net/xfrm/xfrm_state.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * xfrm_state.c
   4  *
   5  * Changes:
   6  *      Mitsuru KANDA @USAGI
   7  *      Kazunori MIYAZAWA @USAGI
   8  *      Kunihiro Ishiguro <kunihiro@ipinfusion.com>
   9  *              IPv6 support
  10  *      YOSHIFUJI Hideaki @USAGI
  11  *              Split up af-specific functions
  12  *      Derek Atkins <derek@ihtfp.com>
  13  *              Add UDP Encapsulation
  14  *
  15  */
  16
  17 #include <linux/workqueue.h>
  18 #include <net/xfrm.h>
  19 #include <linux/pfkeyv2.h>
  20 #include <linux/ipsec.h>
  21 #include <linux/module.h>
  22 #include <linux/cache.h>
  23 #include <linux/audit.h>
  24 #include <linux/uaccess.h>
  25 #include <linux/ktime.h>
  26 #include <linux/slab.h>
  27 #include <linux/interrupt.h>
  28 #include <linux/kernel.h>
  29
  30 #include <crypto/aead.h>
  31
  32 #include "xfrm_hash.h"
  33
  34 #define xfrm_state_deref_prot(table, net) \
  35         rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock))
  36
  37 static void xfrm_state_gc_task(struct work_struct *work);
  38
  39 /* Each xfrm_state may be linked to two tables:
  40
  41    1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl)
  42    2. Hash table by (daddr,family,reqid) to find what SAs exist for given
  43       destination/tunnel endpoint. (output)
  44  */
  45
  46 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024;
  47 static struct kmem_cache *xfrm_state_cache __ro_after_init;
  48
  49 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task);
  50 static HLIST_HEAD(xfrm_state_gc_list);
  51
  52 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x)
  53 {
  54         return refcount_inc_not_zero(&x->refcnt);
  55 }
  56
  57 static inline unsigned int xfrm_dst_hash(struct net *net,
  58                                          const xfrm_address_t *daddr,
  59                                          const xfrm_address_t *saddr,
  60                                          u32 reqid,
  61                                          unsigned short family)
  62 {
  63         return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask);
  64 }
  65
  66 static inline unsigned int xfrm_src_hash(struct net *net,
  67                                          const xfrm_address_t *daddr,
  68                                          const xfrm_address_t *saddr,
  69                                          unsigned short family)
  70 {
  71         return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask);
  72 }
  73
  74 static inline unsigned int
  75 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr,
  76               __be32 spi, u8 proto, unsigned short family)
  77 {
  78         return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask);
  79 }
  80
  81 static unsigned int xfrm_seq_hash(struct net *net, u32 seq)
  82 {
  83         return __xfrm_seq_hash(seq, net->xfrm.state_hmask);
  84 }
  85
  86 static void xfrm_hash_transfer(struct hlist_head *list,
  87                                struct hlist_head *ndsttable,
  88                                struct hlist_head *nsrctable,
  89                                struct hlist_head *nspitable,
  90                                struct hlist_head *nseqtable,
  91                                unsigned int nhashmask)
  92 {
  93         struct hlist_node *tmp;
  94         struct xfrm_state *x;
  95
  96         hlist_for_each_entry_safe(x, tmp, list, bydst) {
  97                 unsigned int h;
  98
  99                 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr,
 100                                     x->props.reqid, x->props.family,
 101                                     nhashmask);
 102                 hlist_add_head_rcu(&x->bydst, ndsttable + h);
 103
 104                 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr,
 105                                     x->props.family,
 106                                     nhashmask);
 107                 hlist_add_head_rcu(&x->bysrc, nsrctable + h);
 108
 109                 if (x->id.spi) {
 110                         h = __xfrm_spi_hash(&x->id.daddr, x->id.spi,
 111                                             x->id.proto, x->props.family,
 112                                             nhashmask);
 113                         hlist_add_head_rcu(&x->byspi, nspitable + h);
 114                 }
 115
 116                 if (x->km.seq) {
 117                         h = __xfrm_seq_hash(x->km.seq, nhashmask);
 118                         hlist_add_head_rcu(&x->byseq, nseqtable + h);
 119                 }
 120         }
 121 }
 122
 123 static unsigned long xfrm_hash_new_size(unsigned int state_hmask)
 124 {
 125         return ((state_hmask + 1) << 1) * sizeof(struct hlist_head);
 126 }
 127
 128 static void xfrm_hash_resize(struct work_struct *work)
 129 {
 130         struct net *net = container_of(work, struct net, xfrm.state_hash_work);
 131         struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq;
 132         unsigned long nsize, osize;
 133         unsigned int nhashmask, ohashmask;
 134         int i;
 135
 136         nsize = xfrm_hash_new_size(net->xfrm.state_hmask);
 137         ndst = xfrm_hash_alloc(nsize);
 138         if (!ndst)
 139                 return;
 140         nsrc = xfrm_hash_alloc(nsize);
 141         if (!nsrc) {
 142                 xfrm_hash_free(ndst, nsize);
 143                 return;
 144         }
 145         nspi = xfrm_hash_alloc(nsize);
 146         if (!nspi) {
 147                 xfrm_hash_free(ndst, nsize);
 148                 xfrm_hash_free(nsrc, nsize);
 149                 return;
 150         }
 151         nseq = xfrm_hash_alloc(nsize);
 152         if (!nseq) {
 153                 xfrm_hash_free(ndst, nsize);
 154                 xfrm_hash_free(nsrc, nsize);
 155                 xfrm_hash_free(nspi, nsize);
 156                 return;
 157         }
 158
 159         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 160         write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
 161
 162         nhashmask = (nsize / sizeof(struct hlist_head)) - 1U;
 163         odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net);
 164         for (i = net->xfrm.state_hmask; i >= 0; i--)
 165                 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask);
 166
 167         osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net);
 168         ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net);
 169         oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net);
 170         ohashmask = net->xfrm.state_hmask;
 171
 172         rcu_assign_pointer(net->xfrm.state_bydst, ndst);
 173         rcu_assign_pointer(net->xfrm.state_bysrc, nsrc);
 174         rcu_assign_pointer(net->xfrm.state_byspi, nspi);
 175         rcu_assign_pointer(net->xfrm.state_byseq, nseq);
 176         net->xfrm.state_hmask = nhashmask;
 177
 178         write_seqcount_end(&net->xfrm.xfrm_state_hash_generation);
 179         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 180
 181         osize = (ohashmask + 1) * sizeof(struct hlist_head);
 182
 183         synchronize_rcu();
 184
 185         xfrm_hash_free(odst, osize);
 186         xfrm_hash_free(osrc, osize);
 187         xfrm_hash_free(ospi, osize);
 188         xfrm_hash_free(oseq, osize);
 189 }
 190
 191 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock);
 192 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO];
 193
 194 static DEFINE_SPINLOCK(xfrm_state_gc_lock);
 195
 196 int __xfrm_state_delete(struct xfrm_state *x);
 197
 198 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol);
 199 static bool km_is_alive(const struct km_event *c);
 200 void km_state_expired(struct xfrm_state *x, int hard, u32 portid);
 201
 202 int xfrm_register_type(const struct xfrm_type *type, unsigned short family)
 203 {
 204         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 205         int err = 0;
 206
 207         if (!afinfo)
 208                 return -EAFNOSUPPORT;
 209
 210 #define X(afi, T, name) do {                    \
 211                 WARN_ON((afi)->type_ ## name);  \
 212                 (afi)->type_ ## name = (T);     \
 213         } while (0)
 214
 215         switch (type->proto) {
 216         case IPPROTO_COMP:
 217                 X(afinfo, type, comp);
 218                 break;
 219         case IPPROTO_AH:
 220                 X(afinfo, type, ah);
 221                 break;
 222         case IPPROTO_ESP:
 223                 X(afinfo, type, esp);
 224                 break;
 225         case IPPROTO_IPIP:
 226                 X(afinfo, type, ipip);
 227                 break;
 228         case IPPROTO_DSTOPTS:
 229                 X(afinfo, type, dstopts);
 230                 break;
 231         case IPPROTO_ROUTING:
 232                 X(afinfo, type, routing);
 233                 break;
 234         case IPPROTO_IPV6:
 235                 X(afinfo, type, ipip6);
 236                 break;
 237         default:
 238                 WARN_ON(1);
 239                 err = -EPROTONOSUPPORT;
 240                 break;
 241         }
 242 #undef X
 243         rcu_read_unlock();
 244         return err;
 245 }
 246 EXPORT_SYMBOL(xfrm_register_type);
 247
 248 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family)
 249 {
 250         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 251
 252         if (unlikely(afinfo == NULL))
 253                 return;
 254
 255 #define X(afi, T, name) do {                            \
 256                 WARN_ON((afi)->type_ ## name != (T));   \
 257                 (afi)->type_ ## name = NULL;            \
 258         } while (0)
 259
 260         switch (type->proto) {
 261         case IPPROTO_COMP:
 262                 X(afinfo, type, comp);
 263                 break;
 264         case IPPROTO_AH:
 265                 X(afinfo, type, ah);
 266                 break;
 267         case IPPROTO_ESP:
 268                 X(afinfo, type, esp);
 269                 break;
 270         case IPPROTO_IPIP:
 271                 X(afinfo, type, ipip);
 272                 break;
 273         case IPPROTO_DSTOPTS:
 274                 X(afinfo, type, dstopts);
 275                 break;
 276         case IPPROTO_ROUTING:
 277                 X(afinfo, type, routing);
 278                 break;
 279         case IPPROTO_IPV6:
 280                 X(afinfo, type, ipip6);
 281                 break;
 282         default:
 283                 WARN_ON(1);
 284                 break;
 285         }
 286 #undef X
 287         rcu_read_unlock();
 288 }
 289 EXPORT_SYMBOL(xfrm_unregister_type);
 290
 291 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family)
 292 {
 293         const struct xfrm_type *type = NULL;
 294         struct xfrm_state_afinfo *afinfo;
 295         int modload_attempted = 0;
 296
 297 retry:
 298         afinfo = xfrm_state_get_afinfo(family);
 299         if (unlikely(afinfo == NULL))
 300                 return NULL;
 301
 302         switch (proto) {
 303         case IPPROTO_COMP:
 304                 type = afinfo->type_comp;
 305                 break;
 306         case IPPROTO_AH:
 307                 type = afinfo->type_ah;
 308                 break;
 309         case IPPROTO_ESP:
 310                 type = afinfo->type_esp;
 311                 break;
 312         case IPPROTO_IPIP:
 313                 type = afinfo->type_ipip;
 314                 break;
 315         case IPPROTO_DSTOPTS:
 316                 type = afinfo->type_dstopts;
 317                 break;
 318         case IPPROTO_ROUTING:
 319                 type = afinfo->type_routing;
 320                 break;
 321         case IPPROTO_IPV6:
 322                 type = afinfo->type_ipip6;
 323                 break;
 324         default:
 325                 break;
 326         }
 327
 328         if (unlikely(type && !try_module_get(type->owner)))
 329                 type = NULL;
 330
 331         rcu_read_unlock();
 332
 333         if (!type && !modload_attempted) {
 334                 request_module("xfrm-type-%d-%d", family, proto);
 335                 modload_attempted = 1;
 336                 goto retry;
 337         }
 338
 339         return type;
 340 }
 341
 342 static void xfrm_put_type(const struct xfrm_type *type)
 343 {
 344         module_put(type->owner);
 345 }
 346
 347 int xfrm_register_type_offload(const struct xfrm_type_offload *type,
 348                                unsigned short family)
