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/compat.h> 18 #include <linux/workqueue.h> 19 #include <net/xfrm.h> 20 #include <linux/pfkeyv2.h> 21 #include <linux/ipsec.h> 22 #include <linux/module.h> 23 #include <linux/cache.h> 24 #include <linux/audit.h> 25 #include <linux/uaccess.h> 26 #include <linux/ktime.h> 27 #include <linux/slab.h> 28 #include <linux/interrupt.h> 29 #include <linux/kernel.h> 30 31 #include <crypto/aead.h> 32 33 #include "xfrm_hash.h" 34 35 #define xfrm_state_deref_prot(table, net) \ 36 rcu_dereference_protected((table), lockdep_is_held(&(net)->xfrm.xfrm_state_lock)) 37 38 static void xfrm_state_gc_task(struct work_struct *work); 39 40 /* Each xfrm_state may be linked to two tables: 41 42 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl) 43 2. Hash table by (daddr,family,reqid) to find what SAs exist for given 44 destination/tunnel endpoint. (output) 45 */ 46 47 static unsigned int xfrm_state_hashmax __read_mostly = 1 * 1024 * 1024; 48 static struct kmem_cache *xfrm_state_cache __ro_after_init; 49 50 static DECLARE_WORK(xfrm_state_gc_work, xfrm_state_gc_task); 51 static HLIST_HEAD(xfrm_state_gc_list); 52 53 static inline bool xfrm_state_hold_rcu(struct xfrm_state __rcu *x) 54 { 55 return refcount_inc_not_zero(&x->refcnt); 56 } 57 58 static inline unsigned int xfrm_dst_hash(struct net *net, 59 const xfrm_address_t *daddr, 60 const xfrm_address_t *saddr, 61 u32 reqid, 62 unsigned short family) 63 { 64 return __xfrm_dst_hash(daddr, saddr, reqid, family, net->xfrm.state_hmask); 65 } 66 67 static inline unsigned int xfrm_src_hash(struct net *net, 68 const xfrm_address_t *daddr, 69 const xfrm_address_t *saddr, 70 unsigned short family) 71 { 72 return __xfrm_src_hash(daddr, saddr, family, net->xfrm.state_hmask); 73 } 74 75 static inline unsigned int 76 xfrm_spi_hash(struct net *net, const xfrm_address_t *daddr, 77 __be32 spi, u8 proto, unsigned short family) 78 { 79 return __xfrm_spi_hash(daddr, spi, proto, family, net->xfrm.state_hmask); 80 } 81 82 static unsigned int xfrm_seq_hash(struct net *net, u32 seq) 83 { 84 return __xfrm_seq_hash(seq, net->xfrm.state_hmask); 85 } 86 87 #define XFRM_STATE_INSERT(by, _n, _h, _type) \ 88 { \ 89 struct xfrm_state *_x = NULL; \ 90 \ 91 if (_type != XFRM_DEV_OFFLOAD_PACKET) { \ 92 hlist_for_each_entry_rcu(_x, _h, by) { \ 93 if (_x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \ 94 continue; \ 95 break; \ 96 } \ 97 } \ 98 \ 99 if (!_x || _x->xso.type == XFRM_DEV_OFFLOAD_PACKET) \ 100 /* SAD is empty or consist from HW SAs only */ \ 101 hlist_add_head_rcu(_n, _h); \ 102 else \ 103 hlist_add_before_rcu(_n, &_x->by); \ 104 } 105 106 static void xfrm_hash_transfer(struct hlist_head *list, 107 struct hlist_head *ndsttable, 108 struct hlist_head *nsrctable, 109 struct hlist_head *nspitable, 110 struct hlist_head *nseqtable, 111 unsigned int nhashmask) 112 { 113 struct hlist_node *tmp; 114 struct xfrm_state *x; 115 116 hlist_for_each_entry_safe(x, tmp, list, bydst) { 117 unsigned int h; 118 119 h = __xfrm_dst_hash(&x->id.daddr, &x->props.saddr, 120 x->props.reqid, x->props.family, 121 nhashmask); 122 XFRM_STATE_INSERT(bydst, &x->bydst, ndsttable + h, x->xso.type); 123 124 h = __xfrm_src_hash(&x->id.daddr, &x->props.saddr, 125 x->props.family, 126 nhashmask); 127 XFRM_STATE_INSERT(bysrc, &x->bysrc, nsrctable + h, x->xso.type); 128 129 if (x->id.spi) { 130 h = __xfrm_spi_hash(&x->id.daddr, x->id.spi, 131 x->id.proto, x->props.family, 132 nhashmask); 133 XFRM_STATE_INSERT(byspi, &x->byspi, nspitable + h, 134 x->xso.type); 135 } 136 137 if (x->km.seq) { 138 h = __xfrm_seq_hash(x->km.seq, nhashmask); 139 XFRM_STATE_INSERT(byseq, &x->byseq, nseqtable + h, 140 x->xso.type); 141 } 142 } 143 } 144 145 static unsigned long xfrm_hash_new_size(unsigned int state_hmask) 146 { 147 return ((state_hmask + 1) << 1) * sizeof(struct hlist_head); 148 } 149 150 static void xfrm_hash_resize(struct work_struct *work) 151 { 152 struct net *net = container_of(work, struct net, xfrm.state_hash_work); 153 struct hlist_head *ndst, *nsrc, *nspi, *nseq, *odst, *osrc, *ospi, *oseq; 154 unsigned long nsize, osize; 155 unsigned int nhashmask, ohashmask; 156 int i; 157 158 nsize = xfrm_hash_new_size(net->xfrm.state_hmask); 159 ndst = xfrm_hash_alloc(nsize); 160 if (!ndst) 161 return; 162 nsrc = xfrm_hash_alloc(nsize); 163 if (!nsrc) { 164 xfrm_hash_free(ndst, nsize); 165 return; 166 } 167 nspi = xfrm_hash_alloc(nsize); 168 if (!nspi) { 169 xfrm_hash_free(ndst, nsize); 170 xfrm_hash_free(nsrc, nsize); 171 return; 172 } 173 nseq = xfrm_hash_alloc(nsize); 174 if (!nseq) { 175 xfrm_hash_free(ndst, nsize); 176 xfrm_hash_free(nsrc, nsize); 177 xfrm_hash_free(nspi, nsize); 178 return; 179 } 180 181 spin_lock_bh(&net->xfrm.xfrm_state_lock); 182 write_seqcount_begin(&net->xfrm.xfrm_state_hash_generation); 183 184 nhashmask = (nsize / sizeof(struct hlist_head)) - 1U; 185 odst = xfrm_state_deref_prot(net->xfrm.state_bydst, net); 186 for (i = net->xfrm.state_hmask; i >= 0; i--) 187 xfrm_hash_transfer(odst + i, ndst, nsrc, nspi, nseq, nhashmask); 188 189 osrc = xfrm_state_deref_prot(net->xfrm.state_bysrc, net); 190 ospi = xfrm_state_deref_prot(net->xfrm.state_byspi, net); 191 oseq = xfrm_state_deref_prot(net->xfrm.state_byseq, net); 192 ohashmask = net->xfrm.state_hmask; 193 194 rcu_assign_pointer(net->xfrm.state_bydst, ndst); 195 rcu_assign_pointer(net->xfrm.state_bysrc, nsrc); 196 rcu_assign_pointer(net->xfrm.state_byspi, nspi); 197 rcu_assign_pointer(net->xfrm.state_byseq, nseq); 198 net->xfrm.state_hmask = nhashmask; 199 200 write_seqcount_end(&net->xfrm.xfrm_state_hash_generation); 201 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 202 203 osize = (ohashmask + 1) * sizeof(struct hlist_head); 204 205 synchronize_rcu(); 206 207 xfrm_hash_free(odst, osize); 208 xfrm_hash_free(osrc, osize); 209 xfrm_hash_free(ospi, osize); 210 xfrm_hash_free(oseq, osize); 211 } 212 213 static DEFINE_SPINLOCK(xfrm_state_afinfo_lock); 214 static struct xfrm_state_afinfo __rcu *xfrm_state_afinfo[NPROTO]; 215 216 static DEFINE_SPINLOCK(xfrm_state_gc_lock); 217 218 int __xfrm_state_delete(struct xfrm_state *x); 219 220 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol); 221 static bool km_is_alive(const struct km_event *c); 222 void km_state_expired(struct xfrm_state *x, int hard, u32 portid); 223 224 int xfrm_register_type(const struct xfrm_type *type, unsigned short family) 225 { 226 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family); 227 int err = 0; 228 229 if (!afinfo) 230 return -EAFNOSUPPORT; 231 232 #define X(afi, T, name) do { \ 233 WARN_ON((afi)->type_ ## name); \ 234 (afi)->type_ ## name = (T); \ 235 } while (0) 236 237 switch (type->proto) { 238 case IPPROTO_COMP: 239 X(afinfo, type, comp); 240 break; 241 