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 + 1;
 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                      (x->sel.family != family ||
 746                       !xfrm_selector_match(&x->sel, fl, family))) ||
 747                     !security_xfrm_state_pol_flow_match(x, pol, fl))
 748                         return;
 749
 750                 if (!*best ||
 751                     (*best)->km.dying > x->km.dying ||
 752                     ((*best)->km.dying == x->km.dying &&
 753                      (*best)->curlft.add_time < x->curlft.add_time))
 754                         *best = x;
 755         } else if (x->km.state == XFRM_STATE_ACQ) {
 756                 *acq_in_progress = 1;
 757         } else if (x->km.state == XFRM_STATE_ERROR ||
 758                    x->km.state == XFRM_STATE_EXPIRED) {
 759                 if ((!x->sel.family ||
 760                      (x->sel.family == family &&
 761                       xfrm_selector_match(&x->sel, fl, family))) &&
 762                     security_xfrm_state_pol_flow_match(x, pol, fl))
 763                         *error = -ESRCH;
 764         }
 765 }
 766
 767 struct xfrm_state *
 768 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 769                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
 770                 struct xfrm_policy *pol, int *err,
 771                 unsigned short family)
 772 {
 773         static xfrm_address_t saddr_wildcard = { };
 774         struct net *net = xp_net(pol);
 775         unsigned int h, h_wildcard;
 776         struct xfrm_state *x, *x0, *to_put;
 777         int acquire_in_progress = 0;
 778         int error = 0;
 779         struct xfrm_state *best = NULL;
 780         u32 mark = pol->mark.v & pol->mark.m;
 781         unsigned short encap_family = tmpl->encap_family;
 782         struct km_event c;
 783
 784         to_put = NULL;
 785
 786         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 787         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
 788         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 789                 if (x->props.family == encap_family &&
 790                     x->props.reqid == tmpl->reqid &&
 791                     (mark & x->mark.m) == x->mark.v &&
 792                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 793                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
 794                     tmpl->mode == x->props.mode &&
 795                     tmpl->id.proto == x->id.proto &&
 796                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 797                         xfrm_state_look_at(pol, x, fl, family,
 798                                            &best, &acquire_in_progress, &error);
 799         }
 800         if (best || acquire_in_progress)
 801                 goto found;
 802
 803         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
 804         hlist_for_each_entry(x, net->xfrm.state_bydst+h_wildcard, bydst) {
 805                 if (x->props.family == encap_family &&
 806                     x->props.reqid == tmpl->reqid &&
 807                     (mark & x->mark.m) == x->mark.v &&
 808                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 809                     xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
 810                     tmpl->mode == x->props.mode &&
 811                     tmpl->id.proto == x->id.proto &&
 812                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
 813                         xfrm_state_look_at(pol, x, fl, family,
 814                                            &best, &acquire_in_progress, &error);
 815         }
 816
 817 found:
 818         x = best;
 819         if (!x && !error && !acquire_in_progress) {
 820                 if (tmpl->id.spi &&
 821                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
 822                                               tmpl->id.proto, encap_family)) != NULL) {
 823                         to_put = x0;
 824                         error = -EEXIST;
 825                         goto out;
 826                 }
 827
 828                 c.net = net;
 829                 /* If the KMs have no listeners (yet...), avoid allocating an SA
 830                  * for each and every packet - garbage collection might not
 831                  * handle the flood.
 832                  */
 833                 if (!km_is_alive(&c)) {
 834                         error = -ESRCH;
 835                         goto out;
 836                 }
 837
 838                 x = xfrm_state_alloc(net);
 839                 if (x == NULL) {
 840                         error = -ENOMEM;
 841                         goto out;
 842                 }
 843                 /* Initialize temporary state matching only
 844                  * to current session. */
 845                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
 846                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
 847
 848                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
 849                 if (error) {
 850                         x->km.state = XFRM_STATE_DEAD;
 851                         to_put = x;
 852                         x = NULL;
 853                         goto out;
 854                 }
 855
 856                 if (km_query(x, tmpl, pol) == 0) {
 857                         x->km.state = XFRM_STATE_ACQ;
 858                         list_add(&x->km.all, &net->xfrm.state_all);
 859                         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 860                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
 861                         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 862                         if (x->id.spi) {
 863                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
 864                                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 865                         }
 866                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
 867                         tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
 868                         net->xfrm.state_num++;
 869                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 870                 } else {
 871                         x->km.state = XFRM_STATE_DEAD;
 872                         to_put = x;
 873                         x = NULL;
 874                         error = -ESRCH;
 875                 }
 876         }
 877 out:
 878         if (x)
 879                 xfrm_state_hold(x);
 880         else
 881                 *err = acquire_in_progress ? -EAGAIN : error;
 882         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 883         if (to_put)
 884                 xfrm_state_put(to_put);
 885         return x;
 886 }
 887
 888 struct xfrm_state *
 889 xfrm_stateonly_find(struct net *net, u32 mark,
 890                     xfrm_address_t *daddr, xfrm_address_t *saddr,
 891                     unsigned short family, u8 mode, u8 proto, u32 reqid)
 892 {
 893         unsigned int h;
 894         struct xfrm_state *rx = NULL, *x = NULL;
 895