 349 {
 350         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 351         int err = 0;
 352
 353         if (unlikely(afinfo == NULL))
 354                 return -EAFNOSUPPORT;
 355
 356         switch (type->proto) {
 357         case IPPROTO_ESP:
 358                 WARN_ON(afinfo->type_offload_esp);
 359                 afinfo->type_offload_esp = type;
 360                 break;
 361         default:
 362                 WARN_ON(1);
 363                 err = -EPROTONOSUPPORT;
 364                 break;
 365         }
 366
 367         rcu_read_unlock();
 368         return err;
 369 }
 370 EXPORT_SYMBOL(xfrm_register_type_offload);
 371
 372 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type,
 373                                   unsigned short family)
 374 {
 375         struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family);
 376
 377         if (unlikely(afinfo == NULL))
 378                 return;
 379
 380         switch (type->proto) {
 381         case IPPROTO_ESP:
 382                 WARN_ON(afinfo->type_offload_esp != type);
 383                 afinfo->type_offload_esp = NULL;
 384                 break;
 385         default:
 386                 WARN_ON(1);
 387                 break;
 388         }
 389         rcu_read_unlock();
 390 }
 391 EXPORT_SYMBOL(xfrm_unregister_type_offload);
 392
 393 static const struct xfrm_type_offload *
 394 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load)
 395 {
 396         const struct xfrm_type_offload *type = NULL;
 397         struct xfrm_state_afinfo *afinfo;
 398
 399 retry:
 400         afinfo = xfrm_state_get_afinfo(family);
 401         if (unlikely(afinfo == NULL))
 402                 return NULL;
 403
 404         switch (proto) {
 405         case IPPROTO_ESP:
 406                 type = afinfo->type_offload_esp;
 407                 break;
 408         default:
 409                 break;
 410         }
 411
 412         if ((type && !try_module_get(type->owner)))
 413                 type = NULL;
 414
 415         rcu_read_unlock();
 416
 417         if (!type && try_load) {
 418                 request_module("xfrm-offload-%d-%d", family, proto);
 419                 try_load = false;
 420                 goto retry;
 421         }
 422
 423         return type;
 424 }
 425
 426 static void xfrm_put_type_offload(const struct xfrm_type_offload *type)
 427 {
 428         module_put(type->owner);
 429 }
 430
 431 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = {
 432         [XFRM_MODE_BEET] = {
 433                 .encap = XFRM_MODE_BEET,
 434                 .flags = XFRM_MODE_FLAG_TUNNEL,
 435                 .family = AF_INET,
 436         },
 437         [XFRM_MODE_TRANSPORT] = {
 438                 .encap = XFRM_MODE_TRANSPORT,
 439                 .family = AF_INET,
 440         },
 441         [XFRM_MODE_TUNNEL] = {
 442                 .encap = XFRM_MODE_TUNNEL,
 443                 .flags = XFRM_MODE_FLAG_TUNNEL,
 444                 .family = AF_INET,
 445         },
 446 };
 447
 448 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = {
 449         [XFRM_MODE_BEET] = {
 450                 .encap = XFRM_MODE_BEET,
 451                 .flags = XFRM_MODE_FLAG_TUNNEL,
 452                 .family = AF_INET6,
 453         },
 454         [XFRM_MODE_ROUTEOPTIMIZATION] = {
 455                 .encap = XFRM_MODE_ROUTEOPTIMIZATION,
 456                 .family = AF_INET6,
 457         },
 458         [XFRM_MODE_TRANSPORT] = {
 459                 .encap = XFRM_MODE_TRANSPORT,
 460                 .family = AF_INET6,
 461         },
 462         [XFRM_MODE_TUNNEL] = {
 463                 .encap = XFRM_MODE_TUNNEL,
 464                 .flags = XFRM_MODE_FLAG_TUNNEL,
 465                 .family = AF_INET6,
 466         },
 467 };
 468
 469 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family)
 470 {
 471         const struct xfrm_mode *mode;
 472
 473         if (unlikely(encap >= XFRM_MODE_MAX))
 474                 return NULL;
 475
 476         switch (family) {
 477         case AF_INET:
 478                 mode = &xfrm4_mode_map[encap];
 479                 if (mode->family == family)
 480                         return mode;
 481                 break;
 482         case AF_INET6:
 483                 mode = &xfrm6_mode_map[encap];
 484                 if (mode->family == family)
 485                         return mode;
 486                 break;
 487         default:
 488                 break;
 489         }
 490
 491         return NULL;
 492 }
 493
 494 void xfrm_state_free(struct xfrm_state *x)
 495 {
 496         kmem_cache_free(xfrm_state_cache, x);
 497 }
 498 EXPORT_SYMBOL(xfrm_state_free);
 499
 500 static void ___xfrm_state_destroy(struct xfrm_state *x)
 501 {
 502         hrtimer_cancel(&x->mtimer);
 503         del_timer_sync(&x->rtimer);
 504         kfree(x->aead);
 505         kfree(x->aalg);
 506         kfree(x->ealg);
 507         kfree(x->calg);
 508         kfree(x->encap);
 509         kfree(x->coaddr);
 510         kfree(x->replay_esn);
 511         kfree(x->preplay_esn);
 512         if (x->type_offload)
 513                 xfrm_put_type_offload(x->type_offload);
 514         if (x->type) {
 515                 x->type->destructor(x);
 516                 xfrm_put_type(x->type);
 517         }
 518         if (x->xfrag.page)
 519                 put_page(x->xfrag.page);
 520         xfrm_dev_state_free(x);
 521         security_xfrm_state_free(x);
 522         xfrm_state_free(x);
 523 }
 524
 525 static void xfrm_state_gc_task(struct work_struct *work)
 526 {
 527         struct xfrm_state *x;
 528         struct hlist_node *tmp;
 529         struct hlist_head gc_list;
 530
 531         spin_lock_bh(&xfrm_state_gc_lock);
 532         hlist_move_list(&xfrm_state_gc_list, &gc_list);
 533         spin_unlock_bh(&xfrm_state_gc_lock);
 534
 535         synchronize_rcu();
 536
 537         hlist_for_each_entry_safe(x, tmp, &gc_list, gclist)
 538                 ___xfrm_state_destroy(x);
 539 }
 540
 541 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me)
 542 {
 543         struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer);
 544         enum hrtimer_restart ret = HRTIMER_NORESTART;
 545         time64_t now = ktime_get_real_seconds();
 546         time64_t next = TIME64_MAX;
 547         int warn = 0;
 548         int err = 0;
 549
 550         spin_lock(&x->lock);
 551         if (x->km.state == XFRM_STATE_DEAD)
 552                 goto out;
 553         if (x->km.state == XFRM_STATE_EXPIRED)
 554                 goto expired;
 555         if (x->lft.hard_add_expires_seconds) {
 556                 long tmo = x->lft.hard_add_expires_seconds +
 557                         x->curlft.add_time - now;
 558                 if (tmo <= 0) {
 559                         if (x->xflags & XFRM_SOFT_EXPIRE) {
 560                                 /* enter hard expire without soft expire first?!
 561                                  * setting a new date could trigger this.
 562                                  * workaround: fix x->curflt.add_time by below:
 563                                  */
 564                                 x->curlft.add_time = now - x->saved_tmo - 1;
 565                                 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo;
 566                         } else
 567                                 goto expired;
 568                 }
 569                 if (tmo < next)
 570                         next = tmo;
 571         }
 572         if (x->lft.hard_use_expires_seconds) {
 573                 long tmo = x->lft.hard_use_expires_seconds +
 574                         (x->curlft.use_time ? : now) - now;
 575                 if (tmo <= 0)
 576                         goto expired;
 577                 if (tmo < next)
 578                         next = tmo;
 579         }
 580         if (x->km.dying)
 581                 goto resched;
 582         if (x->lft.soft_add_expires_seconds) {
 583                 long tmo = x->lft.soft_add_expires_seconds +
 584                         x->curlft.add_time - now;
 585                 if (tmo <= 0) {
 586                         warn = 1;
 587                         x->xflags &= ~XFRM_SOFT_EXPIRE;
 588                 } else if (tmo < next) {
 589                         next = tmo;
 590                         x->xflags |= XFRM_SOFT_EXPIRE;
 591                         x->saved_tmo = tmo;
 592                 }
 593         }
 594         if (x->lft.soft_use_expires_seconds) {
 595                 long tmo = x->lft.soft_use_expires_seconds +
 596                         (x->curlft.use_time ? : now) - now;
 597                 if (tmo <= 0)
 598                         warn = 1;
 599                 else if (tmo < next)
 600                         next = tmo;
 601         }
 602
 603         x->km.dying = warn;
 604         if (warn)
 605                 km_state_expired(x, 0, 0);
 606 resched:
 607         if (next != TIME64_MAX) {
 608                 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0));
 609                 ret = HRTIMER_RESTART;
 610         }
 611
 612         goto out;
 613
 614 expired:
 615         if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0)
 616                 x->km.state = XFRM_STATE_EXPIRED;
 617
 618         err = __xfrm_state_delete(x);
 619         if (!err)
 620                 km_state_expired(x, 1, 0);
 621
 622         xfrm_audit_state_delete(x, err ? 0 : 1, true);
 623
 624 out:
 625         spin_unlock(&x->lock);
 626         return ret;
 627 }
 628
 629 static void xfrm_replay_timer_handler(struct timer_list *t);
 630
 631 struct xfrm_state *xfrm_state_alloc(struct net *net)
 632 {
 633         struct xfrm_state *x;
 634
 635         x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC);
 636
 637         if (x) {
 638                 write_pnet(&x->xs_net, net);
 639                 refcount_set(&x->refcnt, 1);
 640                 atomic_set(&x->tunnel_users, 0);
 641                 INIT_LIST_HEAD(&x->km.all);
 642                 INIT_HLIST_NODE(&x->bydst);
 643                 INIT_HLIST_NODE(&x->bysrc);
 644                 INIT_HLIST_NODE(&x->byspi);
 645                 INIT_HLIST_NODE(&x->byseq);
 646                 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT);
 647                 x->mtimer.function = xfrm_timer_handler;
 648                 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0);
 649                 x->curlft.add_time = ktime_get_real_seconds();
 650                 x->lft.soft_byte_limit = XFRM_INF;
 651                 x->lft.soft_packet_limit = XFRM_INF;
 652                 x->lft.hard_byte_limit = XFRM_INF;
 653                 x->lft.hard_packet_limit = XFRM_INF;
 654                 x->replay_maxage = 0;
 655                 x->replay_maxdiff = 0;
 656                 spin_lock_init(&x->lock);
 657         }
 658         return x;
 659 }
 660 EXPORT_SYMBOL(xfrm_state_alloc);
 661
 662 void __xfrm_state_destroy(struct xfrm_state *x, bool sync)
 663 {
 664         WARN_ON(x->km.state != XFRM_STATE_DEAD);
 665
 666         if (sync) {
 667                 synchronize_rcu();
 668                 ___xfrm_state_destroy(x);
 669         } else {
 670                 spin_lock_bh(&xfrm_state_gc_lock);
 671                 hlist_add_head(&x->gclist, &xfrm_state_gc_list);
 672                 spin_unlock_bh(&xfrm_state_gc_lock);
 673                 schedule_work(&xfrm_state_gc_work);
 674         }
 675 }
 676 EXPORT_SYMBOL(__xfrm_state_destroy);
 677
 678 int __xfrm_state_delete(struct xfrm_state *x)
 679 {
 680         struct net *net = xs_net(x);
 681         int err = -ESRCH;
 682
 683         if (x->km.state != XFRM_STATE_DEAD) {
 684                 x->km.state = XFRM_STATE_DEAD;
 685                 spin_lock(&net->xfrm.xfrm_state_lock);
 686                 list_del(&x->km.all);
 687                 hlist_del_rcu(&x->bydst);
 688                 hlist_del_rcu(&x->bysrc);
 689                 if (x->km.seq)
 690                         hlist_del_rcu(&x->byseq);
 691                 if (x->id.spi)
 692                         hlist_del_rcu(&x->byspi);
 693                 net->xfrm.state_num--;
 694                 spin_unlock(&net->xfrm.xfrm_state_lock);
 695
 696                 if (x->encap_sk)
 697                         sock_put(rcu_dereference_raw(x->encap_sk));
 698
 699                 xfrm_dev_state_delete(x);
 700
 701                 /* All xfrm_state objects are created by xfrm_state_alloc.