case IPPROTO_AH: 242 X(afinfo, type, ah); 243 break; 244 case IPPROTO_ESP: 245 X(afinfo, type, esp); 246 break; 247 case IPPROTO_IPIP: 248 X(afinfo, type, ipip); 249 break; 250 case IPPROTO_DSTOPTS: 251 X(afinfo, type, dstopts); 252 break; 253 case IPPROTO_ROUTING: 254 X(afinfo, type, routing); 255 break; 256 case IPPROTO_IPV6: 257 X(afinfo, type, ipip6); 258 break; 259 default: 260 WARN_ON(1); 261 err = -EPROTONOSUPPORT; 262 break; 263 } 264 #undef X 265 rcu_read_unlock(); 266 return err; 267 } 268 EXPORT_SYMBOL(xfrm_register_type); 269 270 void xfrm_unregister_type(const struct xfrm_type *type, unsigned short family) 271 { 272 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family); 273 274 if (unlikely(afinfo == NULL)) 275 return; 276 277 #define X(afi, T, name) do { \ 278 WARN_ON((afi)->type_ ## name != (T)); \ 279 (afi)->type_ ## name = NULL; \ 280 } while (0) 281 282 switch (type->proto) { 283 case IPPROTO_COMP: 284 X(afinfo, type, comp); 285 break; 286 case IPPROTO_AH: 287 X(afinfo, type, ah); 288 break; 289 case IPPROTO_ESP: 290 X(afinfo, type, esp); 291 break; 292 case IPPROTO_IPIP: 293 X(afinfo, type, ipip); 294 break; 295 case IPPROTO_DSTOPTS: 296 X(afinfo, type, dstopts); 297 break; 298 case IPPROTO_ROUTING: 299 X(afinfo, type, routing); 300 break; 301 case IPPROTO_IPV6: 302 X(afinfo, type, ipip6); 303 break; 304 default: 305 WARN_ON(1); 306 break; 307 } 308 #undef X 309 rcu_read_unlock(); 310 } 311 EXPORT_SYMBOL(xfrm_unregister_type); 312 313 static const struct xfrm_type *xfrm_get_type(u8 proto, unsigned short family) 314 { 315 const struct xfrm_type *type = NULL; 316 struct xfrm_state_afinfo *afinfo; 317 int modload_attempted = 0; 318 319 retry: 320 afinfo = xfrm_state_get_afinfo(family); 321 if (unlikely(afinfo == NULL)) 322 return NULL; 323 324 switch (proto) { 325 case IPPROTO_COMP: 326 type = afinfo->type_comp; 327 break; 328 case IPPROTO_AH: 329 type = afinfo->type_ah; 330 break; 331 case IPPROTO_ESP: 332 type = afinfo->type_esp; 333 break; 334 case IPPROTO_IPIP: 335 type = afinfo->type_ipip; 336 break; 337 case IPPROTO_DSTOPTS: 338 type = afinfo->type_dstopts; 339 break; 340 case IPPROTO_ROUTING: 341 type = afinfo->type_routing; 342 break; 343 case IPPROTO_IPV6: 344 type = afinfo->type_ipip6; 345 break; 346 default: 347 break; 348 } 349 350 if (unlikely(type && !try_module_get(type->owner))) 351 type = NULL; 352 353 rcu_read_unlock(); 354 355 if (!type && !modload_attempted) { 356 request_module("xfrm-type-%d-%d", family, proto); 357 modload_attempted = 1; 358 goto retry; 359 } 360 361 return type; 362 } 363 364 static void xfrm_put_type(const struct xfrm_type *type) 365 { 366 module_put(type->owner); 367 } 368 369 int xfrm_register_type_offload(const struct xfrm_type_offload *type, 370 unsigned short family) 371 { 372 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family); 373 int err = 0; 374 375 if (unlikely(afinfo == NULL)) 376 return -EAFNOSUPPORT; 377 378 switch (type->proto) { 379 case IPPROTO_ESP: 380 WARN_ON(afinfo->type_offload_esp); 381 afinfo->type_offload_esp = type; 382 break; 383 default: 384 WARN_ON(1); 385 err = -EPROTONOSUPPORT; 386 break; 387 } 388 389 rcu_read_unlock(); 390 return err; 391 } 392 EXPORT_SYMBOL(xfrm_register_type_offload); 393 394 void xfrm_unregister_type_offload(const struct xfrm_type_offload *type, 395 unsigned short family) 396 { 397 struct xfrm_state_afinfo *afinfo = xfrm_state_get_afinfo(family); 398 399 if (unlikely(afinfo == NULL)) 400 return; 401 402 switch (type->proto) { 403 case IPPROTO_ESP: 404 WARN_ON(afinfo->type_offload_esp != type); 405 afinfo->type_offload_esp = NULL; 406 break; 407 default: 408 WARN_ON(1); 409 break; 410 } 411 rcu_read_unlock(); 412 } 413 EXPORT_SYMBOL(xfrm_unregister_type_offload); 414 415 static const struct xfrm_type_offload * 416 xfrm_get_type_offload(u8 proto, unsigned short family, bool try_load) 417 { 418 const struct xfrm_type_offload *type = NULL; 419 struct xfrm_state_afinfo *afinfo; 420 421 retry: 422 afinfo = xfrm_state_get_afinfo(family); 423 if (unlikely(afinfo == NULL)) 424 return NULL; 425 426 switch (proto) { 427 case IPPROTO_ESP: 428 type = afinfo->type_offload_esp; 429 break; 430 default: 431 break; 432 } 433 434 if ((type && !try_module_get(type->owner))) 435 type = NULL; 436 437 rcu_read_unlock(); 438 439 if (!type && try_load) { 440 request_module("xfrm-offload-%d-%d", family, proto); 441 try_load = false; 442 goto retry; 443 } 444 445 return type; 446 } 447 448 static void xfrm_put_type_offload(const struct xfrm_type_offload *type) 449 { 450 module_put(type->owner); 451 } 452 453 static const struct xfrm_mode xfrm4_mode_map[XFRM_MODE_MAX] = { 454 [XFRM_MODE_BEET] = { 455 .encap = XFRM_MODE_BEET, 456 .flags = XFRM_MODE_FLAG_TUNNEL, 457 .family = AF_INET, 458 }, 459 [XFRM_MODE_TRANSPORT] = { 460 .encap = XFRM_MODE_TRANSPORT, 461 .family = AF_INET, 462 }, 463 [XFRM_MODE_TUNNEL] = { 464 .encap = XFRM_MODE_TUNNEL, 465 .flags = XFRM_MODE_FLAG_TUNNEL, 466 .family = AF_INET, 467 }, 468 }; 469 470 static const struct xfrm_mode xfrm6_mode_map[XFRM_MODE_MAX] = { 471 [XFRM_MODE_BEET] = { 472 .encap = XFRM_MODE_BEET, 473 .flags = XFRM_MODE_FLAG_TUNNEL, 474 .family = AF_INET6, 475 }, 476 [XFRM_MODE_ROUTEOPTIMIZATION] = { 477 .encap = XFRM_MODE_ROUTEOPTIMIZATION, 478 .family = AF_INET6, 479 }, 480 [XFRM_MODE_TRANSPORT] = { 481 .encap = XFRM_MODE_TRANSPORT, 482 .family = AF_INET6, 483 }, 484 [XFRM_MODE_TUNNEL] = { 485 .encap = XFRM_MODE_TUNNEL, 486 .flags = XFRM_MODE_FLAG_TUNNEL, 487 .family = AF_INET6, 488 }, 489 }; 490 491 static const struct xfrm_mode *xfrm_get_mode(unsigned int encap, int family) 492 { 493 const struct xfrm_mode *mode; 494 495 if (unlikely(encap >= XFRM_MODE_MAX)) 496 return NULL; 497 498 switch (family) { 499 case AF_INET: 500 mode = &xfrm4_mode_map[encap]; 501 if (mode->family == family) 502 return mode; 503 break; 504 case AF_INET6: 505 mode = &xfrm6_mode_map[encap]; 506 if (mode->family == family) 507 return mode; 508 break; 509 default: 510 break; 511 } 512 513 return NULL; 514 } 515 516 void xfrm_state_free(struct xfrm_state *x) 517 { 518 kmem_cache_free(xfrm_state_cache, x); 519 } 520 EXPORT_SYMBOL(xfrm_state_free); 521 522 static void ___xfrm_state_destroy(struct xfrm_state *x) 523 { 524 hrtimer_cancel(&x->mtimer); 525 del_timer_sync(&x->rtimer); 526 kfree(x->aead); 527 kfree(x->aalg); 528 kfree(x->ealg); 529 kfree(x->calg); 530 kfree(x->encap); 531 kfree(x->coaddr); 532 kfree(x->replay_esn); 533 kfree(x->preplay_esn); 534 if (x->type_offload) 535 xfrm_put_type_offload(x->type_offload); 536 if (x->type) { 537 x->type->destructor(x); 538 xfrm_put_type(x->type); 539 } 540 if (x->xfrag.page) 541 put_page(x->xfrag.page); 542 xfrm_dev_state_free(x); 543 security_xfrm_state_free(x); 544 xfrm_state_free(x); 545 } 546 547 static void xfrm_state_gc_task(struct work_struct *work) 548 { 549 struct xfrm_state *x; 550 struct hlist_node *tmp; 551 struct hlist_head gc_list; 552 553 spin_lock_bh(&xfrm_state_gc_lock); 554 hlist_move_list(&xfrm_state_gc_list, &gc_list); 555 spin_unlock_bh(&xfrm_state_gc_lock); 556 557 synchronize_rcu(); 558 559 hlist_for_each_entry_safe(x, tmp, &gc_list, gclist) 560 ___xfrm_state_destroy(x); 561 } 562 563 static enum hrtimer_restart xfrm_timer_handler(struct hrtimer *me) 564 { 565 struct xfrm_state *x = container_of(me, struct xfrm_state, mtimer); 566 enum hrtimer_restart ret = HRTIMER_NORESTART; 567 time64_t now = ktime_get_real_seconds(); 568 time64_t next = TIME64_MAX; 569 int warn = 0; 570 int err = 0; 571 572 spin_lock(&x->lock); 573 xfrm_dev_state_update_curlft(x); 574 575 if (x->km.state == XFRM_STATE_DEAD) 576 goto out; 577 if (x->km.state == XFRM_STATE_EXPIRED) 578 goto expired; 579 if (x->lft.hard_add_expires_seconds) { 580 time64_t tmo = x->lft.hard_add_expires_seconds + 581 x->curlft.add_time - now; 582 if (tmo <= 0) { 583 if (x->xflags & XFRM_SOFT_EXPIRE) { 584 /* enter hard expire without soft expire first?! 585 * setting a new date could trigger this. 586 * workaround: fix x->curflt.add_time by below: 587 */ 588 x->curlft.add_time = now - x->saved_tmo - 1; 589 tmo = x->lft.hard_add_expires_seconds - x->saved_tmo; 590 } else 591 goto expired; 592 } 593 if (tmo < next) 594 next = tmo; 595 } 596 if (x->lft.hard_use_expires_seconds) { 597 time64_t tmo = x->lft.hard_use_expires_seconds + 598 (READ_ONCE(x->curlft.use_time) ? : now) - now; 599 if (tmo <= 0) 600 goto expired; 601 if (tmo < next) 602 next = tmo; 603 } 604 if (x->km.dying) 605 goto resched; 606 if (x->lft.soft_add_expires_seconds) { 607 time64_t tmo = x->lft.soft_add_expires_seconds + 608 x->curlft.add_time - now; 609 if (tmo <= 0) { 610 warn = 1; 611 x->xflags &= ~XFRM_SOFT_EXPIRE; 612 } else if (tmo < next) { 613 next = tmo; 614 x->xflags |= XFRM_SOFT_EXPIRE; 615 x->saved_tmo = tmo; 616 } 617 } 618 if (x->lft.soft_use_expires_seconds) { 619 time64_t tmo = x->lft.soft_use_expires_seconds + 620 (READ_ONCE(x->curlft.use_time) ? : now) - now; 621 if (tmo <= 0) 622 warn = 1; 623 else if (tmo < next) 624 next = tmo; 625 } 626 627 x->km.dying = warn; 628 if (warn) 629 km_state_expired(x, 0, 0); 630 resched: 631 if (next != TIME64_MAX) { 632 hrtimer_forward_now(&x->mtimer, ktime_set(next, 0)); 633 ret = HRTIMER_RESTART; 634 } 635 636 goto out; 637 638 expired: 639 if (x->km.state == XFRM_STATE_ACQ && x->id.spi == 0) 640 x->km.state = XFRM_STATE_EXPIRED; 641 642 err = __xfrm_state_delete(x); 643 if (!err) 644 km_state_expired(x, 1, 0); 645 646 xfrm_audit_state_delete(x, err ? 0 : 1, true); 647 648 out: 649 spin_unlock(&x->lock); 650 return ret; 651 } 652 653 static void xfrm_replay_timer_handler(struct timer_list *t); 654 655 struct xfrm_state *xfrm_state_alloc(struct net *net) 656 { 657 struct xfrm_state *x; 658 659 x = kmem_cache_zalloc(xfrm_state_cache, GFP_ATOMIC); 660 661 if (x) { 662 write_pnet(&x->xs_net, net); 663 refcount_set(&x->refcnt, 1); 664 atomic_set(&x->tunnel_users, 0); 665 INIT_LIST_HEAD(&x->km.all); 666 INIT_HLIST_NODE(&x->bydst); 667 INIT_HLIST_NODE(&x->bysrc); 668 INIT_HLIST_NODE(&x->byspi); 669 INIT_HLIST_NODE(&x->byseq); 670 hrtimer_init(&x->mtimer, CLOCK_BOOTTIME, HRTIMER_MODE_ABS_SOFT); 671 x->mtimer.function = xfrm_timer_handler; 672 timer_setup(&x->rtimer, xfrm_replay_timer_handler, 0); 673 x->curlft.add_time = ktime_get_real_seconds(); 674 x->lft.soft_byte_limit = XFRM_INF; 675 x->lft.soft_packet_limit = XFRM_INF; 676 x->lft.hard_byte_limit = XFRM_INF; 677 x->lft.hard_packet_limit = XFRM_INF; 678 x->replay_maxage = 0; 679 x->replay_maxdiff = 0; 680 spin_lock_init(&x->lock); 681 } 682 return x; 683 } 684 EXPORT_SYMBOL(xfrm_state_alloc); 685 686 void __xfrm_state_destroy(struct xfrm_state *x, bool sync) 687 { 688 WARN_ON(x->km.state != XFRM_STATE_DEAD); 689 690 if (sync) { 691 synchronize_rcu(); 692 ___xfrm_state_destroy(x); 693 } else { 694 spin_lock_bh(&xfrm_state_gc_lock); 695 hlist_add_head(&x->gclist, &xfrm_state_gc_list); 696 spin_unlock_bh(&xfrm_state_gc_lock); 697 schedule_work(&xfrm_state_gc_work); 698 } 699 } 700 EXPORT_SYMBOL(__xfrm_state_destroy); 701 702 int __xfrm_state_delete(struct xfrm_state *x) 703 { 704 struct net *net = xs_net(x); 705 int err = -ESRCH; 706 707 if (x->km.state != XFRM_STATE_DEAD) { 708 x->km.state = XFRM_STATE_DEAD; 709 spin_lock(&net->xfrm.xfrm_state_lock); 710 list_del(&x->km.all); 711 hlist_del_rcu(&x->bydst); 712 hlist_del_rcu(&x->bysrc); 713 if (x->km.seq) 714 hlist_del_rcu(&x->byseq); 715 if (x->id.spi) 716 hlist_del_rcu(&x->byspi); 717 net->xfrm.state_num--; 718 spin_unlock(&net->xfrm.xfrm_state_lock); 719 720 if (x->encap_sk) 721 sock_put(rcu_dereference_raw(x->encap_sk)); 722 723 xfrm_dev_state_delete(x); 724 725 /* All xfrm_state objects are created by xfrm_state_alloc. 726 * The xfrm_state_alloc call gives a reference, and that 727 * is what we are dropping here. 728 */ 729 xfrm_state_put(x); 730 err = 0; 731 } 732 733 return err; 734 } 735 EXPORT_SYMBOL(__xfrm_state_delete); 736 737 int xfrm_state_delete(struct xfrm_state *x) 738 { 739 int err; 740 741 spin_lock_bh(&x->lock); 742 err = __xfrm_state_delete(x); 743 spin_unlock_bh(&x->lock); 744 745 return err; 746 } 747 EXPORT_SYMBOL(xfrm_state_delete); 748 749 #ifdef CONFIG_SECURITY_NETWORK_XFRM 750 static inline int 751 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid) 752 { 753 int i, err = 0; 754 755 for (i = 0; i <= net->xfrm.state_hmask; i++) { 756 struct xfrm_state *x; 757 758 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) { 759 if (xfrm_id_proto_match(x->id.proto, proto) && 760 (err = security_xfrm_state_delete(x)) != 0) { 761 xfrm_audit_state_delete(x, 0, task_valid); 762 return err; 763 } 764 } 765 } 766 767 return err; 768 } 769 770 static inline int 771 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid) 772 { 773 int i, err = 0; 774 775 for (i = 0; i <= net->xfrm.state_hmask; i++) { 776 struct xfrm_state *x; 777 struct xfrm_dev_offload *xso; 778 779 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) { 780 xso = &x->xso; 781 782 if (xso->dev == dev && 783 (err = security_xfrm_state_delete(x)) != 0) { 784 xfrm_audit_state_delete(x, 0, task_valid); 785 return err; 786 } 787 } 788 } 789 790 return err; 791 } 792 #else 793 static inline int 794 xfrm_state_flush_secctx_check(struct net *net, u8 proto, bool task_valid) 795 { 796 return 0; 797 } 798 799 static inline int 800 xfrm_dev_state_flush_secctx_check(struct net *net, struct net_device *dev, bool task_valid) 801 { 802 return 0; 803 } 804 #endif 805 806 int xfrm_state_flush(struct net *net, u8 proto, bool task_valid, bool sync) 807 { 808 int i, err = 