net/xfrm/xfrm_state.c

   1 /*
   2  * xfrm_state.c
   3  *
   4  * Changes:
   5  *      Mitsuru KANDA @USAGI
   6  *      Kazunori MIYAZAWA @USAGI
   7  *      Kunihiro Ishiguro <kunihiro@ipinfusion.com>
   8  *              IPv6 support
   9  *      YOSHIFUJI Hideaki @USAGI
  10  *              Split up af-specific functions
  11  *      Derek Atkins <derek@ihtfp.com>
  12  *              Add UDP Encapsulation
  13  *
  14  */
  15
  16 #include <linux/workqueue.h>
  17 #include <net/xfrm.h>
  18 #include <linux/pfkeyv2.h>
  19 #include <linux/ipsec.h>
  20 #include <linux/module.h>
  21 #include <linux/cache.h>
  22 #include <linux/audit.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/ktime.h>
  25 #include <linux/slab.h>
  26 #include <linux/interrupt.h>
  27 #include <linux/kernel.h>
  28
  29 #include "xfrm_hash.h"
  30
  31 /* Each xfrm_state may be linked to two tables:
  32
  33    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  34    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  35       destination/tunnel endpoint. (output)
  36  */
  37
  38 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  39
  40 static inline unsigned int xfrm_dst_hash(struct net *net,
  41                                          const xfrm_address_t *daddr,
  42                                          const xfrm_address_t *saddr,
  43                                          u32 reqid,
  44                                          unsigned short family)
  45 {
  46         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  47 }
  48
  49 static inline unsigned int xfrm_src_hash(struct net *net,
  50                                          const xfrm_address_t *daddr,
  51                                          const xfrm_address_t *saddr,
  52                                          unsigned short family)
  53 {
  54         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  55 }
  56
  57 static inline unsigned int
  58 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  59               __be32 spi, u8 proto, unsigned short family)
  60 {
  61         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  62 }
  63
  64 static void xfrm_hash_transfer(struct hlist_head *list,
  65                                struct hlist_head *ndsttable,
  66                                struct hlist_head *nsrctable,
  67                                struct hlist_head *nspitable,
  68                                unsigned int nhashmask)
  69 {
  70         struct hlist_node *tmp;
  71         struct xfrm_state *x;
  72
  73         hlist_for_each_entry_safe(x, tmp, list, bydst) {
  74                 unsigned int h;
  75
  76                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
  77                                     x->props.reqid, x->props.family,
  78                                     nhashmask);
  79                 hlist_add_head(&x->bydst, ndsttable+h);
  80
  81                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
  82                                     x->props.family,
  83                                     nhashmask);
  84                 hlist_add_head(&x->bysrc, nsrctable+h);
  85
  86                 if (x->id.spi) {
  87                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
  88                                             x->id.proto, x->props.family,
  89                                             nhashmask);
  90                         hlist_add_head(&x->byspi, nspitable+h);
  91                 }
  92         }
  93 }
  94
  95 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
  96 {
  97         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
  98 }
  99
 100 static void xfrm_hash_resize(struct work_struct *work)
 101 {
 102         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 103         struct hlist_head *ndst, *nsrc, *nspi, *odst, *osrc, *ospi;
 104         unsigned long nsize, osize;
 105         unsigned int nhashmask, ohashmask;
 106         int i;
 107
 108         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 109         ndst = xfrm_hash_alloc(nsize);
 110         if (!ndst)
 111                 return;
 112         nsrc = xfrm_hash_alloc(nsize);
 113         if (!nsrc) {
 114                 xfrm_hash_free(ndst, nsize);
 115                 return;
 116         }
 117         nspi = xfrm_hash_alloc(nsize);
 118         if (!nspi) {
 119                 xfrm_hash_free(ndst, nsize);
 120                 xfrm_hash_free(nsrc, nsize);
 121                 return;
 122         }
 123
 124         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 125
 126         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 127         for (i = net->xfrm.state_hmask; i >= 0; i--)
 128                 xfrm_hash_transfer(net->xfrm.state_bydst+i, ndst, nsrc, nspi,
 129                                    nhashmask);
 130
 131         odst = net->xfrm.state_bydst;
 132         osrc = net->xfrm.state_bysrc;
 133         ospi = net->xfrm.state_byspi;
 134         ohashmask = net->xfrm.state_hmask;
 135
 136         net->xfrm.state_bydst = ndst;
 137         net->xfrm.state_bysrc = nsrc;
 138         net->xfrm.state_byspi = nspi;
 139         net->xfrm.state_hmask = nhashmask;
 140
 141         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 142
 143         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 144         xfrm_hash_free(odst, osize);
 145         xfrm_hash_free(osrc, osize);
 146         xfrm_hash_free(ospi, osize);
 147 }
 148
 149 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 150 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
 151
 152 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 153
 154 int __xfrm_state_delete(struct xfrm_state *x);
 155
 156 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 157 bool km_is_alive(const struct km_event *c);
 158 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 159
 160 static DEFINE_SPINLOCK(xfrm_type_lock);
 161 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 162 {
 163         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 164         const struct xfrm_type **typemap;
 165         int err = 0;
 166
 167         if (unlikely(afinfo == NULL))
 168                 return -EAFNOSUPPORT;
 169         typemap = afinfo->type_map;
 170         spin_lock_bh(&xfrm_type_lock);
 171
 172         if (likely(typemap[type->proto] == NULL))
 173                 typemap[type->proto] = type;
 174         else
 175                 err = -EEXIST;
 176         spin_unlock_bh(&xfrm_type_lock);
 177         xfrm_state_put_afinfo(afinfo);
 178         return err;
 179 }
 180 EXPORT_SYMBOL(xfrm_register_type);
 181
 182 int xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 183 {
 184         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 185         const struct xfrm_type **typemap;
 186         int err = 0;
 187
 188         if (unlikely(afinfo == NULL))
 189                 return -EAFNOSUPPORT;
 190         typemap = afinfo->type_map;
 191         spin_lock_bh(&xfrm_type_lock);
 192
 193         if (unlikely(typemap[type->proto] != type))
 194                 err = -ENOENT;
 195         else
 196                 typemap[type->proto] = NULL;
 197         spin_unlock_bh(&xfrm_type_lock);
 198         xfrm_state_put_afinfo(afinfo);
 199         return err;
 200 }
 201 EXPORT_SYMBOL(xfrm_unregister_type);
 202
 203 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 204 {
 205         struct xfrm_state_afinfo *afinfo;
 206         const struct xfrm_type **typemap;
 207         const struct xfrm_type *type;
 208         int modload_attempted = 0;
 209
 210 retry:
 211         afinfo = xfrm_state_get_afinfo(family);
 212         if (unlikely(afinfo == NULL))
 213                 return NULL;
 214         typemap = afinfo->type_map;
 215
 216         type = typemap[proto];
 217         if (unlikely(type && !try_module_get(type->owner)))
 218                 type = NULL;
 219         if (!type && !modload_attempted) {
 220                 xfrm_state_put_afinfo(afinfo);
 221                 request_module("xfrm-type-%d-%d", family, proto);
 222                 modload_attempted = 1;
 223                 goto retry;
 224         }
 225
 226         xfrm_state_put_afinfo(afinfo);
 227         return type;
 228 }
 229
 230 static void xfrm_put_type(const struct xfrm_type *type)
 231 {
 232         module_put(type->owner);
 233 }
 234
 235 static DEFINE_SPINLOCK(xfrm_mode_lock);
 236 int xfrm_register_mode(struct xfrm_mode *mode, int family)
 237 {
 238         struct xfrm_state_afinfo *afinfo;
 239         struct xfrm_mode **modemap;
 240         int err;
 241
 242         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 243                 return -EINVAL;
 244
 245         afinfo = xfrm_state_get_afinfo(family);
 246         if (unlikely(afinfo == NULL))
 247                 return -EAFNOSUPPORT;
 248
 249         err = -EEXIST;
 250         modemap = afinfo->mode_map;
 251         spin_lock_bh(&xfrm_mode_lock);
 252         if (modemap[mode->encap])
 253                 goto out;
 254
 255         err = -ENOENT;
 256         if (!try_module_get(afinfo->owner))
 257                 goto out;
 258
 259         mode->afinfo = afinfo;
 260         modemap[mode->encap] = mode;
 261         err = 0;
 262
 263 out:
 264         spin_unlock_bh(&xfrm_mode_lock);
 265         xfrm_state_put_afinfo(afinfo);
 266         return err;
 267 }
 268 EXPORT_SYMBOL(xfrm_register_mode);
 269
 270 int xfrm_unregister_mode(struct xfrm_mode *mode, int family)
 271 {
 272         struct xfrm_state_afinfo *afinfo;
 273         struct xfrm_mode **modemap;