 702                  * The xfrm_state_alloc call gives a reference, and that
 703                  * is what we are dropping here.
 704                  */
 705                 xfrm_state_put(x);
 706                 err = 0;
 707         }
 708
 709         return err;
 710 }
 711 EXPORT_SYMBOL(__xfrm_state_delete);
 712
 713 int xfrm_state_delete(struct xfrm_state *x)
 714 {
 715         int err;
 716
 717         spin_lock_bh(&x->lock);
 718         err = __xfrm_state_delete(x);
 719         spin_unlock_bh(&x->lock);
 720
 721         return err;
 722 }
 723 EXPORT_SYMBOL(xfrm_state_delete);
 724
 725 #ifdef CONFIG_SECURITY_NETWORK_XFRM
 726 static inline int
 727 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 728 {
 729         int i, err = 0;
 730
 731         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 732                 struct xfrm_state *x;
 733
 734                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 735                         if (xfrm_id_proto_match(x->id.proto, proto) &&
 736                            (err = security_xfrm_state_delete(x)) != 0) {
 737                                 xfrm_audit_state_delete(x, 0, task_valid);
 738                                 return err;
 739                         }
 740                 }
 741         }
 742
 743         return err;
 744 }
 745
 746 static inline int
 747 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
 748 {
 749         int i, err = 0;
 750
 751         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 752                 struct xfrm_state *x;
 753                 struct xfrm_state_offload *xso;
 754
 755                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 756                         xso = &x->xso;
 757
 758                         if (xso->dev == dev &&
 759                            (err = security_xfrm_state_delete(x)) != 0) {
 760                                 xfrm_audit_state_delete(x, 0, task_valid);
 761                                 return err;
 762                         }
 763                 }
 764         }
 765
 766         return err;
 767 }
 768 #else
 769 static inline int
 770 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid)
 771 {
 772         return 0;
 773 }
 774
 775 static inline int
 776 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid)
 777 {
 778         return 0;
 779 }
 780 #endif
 781
 782 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync)
 783 {
 784         int i, err = 0, cnt = 0;
 785
 786         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 787         err = xfrm_state_flush_secctx_check(net, proto, task_valid);
 788         if (err)
 789                 goto out;
 790
 791         err = -ESRCH;
 792         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 793                 struct xfrm_state *x;
 794 restart:
 795                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 796                         if (!xfrm_state_kern(x) &&
 797                             xfrm_id_proto_match(x->id.proto, proto)) {
 798                                 xfrm_state_hold(x);
 799                                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 800
 801                                 err = xfrm_state_delete(x);
 802                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 803                                                         task_valid);
 804                                 if (sync)
 805                                         xfrm_state_put_sync(x);
 806                                 else
 807                                         xfrm_state_put(x);
 808                                 if (!err)
 809                                         cnt++;
 810
 811                                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
 812                                 goto restart;
 813                         }
 814                 }
 815         }
 816 out:
 817         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 818         if (cnt)
 819                 err = 0;
 820
 821         return err;
 822 }
 823 EXPORT_SYMBOL(xfrm_state_flush);
 824
 825 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid)
 826 {
 827         int i, err = 0, cnt = 0;
 828
 829         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 830         err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid);
 831         if (err)
 832                 goto out;
 833
 834         err = -ESRCH;
 835         for (i = 0; i <= net->xfrm.state_hmask; i++) {
 836                 struct xfrm_state *x;
 837                 struct xfrm_state_offload *xso;
 838 restart:
 839                 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) {
 840                         xso = &x->xso;
 841
 842                         if (!xfrm_state_kern(x) && xso->dev == dev) {
 843                                 xfrm_state_hold(x);
 844                                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 845
 846                                 err = xfrm_state_delete(x);
 847                                 xfrm_audit_state_delete(x, err ? 0 : 1,
 848                                                         task_valid);
 849                                 xfrm_state_put(x);
 850                                 if (!err)
 851                                         cnt++;
 852
 853                                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
 854                                 goto restart;
 855                         }
 856                 }
 857         }
 858         if (cnt)
 859                 err = 0;
 860
 861 out:
 862         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 863         return err;
 864 }
 865 EXPORT_SYMBOL(xfrm_dev_state_flush);
 866
 867 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si)
 868 {
 869         spin_lock_bh(&net->xfrm.xfrm_state_lock);
 870         si->sadcnt = net->xfrm.state_num;
 871         si->sadhcnt = net->xfrm.state_hmask + 1;
 872         si->sadhmcnt = xfrm_state_hashmax;
 873         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
 874 }
 875 EXPORT_SYMBOL(xfrm_sad_getinfo);
 876
 877 static void
 878 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
 879 {
 880         const struct flowi4 *fl4 = &fl->u.ip4;
 881
 882         sel->daddr.a4 = fl4->daddr;
 883         sel->saddr.a4 = fl4->saddr;
 884         sel->dport = xfrm_flowi_dport(fl, &fl4->uli);
 885         sel->dport_mask = htons(0xffff);
 886         sel->sport = xfrm_flowi_sport(fl, &fl4->uli);
 887         sel->sport_mask = htons(0xffff);
 888         sel->family = AF_INET;
 889         sel->prefixlen_d = 32;
 890         sel->prefixlen_s = 32;
 891         sel->proto = fl4->flowi4_proto;
 892         sel->ifindex = fl4->flowi4_oif;
 893 }
 894
 895 static void
 896 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl)
 897 {
 898         const struct flowi6 *fl6 = &fl->u.ip6;
 899
 900         /* Initialize temporary selector matching only to current session. */
 901         *(struct in6_addr *)&sel->daddr = fl6->daddr;
 902         *(struct in6_addr *)&sel->saddr = fl6->saddr;
 903         sel->dport = xfrm_flowi_dport(fl, &fl6->uli);
 904         sel->dport_mask = htons(0xffff);
 905         sel->sport = xfrm_flowi_sport(fl, &fl6->uli);
 906         sel->sport_mask = htons(0xffff);
 907         sel->family = AF_INET6;
 908         sel->prefixlen_d = 128;
 909         sel->prefixlen_s = 128;
 910         sel->proto = fl6->flowi6_proto;
 911         sel->ifindex = fl6->flowi6_oif;
 912 }
 913
 914 static void
 915 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl,
 916                     const struct xfrm_tmpl *tmpl,
 917                     const xfrm_address_t *daddr, const xfrm_address_t *saddr,
 918                     unsigned short family)
 919 {
 920         switch (family) {
 921         case AF_INET:
 922                 __xfrm4_init_tempsel(&x->sel, fl);
 923                 break;
 924         case AF_INET6:
 925                 __xfrm6_init_tempsel(&x->sel, fl);
 926                 break;
 927         }
 928
 929         x->id = tmpl->id;
 930
 931         switch (tmpl->encap_family) {
 932         case AF_INET:
 933                 if (x->id.daddr.a4 == 0)
 934                         x->id.daddr.a4 = daddr->a4;
 935                 x->props.saddr = tmpl->saddr;
 936                 if (x->props.saddr.a4 == 0)
 937                         x->props.saddr.a4 = saddr->a4;
 938                 break;
 939         case AF_INET6:
 940                 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr))
 941                         memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr));
 942                 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr));
 943                 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr))
 944                         memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr));
 945                 break;
 946         }
 947
 948         x->props.mode = tmpl->mode;
 949         x->props.reqid = tmpl->reqid;
 950         x->props.family = tmpl->encap_family;
 951 }
 952
 953 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark,
 954                                               const xfrm_address_t *daddr,
 955                                               __be32 spi, u8 proto,
 956                                               unsigned short family)
 957 {
 958         unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family);
 959         struct xfrm_state *x;
 960
 961         hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) {
 962                 if (x->props.family != family ||
 963                     x->id.spi       != spi ||
 964                     x->id.proto     != proto ||
 965                     !xfrm_addr_equal(&x->id.daddr, daddr, family))
 966                         continue;
 967
 968                 if ((mark & x->mark.m) != x->mark.v)
 969                         continue;
 970                 if (!xfrm_state_hold_rcu(x))
 971                         continue;
 972                 return x;
 973         }
 974
 975         return NULL;
 976 }
 977
 978 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark,
 979                                                      const xfrm_address_t *daddr,
 980                                                      const xfrm_address_t *saddr,
 981                                                      u8 proto, unsigned short family)
 982 {
 983         unsigned int h = xfrm_src_hash(net, daddr, saddr, family);
 984         struct xfrm_state *x;
 985
 986         hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) {
 987                 if (x->props.family != family ||
 988                     x->id.proto     != proto ||
 989                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
 990                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
 991                         continue;
 992
 993                 if ((mark & x->mark.m) != x->mark.v)
 994                         continue;
 995                 if (!xfrm_state_hold_rcu(x))
 996                         continue;
 997                 return x;
 998         }
 999
1000         return NULL;
1001 }
1002
1003 static inline struct xfrm_state *
1004 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family)
1005 {
1006         struct net *net = xs_net(x);
1007         u32 mark = x->mark.v & x->mark.m;
1008
1009         if (use_spi)
1010                 return __xfrm_state_lookup(net, mark, &x->id.daddr,
1011                                            x->id.spi, x->id.proto, family);
1012         else
1013                 return __xfrm_state_lookup_byaddr(net, mark,
1014                                                   &x->id.daddr,
1015                                                   &x->props.saddr,
1016                                                   x->id.proto, family);
1017 }
1018
1019 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision)
1020 {
1021         if (have_hash_collision &&
1022             (net->xfrm.state_hmask + 1) < xfrm_state_hashmax &&
1023             net->xfrm.state_num > net->xfrm.state_hmask)
1024                 schedule_work(&net->xfrm.state_hash_work);
1025 }
1026
1027 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x,
1028                                const struct flowi *fl, unsigned short family,
1029                                struct xfrm_state **best, int *acq_in_progress,
1030                                int *error)
1031 {
1032         /* Resolution logic:
1033          * 1. There is a valid state with matching selector. Done.
1034          * 2. Valid state with inappropriate selector. Skip.
1035          *
1036          * Entering area of "sysdeps".
1037          *
1038          * 3. If state is not valid, selector is temporary, it selects
1039          *    only session which triggered previous resolution. Key
1040          *    manager will do something to install a state with proper
1041          *    selector.
