h;
  85
  86                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  87                                     x->props.reqid, x->props.family,
  88                                     nhashmask);
  89                 hlist_add_head(&x->bydst, ndsttable+h);
  90
  91                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  92                                     x->props.family,
  93                                     nhashmask);
  94                 hlist_add_head(&x->bysrc, nsrctable+h);
  95
  96                 if (x->id.spi) {
  97                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  98                                             x->id.proto, x->props.family,
  99                                             nhashmask);
 100                         hlist_add_head(&x->byspi, nspitable+h);
 101                 }
 102         }
 103 }
 104
 105 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 106 {
 107         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 108 }
 109
 110 static DEFINE_MUTEX(hash_resize_mutex);
 111
 112 static void xfrm_hash_resize(struct work_struct *work)
 113 {
 114         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 115         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 116         unsigned long nsize, osize;
 117         unsigned int nhashmask, ohashmask;
 118         int i;
 119
 120         mutex_lock(&hash_resize_mutex);
 121
 122         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 123         ndst = xfrm_hash_alloc(nsize);
 124         if (!ndst)
 125                 goto out_unlock;
 126         nsrc = xfrm_hash_alloc(nsize);
 127         if (!nsrc) {
 128                 xfrm_hash_free(ndst, nsize);
 129                 goto out_unlock;
 130         }
 131         nspi = xfrm_hash_alloc(nsize);
 132         if (!nspi) {
 133                 xfrm_hash_free(ndst, nsize);
 134                 xfrm_hash_free(nsrc, nsize);
 135                 goto out_unlock;
 136         }
 137
 138         spin_lock_bh(&xfrm_state_lock);
 139
 140         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 141         for (i = net->xfrm.state_hmask; i >= 0; i--)
 142                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 143                                    nhashmask);
 144
 145         odst = net->xfrm.state_bydst;
 146         osrc = net->xfrm.state_bysrc;
 147         ospi = net->xfrm.state_byspi;
 148         ohashmask = net->xfrm.state_hmask;
 149
 150         net->xfrm.state_bydst = ndst;
 151         net->xfrm.state_bysrc = nsrc;
 152         net->xfrm.state_byspi = nspi;
 153         net->xfrm.state_hmask = nhashmask;
 154
 155         spin_unlock_bh(&xfrm_state_lock);
 156
 157         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 158         xfrm_hash_free(odst, osize);
 159         xfrm_hash_free(osrc, osize);
 160         xfrm_hash_free(ospi, osize);
 161
 162 out_unlock:
 163         mutex_unlock(&hash_resize_mutex);
 164 }
 165
 166 static DEFINE_RWLOCK(xfrm_state_afinfo_lock);
 167 static struct xfrm_state_afinfo *xfrm_state_afinfo[NPROTO];
 168
 169 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 170
 171 int __xfrm_state_delete(struct xfrm_state *x);
 172
 173 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 174 void km_state_expired(struct xfrm_state *x, int hard, u32 pid);
 175
 176 static struct xfrm_state_afinfo *xfrm_state_lock_afinfo(unsigned int family)
 177 {
 178         struct xfrm_state_afinfo *afinfo;
 179         if (unlikely(family >= NPROTO))
 180                 return NULL;
 181         write_lock_bh(&xfrm_state_afinfo_lock);
 182         afinfo = xfrm_state_afinfo[family];
 183         if (unlikely(!afinfo))
 184                 write_unlock_bh(&xfrm_state_afinfo_lock);
 185         return afinfo;
 186 }
 187
 188 static void xfrm_state_unlock_afinfo(struct xfrm_state_afinfo *afinfo)
 189         __releases(xfrm_state_afinfo_lock)
 190 {
 191         write_unlock_bh(&xfrm_state_afinfo_lock);
 192 }
 193
 194 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 195 {
 196         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 197         const struct xfrm_type **typemap;
 198         int err = 0;
 199
 200         if (unlikely(afinfo == NULL))
 201                 return -EAFNOSUPPORT;
 202         typemap = afinfo->type_map;
 203
 204         if (likely(typemap[type->proto] == NULL))
 205                 typemap[type->proto] = type;
 206         else
 207                 err = -EEXIST;
 208         xfrm_state_unlock_afinfo(afinfo);
 209         return err;
 210 }
 211 EXPORT_SYMBOL(xfrm_register_type);
 212
 213 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 214 {
 215         struct xfrm_state_afinfo *afinfo = xfrm_state_lock_afinfo(family);
 216         const struct xfrm_type **typemap;
 217         int err = 0;
 218
 219         if (unlikely(afinfo == NULL))
 220                 return -EAFNOSUPPORT;
 221         typemap = afinfo->type_map;
 222
 223         if (unlikely(typemap[type->proto] != type))
 224                 err = -ENOENT;
 225         else
 226                 typemap[type->proto] = NULL;
 227         xfrm_state_unlock_afinfo(afinfo);
 228         return err;
 229 }
 230 EXPORT_SYMBOL(xfrm_unregister_type);
 231
 232 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 233 {
 234         struct xfrm_state_afinfo *afinfo;
 235         const struct xfrm_type **typemap;
 236         const struct xfrm_type *type;
 237         int modload_attempted = 0;
 238
 239 retry:
 240         afinfo = xfrm_state_get_afinfo(family);
 241         if (unlikely(afinfo == NULL))
 242                 return NULL;
 243         typemap = afinfo->type_map;
 244
 245         type = typemap[proto];
 246         if (unlikely(type && !try_module_get(type->owner)))
 247                 type = NULL;
 248         if (!type && !modload_attempted) {
 249                 xfrm_state_put_afinfo(afinfo);
 250                 request_module("xfrm-type-%d-%d", family, proto);
 251                 modload_attempted = 1;
 252                 goto retry;
 253         }
 254
 255         xfrm_state_put_afinfo(afinfo);
 256         return type;
 257 }
 258
 259 static void xfrm_put_type(const struct xfrm_type *type)
 260 {
 261         module_put(type->owner);
 262 }
 263
 264 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 265 {
 266         struct xfrm_state_afinfo *afinfo;
 267         struct xfrm_mode **modemap;
 268         int err;
 269
 270         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 271                 return -EINVAL;
 272
 273         afinfo = xfrm_state_lock_afinfo(family);
 274         if (unlikely(afinfo == NULL))
 275                 return -EAFNOSUPPORT;
 276
 277         err = -EEXIST;
 278         modemap = afinfo->mode_map;
 279         if (modemap[mode->encap])
 280                 goto out;
 281
 282         err = -ENOENT;
 283         if (!try_module_get(afinfo->owner))
 284                 goto out;
 285
 286         mode->afinfo = afinfo;
 287         modemap[mode->encap] = mode;
 288         err = 0;
 289
 290 out:
 291         xfrm_state_unlock_afinfo(afinfo);
 292         return err;
 293 }
 294 EXPORT_SYMBOL(xfrm_register_mode);
 295
 296 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 297 {
 298         struct xfrm_state_afinfo *afinfo;
 299         struct xfrm_mode **modemap;
 300         int err;
 301
 302         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 303                 return -EINVAL;
 304
 305         afinfo = xfrm_state_lock_afinfo(family);
 306         if (unlikely(afinfo == NULL))
 307                 return -EAFNOSUPPORT;
 308
 309         err = -ENOENT;
 310         modemap = afinfo->mode_map;
 311         if (likely(modemap[mode->encap] == mode)) {
 312                 modemap[mode->encap] = NULL;
 313                 module_put(mode->afinfo->owner);
 314                 err = 0;
 315         }
 316
 317         xfrm_state_unlock_afinfo(afinfo);
 318         return err;
 319 }
 320 EXPORT_SYMBOL(xfrm_unregister_mode);
 321
 322 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 323 {
 324         struct xfrm_state_afinfo *afinfo;
 325         struct xfrm_mode *mode;
