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