1042          */
1043         if (x->km.state == XFRM_STATE_VALID) {
1044                 if ((x->sel.family &&
1045                      (x->sel.family != family ||
1046                       !xfrm_selector_match(&x->sel, fl, family))) ||
1047                     !security_xfrm_state_pol_flow_match(x, pol,
1048                                                         &fl->u.__fl_common))
1049                         return;
1050
1051                 if (!*best ||
1052                     (*best)->km.dying > x->km.dying ||
1053                     ((*best)->km.dying == x->km.dying &&
1054                      (*best)->curlft.add_time < x->curlft.add_time))
1055                         *best = x;
1056         } else if (x->km.state == XFRM_STATE_ACQ) {
1057                 *acq_in_progress = 1;
1058         } else if (x->km.state == XFRM_STATE_ERROR ||
1059                    x->km.state == XFRM_STATE_EXPIRED) {
1060                 if ((!x->sel.family ||
1061                      (x->sel.family == family &&
1062                       xfrm_selector_match(&x->sel, fl, family))) &&
1063                     security_xfrm_state_pol_flow_match(x, pol,
1064                                                        &fl->u.__fl_common))
1065                         *error = -ESRCH;
1066         }
1067 }
1068
1069 struct xfrm_state *
1070 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1071                 const struct flowi *fl, struct xfrm_tmpl *tmpl,
1072                 struct xfrm_policy *pol, int *err,
1073                 unsigned short family, u32 if_id)
1074 {
1075         static xfrm_address_t saddr_wildcard = { };
1076         struct net *net = xp_net(pol);
1077         unsigned int h, h_wildcard;
1078         struct xfrm_state *x, *x0, *to_put;
1079         int acquire_in_progress = 0;
1080         int error = 0;
1081         struct xfrm_state *best = NULL;
1082         u32 mark = pol->mark.v & pol->mark.m;
1083         unsigned short encap_family = tmpl->encap_family;
1084         unsigned int sequence;
1085         struct km_event c;
1086
1087         to_put = NULL;
1088
1089         sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation);
1090
1091         rcu_read_lock();
1092         h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family);
1093         hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) {
1094                 if (x->props.family == encap_family &&
1095                     x->props.reqid == tmpl->reqid &&
1096                     (mark & x->mark.m) == x->mark.v &&
1097                     x->if_id == if_id &&
1098                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
1099                     xfrm_state_addr_check(x, daddr, saddr, encap_family) &&
1100                     tmpl->mode == x->props.mode &&
1101                     tmpl->id.proto == x->id.proto &&
1102                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1103                         xfrm_state_look_at(pol, x, fl, family,
1104                                            &best, &acquire_in_progress, &error);
1105         }
1106         if (best || acquire_in_progress)
1107                 goto found;
1108
1109         h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family);
1110         hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) {
1111                 if (x->props.family == encap_family &&
1112                     x->props.reqid == tmpl->reqid &&
1113                     (mark & x->mark.m) == x->mark.v &&
1114                     x->if_id == if_id &&
1115                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
1116                     xfrm_addr_equal(&x->id.daddr, daddr, encap_family) &&
1117                     tmpl->mode == x->props.mode &&
1118                     tmpl->id.proto == x->id.proto &&
1119                     (tmpl->id.spi == x->id.spi || !tmpl->id.spi))
1120                         xfrm_state_look_at(pol, x, fl, family,
1121                                            &best, &acquire_in_progress, &error);
1122         }
1123
1124 found:
1125         x = best;
1126         if (!x && !error && !acquire_in_progress) {
1127                 if (tmpl->id.spi &&
1128                     (x0 = __xfrm_state_lookup(net, mark, daddr, tmpl->id.spi,
1129                                               tmpl->id.proto, encap_family)) != NULL) {
1130                         to_put = x0;
1131                         error = -EEXIST;
1132                         goto out;
1133                 }
1134
1135                 c.net = net;
1136                 /* If the KMs have no listeners (yet...), avoid allocating an SA
1137                  * for each and every packet - garbage collection might not
1138                  * handle the flood.
1139                  */
1140                 if (!km_is_alive(&c)) {
1141                         error = -ESRCH;
1142                         goto out;
1143                 }
1144
1145                 x = xfrm_state_alloc(net);
1146                 if (x == NULL) {
1147                         error = -ENOMEM;
1148                         goto out;
1149                 }
1150                 /* Initialize temporary state matching only
1151                  * to current session. */
1152                 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family);
1153                 memcpy(&x->mark, &pol->mark, sizeof(x->mark));
1154                 x->if_id = if_id;
1155
1156                 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid);
1157                 if (error) {
1158                         x->km.state = XFRM_STATE_DEAD;
1159                         to_put = x;
1160                         x = NULL;
1161                         goto out;
1162                 }
1163
1164                 if (km_query(x, tmpl, pol) == 0) {
1165                         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1166                         x->km.state = XFRM_STATE_ACQ;
1167                         list_add(&x->km.all, &net->xfrm.state_all);
1168                         hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1169                         h = xfrm_src_hash(net, daddr, saddr, encap_family);
1170                         hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1171                         if (x->id.spi) {
1172                                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family);
1173                                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1174                         }
1175                         if (x->km.seq) {
1176                                 h = xfrm_seq_hash(net, x->km.seq);
1177                                 hlist_add_head_rcu(&x->byseq, net->xfrm.state_byseq + h);
1178                         }
1179                         x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1180                         hrtimer_start(&x->mtimer,
1181                                       ktime_set(net->xfrm.sysctl_acq_expires, 0),
1182                                       HRTIMER_MODE_REL_SOFT);
1183                         net->xfrm.state_num++;
1184                         xfrm_hash_grow_check(net, x->bydst.next != NULL);
1185                         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1186                 } else {
1187                         x->km.state = XFRM_STATE_DEAD;
1188                         to_put = x;
1189                         x = NULL;
1190                         error = -ESRCH;
1191                 }
1192         }
1193 out:
1194         if (x) {
1195                 if (!xfrm_state_hold_rcu(x)) {
1196                         *err = -EAGAIN;
1197                         x = NULL;
1198                 }
1199         } else {
1200                 *err = acquire_in_progress ? -EAGAIN : error;
1201         }
1202         rcu_read_unlock();
1203         if (to_put)
1204                 xfrm_state_put(to_put);
1205
1206         if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) {
1207                 *err = -EAGAIN;
1208                 if (x) {
1209                         xfrm_state_put(x);
1210                         x = NULL;
1211                 }
1212         }
1213
1214         return x;
1215 }
1216
1217 struct xfrm_state *
1218 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id,
1219                     xfrm_address_t *daddr, xfrm_address_t *saddr,
1220                     unsigned short family, u8 mode, u8 proto, u32 reqid)
1221 {
1222         unsigned int h;
1223         struct xfrm_state *rx = NULL, *x = NULL;
1224
1225         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1226         h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1227         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1228                 if (x->props.family == family &&
1229                     x->props.reqid == reqid &&
1230                     (mark & x->mark.m) == x->mark.v &&
1231                     x->if_id == if_id &&
1232                     !(x->props.flags & XFRM_STATE_WILDRECV) &&
1233                     xfrm_state_addr_check(x, daddr, saddr, family) &&
1234                     mode == x->props.mode &&
1235                     proto == x->id.proto &&
1236                     x->km.state == XFRM_STATE_VALID) {
1237                         rx = x;
1238                         break;
1239                 }
1240         }
1241
1242         if (rx)
1243                 xfrm_state_hold(rx);
1244         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1245
1246
1247         return rx;
1248 }
1249 EXPORT_SYMBOL(xfrm_stateonly_find);
1250
1251 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi,
1252                                               unsigned short family)
1253 {
1254         struct xfrm_state *x;
1255         struct xfrm_state_walk *w;
1256
1257         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1258         list_for_each_entry(w, &net->xfrm.state_all, all) {
1259                 x = container_of(w, struct xfrm_state, km);
1260                 if (x->props.family != family ||
1261                         x->id.spi != spi)
1262                         continue;
1263
1264                 xfrm_state_hold(x);
1265                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1266                 return x;
1267         }
1268         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1269         return NULL;
1270 }
1271 EXPORT_SYMBOL(xfrm_state_lookup_byspi);
1272
1273 static void __xfrm_state_insert(struct xfrm_state *x)
1274 {
1275         struct net *net = xs_net(x);
1276         unsigned int h;
1277
1278         list_add(&x->km.all, &net->xfrm.state_all);
1279
1280         h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr,
1281                           x->props.reqid, x->props.family);
1282         hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1283
1284         h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family);
1285         hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1286
1287         if (x->id.spi) {
1288                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto,
1289                                   x->props.family);
1290
1291                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
1292         }
1293
1294         if (x->km.seq) {
1295                 h = xfrm_seq_hash(net, x->km.seq);
1296
1297                 hlist_add_head_rcu(&x->byseq, net->xfrm.state_byseq + h);
1298         }
1299
1300         hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT);
1301         if (x->replay_maxage)
1302                 mod_timer(&x->rtimer, jiffies + x->replay_maxage);
1303
1304         net->xfrm.state_num++;
1305
1306         xfrm_hash_grow_check(net, x->bydst.next != NULL);
1307 }
1308
1309 /* net->xfrm.xfrm_state_lock is held */
1310 static void __xfrm_state_bump_genids(struct xfrm_state *xnew)
1311 {
1312         struct net *net = xs_net(xnew);
1313         unsigned short family = xnew->props.family;
1314         u32 reqid = xnew->props.reqid;
1315         struct xfrm_state *x;
1316         unsigned int h;
1317         u32 mark = xnew->mark.v & xnew->mark.m;
1318         u32 if_id = xnew->if_id;
1319
1320         h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family);
1321         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1322                 if (x->props.family     == family &&
1323                     x->props.reqid      == reqid &&
1324                     x->if_id            == if_id &&
1325                     (mark & x->mark.m) == x->mark.v &&
1326                     xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) &&
1327                     xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family))
1328                         x->genid++;
1329         }
1330 }
1331
1332 void xfrm_state_insert(struct xfrm_state *x)
1333 {
1334         struct net *net = xs_net(x);
1335
1336         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1337         __xfrm_state_bump_genids(x);
1338         __xfrm_state_insert(x);
1339         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1340 }
1341 EXPORT_SYMBOL(xfrm_state_insert);
1342
1343 /* net->xfrm.xfrm_state_lock is held */
1344 static struct xfrm_state *__find_acq_core(struct net *net,
1345                                           const struct xfrm_mark *m,
1346                                           unsigned short family, u8 mode,
1347                                           u32 reqid, u32 if_id, u8 proto,
1348                                           const xfrm_address_t *daddr,
1349                                           const xfrm_address_t *saddr,
1350                                           int create)
1351 {
1352         unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family);