 326         int modload_attempted = 0;
 327
 328         if (unlikely(encap >= XFRM_MODE_MAX))
 329                 return NULL;
 330
 331 retry:
 332         afinfo = xfrm_state_get_afinfo(family);
 333         if (unlikely(afinfo == NULL))
 334                 return NULL;
 335
 336         mode = afinfo->mode_map[encap];
 337         if (unlikely(mode && !try_module_get(mode->owner)))
 338                 mode = NULL;
 339         if (!mode && !modload_attempted) {
 340                 xfrm_state_put_afinfo(afinfo);
 341                 request_module("xfrm-mode-%d-%d", family, encap);
 342                 modload_attempted = 1;
 343                 goto retry;
 344         }
 345
 346         xfrm_state_put_afinfo(afinfo);
 347         return mode;
 348 }
 349
 350 static void xfrm_put_mode(struct xfrm_mode *mode)
 351 {
 352         module_put(mode->owner);
 353 }
 354
 355 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 356 {
 357         tasklet_hrtimer_cancel(&x->mtimer);
 358         del_timer_sync(&x->rtimer);
 359         kfree(x->aalg);
 360         kfree(x->ealg);
 361         kfree(x->calg);
 362         kfree(x->encap);
 363         kfree(x->coaddr);
 364         if (x->inner_mode)
 365                 xfrm_put_mode(x->inner_mode);
 366         if (x->inner_mode_iaf)
 367                 xfrm_put_mode(x->inner_mode_iaf);
 368         if (x->outer_mode)
 369                 xfrm_put_mode(x->outer_mode);
 370         if (x->type) {
 371                 x->type->destructor(x);
 372                 xfrm_put_type(x->type);
 373         }
 374         security_xfrm_state_free(x);
 375         kfree(x);
 376 }
 377
 378 static void xfrm_state_gc_task(struct work_struct *work)
 379 {
 380         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 381         struct xfrm_state *x;
 382         struct hlist_node *entry, *tmp;
 383         struct hlist_head gc_list;
 384
 385         spin_lock_bh(&xfrm_state_gc_lock);
 386         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 387         spin_unlock_bh(&xfrm_state_gc_lock);
 388
 389         hlist_for_each_entry_safe(x, entry, tmp, &gc_list, gclist)
 390                 xfrm_state_gc_destroy(x);
 391
 392         wake_up(&net->xfrm.km_waitq);
 393 }
 394
 395 static inline unsigned long make_jiffies(long secs)
 396 {
 397         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 398                 return MAX_SCHEDULE_TIMEOUT-1;
 399         else
 400                 return secs*HZ;
 401 }
 402
 403 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer * me)
 404 {
 405         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 406         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 407         struct net *net = xs_net(x);
 408         unsigned long now = get_seconds();
 409         long next = LONG_MAX;
 410         int warn = 0;
 411         int err = 0;
 412
 413         spin_lock(&x->lock);
 414         if (x->km.state == XFRM_STATE_DEAD)
 415                 goto out;
 416         if (x->km.state == XFRM_STATE_EXPIRED)
 417                 goto expired;
 418         if (x->lft.hard_add_expires_seconds) {
 419                 long tmo = x->lft.hard_add_expires_seconds +
 420                         x->curlft.add_time - now;
 421                 if (tmo <= 0)
 422                         goto expired;
 423                 if (tmo < next)
 424                         next = tmo;
 425         }
 426         if (x->lft.hard_use_expires_seconds) {
 427                 long tmo = x->lft.hard_use_expires_seconds +
 428                         (x->curlft.use_time ? : now) - now;
 429                 if (tmo <= 0)
 430                         goto expired;
 431                 if (tmo < next)
 432                         next = tmo;
 433         }
 434         if (x->km.dying)
 435                 goto resched;
 436         if (x->lft.soft_add_expires_seconds) {
 437                 long tmo = x->lft.soft_add_expires_seconds +
 438                         x->curlft.add_time - now;
 439                 if (tmo <= 0)
 440                         warn = 1;
 441                 else if (tmo < next)
 442                         next = tmo;
 443         }
 444         if (x->lft.soft_use_expires_seconds) {
 445                 long tmo = x->lft.soft_use_expires_seconds +
 446                         (x->curlft.use_time ? : now) - now;
 447                 if (tmo <= 0)
 448                         warn = 1;
 449                 else if (tmo < next)
 450                         next = tmo;
 451         }
 452
 453         x->km.dying = warn;
 454         if (warn)
 455                 km_state_expired(x, 0, 0);
 456 resched:
 457         if (next != LONG_MAX){
 458                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 459         }
 460
 461         goto out;
 462
 463 expired:
 464         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) {
 465                 x->km.state = XFRM_STATE_EXPIRED;
 466                 wake_up(&net->xfrm.km_waitq);
 467                 next = 2;
 468                 goto resched;
 469         }
 470
 471         err = __xfrm_state_delete(x);
 472         if (!err && x->id.spi)
 473                 km_state_expired(x, 1, 0);
 474
 475         xfrm_audit_state_delete(x, err ? 0 : 1,
 476                                 audit_get_loginuid(current),
 477                                 audit_get_sessionid(current), 0);
 478
 479 out:
 480         spin_unlock(&x->lock);
 481         return HRTIMER_NORESTART;
 482 }
 483
 484 static void xfrm_replay_timer_handler(unsigned long data);
 485
 486 struct xfrm_state *xfrm_state_alloc(struct net *net)
 487 {
 488         struct xfrm_state *x;
 489
 490         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 491
 492         if (x) {
 493                 write_pnet(&x->xs_net, net);
 494                 atomic_set(&x->refcnt, 1);
 495                 atomic_set(&x->tunnel_users, 0);
 496                 INIT_LIST_HEAD(&x->km.all);
 497                 INIT_HLIST_NODE(&x->bydst);
 498                 INIT_HLIST_NODE(&x->bysrc);
 499                 INIT_HLIST_NODE(&x->byspi);
 500                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler, CLOCK_REALTIME, HRTIMER_MODE_ABS);
 501                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 502                                 (unsigned long)x);
 503                 x->curlft.add_time = get_seconds();
 504                 x->lft.soft_byte_limit = XFRM_INF;
 505                 x->lft.soft_packet_limit = XFRM_INF;
 506                 x->lft.hard_byte_limit = XFRM_INF;
 507                 x->lft.hard_packet_limit = XFRM_INF;
 508                 x->replay_maxage = 0;
 509                 x->replay_maxdiff = 0;
 510                 x->inner_mode = NULL;
 511                 x->inner_mode_iaf = NULL;
 512                 spin_lock_init(&x->lock);
 513         }
 514         return x;
 515 }
 516 EXPORT_SYMBOL(xfrm_state_alloc);
 517
 518 void __xfrm_state_destroy(struct xfrm_state *x)
 519 {
 520         struct net *net = xs_net(x);
 521
 522         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 523
 524         spin_lock_bh(&xfrm_state_gc_lock);
 525         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 526         spin_unlock_bh(&xfrm_state_gc_lock);
 527         schedule_work(&net->xfrm.state_gc_work);
 528 }
 529 EXPORT_SYMBOL(__xfrm_state_destroy);
 530
 531 int __xfrm_state_delete(struct xfrm_state *x)
 532 {
 533         struct net *net = xs_net(x);
 534         int err = -ESRCH;
 535
 536         if (x->km.state != XFRM_STATE_DEAD) {
 537                 x->km.state = XFRM_STATE_DEAD;
 538                 spin_lock(&xfrm_state_lock);
 539                 list_del(&x->km.all);
 540                 hlist_del(&x->bydst);
 541                 hlist_del(&x->bysrc);
 542                 if (x->id.spi)
 543                         hlist_del(&x->byspi);
 544                 net->xfrm.state_num--;
 545                 spin_unlock(&xfrm_state_lock);
 546
 547                 /* All xfrm_state objects are created by xfrm_state_alloc.
 548                  * The xfrm_state_alloc call gives a reference, and that
 549                  * is what we are dropping here.