 896         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 897         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
 898         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 899                 if (x->props.family == family &&
 900                     x->props.reqid == reqid &&
 901                     (mark & x->mark.m) == x->mark.v &&
 902                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
 903                     xfrm_state_addr_check(x, daddr, saddr, family) &&
 904                     mode == x->props.mode &&
 905                     proto == x->id.proto &&
 906                     x->km.state == XFRM_STATE_VALID) {
 907                         rx = x;
 908                         break;
 909                 }
 910         }
 911
 912         if (rx)
 913                 xfrm_state_hold(rx);
 914         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 915
 916
 917         return rx;
 918 }
 919 EXPORT_SYMBOL(xfrm_stateonly_find);
 920
 921 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
 922                                               unsigned short family)
 923 {
 924         struct xfrm_state *x;
 925         struct xfrm_state_walk *w;
 926
 927         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 928         list_for_each_entry(w, &net->xfrm.state_all, all) {
 929                 x = container_of(w, struct xfrm_state, km);
 930                 if (x->props.family != family ||
 931                         x->id.spi != spi)
 932                         continue;
 933
 934                 xfrm_state_hold(x);
 935                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 936                 return x;
 937         }
 938         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 939         return NULL;
 940 }
 941 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
 942
 943 static void __xfrm_state_insert(struct xfrm_state *x)
 944 {
 945         struct net *net = xs_net(x);
 946         unsigned int h;
 947
 948         list_add(&x->km.all, &net->xfrm.state_all);
 949
 950         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
 951                           x->props.reqid, x->props.family);
 952         hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
 953
 954         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
 955         hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
 956
 957         if (x->id.spi) {
 958                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
 959                                   x->props.family);
 960
 961                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
 962         }
 963
 964         tasklet_hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
 965         if (x->replay_maxage)
 966                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
 967
 968         net->xfrm.state_num++;
 969
 970         xfrm_hash_grow_check(net, x->bydst.next != NULL);
 971 }
 972
 973 /* net->xfrm.xfrm_state_lock is held */
 974 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
 975 {
 976         struct net *net = xs_net(xnew);
 977         unsigned short family = xnew->props.family;
 978         u32 reqid = xnew->props.reqid;
 979         struct xfrm_state *x;
 980         unsigned int h;
 981         u32 mark = xnew->mark.v & xnew->mark.m;
 982
 983         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
 984         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
 985                 if (x->props.family     == family &&
 986                     x->props.reqid      == reqid &&
 987                     (mark & x->mark.m) == x->mark.v &&
 988                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
 989                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
 990                         x->genid++;
 991         }
 992 }
 993
 994 void xfrm_state_insert(struct xfrm_state *x)
 995 {
 996         struct net *net = xs_net(x);
 997
 998         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 999         __xfrm_state_bump_genids(x);
1000         __xfrm_state_insert(x);
1001         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1002 }
1003 EXPORT_SYMBOL(xfrm_state_insert);
1004
1005 /* net->xfrm.xfrm_state_lock is held */
1006 static struct xfrm_state *__find_acq_core(struct net *net,
1007                                           const struct xfrm_mark *m,
1008                                           unsigned short family, u8 mode,
1009                                           u32 reqid, u8 proto,
1010                                           const xfrm_address_t *daddr,
1011                                           const xfrm_address_t *saddr,
1012                                           int create)
1013 {
1014         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1015         struct xfrm_state *x;
1016         u32 mark = m->v & m->m;
1017
1018         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1019                 if (x->props.reqid  != reqid ||
1020                     x->props.mode   != mode ||
1021                     x->props.family != family ||
1022                     x->km.state     != XFRM_STATE_ACQ ||
1023                     x->id.spi       != 0 ||
1024                     x->id.proto     != proto ||
1025                     (mark & x->mark.m) != x->mark.v ||
1026                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1027                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1028                         continue;
1029
1030                 xfrm_state_hold(x);
1031                 return x;
1032         }
1033
1034         if (!create)
1035                 return NULL;
1036
1037         x = xfrm_state_alloc(net);
1038         if (likely(x)) {
1039                 switch (family) {
1040                 case AF_INET:
1041                         x->sel.daddr.a4 = daddr->a4;
1042                         x->sel.saddr.a4 = saddr->a4;
1043                         x->sel.prefixlen_d = 32;
1044                         x->sel.prefixlen_s = 32;
1045                         x->props.saddr.a4 = saddr->a4;
1046                         x->id.daddr.a4 = daddr->a4;
1047                         break;
1048
1049                 case AF_INET6:
1050                         x->sel.daddr.in6 = daddr->in6;
1051                         x->sel.saddr.in6 = saddr->in6;
1052                         x->sel.prefixlen_d = 128;
1053                         x->sel.prefixlen_s = 128;
1054                         x->props.saddr.in6 = saddr->in6;
1055                         x->id.daddr.in6 = daddr->in6;
1056                         break;
1057                 }
1058
1059                 x->km.state = XFRM_STATE_ACQ;
1060                 x->id.proto = proto;
1061                 x->props.family = family;
1062                 x->props.mode = mode;
1063                 x->props.reqid = reqid;
1064                 x->mark.v = m->v;
1065                 x->mark.m = m->m;
1066                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1067                 xfrm_state_hold(x);
1068                 tasklet_hrtimer_start(&x->mtimer, ktime_set(net->xfrm.sysctl_acq_expires, 0), HRTIMER_MODE_REL);
