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