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 #define xfrm_state_deref_prot(table, net) \
  32         rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
  33
  34 static void xfrm_state_gc_task(struct work_struct *work);
  35
  36 /* Each xfrm_state may be linked to two tables:
  37
  38    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  39    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  40       destination/tunnel endpoint. (output)
  41  */
  42
  43 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  44 static __read_mostly seqcount_t xfrm_state_hash_generation = SEQCNT_ZERO(xfrm_state_hash_generation);
  45
  46 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
  47 static HLIST_HEAD(xfrm_state_gc_list);
  48
  49 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
  50 {
  51         return atomic_inc_not_zero(&x->refcnt);
  52 }
  53
  54 static inline unsigned int xfrm_dst_hash(struct net *net,
  55                                          const xfrm_address_t *daddr,
  56                                          const xfrm_address_t *saddr,
  57                                          u32 reqid,
  58                                          unsigned short family)
  59 {
  60         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  61 }
  62
  63 static inline unsigned int xfrm_src_hash(struct net *net,
  64                                          const xfrm_address_t *daddr,
  65                                          const xfrm_address_t *saddr,
  66                                          unsigned short family)
  67 {
  68         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  69 }
  70
  71 static inline unsigned int
  72 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  73               __be32 spi, u8 proto, unsigned short family)
  74 {
  75         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  76 }
  77
  78 static void xfrm_hash_transfer(struct hlist_head *list,
  79                                struct hlist_head *ndsttable,
  80                                struct hlist_head *nsrctable,
  81                                struct hlist_head *nspitable,
  82                                unsigned int nhashmask)
  83 {
  84         struct hlist_node *tmp;
  85         struct xfrm_state *x;
  86
  87         hlist_for_each_entry_safe(x, tmp, list, bydst) {
  88                 unsigned int h;
  89
  90                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  91                                     x->props.reqid, x->props.family,
  92                                     nhashmask);
  93                 hlist_add_head_rcu(&x->bydst, ndsttable + h);
  94
  95                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  96                                     x->props.family,
  97                                     nhashmask);
  98                 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
  99
 100                 if (x->id.spi) {
 101                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
 102                                             x->id.proto, x->props.family,
 103                                             nhashmask);
 104                         hlist_add_head_rcu(&x->byspi, nspitable + h);
 105                 }
 106         }
 107 }
 108
 109 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 110 {
 111         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 112 }
 113
 114 static void xfrm_hash_resize(struct work_struct *work)
 115 {
 116         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 117         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 118         unsigned long nsize, osize;
 119         unsigned int nhashmask, ohashmask;
 120         int i;
 121
 122         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 123         ndst = xfrm_hash_alloc(nsize);
 124         if (!ndst)
 125                 return;
 126         nsrc = xfrm_hash_alloc(nsize);
 127         if (!nsrc) {
 128                 xfrm_hash_free(ndst, nsize);
 129                 return;
 130         }
 131         nspi = xfrm_hash_alloc(nsize);
 132         if (!nspi) {
 133                 xfrm_hash_free(ndst, nsize);
 134                 xfrm_hash_free(nsrc, nsize);
 135                 return;
 136         }
 137
 138         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 139         write_seqcount_begin(&xfrm_state_hash_generation);
 140
 141         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 142         odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
 143         for (i = net->xfrm.state_hmask; i >= 0; i--)
 144                 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nhashmask);
 145
 146         osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
 147         ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
 148         ohashmask = net->xfrm.state_hmask;
 149
 150         rcu_assign_pointer(net->xfrm.state_bydst, ndst);
 151         rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
 152         rcu_assign_pointer(net->xfrm.state_byspi, nspi);
 153         net->xfrm.state_hmask = nhashmask;
 154
 155         write_seqcount_end(&xfrm_state_hash_generation);
 156         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 157
 158         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 159
 160         synchronize_rcu();
 161
 162         xfrm_hash_free(odst, osize);
 163         xfrm_hash_free(osrc, osize);
 164         xfrm_hash_free(ospi, osize);
 165 }
 166
 167 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 168 static struct xfrm_state_afinfo __rcu *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 bool km_is_alive(const struct km_event *c);
 176 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 177
 178 static DEFINE_SPINLOCK(xfrm_type_lock);
 179 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 180 {
 181         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 182         const struct xfrm_type **typemap;
 183         int err = 0;
 184
 185         if (unlikely(afinfo == NULL))
 186                 return -EAFNOSUPPORT;
 187         typemap = afinfo->type_map;
 188         spin_lock_bh(&xfrm_type_lock);
 189
 190         if (likely(typemap[type->proto] == NULL))
 191                 typemap[type->proto] = type;
 192         else
 193                 err = -EEXIST;
 194         spin_unlock_bh(&xfrm_type_lock);
 195         xfrm_state_put_afinfo(afinfo);
 196         return err;
 197 }
 198 EXPORT_SYMBOL(xfrm_register_type);
 199
 200 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 201 {
 202         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 203         const struct xfrm_type **typemap;
 204         int err = 0;
 205
 206         if (unlikely(afinfo == NULL))
 207                 return -EAFNOSUPPORT;
 208         typemap = afinfo->type_map;
 209         spin_lock_bh(&xfrm_type_lock);
 210
 211         if (unlikely(typemap[type->proto] != type))
 212                 err = -ENOENT;
 213         else
 214                 typemap[type->proto] = NULL;
 215         spin_unlock_bh(&xfrm_type_lock);
 216         xfrm_state_put_afinfo(afinfo);
 217         return err;
 218 }
 219 EXPORT_SYMBOL(xfrm_unregister_type);
 220
 221 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 222 {
 223         struct xfrm_state_afinfo *afinfo;
 224         const struct xfrm_type **typemap;
 225         const struct xfrm_type *type;
 226         int modload_attempted = 0;
 227
 228 retry:
 229         afinfo = xfrm_state_get_afinfo(family);
 230         if (unlikely(afinfo == NULL))
 231                 return NULL;
 232         typemap = afinfo->type_map;
 233
 234         type = typemap[proto];
 235         if (unlikely(type && !try_module_get(type->owner)))
 236                 type = NULL;
 237         if (!type && !modload_attempted) {
 238                 xfrm_state_put_afinfo(afinfo);
 239                 request_module("xfrm-type-%d-%d", family, proto);
 240                 modload_attempted = 1;
 241                 goto retry;
 242         }
 243
 244         xfrm_state_put_afinfo(afinfo);
 245         return type;
 246 }
 247
 248 static void xfrm_put_type(const struct xfrm_type *type)
 249 {
 250         module_put(type->owner);
 251 }
 252
 253 static DEFINE_SPINLOCK(xfrm_mode_lock);
 254 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 255 {
 256         struct xfrm_state_afinfo *afinfo;
 257         struct xfrm_mode **modemap;
 258         int err;
 259
 260         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 261                 return -EINVAL;
 262
 263         afinfo = xfrm_state_get_afinfo(family);
 264         if (unlikely(afinfo == NULL))
 265                 return -EAFNOSUPPORT;
 266
 267         err = -EEXIST;
 268         modemap = afinfo->mode_map;
 269         spin_lock_bh(&xfrm_mode_lock);
 270         if (modemap[mode->encap])
 271                 goto out;
 272
 273         err = -ENOENT;
 274         if (!try_module_get(afinfo->owner))
 275                 goto out;
 276
 277         mode->afinfo = afinfo;
 278         modemap[mode->encap] = mode;
 279         err = 0;
 280
 281 out:
 282         spin_unlock_bh(&xfrm_mode_lock);
 283         xfrm_state_put_afinfo(afinfo);
 284         return err;
 285 }
 286 EXPORT_SYMBOL(xfrm_register_mode);
 287
 288 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 289 {
 290         struct xfrm_state_afinfo *afinfo;
 291         struct xfrm_mode **modemap;
 292         int err;
 293
 294         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 295                 return -EINVAL;
 296
 297         afinfo = xfrm_state_get_afinfo(family);
 298         if (unlikely(afinfo == NULL))
 299                 return -EAFNOSUPPORT;
 300
 301         err = -ENOENT;
 302         modemap = afinfo->mode_map;
 303         spin_lock_bh(&xfrm_mode_lock);
 304         if (likely(modemap[mode->encap] == mode)) {
 305                 modemap[mode->encap] = NULL;
 306                 module_put(mode->afinfo->owner);