1353         struct xfrm_state *x;
1354         u32 mark = m->v & m->m;
1355
1356         hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1357                 if (x->props.reqid  != reqid ||
1358                     x->props.mode   != mode ||
1359                     x->props.family != family ||
1360                     x->km.state     != XFRM_STATE_ACQ ||
1361                     x->id.spi       != 0 ||
1362                     x->id.proto     != proto ||
1363                     (mark & x->mark.m) != x->mark.v ||
1364                     !xfrm_addr_equal(&x->id.daddr, daddr, family) ||
1365                     !xfrm_addr_equal(&x->props.saddr, saddr, family))
1366                         continue;
1367
1368                 xfrm_state_hold(x);
1369                 return x;
1370         }
1371
1372         if (!create)
1373                 return NULL;
1374
1375         x = xfrm_state_alloc(net);
1376         if (likely(x)) {
1377                 switch (family) {
1378                 case AF_INET:
1379                         x->sel.daddr.a4 = daddr->a4;
1380                         x->sel.saddr.a4 = saddr->a4;
1381                         x->sel.prefixlen_d = 32;
1382                         x->sel.prefixlen_s = 32;
1383                         x->props.saddr.a4 = saddr->a4;
1384                         x->id.daddr.a4 = daddr->a4;
1385                         break;
1386
1387                 case AF_INET6:
1388                         x->sel.daddr.in6 = daddr->in6;
1389                         x->sel.saddr.in6 = saddr->in6;
1390                         x->sel.prefixlen_d = 128;
1391                         x->sel.prefixlen_s = 128;
1392                         x->props.saddr.in6 = saddr->in6;
1393                         x->id.daddr.in6 = daddr->in6;
1394                         break;
1395                 }
1396
1397                 x->km.state = XFRM_STATE_ACQ;
1398                 x->id.proto = proto;
1399                 x->props.family = family;
1400                 x->props.mode = mode;
1401                 x->props.reqid = reqid;
1402                 x->if_id = if_id;
1403                 x->mark.v = m->v;
1404                 x->mark.m = m->m;
1405                 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires;
1406                 xfrm_state_hold(x);
1407                 hrtimer_start(&x->mtimer,
1408                               ktime_set(net->xfrm.sysctl_acq_expires, 0),
1409                               HRTIMER_MODE_REL_SOFT);
1410                 list_add(&x->km.all, &net->xfrm.state_all);
1411                 hlist_add_head_rcu(&x->bydst, net->xfrm.state_bydst + h);
1412                 h = xfrm_src_hash(net, daddr, saddr, family);
1413                 hlist_add_head_rcu(&x->bysrc, net->xfrm.state_bysrc + h);
1414
1415                 net->xfrm.state_num++;
1416
1417                 xfrm_hash_grow_check(net, x->bydst.next != NULL);
1418         }
1419
1420         return x;
1421 }
1422
1423 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq);
1424
1425 int xfrm_state_add(struct xfrm_state *x)
1426 {
1427         struct net *net = xs_net(x);
1428         struct xfrm_state *x1, *to_put;
1429         int family;
1430         int err;
1431         u32 mark = x->mark.v & x->mark.m;
1432         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1433
1434         family = x->props.family;
1435
1436         to_put = NULL;
1437
1438         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1439
1440         x1 = __xfrm_state_locate(x, use_spi, family);
1441         if (x1) {
1442                 to_put = x1;
1443                 x1 = NULL;
1444                 err = -EEXIST;
1445                 goto out;
1446         }
1447
1448         if (use_spi && x->km.seq) {
1449                 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq);
1450                 if (x1 && ((x1->id.proto != x->id.proto) ||
1451                     !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) {
1452                         to_put = x1;
1453                         x1 = NULL;
1454                 }
1455         }
1456
1457         if (use_spi && !x1)
1458                 x1 = __find_acq_core(net, &x->mark, family, x->props.mode,
1459                                      x->props.reqid, x->if_id, x->id.proto,
1460                                      &x->id.daddr, &x->props.saddr, 0);
1461
1462         __xfrm_state_bump_genids(x);
1463         __xfrm_state_insert(x);
1464         err = 0;
1465
1466 out:
1467         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1468
1469         if (x1) {
1470                 xfrm_state_delete(x1);
1471                 xfrm_state_put(x1);
1472         }
1473
1474         if (to_put)
1475                 xfrm_state_put(to_put);
1476
1477         return err;
1478 }
1479 EXPORT_SYMBOL(xfrm_state_add);
1480
1481 #ifdef CONFIG_XFRM_MIGRATE
1482 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security)
1483 {
1484         struct xfrm_user_sec_ctx *uctx;
1485         int size = sizeof(*uctx) + security->ctx_len;
1486         int err;
1487
1488         uctx = kmalloc(size, GFP_KERNEL);
1489         if (!uctx)
1490                 return -ENOMEM;
1491
1492         uctx->exttype = XFRMA_SEC_CTX;
1493         uctx->len = size;
1494         uctx->ctx_doi = security->ctx_doi;
1495         uctx->ctx_alg = security->ctx_alg;
1496         uctx->ctx_len = security->ctx_len;
1497         memcpy(uctx + 1, security->ctx_str, security->ctx_len);
1498         err = security_xfrm_state_alloc(x, uctx);
1499         kfree(uctx);
1500         if (err)
1501                 return err;
1502
1503         return 0;
1504 }
1505
1506 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig,
1507                                            struct xfrm_encap_tmpl *encap)
1508 {
1509         struct net *net = xs_net(orig);
1510         struct xfrm_state *x = xfrm_state_alloc(net);
1511         if (!x)
1512                 goto out;
1513
1514         memcpy(&x->id, &orig->id, sizeof(x->id));
1515         memcpy(&x->sel, &orig->sel, sizeof(x->sel));
1516         memcpy(&x->lft, &orig->lft, sizeof(x->lft));
1517         x->props.mode = orig->props.mode;
1518         x->props.replay_window = orig->props.replay_window;
1519         x->props.reqid = orig->props.reqid;
1520         x->props.family = orig->props.family;
1521         x->props.saddr = orig->props.saddr;
1522
1523         if (orig->aalg) {
1524                 x->aalg = xfrm_algo_auth_clone(orig->aalg);
1525                 if (!x->aalg)
1526                         goto error;
1527         }
1528         x->props.aalgo = orig->props.aalgo;
1529
1530         if (orig->aead) {
1531                 x->aead = xfrm_algo_aead_clone(orig->aead);
1532                 x->geniv = orig->geniv;
1533                 if (!x->aead)
1534                         goto error;
1535         }
1536         if (orig->ealg) {
1537                 x->ealg = xfrm_algo_clone(orig->ealg);
1538                 if (!x->ealg)
1539                         goto error;
1540         }
1541         x->props.ealgo = orig->props.ealgo;
1542
1543         if (orig->calg) {
1544                 x->calg = xfrm_algo_clone(orig->calg);
1545                 if (!x->calg)
1546                         goto error;
1547         }
1548         x->props.calgo = orig->props.calgo;
1549
1550         if (encap || orig->encap) {
1551                 if (encap)
1552                         x->encap = kmemdup(encap, sizeof(*x->encap),
1553                                         GFP_KERNEL);
1554                 else
1555                         x->encap = kmemdup(orig->encap, sizeof(*x->encap),
1556                                         GFP_KERNEL);
1557
1558                 if (!x->encap)
1559                         goto error;
1560         }
1561
1562         if (orig->security)
1563                 if (clone_security(x, orig->security))
1564                         goto error;
1565
1566         if (orig->coaddr) {
1567                 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr),
1568                                     GFP_KERNEL);
1569                 if (!x->coaddr)
1570                         goto error;
1571         }
1572
1573         if (orig->replay_esn) {
1574                 if (xfrm_replay_clone(x, orig))
1575                         goto error;
1576         }
1577
1578         memcpy(&x->mark, &orig->mark, sizeof(x->mark));
1579         memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark));
1580
1581         x->props.flags = orig->props.flags;
1582         x->props.extra_flags = orig->props.extra_flags;
1583
1584         x->if_id = orig->if_id;
1585         x->tfcpad = orig->tfcpad;
1586         x->replay_maxdiff = orig->replay_maxdiff;
1587         x->replay_maxage = orig->replay_maxage;
1588         memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft));
1589         x->km.state = orig->km.state;
1590         x->km.seq = orig->km.seq;
1591         x->replay = orig->replay;
1592         x->preplay = orig->preplay;
1593         x->mapping_maxage = orig->mapping_maxage;
1594         x->lastused = orig->lastused;
1595         x->new_mapping = 0;
1596         x->new_mapping_sport = 0;
1597
1598         return x;
1599
1600  error:
1601         xfrm_state_put(x);
1602 out:
1603         return NULL;
1604 }
1605
1606 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net,
1607                                                 u32 if_id)
1608 {
1609         unsigned int h;
1610         struct xfrm_state *x = NULL;
1611
1612         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1613
1614         if (m->reqid) {
1615                 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr,
1616                                   m->reqid, m->old_family);
1617                 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) {
1618                         if (x->props.mode != m->mode ||
1619                             x->id.proto != m->proto)
1620                                 continue;
1621                         if (m->reqid && x->props.reqid != m->reqid)
1622                                 continue;
1623                         if (if_id != 0 && x->if_id != if_id)
1624                                 continue;
1625                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1626                                              m->old_family) ||
1627                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1628                                              m->old_family))
1629                                 continue;
1630                         xfrm_state_hold(x);
1631                         break;
1632                 }
1633         } else {
1634                 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr,
1635                                   m->old_family);
1636                 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) {
1637                         if (x->props.mode != m->mode ||
1638                             x->id.proto != m->proto)
1639                                 continue;
1640                         if (if_id != 0 && x->if_id != if_id)
1641                                 continue;
1642                         if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr,
1643                                              m->old_family) ||
1644                             !xfrm_addr_equal(&x->props.saddr, &m->old_saddr,
1645                                              m->old_family))
1646                                 continue;
1647                         xfrm_state_hold(x);
1648                         break;
1649                 }
1650         }
1651
1652         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1653
1654         return x;
1655 }
1656 EXPORT_SYMBOL(xfrm_migrate_state_find);
1657
1658 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x,
1659                                       struct xfrm_migrate *m,
1660                                       struct xfrm_encap_tmpl *encap)
1661 {
1662         struct xfrm_state *xc;
1663
1664         xc = xfrm_state_clone(x, encap);
1665         if (!xc)
1666                 return NULL;
1667
1668         xc->props.family = m->new_family;
1669
1670         if (xfrm_init_state(xc) < 0)
1671                 goto error;
1672
1673         memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr));
1674         memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr));
1675
1676         /* add state */
1677         if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) {
1678                 /* a care is needed when the destination address of the
1679                    state is to be updated as it is a part of triplet */
1680                 xfrm_state_insert(xc);
1681         } else {
1682                 if (xfrm_state_add(xc) < 0)
1683                         goto error;
1684         }
1685
1686         return xc;
1687 error:
1688         xfrm_state_put(xc);
1689         return NULL;
1690 }
1691 EXPORT_SYMBOL(xfrm_state_migrate);
1692 #endif
1693
1694 int xfrm_state_update(struct xfrm_state *x)
1695 {
1696         struct xfrm_state *x1, *to_put;
1697         int err;
1698         int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY);
1699         struct net *net = xs_net(x);
1700
1701         to_put = NULL;
1702
1703         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1704         x1 = __xfrm_state_locate(x, use_spi, x->props.family);
1705
1706         err = -ESRCH;
1707         if (!x1)
1708                 goto out;
1709
1710         if (xfrm_state_kern(x1)) {
1711                 to_put = x1;
1712                 err = -EEXIST;
1713                 goto out;
1714         }
1715
1716         if (x1->km.state == XFRM_STATE_ACQ) {