1069                 list_add(&x->km.all, &net->xfrm.state_all);
1070                 hlist_add_head(&x->bydst, net->xfrm.state_bydst+h);
1071                 h = xfrm_src_hash(net, daddr, saddr, family);
1072                 hlist_add_head(&x->bysrc, net->xfrm.state_bysrc+h);
1073
1074                 net->xfrm.state_num++;
1075
1076                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1077         }
1078
1079         return x;
1080 }
1081
1082 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1083
1084 int xfrm_state_add(struct xfrm_state *x)
1085 {
1086         struct net *net = xs_net(x);
1087         struct xfrm_state *x1, *to_put;
1088         int family;
1089         int err;
1090         u32 mark = x->mark.v & x->mark.m;
1091         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1092
1093         family = x->props.family;
1094
1095         to_put = NULL;
1096
1097         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1098
1099         x1 = __xfrm_state_locate(x, use_spi, family);
1100         if (x1) {
1101                 to_put = x1;
1102                 x1 = NULL;
1103                 err = -EEXIST;
1104                 goto out;
1105         }
1106
1107         if (use_spi && x->km.seq) {
1108                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1109                 if (x1 && ((x1->id.proto != x->id.proto) ||
1110                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1111                         to_put = x1;
1112                         x1 = NULL;
1113                 }
1114         }
1115
1116         if (use_spi && !x1)
1117                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1118                                      x->props.reqid, x->id.proto,
1119                                      &x->id.daddr, &x->props.saddr, 0);
1120
1121         __xfrm_state_bump_genids(x);
1122         __xfrm_state_insert(x);
1123         err = 0;
1124
1125 out:
1126         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1127
1128         if (x1) {
1129                 xfrm_state_delete(x1);
1130                 xfrm_state_put(x1);
1131         }
1132
1133         if (to_put)
1134                 xfrm_state_put(to_put);
1135
1136         return err;
1137 }
1138 EXPORT_SYMBOL(xfrm_state_add);
1139
1140 #ifdef CONFIG_XFRM_MIGRATE
1141 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig)
1142 {
1143         struct net *net = xs_net(orig);
1144         struct xfrm_state *x = xfrm_state_alloc(net);
1145         if (!x)
1146                 goto out;
1147
1148         memcpy(&x->id, &orig->id, sizeof(x->id));
1149         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1150         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1151         x->props.mode = orig->props.mode;
1152         x->props.replay_window = orig->props.replay_window;
1153         x->props.reqid = orig->props.reqid;
1154         x->props.family = orig->props.family;
1155         x->props.saddr = orig->props.saddr;
1156
1157         if (orig->aalg) {
1158                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1159                 if (!x->aalg)
1160                         goto error;
1161         }
1162         x->props.aalgo = orig->props.aalgo;
1163
1164         if (orig->aead) {
1165                 x->aead = xfrm_algo_aead_clone(orig->aead);
1166                 x->geniv = orig->geniv;
1167                 if (!x->aead)
1168                         goto error;
1169         }
1170         if (orig->ealg) {
1171                 x->ealg = xfrm_algo_clone(orig->ealg);
1172                 if (!x->ealg)
1173                         goto error;
1174         }
1175         x->props.ealgo = orig->props.ealgo;
1176
1177         if (orig->calg) {
1178                 x->calg = xfrm_algo_clone(orig->calg);
1179                 if (!x->calg)
1180                         goto error;
1181         }
1182         x->props.calgo = orig->props.calgo;
1183
1184         if (orig->encap) {
1185                 x->encap = kmemdup(orig->encap, sizeof(*x->encap), GFP_KERNEL);
1186                 if (!x->encap)
1187                         goto error;
1188         }
1189
1190         if (orig->coaddr) {
1191                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1192                                     GFP_KERNEL);
1193                 if (!x->coaddr)
1194                         goto error;
1195         }
1196
1197         if (orig->replay_esn) {
1198                 if (xfrm_replay_clone(x, orig))
1199                         goto error;
1200         }
1201
1202         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1203
1204         if (xfrm_init_state(x) < 0)
1205                 goto error;
1206
1207         x->props.flags = orig->props.flags;
1208         x->props.extra_flags = orig->props.extra_flags;
1209
1210         x->tfcpad = orig->tfcpad;
1211         x->replay_maxdiff = orig->replay_maxdiff;
1212         x->replay_maxage = orig->replay_maxage;
1213         memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1214         x->km.state = orig->km.state;
1215         x->km.seq = orig->km.seq;
1216         x->replay = orig->replay;
1217         x->preplay = orig->preplay;
1218
1219         return x;
1220
1221  error:
1222         xfrm_state_put(x);
1223 out:
1224         return NULL;
1225 }
1226
1227 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net)
1228 {
1229         unsigned int h;
1230         struct xfrm_state *x = NULL;
1231
1232         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1233
1234         if (m->reqid) {
1235                 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1236                                   m->reqid, m->old_family);
1237                 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1238                         if (x->props.mode != m->mode ||
1239                             x->id.proto != m->proto)
1240                                 continue;
1241                         if (m->reqid && x->props.reqid != m->reqid)
1242                                 continue;
1243                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1244                                              m->old_family) ||
1245                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1246                                              m->old_family))
1247                                 continue;
1248                         xfrm_state_hold(x);
1249                         break;
1250                 }
1251         } else {
1252                 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1253                                   m->old_family);
1254                 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1255                         if (x->props.mode != m->mode ||
1256                             x->id.proto != m->proto)
1257                                 continue;
1258                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1259                                              m->old_family) ||
1260                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1261                                              m->old_family))