 274         int err;
 275
 276         if (unlikely(mode->encap >= XFRM_MODE_MAX))
 277                 return -EINVAL;
 278
 279         afinfo = xfrm_state_get_afinfo(family);
 280         if (unlikely(afinfo == NULL))
 281                 return -EAFNOSUPPORT;
 282
 283         err = -ENOENT;
 284         modemap = afinfo->mode_map;
 285         spin_lock_bh(&xfrm_mode_lock);
 286         if (likely(modemap[mode->encap] == mode)) {
 287                 modemap[mode->encap] = NULL;
 288                 module_put(mode->afinfo->owner);
 289                 err = 0;
 290         }
 291
 292         spin_unlock_bh(&xfrm_mode_lock);
 293         xfrm_state_put_afinfo(afinfo);
 294         return err;
 295 }
 296 EXPORT_SYMBOL(xfrm_unregister_mode);
 297
 298 static struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 299 {
 300         struct xfrm_state_afinfo *afinfo;
 301         struct xfrm_mode *mode;
 302         int modload_attempted = 0;
 303
 304         if (unlikely(encap >= XFRM_MODE_MAX))
 305                 return NULL;
 306
 307 retry:
 308         afinfo = xfrm_state_get_afinfo(family);
 309         if (unlikely(afinfo == NULL))
 310                 return NULL;
 311
 312         mode = afinfo->mode_map[encap];
 313         if (unlikely(mode && !try_module_get(mode->owner)))
 314                 mode = NULL;
 315         if (!mode && !modload_attempted) {
 316                 xfrm_state_put_afinfo(afinfo);
 317                 request_module("xfrm-mode-%d-%d", family, encap);
 318                 modload_attempted = 1;
 319                 goto retry;
 320         }
 321
 322         xfrm_state_put_afinfo(afinfo);
 323         return mode;
 324 }
 325
 326 static void xfrm_put_mode(struct xfrm_mode *mode)
 327 {
 328         module_put(mode->owner);
 329 }
 330
 331 static void xfrm_state_gc_destroy(struct xfrm_state *x)
 332 {
 333         tasklet_hrtimer_cancel(&x->mtimer);
 334         del_timer_sync(&x->rtimer);
 335         kfree(x->aead);
 336         kfree(x->aalg);
 337         kfree(x->ealg);
 338         kfree(x->calg);
 339         kfree(x->encap);
 340         kfree(x->coaddr);
 341         kfree(x->replay_esn);
 342         kfree(x->preplay_esn);
 343         if (x->inner_mode)
 344                 xfrm_put_mode(x->inner_mode);
 345         if (x->inner_mode_iaf)
 346                 xfrm_put_mode(x->inner_mode_iaf);
 347         if (x->outer_mode)
 348                 xfrm_put_mode(x->outer_mode);
 349         if (x->type) {
 350                 x->type->destructor(x);
 351                 xfrm_put_type(x->type);
 352         }
 353         security_xfrm_state_free(x);
 354         kfree(x);
 355 }
 356
 357 static void xfrm_state_gc_task(struct work_struct *work)
 358 {
 359         struct net *net = container_of(work, struct net, xfrm.state_gc_work);
 360         struct xfrm_state *x;
 361         struct hlist_node *tmp;
 362         struct hlist_head gc_list;
 363
 364         spin_lock_bh(&xfrm_state_gc_lock);
 365         hlist_move_list(&net->xfrm.state_gc_list, &gc_list);
 366         spin_unlock_bh(&xfrm_state_gc_lock);
 367
 368         hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 369                 xfrm_state_gc_destroy(x);
 370 }
 371
 372 static inline unsigned long make_jiffies(long secs)
 373 {
 374         if (secs >= (MAX_SCHEDULE_TIMEOUT-1)/HZ)
 375                 return MAX_SCHEDULE_TIMEOUT-1;
 376         else
 377                 return secs*HZ;
 378 }
 379
 380 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
 381 {
 382         struct tasklet_hrtimer *thr = container_of(me, struct tasklet_hrtimer, timer);
 383         struct xfrm_state *x = container_of(thr, struct xfrm_state, mtimer);
 384         unsigned long now = get_seconds();
 385         long next = LONG_MAX;
 386         int warn = 0;
 387         int err = 0;
 388
 389         spin_lock(&x->lock);
 390         if (x->km.state == XFRM_STATE_DEAD)
 391                 goto out;
 392         if (x->km.state == XFRM_STATE_EXPIRED)
 393                 goto expired;
 394         if (x->lft.hard_add_expires_seconds) {
 395                 long tmo = x->lft.hard_add_expires_seconds +
 396                         x->curlft.add_time - now;
 397                 if (tmo <= 0) {
 398                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 399                                 /* enter hard expire without soft expire first?!
 400                                  * setting a new date could trigger this.
 401                                  * workarbound: fix x->curflt.add_time by below:
 402                                  */
 403                                 x->curlft.add_time = now - x->saved_tmo - 1;
 404                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 405                         } else
 406                                 goto expired;
 407                 }
 408                 if (tmo < next)
 409                         next = tmo;
 410         }
 411         if (x->lft.hard_use_expires_seconds) {
 412                 long tmo = x->lft.hard_use_expires_seconds +
 413                         (x->curlft.use_time ? : now) - now;
 414                 if (tmo <= 0)
 415                         goto expired;
 416                 if (tmo < next)
 417                         next = tmo;
 418         }
 419         if (x->km.dying)
 420                 goto resched;
 421         if (x->lft.soft_add_expires_seconds) {
 422                 long tmo = x->lft.soft_add_expires_seconds +
 423                         x->curlft.add_time - now;
 424                 if (tmo <= 0) {
 425                         warn = 1;
 426                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 427                 } else if (tmo < next) {
 428                         next = tmo;
 429                         x->xflags |= XFRM_SOFT_EXPIRE;
 430                         x->saved_tmo = tmo;
 431                 }
 432         }
 433         if (x->lft.soft_use_expires_seconds) {
 434                 long tmo = x->lft.soft_use_expires_seconds +
 435                         (x->curlft.use_time ? : now) - now;
 436                 if (tmo <= 0)
 437                         warn = 1;
 438                 else if (tmo < next)
 439                         next = tmo;
 440         }
 441
 442         x->km.dying = warn;
 443         if (warn)
 444                 km_state_expired(x, 0, 0);
 445 resched:
 446         if (next != LONG_MAX) {
 447                 tasklet_hrtimer_start(&x->mtimer, ktime_set(next, 0), HRTIMER_MODE_REL);
 448         }
 449
 450         goto out;
 451
 452 expired:
 453         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
 454                 x->km.state = XFRM_STATE_EXPIRED;
 455
 456         err = __xfrm_state_delete(x);
 457         if (!err)
 458                 km_state_expired(x, 1, 0);
 459
 460         xfrm_audit_state_delete(x, err ? 0 : 1, true);
 461
 462 out:
 463         spin_unlock(&x->lock);
 464         return HRTIMER_NORESTART;
 465 }
 466
 467 static void xfrm_replay_timer_handler(unsigned long data);
 468
 469 struct xfrm_state *xfrm_state_alloc(struct net *net)
 470 {
 471         struct xfrm_state *x;
 472
 473         x = kzalloc(sizeof(struct xfrm_state), GFP_ATOMIC);
 474
 475         if (x) {
 476                 write_pnet(&x->xs_net, net);
 477                 atomic_set(&x->refcnt, 1);
 478                 atomic_set(&x->tunnel_users, 0);
 479                 INIT_LIST_HEAD(&x->km.all);
 480                 INIT_HLIST_NODE(&x->bydst);
 481                 INIT_HLIST_NODE(&x->bysrc);
 482                 INIT_HLIST_NODE(&x->byspi);
 483                 tasklet_hrtimer_init(&x->mtimer, xfrm_timer_handler,
 484                                         CLOCK_BOOTTIME, HRTIMER_MODE_ABS);
 485                 setup_timer(&x->rtimer, xfrm_replay_timer_handler,
 486                                 (unsigned long)x);
 487                 x->curlft.add_time = get_seconds();
 488                 x->lft.soft_byte_limit = XFRM_INF;
 489                 x->lft.soft_packet_limit = XFRM_INF;
 490                 x->lft.hard_byte_limit = XFRM_INF;
 491                 x->lft.hard_packet_limit = XFRM_INF;
 492                 x->replay_maxage = 0;
 493                 x->replay_maxdiff = 0;
 494                 x->inner_mode = NULL;
 495                 x->inner_mode_iaf = NULL;
 496                 spin_lock_init(&x->lock);
 497         }
 498         return x;
 499 }
 500 EXPORT_SYMBOL(xfrm_state_alloc);
 501
 502 void __xfrm_state_destroy(struct xfrm_state *x)
 503 {
 504         struct net *net = xs_net(x);
 505
 506         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 507
 508         spin_lock_bh(&xfrm_state_gc_lock);
 509         hlist_add_head(&x->gclist, &net->xfrm.state_gc_list);
 510         spin_unlock_bh(&xfrm_state_gc_lock);
 511         schedule_work(&net->xfrm.state_gc_work);
 512 }
 513 EXPORT_SYMBOL(__xfrm_state_destroy);
 514
 515 int __xfrm_state_delete(struct xfrm_state *x)
 516 {
 517         struct net *net = xs_net(x);
 518         int err = -ESRCH;
 519
 520         if (x->km.state != XFRM_STATE_DEAD) {
 521                 x->km.state = XFRM_STATE_DEAD;
 522                 spin_lock(&net->xfrm.xfrm_state_lock);
 523                 list_del(&x->km.all);
 524                 hlist_del(&x->bydst);
 525                 hlist_del(&x->bysrc);
 526                 if (x->id.spi)
 527                         hlist_del(&x->byspi);
 528                 net->xfrm.state_num--;
 529                 spin_unlock(&net->xfrm.xfrm_state_lock);
 530
 531                 /* All xfrm_state objects are created by xfrm_state_alloc.