1717                 __xfrm_state_insert(x);
1718                 x = NULL;
1719         }
1720         err = 0;
1721
1722 out:
1723         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1724
1725         if (to_put)
1726                 xfrm_state_put(to_put);
1727
1728         if (err)
1729                 return err;
1730
1731         if (!x) {
1732                 xfrm_state_delete(x1);
1733                 xfrm_state_put(x1);
1734                 return 0;
1735         }
1736
1737         err = -EINVAL;
1738         spin_lock_bh(&x1->lock);
1739         if (likely(x1->km.state == XFRM_STATE_VALID)) {
1740                 if (x->encap && x1->encap &&
1741                     x->encap->encap_type == x1->encap->encap_type)
1742                         memcpy(x1->encap, x->encap, sizeof(*x1->encap));
1743                 else if (x->encap || x1->encap)
1744                         goto fail;
1745
1746                 if (x->coaddr && x1->coaddr) {
1747                         memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr));
1748                 }
1749                 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel)))
1750                         memcpy(&x1->sel, &x->sel, sizeof(x1->sel));
1751                 memcpy(&x1->lft, &x->lft, sizeof(x1->lft));
1752                 x1->km.dying = 0;
1753
1754                 hrtimer_start(&x1->mtimer, ktime_set(1, 0),
1755                               HRTIMER_MODE_REL_SOFT);
1756                 if (x1->curlft.use_time)
1757                         xfrm_state_check_expire(x1);
1758
1759                 if (x->props.smark.m || x->props.smark.v || x->if_id) {
1760                         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1761
1762                         if (x->props.smark.m || x->props.smark.v)
1763                                 x1->props.smark = x->props.smark;
1764
1765                         if (x->if_id)
1766                                 x1->if_id = x->if_id;
1767
1768                         __xfrm_state_bump_genids(x1);
1769                         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1770                 }
1771
1772                 err = 0;
1773                 x->km.state = XFRM_STATE_DEAD;
1774                 __xfrm_state_put(x);
1775         }
1776
1777 fail:
1778         spin_unlock_bh(&x1->lock);
1779
1780         xfrm_state_put(x1);
1781
1782         return err;
1783 }
1784 EXPORT_SYMBOL(xfrm_state_update);
1785
1786 int xfrm_state_check_expire(struct xfrm_state *x)
1787 {
1788         if (!x->curlft.use_time)
1789                 x->curlft.use_time = ktime_get_real_seconds();
1790
1791         if (x->curlft.bytes >= x->lft.hard_byte_limit ||
1792             x->curlft.packets >= x->lft.hard_packet_limit) {
1793                 x->km.state = XFRM_STATE_EXPIRED;
1794                 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT);
1795                 return -EINVAL;
1796         }
1797
1798         if (!x->km.dying &&
1799             (x->curlft.bytes >= x->lft.soft_byte_limit ||
1800              x->curlft.packets >= x->lft.soft_packet_limit)) {
1801                 x->km.dying = 1;
1802                 km_state_expired(x, 0, 0);
1803         }
1804         return 0;
1805 }
1806 EXPORT_SYMBOL(xfrm_state_check_expire);
1807
1808 struct xfrm_state *
1809 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi,
1810                   u8 proto, unsigned short family)
1811 {
1812         struct xfrm_state *x;
1813
1814         rcu_read_lock();
1815         x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family);
1816         rcu_read_unlock();
1817         return x;
1818 }
1819 EXPORT_SYMBOL(xfrm_state_lookup);
1820
1821 struct xfrm_state *
1822 xfrm_state_lookup_byaddr(struct net *net, u32 mark,
1823                          const xfrm_address_t *daddr, const xfrm_address_t *saddr,
1824                          u8 proto, unsigned short family)
1825 {
1826         struct xfrm_state *x;
1827
1828         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1829         x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family);
1830         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1831         return x;
1832 }
1833 EXPORT_SYMBOL(xfrm_state_lookup_byaddr);
1834
1835 struct xfrm_state *
1836 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid,
1837               u32 if_id, u8 proto, const xfrm_address_t *daddr,
1838               const xfrm_address_t *saddr, int create, unsigned short family)
1839 {
1840         struct xfrm_state *x;
1841
1842         spin_lock_bh(&net->xfrm.xfrm_state_lock);
1843         x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create);
1844         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
1845
1846         return x;
1847 }
1848 EXPORT_SYMBOL(xfrm_find_acq);
1849
1850 #ifdef CONFIG_XFRM_SUB_POLICY
1851 #if IS_ENABLED(CONFIG_IPV6)
1852 /* distribution counting sort function for xfrm_state and xfrm_tmpl */
1853 static void
1854 __xfrm6_sort(void **dst, void **src, int n,
1855              int (*cmp)(const void *p), int maxclass)
1856 {
1857         int count[XFRM_MAX_DEPTH] = { };
1858         int class[XFRM_MAX_DEPTH];
1859         int i;
1860
1861         for (i = 0; i < n; i++) {
1862                 int c = cmp(src[i]);
1863
1864                 class[i] = c;
1865                 count[c]++;
1866         }
1867
1868         for (i = 2; i < maxclass; i++)
1869                 count[i] += count[i - 1];
1870
1871         for (i = 0; i < n; i++) {
1872                 dst[count[class[i] - 1]++] = src[i];
1873                 src[i] = NULL;
1874         }
1875 }
1876
1877 /* Rule for xfrm_state:
1878  *
1879  * rule 1: select IPsec transport except AH
1880  * rule 2: select MIPv6 RO or inbound trigger
1881  * rule 3: select IPsec transport AH
1882  * rule 4: select IPsec tunnel
1883  * rule 5: others
1884  */
1885 static int __xfrm6_state_sort_cmp(const void *p)
1886 {
1887         const struct xfrm_state *v = p;
1888
1889         switch (v->props.mode) {
1890         case XFRM_MODE_TRANSPORT:
1891                 if (v->id.proto != IPPROTO_AH)
1892                         return 1;
1893                 else
1894                         return 3;
1895 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1896         case XFRM_MODE_ROUTEOPTIMIZATION:
1897         case XFRM_MODE_IN_TRIGGER:
1898                 return 2;
1899 #endif
1900         case XFRM_MODE_TUNNEL:
1901         case XFRM_MODE_BEET:
1902                 return 4;
1903         }
1904         return 5;
1905 }
1906
1907 /* Rule for xfrm_tmpl:
1908  *
1909  * rule 1: select IPsec transport
1910  * rule 2: select MIPv6 RO or inbound trigger
1911  * rule 3: select IPsec tunnel
1912  * rule 4: others
1913  */
1914 static int __xfrm6_tmpl_sort_cmp(const void *p)
1915 {
1916         const struct xfrm_tmpl *v = p;
1917
1918         switch (v->mode) {
1919         case XFRM_MODE_TRANSPORT:
1920                 return 1;
1921 #if IS_ENABLED(CONFIG_IPV6_MIP6)
1922         case XFRM_MODE_ROUTEOPTIMIZATION:
1923         case XFRM_MODE_IN_TRIGGER:
1924                 return 2;
1925 #endif
1926         case XFRM_MODE_TUNNEL:
1927         case XFRM_MODE_BEET:
1928                 return 3;
1929         }
1930         return 4;
1931 }
1932 #else
1933 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; }
1934 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; }
1935
1936 static inline void
1937 __xfrm6_sort(void **dst, void **src, int n,
1938              int (*cmp)(const void *p), int maxclass)
1939 {
1940         int i;
1941
1942         for (i = 0; i < n; i++)
1943                 dst[i] = src[i];
1944 }
1945 #endif /* CONFIG_IPV6 */
1946
1947 void
1948 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n,
1949                unsigned short family)
1950 {
1951         int i;
1952
1953         if (family == AF_INET6)
1954                 __xfrm6_sort((void **)dst, (void **)src, n,
1955                              __xfrm6_tmpl_sort_cmp, 5);
1956         else
1957                 for (i = 0; i < n; i++)
1958                         dst[i] = src[i];
1959 }
1960
1961 void
1962 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n,
1963                 unsigned short family)
1964 {
1965         int i;
1966
1967         if (family == AF_INET6)
1968                 __xfrm6_sort((void **)dst, (void **)src, n,
1969                              __xfrm6_state_sort_cmp, 6);
1970         else
1971                 for (i = 0; i < n; i++)
1972                         dst[i] = src[i];
1973 }
1974 #endif
1975
1976 /* Silly enough, but I'm lazy to build resolution list */
1977
1978 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1979 {
1980         unsigned int h = xfrm_seq_hash(net, seq);
1981         struct xfrm_state *x;
1982
1983         hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) {
1984                 if (x->km.seq == seq &&
1985                     (mark & x->mark.m) == x->mark.v &&
1986                     x->km.state == XFRM_STATE_ACQ) {
1987                         xfrm_state_hold(x);
1988                         return x;
1989                 }
1990         }
1991
1992         return NULL;
1993 }
1994
1995 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq)
1996 {
1997         struct xfrm_state *x;
1998
1999         spin_lock_bh(&net->xfrm.xfrm_state_lock);
2000         x = __xfrm_find_acq_byseq(net, mark, seq);
2001         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2002         return x;
2003 }
2004 EXPORT_SYMBOL(xfrm_find_acq_byseq);
2005
2006 u32 xfrm_get_acqseq(void)
2007 {
2008         u32 res;
2009         static atomic_t acqseq;
2010
2011         do {
2012                 res = atomic_inc_return(&acqseq);
2013         } while (!res);
2014
2015         return res;
2016 }
2017 EXPORT_SYMBOL(xfrm_get_acqseq);
2018
2019 int verify_spi_info(u8 proto, u32 min, u32 max)
2020 {
2021         switch (proto) {
2022         case IPPROTO_AH:
2023         case IPPROTO_ESP:
2024                 break;
2025
2026         case IPPROTO_COMP:
2027                 /* IPCOMP spi is 16-bits. */
2028                 if (max >= 0x10000)
2029                         return -EINVAL;
2030                 break;
2031
2032         default:
2033                 return -EINVAL;
2034         }
2035
2036         if (min > max)
2037                 return -EINVAL;
2038
2039         return 0;
2040 }
2041 EXPORT_SYMBOL(verify_spi_info);
2042
2043 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high)
2044 {
2045         struct net *net = xs_net(x);
2046         unsigned int h;
2047         struct xfrm_state *x0;
2048         int err = -ENOENT;
2049         __be32 minspi = htonl(low);
2050         __be32 maxspi = htonl(high);
2051         __be32 newspi = 0;
2052         u32 mark = x->mark.v & x->mark.m;
2053
2054         spin_lock_bh(&x->lock);
2055         if (x->km.state == XFRM_STATE_DEAD)
2056                 goto unlock;
2057
2058         err = 0;
2059         if (x->id.spi)
2060                 goto unlock;
2061
2062         err = -ENOENT;
2063
2064         if (minspi == maxspi) {
2065                 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family);
2066                 if (x0) {
2067                         xfrm_state_put(x0);
2068                         goto unlock;
2069                 }
2070                 newspi = minspi;
2071         } else {
2072                 u32 spi = 0;
2073                 for (h = 0; h < high-low+1; h++) {
2074                         spi = low + prandom_u32()%(high-low+1);
2075                         x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family);
2076                         if (x0 == NULL) {
2077                                 newspi = htonl(spi);
2078                                 break;
2079                         }
2080                         xfrm_state_put(x0);
2081                 }
2082         }
2083         if (newspi) {
2084                 spin_lock_bh(&net->xfrm.xfrm_state_lock);
2085                 x->id.spi = newspi;
2086                 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family);
2087                 hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h);
2088                 spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2089
2090                 err = 0;
2091         }
2092
2093 unlock:
2094         spin_unlock_bh(&x->lock);
2095
2096         return err;
2097 }
2098 EXPORT_SYMBOL(xfrm_alloc_spi);
2099
2100 static bool __xfrm_state_filter_match(struct xfrm_state *x,
2101                                       struct xfrm_address_filter *filter)
2102 {
2103         if (filter) {
2104                 if ((filter->family == AF_INET ||
2105                      filter->family == AF_INET6) &&
2106                     x->props.family != filter->family)
2107                         return false;
2108
2109                 return addr_match(&x->props.saddr, &filter->saddr,
2110                                   filter->splen) &&
2111                        addr_match(&x->id.daddr, &filter->daddr,
2112                                   filter->dplen);
2113         }
2114         return true;
2115 }
2116
2117 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk,
2118                     int (*func)(struct xfrm_state *, int, void*),
2119                     void *data)
2120 {
2121         struct xfrm_state *state;
2122         struct xfrm_state_walk *x;
2123         int err = 0;
2124
2125         if (walk->seq != 0 && list_empty(&walk->all))
2126                 return 0;
2127
2128         spin_lock_bh(&net->xfrm.xfrm_state_lock);
2129         if (list_empty(&walk->all))
2130                 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all);
2131         else
2132                 x = list_first_entry(&walk->all, struct xfrm_state_walk, all);