 307                 err = 0;
 308         }
 309
 310         spin_unlock_bh(&xfrm_mode_lock);
 311         xfrm_state_put_afinfo(afinfo);
 312         return err;
 313 }
 314 EXPORT_SYMBOL(xfrm_unregister_mode);
 315
 316 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 317 {
 318         struct xfrm_state_afinfo *afinfo;
 319         struct xfrm_mode *mode;
 320         int modload_attempted = 0;
 321
 322         if (unlikely(encap >= XFRM_MODE_MAX))
 323                 return NULL;
 324
 325 retry:
 326         afinfo = xfrm_state_get_afinfo(family);
 327         if (unlikely(afinfo == NULL))
 328                 return NULL;
 329
 330         mode = afinfo->mode_map[encap];
 331         if (unlikely(mode && !try_module_get(mode->owner)))
 332                 mode = NULL;
 333         if (!mode && !modload_attempted) {
 334                 xfrm_state_put_afinfo(afinfo);
 335                 request_module("xfrm-mode-%d-%d", family, encap);
 336                 modload_attempted = 1;
 337                 goto retry;
 338         }
 339
 340         xfrm_state_put_afinfo(afinfo);
 341         return mode;
 342 }
 343
 344 static void xfrm_put_mode(struct xfrm_mode *mode)
 345 {
 346         module_put(mode->owner);
 347 }
 348
 349 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 350 {
 351         tasklet_hrtimer_cancel(&x->mtimer);
 352         del_timer_sync(&x->rtimer);
 353         kfree(x->aead);
 354         kfree(x->aalg);
 355         kfree(x->ealg);
 356         kfree(x->calg);
 357         kfree(x->encap);
 358         kfree(x->coaddr);
 359         kfree(x->replay_esn);
 360         kfree(x->preplay_esn);
 361         if (x->inner_mode)
 362                 xfrm_put_mode(x->inner_mode);
 363         if (x->inner_mode_iaf)
 364                 xfrm_put_mode(x->inner_mode_iaf);
 365         if (x->outer_mode)
 366                 xfrm_put_mode(x->outer_mode);
 367         if (x->type) {
 368                 x->type->destructor(x);
 369                 xfrm_put_type(x->type);
 370         }
 371         security_xfrm_state_free(x);
 372         kfree(x);
 373 }
 374
 375 static void xfrm_state_gc_task(struct work_struct *work)
 376 {
 377         struct xfrm_state *x;
 378         struct hlist_node *tmp;
 379         struct hlist_head gc_list;
 380
 381         spin_lock_bh(&xfrm_state_gc_lock);
 382         hlist_move_list(&xfrm_state_gc_list, &gc_list);
 383         spin_unlock_bh(&xfrm_state_gc_lock);
 384
 385         synchronize_rcu();
 386
 387         hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 388                 xfrm_state_gc_destroy(x);
 389 }
 390
 391 static inline unsigned long make_jiffies(long secs)
 392 {
 393         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 394                 return MAX_SCHEDULE_TIMEOUT-1;
 395         else
 396                 return secs*HZ;
 397 }
 398
 399 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
 400 {
 401         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 402         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 403         unsigned long now = get_seconds();
 404         long next = LONG_MAX;
 405         int warn = 0;
 406         int err = 0;
 407
 408         spin_lock(&x->lock);
 409         if (x->km.state == XFRM_STATE_DEAD)
 410                 goto out;
 411         if (x->km.state == XFRM_STATE_EXPIRED)
 412                 goto expired;
 413         if (x->lft.hard_add_expires_seconds) {
 414                 long tmo = x->lft.hard_add_expires_seconds +
 415                         x->curlft.add_time - now;
 416                 if (tmo <= 0) {
 417                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 418                                 /* enter hard expire without soft expire first?!
 419                                  * setting a new date could trigger this.
 420                                  * workarbound: fix x->curflt.add_time by below:
 421                                  */
 422                                 x->curlft.add_time = now - x->saved_tmo - 1;
 423                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 424                         } else
 425                                 goto expired;
 426                 }
 427                 if (tmo < next)
 428                         next = tmo;
 429         }
 430         if (x->lft.hard_use_expires_seconds) {
 431                 long tmo = x->lft.hard_use_expires_seconds +
 432                         (x->curlft.use_time ? : now) - now;
 433                 if (tmo <= 0)
 434                         goto expired;
 435                 if (tmo < next)
 436                         next = tmo;
 437         }
 438         if (x->km.dying)
 439                 goto resched;
 440         if (x->lft.soft_add_expires_seconds) {
 441                 long tmo = x->lft.soft_add_expires_seconds +
 442                         x->curlft.add_time - now;
 443                 if (tmo <= 0) {
 444                         warn = 1;
 445                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 446                 } else if (tmo < next) {
 447                         next = tmo;
 448                         x->xflags |= XFRM_SOFT_EXPIRE;
 449                         x->saved_tmo = tmo;
 450                 }
 451         }
 452         if (x->lft.soft_use_expires_seconds) {
 453                 long tmo = x->lft.soft_use_expires_seconds +
 454                         (x->curlft.use_time ? : now) - now;
 455                 if (tmo <= 0)
 456                         warn = 1;
 457                 else if (tmo < next)
 458                         next = tmo;
 459         }
 460
 461         x->km.dying = warn;
 462         if (warn)
 463                 km_state_expired(x, 0, 0);
 464 resched:
 465         if (next != LONG_MAX) {
 466                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 467         }
 468
 469         goto out;
 470
 471 expired:
 472         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
 473                 x->km.state = XFRM_STATE_EXPIRED;
 474
 475         err = __xfrm_state_delete(x);
 476         if (!err)
 477                 km_state_expired(x, 1, 0);
 478
 479         xfrm_audit_state_delete(x, err ? 0 : 1, true);
 480
 481 out:
 482         spin_unlock(&x->lock);
 483         return HRTIMER_NORESTART;
 484 }
 485
 486 static void xfrm_replay_timer_handler(unsigned long data);
 487
 488 struct xfrm_state *xfrm_state_alloc(struct net *net)
 489 {
 490         struct xfrm_state *x;
 491
 492         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 493
 494         if (x) {
 495                 write_pnet(&x->xs_net, net);
 496                 atomic_set(&x->refcnt, 1);
 497                 atomic_set(&x->tunnel_users, 0);
 498                 INIT_LIST_HEAD(&x->km.all);
 499                 INIT_HLIST_NODE(&x->bydst);
 500                 INIT_HLIST_NODE(&x->bysrc);
 501                 INIT_HLIST_NODE(&x->byspi);
 502                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
 503                                         CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
 504                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 505                                 (unsigned long)x);
 506                 x->curlft.add_time = get_seconds();
 507                 x->lft.soft_byte_limit = XFRM_INF;
 508                 x->lft.soft_packet_limit = XFRM_INF;
 509                 x->lft.hard_byte_limit = XFRM_INF;
 510                 x->lft.hard_packet_limit = XFRM_INF;
 511                 x->replay_maxage = 0;
 512                 x->replay_maxdiff = 0;
 513                 x->inner_mode = NULL;
 514                 x->inner_mode_iaf = NULL;
 515                 spin_lock_init(&x->lock);
 516         }
 517         return x;
 518 }
 519 EXPORT_SYMBOL(xfrm_state_alloc);
 520
 521 void __xfrm_state_destroy(struct xfrm_state *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, &xfrm_state_gc_list);
 527         spin_unlock_bh(&xfrm_state_gc_lock);
 528         schedule_work(&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(&net->xfrm.xfrm_state_lock);
 540                 list_del(&x->km.all);
 541                 hlist_del_rcu(&x->bydst);
 542                 hlist_del_rcu(&x->bysrc);
 543                 if (x->id.spi)
 544                         hlist_del_rcu(&x->byspi);
 545                 net->xfrm.state_num--;
 546                 spin_unlock(&net->xfrm.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, bool task_valid)
 575 {
 576         int i, err = 0;
 577
 578         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 579                 struct xfrm_state *x;
 580
 581                 hlist_for_each_entry(x, 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, task_valid);
 585                                 return err;
 586                         }
 587                 }
 588         }
 589
 590         return err;
 591 }
 592 #else
 593 static inline int
 594 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 595 {
 596         return 0;
 597 }
 598 #endif
 599
 600 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
 601 {
 602         int i, err = 0, cnt = 0;
 603
 604         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 605         err = xfrm_state_flush_secctx_check(net, proto, task_valid);
 606         if (err)
 607                 goto out;
 608
 609         err = -ESRCH;
 610         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 611                 struct xfrm_state *x;
 612 restart:
 613                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 614                         if (!xfrm_state_kern(x) &&
 615                             xfrm_id_proto_match(x->id.proto, proto)) {
 616                                 xfrm_state_hold(x);
 617                                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 618
 619                                 err = xfrm_state_delete(x);
 620                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 621                                                         task_valid);