 550                  */
 551                 xfrm_state_put(x);
 552                 err = 0;
 553         }
 554
 555         return err;
 556 }
 557 EXPORT_SYMBOL(__xfrm_state_delete);
 558
 559 int xfrm_state_delete(struct xfrm_state *x)
 560 {
 561         int err;
 562
 563         spin_lock_bh(&x->lock);
 564         err = __xfrm_state_delete(x);
 565         spin_unlock_bh(&x->lock);
 566
 567         return err;
 568 }
 569 EXPORT_SYMBOL(xfrm_state_delete);
 570
 571 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 572 static inline int
 573 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 574 {
 575         int i, err = 0;
 576
 577         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 578                 struct hlist_node *entry;
 579                 struct xfrm_state *x;
 580
 581                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
 582                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 583                            (err = security_xfrm_state_delete(x)) != 0) {
 584                                 xfrm_audit_state_delete(x, 0,
 585                                                         audit_info->loginuid,
 586                                                         audit_info->sessionid,
 587                                                         audit_info->secid);
 588                                 return err;
 589                         }
 590                 }
 591         }
 592
 593         return err;
 594 }
 595 #else
 596 static inline int
 597 xfrm_state_flush_secctx_check(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 598 {
 599         return 0;
 600 }
 601 #endif
 602
 603 int xfrm_state_flush(struct net *net, u8 proto, struct xfrm_audit *audit_info)
 604 {
 605         int i, err = 0, cnt = 0;
 606
 607         spin_lock_bh(&xfrm_state_lock);
 608         err = xfrm_state_flush_secctx_check(net, proto, audit_info);
 609         if (err)
 610                 goto out;
 611
 612         err = -ESRCH;
 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                                 if (!err)
 630                                         cnt++;
 631
 632                                 spin_lock_bh(&xfrm_state_lock);
 633                                 goto restart;
 634                         }
 635                 }
 636         }
 637         if (cnt)
 638                 err = 0;
 639
 640 out:
 641         spin_unlock_bh(&xfrm_state_lock);
 642         wake_up(&net->xfrm.km_waitq);
 643         return err;
 644 }
 645 EXPORT_SYMBOL(xfrm_state_flush);
 646
 647 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 648 {
 649         spin_lock_bh(&xfrm_state_lock);
 650         si->sadcnt = net->xfrm.state_num;
 651         si->sadhcnt = net->xfrm.state_hmask;
 652         si->sadhmcnt = xfrm_state_hashmax;
 653         spin_unlock_bh(&xfrm_state_lock);
 654 }
 655 EXPORT_SYMBOL(xfrm_sad_getinfo);
 656
 657 static int
 658 xfrm_init_tempsel(struct xfrm_state *x, struct flowi *fl,
 659                   struct xfrm_tmpl *tmpl,
 660                   xfrm_address_t *daddr, xfrm_address_t *saddr,
 661                   unsigned short family)
 662 {
 663         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 664         if (!afinfo)
 665                 return -1;
 666         afinfo->init_tempsel(x, fl, tmpl, daddr, saddr);
 667         xfrm_state_put_afinfo(afinfo);
 668         return 0;
 669 }
 670
 671 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi, u8 proto, unsigned short family)
 672 {
 673         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 674         struct xfrm_state *x;
 675         struct hlist_node *entry;
 676
 677         hlist_for_each_entry(x, entry, net->xfrm.state_byspi+h, byspi) {
 678                 if (x->props.family != family ||
 679                     x->id.spi       != spi ||
 680                     x->id.proto     != proto ||
 681                     xfrm_addr_cmp(&x->id.daddr, daddr, family))
 682                         continue;
 683
 684                 if ((mark & x->mark.m) != x->mark.v)
 685                         continue;
 686                 xfrm_state_hold(x);
 687                 return x;
 688         }
 689
 690         return NULL;
 691 }
 692
 693 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)
 694 {
 695         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 696         struct xfrm_state *x;
 697         struct hlist_node *entry;
 698
 699         hlist_for_each_entry(x, entry, net->xfrm.state_bysrc+h, bysrc) {
 700                 if (x->props.family != family ||
 701                     x->id.proto     != proto ||
 702                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
 703                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
 704                         continue;
 705
 706                 if ((mark & x->mark.m) != x->mark.v)
 707                         continue;
 708                 xfrm_state_hold(x);
 709                 return x;
 710         }
 711
 712         return NULL;
 713 }
 714
 715 static inline struct xfrm_state *
 716 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 717 {
 718         struct net *net = xs_net(x);
 719         u32 mark = x->mark.v & x->mark.m;
 720
 721         if (use_spi)
 722                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 723                                            x->id.spi, x->id.proto, family);
 724         else
 725                 return __xfrm_state_lookup_byaddr(net, mark,
 726                                                   &x->id.daddr,
 727                                                   &x->props.saddr,
 728                                                   x->id.proto, family);
 729 }
 730
 731 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 732 {
 733         if (have_hash_collision &&
 734             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 735             net->xfrm.state_num > net->xfrm.state_hmask)
 736                 schedule_work(&net->xfrm.state_hash_work);
 737 }
 738
 739 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 740                                struct flowi *fl, unsigned short family,
 741                                xfrm_address_t *daddr, xfrm_address_t *saddr,
 742                                struct xfrm_state **best, int *acq_in_progress,
 743                                int *error)
 744 {
 745         /* Resolution logic:
 746          * 1. There is a valid state with matching selector. Done.
 747          * 2. Valid state with inappropriate selector. Skip.
 748          *
 749          * Entering area of "sysdeps".
 750          *
 751          * 3. If state is not valid, selector is temporary, it selects
 752          *    only session which triggered previous resolution. Key
 753          *    manager will do something to install a state with proper
 754          *    selector.
 755          */
 756         if (x->km.state == XFRM_STATE_VALID) {
 757                 if ((x->sel.family &&
 758                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 759                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 760                         return;
 761
 762                 if (!*best ||
 763                     (*best)->km.dying > x->km.dying ||
 764                     ((*best)->km.dying == x->km.dying &&
 765                      (*best)->curlft.add_time < x->curlft.add_time))
 766                         *best = x;
 767         } else if (x->km.state == XFRM_STATE_ACQ) {
 768                 *acq_in_progress = 1;
 769         } else if (x->km.state == XFRM_STATE_ERROR ||
 770                    x->km.state == XFRM_STATE_EXPIRED) {
 771                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 772                     security_xfrm_state_pol_flow_match(x, pol, fl))
 773                         *error = -ESRCH;
 774         }
 775 }
 776
 777 struct xfrm_state *
 778 xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr,
 779                 struct flowi *fl, struct xfrm_tmpl *tmpl,
 780                 struct xfrm_policy *pol, int *err,
 781                 unsigned short family)
 782 {
 783         static xfrm_address_t saddr_wildcard = { };
 784         struct net *net = xp_net(pol);
 785         unsigned int h, h_wildcard;
 786         struct hlist_node *entry;
 787         struct xfrm_state *x, *x0, *to_put;
 788         int acquire_in_progress = 0;
 789         int error = 0;
 790         struct xfrm_state *best = NULL;
 791         u32 mark = pol->mark.v & pol->mark.m;
 792
 793         to_put = NULL;
 794
 795         spin_lock_bh(&xfrm_state_lock);
 796         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, family);
 797         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 798                 if (x->props.family == family &&
 799                     x->props.reqid == tmpl->reqid &&
 800                     (mark & x->mark.m) == x->mark.v &&
 801                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 802                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 803                     tmpl->mode == x->props.mode &&
 804                     tmpl->id.proto == x->id.proto &&
 805                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 806                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 807                                            &best, &acquire_in_progress, &error);
 808         }
 809         if (best)
 810                 goto found;
 811
 812         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, family);
 813         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h_wildcard, bydst) {
 814                 if (x->props.family == family &&
 815                     x->props.reqid == tmpl->reqid &&
 816                     (mark & x->mark.m) == x->mark.v &&
 817                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 818                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 819                     tmpl->mode == x->props.mode &&
 820                     tmpl->id.proto == x->id.proto &&
 821                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 822                         xfrm_state_look_at(pol, x, fl, family, daddr, saddr,
 823                                            &best, &acquire_in_progress, &error);
 824         }
 825
 826 found:
 827         x = best;
 828         if (!x && !error && !acquire_in_progress) {