 532                  * The xfrm_state_alloc call gives a reference, and that
 533                  * is what we are dropping here.
 534                  */
 535                 xfrm_state_put(x);
 536                 err = 0;
 537         }
 538
 539         return err;
 540 }
 541 EXPORT_SYMBOL(__xfrm_state_delete);
 542
 543 int xfrm_state_delete(struct xfrm_state *x)
 544 {
 545         int err;
 546
 547         spin_lock_bh(&x->lock);
 548         err = __xfrm_state_delete(x);
 549         spin_unlock_bh(&x->lock);
 550
 551         return err;
 552 }
 553 EXPORT_SYMBOL(xfrm_state_delete);
 554
 555 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 556 static inline int
 557 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 558 {
 559         int i, err = 0;
 560
 561         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 562                 struct xfrm_state *x;
 563
 564                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 565                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 566                            (err = security_xfrm_state_delete(x)) != 0) {
 567                                 xfrm_audit_state_delete(x, 0, task_valid);
 568                                 return err;
 569                         }
 570                 }
 571         }
 572
 573         return err;
 574 }
 575 #else
 576 static inline int
 577 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 578 {
 579         return 0;
 580 }
 581 #endif
 582
 583 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid)
 584 {
 585         int i, err = 0, cnt = 0;
 586
 587         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 588         err = xfrm_state_flush_secctx_check(net, proto, task_valid);
 589         if (err)
 590                 goto out;
 591
 592         err = -ESRCH;
 593         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 594                 struct xfrm_state *x;
 595 restart:
 596                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 597                         if (!xfrm_state_kern(x) &&
 598                             xfrm_id_proto_match(x->id.proto, proto)) {
 599                                 xfrm_state_hold(x);
 600                                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 601
 602                                 err = xfrm_state_delete(x);
 603                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 604                                                         task_valid);
 605                                 xfrm_state_put(x);
 606                                 if (!err)
 607                                         cnt++;
 608
 609                                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
 610                                 goto restart;
 611                         }
 612                 }
 613         }
 614         if (cnt)
 615                 err = 0;
 616
 617 out:
 618         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 619         return err;
 620 }
 621 EXPORT_SYMBOL(xfrm_state_flush);
 622
 623 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 624 {
 625         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 626         si->sadcnt = net->xfrm.state_num;
 627         si->sadhcnt = net->xfrm.state_hmask;
 628         si->sadhmcnt = xfrm_state_hashmax;
 629         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 630 }
 631 EXPORT_SYMBOL(xfrm_sad_getinfo);
 632
 633 static int
 634 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 635                     const struct xfrm_tmpl *tmpl,
 636                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 637                     unsigned short family)
 638 {
 639         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 640         if (!afinfo)
 641                 return -1;
 642         afinfo->init_tempsel(&x->sel, fl);
 643
 644         if (family != tmpl->encap_family) {
 645                 xfrm_state_put_afinfo(afinfo);
 646                 afinfo = xfrm_state_get_afinfo(tmpl->encap_family);
 647                 if (!afinfo)
 648                         return -1;
 649         }
 650         afinfo->init_temprop(x, tmpl, daddr, saddr);
 651         xfrm_state_put_afinfo(afinfo);
 652         return 0;
 653 }
 654
 655 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 656                                               const xfrm_address_t *daddr,
 657                                               __be32 spi, u8 proto,
 658                                               unsigned short family)
 659 {
 660         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 661         struct xfrm_state *x;
 662
 663         hlist_for_each_entry(x, net->xfrm.state_byspi+h, byspi) {
 664                 if (x->props.family != family ||
 665                     x->id.spi       != spi ||
 666                     x->id.proto     != proto ||
 667                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 668                         continue;
 669
 670                 if ((mark & x->mark.m) != x->mark.v)
 671                         continue;
 672                 xfrm_state_hold(x);
 673                 return x;
 674         }
 675
 676         return NULL;
 677 }
 678
 679 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 680                                                      const xfrm_address_t *daddr,
 681                                                      const xfrm_address_t *saddr,
 682                                                      u8 proto, unsigned short family)
 683 {
 684         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 685         struct xfrm_state *x;
 686
 687         hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
 688                 if (x->props.family != family ||
 689                     x->id.proto     != proto ||
 690                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 691                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 692                         continue;
 693
 694                 if ((mark & x->mark.m) != x->mark.v)
 695                         continue;
 696                 xfrm_state_hold(x);
 697                 return x;
 698         }
 699
 700         return NULL;
 701 }
 702
 703 static inline struct xfrm_state *
 704 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
 705 {
 706         struct net *net = xs_net(x);
 707         u32 mark = x->mark.v & x->mark.m;
 708
 709         if (use_spi)
 710                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
 711                                            x->id.spi, x->id.proto, family);
 712         else
 713                 return __xfrm_state_lookup_byaddr(net, mark,
 714                                                   &x->id.daddr,
 715                                                   &x->props.saddr,
 716                                                   x->id.proto, family);
 717 }
 718
 719 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
 720 {
 721         if (have_hash_collision &&
 722             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
 723             net->xfrm.state_num > net->xfrm.state_hmask)
 724                 schedule_work(&net->xfrm.state_hash_work);
 725 }
 726
 727 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
 728                                const struct flowi *fl, unsigned short family,
 729                                struct xfrm_state **best, int *acq_in_progress,
 730                                int *error)
 731 {
 732         /* Resolution logic:
 733          * 1. There is a valid state with matching selector. Done.
 734          * 2. Valid state with inappropriate selector. Skip.
 735          *
 736          * Entering area of "sysdeps".
 737          *
 738          * 3. If state is not valid, selector is temporary, it selects
 739          *    only session which triggered previous resolution. Key
 740          *    manager will do something to install a state with proper
 741          *    selector.