 622                                 xfrm_state_put(x);
 623                                 if (!err)
 624                                         cnt++;
 625
 626                                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
 627                                 goto restart;
 628                         }
 629                 }
 630         }
 631         if (cnt)
 632                 err = 0;
 633
 634 out:
 635         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 636         return err;
 637 }
 638 EXPORT_SYMBOL(xfrm_state_flush);
 639
 640 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 641 {
 642         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 643         si->sadcnt = net->xfrm.state_num;
 644         si->sadhcnt = net->xfrm.state_hmask + 1;
 645         si->sadhmcnt = xfrm_state_hashmax;
 646         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 647 }
 648 EXPORT_SYMBOL(xfrm_sad_getinfo);
 649
 650 static int
 651 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 652                     const struct xfrm_tmpl *tmpl,
 653                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 654                     unsigned short family)
 655 {
 656         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 657         if (!afinfo)
 658                 return -1;
 659         afinfo->init_tempsel(&x->sel, fl);
 660
 661         if (family != tmpl->encap_family) {
 662                 xfrm_state_put_afinfo(afinfo);
 663                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 664                 if (!afinfo)
 665                         return -1;
 666         }
 667         afinfo->init_temprop(x, 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,
 673                                               const xfrm_address_t *daddr,
 674                                               __be32 spi, u8 proto,
 675                                               unsigned short family)
 676 {
 677         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 678         struct xfrm_state *x;
 679
 680         hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
 681                 if (x->props.family != family ||
 682                     x->id.spi       != spi ||
 683                     x->id.proto     != proto ||
 684                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 685                         continue;
 686
 687                 if ((mark & x->mark.m) != x->mark.v)
 688                         continue;
 689                 if (!xfrm_state_hold_rcu(x))
 690                         continue;
 691                 return x;
 692         }
 693
 694         return NULL;
 695 }
 696
 697 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 698                                                      const xfrm_address_t *daddr,
 699                                                      const xfrm_address_t *saddr,
 700                                                      u8 proto, unsigned short family)
 701 {
 702         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 703         struct xfrm_state *x;
 704
 705         hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
 706                 if (x->props.family != family ||
 707                     x->id.proto     != proto ||
 708                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 709                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 710                         continue;
 711
 712                 if ((mark & x->mark.m) != x->mark.v)
 713                         continue;
 714                 if (!xfrm_state_hold_rcu(x))
 715                         continue;
 716                 return x;
 717         }
 718
 719         return NULL;
 720 }
 721
 722 static inline struct xfrm_state *
 723 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 724 {
 725         struct net *net = xs_net(x);
 726         u32 mark = x->mark.v & x->mark.m;
 727
 728         if (use_spi)
 729                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 730                                            x->id.spi, x->id.proto, family);
 731         else
 732                 return __xfrm_state_lookup_byaddr(net, mark,
 733                                                   &x->id.daddr,
 734                                                   &x->props.saddr,
 735                                                   x->id.proto, family);
 736 }
 737
 738 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 739 {
 740         if (have_hash_collision &&
 741             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 742             net->xfrm.state_num > net->xfrm.state_hmask)
 743                 schedule_work(&net->xfrm.state_hash_work);
 744 }
 745
 746 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 747                                const struct flowi *fl, unsigned short family,
 748                                struct xfrm_state **best, int *acq_in_progress,
 749                                int *error)
 750 {
 751         /* Resolution logic:
 752          * 1. There is a valid state with matching selector. Done.
 753          * 2. Valid state with inappropriate selector. Skip.
 754          *
 755          * Entering area of "sysdeps".
 756          *
 757          * 3. If state is not valid, selector is temporary, it selects
 758          *    only session which triggered previous resolution. Key
 759          *    manager will do something to install a state with proper
 760          *    selector.
 761          */
 762         if (x->km.state == XFRM_STATE_VALID) {
 763                 if ((x->sel.family &&
 764                      (x->sel.family != family ||
 765                       !xfrm_selector_match(&x->sel, fl, family))) ||
 766                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 767                         return;
 768
 769                 if (!*best ||
 770                     (*best)->km.dying > x->km.dying ||
 771                     ((*best)->km.dying == x->km.dying &&
 772                      (*best)->curlft.add_time < x->curlft.add_time))
 773                         *best = x;
 774         } else if (x->km.state == XFRM_STATE_ACQ) {
 775                 *acq_in_progress = 1;
 776         } else if (x->km.state == XFRM_STATE_ERROR ||
 777                    x->km.state == XFRM_STATE_EXPIRED) {
 778                 if ((!x->sel.family ||
 779                      (x->sel.family == family &&
 780                       xfrm_selector_match(&x->sel, fl, family))) &&
 781                     security_xfrm_state_pol_flow_match(x, pol, fl))
 782                         *error = -ESRCH;
 783         }
 784 }
 785
 786 struct xfrm_state *
 787 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 788                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 789                 struct xfrm_policy *pol, int *err,
 790                 unsigned short family)
 791 {
 792         static xfrm_address_t saddr_wildcard = { };
 793         struct net *net = xp_net(pol);
 794         unsigned int h, h_wildcard;
 795         struct xfrm_state *x, *x0, *to_put;
 796         int acquire_in_progress = 0;
 797         int error = 0;
 798         struct xfrm_state *best = NULL;
 799         u32 mark = pol->mark.v & pol->mark.m;
 800         unsigned short encap_family = tmpl->encap_family;
 801         unsigned int sequence;
 802         struct km_event c;
 803
 804         to_put = NULL;
 805
 806         sequence = read_seqcount_begin(&xfrm_state_hash_generation);
 807
 808         rcu_read_lock();
 809         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 810         hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
 811                 if (x->props.family == encap_family &&
 812                     x->props.reqid == tmpl->reqid &&
 813                     (mark & x->mark.m) == x->mark.v &&
 814                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 815                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 816                     tmpl->mode == x->props.mode &&
 817                     tmpl->id.proto == x->id.proto &&
 818                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 819                         xfrm_state_look_at(pol, x, fl, family,
 820                                            &best, &acquire_in_progress, &error);
 821         }
 822         if (best || acquire_in_progress)
 823                 goto found;
 824
 825         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 826         hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
 827                 if (x->props.family == encap_family &&
 828                     x->props.reqid == tmpl->reqid &&
 829                     (mark & x->mark.m) == x->mark.v &&
 830                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 831                     xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
 832                     tmpl->mode == x->props.mode &&
 833                     tmpl->id.proto == x->id.proto &&
 834                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 835                         xfrm_state_look_at(pol, x, fl, family,
 836                                            &best, &acquire_in_progress, &error);
 837         }
 838
 839 found:
 840         x = best;
 841         if (!x && !error && !acquire_in_progress) {
 842                 if (tmpl->id.spi &&
 843                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 844                                               tmpl->id.proto, encap_family)) != NULL) {
 845                         to_put = x0;
 846                         error = -EEXIST;
 847                         goto out;
 848                 }
 849
 850                 c.net = net;
 851                 /* If the KMs have no listeners (yet...), avoid allocating an SA
 852                  * for each and every packet - garbage collection might not
 853                  * handle the flood.