 829                 if (tmpl->id.spi &&
 830                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 831                                               tmpl->id.proto, family)) != NULL) {
 832                         to_put = x0;
 833                         error = -EEXIST;
 834                         goto out;
 835                 }
 836                 x = xfrm_state_alloc(net);
 837                 if (x == NULL) {
 838                         error = -ENOMEM;
 839                         goto out;
 840                 }
 841                 /* Initialize temporary selector matching only
 842                  * to current session. */
 843                 xfrm_init_tempsel(x, fl, tmpl, daddr, saddr, family);
 844                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 845
 846                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->secid);
 847                 if (error) {
 848                         x->km.state = XFRM_STATE_DEAD;
 849                         to_put = x;
 850                         x = NULL;
 851                         goto out;
 852                 }
 853
 854                 if (km_query(x, tmpl, pol) == 0) {
 855                         x->km.state = XFRM_STATE_ACQ;
 856                         list_add(&x->km.all, &net->xfrm.state_all);
 857                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 858                         h = xfrm_src_hash(net, daddr, saddr, family);
 859                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 860                         if (x->id.spi) {
 861                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family);
 862                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 863                         }
 864                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 865                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 866                         net->xfrm.state_num++;
 867                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 868                 } else {
 869                         x->km.state = XFRM_STATE_DEAD;
 870                         to_put = x;
 871                         x = NULL;
 872                         error = -ESRCH;
 873                 }
 874         }
 875 out:
 876         if (x)
 877                 xfrm_state_hold(x);
 878         else
 879                 *err = acquire_in_progress ? -EAGAIN : error;
 880         spin_unlock_bh(&xfrm_state_lock);
 881         if (to_put)
 882                 xfrm_state_put(to_put);
 883         return x;
 884 }
 885
 886 struct xfrm_state *
 887 xfrm_stateonly_find(struct net *net, u32 mark,
 888                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 889                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 890 {
 891         unsigned int h;
 892         struct xfrm_state *rx = NULL, *x = NULL;
 893         struct hlist_node *entry;
 894
 895         spin_lock(&xfrm_state_lock);
 896         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 897         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 898                 if (x->props.family == family &&
 899                     x->props.reqid == reqid &&
 900                     (mark & x->mark.m) == x->mark.v &&
 901                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 902                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 903                     mode == x->props.mode &&
 904                     proto == x->id.proto &&
 905                     x->km.state == XFRM_STATE_VALID) {
 906                         rx = x;
 907                         break;
 908                 }
 909         }
 910
 911         if (rx)
 912                 xfrm_state_hold(rx);
 913         spin_unlock(&xfrm_state_lock);
 914
 915
 916         return rx;
 917 }
 918 EXPORT_SYMBOL(xfrm_stateonly_find);
 919
 920 static void __xfrm_state_insert(struct xfrm_state *x)
 921 {
 922         struct net *net = xs_net(x);
 923         unsigned int h;
 924
 925         list_add(&x->km.all, &net->xfrm.state_all);
 926
 927         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 928                           x->props.reqid, x->props.family);
 929         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 930
 931         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 932         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 933
 934         if (x->id.spi) {
 935                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 936                                   x->props.family);
 937
 938                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 939         }
 940
 941         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 942         if (x->replay_maxage)
 943                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 944
 945         wake_up(&net->xfrm.km_waitq);
 946
 947         net->xfrm.state_num++;
 948
 949         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 950 }
 951
 952 /* xfrm_state_lock is held */
 953 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 954 {
 955         struct net *net = xs_net(xnew);
 956         unsigned short family = xnew->props.family;
 957         u32 reqid = xnew->props.reqid;
 958         struct xfrm_state *x;
 959         struct hlist_node *entry;
 960         unsigned int h;
 961         u32 mark = xnew->mark.v & xnew->mark.m;
 962
 963         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 964         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 965                 if (x->props.family     == family &&
 966                     x->props.reqid      == reqid &&
 967                     (mark & x->mark.m) == x->mark.v &&
 968                     !xfrm_addr_cmp(&x->id.daddr, &xnew->id.daddr, family) &&
 969                     !xfrm_addr_cmp(&x->props.saddr, &xnew->props.saddr, family))
 970                         x->genid++;
 971         }
 972 }
 973
 974 void xfrm_state_insert(struct xfrm_state *x)
 975 {
 976         spin_lock_bh(&xfrm_state_lock);
 977         __xfrm_state_bump_genids(x);
 978         __xfrm_state_insert(x);
 979         spin_unlock_bh(&xfrm_state_lock);
 980 }
 981 EXPORT_SYMBOL(xfrm_state_insert);
 982
 983 /* xfrm_state_lock is held */
 984 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)
 985 {
 986         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 987         struct hlist_node *entry;
 988         struct xfrm_state *x;
 989         u32 mark = m->v & m->m;
 990
 991         hlist_for_each_entry(x, entry, net->xfrm.state_bydst+h, bydst) {
 992                 if (x->props.reqid  != reqid ||
 993                     x->props.mode   != mode ||
 994                     x->props.family != family ||
 995                     x->km.state     != XFRM_STATE_ACQ ||
 996                     x->id.spi       != 0 ||
 997                     x->id.proto     != proto ||
 998                     (mark & x->mark.m) != x->mark.v ||
 999                     xfrm_addr_cmp(&x->id.daddr, daddr, family) ||
1000                     xfrm_addr_cmp(&x->props.saddr, saddr, family))
1001                         continue;
1002
1003                 xfrm_state_hold(x);
1004                 return x;
1005         }
1006
1007         if (!create)
1008                 return NULL;
1009
1010         x = xfrm_state_alloc(net);
1011         if (likely(x)) {
1012                 switch (family) {
1013                 case AF_INET:
1014                         x->sel.daddr.a4 = daddr->a4;
1015                         x->sel.saddr.a4 = saddr->a4;
1016                         x->sel.prefixlen_d = 32;
1017                         x->sel.prefixlen_s = 32;
1018                         x->props.saddr.a4 = saddr->a4;
1019                         x->id.daddr.a4 = daddr->a4;
1020                         break;
1021
1022                 case AF_INET6:
1023                         ipv6_addr_copy((struct in6_addr *)x->sel.daddr.a6,
1024                                        (struct in6_addr *)daddr);
1025                         ipv6_addr_copy((struct in6_addr *)x->sel.saddr.a6,
1026                                        (struct in6_addr *)saddr);
1027                         x->sel.prefixlen_d = 128;
1028                         x->sel.prefixlen_s = 128;
1029                         ipv6_addr_copy((struct in6_addr *)x->props.saddr.a6,
1030                                        (struct in6_addr *)saddr);
1031                         ipv6_addr_copy((struct in6_addr *)x->id.daddr.a6,
1032                                        (struct in6_addr *)daddr);
1033                         break;
1034                 }
1035
1036                 x->km.state = XFRM_STATE_ACQ;
1037                 x->id.proto = proto;
1038                 x->props.family = family;
1039                 x->props.mode = mode;
1040                 x->props.reqid = reqid;
1041                 x->mark.v = m->v;
1042                 x->mark.m = m->m;
1043                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1044                 xfrm_state_hold(x);
1045                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1046                 list_add(&x->km.all, &net->xfrm.state_all);
1047                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1048                 h = xfrm_src_hash(net, daddr, saddr, family);
1049                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1050
1051                 net->xfrm.state_num++;
1052
1053                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1054         }
1055
1056         return x;
1057 }
1058
1059 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1060
1061 int xfrm_state_add(struct xfrm_state *x)
1062 {
1063         struct net *net = xs_net(x);
1064         struct xfrm_state *x1, *to_put;
1065         int family;
1066         int err;
1067         u32 mark = x->mark.v & x->mark.m;
1068         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1069
1070         family = x->props.family;
1071
1072         to_put = NULL;
1073
1074         spin_lock_bh(&xfrm_state_lock);
1075
1076         x1 = __xfrm_state_locate(x, use_spi, family);
1077         if (x1) {
1078                 to_put = x1;
1079                 x1 = NULL;
1080                 err = -EEXIST;
1081                 goto out;
1082         }
1083
1084         if (use_spi && x->km.seq) {
1085                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1086                 if (x1 && ((x1->id.proto != x->id.proto) ||