2133         list_for_each_entry_from(x, &net->xfrm.state_all, all) {
2134                 if (x->state == XFRM_STATE_DEAD)
2135                         continue;
2136                 state = container_of(x, struct xfrm_state, km);
2137                 if (!xfrm_id_proto_match(state->id.proto, walk->proto))
2138                         continue;
2139                 if (!__xfrm_state_filter_match(state, walk->filter))
2140                         continue;
2141                 err = func(state, walk->seq, data);
2142                 if (err) {
2143                         list_move_tail(&walk->all, &x->all);
2144                         goto out;
2145                 }
2146                 walk->seq++;
2147         }
2148         if (walk->seq == 0) {
2149                 err = -ENOENT;
2150                 goto out;
2151         }
2152         list_del_init(&walk->all);
2153 out:
2154         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2155         return err;
2156 }
2157 EXPORT_SYMBOL(xfrm_state_walk);
2158
2159 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto,
2160                           struct xfrm_address_filter *filter)
2161 {
2162         INIT_LIST_HEAD(&walk->all);
2163         walk->proto = proto;
2164         walk->state = XFRM_STATE_DEAD;
2165         walk->seq = 0;
2166         walk->filter = filter;
2167 }
2168 EXPORT_SYMBOL(xfrm_state_walk_init);
2169
2170 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net)
2171 {
2172         kfree(walk->filter);
2173
2174         if (list_empty(&walk->all))
2175                 return;
2176
2177         spin_lock_bh(&net->xfrm.xfrm_state_lock);
2178         list_del(&walk->all);
2179         spin_unlock_bh(&net->xfrm.xfrm_state_lock);
2180 }
2181 EXPORT_SYMBOL(xfrm_state_walk_done);
2182
2183 static void xfrm_replay_timer_handler(struct timer_list *t)
2184 {
2185         struct xfrm_state *x = from_timer(x, t, rtimer);
2186
2187         spin_lock(&x->lock);
2188
2189         if (x->km.state == XFRM_STATE_VALID) {
2190                 if (xfrm_aevent_is_on(xs_net(x)))
2191                         xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT);
2192                 else
2193                         x->xflags |= XFRM_TIME_DEFER;
2194         }
2195
2196         spin_unlock(&x->lock);
2197 }
2198
2199 static LIST_HEAD(xfrm_km_list);
2200
2201 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c)
2202 {
2203         struct xfrm_mgr *km;
2204
2205         rcu_read_lock();
2206         list_for_each_entry_rcu(km, &xfrm_km_list, list)
2207                 if (km->notify_policy)
2208                         km->notify_policy(xp, dir, c);
2209         rcu_read_unlock();
2210 }
2211
2212 void km_state_notify(struct xfrm_state *x, const struct km_event *c)
2213 {
2214         struct xfrm_mgr *km;
2215         rcu_read_lock();
2216         list_for_each_entry_rcu(km, &xfrm_km_list, list)
2217                 if (km->notify)
2218                         km->notify(x, c);
2219         rcu_read_unlock();
2220 }
2221
2222 EXPORT_SYMBOL(km_policy_notify);
2223 EXPORT_SYMBOL(km_state_notify);
2224
2225 void km_state_expired(struct xfrm_state *x, int hard, u32 portid)
2226 {
2227         struct km_event c;
2228
2229         c.data.hard = hard;
2230         c.portid = portid;
2231         c.event = XFRM_MSG_EXPIRE;
2232         km_state_notify(x, &c);
2233 }
2234
2235 EXPORT_SYMBOL(km_state_expired);
2236 /*
2237  * We send to all registered managers regardless of failure
2238  * We are happy with one success
2239 */
2240 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol)
2241 {
2242         int err = -EINVAL, acqret;
2243         struct xfrm_mgr *km;
2244
2245         rcu_read_lock();
2246         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2247                 acqret = km->acquire(x, t, pol);
2248                 if (!acqret)
2249                         err = acqret;
2250         }
2251         rcu_read_unlock();
2252         return err;
2253 }
2254 EXPORT_SYMBOL(km_query);
2255
2256 static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2257 {
2258         int err = -EINVAL;
2259         struct xfrm_mgr *km;
2260
2261         rcu_read_lock();
2262         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2263                 if (km->new_mapping)
2264                         err = km->new_mapping(x, ipaddr, sport);
2265                 if (!err)
2266                         break;
2267         }
2268         rcu_read_unlock();
2269         return err;
2270 }
2271
2272 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport)
2273 {
2274         int ret = 0;
2275
2276         if (x->mapping_maxage) {
2277                 if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage ||
2278                     x->new_mapping_sport != sport) {
2279                         x->new_mapping_sport = sport;
2280                         x->new_mapping = jiffies / HZ;
2281                         ret = __km_new_mapping(x, ipaddr, sport);
2282                 }
2283         } else {
2284                 ret = __km_new_mapping(x, ipaddr, sport);
2285         }
2286
2287         return ret;
2288 }
2289 EXPORT_SYMBOL(km_new_mapping);
2290
2291 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid)
2292 {
2293         struct km_event c;
2294
2295         c.data.hard = hard;
2296         c.portid = portid;
2297         c.event = XFRM_MSG_POLEXPIRE;
2298         km_policy_notify(pol, dir, &c);
2299 }
2300 EXPORT_SYMBOL(km_policy_expired);
2301
2302 #ifdef CONFIG_XFRM_MIGRATE
2303 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type,
2304                const struct xfrm_migrate *m, int num_migrate,
2305                const struct xfrm_kmaddress *k,
2306                const struct xfrm_encap_tmpl *encap)
2307 {
2308         int err = -EINVAL;
2309         int ret;
2310         struct xfrm_mgr *km;
2311
2312         rcu_read_lock();
2313         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2314                 if (km->migrate) {
2315                         ret = km->migrate(sel, dir, type, m, num_migrate, k,
2316                                           encap);
2317                         if (!ret)
2318                                 err = ret;
2319                 }
2320         }
2321         rcu_read_unlock();
2322         return err;
2323 }
2324 EXPORT_SYMBOL(km_migrate);
2325 #endif
2326
2327 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr)
2328 {
2329         int err = -EINVAL;
2330         int ret;
2331         struct xfrm_mgr *km;
2332
2333         rcu_read_lock();
2334         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2335                 if (km->report) {
2336                         ret = km->report(net, proto, sel, addr);
2337                         if (!ret)
2338                                 err = ret;
2339                 }
2340         }
2341         rcu_read_unlock();
2342         return err;
2343 }
2344 EXPORT_SYMBOL(km_report);
2345
2346 static bool km_is_alive(const struct km_event *c)
2347 {
2348         struct xfrm_mgr *km;
2349         bool is_alive = false;
2350
2351         rcu_read_lock();
2352         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2353                 if (km->is_alive && km->is_alive(c)) {
2354                         is_alive = true;
2355                         break;
2356                 }
2357         }
2358         rcu_read_unlock();
2359
2360         return is_alive;
2361 }
2362
2363 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT)
2364 static DEFINE_SPINLOCK(xfrm_translator_lock);
2365 static struct xfrm_translator __rcu *xfrm_translator;
2366
2367 struct xfrm_translator *xfrm_get_translator(void)
2368 {
2369         struct xfrm_translator *xtr;
2370
2371         rcu_read_lock();
2372         xtr = rcu_dereference(xfrm_translator);
2373         if (unlikely(!xtr))
2374                 goto out;
2375         if (!try_module_get(xtr->owner))
2376                 xtr = NULL;
2377 out:
2378         rcu_read_unlock();
2379         return xtr;
2380 }
2381 EXPORT_SYMBOL_GPL(xfrm_get_translator);
2382
2383 void xfrm_put_translator(struct xfrm_translator *xtr)
2384 {
2385         module_put(xtr->owner);
2386 }
2387 EXPORT_SYMBOL_GPL(xfrm_put_translator);
2388
2389 int xfrm_register_translator(struct xfrm_translator *xtr)
2390 {
2391         int err = 0;
2392
2393         spin_lock_bh(&xfrm_translator_lock);
2394         if (unlikely(xfrm_translator != NULL))
2395                 err = -EEXIST;
2396         else
2397                 rcu_assign_pointer(xfrm_translator, xtr);
2398         spin_unlock_bh(&xfrm_translator_lock);
2399
2400         return err;
2401 }
2402 EXPORT_SYMBOL_GPL(xfrm_register_translator);
2403
2404 int xfrm_unregister_translator(struct xfrm_translator *xtr)
2405 {
2406         int err = 0;
2407
2408         spin_lock_bh(&xfrm_translator_lock);
2409         if (likely(xfrm_translator != NULL)) {
2410                 if (rcu_access_pointer(xfrm_translator) != xtr)
2411                         err = -EINVAL;
2412                 else
2413                         RCU_INIT_POINTER(xfrm_translator, NULL);
2414         }
2415         spin_unlock_bh(&xfrm_translator_lock);
2416         synchronize_rcu();
2417
2418         return err;
2419 }
2420 EXPORT_SYMBOL_GPL(xfrm_unregister_translator);
2421 #endif
2422
2423 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen)
2424 {
2425         int err;
2426         u8 *data;
2427         struct xfrm_mgr *km;
2428         struct xfrm_policy *pol = NULL;
2429
2430         if (sockptr_is_null(optval) && !optlen) {
2431                 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL);
2432                 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL);
2433                 __sk_dst_reset(sk);
2434                 return 0;
2435         }
2436
2437         if (optlen <= 0 || optlen > PAGE_SIZE)
2438                 return -EMSGSIZE;
2439
2440         data = memdup_sockptr(optval, optlen);
2441         if (IS_ERR(data))
2442                 return PTR_ERR(data);
2443
2444         if (in_compat_syscall()) {
2445                 struct xfrm_translator *xtr = xfrm_get_translator();
2446
2447                 if (!xtr) {
2448                         kfree(data);
2449                         return -EOPNOTSUPP;
2450                 }
2451
2452                 err = xtr->xlate_user_policy_sockptr(&data, optlen);
2453                 xfrm_put_translator(xtr);
2454                 if (err) {
2455                         kfree(data);
2456                         return err;
2457                 }
2458         }
2459
2460         err = -EINVAL;
2461         rcu_read_lock();
2462         list_for_each_entry_rcu(km, &xfrm_km_list, list) {
2463                 pol = km->compile_policy(sk, optname, data,
2464                                          optlen, &err);
2465                 if (err >= 0)
2466                         break;
2467         }
2468         rcu_read_unlock();
2469
2470         if (err >= 0) {
2471                 xfrm_sk_policy_insert(sk, err, pol);
2472                 xfrm_pol_put(pol);
2473                 __sk_dst_reset(sk);
2474                 err = 0;
2475         }
2476
2477         kfree(data);
2478         return err;
2479 }
2480 EXPORT_SYMBOL(xfrm_user_policy);
2481
2482 static DEFINE_SPINLOCK(xfrm_km_lock);
2483
2484 int xfrm_register_km(struct xfrm_mgr *km)
2485 {
2486         spin_lock_bh(&xfrm_km_lock);
2487         list_add_tail_rcu(&km->list, &xfrm_km_list);
2488         spin_unlock_bh(&xfrm_km_lock);
2489         return 0;
2490 }
2491 EXPORT_SYMBOL(xfrm_register_km);
2492
2493 int xfrm_unregister_km(struct xfrm_mgr *km)
2494 {
2495         spin_lock_bh(&xfrm_km_lock);
2496         list_del_rcu(&km->list);
2497         spin_unlock_bh(&xfrm_km_lock);
2498         synchronize_rcu();
2499         return 0;
2500 }
2501 EXPORT_SYMBOL(xfrm_unregister_km);
2502
2503 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo)
2504 {
2505         int err = 0;
2506
2507         if (WARN_ON(afinfo->family >= NPROTO))
2508                 return -EAFNOSUPPORT;
2509
2510         spin_lock_bh(&xfrm_state_afinfo_lock);
2511         if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL))
2512                 err = -EEXIST;
2513         else
2514                 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo);
2515         spin_unlock_bh(&xfrm_state_afinfo_lock);
2516         return err;
2517 }
2518 EXPORT_SYMBOL(xfrm_state_register_afinfo);
2519
2520 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo)
2521 {
2522         int err = 0, family = afinfo->family;
2523
2524         if (WARN_ON(family >= NPROTO))
2525                 return -EAFNOSUPPORT;
2526
2527         spin_lock_bh(&xfrm_state_afinfo_lock);
2528         if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) {
2529                 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo)
2530                         err = -EINVAL;
2531                 else
2532                         RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL);
2533         }
2534         spin_unlock_bh(&xfrm_state_afinfo_lock);
2535         synchronize_rcu();
2536         return err;
2537 }
2538 EXPORT_SYMBOL(xfrm_state_unregister_afinfo);
2539
2540 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family)
2541 {
2542         if (unlikely(family >= NPROTO))
2543                 return NULL;
2544
2545         return rcu_dereference(xfrm_state_afinfo[family]);
2546 }
2547 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu);
2548
2549 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family)
2550 {
2551         struct xfrm_state_afinfo *afinfo;
2552         if (unlikely(family >= NPROTO))
2553                 return NULL;
2554         rcu_read_lock();