 854                  */
 855                 if (!km_is_alive(&c)) {
 856                         error = -ESRCH;
 857                         goto out;
 858                 }
 859
 860                 x = xfrm_state_alloc(net);
 861                 if (x == NULL) {
 862                         error = -ENOMEM;
 863                         goto out;
 864                 }
 865                 /* Initialize temporary state matching only
 866                  * to current session. */
 867                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 868                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 869
 870                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 871                 if (error) {
 872                         x->km.state = XFRM_STATE_DEAD;
 873                         to_put = x;
 874                         x = NULL;
 875                         goto out;
 876                 }
 877
 878                 if (km_query(x, tmpl, pol) == 0) {
 879                         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 880                         x->km.state = XFRM_STATE_ACQ;
 881                         list_add(&x->km.all, &net->xfrm.state_all);
 882                         hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
 883                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 884                         hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
 885                         if (x->id.spi) {
 886                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 887                                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
 888                         }
 889                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 890                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 891                         net->xfrm.state_num++;
 892                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 893                         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 894                 } else {
 895                         x->km.state = XFRM_STATE_DEAD;
 896                         to_put = x;
 897                         x = NULL;
 898                         error = -ESRCH;
 899                 }
 900         }
 901 out:
 902         if (x) {
 903                 if (!xfrm_state_hold_rcu(x)) {
 904                         *err = -EAGAIN;
 905                         x = NULL;
 906                 }
 907         } else {
 908                 *err = acquire_in_progress ? -EAGAIN : error;
 909         }
 910         rcu_read_unlock();
 911         if (to_put)
 912                 xfrm_state_put(to_put);
 913
 914         if (read_seqcount_retry(&xfrm_state_hash_generation, sequence)) {
 915                 *err = -EAGAIN;
 916                 if (x) {
 917                         xfrm_state_put(x);
 918                         x = NULL;
 919                 }
 920         }
 921
 922         return x;
 923 }
 924
 925 struct xfrm_state *
 926 xfrm_stateonly_find(struct net *net, u32 mark,
 927                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 928                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 929 {
 930         unsigned int h;
 931         struct xfrm_state *rx = NULL, *x = NULL;
 932
 933         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 934         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 935         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 936                 if (x->props.family == family &&
 937                     x->props.reqid == reqid &&
 938                     (mark & x->mark.m) == x->mark.v &&
 939                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 940                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 941                     mode == x->props.mode &&
 942                     proto == x->id.proto &&
 943                     x->km.state == XFRM_STATE_VALID) {
 944                         rx = x;
 945                         break;
 946                 }
 947         }
 948
 949         if (rx)
 950                 xfrm_state_hold(rx);
 951         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 952
 953
 954         return rx;
 955 }
 956 EXPORT_SYMBOL(xfrm_stateonly_find);
 957
 958 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
 959                                               unsigned short family)
 960 {
 961         struct xfrm_state *x;
 962         struct xfrm_state_walk *w;
 963
 964         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 965         list_for_each_entry(w, &net->xfrm.state_all, all) {
 966                 x = container_of(w, struct xfrm_state, km);
 967                 if (x->props.family != family ||
 968                         x->id.spi != spi)
 969                         continue;
 970
 971                 xfrm_state_hold(x);
 972                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 973                 return x;
 974         }
 975         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 976         return NULL;
 977 }
 978 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
 979
 980 static void __xfrm_state_insert(struct xfrm_state *x)
 981 {
 982         struct net *net = xs_net(x);
 983         unsigned int h;
 984
 985         list_add(&x->km.all, &net->xfrm.state_all);
 986
 987         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 988                           x->props.reqid, x->props.family);
 989         hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
 990
 991         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 992         hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
 993
 994         if (x->id.spi) {
 995                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 996                                   x->props.family);
 997
 998                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
 999         }
1000
1001         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1002         if (x->replay_maxage)
1003                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1004
1005         net->xfrm.state_num++;
1006
1007         xfrm_hash_grow_check(net, x->bydst.next != NULL);
1008 }
1009
1010 /* net->xfrm.xfrm_state_lock is held */
1011 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1012 {
1013         struct net *net = xs_net(xnew);
1014         unsigned short family = xnew->props.family;
1015         u32 reqid = xnew->props.reqid;
1016         struct xfrm_state *x;
1017         unsigned int h;
1018         u32 mark = xnew->mark.v & xnew->mark.m;
1019
1020         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1021         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1022                 if (x->props.family     == family &&
1023                     x->props.reqid      == reqid &&
1024                     (mark & x->mark.m) == x->mark.v &&
1025                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1026                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1027                         x->genid++;
1028         }
1029 }
1030
1031 void xfrm_state_insert(struct xfrm_state *x)
1032 {
1033         struct net *net = xs_net(x);
1034
1035         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1036         __xfrm_state_bump_genids(x);
1037         __xfrm_state_insert(x);
1038         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1039 }
1040 EXPORT_SYMBOL(xfrm_state_insert);
1041
1042 /* net->xfrm.xfrm_state_lock is held */
1043 static struct xfrm_state *__find_acq_core(struct net *net,
1044                                           const struct xfrm_mark *m,
1045                                           unsigned short family, u8 mode,
1046                                           u32 reqid, u8 proto,
1047                                           const xfrm_address_t *daddr,
1048                                           const xfrm_address_t *saddr,
1049                                           int create)
1050 {
1051         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1052         struct xfrm_state *x;
1053         u32 mark = m->v & m->m;
1054
1055         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1056                 if (x->props.reqid  != reqid ||
1057                     x->props.mode   != mode ||
1058                     x->props.family != family ||
1059                     x->km.state     != XFRM_STATE_ACQ ||
1060                     x->id.spi       != 0 ||
1061                     x->id.proto     != proto ||
1062                     (mark & x->mark.m) != x->mark.v ||
1063                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1064                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1065                         continue;
1066
1067                 xfrm_state_hold(x);
1068                 return x;
1069         }
1070
1071         if (!create)
1072                 return NULL;
1073
1074         x = xfrm_state_alloc(net);
1075         if (likely(x)) {
1076                 switch (family) {
1077                 case AF_INET:
1078                         x->sel.daddr.a4 = daddr->a4;
1079                         x->sel.saddr.a4 = saddr->a4;
1080                         x->sel.prefixlen_d = 32;
1081                         x->sel.prefixlen_s = 32;
1082                         x->props.saddr.a4 = saddr->a4;
1083                         x->id.daddr.a4 = daddr->a4;
1084                         break;
1085
1086                 case AF_INET6:
1087                         x->sel.daddr.in6 = daddr->in6;
1088                         x->sel.saddr.in6 = saddr->in6;
1089                         x->sel.prefixlen_d = 128;
1090                         x->sel.prefixlen_s = 128;
1091                         x->props.saddr.in6 = saddr->in6;
1092                         x->id.daddr.in6 = daddr->in6;
1093                         break;
1094                 }
1095
1096                 x->km.state = XFRM_STATE_ACQ;
1097                 x->id.proto = proto;
1098                 x->props.family = family;
1099                 x->props.mode = mode;
1100                 x->props.reqid = reqid;
1101                 x->mark.v = m->v;
1102                 x->mark.m = m->m;
1103                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1104                 xfrm_state_hold(x);
1105                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1106                 list_add(&x->km.all, &net->xfrm.state_all);
1107                 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1108                 h = xfrm_src_hash(net, daddr, saddr, family);