0, cnt = 0; 809 810 spin_lock_bh(&net->xfrm.xfrm_state_lock); 811 err = xfrm_state_flush_secctx_check(net, proto, task_valid); 812 if (err) 813 goto out; 814 815 err = -ESRCH; 816 for (i = 0; i <= net->xfrm.state_hmask; i++) { 817 struct xfrm_state *x; 818 restart: 819 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) { 820 if (!xfrm_state_kern(x) && 821 xfrm_id_proto_match(x->id.proto, proto)) { 822 xfrm_state_hold(x); 823 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 824 825 err = xfrm_state_delete(x); 826 xfrm_audit_state_delete(x, err ? 0 : 1, 827 task_valid); 828 if (sync) 829 xfrm_state_put_sync(x); 830 else 831 xfrm_state_put(x); 832 if (!err) 833 cnt++; 834 835 spin_lock_bh(&net->xfrm.xfrm_state_lock); 836 goto restart; 837 } 838 } 839 } 840 out: 841 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 842 if (cnt) 843 err = 0; 844 845 return err; 846 } 847 EXPORT_SYMBOL(xfrm_state_flush); 848 849 int xfrm_dev_state_flush(struct net *net, struct net_device *dev, bool task_valid) 850 { 851 int i, err = 0, cnt = 0; 852 853 spin_lock_bh(&net->xfrm.xfrm_state_lock); 854 err = xfrm_dev_state_flush_secctx_check(net, dev, task_valid); 855 if (err) 856 goto out; 857 858 err = -ESRCH; 859 for (i = 0; i <= net->xfrm.state_hmask; i++) { 860 struct xfrm_state *x; 861 struct xfrm_dev_offload *xso; 862 restart: 863 hlist_for_each_entry(x, net->xfrm.state_bydst+i, bydst) { 864 xso = &x->xso; 865 866 if (!xfrm_state_kern(x) && xso->dev == dev) { 867 xfrm_state_hold(x); 868 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 869 870 err = xfrm_state_delete(x); 871 xfrm_audit_state_delete(x, err ? 0 : 1, 872 task_valid); 873 xfrm_state_put(x); 874 if (!err) 875 cnt++; 876 877 spin_lock_bh(&net->xfrm.xfrm_state_lock); 878 goto restart; 879 } 880 } 881 } 882 if (cnt) 883 err = 0; 884 885 out: 886 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 887 return err; 888 } 889 EXPORT_SYMBOL(xfrm_dev_state_flush); 890 891 void xfrm_sad_getinfo(struct net *net, struct xfrmk_sadinfo *si) 892 { 893 spin_lock_bh(&net->xfrm.xfrm_state_lock); 894 si->sadcnt = net->xfrm.state_num; 895 si->sadhcnt = net->xfrm.state_hmask + 1; 896 si->sadhmcnt = xfrm_state_hashmax; 897 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 898 } 899 EXPORT_SYMBOL(xfrm_sad_getinfo); 900 901 static void 902 __xfrm4_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl) 903 { 904 const struct flowi4 *fl4 = &fl->u.ip4; 905 906 sel->daddr.a4 = fl4->daddr; 907 sel->saddr.a4 = fl4->saddr; 908 sel->dport = xfrm_flowi_dport(fl, &fl4->uli); 909 sel->dport_mask = htons(0xffff); 910 sel->sport = xfrm_flowi_sport(fl, &fl4->uli); 911 sel->sport_mask = htons(0xffff); 912 sel->family = AF_INET; 913 sel->prefixlen_d = 32; 914 sel->prefixlen_s = 32; 915 sel->proto = fl4->flowi4_proto; 916 sel->ifindex = fl4->flowi4_oif; 917 } 918 919 static void 920 __xfrm6_init_tempsel(struct xfrm_selector *sel, const struct flowi *fl) 921 { 922 const struct flowi6 *fl6 = &fl->u.ip6; 923 924 /* Initialize temporary selector matching only to current session. */ 925 *(struct in6_addr *)&sel->daddr = fl6->daddr; 926 *(struct in6_addr *)&sel->saddr = fl6->saddr; 927 sel->dport = xfrm_flowi_dport(fl, &fl6->uli); 928 sel->dport_mask = htons(0xffff); 929 sel->sport = xfrm_flowi_sport(fl, &fl6->uli); 930 sel->sport_mask = htons(0xffff); 931 sel->family = AF_INET6; 932 sel->prefixlen_d = 