 742          */
 743         if (x->km.state == XFRM_STATE_VALID) {
 744                 if ((x->sel.family &&
 745                      !xfrm_selector_match(&x->sel, fl, x->sel.family)) ||
 746                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 747                         return;
 748
 749                 if (!*best ||
 750                     (*best)->km.dying > x->km.dying ||
 751                     ((*best)->km.dying == x->km.dying &&
 752                      (*best)->curlft.add_time < x->curlft.add_time))
 753                         *best = x;
 754         } else if (x->km.state == XFRM_STATE_ACQ) {
 755                 *acq_in_progress = 1;
 756         } else if (x->km.state == XFRM_STATE_ERROR ||
 757                    x->km.state == XFRM_STATE_EXPIRED) {
 758                 if (xfrm_selector_match(&x->sel, fl, x->sel.family) &&
 759                     security_xfrm_state_pol_flow_match(x, pol, fl))
 760                         *error = -ESRCH;
 761         }
 762 }
 763
 764 struct xfrm_state *
 765 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 766                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 767                 struct xfrm_policy *pol, int *err,
 768                 unsigned short family)
 769 {
 770         static xfrm_address_t saddr_wildcard = { };
 771         struct net *net = xp_net(pol);
 772         unsigned int h, h_wildcard;
 773         struct xfrm_state *x, *x0, *to_put;
 774         int acquire_in_progress = 0;
 775         int error = 0;
 776         struct xfrm_state *best = NULL;
 777         u32 mark = pol->mark.v & pol->mark.m;
 778         unsigned short encap_family = tmpl->encap_family;
 779         struct km_event c;
 780
 781         to_put = NULL;
 782
 783         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 784         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 785         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 786                 if (x->props.family == encap_family &&
 787                     x->props.reqid == tmpl->reqid &&
 788                     (mark & x->mark.m) == x->mark.v &&
 789                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 790                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 791                     tmpl->mode == x->props.mode &&
 792                     tmpl->id.proto == x->id.proto &&
 793                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 794                         xfrm_state_look_at(pol, x, fl, encap_family,
 795                                            &best, &acquire_in_progress, &error);
 796         }
 797         if (best || acquire_in_progress)
 798                 goto found;
 799
 800         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 801         hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
 802                 if (x->props.family == encap_family &&
 803                     x->props.reqid == tmpl->reqid &&
 804                     (mark & x->mark.m) == x->mark.v &&
 805                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 806                     xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
 807                     tmpl->mode == x->props.mode &&
 808                     tmpl->id.proto == x->id.proto &&
 809                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 810                         xfrm_state_look_at(pol, x, fl, encap_family,
 811                                            &best, &acquire_in_progress, &error);
 812         }
 813
 814 found:
 815         x = best;
 816         if (!x && !error && !acquire_in_progress) {
 817                 if (tmpl->id.spi &&
 818                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 819                                               tmpl->id.proto, encap_family)) != NULL) {
 820                         to_put = x0;
 821                         error = -EEXIST;
 822                         goto out;
 823                 }
 824
 825                 c.net = net;
 826                 /* If the KMs have no listeners (yet...), avoid allocating an SA
 827                  * for each and every packet - garbage collection might not
 828                  * handle the flood.
 829                  */
 830                 if (!km_is_alive(&c)) {
 831                         error = -ESRCH;
 832                         goto out;
 833                 }
 834
 835                 x = xfrm_state_alloc(net);
 836                 if (x == NULL) {
 837                         error = -ENOMEM;
 838                         goto out;
 839                 }
 840                 /* Initialize temporary state matching only
 841                  * to current session. */
 842                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 843                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 844
 845                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 846                 if (error) {
 847                         x->km.state = XFRM_STATE_DEAD;
 848                         to_put = x;
 849                         x = NULL;
 850                         goto out;
 851                 }
 852
 853                 if (km_query(x, tmpl, pol) == 0) {
 854                         x->km.state = XFRM_STATE_ACQ;
 855                         list_add(&x->km.all, &net->xfrm.state_all);
 856                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 857                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 858                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 859                         if (x->id.spi) {
 860                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 861                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 862                         }
 863                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 864                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 865                         net->xfrm.state_num++;
 866                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 867                 } else {
 868                         x->km.state = XFRM_STATE_DEAD;
 869                         to_put = x;
 870                         x = NULL;
 871                         error = -ESRCH;
 872                 }
 873         }
 874 out:
 875         if (x)
 876                 xfrm_state_hold(x);
 877         else
 878                 *err = acquire_in_progress ? -EAGAIN : error;
 879         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 880         if (to_put)
 881                 xfrm_state_put(to_put);
 882         return x;
 883 }
 884
 885 struct xfrm_state *
 886 xfrm_stateonly_find(struct net *net, u32 mark,
 887                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 888                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 889 {
 890         unsigned int h;
 891         struct xfrm_state *rx = NULL, *x = NULL;
 892
 893         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 894         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 895         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 896                 if (x->props.family == family &&
 897                     x->props.reqid == reqid &&
 898                     (mark & x->mark.m) == x->mark.v &&
 899                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 900                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 901                     mode == x->props.mode &&
 902                     proto == x->id.proto &&
 903                     x->km.state == XFRM_STATE_VALID) {
 904                         rx = x;
 905                         break;
 906                 }
 907         }
 908
 909         if (rx)
 910                 xfrm_state_hold(rx);
 911         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 912
 913
 914         return rx;
 915 }
 916 EXPORT_SYMBOL(xfrm_stateonly_find);
 917
 918 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
 919                                               unsigned short family)
 920 {
 921         struct xfrm_state *x;
 922         struct xfrm_state_walk *w;
 923
 924         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 925         list_for_each_entry(w, &net->xfrm.state_all, all) {
 926                 x = container_of(w, struct xfrm_state, km);
 927                 if (x->props.family != family ||
 928                         x->id.spi != spi)
 929                         continue;
 930
 931                 xfrm_state_hold(x);
 932                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 933                 return x;
 934         }
 935         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 936         return NULL;
 937 }
 938 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
 939
 940 static void __xfrm_state_insert(struct xfrm_state *x)
 941 {
 942         struct net *net = xs_net(x);
 943         unsigned int h;
 944
 945         list_add(&x->km.all, &net->xfrm.state_all);
 946
 947         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 948                           x->props.reqid, x->props.family);
 949         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 950
 951         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 952         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 953
 954         if (x->id.spi) {
 955                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 956                                   x->props.family);
 957
 958                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 959         }
 960
 961         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 962         if (x->replay_maxage)
 963                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 964
 965         net->xfrm.state_num++;
 966
 967         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 968 }
 969
 970 /* net->xfrm.xfrm_state_lock is held */
 971 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 972 {
 973         struct net *net = xs_net(xnew);
 974         unsigned short family = xnew->props.family;
 975         u32 reqid = xnew->props.reqid;
 976         struct xfrm_state *x;
 977         unsigned int h;
 978         u32 mark = xnew->mark.v & xnew->mark.m;
 979
 980         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 981         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 982                 if (x->props.family     == family &&
 983                     x->props.reqid      == reqid &&
 984                     (mark & x->mark.m) == x->mark.v &&
 985                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 986                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
 987                         x->genid++;
 988         }
 989 }
 990
 991 void xfrm_state_insert(struct xfrm_state *x)
 992 {
 993         struct net *net = xs_net(x);
 994
 995         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 996         __xfrm_state_bump_genids(x);
 997         __xfrm_state_insert(x);
 998         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 999 }
1000 EXPORT_SYMBOL(xfrm_state_insert);
1001
1002 /* net->xfrm.xfrm_state_lock is held */
1003 static struct xfrm_state *__find_acq_core(struct net *net,