1109                 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1110
1111                 net->xfrm.state_num++;
1112
1113                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1114         }
1115
1116         return x;
1117 }
1118
1119 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1120
1121 int xfrm_state_add(struct xfrm_state *x)
1122 {
1123         struct net *net = xs_net(x);
1124         struct xfrm_state *x1, *to_put;
1125         int family;
1126         int err;
1127         u32 mark = x->mark.v & x->mark.m;
1128         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1129
1130         family = x->props.family;
1131
1132         to_put = NULL;
1133
1134         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1135
1136         x1 = __xfrm_state_locate(x, use_spi, family);
1137         if (x1) {
1138                 to_put = x1;
1139                 x1 = NULL;
1140                 err = -EEXIST;
1141                 goto out;
1142         }
1143
1144         if (use_spi && x->km.seq) {
1145                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1146                 if (x1 && ((x1->id.proto != x->id.proto) ||
1147                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1148                         to_put = x1;
1149                         x1 = NULL;
1150                 }
1151         }
1152
1153         if (use_spi && !x1)
1154                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1155                                      x->props.reqid, x->id.proto,
1156                                      &x->id.daddr, &x->props.saddr, 0);
1157
1158         __xfrm_state_bump_genids(x);
1159         __xfrm_state_insert(x);
1160         err = 0;
1161
1162 out:
1163         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1164
1165         if (x1) {
1166                 xfrm_state_delete(x1);
1167                 xfrm_state_put(x1);
1168         }
1169
1170         if (to_put)
1171                 xfrm_state_put(to_put);
1172
1173         return err;
1174 }
1175 EXPORT_SYMBOL(xfrm_state_add);
1176
1177 #ifdef CONFIG_XFRM_MIGRATE
1178 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1179 {
1180         struct net *net = xs_net(orig);
1181         struct xfrm_state *x = xfrm_state_alloc(net);
1182         if (!x)
1183                 goto out;
1184
1185         memcpy(&x->id, &orig->id, sizeof(x->id));
1186         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1187         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1188         x->props.mode = orig->props.mode;
1189         x->props.replay_window = orig->props.replay_window;
1190         x->props.reqid = orig->props.reqid;
1191         x->props.family = orig->props.family;
1192         x->props.saddr = orig->props.saddr;
1193
1194         if (orig->aalg) {
1195                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1196                 if (!x->aalg)
1197                         goto error;
1198         }
1199         x->props.aalgo = orig->props.aalgo;
1200
1201         if (orig->aead) {
1202                 x->aead = xfrm_algo_aead_clone(orig->aead);
1203                 x->geniv = orig->geniv;
1204                 if (!x->aead)
1205                         goto error;
1206         }
1207         if (orig->ealg) {
1208                 x->ealg = xfrm_algo_clone(orig->ealg);
1209                 if (!x->ealg)
1210                         goto error;
1211         }
1212         x->props.ealgo = orig->props.ealgo;
1213
1214         if (orig->calg) {
1215                 x->calg = xfrm_algo_clone(orig->calg);
1216                 if (!x->calg)
1217                         goto error;
1218         }
1219         x->props.calgo = orig->props.calgo;
1220
1221         if (orig->encap) {
1222                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1223                 if (!x->encap)
1224                         goto error;
1225         }
1226
1227         if (orig->coaddr) {
1228                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1229                                     GFP_KERNEL);
1230                 if (!x->coaddr)
1231                         goto error;
1232         }
1233
1234         if (orig->replay_esn) {
1235                 if (xfrm_replay_clone(x, orig))
1236                         goto error;
1237         }
1238
1239         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1240
1241         if (xfrm_init_state(x) < 0)
1242                 goto error;
1243
1244         x->props.flags = orig->props.flags;
1245         x->props.extra_flags = orig->props.extra_flags;
1246
1247         x->tfcpad = orig->tfcpad;
1248         x->replay_maxdiff = orig->replay_maxdiff;
1249         x->replay_maxage = orig->replay_maxage;
1250         memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1251         x->km.state = orig->km.state;
1252         x->km.seq = orig->km.seq;
1253         x->replay = orig->replay;
1254         x->preplay = orig->preplay;
1255
1256         return x;
1257
1258  error:
1259         xfrm_state_put(x);
1260 out:
1261         return NULL;
1262 }
1263
1264 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1265 {
1266         unsigned int h;
1267         struct xfrm_state *x = NULL;
1268
1269         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1270
1271         if (m->reqid) {
1272                 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1273                                   m->reqid, m->old_family);
1274                 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1275                         if (x->props.mode != m->mode ||
1276                             x->id.proto != m->proto)
1277                                 continue;
1278                         if (m->reqid && x->props.reqid != m->reqid)
1279                                 continue;
1280                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1281                                              m->old_family) ||
1282                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1283                                              m->old_family))
1284                                 continue;
1285                         xfrm_state_hold(x);
1286                         break;
1287                 }
1288         } else {
1289                 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1290                                   m->old_family);
1291                 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1292                         if (x->props.mode != m->mode ||
1293                             x->id.proto != m->proto)
1294                                 continue;
1295                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1296                                              m->old_family) ||
1297                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1298                                              m->old_family))
1299                                 continue;
1300                         xfrm_state_hold(x);
1301                         break;
1302                 }
1303         }
1304
1305         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1306
1307         return x;
1308 }
1309 EXPORT_SYMBOL(xfrm_migrate_state_find);
1310
1311 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1312                                       struct xfrm_migrate *m)
1313 {
1314         struct xfrm_state *xc;
1315
1316         xc = xfrm_state_clone(x);
1317         if (!xc)
1318                 return NULL;
1319
1320         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1321         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1322
1323         /* add state */
1324         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1325                 /* a care is needed when the destination address of the
1326                    state is to be updated as it is a part of triplet */
1327                 xfrm_state_insert(xc);
1328         } else {
1329                 if (xfrm_state_add(xc) < 0)
1330                         goto error;
1331         }
1332
1333         return xc;
1334 error:
1335         xfrm_state_put(xc);
1336         return NULL;
1337 }
1338 EXPORT_SYMBOL(xfrm_state_migrate);
1339 #endif
1340
1341 int xfrm_state_update(struct xfrm_state *x)
1342 {
1343         struct xfrm_state *x1, *to_put;
1344         int err;
1345         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1346         struct net *net = xs_net(x);
1347
1348         to_put = NULL;
1349
1350         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1351         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1352
1353         err = -ESRCH;
1354         if (!x1)
1355                 goto out;
1356
1357         if (xfrm_state_kern(x1)) {
1358                 to_put = x1;
1359                 err = -EEXIST;
1360                 goto out;
1361         }
1362
1363         if (x1->km.state == XFRM_STATE_ACQ) {
1364                 __xfrm_state_insert(x);
1365                 x = NULL;
1366         }
1367         err = 0;
1368
1369 out:
1370         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1371
1372         if (to_put)
1373                 xfrm_state_put(to_put);
1374
1375         if (err)
1376                 return err;
1377
1378         if (!x) {
1379                 xfrm_state_delete(x1);
1380                 xfrm_state_put(x1);
1381                 return 0;
1382         }
1383
1384         err = -EINVAL;
1385         spin_lock_bh(&x1->lock);
1386         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1387                 if (x->encap && x1->encap)
1388                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1389                 if (x->coaddr && x1->coaddr) {
1390                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1391                 }
1392                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1393                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1394                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1395                 x1->km.dying = 0;
1396
1397                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1398                 if (x1->curlft.use_time)
1399                         xfrm_state_check_expire(x1);
1400
1401                 err = 0;
1402                 x->km.state = XFRM_STATE_DEAD;
1403                 __xfrm_state_put(x);
1404         }
1405         spin_unlock_bh(&x1->lock);
1406
1407         xfrm_state_put(x1);
1408
1409         return err;
1410 }
1411 EXPORT_SYMBOL(xfrm_state_update);
1412