1262                                 continue;
1263                         xfrm_state_hold(x);
1264                         break;
1265                 }
1266         }
1267
1268         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1269
1270         return x;
1271 }
1272 EXPORT_SYMBOL(xfrm_migrate_state_find);
1273
1274 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1275                                       struct xfrm_migrate *m)
1276 {
1277         struct xfrm_state *xc;
1278
1279         xc = xfrm_state_clone(x);
1280         if (!xc)
1281                 return NULL;
1282
1283         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1284         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1285
1286         /* add state */
1287         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1288                 /* a care is needed when the destination address of the
1289                    state is to be updated as it is a part of triplet */
1290                 xfrm_state_insert(xc);
1291         } else {
1292                 if (xfrm_state_add(xc) < 0)
1293                         goto error;
1294         }
1295
1296         return xc;
1297 error:
1298         xfrm_state_put(xc);
1299         return NULL;
1300 }
1301 EXPORT_SYMBOL(xfrm_state_migrate);
1302 #endif
1303
1304 int xfrm_state_update(struct xfrm_state *x)
1305 {
1306         struct xfrm_state *x1, *to_put;
1307         int err;
1308         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1309         struct net *net = xs_net(x);
1310
1311         to_put = NULL;
1312
1313         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1314         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1315
1316         err = -ESRCH;
1317         if (!x1)
1318                 goto out;
1319
1320         if (xfrm_state_kern(x1)) {
1321                 to_put = x1;
1322                 err = -EEXIST;
1323                 goto out;
1324         }
1325
1326         if (x1->km.state == XFRM_STATE_ACQ) {
1327                 __xfrm_state_insert(x);
1328                 x = NULL;
1329         }
1330         err = 0;
1331
1332 out:
1333         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1334
1335         if (to_put)
1336                 xfrm_state_put(to_put);
1337
1338         if (err)
1339                 return err;
1340
1341         if (!x) {
1342                 xfrm_state_delete(x1);
1343                 xfrm_state_put(x1);
1344                 return 0;
1345         }
1346
1347         err = -EINVAL;
1348         spin_lock_bh(&x1->lock);
1349         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1350                 if (x->encap && x1->encap)
1351                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1352                 if (x->coaddr && x1->coaddr) {
1353                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1354                 }
1355                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1356                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1357                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1358                 x1->km.dying = 0;
1359
1360                 tasklet_hrtimer_start(&x1->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL);
1361                 if (x1->curlft.use_time)
1362                         xfrm_state_check_expire(x1);
1363
1364                 err = 0;
1365                 x->km.state = XFRM_STATE_DEAD;
1366                 __xfrm_state_put(x);
1367         }
1368         spin_unlock_bh(&x1->lock);
1369
1370         xfrm_state_put(x1);
1371
1372         return err;
1373 }
1374 EXPORT_SYMBOL(xfrm_state_update);
1375
1376 int xfrm_state_check_expire(struct xfrm_state *x)
1377 {
1378         if (!x->curlft.use_time)
1379                 x->curlft.use_time = get_seconds();
1380
1381         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1382             x->curlft.packets >= x->lft.hard_packet_limit) {
1383                 x->km.state = XFRM_STATE_EXPIRED;
1384                 tasklet_hrtimer_start(&x->mtimer, ktime_set(0, 0), HRTIMER_MODE_REL);
1385                 return -EINVAL;
1386         }
1387
1388         if (!x->km.dying &&
1389             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1390              x->curlft.packets >= x->lft.soft_packet_limit)) {
1391                 x->km.dying = 1;
1392                 km_state_expired(x, 0, 0);
1393         }
1394         return 0;
1395 }
1396 EXPORT_SYMBOL(xfrm_state_check_expire);
1397
1398 struct xfrm_state *
1399 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1400                   u8 proto, unsigned short family)
1401 {
1402         struct xfrm_state *x;
1403
1404         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1405         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1406         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1407         return x;
1408 }
1409 EXPORT_SYMBOL(xfrm_state_lookup);
1410
1411 struct xfrm_state *
1412 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1413                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1414                          u8 proto, unsigned short family)
1415 {
1416         struct xfrm_state *x;
1417
1418         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1419         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1420         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1421         return x;
1422 }
1423 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1424
1425 struct xfrm_state *
1426 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1427               u8 proto, const xfrm_address_t *daddr,
1428               const xfrm_address_t *saddr, int create, unsigned short family)
1429 {
1430         struct xfrm_state *x;
1431
1432         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1433         x = __find_acq_core(net, mark, family, mode, reqid, proto, daddr, saddr, create);
1434         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1435
1436         return x;
1437 }
1438 EXPORT_SYMBOL(xfrm_find_acq);
1439
1440 #ifdef CONFIG_XFRM_SUB_POLICY
1441 int
1442 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1443                unsigned short family, struct net *net)
1444 {
1445         int err = 0;
1446         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1447         if (!afinfo)
1448                 return -EAFNOSUPPORT;
1449
1450         spin_lock_bh(&net->xfrm.xfrm_state_lock); /*FIXME*/
1451         if (afinfo->tmpl_sort)
1452                 err = afinfo->tmpl_sort(dst, src, n);
1453         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1454         xfrm_state_put_afinfo(afinfo);
1455         return err;
1456 }
1457 EXPORT_SYMBOL(xfrm_tmpl_sort);
1458
1459 int
1460 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1461                 unsigned short family)
1462 {
1463         int err = 0;
1464         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
1465         struct net *net = xs_net(*src);
1466
1467         if (!afinfo)
1468                 return -EAFNOSUPPORT;
1469
1470         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1471         if (afinfo->state_sort)