1087                     xfrm_addr_cmp(&x1->id.daddr, &x->id.daddr, family))) {
1088                         to_put = x1;
1089                         x1 = NULL;
1090                 }
1091         }
1092
1093         if (use_spi && !x1)
1094                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1095                                      x->props.reqid, x->id.proto,
1096                                      &x->id.daddr, &x->props.saddr, 0);
1097
1098         __xfrm_state_bump_genids(x);
1099         __xfrm_state_insert(x);
1100         err = 0;
1101
1102 out:
1103         spin_unlock_bh(&xfrm_state_lock);
1104
1105         if (x1) {
1106                 xfrm_state_delete(x1);
1107                 xfrm_state_put(x1);
1108         }
1109
1110         if (to_put)
1111                 xfrm_state_put(to_put);
1112
1113         return err;
1114 }
1115 EXPORT_SYMBOL(xfrm_state_add);
1116
1117 #ifdef CONFIG_XFRM_MIGRATE
1118 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, int *errp)
1119 {
1120         struct net *net = xs_net(orig);
1121         int err = -ENOMEM;
1122         struct xfrm_state *x = xfrm_state_alloc(net);
1123         if (!x)
1124                 goto out;
1125
1126         memcpy(&x->id, &orig->id, sizeof(x->id));
1127         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1128         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1129         x->props.mode = orig->props.mode;
1130         x->props.replay_window = orig->props.replay_window;
1131         x->props.reqid = orig->props.reqid;
1132         x->props.family = orig->props.family;
1133         x->props.saddr = orig->props.saddr;
1134
1135         if (orig->aalg) {
1136                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1137                 if (!x->aalg)
1138                         goto error;
1139         }
1140         x->props.aalgo = orig->props.aalgo;
1141
1142         if (orig->ealg) {
1143                 x->ealg = xfrm_algo_clone(orig->ealg);
1144                 if (!x->ealg)
1145                         goto error;
1146         }
1147         x->props.ealgo = orig->props.ealgo;
1148
1149         if (orig->calg) {
1150                 x->calg = xfrm_algo_clone(orig->calg);
1151                 if (!x->calg)
1152                         goto error;
1153         }
1154         x->props.calgo = orig->props.calgo;
1155
1156         if (orig->encap) {
1157                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1158                 if (!x->encap)
1159                         goto error;
1160         }
1161
1162         if (orig->coaddr) {
1163                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1164                                     GFP_KERNEL);
1165                 if (!x->coaddr)
1166                         goto error;
1167         }
1168
1169         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1170
1171         err = xfrm_init_state(x);
1172         if (err)
1173                 goto error;
1174
1175         x->props.flags = orig->props.flags;
1176
1177         x->curlft.add_time = orig->curlft.add_time;
1178         x->km.state = orig->km.state;
1179         x->km.seq = orig->km.seq;
1180
1181         return x;
1182
1183  error:
1184         xfrm_state_put(x);
1185 out:
1186         if (errp)
1187                 *errp = err;
1188         return NULL;
1189 }
1190
1191 /* xfrm_state_lock is held */
1192 struct xfrm_state * xfrm_migrate_state_find(struct xfrm_migrate *m)
1193 {
1194         unsigned int h;
1195         struct xfrm_state *x;
1196         struct hlist_node *entry;
1197
1198         if (m->reqid) {
1199                 h = xfrm_dst_hash(&init_net, &m->old_daddr, &m->old_saddr,
1200                                   m->reqid, m->old_family);
1201                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bydst+h, bydst) {
1202                         if (x->props.mode != m->mode ||
1203                             x->id.proto != m->proto)
1204                                 continue;
1205                         if (m->reqid && x->props.reqid != m->reqid)
1206                                 continue;
1207                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1208                                           m->old_family) ||
1209                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1210                                           m->old_family))
1211                                 continue;
1212                         xfrm_state_hold(x);
1213                         return x;
1214                 }
1215         } else {
1216                 h = xfrm_src_hash(&init_net, &m->old_daddr, &m->old_saddr,
1217                                   m->old_family);
1218                 hlist_for_each_entry(x, entry, init_net.xfrm.state_bysrc+h, bysrc) {
1219                         if (x->props.mode != m->mode ||
1220                             x->id.proto != m->proto)
1221                                 continue;
1222                         if (xfrm_addr_cmp(&x->id.daddr, &m->old_daddr,
1223                                           m->old_family) ||
1224                             xfrm_addr_cmp(&x->props.saddr, &m->old_saddr,
1225                                           m->old_family))
1226                                 continue;
1227                         xfrm_state_hold(x);
1228                         return x;
1229                 }
1230         }
1231
1232         return NULL;
1233 }
1234 EXPORT_SYMBOL(xfrm_migrate_state_find);
1235
1236 struct xfrm_state * xfrm_state_migrate(struct xfrm_state *x,
1237                                        struct xfrm_migrate *m)
1238 {
1239         struct xfrm_state *xc;
1240         int err;
1241
1242         xc = xfrm_state_clone(x, &err);
1243         if (!xc)
1244                 return NULL;
1245
1246         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1247         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1248
1249         /* add state */
1250         if (!xfrm_addr_cmp(&x->id.daddr, &m->new_daddr, m->new_family)) {
1251                 /* a care is needed when the destination address of the
1252                    state is to be updated as it is a part of triplet */
1253                 xfrm_state_insert(xc);
1254         } else {
1255                 if ((err = xfrm_state_add(xc)) < 0)
1256                         goto error;
1257         }
1258
1259         return xc;
1260 error:
1261         kfree(xc);
1262         return NULL;
1263 }
1264 EXPORT_SYMBOL(xfrm_state_migrate);
1265 #endif
1266
1267 int xfrm_state_update(struct xfrm_state *x)
1268 {
1269         struct xfrm_state *x1, *to_put;
1270         int err;
1271         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1272
1273         to_put = NULL;
1274
1275         spin_lock_bh(&xfrm_state_lock);
1276         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1277
1278         err = -ESRCH;
1279         if (!x1)
1280                 goto out;
1281
1282         if (xfrm_state_kern(x1)) {
1283                 to_put = x1;
1284                 err = -EEXIST;
1285                 goto out;
1286         }
1287
1288         if (x1->km.state == XFRM_STATE_ACQ) {
1289                 __xfrm_state_insert(x);
1290                 x = NULL;
1291         }
1292         err = 0;
1293
1294 out:
1295         spin_unlock_bh(&xfrm_state_lock);
1296
1297         if (to_put)
1298                 xfrm_state_put(to_put);
1299
1300         if (err)
1301                 return err;
1302
1303         if (!x) {
1304                 xfrm_state_delete(x1);
1305                 xfrm_state_put(x1);
1306                 return 0;
1307         }
1308
1309         err = -EINVAL;
1310         spin_lock_bh(&x1->lock);
1311         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1312                 if (x->encap && x1->encap)
1313                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1314                 if (x->coaddr && x1->coaddr) {
1315                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1316                 }
1317                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1318                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1319                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1320                 x1->km.dying = 0;
1321
1322                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1323                 if (x1->curlft.use_time)
1324                         xfrm_state_check_expire(x1);
1325
1326                 err = 0;
1327         }
1328         spin_unlock_bh(&x1->lock);
1329
1330         xfrm_state_put(x1);
1331
1332         return err;
1333 }
1334 EXPORT_SYMBOL(xfrm_state_update);
1335
1336 int xfrm_state_check_expire(struct xfrm_state *x)
1337 {
1338         if (!x->curlft.use_time)
1339                 x->curlft.use_time = get_seconds();
1340
1341         if (x->km.state != XFRM_STATE_VALID)
1342                 return -EINVAL;
1343
1344         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1345             x->curlft.packets >= x->lft.hard_packet_limit) {
1346                 x->km.state = XFRM_STATE_EXPIRED;
1347                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0,0), HRTIMER_MODE_REL);
1348                 return -EINVAL;
1349         }
1350
1351         if (!x->km.dying &&
1352             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1353              x->curlft.packets >= x->lft.soft_packet_limit)) {
1354                 x->km.dying = 1;
1355                 km_state_expired(x, 0, 0);
1356         }
1357         return 0;
1358 }
1359 EXPORT_SYMBOL(xfrm_state_check_expire);
1360
1361 struct xfrm_state *
1362 xfrm_state_lookup(struct net *net, u32 mark, xfrm_address_t *daddr, __be32 spi,
1363                   u8 proto, unsigned short family)
1364 {
1365         struct xfrm_state *x;
1366
1367         spin_lock_bh(&xfrm_state_lock);
1368         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1369         spin_unlock_bh(&xfrm_state_lock);
1370         return x;
1371 }
1372 EXPORT_SYMBOL(xfrm_state_lookup);
1373
1374 struct xfrm_state *
1375 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1376                          xfrm_address_t *daddr, xfrm_address_t *saddr,
1377                          u8 proto, unsigned short family)
1378 {
1379         struct xfrm_state *x;
1380
1381         spin_lock_bh(&xfrm_state_lock);
1382         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1383         spin_unlock_bh(&xfrm_state_lock);