1413 int xfrm_state_check_expire(struct xfrm_state *x)
1414 {
1415         if (!x->curlft.use_time)
1416                 x->curlft.use_time = get_seconds();
1417
1418         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1419             x->curlft.packets >= x->lft.hard_packet_limit) {
1420                 x->km.state = XFRM_STATE_EXPIRED;
1421                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1422                 return -EINVAL;
1423         }
1424
1425         if (!x->km.dying &&
1426             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1427              x->curlft.packets >= x->lft.soft_packet_limit)) {
1428                 x->km.dying = 1;
1429                 km_state_expired(x, 0, 0);
1430         }
1431         return 0;
1432 }
1433 EXPORT_SYMBOL(xfrm_state_check_expire);
1434
1435 struct xfrm_state *
1436 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1437                   u8 proto, unsigned short family)
1438 {
1439         struct xfrm_state *x;
1440
1441         rcu_read_lock();
1442         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1443         rcu_read_unlock();
1444         return x;
1445 }
1446 EXPORT_SYMBOL(xfrm_state_lookup);
1447
1448 struct xfrm_state *
1449 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1450                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1451                          u8 proto, unsigned short family)
1452 {
1453         struct xfrm_state *x;
1454
1455         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1456         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1457         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1458         return x;
1459 }
1460 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1461
1462 struct xfrm_state *
1463 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1464               u8 proto, const xfrm_address_t *daddr,
1465               const xfrm_address_t *saddr, int create, unsigned short family)
1466 {
1467         struct xfrm_state *x;
1468
1469         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1470         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1471         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1472
1473         return x;
1474 }
1475 EXPORT_SYMBOL(xfrm_find_acq);
1476
1477 #ifdef CONFIG_XFRM_SUB_POLICY
1478 int
1479 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1480                unsigned short family, struct net *net)
1481 {
1482         int err = 0;
1483         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1484         if (!afinfo)
1485                 return -EAFNOSUPPORT;
1486
1487         spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1488         if (afinfo->tmpl_sort)
1489                 err = afinfo->tmpl_sort(dst, src, n);
1490         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1491         xfrm_state_put_afinfo(afinfo);
1492         return err;
1493 }
1494 EXPORT_SYMBOL(xfrm_tmpl_sort);
1495
1496 int
1497 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1498                 unsigned short family)
1499 {
1500         int err = 0;
1501         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1502         struct net *net = xs_net(*src);
1503
1504         if (!afinfo)
1505                 return -EAFNOSUPPORT;
1506
1507         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1508         if (afinfo->state_sort)
1509                 err = afinfo->state_sort(dst, src, n);
1510         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1511         xfrm_state_put_afinfo(afinfo);
1512         return err;
1513 }
1514 EXPORT_SYMBOL(xfrm_state_sort);
1515 #endif
1516
1517 /* Silly enough, but I'm lazy to build resolution list */
1518
1519 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1520 {
1521         int i;
1522
1523         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1524                 struct xfrm_state *x;
1525
1526                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1527                         if (x->km.seq == seq &&
1528                             (mark & x->mark.m) == x->mark.v &&
1529                             x->km.state == XFRM_STATE_ACQ) {
1530                                 xfrm_state_hold(x);
1531                                 return x;
1532                         }
1533                 }
1534         }
1535         return NULL;
1536 }
1537
1538 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1539 {
1540         struct xfrm_state *x;
1541
1542         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1543         x = __xfrm_find_acq_byseq(net, mark, seq);
1544         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1545         return x;
1546 }
1547 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1548
1549 u32 xfrm_get_acqseq(void)
1550 {
1551         u32 res;
1552         static atomic_t acqseq;
1553
1554         do {
1555                 res = atomic_inc_return(&acqseq);
1556         } while (!res);
1557
1558         return res;
1559 }
1560 EXPORT_SYMBOL(xfrm_get_acqseq);
1561
1562 int verify_spi_info(u8 proto, u32 min, u32 max)
1563 {
1564         switch (proto) {
1565         case IPPROTO_AH:
1566         case IPPROTO_ESP:
1567                 break;
1568
1569         case IPPROTO_COMP:
1570                 /* IPCOMP spi is 16-bits. */
1571                 if (max >= 0x10000)
1572                         return -EINVAL;
1573                 break;
1574
1575         default:
1576                 return -EINVAL;
1577         }
1578
1579         if (min > max)
1580                 return -EINVAL;
1581
1582         return 0;
1583 }
1584 EXPORT_SYMBOL(verify_spi_info);
1585
1586 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1587 {
1588         struct net *net = xs_net(x);
1589         unsigned int h;
1590         struct xfrm_state *x0;
1591         int err = -ENOENT;
1592         __be32 minspi = htonl(low);
1593         __be32 maxspi = htonl(high);
1594         __be32 newspi = 0;
1595         u32 mark = x->mark.v & x->mark.m;
1596
1597         spin_lock_bh(&x->lock);
1598         if (x->km.state == XFRM_STATE_DEAD)
1599                 goto unlock;
1600
1601         err = 0;
1602         if (x->id.spi)
1603                 goto unlock;
1604
1605         err = -ENOENT;
1606
1607         if (minspi == maxspi) {
1608                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1609                 if (x0) {
1610                         xfrm_state_put(x0);
1611                         goto unlock;
1612                 }
1613                 newspi = minspi;
1614         } else {
1615                 u32 spi = 0;
1616                 for (h = 0; h < high-low+1; h++) {
1617                         spi = low + prandom_u32()%(high-low+1);
1618                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1619                         if (x0 == NULL) {
1620                                 newspi = htonl(spi);
1621                                 break;
1622                         }
1623                         xfrm_state_put(x0);
1624                 }
1625         }
1626         if (newspi) {
1627                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1628                 x->id.spi = newspi;
1629                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1630                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1631                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1632
1633                 err = 0;
1634         }
1635
1636 unlock:
1637         spin_unlock_bh(&x->lock);
1638
1639         return err;
1640 }
1641 EXPORT_SYMBOL(xfrm_alloc_spi);
1642
1643 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1644                                       struct xfrm_address_filter *filter)
1645 {
1646         if (filter) {
1647                 if ((filter->family == AF_INET ||
1648                      filter->family == AF_INET6) &&
1649                     x->props.family != filter->family)
1650                         return false;
1651
1652                 return addr_match(&x->props.saddr, &filter->saddr,
1653                                   filter->splen) &&
1654                        addr_match(&x->id.daddr, &filter->daddr,
1655                                   filter->dplen);
1656         }
1657         return true;
1658 }
1659
1660 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1661                     int (*func)(struct xfrm_state *, int, void*),
1662                     void *data)
1663 {
1664         struct xfrm_state *state;
1665         struct xfrm_state_walk *x;
1666         int err = 0;
1667
1668         if (walk->seq != 0 && list_empty(&walk->all))
1669                 return 0;
1670
1671         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1672         if (list_empty(&walk->all))
1673                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1674         else
1675                 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1676         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1677                 if (x->state == XFRM_STATE_DEAD)
1678                         continue;
1679                 state = container_of(x, struct xfrm_state, km);
1680                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1681                         continue;
1682                 if (!__xfrm_state_filter_match(state, walk->filter))
1683                         continue;
1684                 err = func(state, walk->seq, data);
1685                 if (err) {
1686                         list_move_tail(&walk->all, &x->all);
1687                         goto out;
1688                 }
1689                 walk->seq++;
1690         }
1691         if (walk->seq == 0) {
1692                 err = -ENOENT;
1693                 goto out;
1694         }
1695         list_del_init(&walk->all);
1696 out:
1697         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1698         return err;
1699 }
1700 EXPORT_SYMBOL(xfrm_state_walk);
1701
1702 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1703                           struct xfrm_address_filter *filter)
1704 {
1705         INIT_LIST_HEAD(&walk->all);
1706         walk->proto = proto;
1707         walk->state = XFRM_STATE_DEAD;
1708         walk->seq = 0;
1709         walk->filter = filter;
1710 }
1711 EXPORT_SYMBOL(xfrm_state_walk_init);
1712
1713 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1714 {
1715         kfree(walk->filter);
1716
1717         if (list_empty(&walk->all))
1718                 return;
1719
1720         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1721         list_del(&walk->all);