1004                                           const struct xfrm_mark *m,
1005                                           unsigned short family, u8 mode,
1006                                           u32 reqid, u8 proto,
1007                                           const xfrm_address_t *daddr,
1008                                           const xfrm_address_t *saddr,
1009                                           int create)
1010 {
1011         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1012         struct xfrm_state *x;
1013         u32 mark = m->v & m->m;
1014
1015         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1016                 if (x->props.reqid  != reqid ||
1017                     x->props.mode   != mode ||
1018                     x->props.family != family ||
1019                     x->km.state     != XFRM_STATE_ACQ ||
1020                     x->id.spi       != 0 ||
1021                     x->id.proto     != proto ||
1022                     (mark & x->mark.m) != x->mark.v ||
1023                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1024                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1025                         continue;
1026
1027                 xfrm_state_hold(x);
1028                 return x;
1029         }
1030
1031         if (!create)
1032                 return NULL;
1033
1034         x = xfrm_state_alloc(net);
1035         if (likely(x)) {
1036                 switch (family) {
1037                 case AF_INET:
1038                         x->sel.daddr.a4 = daddr->a4;
1039                         x->sel.saddr.a4 = saddr->a4;
1040                         x->sel.prefixlen_d = 32;
1041                         x->sel.prefixlen_s = 32;
1042                         x->props.saddr.a4 = saddr->a4;
1043                         x->id.daddr.a4 = daddr->a4;
1044                         break;
1045
1046                 case AF_INET6:
1047                         x->sel.daddr.in6 = daddr->in6;
1048                         x->sel.saddr.in6 = saddr->in6;
1049                         x->sel.prefixlen_d = 128;
1050                         x->sel.prefixlen_s = 128;
1051                         x->props.saddr.in6 = saddr->in6;
1052                         x->id.daddr.in6 = daddr->in6;
1053                         break;
1054                 }
1055
1056                 x->km.state = XFRM_STATE_ACQ;
1057                 x->id.proto = proto;
1058                 x->props.family = family;
1059                 x->props.mode = mode;
1060                 x->props.reqid = reqid;
1061                 x->mark.v = m->v;
1062                 x->mark.m = m->m;
1063                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1064                 xfrm_state_hold(x);
1065                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1066                 list_add(&x->km.all, &net->xfrm.state_all);
1067                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1068                 h = xfrm_src_hash(net, daddr, saddr, family);
1069                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1070
1071                 net->xfrm.state_num++;
1072
1073                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1074         }
1075
1076         return x;
1077 }
1078
1079 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1080
1081 int xfrm_state_add(struct xfrm_state *x)
1082 {
1083         struct net *net = xs_net(x);
1084         struct xfrm_state *x1, *to_put;
1085         int family;
1086         int err;
1087         u32 mark = x->mark.v & x->mark.m;
1088         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1089
1090         family = x->props.family;
1091
1092         to_put = NULL;
1093
1094         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1095
1096         x1 = __xfrm_state_locate(x, use_spi, family);
1097         if (x1) {
1098                 to_put = x1;
1099                 x1 = NULL;
1100                 err = -EEXIST;
1101                 goto out;
1102         }
1103
1104         if (use_spi && x->km.seq) {
1105                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1106                 if (x1 && ((x1->id.proto != x->id.proto) ||
1107                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1108                         to_put = x1;
1109                         x1 = NULL;
1110                 }
1111         }
1112
1113         if (use_spi && !x1)
1114                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1115                                      x->props.reqid, x->id.proto,
1116                                      &x->id.daddr, &x->props.saddr, 0);
1117
1118         __xfrm_state_bump_genids(x);
1119         __xfrm_state_insert(x);
1120         err = 0;
1121
1122 out:
1123         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1124
1125         if (x1) {
1126                 xfrm_state_delete(x1);
1127                 xfrm_state_put(x1);
1128         }
1129
1130         if (to_put)
1131                 xfrm_state_put(to_put);
1132
1133         return err;
1134 }
1135 EXPORT_SYMBOL(xfrm_state_add);
1136
1137 #ifdef CONFIG_XFRM_MIGRATE
1138 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1139 {
1140         struct net *net = xs_net(orig);
1141         struct xfrm_state *x = xfrm_state_alloc(net);
1142         if (!x)
1143                 goto out;
1144
1145         memcpy(&x->id, &orig->id, sizeof(x->id));
1146         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1147         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1148         x->props.mode = orig->props.mode;
1149         x->props.replay_window = orig->props.replay_window;
1150         x->props.reqid = orig->props.reqid;
1151         x->props.family = orig->props.family;
1152         x->props.saddr = orig->props.saddr;
1153
1154         if (orig->aalg) {
1155                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1156                 if (!x->aalg)
1157                         goto error;
1158         }
1159         x->props.aalgo = orig->props.aalgo;
1160
1161         if (orig->aead) {
1162                 x->aead = xfrm_algo_aead_clone(orig->aead);
1163                 if (!x->aead)
1164                         goto error;
1165         }
1166         if (orig->ealg) {
1167                 x->ealg = xfrm_algo_clone(orig->ealg);
1168                 if (!x->ealg)
1169                         goto error;
1170         }
1171         x->props.ealgo = orig->props.ealgo;
1172
1173         if (orig->calg) {
1174                 x->calg = xfrm_algo_clone(orig->calg);
1175                 if (!x->calg)
1176                         goto error;
1177         }
1178         x->props.calgo = orig->props.calgo;
1179
1180         if (orig->encap) {
1181                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1182                 if (!x->encap)
1183                         goto error;
1184         }
1185
1186         if (orig->coaddr) {
1187                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1188                                     GFP_KERNEL);
1189                 if (!x->coaddr)
1190                         goto error;
1191         }
1192
1193         if (orig->replay_esn) {
1194                 if (xfrm_replay_clone(x, orig))
1195                         goto error;
1196         }
1197
1198         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1199
1200         if (xfrm_init_state(x) < 0)
1201                 goto error;
1202
1203         x->props.flags = orig->props.flags;
1204         x->props.extra_flags = orig->props.extra_flags;
1205
1206         x->tfcpad = orig->tfcpad;
1207         x->replay_maxdiff = orig->replay_maxdiff;
1208         x->replay_maxage = orig->replay_maxage;
1209         x->curlft.add_time = orig->curlft.add_time;
1210         x->km.state = orig->km.state;
1211         x->km.seq = orig->km.seq;
1212
1213         return x;
1214
1215  error:
1216         xfrm_state_put(x);
1217 out:
1218         return NULL;
1219 }
1220
1221 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1222 {
1223         unsigned int h;
1224         struct xfrm_state *x = NULL;
1225
1226         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1227
1228         if (m->reqid) {
1229                 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1230                                   m->reqid, m->old_family);
1231                 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1232                         if (x->props.mode != m->mode ||
1233                             x->id.proto != m->proto)
1234                                 continue;
1235                         if (m->reqid && x->props.reqid != m->reqid)
1236                                 continue;
1237                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1238                                              m->old_family) ||
1239                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1240                                              m->old_family))
1241                                 continue;
1242                         xfrm_state_hold(x);
1243                         break;
1244                 }
1245         } else {
1246                 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1247                                   m->old_family);
1248                 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1249                         if (x->props.mode != m->mode ||
1250                             x->id.proto != m->proto)
1251                                 continue;
1252                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1253                                              m->old_family) ||
1254                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1255                                              m->old_family))
1256                                 continue;
1257                         xfrm_state_hold(x);
1258                         break;
1259                 }
1260         }
1261
1262         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1263
1264         return x;
1265 }
1266 EXPORT_SYMBOL(xfrm_migrate_state_find);
1267
1268 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1269                                       struct xfrm_migrate *m)
1270 {
1271         struct xfrm_state *xc;
1272
1273         xc = xfrm_state_clone(x);
1274         if (!xc)
1275                 return NULL;
1276
1277         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1278         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1279
1280         /* add state */
1281         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1282                 /* a care is needed when the destination address of the
1283                    state is to be updated as it is a part of triplet */
1284                 xfrm_state_insert(xc);
1285         } else {
1286                 if (xfrm_state_add(xc) < 0)
1287                         goto error;
1288         }
1289
1290         return xc;
1291 error:
1292         xfrm_state_put(xc);
1293         return NULL;
1294 }
1295 EXPORT_SYMBOL(xfrm_state_migrate);
1296 #endif
1297
1298 int xfrm_state_update(struct xfrm_state *x)
1299 {
1300         struct xfrm_state *x1, *to_put;