1472                 err = afinfo->state_sort(dst, src, n);
1473         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1474         xfrm_state_put_afinfo(afinfo);
1475         return err;
1476 }
1477 EXPORT_SYMBOL(xfrm_state_sort);
1478 #endif
1479
1480 /* Silly enough, but I'm lazy to build resolution list */
1481
1482 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1483 {
1484         int i;
1485
1486         for (i = 0; i <= net->xfrm.state_hmask; i++) {
1487                 struct xfrm_state *x;
1488
1489                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
1490                         if (x->km.seq == seq &&
1491                             (mark & x->mark.m) == x->mark.v &&
1492                             x->km.state == XFRM_STATE_ACQ) {
1493                                 xfrm_state_hold(x);
1494                                 return x;
1495                         }
1496                 }
1497         }
1498         return NULL;
1499 }
1500
1501 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1502 {
1503         struct xfrm_state *x;
1504
1505         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1506         x = __xfrm_find_acq_byseq(net, mark, seq);
1507         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1508         return x;
1509 }
1510 EXPORT_SYMBOL(xfrm_find_acq_byseq);
1511
1512 u32 xfrm_get_acqseq(void)
1513 {
1514         u32 res;
1515         static atomic_t acqseq;
1516
1517         do {
1518                 res = atomic_inc_return(&acqseq);
1519         } while (!res);
1520
1521         return res;
1522 }
1523 EXPORT_SYMBOL(xfrm_get_acqseq);
1524
1525 int verify_spi_info(u8 proto, u32 min, u32 max)
1526 {
1527         switch (proto) {
1528         case IPPROTO_AH:
1529         case IPPROTO_ESP:
1530                 break;
1531
1532         case IPPROTO_COMP:
1533                 /* IPCOMP spi is 16-bits. */
1534                 if (max >= 0x10000)
1535                         return -EINVAL;
1536                 break;
1537
1538         default:
1539                 return -EINVAL;
1540         }
1541
1542         if (min > max)
1543                 return -EINVAL;
1544
1545         return 0;
1546 }
1547 EXPORT_SYMBOL(verify_spi_info);
1548
1549 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
1550 {
1551         struct net *net = xs_net(x);
1552         unsigned int h;
1553         struct xfrm_state *x0;
1554         int err = -ENOENT;
1555         __be32 minspi = htonl(low);
1556         __be32 maxspi = htonl(high);
1557         __be32 newspi = 0;
1558         u32 mark = x->mark.v & x->mark.m;
1559
1560         spin_lock_bh(&x->lock);
1561         if (x->km.state == XFRM_STATE_DEAD)
1562                 goto unlock;
1563
1564         err = 0;
1565         if (x->id.spi)
1566                 goto unlock;
1567
1568         err = -ENOENT;
1569
1570         if (minspi == maxspi) {
1571                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
1572                 if (x0) {
1573                         xfrm_state_put(x0);
1574                         goto unlock;
1575                 }
1576                 newspi = minspi;
1577         } else {
1578                 u32 spi = 0;
1579                 for (h = 0; h < high-low+1; h++) {
1580                         spi = low + prandom_u32()%(high-low+1);
1581                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
1582                         if (x0 == NULL) {
1583                                 newspi = htonl(spi);
1584                                 break;
1585                         }
1586                         xfrm_state_put(x0);
1587                 }
1588         }
1589         if (newspi) {
1590                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
1591                 x->id.spi = newspi;
1592                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
1593                 hlist_add_head(&x->byspi, net->xfrm.state_byspi+h);
1594                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1595
1596                 err = 0;
1597         }
1598
1599 unlock:
1600         spin_unlock_bh(&x->lock);
1601
1602         return err;
1603 }
1604 EXPORT_SYMBOL(xfrm_alloc_spi);
1605
1606 static bool __xfrm_state_filter_match(struct xfrm_state *x,
1607                                       struct xfrm_address_filter *filter)
1608 {
1609         if (filter) {
1610                 if ((filter->family == AF_INET ||
1611                      filter->family == AF_INET6) &&
1612                     x->props.family != filter->family)
1613                         return false;
1614
1615                 return addr_match(&x->props.saddr, &filter->saddr,
1616                                   filter->splen) &&
1617                        addr_match(&x->id.daddr, &filter->daddr,
1618                                   filter->dplen);
1619         }
1620         return true;
1621 }
1622
1623 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
1624                     int (*func)(struct xfrm_state *, int, void*),
1625                     void *data)
1626 {
1627         struct xfrm_state *state;
1628         struct xfrm_state_walk *x;
1629         int err = 0;
1630
1631         if (walk->seq != 0 && list_empty(&walk->all))
1632                 return 0;
1633
1634         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1635         if (list_empty(&walk->all))
1636                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
1637         else
1638                 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
1639         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
1640                 if (x->state == XFRM_STATE_DEAD)
1641                         continue;
1642                 state = container_of(x, struct xfrm_state, km);
1643                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
1644                         continue;
1645                 if (!__xfrm_state_filter_match(state, walk->filter))
1646                         continue;
1647                 err = func(state, walk->seq, data);
1648                 if (err) {
1649                         list_move_tail(&walk->all, &x->all);
1650                         goto out;
1651                 }
1652                 walk->seq++;
1653         }
1654         if (walk->seq == 0) {
1655                 err = -ENOENT;
1656                 goto out;
1657         }
1658         list_del_init(&walk->all);
1659 out:
1660         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1661         return err;
1662 }
1663 EXPORT_SYMBOL(xfrm_state_walk);
1664
1665 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
1666                           struct xfrm_address_filter *filter)
1667 {
1668         INIT_LIST_HEAD(&walk->all);
1669         walk->proto = proto;
1670         walk->state = XFRM_STATE_DEAD;
1671         walk->seq = 0;
1672         walk->filter = filter;
1673 }
1674 EXPORT_SYMBOL(xfrm_state_walk_init);
1675
1676 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
1677 {
1678         kfree(walk->filter);
1679
1680         if (list_empty(&walk->all))
1681                 return;
1682