1722         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1723 }
1724 EXPORT_SYMBOL(xfrm_state_walk_done);
1725
1726 static void xfrm_replay_timer_handler(unsigned long data)
1727 {
1728         struct xfrm_state *x = (struct xfrm_state *)data;
1729
1730         spin_lock(&x->lock);
1731
1732         if (x->km.state == XFRM_STATE_VALID) {
1733                 if (xfrm_aevent_is_on(xs_net(x)))
1734                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1735                 else
1736                         x->xflags |= XFRM_TIME_DEFER;
1737         }
1738
1739         spin_unlock(&x->lock);
1740 }
1741
1742 static LIST_HEAD(xfrm_km_list);
1743
1744 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1745 {
1746         struct xfrm_mgr *km;
1747
1748         rcu_read_lock();
1749         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1750                 if (km->notify_policy)
1751                         km->notify_policy(xp, dir, c);
1752         rcu_read_unlock();
1753 }
1754
1755 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1756 {
1757         struct xfrm_mgr *km;
1758         rcu_read_lock();
1759         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1760                 if (km->notify)
1761                         km->notify(x, c);
1762         rcu_read_unlock();
1763 }
1764
1765 EXPORT_SYMBOL(km_policy_notify);
1766 EXPORT_SYMBOL(km_state_notify);
1767
1768 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1769 {
1770         struct km_event c;
1771
1772         c.data.hard = hard;
1773         c.portid = portid;
1774         c.event = XFRM_MSG_EXPIRE;
1775         km_state_notify(x, &c);
1776 }
1777
1778 EXPORT_SYMBOL(km_state_expired);
1779 /*
1780  * We send to all registered managers regardless of failure
1781  * We are happy with one success
1782 */
1783 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1784 {
1785         int err = -EINVAL, acqret;
1786         struct xfrm_mgr *km;
1787
1788         rcu_read_lock();
1789         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1790                 acqret = km->acquire(x, t, pol);
1791                 if (!acqret)
1792                         err = acqret;
1793         }
1794         rcu_read_unlock();
1795         return err;
1796 }
1797 EXPORT_SYMBOL(km_query);
1798
1799 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1800 {
1801         int err = -EINVAL;
1802         struct xfrm_mgr *km;
1803
1804         rcu_read_lock();
1805         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1806                 if (km->new_mapping)
1807                         err = km->new_mapping(x, ipaddr, sport);
1808                 if (!err)
1809                         break;
1810         }
1811         rcu_read_unlock();
1812         return err;
1813 }
1814 EXPORT_SYMBOL(km_new_mapping);
1815
1816 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1817 {
1818         struct km_event c;
1819
1820         c.data.hard = hard;
1821         c.portid = portid;
1822         c.event = XFRM_MSG_POLEXPIRE;
1823         km_policy_notify(pol, dir, &c);
1824 }
1825 EXPORT_SYMBOL(km_policy_expired);
1826
1827 #ifdef CONFIG_XFRM_MIGRATE
1828 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1829                const struct xfrm_migrate *m, int num_migrate,
1830                const struct xfrm_kmaddress *k)
1831 {
1832         int err = -EINVAL;
1833         int ret;
1834         struct xfrm_mgr *km;
1835
1836         rcu_read_lock();
1837         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1838                 if (km->migrate) {
1839                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1840                         if (!ret)
1841                                 err = ret;
1842                 }
1843         }
1844         rcu_read_unlock();
1845         return err;
1846 }
1847 EXPORT_SYMBOL(km_migrate);
1848 #endif
1849
1850 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1851 {
1852         int err = -EINVAL;
1853         int ret;
1854         struct xfrm_mgr *km;
1855
1856         rcu_read_lock();
1857         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1858                 if (km->report) {
1859                         ret = km->report(net, proto, sel, addr);
1860                         if (!ret)
1861                                 err = ret;
1862                 }
1863         }
1864         rcu_read_unlock();
1865         return err;
1866 }
1867 EXPORT_SYMBOL(km_report);
1868
1869 bool km_is_alive(const struct km_event *c)
1870 {
1871         struct xfrm_mgr *km;
1872         bool is_alive = false;
1873
1874         rcu_read_lock();
1875         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1876                 if (km->is_alive && km->is_alive(c)) {
1877                         is_alive = true;
1878                         break;
1879                 }
1880         }
1881         rcu_read_unlock();
1882
1883         return is_alive;
1884 }
1885 EXPORT_SYMBOL(km_is_alive);
1886
1887 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1888 {
1889         int err;
1890         u8 *data;
1891         struct xfrm_mgr *km;
1892         struct xfrm_policy *pol = NULL;
1893
1894 #ifdef CONFIG_COMPAT
1895         if (in_compat_syscall())
1896                 return -EOPNOTSUPP;
1897 #endif
1898
1899         if (!optval && !optlen) {
1900                 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
1901                 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
1902                 __sk_dst_reset(sk);
1903                 return 0;
1904         }
1905
1906         if (optlen <= 0 || optlen > PAGE_SIZE)
1907                 return -EMSGSIZE;
1908
1909         data = kmalloc(optlen, GFP_KERNEL);
1910         if (!data)
1911                 return -ENOMEM;
1912
1913         err = -EFAULT;
1914         if (copy_from_user(data, optval, optlen))
1915                 goto out;
1916
1917         err = -EINVAL;
1918         rcu_read_lock();
1919         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1920                 pol = km->compile_policy(sk, optname, data,
1921                                          optlen, &err);
1922                 if (err >= 0)
1923                         break;
1924         }
1925         rcu_read_unlock();
1926
1927         if (err >= 0) {
1928                 xfrm_sk_policy_insert(sk, err, pol);
1929                 xfrm_pol_put(pol);
1930                 err = 0;
1931         }
1932
1933 out:
1934         kfree(data);
1935         return err;
1936 }
1937 EXPORT_SYMBOL(xfrm_user_policy);
1938
1939 static DEFINE_SPINLOCK(xfrm_km_lock);
1940
1941 int xfrm_register_km(struct xfrm_mgr *km)
1942 {
1943         spin_lock_bh(&xfrm_km_lock);
1944         list_add_tail_rcu(&km->list, &xfrm_km_list);
1945         spin_unlock_bh(&xfrm_km_lock);
1946         return 0;
1947 }
1948 EXPORT_SYMBOL(xfrm_register_km);
1949
1950 int xfrm_unregister_km(struct xfrm_mgr *km)
1951 {
1952         spin_lock_bh(&xfrm_km_lock);
1953         list_del_rcu(&km->list);
1954         spin_unlock_bh(&xfrm_km_lock);
1955         synchronize_rcu();
1956         return 0;
1957 }
1958 EXPORT_SYMBOL(xfrm_unregister_km);
1959
1960 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1961 {
1962         int err = 0;
1963         if (unlikely(afinfo == NULL))
1964                 return -EINVAL;
1965         if (unlikely(afinfo->family >= NPROTO))
1966                 return -EAFNOSUPPORT;
1967         spin_lock_bh(&xfrm_state_afinfo_lock);
1968         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1969                 err = -EEXIST;
1970         else
1971                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1972         spin_unlock_bh(&xfrm_state_afinfo_lock);
1973         return err;
1974 }
1975 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1976
1977 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1978 {
1979         int err = 0;
1980         if (unlikely(afinfo == NULL))
1981                 return -EINVAL;
1982         if (unlikely(afinfo->family >= NPROTO))
1983                 return -EAFNOSUPPORT;
1984         spin_lock_bh(&xfrm_state_afinfo_lock);
1985         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1986                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1987                         err = -EINVAL;
1988                 else
1989                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1990         }
1991         spin_unlock_bh(&xfrm_state_afinfo_lock);
1992         synchronize_rcu();
1993         return err;
1994 }
1995 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1996
1997 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1998 {
1999         struct xfrm_state_afinfo *afinfo;
2000         if (unlikely(family >= NPROTO))
2001                 return NULL;
2002         rcu_read_lock();
2003         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2004         if (unlikely(!afinfo))
2005                 rcu_read_unlock();
2006         return afinfo;
2007 }
2008
2009 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
2010 {
2011         rcu_read_unlock();
2012 }
2013
2014 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2015 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2016 {
2017         if (x->tunnel) {
2018                 struct xfrm_state *t = x->tunnel;
2019
2020                 if (atomic_read(&t->tunnel_users) == 2)
2021                         xfrm_state_delete(t);
2022                 atomic_dec(&t->tunnel_users);
2023                 xfrm_state_put(t);
2024                 x->tunnel = NULL;
2025         }
2026 }
2027 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2028
2029 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
2030 {
2031         int res;
2032
2033         spin_lock_bh(&x->lock);
2034         if (x->km.state == XFRM_STATE_VALID &&
2035             x->type && x->type->get_mtu)
2036                 res = x->type->get_mtu(x, mtu);
2037         else
2038                 res = mtu - x->props.header_len;
2039         spin_unlock_bh(&x->lock);
2040         return res;
2041 }
2042
2043 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
2044 {
2045         struct xfrm_state_afinfo *afinfo;
2046         struct xfrm_mode *inner_mode;
2047         int family = x->props.family;
2048         int err;
2049
2050         err = -EAFNOSUPPORT;
2051         afinfo = xfrm_state_get_afinfo(family);
2052         if (!afinfo)
2053                 goto error;
2054
2055         err = 0;
2056         if (afinfo->init_flags)
2057                 err = afinfo->init_flags(x);
2058
2059         xfrm_state_put_afinfo(afinfo);
2060
2061         if (err)
2062                 goto error;