1301         int err;
1302         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1303         struct net *net = xs_net(x);
1304
1305         to_put = NULL;
1306
1307         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1308         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1309
1310         err = -ESRCH;
1311         if (!x1)
1312                 goto out;
1313
1314         if (xfrm_state_kern(x1)) {
1315                 to_put = x1;
1316                 err = -EEXIST;
1317                 goto out;
1318         }
1319
1320         if (x1->km.state == XFRM_STATE_ACQ) {
1321                 __xfrm_state_insert(x);
1322                 x = NULL;
1323         }
1324         err = 0;
1325
1326 out:
1327         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1328
1329         if (to_put)
1330                 xfrm_state_put(to_put);
1331
1332         if (err)
1333                 return err;
1334
1335         if (!x) {
1336                 xfrm_state_delete(x1);
1337                 xfrm_state_put(x1);
1338                 return 0;
1339         }
1340
1341         err = -EINVAL;
1342         spin_lock_bh(&x1->lock);
1343         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1344                 if (x->encap && x1->encap)
1345                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1346                 if (x->coaddr && x1->coaddr) {
1347                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1348                 }
1349                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1350                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1351                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1352                 x1->km.dying = 0;
1353
1354                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1355                 if (x1->curlft.use_time)
1356                         xfrm_state_check_expire(x1);
1357
1358                 err = 0;
1359                 x->km.state = XFRM_STATE_DEAD;
1360                 __xfrm_state_put(x);
1361         }
1362         spin_unlock_bh(&x1->lock);
1363
1364         xfrm_state_put(x1);
1365
1366         return err;
1367 }
1368 EXPORT_SYMBOL(xfrm_state_update);
1369
1370 int xfrm_state_check_expire(struct xfrm_state *x)
1371 {
1372         if (!x->curlft.use_time)
1373                 x->curlft.use_time = get_seconds();
1374
1375         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1376             x->curlft.packets >= x->lft.hard_packet_limit) {
1377                 x->km.state = XFRM_STATE_EXPIRED;
1378                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1379                 return -EINVAL;
1380         }
1381
1382         if (!x->km.dying &&
1383             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1384              x->curlft.packets >= x->lft.soft_packet_limit)) {
1385                 x->km.dying = 1;
1386                 km_state_expired(x, 0, 0);
1387         }
1388         return 0;
1389 }
1390 EXPORT_SYMBOL(xfrm_state_check_expire);
1391
1392 struct xfrm_state *
1393 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1394                   u8 proto, unsigned short family)
1395 {
1396         struct xfrm_state *x;
1397
1398         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1399         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1400         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1401         return x;
1402 }
1403 EXPORT_SYMBOL(xfrm_state_lookup);
1404
1405 struct xfrm_state *
1406 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1407                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1408                          u8 proto, unsigned short family)
1409 {
1410         struct xfrm_state *x;
1411
1412         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1413         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1414         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1415         return x;
1416 }
1417 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1418
1419 struct xfrm_state *
1420 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1421               u8 proto, const xfrm_address_t *daddr,
1422               const xfrm_address_t *saddr, int create, unsigned short family)
1423 {
1424         struct xfrm_state *x;
1425
1426         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1427         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1428         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1429
1430         return x;
1431 }
1432 EXPORT_SYMBOL(xfrm_find_acq);
1433
1434 #ifdef CONFIG_XFRM_SUB_POLICY
1435 int
1436 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1437                unsigned short family, struct net *net)
1438 {
1439         int err = 0;
1440         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1441         if (!afinfo)
1442                 return -EAFNOSUPPORT;
1443
1444         spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1445         if (afinfo->tmpl_sort)
1446                 err = afinfo->tmpl_sort(dst, src, n);
1447         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1448         xfrm_state_put_afinfo(afinfo);
1449         return err;
1450 }
1451 EXPORT_SYMBOL(xfrm_tmpl_sort);
1452
1453 int
1454 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1455                 unsigned short family)
1456 {
1457         int err = 0;
1458         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1459         struct net *net = xs_net(*src);
1460
1461         if (!afinfo)
1462                 return -EAFNOSUPPORT;
1463
1464         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1465         if (afinfo->state_sort)
1466                 err = afinfo->state_sort(dst, src, n);
1467         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1468         xfrm_state_put_afinfo(afinfo);
1469         return err;
1470 }
1471 EXPORT_SYMBOL(xfrm_state_sort);
1472 #endif
1473
1474 /* Silly enough, but I'm lazy to build resolution list */
1475
1476 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1477 {
1478         int i;
1479
1480         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1481                 struct xfrm_state *x;
1482
1483                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1484                         if (x->km.seq == seq &&
1485                             (mark & x->mark.m) == x->mark.v &&
1486                             x->km.state == XFRM_STATE_ACQ) {
1487                                 xfrm_state_hold(x);
1488                                 return x;
1489                         }
1490                 }
1491         }
1492         return NULL;
1493 }
1494
1495 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1496 {
1497         struct xfrm_state *x;
1498
1499         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1500         x = __xfrm_find_acq_byseq(net, mark, seq);
1501         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1502         return x;
1503 }
1504 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1505
1506 u32 xfrm_get_acqseq(void)
1507 {
1508         u32 res;
1509         static atomic_t acqseq;
1510
1511         do {
1512                 res = atomic_inc_return(&acqseq);
1513         } while (!res);
1514
1515         return res;
1516 }
1517 EXPORT_SYMBOL(xfrm_get_acqseq);
1518
1519 int verify_spi_info(u8 proto, u32 min, u32 max)
1520 {
1521         switch (proto) {
1522         case IPPROTO_AH:
1523         case IPPROTO_ESP:
1524                 break;
1525
1526         case IPPROTO_COMP:
1527                 /* IPCOMP spi is 16-bits. */
1528                 if (max >= 0x10000)
1529                         return -EINVAL;
1530                 break;
1531
1532         default:
1533                 return -EINVAL;
1534         }
1535
1536         if (min > max)
1537                 return -EINVAL;
1538
1539         return 0;
1540 }
1541 EXPORT_SYMBOL(verify_spi_info);
1542
1543 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1544 {
1545         struct net *net = xs_net(x);
1546         unsigned int h;
1547         struct xfrm_state *x0;
1548         int err = -ENOENT;
1549         __be32 minspi = htonl(low);
1550         __be32 maxspi = htonl(high);
1551         u32 mark = x->mark.v & x->mark.m;
1552
1553         spin_lock_bh(&x->lock);
1554         if (x->km.state == XFRM_STATE_DEAD)
1555                 goto unlock;
1556
1557         err = 0;
1558         if (x->id.spi)
1559                 goto unlock;
1560
1561         err = -ENOENT;
1562
1563         if (minspi == maxspi) {
1564                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1565                 if (x0) {
1566                         xfrm_state_put(x0);
1567                         goto unlock;
1568                 }
1569                 x->id.spi = minspi;
1570         } else {
1571                 u32 spi = 0;
1572                 for (h = 0; h < high-low+1; h++) {
1573                         spi = low + prandom_u32()%(high-low+1);
1574                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1575                         if (x0 == NULL) {
1576                                 x->id.spi = htonl(spi);
1577                                 break;
1578                         }
1579                         xfrm_state_put(x0);
1580                 }
1581         }
1582         if (x->id.spi) {
1583                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1584                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1585                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1586                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1587
1588                 err = 0;
1589         }
1590
1591 unlock:
1592         spin_unlock_bh(&x->lock);
1593
1594         return err;
1595 }
1596 EXPORT_SYMBOL(xfrm_alloc_spi);
1597
1598 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1599                                       struct xfrm_address_filter *filter)
1600 {
1601         if (filter) {
1602                 if ((filter->family == AF_INET ||
1603                      filter->family == AF_INET6) &&
1604                     x->props.family != filter->family)
1605                         return false;
1606
1607                 return addr_match(&x->props.saddr, &filter->saddr,
1608                                   filter->splen) &&
1609                        addr_match(&x->id.daddr, &filter->daddr,
1610                                   filter->dplen);
1611         }
1612         return true;
1613 }
1614
1615 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1616                     int (*func)(struct xfrm_state *, int, void*),
1617                     void *data)
1618 {
1619         struct xfrm_state *state;
1620         struct xfrm_state_walk *x;
1621         int err = 0;
1622
1623         if (walk->seq != 0 && list_empty(&walk->all))
1624                 return 0;
1625
1626         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1627         if (list_empty(&walk->all))
1628                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1629         else
1630                 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1631         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1632                 if (x->state == XFRM_STATE_DEAD)