1384         return x;
1385 }
1386 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1387
1388 struct xfrm_state *
1389 xfrm_find_acq(struct net *net, struct xfrm_mark *mark, u8 mode, u32 reqid, u8 proto,
1390               xfrm_address_t *daddr, xfrm_address_t *saddr,
1391               int create, unsigned short family)
1392 {
1393         struct xfrm_state *x;
1394
1395         spin_lock_bh(&xfrm_state_lock);
1396         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1397         spin_unlock_bh(&xfrm_state_lock);
1398
1399         return x;
1400 }
1401 EXPORT_SYMBOL(xfrm_find_acq);
1402
1403 #ifdef CONFIG_XFRM_SUB_POLICY
1404 int
1405 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1406                unsigned short family)
1407 {
1408         int err = 0;
1409         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1410         if (!afinfo)
1411                 return -EAFNOSUPPORT;
1412
1413         spin_lock_bh(&xfrm_state_lock);
1414         if (afinfo->tmpl_sort)
1415                 err = afinfo->tmpl_sort(dst, src, n);
1416         spin_unlock_bh(&xfrm_state_lock);
1417         xfrm_state_put_afinfo(afinfo);
1418         return err;
1419 }
1420 EXPORT_SYMBOL(xfrm_tmpl_sort);
1421
1422 int
1423 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1424                 unsigned short family)
1425 {
1426         int err = 0;
1427         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1428         if (!afinfo)
1429                 return -EAFNOSUPPORT;
1430
1431         spin_lock_bh(&xfrm_state_lock);
1432         if (afinfo->state_sort)
1433                 err = afinfo->state_sort(dst, src, n);
1434         spin_unlock_bh(&xfrm_state_lock);
1435         xfrm_state_put_afinfo(afinfo);
1436         return err;
1437 }
1438 EXPORT_SYMBOL(xfrm_state_sort);
1439 #endif
1440
1441 /* Silly enough, but I'm lazy to build resolution list */
1442
1443 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1444 {
1445         int i;
1446
1447         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1448                 struct hlist_node *entry;
1449                 struct xfrm_state *x;
1450
1451                 hlist_for_each_entry(x, entry, net->xfrm.state_bydst+i, bydst) {
1452                         if (x->km.seq == seq &&
1453                             (mark & x->mark.m) == x->mark.v &&
1454                             x->km.state == XFRM_STATE_ACQ) {
1455                                 xfrm_state_hold(x);
1456                                 return x;
1457                         }
1458                 }
1459         }
1460         return NULL;
1461 }
1462
1463 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1464 {
1465         struct xfrm_state *x;
1466
1467         spin_lock_bh(&xfrm_state_lock);
1468         x = __xfrm_find_acq_byseq(net, mark, seq);
1469         spin_unlock_bh(&xfrm_state_lock);
1470         return x;
1471 }
1472 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1473
1474 u32 xfrm_get_acqseq(void)
1475 {
1476         u32 res;
1477         static atomic_t acqseq;
1478
1479         do {
1480                 res = atomic_inc_return(&acqseq);
1481         } while (!res);
1482
1483         return res;
1484 }
1485 EXPORT_SYMBOL(xfrm_get_acqseq);
1486
1487 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1488 {
1489         struct net *net = xs_net(x);
1490         unsigned int h;
1491         struct xfrm_state *x0;
1492         int err = -ENOENT;
1493         __be32 minspi = htonl(low);
1494         __be32 maxspi = htonl(high);
1495         u32 mark = x->mark.v & x->mark.m;
1496
1497         spin_lock_bh(&x->lock);
1498         if (x->km.state == XFRM_STATE_DEAD)
1499                 goto unlock;
1500
1501         err = 0;
1502         if (x->id.spi)
1503                 goto unlock;
1504
1505         err = -ENOENT;
1506
1507         if (minspi == maxspi) {
1508                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1509                 if (x0) {
1510                         xfrm_state_put(x0);
1511                         goto unlock;
1512                 }
1513                 x->id.spi = minspi;
1514         } else {
1515                 u32 spi = 0;
1516                 for (h=0; h<high-low+1; h++) {
1517                         spi = low + net_random()%(high-low+1);
1518                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1519                         if (x0 == NULL) {
1520                                 x->id.spi = htonl(spi);
1521                                 break;
1522                         }
1523                         xfrm_state_put(x0);
1524                 }
1525         }
1526         if (x->id.spi) {
1527                 spin_lock_bh(&xfrm_state_lock);
1528                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1529                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1530                 spin_unlock_bh(&xfrm_state_lock);
1531
1532                 err = 0;
1533         }
1534
1535 unlock:
1536         spin_unlock_bh(&x->lock);
1537
1538         return err;
1539 }
1540 EXPORT_SYMBOL(xfrm_alloc_spi);
1541
1542 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1543                     int (*func)(struct xfrm_state *, int, void*),
1544                     void *data)
1545 {
1546         struct xfrm_state *state;
1547         struct xfrm_state_walk *x;
1548         int err = 0;
1549
1550         if (walk->seq != 0 && list_empty(&walk->all))
1551                 return 0;
1552
1553         spin_lock_bh(&xfrm_state_lock);
1554         if (list_empty(&walk->all))
1555                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1556         else
1557                 x = list_entry(&walk->all, struct xfrm_state_walk, all);
1558         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1559                 if (x->state == XFRM_STATE_DEAD)
1560                         continue;
1561                 state = container_of(x, struct xfrm_state, km);
1562                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1563                         continue;
1564                 err = func(state, walk->seq, data);
1565                 if (err) {
1566                         list_move_tail(&walk->all, &x->all);
1567                         goto out;
1568                 }
1569                 walk->seq++;
1570         }
1571         if (walk->seq == 0) {
1572                 err = -ENOENT;
1573                 goto out;
1574         }
1575         list_del_init(&walk->all);
1576 out:
1577         spin_unlock_bh(&xfrm_state_lock);
1578         return err;
1579 }
1580 EXPORT_SYMBOL(xfrm_state_walk);
1581
1582 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto)
1583 {
1584         INIT_LIST_HEAD(&walk->all);
1585         walk->proto = proto;
1586         walk->state = XFRM_STATE_DEAD;
1587         walk->seq = 0;
1588 }
1589 EXPORT_SYMBOL(xfrm_state_walk_init);
1590
1591 void xfrm_state_walk_done(struct xfrm_state_walk *walk)
1592 {
1593         if (list_empty(&walk->all))
1594                 return;
1595
1596         spin_lock_bh(&xfrm_state_lock);
1597         list_del(&walk->all);
1598         spin_unlock_bh(&xfrm_state_lock);
1599 }
1600 EXPORT_SYMBOL(xfrm_state_walk_done);
1601
1602
1603 void xfrm_replay_notify(struct xfrm_state *x, int event)
1604 {
1605         struct km_event c;
1606         /* we send notify messages in case
1607          *  1. we updated on of the sequence numbers, and the seqno difference
1608          *     is at least x->replay_maxdiff, in this case we also update the
1609          *     timeout of our timer function
1610          *  2. if x->replay_maxage has elapsed since last update,
1611          *     and there were changes
1612          *
1613          *  The state structure must be locked!
1614          */
1615
1616         switch (event) {
1617         case XFRM_REPLAY_UPDATE:
1618                 if (x->replay_maxdiff &&
1619                     (x->replay.seq - x->preplay.seq < x->replay_maxdiff) &&
1620                     (x->replay.oseq - x->preplay.oseq < x->replay_maxdiff)) {
1621                         if (x->xflags & XFRM_TIME_DEFER)
1622                                 event = XFRM_REPLAY_TIMEOUT;
1623                         else
1624                                 return;
1625                 }
1626
1627                 break;
1628
1629         case XFRM_REPLAY_TIMEOUT:
1630                 if ((x->replay.seq == x->preplay.seq) &&
1631                     (x->replay.bitmap == x->preplay.bitmap) &&
1632                     (x->replay.oseq == x->preplay.oseq)) {
1633                         x->xflags |= XFRM_TIME_DEFER;
1634                         return;
1635                 }
1636
1637                 break;
1638         }
1639
1640         memcpy(&x->preplay, &x->replay, sizeof(struct xfrm_replay_state));
1641         c.event = XFRM_MSG_NEWAE;
1642         c.data.aevent = event;
1643         km_state_notify(x, &c);
1644
1645         if (x->replay_maxage &&
1646             !mod_timer(&x->rtimer, jiffies + x->replay_maxage))
1647                 x->xflags &= ~XFRM_TIME_DEFER;
1648 }
1649
1650 static void xfrm_replay_timer_handler(unsigned long data)
1651 {
1652         struct xfrm_state *x = (struct xfrm_state*)data;
1653
1654         spin_lock(&x->lock);
1655
1656         if (x->km.state == XFRM_STATE_VALID) {
1657                 if (xfrm_aevent_is_on(xs_net(x)))
1658                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
1659                 else
1660                         x->xflags |= XFRM_TIME_DEFER;
1661         }
1662
1663         spin_unlock(&x->lock);
1664 }
1665
1666 int xfrm_replay_check(struct xfrm_state *x,
1667                       struct sk_buff *skb, __be32 net_seq)
1668 {
1669         u32 diff;
1670         u32 seq = ntohl(net_seq);
1671
1672         if (unlikely(seq == 0))
1673                 goto err;
1674
1675         if (likely(seq > x->replay.seq))
1676                 return 0;
1677
1678         diff = x->replay.seq - seq;
1679         if (diff >= min_t(unsigned int, x->props.replay_window,
1680                           sizeof(x->replay.bitmap) * 8)) {
1681                 x->stats.replay_window++;
1682                 goto err;
1683         }
1684
1685         if (x->replay.bitmap & (1U << diff)) {
1686                 x->stats.replay++;
1687                 goto err;
1688         }
1689         return 0;
1690
1691 err:
1692         xfrm_audit_state_replay(x, skb, net_seq);