2063
2064         err = -EPROTONOSUPPORT;
2065
2066         if (x->sel.family != AF_UNSPEC) {
2067                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2068                 if (inner_mode == NULL)
2069                         goto error;
2070
2071                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2072                     family != x->sel.family) {
2073                         xfrm_put_mode(inner_mode);
2074                         goto error;
2075                 }
2076
2077                 x->inner_mode = inner_mode;
2078         } else {
2079                 struct xfrm_mode *inner_mode_iaf;
2080                 int iafamily = AF_INET;
2081
2082                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2083                 if (inner_mode == NULL)
2084                         goto error;
2085
2086                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2087                         xfrm_put_mode(inner_mode);
2088                         goto error;
2089                 }
2090                 x->inner_mode = inner_mode;
2091
2092                 if (x->props.family == AF_INET)
2093                         iafamily = AF_INET6;
2094
2095                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2096                 if (inner_mode_iaf) {
2097                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2098                                 x->inner_mode_iaf = inner_mode_iaf;
2099                         else
2100                                 xfrm_put_mode(inner_mode_iaf);
2101                 }
2102         }
2103
2104         x->type = xfrm_get_type(x->id.proto, family);
2105         if (x->type == NULL)
2106                 goto error;
2107
2108         err = x->type->init_state(x);
2109         if (err)
2110                 goto error;
2111
2112         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2113         if (x->outer_mode == NULL) {
2114                 err = -EPROTONOSUPPORT;
2115                 goto error;
2116         }
2117
2118         if (init_replay) {
2119                 err = xfrm_init_replay(x);
2120                 if (err)
2121                         goto error;
2122         }
2123
2124         x->km.state = XFRM_STATE_VALID;
2125
2126 error:
2127         return err;
2128 }
2129
2130 EXPORT_SYMBOL(__xfrm_init_state);
2131
2132 int xfrm_init_state(struct xfrm_state *x)
2133 {
2134         return __xfrm_init_state(x, true);
2135 }
2136
2137 EXPORT_SYMBOL(xfrm_init_state);
2138
2139 int __net_init xfrm_state_init(struct net *net)
2140 {
2141         unsigned int sz;
2142
2143         INIT_LIST_HEAD(&net->xfrm.state_all);
2144
2145         sz = sizeof(struct hlist_head) * 8;
2146
2147         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2148         if (!net->xfrm.state_bydst)
2149                 goto out_bydst;
2150         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2151         if (!net->xfrm.state_bysrc)
2152                 goto out_bysrc;
2153         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2154         if (!net->xfrm.state_byspi)
2155                 goto out_byspi;
2156         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2157
2158         net->xfrm.state_num = 0;
2159         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2160         spin_lock_init(&net->xfrm.xfrm_state_lock);
2161         return 0;
2162
2163 out_byspi:
2164         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2165 out_bysrc:
2166         xfrm_hash_free(net->xfrm.state_bydst, sz);
2167 out_bydst:
2168         return -ENOMEM;
2169 }
2170
2171 void xfrm_state_fini(struct net *net)
2172 {
2173         unsigned int sz;
2174
2175         flush_work(&net->xfrm.state_hash_work);
2176         xfrm_state_flush(net, 0, false);
2177         flush_work(&xfrm_state_gc_work);
2178
2179         WARN_ON(!list_empty(&net->xfrm.state_all));
2180
2181         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2182         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2183         xfrm_hash_free(net->xfrm.state_byspi, sz);
2184         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2185         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2186         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2187         xfrm_hash_free(net->xfrm.state_bydst, sz);
2188 }
2189
2190 #ifdef CONFIG_AUDITSYSCALL
2191 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2192                                      struct audit_buffer *audit_buf)
2193 {
2194         struct xfrm_sec_ctx *ctx = x->security;
2195         u32 spi = ntohl(x->id.spi);
2196
2197         if (ctx)
2198                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2199                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2200
2201         switch (x->props.family) {
2202         case AF_INET:
2203                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2204                                  &x->props.saddr.a4, &x->id.daddr.a4);
2205                 break;
2206         case AF_INET6:
2207                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2208                                  x->props.saddr.a6, x->id.daddr.a6);
2209                 break;
2210         }
2211
2212         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2213 }
2214
2215 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2216                                       struct audit_buffer *audit_buf)
2217 {
2218         const struct iphdr *iph4;
2219         const struct ipv6hdr *iph6;
2220
2221         switch (family) {
2222         case AF_INET:
2223                 iph4 = ip_hdr(skb);
2224                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2225                                  &iph4->saddr, &iph4->daddr);
2226                 break;
2227         case AF_INET6:
2228                 iph6 = ipv6_hdr(skb);
2229                 audit_log_format(audit_buf,
2230                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2231                                  &iph6->saddr, &iph6->daddr,
2232                                  iph6->flow_lbl[0] & 0x0f,
2233                                  iph6->flow_lbl[1],
2234                                  iph6->flow_lbl[2]);
2235                 break;
2236         }
2237 }
2238
2239 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2240 {
2241         struct audit_buffer *audit_buf;
2242
2243         audit_buf = xfrm_audit_start("SAD-add");
2244         if (audit_buf == NULL)
2245                 return;
2246         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2247         xfrm_audit_helper_sainfo(x, audit_buf);
2248         audit_log_format(audit_buf, " res=%u", result);
2249         audit_log_end(audit_buf);
2250 }
2251 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2252
2253 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2254 {
2255         struct audit_buffer *audit_buf;
2256
2257         audit_buf = xfrm_audit_start("SAD-delete");
2258         if (audit_buf == NULL)
2259                 return;
2260         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2261         xfrm_audit_helper_sainfo(x, audit_buf);
2262         audit_log_format(audit_buf, " res=%u", result);
2263         audit_log_end(audit_buf);
2264 }
2265 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2266
2267 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2268                                       struct sk_buff *skb)
2269 {
2270         struct audit_buffer *audit_buf;
2271         u32 spi;
2272
2273         audit_buf = xfrm_audit_start("SA-replay-overflow");
2274         if (audit_buf == NULL)
2275                 return;
2276         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2277         /* don't record the sequence number because it's inherent in this kind
2278          * of audit message */
2279         spi = ntohl(x->id.spi);
2280         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2281         audit_log_end(audit_buf);
2282 }
2283 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2284
2285 void xfrm_audit_state_replay(struct xfrm_state *x,
2286                              struct sk_buff *skb, __be32 net_seq)
2287 {
2288         struct audit_buffer *audit_buf;
2289         u32 spi;
2290
2291         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2292         if (audit_buf == NULL)
2293                 return;
2294         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2295         spi = ntohl(x->id.spi);
2296         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2297                          spi, spi, ntohl(net_seq));
2298         audit_log_end(audit_buf);
2299 }
2300 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2301
2302 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2303 {
2304         struct audit_buffer *audit_buf;
2305
2306         audit_buf = xfrm_audit_start("SA-notfound");
2307         if (audit_buf == NULL)
2308                 return;
2309         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2310         audit_log_end(audit_buf);
2311 }
2312 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2313
2314 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2315                                __be32 net_spi, __be32 net_seq)
2316 {
2317         struct audit_buffer *audit_buf;
2318         u32 spi;
2319
2320         audit_buf = xfrm_audit_start("SA-notfound");
2321         if (audit_buf == NULL)
2322                 return;
2323         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2324         spi = ntohl(net_spi);
2325         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2326                          spi, spi, ntohl(net_seq));
2327         audit_log_end(audit_buf);
2328 }
2329 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2330
2331 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2332                               struct sk_buff *skb, u8 proto)
2333 {
2334         struct audit_buffer *audit_buf;
2335         __be32 net_spi;
2336         __be32 net_seq;
2337
2338         audit_buf = xfrm_audit_start("SA-icv-failure");
2339         if (audit_buf == NULL)
2340                 return;
2341         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2342         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2343                 u32 spi = ntohl(net_spi);
2344                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2345                                  spi, spi, ntohl(net_seq));
2346         }
2347         audit_log_end(audit_buf);
2348 }
2349 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2350 #endif /* CONFIG_AUDITSYSCALL */