2555         afinfo = rcu_dereference(xfrm_state_afinfo[family]);
2556         if (unlikely(!afinfo))
2557                 rcu_read_unlock();
2558         return afinfo;
2559 }
2560
2561 void xfrm_flush_gc(void)
2562 {
2563         flush_work(&xfrm_state_gc_work);
2564 }
2565 EXPORT_SYMBOL(xfrm_flush_gc);
2566
2567 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */
2568 void xfrm_state_delete_tunnel(struct xfrm_state *x)
2569 {
2570         if (x->tunnel) {
2571                 struct xfrm_state *t = x->tunnel;
2572
2573                 if (atomic_read(&t->tunnel_users) == 2)
2574                         xfrm_state_delete(t);
2575                 atomic_dec(&t->tunnel_users);
2576                 xfrm_state_put_sync(t);
2577                 x->tunnel = NULL;
2578         }
2579 }
2580 EXPORT_SYMBOL(xfrm_state_delete_tunnel);
2581
2582 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu)
2583 {
2584         const struct xfrm_type *type = READ_ONCE(x->type);
2585         struct crypto_aead *aead;
2586         u32 blksize, net_adj = 0;
2587
2588         if (x->km.state != XFRM_STATE_VALID ||
2589             !type || type->proto != IPPROTO_ESP)
2590                 return mtu - x->props.header_len;
2591
2592         aead = x->data;
2593         blksize = ALIGN(crypto_aead_blocksize(aead), 4);
2594
2595         switch (x->props.mode) {
2596         case XFRM_MODE_TRANSPORT:
2597         case XFRM_MODE_BEET:
2598                 if (x->props.family == AF_INET)
2599                         net_adj = sizeof(struct iphdr);
2600                 else if (x->props.family == AF_INET6)
2601                         net_adj = sizeof(struct ipv6hdr);
2602                 break;
2603         case XFRM_MODE_TUNNEL:
2604                 break;
2605         default:
2606                 WARN_ON_ONCE(1);
2607                 break;
2608         }
2609
2610         return ((mtu - x->props.header_len - crypto_aead_authsize(aead) -
2611                  net_adj) & ~(blksize - 1)) + net_adj - 2;
2612 }
2613 EXPORT_SYMBOL_GPL(xfrm_state_mtu);
2614
2615 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload)
2616 {
2617         const struct xfrm_mode *inner_mode;
2618         const struct xfrm_mode *outer_mode;
2619         int family = x->props.family;
2620         int err;
2621
2622         if (family == AF_INET &&
2623             READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc))
2624                 x->props.flags |= XFRM_STATE_NOPMTUDISC;
2625
2626         err = -EPROTONOSUPPORT;
2627
2628         if (x->sel.family != AF_UNSPEC) {
2629                 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family);
2630                 if (inner_mode == NULL)
2631                         goto error;
2632
2633                 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) &&
2634                     family != x->sel.family)
2635                         goto error;
2636
2637                 x->inner_mode = *inner_mode;
2638         } else {
2639                 const struct xfrm_mode *inner_mode_iaf;
2640                 int iafamily = AF_INET;
2641
2642                 inner_mode = xfrm_get_mode(x->props.mode, x->props.family);
2643                 if (inner_mode == NULL)
2644                         goto error;
2645
2646                 x->inner_mode = *inner_mode;
2647
2648                 if (x->props.family == AF_INET)
2649                         iafamily = AF_INET6;
2650
2651                 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily);
2652                 if (inner_mode_iaf) {
2653                         if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL)
2654                                 x->inner_mode_iaf = *inner_mode_iaf;
2655                 }
2656         }
2657
2658         x->type = xfrm_get_type(x->id.proto, family);
2659         if (x->type == NULL)
2660                 goto error;
2661
2662         x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload);
2663
2664         err = x->type->init_state(x);
2665         if (err)
2666                 goto error;
2667
2668         outer_mode = xfrm_get_mode(x->props.mode, family);
2669         if (!outer_mode) {
2670                 err = -EPROTONOSUPPORT;
2671                 goto error;
2672         }
2673
2674         x->outer_mode = *outer_mode;
2675         if (init_replay) {
2676                 err = xfrm_init_replay(x);
2677                 if (err)
2678                         goto error;
2679         }
2680
2681 error:
2682         return err;
2683 }
2684
2685 EXPORT_SYMBOL(__xfrm_init_state);
2686
2687 int xfrm_init_state(struct xfrm_state *x)
2688 {
2689         int err;
2690
2691         err = __xfrm_init_state(x, true, false);
2692         if (!err)
2693                 x->km.state = XFRM_STATE_VALID;
2694
2695         return err;
2696 }
2697
2698 EXPORT_SYMBOL(xfrm_init_state);
2699
2700 int __net_init xfrm_state_init(struct net *net)
2701 {
2702         unsigned int sz;
2703
2704         if (net_eq(net, &init_net))
2705                 xfrm_state_cache = KMEM_CACHE(xfrm_state,
2706                                               SLAB_HWCACHE_ALIGN | SLAB_PANIC);
2707
2708         INIT_LIST_HEAD(&net->xfrm.state_all);
2709
2710         sz = sizeof(struct hlist_head) * 8;
2711
2712         net->xfrm.state_bydst = xfrm_hash_alloc(sz);
2713         if (!net->xfrm.state_bydst)
2714                 goto out_bydst;
2715         net->xfrm.state_bysrc = xfrm_hash_alloc(sz);
2716         if (!net->xfrm.state_bysrc)
2717                 goto out_bysrc;
2718         net->xfrm.state_byspi = xfrm_hash_alloc(sz);
2719         if (!net->xfrm.state_byspi)
2720                 goto out_byspi;
2721         net->xfrm.state_byseq = xfrm_hash_alloc(sz);
2722         if (!net->xfrm.state_byseq)
2723                 goto out_byseq;
2724         net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1);
2725
2726         net->xfrm.state_num = 0;
2727         INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize);
2728         spin_lock_init(&net->xfrm.xfrm_state_lock);
2729         seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation,
2730                                &net->xfrm.xfrm_state_lock);
2731         return 0;
2732
2733 out_byseq:
2734         xfrm_hash_free(net->xfrm.state_byspi, sz);
2735 out_byspi:
2736         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2737 out_bysrc:
2738         xfrm_hash_free(net->xfrm.state_bydst, sz);
2739 out_bydst:
2740         return -ENOMEM;
2741 }
2742
2743 void xfrm_state_fini(struct net *net)
2744 {
2745         unsigned int sz;
2746
2747         flush_work(&net->xfrm.state_hash_work);
2748         flush_work(&xfrm_state_gc_work);
2749         xfrm_state_flush(net, 0, false, true);
2750
2751         WARN_ON(!list_empty(&net->xfrm.state_all));
2752
2753         sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head);
2754         WARN_ON(!hlist_empty(net->xfrm.state_byseq));
2755         xfrm_hash_free(net->xfrm.state_byseq, sz);
2756         WARN_ON(!hlist_empty(net->xfrm.state_byspi));
2757         xfrm_hash_free(net->xfrm.state_byspi, sz);
2758         WARN_ON(!hlist_empty(net->xfrm.state_bysrc));
2759         xfrm_hash_free(net->xfrm.state_bysrc, sz);
2760         WARN_ON(!hlist_empty(net->xfrm.state_bydst));
2761         xfrm_hash_free(net->xfrm.state_bydst, sz);
2762 }
2763
2764 #ifdef CONFIG_AUDITSYSCALL
2765 static void xfrm_audit_helper_sainfo(struct xfrm_state *x,
2766                                      struct audit_buffer *audit_buf)
2767 {
2768         struct xfrm_sec_ctx *ctx = x->security;
2769         u32 spi = ntohl(x->id.spi);
2770
2771         if (ctx)
2772                 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s",
2773                                  ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str);
2774
2775         switch (x->props.family) {
2776         case AF_INET:
2777                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2778                                  &x->props.saddr.a4, &x->id.daddr.a4);
2779                 break;
2780         case AF_INET6:
2781                 audit_log_format(audit_buf, " src=%pI6 dst=%pI6",
2782                                  x->props.saddr.a6, x->id.daddr.a6);
2783                 break;
2784         }
2785
2786         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2787 }
2788
2789 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family,
2790                                       struct audit_buffer *audit_buf)
2791 {
2792         const struct iphdr *iph4;
2793         const struct ipv6hdr *iph6;
2794
2795         switch (family) {
2796         case AF_INET:
2797                 iph4 = ip_hdr(skb);
2798                 audit_log_format(audit_buf, " src=%pI4 dst=%pI4",
2799                                  &iph4->saddr, &iph4->daddr);
2800                 break;
2801         case AF_INET6:
2802                 iph6 = ipv6_hdr(skb);
2803                 audit_log_format(audit_buf,
2804                                  " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x",
2805                                  &iph6->saddr, &iph6->daddr,
2806                                  iph6->flow_lbl[0] & 0x0f,
2807                                  iph6->flow_lbl[1],
2808                                  iph6->flow_lbl[2]);
2809                 break;
2810         }
2811 }
2812
2813 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid)
2814 {
2815         struct audit_buffer *audit_buf;
2816
2817         audit_buf = xfrm_audit_start("SAD-add");
2818         if (audit_buf == NULL)
2819                 return;
2820         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2821         xfrm_audit_helper_sainfo(x, audit_buf);
2822         audit_log_format(audit_buf, " res=%u", result);
2823         audit_log_end(audit_buf);
2824 }
2825 EXPORT_SYMBOL_GPL(xfrm_audit_state_add);
2826
2827 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid)
2828 {
2829         struct audit_buffer *audit_buf;
2830
2831         audit_buf = xfrm_audit_start("SAD-delete");
2832         if (audit_buf == NULL)
2833                 return;
2834         xfrm_audit_helper_usrinfo(task_valid, audit_buf);
2835         xfrm_audit_helper_sainfo(x, audit_buf);
2836         audit_log_format(audit_buf, " res=%u", result);
2837         audit_log_end(audit_buf);
2838 }
2839 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete);
2840
2841 void xfrm_audit_state_replay_overflow(struct xfrm_state *x,
2842                                       struct sk_buff *skb)
2843 {
2844         struct audit_buffer *audit_buf;
2845         u32 spi;
2846
2847         audit_buf = xfrm_audit_start("SA-replay-overflow");
2848         if (audit_buf == NULL)
2849                 return;
2850         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2851         /* don't record the sequence number because it's inherent in this kind
2852          * of audit message */
2853         spi = ntohl(x->id.spi);
2854         audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi);
2855         audit_log_end(audit_buf);
2856 }
2857 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow);
2858
2859 void xfrm_audit_state_replay(struct xfrm_state *x,
2860                              struct sk_buff *skb, __be32 net_seq)
2861 {
2862         struct audit_buffer *audit_buf;
2863         u32 spi;
2864
2865         audit_buf = xfrm_audit_start("SA-replayed-pkt");
2866         if (audit_buf == NULL)
2867                 return;
2868         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2869         spi = ntohl(x->id.spi);
2870         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2871                          spi, spi, ntohl(net_seq));
2872         audit_log_end(audit_buf);
2873 }
2874 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay);
2875
2876 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family)
2877 {
2878         struct audit_buffer *audit_buf;
2879
2880         audit_buf = xfrm_audit_start("SA-notfound");
2881         if (audit_buf == NULL)
2882                 return;
2883         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2884         audit_log_end(audit_buf);
2885 }
2886 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple);
2887
2888 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family,
2889                                __be32 net_spi, __be32 net_seq)
2890 {
2891         struct audit_buffer *audit_buf;
2892         u32 spi;
2893
2894         audit_buf = xfrm_audit_start("SA-notfound");
2895         if (audit_buf == NULL)
2896                 return;
2897         xfrm_audit_helper_pktinfo(skb, family, audit_buf);
2898         spi = ntohl(net_spi);
2899         audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2900                          spi, spi, ntohl(net_seq));
2901         audit_log_end(audit_buf);
2902 }
2903 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound);
2904
2905 void xfrm_audit_state_icvfail(struct xfrm_state *x,
2906                               struct sk_buff *skb, u8 proto)
2907 {
2908         struct audit_buffer *audit_buf;
2909         __be32 net_spi;
2910         __be32 net_seq;
2911
2912         audit_buf = xfrm_audit_start("SA-icv-failure");
2913         if (audit_buf == NULL)
2914                 return;
2915         xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf);
2916         if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) {
2917                 u32 spi = ntohl(net_spi);
2918                 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u",
2919                                  spi, spi, ntohl(net_seq));
2920         }
2921         audit_log_end(audit_buf);
2922 }
2923 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail);
2924 #endif /* CONFIG_AUDITSYSCALL */