1693         return -EINVAL;
1694 }
1695
1696 void xfrm_replay_advance(struct xfrm_state *x, __be32 net_seq)
1697 {
1698         u32 diff;
1699         u32 seq = ntohl(net_seq);
1700
1701         if (seq > x->replay.seq) {
1702                 diff = seq - x->replay.seq;
1703                 if (diff < x->props.replay_window)
1704                         x->replay.bitmap = ((x->replay.bitmap) << diff) | 1;
1705                 else
1706                         x->replay.bitmap = 1;
1707                 x->replay.seq = seq;
1708         } else {
1709                 diff = x->replay.seq - seq;
1710                 x->replay.bitmap |= (1U << diff);
1711         }
1712
1713         if (xfrm_aevent_is_on(xs_net(x)))
1714                 xfrm_replay_notify(x, XFRM_REPLAY_UPDATE);
1715 }
1716
1717 static LIST_HEAD(xfrm_km_list);
1718 static DEFINE_RWLOCK(xfrm_km_lock);
1719
1720 void km_policy_notify(struct xfrm_policy *xp, int dir, struct km_event *c)
1721 {
1722         struct xfrm_mgr *km;
1723
1724         read_lock(&xfrm_km_lock);
1725         list_for_each_entry(km, &xfrm_km_list, list)
1726                 if (km->notify_policy)
1727                         km->notify_policy(xp, dir, c);
1728         read_unlock(&xfrm_km_lock);
1729 }
1730
1731 void km_state_notify(struct xfrm_state *x, struct km_event *c)
1732 {
1733         struct xfrm_mgr *km;
1734         read_lock(&xfrm_km_lock);
1735         list_for_each_entry(km, &xfrm_km_list, list)
1736                 if (km->notify)
1737                         km->notify(x, c);
1738         read_unlock(&xfrm_km_lock);
1739 }
1740
1741 EXPORT_SYMBOL(km_policy_notify);
1742 EXPORT_SYMBOL(km_state_notify);
1743
1744 void km_state_expired(struct xfrm_state *x, int hard, u32 pid)
1745 {
1746         struct net *net = xs_net(x);
1747         struct km_event c;
1748
1749         c.data.hard = hard;
1750         c.pid = pid;
1751         c.event = XFRM_MSG_EXPIRE;
1752         km_state_notify(x, &c);
1753
1754         if (hard)
1755                 wake_up(&net->xfrm.km_waitq);
1756 }
1757
1758 EXPORT_SYMBOL(km_state_expired);
1759 /*
1760  * We send to all registered managers regardless of failure
1761  * We are happy with one success
1762 */
1763 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1764 {
1765         int err = -EINVAL, acqret;
1766         struct xfrm_mgr *km;
1767
1768         read_lock(&xfrm_km_lock);
1769         list_for_each_entry(km, &xfrm_km_list, list) {
1770                 acqret = km->acquire(x, t, pol, XFRM_POLICY_OUT);
1771                 if (!acqret)
1772                         err = acqret;
1773         }
1774         read_unlock(&xfrm_km_lock);
1775         return err;
1776 }
1777 EXPORT_SYMBOL(km_query);
1778
1779 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1780 {
1781         int err = -EINVAL;
1782         struct xfrm_mgr *km;
1783
1784         read_lock(&xfrm_km_lock);
1785         list_for_each_entry(km, &xfrm_km_list, list) {
1786                 if (km->new_mapping)
1787                         err = km->new_mapping(x, ipaddr, sport);
1788                 if (!err)
1789                         break;
1790         }
1791         read_unlock(&xfrm_km_lock);
1792         return err;
1793 }
1794 EXPORT_SYMBOL(km_new_mapping);
1795
1796 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 pid)
1797 {
1798         struct net *net = xp_net(pol);
1799         struct km_event c;
1800
1801         c.data.hard = hard;
1802         c.pid = pid;
1803         c.event = XFRM_MSG_POLEXPIRE;
1804         km_policy_notify(pol, dir, &c);
1805
1806         if (hard)
1807                 wake_up(&net->xfrm.km_waitq);
1808 }
1809 EXPORT_SYMBOL(km_policy_expired);
1810
1811 #ifdef CONFIG_XFRM_MIGRATE
1812 int km_migrate(struct xfrm_selector *sel, u8 dir, u8 type,
1813                struct xfrm_migrate *m, int num_migrate,
1814                struct xfrm_kmaddress *k)
1815 {
1816         int err = -EINVAL;
1817         int ret;
1818         struct xfrm_mgr *km;
1819
1820         read_lock(&xfrm_km_lock);
1821         list_for_each_entry(km, &xfrm_km_list, list) {
1822                 if (km->migrate) {
1823                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1824                         if (!ret)
1825                                 err = ret;
1826                 }
1827         }
1828         read_unlock(&xfrm_km_lock);
1829         return err;
1830 }
1831 EXPORT_SYMBOL(km_migrate);
1832 #endif
1833
1834 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1835 {
1836         int err = -EINVAL;
1837         int ret;
1838         struct xfrm_mgr *km;
1839
1840         read_lock(&xfrm_km_lock);
1841         list_for_each_entry(km, &xfrm_km_list, list) {
1842                 if (km->report) {
1843                         ret = km->report(net, proto, sel, addr);
1844                         if (!ret)
1845                                 err = ret;
1846                 }
1847         }
1848         read_unlock(&xfrm_km_lock);
1849         return err;
1850 }
1851 EXPORT_SYMBOL(km_report);
1852
1853 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1854 {
1855         int err;
1856         u8 *data;
1857         struct xfrm_mgr *km;
1858         struct xfrm_policy *pol = NULL;
1859
1860         if (optlen <= 0 || optlen > PAGE_SIZE)
1861                 return -EMSGSIZE;
1862
1863         data = kmalloc(optlen, GFP_KERNEL);
1864         if (!data)
1865                 return -ENOMEM;
1866
1867         err = -EFAULT;
1868         if (copy_from_user(data, optval, optlen))
1869                 goto out;
1870
1871         err = -EINVAL;
1872         read_lock(&xfrm_km_lock);
1873         list_for_each_entry(km, &xfrm_km_list, list) {
1874                 pol = km->compile_policy(sk, optname, data,
1875                                          optlen, &err);
1876                 if (err >= 0)
1877                         break;
1878         }
1879         read_unlock(&xfrm_km_lock);
1880
1881         if (err >= 0) {
1882                 xfrm_sk_policy_insert(sk, err, pol);
1883                 xfrm_pol_put(pol);
1884                 err = 0;
1885         }
1886
1887 out:
1888         kfree(data);
1889         return err;
1890 }
1891 EXPORT_SYMBOL(xfrm_user_policy);
1892
1893 int xfrm_register_km(struct xfrm_mgr *km)
1894 {
1895         write_lock_bh(&xfrm_km_lock);
1896         list_add_tail(&km->list, &xfrm_km_list);
1897         write_unlock_bh(&xfrm_km_lock);
1898         return 0;
1899 }
1900 EXPORT_SYMBOL(xfrm_register_km);
1901
1902 int xfrm_unregister_km(struct xfrm_mgr *km)
1903 {
1904         write_lock_bh(&xfrm_km_lock);
1905         list_del(&km->list);
1906         write_unlock_bh(&xfrm_km_lock);
1907         return 0;
1908 }
1909 EXPORT_SYMBOL(xfrm_unregister_km);
1910
1911 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1912 {
1913         int err = 0;
1914         if (unlikely(afinfo == NULL))
1915                 return -EINVAL;
1916         if (unlikely(afinfo->family >= NPROTO))
1917                 return -EAFNOSUPPORT;
1918         write_lock_bh(&xfrm_state_afinfo_lock);
1919         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1920                 err = -ENOBUFS;
1921         else
1922                 xfrm_state_afinfo[afinfo->family] = afinfo;
1923         write_unlock_bh(&xfrm_state_afinfo_lock);
1924         return err;
1925 }
1926 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1927
1928 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1929 {
1930         int err = 0;
1931         if (unlikely(afinfo == NULL))
1932                 return -EINVAL;
1933         if (unlikely(afinfo->family >= NPROTO))
1934                 return -EAFNOSUPPORT;
1935         write_lock_bh(&xfrm_state_afinfo_lock);
1936         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1937                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1938                         err = -EINVAL;
1939                 else
1940                         xfrm_state_afinfo[afinfo->family] = NULL;
1941         }
1942         write_unlock_bh(&xfrm_state_afinfo_lock);
1943         return err;
1944 }
1945 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1946
1947 static struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1948 {
1949         struct xfrm_state_afinfo *afinfo;
1950         if (unlikely(family >= NPROTO))
1951                 return NULL;
1952         read_lock(&xfrm_state_afinfo_lock);
1953         afinfo = xfrm_state_afinfo[family];
1954         if (unlikely(!afinfo))
1955                 read_unlock(&xfrm_state_afinfo_lock);
1956         return afinfo;
1957 }
1958
1959 static void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1960         __releases(xfrm_state_afinfo_lock)
1961 {
1962         read_unlock(&xfrm_state_afinfo_lock);
1963 }
1964
1965 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1966 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1967 {
1968         if (x->tunnel) {
1969                 struct xfrm_state *t = x->tunnel;
1970
1971                 if (atomic_read(&t->tunnel_users) == 2)
1972                         xfrm_state_delete(t);
1973                 atomic_dec(&t->tunnel_users);
1974                 xfrm_state_put(t);
1975                 x->tunnel = NULL;
1976         }
1977 }
1978 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1979
1980 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1981 {
1982         int res;
1983
1984         spin_lock_bh(&x->lock);
1985         if (x->km.state == XFRM_STATE_VALID &&
1986             x->type && x->type->get_mtu)
1987                 res = x->type->get_mtu(x, mtu);
1988         else
1989                 res = mtu - x->props.header_len;
1990         spin_unlock_bh(&x->lock);
1991         return res;
1992 }
1993
1994 int xfrm_init_state(struct xfrm_state *x)
1995 {
1996         struct xfrm_state_afinfo *afinfo;
1997         struct xfrm_mode *inner_mode;
1998         int family = x->props.family;
1999         int err;
2000
2001         err = -EAFNOSUPPORT;
2002         afinfo = xfrm_state_get_afinfo(family);
2003         if (!afinfo)
2004                 goto error;
2005
2006         err = 0;
2007         if (afinfo->init_flags)
2008                 err = afinfo->init_flags(x);
2009
2010         xfrm_state_put_afinfo(afinfo);
2011
2012         if (err)
2013                 goto error;
2014
2015         err = -EPROTONOSUPPORT;
2016
2017         if (x->sel.family != AF_UNSPEC) {
2018                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2019                 if (inner_mode == NULL)
