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