1683         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1684         list_del(&walk->all);
1685         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1686 }
1687 EXPORT_SYMBOL(xfrm_state_walk_done);
1688
1689 static void xfrm_replay_timer_handler(unsigned long data)
1690 {
1691         struct xfrm_state *x = (struct xfrm_state *)data;
1692
1693         spin_lock(&x->lock);
1694
1695         if (x->km.state == XFRM_STATE_VALID) {
1696                 if (xfrm_aevent_is_on(xs_net(x)))
1697                         x->repl->notify(x, XFRM_REPLAY_TIMEOUT);
1698                 else
1699                         x->xflags |= XFRM_TIME_DEFER;
1700         }
1701
1702         spin_unlock(&x->lock);
1703 }
1704
1705 static LIST_HEAD(xfrm_km_list);
1706
1707 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
1708 {
1709         struct xfrm_mgr *km;
1710
1711         rcu_read_lock();
1712         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1713                 if (km->notify_policy)
1714                         km->notify_policy(xp, dir, c);
1715         rcu_read_unlock();
1716 }
1717
1718 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
1719 {
1720         struct xfrm_mgr *km;
1721         rcu_read_lock();
1722         list_for_each_entry_rcu(km, &xfrm_km_list, list)
1723                 if (km->notify)
1724                         km->notify(x, c);
1725         rcu_read_unlock();
1726 }
1727
1728 EXPORT_SYMBOL(km_policy_notify);
1729 EXPORT_SYMBOL(km_state_notify);
1730
1731 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
1732 {
1733         struct km_event c;
1734
1735         c.data.hard = hard;
1736         c.portid = portid;
1737         c.event = XFRM_MSG_EXPIRE;
1738         km_state_notify(x, &c);
1739 }
1740
1741 EXPORT_SYMBOL(km_state_expired);
1742 /*
1743  * We send to all registered managers regardless of failure
1744  * We are happy with one success
1745 */
1746 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
1747 {
1748         int err = -EINVAL, acqret;
1749         struct xfrm_mgr *km;
1750
1751         rcu_read_lock();
1752         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1753                 acqret = km->acquire(x, t, pol);
1754                 if (!acqret)
1755                         err = acqret;
1756         }
1757         rcu_read_unlock();
1758         return err;
1759 }
1760 EXPORT_SYMBOL(km_query);
1761
1762 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
1763 {
1764         int err = -EINVAL;
1765         struct xfrm_mgr *km;
1766
1767         rcu_read_lock();
1768         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1769                 if (km->new_mapping)
1770                         err = km->new_mapping(x, ipaddr, sport);
1771                 if (!err)
1772                         break;
1773         }
1774         rcu_read_unlock();
1775         return err;
1776 }
1777 EXPORT_SYMBOL(km_new_mapping);
1778
1779 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
1780 {
1781         struct km_event c;
1782
1783         c.data.hard = hard;
1784         c.portid = portid;
1785         c.event = XFRM_MSG_POLEXPIRE;
1786         km_policy_notify(pol, dir, &c);
1787 }
1788 EXPORT_SYMBOL(km_policy_expired);
1789
1790 #ifdef CONFIG_XFRM_MIGRATE
1791 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
1792                const struct xfrm_migrate *m, int num_migrate,
1793                const struct xfrm_kmaddress *k)
1794 {
1795         int err = -EINVAL;
1796         int ret;
1797         struct xfrm_mgr *km;
1798
1799         rcu_read_lock();
1800         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1801                 if (km->migrate) {
1802                         ret = km->migrate(sel, dir, type, m, num_migrate, k);
1803                         if (!ret)
1804                                 err = ret;
1805                 }
1806         }
1807         rcu_read_unlock();
1808         return err;
1809 }
1810 EXPORT_SYMBOL(km_migrate);
1811 #endif
1812
1813 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
1814 {
1815         int err = -EINVAL;
1816         int ret;
1817         struct xfrm_mgr *km;
1818
1819         rcu_read_lock();
1820         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1821                 if (km->report) {
1822                         ret = km->report(net, proto, sel, addr);
1823                         if (!ret)
1824                                 err = ret;
1825                 }
1826         }
1827         rcu_read_unlock();
1828         return err;
1829 }
1830 EXPORT_SYMBOL(km_report);
1831
1832 bool km_is_alive(const struct km_event *c)
1833 {
1834         struct xfrm_mgr *km;
1835         bool is_alive = false;
1836
1837         rcu_read_lock();
1838         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1839                 if (km->is_alive && km->is_alive(c)) {
1840                         is_alive = true;
1841                         break;
1842                 }
1843         }
1844         rcu_read_unlock();
1845
1846         return is_alive;
1847 }
1848 EXPORT_SYMBOL(km_is_alive);
1849
1850 int xfrm_user_policy(struct sock *sk, int optname, u8 __user *optval, int optlen)
1851 {
1852         int err;
1853         u8 *data;
1854         struct xfrm_mgr *km;
1855         struct xfrm_policy *pol = NULL;
1856
1857 #ifdef CONFIG_COMPAT
1858         if (is_compat_task())
1859                 return -EOPNOTSUPP;
1860 #endif
1861
1862         if (!optval && !optlen) {
1863                 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
1864                 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
1865                 __sk_dst_reset(sk);
1866                 return 0;
1867         }
1868
1869         if (optlen <= 0 || optlen > PAGE_SIZE)
1870                 return -EMSGSIZE;
1871
1872         data = kmalloc(optlen, GFP_KERNEL);
1873         if (!data)
1874                 return -ENOMEM;
1875
1876         err = -EFAULT;
1877         if (copy_from_user(data, optval, optlen))
1878                 goto out;
1879
1880         err = -EINVAL;
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
1883                 pol = km->compile_policy(sk, optname, data,
1884                                          optlen, &err);
1885                 if (err >= 0)
1886                         break;
1887         }
1888         rcu_read_unlock();
1889
1890         if (err >= 0) {
1891                 xfrm_sk_policy_insert(sk, err, pol);
1892                 xfrm_pol_put(pol);
1893                 __sk_dst_reset(sk);
1894                 err = 0;
1895         }
1896
1897 out:
1898         kfree(data);
1899         return err;
1900 }
1901 EXPORT_SYMBOL(xfrm_user_policy);
1902
1903 static DEFINE_SPINLOCK(xfrm_km_lock);
1904
1905 int xfrm_register_km(struct xfrm_mgr *km)
1906 {
1907         spin_lock_bh(&xfrm_km_lock);
1908         list_add_tail_rcu(&km->list, &xfrm_km_list);
1909         spin_unlock_bh(&xfrm_km_lock);
1910         return 0;
1911 }
1912 EXPORT_SYMBOL(xfrm_register_km);
1913
1914 int xfrm_unregister_km(struct xfrm_mgr *km)
1915 {
1916         spin_lock_bh(&xfrm_km_lock);
1917         list_del_rcu(&km->list);
1918         spin_unlock_bh(&xfrm_km_lock);
1919         synchronize_rcu();