1633                         continue;
1634                 state = container_of(x, struct xfrm_state, km);
1635                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1636                         continue;
1637                 if (!__xfrm_state_filter_match(state, walk->filter))
1638                         continue;
1639                 err = func(state, walk->seq, data);
1640                 if (err) {
1641                         list_move_tail(&walk->all, &x->all);
1642                         goto out;
1643                 }
1644                 walk->seq++;
1645         }
1646         if (walk->seq == 0) {
1647                 err = -ENOENT;
1648                 goto out;
1649         }
1650         list_del_init(&walk->all);
1651 out:
1652         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1653         return err;
1654 }
1655 EXPORT_SYMBOL(xfrm_state_walk);
1656
1657 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1658                           struct xfrm_address_filter *filter)
1659 {
1660         INIT_LIST_HEAD(&walk->all);
1661         walk->proto = proto;
1662         walk->state = XFRM_STATE_DEAD;
1663         walk->seq = 0;
1664         walk->filter = filter;
1665 }
1666 EXPORT_SYMBOL(xfrm_state_walk_init);
1667
1668 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1669 {
1670         kfree(walk->filter);
1671
1672         if (list_empty(&walk->all))
1673                 return;
1674
1675         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1676         list_del(&walk->all);
1677         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1678 }
1679 EXPORT_SYMBOL(xfrm_state_walk_done);
1680
1681 static void xfrm_replay_timer_handler(unsigned long data)
1682 {
1683         struct xfrm_state *x = (struct xfrm_state *)data;
1684
1685         spin_lock(&x->lock);
1686
1687         if (x->km.state == XFRM_STATE_VALID) {
1688                 if (xfrm_aevent_is_on(xs_net(x)))
1689                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1690                 else
1691                         x->xflags |= XFRM_TIME_DEFER;
1692         }
1693
1694         spin_unlock(&x->lock);
1695 }
1696
1697 static LIST_HEAD(xfrm_km_list);
1698
1699 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1700 {
1701         struct xfrm_mgr *km;
1702
1703         rcu_read_lock();
1704         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1705                 if (km->notify_policy)
1706                         km->notify_policy(xp, dir, c);
1707         rcu_read_unlock();
1708 }
1709
1710 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1711 {
1712         struct xfrm_mgr *km;
1713         rcu_read_lock();
1714         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1715                 if (km->notify)
1716                         km->notify(x, c);
1717         rcu_read_unlock();
1718 }
1719
1720 EXPORT_SYMBOL(km_policy_notify);
1721 EXPORT_SYMBOL(km_state_notify);
1722
1723 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1724 {
1725         struct km_event c;
1726
1727         c.data.hard = hard;
1728         c.portid = portid;
1729         c.event = XFRM_MSG_EXPIRE;
1730         km_state_notify(x, &c);
1731 }
1732
1733 EXPORT_SYMBOL(km_state_expired);
1734 /*
1735  * We send to all registered managers regardless of failure
1736  * We are happy with one success
1737 */
1738 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1739 {
1740         int err = -EINVAL, acqret;
1741         struct xfrm_mgr *km;
1742
1743         rcu_read_lock();
1744         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1745                 acqret = km->acquire(x, t, pol);
1746                 if (!acqret)
1747                         err = acqret;
1748         }
1749         rcu_read_unlock();
1750         return err;
1751 }
1752 EXPORT_SYMBOL(km_query);
1753
1754 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1755 {
1756         int err = -EINVAL;
1757         struct xfrm_mgr *km;
1758
1759         rcu_read_lock();
1760         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1761                 if (km->new_mapping)
1762                         err = km->new_mapping(x, ipaddr, sport);
1763                 if (!err)
1764                         break;
1765         }
1766         rcu_read_unlock();
1767         return err;
1768 }
1769 EXPORT_SYMBOL(km_new_mapping);
1770
1771 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1772 {
1773         struct km_event c;
1774
1775         c.data.hard = hard;
1776         c.portid = portid;
1777         c.event = XFRM_MSG_POLEXPIRE;
1778         km_policy_notify(pol, dir, &c);
1779 }
1780 EXPORT_SYMBOL(km_policy_expired);
1781
1782 #ifdef CONFIG_XFRM_MIGRATE
1783 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1784                const struct xfrm_migrate *m, int num_migrate,
1785                const struct xfrm_kmaddress *k)
1786 {
1787         int err = -EINVAL;
1788         int ret;
1789         struct xfrm_mgr *km;
1790
1791         rcu_read_lock();
1792         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1793                 if (km->migrate) {
1794                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1795                         if (!ret)
1796                                 err = ret;
1797                 }
1798         }
1799         rcu_read_unlock();
1800         return err;
1801 }
1802 EXPORT_SYMBOL(km_migrate);
1803 #endif
1804
1805 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1806 {
1807         int err = -EINVAL;
1808         int ret;
1809         struct xfrm_mgr *km;
1810
1811         rcu_read_lock();
1812         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1813                 if (km->report) {
1814                         ret = km->report(net, proto, sel, addr);
1815                         if (!ret)
1816                                 err = ret;
1817                 }
1818         }
1819         rcu_read_unlock();
1820         return err;
1821 }
1822 EXPORT_SYMBOL(km_report);
1823
1824 bool km_is_alive(const struct km_event *c)
1825 {
1826         struct xfrm_mgr *km;
1827         bool is_alive = false;
1828
1829         rcu_read_lock();
1830         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1831                 if (km->is_alive && km->is_alive(c)) {
1832                         is_alive = true;
1833                         break;
1834                 }
1835         }
1836         rcu_read_unlock();
1837
1838         return is_alive;
1839 }
1840 EXPORT_SYMBOL(km_is_alive);
1841
1842 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1843 {
1844         int err;
1845         u8 *data;
1846         struct xfrm_mgr *km;
1847         struct xfrm_policy *pol = NULL;
1848
1849         if (optlen <= 0 || optlen > PAGE_SIZE)
1850                 return -EMSGSIZE;
1851
1852         data = kmalloc(optlen, GFP_KERNEL);
1853         if (!data)
1854                 return -ENOMEM;
1855
1856         err = -EFAULT;
1857         if (copy_from_user(data, optval, optlen))
1858                 goto out;
1859
1860         err = -EINVAL;
1861         rcu_read_lock();
1862         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1863                 pol = km->compile_policy(sk, optname, data,
1864                                          optlen, &err);
1865                 if (err >= 0)
1866                         break;
1867         }
1868         rcu_read_unlock();
1869
1870         if (err >= 0) {
1871                 xfrm_sk_policy_insert(sk, err, pol);
1872                 xfrm_pol_put(pol);
1873                 err = 0;
1874         }
1875
1876 out:
1877         kfree(data);
1878         return err;
1879 }
1880 EXPORT_SYMBOL(xfrm_user_policy);
1881
1882 static DEFINE_SPINLOCK(xfrm_km_lock);
1883
1884 int xfrm_register_km(struct xfrm_mgr *km)
1885 {
1886         spin_lock_bh(&xfrm_km_lock);
1887         list_add_tail_rcu(&km->list, &xfrm_km_list);
1888         spin_unlock_bh(&xfrm_km_lock);
1889         return 0;
1890 }
1891 EXPORT_SYMBOL(xfrm_register_km);
1892
1893 int xfrm_unregister_km(struct xfrm_mgr *km)
1894 {
1895         spin_lock_bh(&xfrm_km_lock);
1896         list_del_rcu(&km->list);
1897         spin_unlock_bh(&xfrm_km_lock);
1898         synchronize_rcu();
1899         return 0;
1900 }
1901 EXPORT_SYMBOL(xfrm_unregister_km);
1902
1903 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1904 {
1905         int err = 0;
1906         if (unlikely(afinfo == NULL))
1907                 return -EINVAL;
1908         if (unlikely(afinfo->family >= NPROTO))
1909                 return -EAFNOSUPPORT;
1910         spin_lock_bh(&xfrm_state_afinfo_lock);
1911         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1912                 err = -EEXIST;
1913         else
1914                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1915         spin_unlock_bh(&xfrm_state_afinfo_lock);
1916         return err;
1917 }
1918 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1919
1920 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1921 {
1922         int err = 0;
1923         if (unlikely(afinfo == NULL))
1924                 return -EINVAL;
1925         if (unlikely(afinfo->family >= NPROTO))
1926                 return -EAFNOSUPPORT;
1927         spin_lock_bh(&xfrm_state_afinfo_lock);
1928         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1929                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1930                         err = -EINVAL;
1931                 else
1932                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1933         }
1934         spin_unlock_bh(&xfrm_state_afinfo_lock);
1935         synchronize_rcu();
1936         return err;
1937 }
1938 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1939
1940 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1941 {
1942         struct xfrm_state_afinfo *afinfo;
1943         if (unlikely(family >= NPROTO))
1944                 return NULL;
1945         rcu_read_lock();
1946         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1947         if (unlikely(!afinfo))
1948                 rcu_read_unlock();
1949         return afinfo;
1950 }
1951
1952 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1953 {
1954         rcu_read_unlock();
1955 }
1956
1957 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1958 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1959 {
1960         if (x->tunnel) {
1961                 struct xfrm_state *t = x->tunnel;
1962
1963                 if (atomic_read(&t->tunnel_users) == 2)
1964                         xfrm_state_delete(t);
1965                 atomic_dec(&t->tunnel_users);
1966                 xfrm_state_put(t);
1967                 x->tunnel = NULL;
1968         }
1969 }
1970 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1971
1972 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1973 {
1974         int res;
1975
1976         spin_lock_bh(&x->lock);
1977         if (x->km.state == XFRM_STATE_VALID &&
1978             x->type && x->type->get_mtu)
1979                 res = x->type->get_mtu(x, mtu);
1980         else
1981                 res = mtu - x->props.header_len;
1982         spin_unlock_bh(&x->lock);
1983         return res;
1984 }
1985