2020                         goto error;
2021
2022                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2023                     family != x->sel.family) {
2024                         xfrm_put_mode(inner_mode);
2025                         goto error;
2026                 }
2027
2028                 x->inner_mode = inner_mode;
2029         } else {
2030                 struct xfrm_mode *inner_mode_iaf;
2031                 int iafamily = AF_INET;
2032
2033                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2034                 if (inner_mode == NULL)
2035                         goto error;
2036
2037                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2038                         xfrm_put_mode(inner_mode);
2039                         goto error;
2040                 }
2041                 x->inner_mode = inner_mode;
2042
2043                 if (x->props.family == AF_INET)
2044                         iafamily = AF_INET6;
2045
2046                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2047                 if (inner_mode_iaf) {
2048                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2049                                 x->inner_mode_iaf = inner_mode_iaf;
2050                         else
2051                                 xfrm_put_mode(inner_mode_iaf);
2052                 }
2053         }
2054
2055         x->type = xfrm_get_type(x->id.proto, family);
2056         if (x->type == NULL)
2057                 goto error;
2058
2059         err = x->type->init_state(x);
2060         if (err)
2061                 goto error;
2062
2063         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2064         if (x->outer_mode == NULL)
2065                 goto error;
2066
2067         x->km.state = XFRM_STATE_VALID;
2068
2069 error:
2070         return err;
2071 }
2072
2073 EXPORT_SYMBOL(xfrm_init_state);
2074
2075 int __net_init xfrm_state_init(struct net *net)
2076 {
2077         unsigned int sz;
2078
2079         INIT_LIST_HEAD(&net->xfrm.state_all);
2080
2081         sz = sizeof(struct hlist_head) * 8;
2082
2083         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2084         if (!net->xfrm.state_bydst)
2085                 goto out_bydst;
2086         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2087         if (!net->xfrm.state_bysrc)
2088                 goto out_bysrc;
2089         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2090         if (!net->xfrm.state_byspi)
2091                 goto out_byspi;
2092         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2093
2094         net->xfrm.state_num = 0;
2095         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2096         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2097         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2098         init_waitqueue_head(&net->xfrm.km_waitq);
2099         return 0;
2100
2101 out_byspi:
2102         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2103 out_bysrc:
2104         xfrm_hash_free(net->xfrm.state_bydst, sz);
2105 out_bydst:
2106         return -ENOMEM;
2107 }
2108
2109 void xfrm_state_fini(struct net *net)
2110 {
2111         struct xfrm_audit audit_info;
2112         unsigned int sz;
2113
2114         flush_work(&net->xfrm.state_hash_work);
2115         audit_info.loginuid = -1;
2116         audit_info.sessionid = -1;
2117         audit_info.secid = 0;
2118         xfrm_state_flush(net, IPSEC_PROTO_ANY, &audit_info);
2119         flush_work(&net->xfrm.state_gc_work);
2120
2121         WARN_ON(!list_empty(&net->xfrm.state_all));
2122
2123         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2124         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2125         xfrm_hash_free(net->xfrm.state_byspi, sz);
2126         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2127         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2128         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2129         xfrm_hash_free(net->xfrm.state_bydst, sz);
2130 }
2131
2132 #ifdef CONFIG_AUDITSYSCALL
2133 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2134                                      struct audit_buffer *audit_buf)
2135 {
2136         struct xfrm_sec_ctx *ctx = x->security;
2137         u32 spi = ntohl(x->id.spi);
2138
2139         if (ctx)
2140                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2141                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2142
2143         switch(x->props.family) {
2144         case AF_INET:
2145                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2146                                  &x->props.saddr.a4, &x->id.daddr.a4);
2147                 break;
2148         case AF_INET6:
2149                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2150                                  x->props.saddr.a6, x->id.daddr.a6);
2151                 break;
2152         }
2153
2154         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2155 }
2156
2157 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2158                                       struct audit_buffer *audit_buf)
2159 {
2160         struct iphdr *iph4;
2161         struct ipv6hdr *iph6;
2162
2163         switch (family) {
2164         case AF_INET:
2165                 iph4 = ip_hdr(skb);
2166                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2167                                  &iph4->saddr, &iph4->daddr);
2168                 break;
2169         case AF_INET6:
2170                 iph6 = ipv6_hdr(skb);
2171                 audit_log_format(audit_buf,
2172                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2173                                  &iph6->saddr,&iph6->daddr,
2174                                  iph6->flow_lbl[0] & 0x0f,
2175                                  iph6->flow_lbl[1],
2176                                  iph6->flow_lbl[2]);
2177                 break;
2178         }
2179 }
2180
2181 void xfrm_audit_state_add(struct xfrm_state *x, int result,
2182                           uid_t auid, u32 sessionid, u32 secid)
2183 {
2184         struct audit_buffer *audit_buf;
2185
2186         audit_buf = xfrm_audit_start("SAD-add");
2187         if (audit_buf == NULL)
2188                 return;
2189         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2190         xfrm_audit_helper_sainfo(x, audit_buf);
2191         audit_log_format(audit_buf, " res=%u", result);
2192         audit_log_end(audit_buf);
2193 }
2194 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2195
2196 void xfrm_audit_state_delete(struct xfrm_state *x, int result,
2197                              uid_t auid, u32 sessionid, u32 secid)
2198 {
2199         struct audit_buffer *audit_buf;
2200
2201         audit_buf = xfrm_audit_start("SAD-delete");
2202         if (audit_buf == NULL)
2203                 return;
2204         xfrm_audit_helper_usrinfo(auid, sessionid, secid, audit_buf);
2205         xfrm_audit_helper_sainfo(x, audit_buf);
2206         audit_log_format(audit_buf, " res=%u", result);
2207         audit_log_end(audit_buf);
2208 }
2209 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2210
2211 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2212                                       struct sk_buff *skb)
2213 {
2214         struct audit_buffer *audit_buf;
2215         u32 spi;
2216
2217         audit_buf = xfrm_audit_start("SA-replay-overflow");
2218         if (audit_buf == NULL)
2219                 return;
2220         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2221         /* don't record the sequence number because it's inherent in this kind
2222          * of audit message */
2223         spi = ntohl(x->id.spi);
2224         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2225         audit_log_end(audit_buf);
2226 }
2227 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2228
2229 static void xfrm_audit_state_replay(struct xfrm_state *x,
2230                              struct sk_buff *skb, __be32 net_seq)
2231 {
2232         struct audit_buffer *audit_buf;
2233         u32 spi;
2234
2235         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2236         if (audit_buf == NULL)
2237                 return;
2238         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2239         spi = ntohl(x->id.spi);
2240         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2241                          spi, spi, ntohl(net_seq));
2242         audit_log_end(audit_buf);
2243 }
2244
2245 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2246 {
2247         struct audit_buffer *audit_buf;
2248
2249         audit_buf = xfrm_audit_start("SA-notfound");
2250         if (audit_buf == NULL)
2251                 return;
2252         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2253         audit_log_end(audit_buf);
2254 }
2255 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2256
2257 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2258                                __be32 net_spi, __be32 net_seq)
2259 {
2260         struct audit_buffer *audit_buf;
2261         u32 spi;
2262
2263         audit_buf = xfrm_audit_start("SA-notfound");
2264         if (audit_buf == NULL)
2265                 return;
2266         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2267         spi = ntohl(net_spi);
2268         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2269                          spi, spi, ntohl(net_seq));
2270         audit_log_end(audit_buf);
2271 }
2272 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2273
2274 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2275                               struct sk_buff *skb, u8 proto)
2276 {
2277         struct audit_buffer *audit_buf;
2278         __be32 net_spi;
2279         __be32 net_seq;
2280
2281         audit_buf = xfrm_audit_start("SA-icv-failure");
2282         if (audit_buf == NULL)
2283                 return;
2284         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2285         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2286                 u32 spi = ntohl(net_spi);
2287                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2288                                  spi, spi, ntohl(net_seq));
2289         }
2290         audit_log_end(audit_buf);
2291 }
2292 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2293 #endif /* CONFIG_AUDITSYSCALL */