1920         return 0;
1921 }
1922 EXPORT_SYMBOL(xfrm_unregister_km);
1923
1924 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
1925 {
1926         int err = 0;
1927         if (unlikely(afinfo == NULL))
1928                 return -EINVAL;
1929         if (unlikely(afinfo->family >= NPROTO))
1930                 return -EAFNOSUPPORT;
1931         spin_lock_bh(&xfrm_state_afinfo_lock);
1932         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
1933                 err = -EEXIST;
1934         else
1935                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
1936         spin_unlock_bh(&xfrm_state_afinfo_lock);
1937         return err;
1938 }
1939 EXPORT_SYMBOL(xfrm_state_register_afinfo);
1940
1941 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
1942 {
1943         int err = 0;
1944         if (unlikely(afinfo == NULL))
1945                 return -EINVAL;
1946         if (unlikely(afinfo->family >= NPROTO))
1947                 return -EAFNOSUPPORT;
1948         spin_lock_bh(&xfrm_state_afinfo_lock);
1949         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
1950                 if (unlikely(xfrm_state_afinfo[afinfo->family] != afinfo))
1951                         err = -EINVAL;
1952                 else
1953                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
1954         }
1955         spin_unlock_bh(&xfrm_state_afinfo_lock);
1956         synchronize_rcu();
1957         return err;
1958 }
1959 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
1960
1961 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
1962 {
1963         struct xfrm_state_afinfo *afinfo;
1964         if (unlikely(family >= NPROTO))
1965                 return NULL;
1966         rcu_read_lock();
1967         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
1968         if (unlikely(!afinfo))
1969                 rcu_read_unlock();
1970         return afinfo;
1971 }
1972
1973 void xfrm_state_put_afinfo(struct xfrm_state_afinfo *afinfo)
1974 {
1975         rcu_read_unlock();
1976 }
1977
1978 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
1979 void xfrm_state_delete_tunnel(struct xfrm_state *x)
1980 {
1981         if (x->tunnel) {
1982                 struct xfrm_state *t = x->tunnel;
1983
1984                 if (atomic_read(&t->tunnel_users) == 2)
1985                         xfrm_state_delete(t);
1986                 atomic_dec(&t->tunnel_users);
1987                 xfrm_state_put(t);
1988                 x->tunnel = NULL;
1989         }
1990 }
1991 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
1992
1993 int xfrm_state_mtu(struct xfrm_state *x, int mtu)
1994 {
1995         int res;
1996
1997         spin_lock_bh(&x->lock);
1998         if (x->km.state == XFRM_STATE_VALID &&
1999             x->type && x->type->get_mtu)
2000                 res = x->type->get_mtu(x, mtu);
2001         else
2002                 res = mtu - x->props.header_len;
2003         spin_unlock_bh(&x->lock);
2004         return res;
2005 }
2006
2007 int __xfrm_init_state(struct xfrm_state *x, bool init_replay)
2008 {
2009         struct xfrm_state_afinfo *afinfo;
2010         struct xfrm_mode *inner_mode;
2011         int family = x->props.family;
2012         int err;
2013
2014         err = -EAFNOSUPPORT;
2015         afinfo = xfrm_state_get_afinfo(family);
2016         if (!afinfo)
2017                 goto error;
2018
2019         err = 0;
2020         if (afinfo->init_flags)
2021                 err = afinfo->init_flags(x);
2022
2023         xfrm_state_put_afinfo(afinfo);
2024
2025         if (err)
2026                 goto error;
2027
2028         err = -EPROTONOSUPPORT;
2029
2030         if (x->sel.family != AF_UNSPEC) {
2031                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2032                 if (inner_mode == NULL)
2033                         goto error;
2034
2035                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2036                     family != x->sel.family) {
2037                         xfrm_put_mode(inner_mode);
2038                         goto error;
2039                 }
2040
2041                 x->inner_mode = inner_mode;
2042         } else {
2043                 struct xfrm_mode *inner_mode_iaf;
2044                 int iafamily = AF_INET;
2045
2046                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2047                 if (inner_mode == NULL)
2048                         goto error;
2049
2050                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL)) {
2051                         xfrm_put_mode(inner_mode);
2052                         goto error;
2053                 }
2054                 x->inner_mode = inner_mode;
2055
2056                 if (x->props.family == AF_INET)
2057                         iafamily = AF_INET6;
2058
2059                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2060                 if (inner_mode_iaf) {
2061                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2062                                 x->inner_mode_iaf = inner_mode_iaf;
2063                         else
2064                                 xfrm_put_mode(inner_mode_iaf);
2065                 }
2066         }
2067
2068         x->type = xfrm_get_type(x->id.proto, family);
2069         if (x->type == NULL)
2070                 goto error;
2071
2072         err = x->type->init_state(x);
2073         if (err)
2074                 goto error;
2075
2076         x->outer_mode = xfrm_get_mode(x->props.mode, family);
2077         if (x->outer_mode == NULL) {
2078                 err = -EPROTONOSUPPORT;
2079                 goto error;
2080         }
2081
2082         if (init_replay) {
2083                 err = xfrm_init_replay(x);
2084                 if (err)
2085                         goto error;
2086         }
2087
2088         x->km.state = XFRM_STATE_VALID;
2089
2090 error:
2091         return err;
2092 }
2093
2094 EXPORT_SYMBOL(__xfrm_init_state);
2095
2096 int xfrm_init_state(struct xfrm_state *x)
2097 {
2098         return __xfrm_init_state(x, true);
2099 }
2100
2101 EXPORT_SYMBOL(xfrm_init_state);
2102
2103 int __net_init xfrm_state_init(struct net *net)
2104 {
2105         unsigned int sz;
2106
2107         INIT_LIST_HEAD(&net->xfrm.state_all);
2108
2109         sz = sizeof(struct hlist_head) * 8;
2110
2111         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2112         if (!net->xfrm.state_bydst)
2113                 goto out_bydst;
2114         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2115         if (!net->xfrm.state_bysrc)
2116                 goto out_bysrc;
2117         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2118         if (!net->xfrm.state_byspi)
2119                 goto out_byspi;
2120         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2121
2122         net->xfrm.state_num = 0;
2123         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2124         INIT_HLIST_HEAD(&net->xfrm.state_gc_list);
2125         INIT_WORK(&net->xfrm.state_gc_work, xfrm_state_gc_task);
2126         spin_lock_init(&net->xfrm.xfrm_state_lock);
2127         return 0;
2128
2129 out_byspi:
2130         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2131 out_bysrc:
2132         xfrm_hash_free(net->xfrm.state_bydst, sz);
2133 out_bydst:
2134         return -ENOMEM;
2135 }
2136
2137 void xfrm_state_fini(struct net *net)
2138 {
2139         unsigned int sz;
2140
2141         flush_work(&net->xfrm.state_hash_work);
2142         xfrm_state_flush(net, 0, false);
2143         flush_work(&net->xfrm.state_gc_work);
2144
2145         WARN_ON(!list_empty(&net->xfrm.state_all));
2146
2147         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2148         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2149         xfrm_hash_free(net->xfrm.state_byspi, sz);
2150         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2151         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2152         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2153         xfrm_hash_free(net->xfrm.state_bydst, sz);
2154 }
2155
2156 #ifdef CONFIG_AUDITSYSCALL
2157 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2158                                      struct audit_buffer *audit_buf)
2159 {
2160         struct xfrm_sec_ctx *ctx = x->security;
2161         u32 spi = ntohl(x->id.spi);
2162
2163         if (ctx)
2164                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2165                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2166
2167         switch (x->props.family) {
2168         case AF_INET:
2169                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2170                                  &x->props.saddr.a4, &x->id.daddr.a4);
2171                 break;
2172         case AF_INET6:
2173                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2174                                  x->props.saddr.a6, x->id.daddr.a6);
2175                 break;
2176         }
2177
2178         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2179 }
2180
2181 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2182                                       struct audit_buffer *audit_buf)
2183 {
2184         const struct iphdr *iph4;
2185         const struct ipv6hdr *iph6;
2186
2187         switch (family) {
2188         case AF_INET:
2189                 iph4 = ip_hdr(skb);
2190                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2191                                  &iph4->saddr, &iph4->daddr);
2192                 break;
2193         case AF_INET6:
2194                 iph6 = ipv6_hdr(skb);
2195                 audit_log_format(audit_buf,
2196                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2197                                  &iph6->saddr, &iph6->daddr,
2198                                  iph6->flow_lbl[0] & 0x0f,
2199                                  iph6->flow_lbl[1],
2200                                  iph6->flow_lbl[2]);
2201                 break;
2202         }
2203 }
2204
2205 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2206 {
2207         struct audit_buffer *audit_buf;
2208
2209         audit_buf = xfrm_audit_start("SAD-add");
2210         if (audit_buf == NULL)
2211                 return;
2212         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2213         xfrm_audit_helper_sainfo(x, audit_buf);
2214         audit_log_format(audit_buf, " res=%u", result);
2215         audit_log_end(audit_buf);
2216 }
2217 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2218
2219 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2220 {
2221         struct audit_buffer *audit_buf;
2222
2223         audit_buf = xfrm_audit_start("SAD-delete");
2224         if (audit_buf == NULL)
2225                 return;
2226         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2227         xfrm_audit_helper_sainfo(x, audit_buf);
2228         audit_log_format(audit_buf, " res=%u", result);
2229         audit_log_end(audit_buf);
2230 }
2231 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2232
2233 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2234                                       struct sk_buff *skb)
2235 {
2236         struct audit_buffer *audit_buf;
2237         u32 spi;
2238
2239         audit_buf = xfrm_audit_start("SA-replay-overflow");
2240         if (audit_buf == NULL)
2241                 return;
2242         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2243         /* don't record the sequence number because it's inherent in this kind
2244          * of audit message */
2245         spi = ntohl(x->id.spi);
2246         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2247         audit_log_end(audit_buf);
2248 }
2249 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2250
2251 void xfrm_audit_state_replay(struct xfrm_state *x,
2252                              struct sk_buff *skb, __be32 net_seq)
2253 {
2254         struct audit_buffer *audit_buf;
2255         u32 spi;
2256
2257         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2258         if (audit_buf == NULL)
2259                 return;
2260         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2261         spi = ntohl(x->id.spi);
2262         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2263                          spi, spi, ntohl(net_seq));
2264         audit_log_end(audit_buf);
2265 }
2266 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2267
2268 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2269 {
2270         struct audit_buffer *audit_buf;
2271
2272         audit_buf = xfrm_audit_start("SA-notfound");
2273         if (audit_buf == NULL)
2274                 return;
2275         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2276         audit_log_end(audit_buf);
2277 }
2278 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2279
2280 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2281                                __be32 net_spi, __be32 net_seq)
2282 {
2283         struct audit_buffer *audit_buf;
2284         u32 spi;
2285
2286         audit_buf = xfrm_audit_start("SA-notfound");
2287         if (audit_buf == NULL)
2288                 return;
2289         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2290         spi = ntohl(net_spi);
2291         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2292                          spi, spi, ntohl(net_seq));
2293         audit_log_end(audit_buf);
2294 }
2295 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2296
2297 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2298                               struct sk_buff *skb, u8 proto)
2299 {
2300         struct audit_buffer *audit_buf;
2301         __be32 net_spi;
2302         __be32 net_seq;
2303
2304         audit_buf = xfrm_audit_start("SA-icv-failure");
2305         if (audit_buf == NULL)
2306                 return;
2307         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2308         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2309                 u32 spi = ntohl(net_spi);
2310                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2311                                  spi, spi, ntohl(net_seq));
2312         }
2313         audit_log_end(audit_buf);
2314 }
2315 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2316 #endif /* CONFIG_AUDITSYSCALL */