1986 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
1987 {
1988         struct xfrm_state_afinfo *afinfo;
1989         struct xfrm_mode *inner_mode;
1990         int family = x->props.family;
1991         int err;
1992
1993         err = -EAFNOSUPPORT;
1994         afinfo = xfrm_state_get_afinfo(family);
1995         if (!afinfo)
1996                 goto error;
1997
1998         err = 0;
1999         if (afinfo->init_flags)
2000                 err = afinfo->init_flags(x);
2001
2002         xfrm_state_put_afinfo(afinfo);
2003
2004         if (err)
2005                 goto error;
2006
2007         err = -EPROTONOSUPPORT;
2008
2009         if (x->sel.family != AF_UNSPEC) {
2010                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2011                 if (inner_mode == NULL)
2012                         goto error;
2013
2014                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2015                     family != x->sel.family) {
2016                         xfrm_put_mode(inner_mode);
2017                         goto error;
2018                 }
2019
2020                 x->inner_mode = inner_mode;
2021         } else {
2022                 struct xfrm_mode *inner_mode_iaf;
2023                 int iafamily = AF_INET;
2024
2025                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2026                 if (inner_mode == NULL)
2027                         goto error;
2028
2029                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2030                         xfrm_put_mode(inner_mode);
2031                         goto error;
2032                 }
2033                 x->inner_mode = inner_mode;
2034
2035                 if (x->props.family == AF_INET)
2036                         iafamily = AF_INET6;
2037
2038                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2039                 if (inner_mode_iaf) {
2040                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2041                                 x->inner_mode_iaf = inner_mode_iaf;
2042                         else
2043                                 xfrm_put_mode(inner_mode_iaf);
2044                 }
2045         }
2046
2047         x->type = xfrm_get_type(x->id.proto, family);
2048         if (x->type == NULL)
2049                 goto error;
2050
2051         err = x->type->init_state(x);
2052         if (err)
2053                 goto error;
2054
2055         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2056         if (x->outer_mode == NULL) {
2057                 err = -EPROTONOSUPPORT;
2058                 goto error;
2059         }
2060
2061         if (init_replay) {
2062                 err = xfrm_init_replay(x);
2063                 if (err)
2064                         goto error;
2065         }
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 xfrm_init_state(struct xfrm_state *x)
2076 {
2077         return __xfrm_init_state(x, true);
2078 }
2079
2080 EXPORT_SYMBOL(xfrm_init_state);
2081
2082 int __net_init xfrm_state_init(struct net *net)
2083 {
2084         unsigned int sz;
2085
2086         INIT_LIST_HEAD(&net->xfrm.state_all);
2087
2088         sz = sizeof(struct hlist_head) * 8;
2089
2090         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2091         if (!net->xfrm.state_bydst)
2092                 goto out_bydst;
2093         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2094         if (!net->xfrm.state_bysrc)
2095                 goto out_bysrc;
2096         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2097         if (!net->xfrm.state_byspi)
2098                 goto out_byspi;
2099         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2100
2101         net->xfrm.state_num = 0;
2102         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2103         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2104         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2105         spin_lock_init(&net->xfrm.xfrm_state_lock);
2106         return 0;
2107
2108 out_byspi:
2109         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2110 out_bysrc:
2111         xfrm_hash_free(net->xfrm.state_bydst, sz);
2112 out_bydst:
2113         return -ENOMEM;
2114 }
2115
2116 void xfrm_state_fini(struct net *net)
2117 {
2118         unsigned int sz;
2119
2120         flush_work(&net->xfrm.state_hash_work);
2121         xfrm_state_flush(net, IPSEC_PROTO_ANY, false);
2122         flush_work(&net->xfrm.state_gc_work);
2123
2124         WARN_ON(!list_empty(&net->xfrm.state_all));
2125
2126         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2127         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2128         xfrm_hash_free(net->xfrm.state_byspi, sz);
2129         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2130         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2131         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2132         xfrm_hash_free(net->xfrm.state_bydst, sz);
2133 }
2134
2135 #ifdef CONFIG_AUDITSYSCALL
2136 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2137                                      struct audit_buffer *audit_buf)
2138 {
2139         struct xfrm_sec_ctx *ctx = x->security;
2140         u32 spi = ntohl(x->id.spi);
2141
2142         if (ctx)
2143                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2144                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2145
2146         switch (x->props.family) {
2147         case AF_INET:
2148                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2149                                  &x->props.saddr.a4, &x->id.daddr.a4);
2150                 break;
2151         case AF_INET6:
2152                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2153                                  x->props.saddr.a6, x->id.daddr.a6);
2154                 break;
2155         }
2156
2157         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2158 }
2159
2160 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2161                                       struct audit_buffer *audit_buf)
2162 {
2163         const struct iphdr *iph4;
2164         const struct ipv6hdr *iph6;
2165
2166         switch (family) {
2167         case AF_INET:
2168                 iph4 = ip_hdr(skb);
2169                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2170                                  &iph4->saddr, &iph4->daddr);
2171                 break;
2172         case AF_INET6:
2173                 iph6 = ipv6_hdr(skb);
2174                 audit_log_format(audit_buf,
2175                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2176                                  &iph6->saddr, &iph6->daddr,
2177                                  iph6->flow_lbl[0] & 0x0f,
2178                                  iph6->flow_lbl[1],
2179                                  iph6->flow_lbl[2]);
2180                 break;
2181         }
2182 }
2183
2184 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2185 {
2186         struct audit_buffer *audit_buf;
2187
2188         audit_buf = xfrm_audit_start("SAD-add");
2189         if (audit_buf == NULL)
2190                 return;
2191         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2192         xfrm_audit_helper_sainfo(x, audit_buf);
2193         audit_log_format(audit_buf, " res=%u", result);
2194         audit_log_end(audit_buf);
2195 }
2196 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2197
2198 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2199 {
2200         struct audit_buffer *audit_buf;
2201
2202         audit_buf = xfrm_audit_start("SAD-delete");
2203         if (audit_buf == NULL)
2204                 return;
2205         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2206         xfrm_audit_helper_sainfo(x, audit_buf);
2207         audit_log_format(audit_buf, " res=%u", result);
2208         audit_log_end(audit_buf);
2209 }
2210 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2211
2212 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2213                                       struct sk_buff *skb)
2214 {
2215         struct audit_buffer *audit_buf;
2216         u32 spi;
2217
2218         audit_buf = xfrm_audit_start("SA-replay-overflow");
2219         if (audit_buf == NULL)
2220                 return;
2221         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2222         /* don't record the sequence number because it's inherent in this kind
2223          * of audit message */
2224         spi = ntohl(x->id.spi);
2225         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2226         audit_log_end(audit_buf);
2227 }
2228 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2229
2230 void xfrm_audit_state_replay(struct xfrm_state *x,
2231                              struct sk_buff *skb, __be32 net_seq)
2232 {
2233         struct audit_buffer *audit_buf;
2234         u32 spi;
2235
2236         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2237         if (audit_buf == NULL)
2238                 return;
2239         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2240         spi = ntohl(x->id.spi);
2241         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2242                          spi, spi, ntohl(net_seq));
2243         audit_log_end(audit_buf);
2244 }
2245 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2246
2247 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2248 {
2249         struct audit_buffer *audit_buf;
2250
2251         audit_buf = xfrm_audit_start("SA-notfound");
2252         if (audit_buf == NULL)
2253                 return;
2254         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2255         audit_log_end(audit_buf);
2256 }
2257 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2258
2259 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2260                                __be32 net_spi, __be32 net_seq)
2261 {
2262         struct audit_buffer *audit_buf;
2263         u32 spi;
2264
2265         audit_buf = xfrm_audit_start("SA-notfound");
2266         if (audit_buf == NULL)
2267                 return;
2268         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2269         spi = ntohl(net_spi);
2270         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2271                          spi, spi, ntohl(net_seq));
2272         audit_log_end(audit_buf);
2273 }
2274 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2275
2276 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2277                               struct sk_buff *skb, u8 proto)
2278 {
2279         struct audit_buffer *audit_buf;
2280         __be32 net_spi;
2281         __be32 net_seq;
2282
2283         audit_buf = xfrm_audit_start("SA-icv-failure");
2284         if (audit_buf == NULL)
2285                 return;
2286         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2287         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2288                 u32 spi = ntohl(net_spi);
2289                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2290                                  spi, spi, ntohl(net_seq));
2291         }
2292         audit_log_end(audit_buf);
2293 }
2294 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2295 #endif /* CONFIG_AUDITSYSCALL */