128; 933 sel->prefixlen_s = 128; 934 sel->proto = fl6->flowi6_proto; 935 sel->ifindex = fl6->flowi6_oif; 936 } 937 938 static void 939 xfrm_init_tempstate(struct xfrm_state *x, const struct flowi *fl, 940 const struct xfrm_tmpl *tmpl, 941 const xfrm_address_t *daddr, const xfrm_address_t *saddr, 942 unsigned short family) 943 { 944 switch (family) { 945 case AF_INET: 946 __xfrm4_init_tempsel(&x->sel, fl); 947 break; 948 case AF_INET6: 949 __xfrm6_init_tempsel(&x->sel, fl); 950 break; 951 } 952 953 x->id = tmpl->id; 954 955 switch (tmpl->encap_family) { 956 case AF_INET: 957 if (x->id.daddr.a4 == 0) 958 x->id.daddr.a4 = daddr->a4; 959 x->props.saddr = tmpl->saddr; 960 if (x->props.saddr.a4 == 0) 961 x->props.saddr.a4 = saddr->a4; 962 break; 963 case AF_INET6: 964 if (ipv6_addr_any((struct in6_addr *)&x->id.daddr)) 965 memcpy(&x->id.daddr, daddr, sizeof(x->sel.daddr)); 966 memcpy(&x->props.saddr, &tmpl->saddr, sizeof(x->props.saddr)); 967 if (ipv6_addr_any((struct in6_addr *)&x->props.saddr)) 968 memcpy(&x->props.saddr, saddr, sizeof(x->props.saddr)); 969 break; 970 } 971 972 x->props.mode = tmpl->mode; 973 x->props.reqid = tmpl->reqid; 974 x->props.family = tmpl->encap_family; 975 } 976 977 static struct xfrm_state *__xfrm_state_lookup_all(struct net *net, u32 mark, 978 const xfrm_address_t *daddr, 979 __be32 spi, u8 proto, 980 unsigned short family, 981 struct xfrm_dev_offload *xdo) 982 { 983 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family); 984 struct xfrm_state *x; 985 986 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) { 987 #ifdef CONFIG_XFRM_OFFLOAD 988 if (xdo->type == XFRM_DEV_OFFLOAD_PACKET) { 989 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET) 990 /* HW states are in the head of list, there is 991 * no need to iterate further. 992 */ 993 break; 994 995 /* Packet offload: both policy and SA should 996 * have same device. 997 */ 998 if (xdo->dev != x->xso.dev) 999 continue; 1000 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET) 1001 /* Skip HW policy for SW lookups */ 1002 continue; 1003 #endif 1004 if (x->props.family != family || 1005 x->id.spi != spi || 1006 x->id.proto != proto || 1007 !xfrm_addr_equal(&x->id.daddr, daddr, family)) 1008 continue; 1009 1010 if ((mark & x->mark.m) != x->mark.v) 1011 continue; 1012 if (!xfrm_state_hold_rcu(x)) 1013 continue; 1014 return x; 1015 } 1016 1017 return NULL; 1018 } 1019 1020 static struct xfrm_state *__xfrm_state_lookup(struct net *net, u32 mark, 1021 const xfrm_address_t *daddr, 1022 __be32 spi, u8 proto, 1023 unsigned short family) 1024 { 1025 unsigned int h = xfrm_spi_hash(net, daddr, spi, proto, family); 1026 struct xfrm_state *x; 1027 1028 hlist_for_each_entry_rcu(x, net->xfrm.state_byspi + h, byspi) { 1029 if (x->props.family != family || 1030 x->id.spi != spi || 1031 x->id.proto != proto || 1032 !xfrm_addr_equal(&x->id.daddr, daddr, family)) 1033 continue; 1034 1035 if ((mark & x->mark.m) != x->mark.v) 1036 continue; 1037 if (!xfrm_state_hold_rcu(x)) 1038 continue; 1039 return x; 1040 } 1041 1042 return NULL; 1043 } 1044 1045 static struct xfrm_state *__xfrm_state_lookup_byaddr(struct net *net, u32 mark, 1046 const xfrm_address_t *daddr, 1047 const xfrm_address_t *saddr, 1048 u8 proto, unsigned short family) 1049 { 1050 unsigned int h = xfrm_src_hash(net, daddr, saddr, family); 1051 struct xfrm_state *x; 1052 1053 hlist_for_each_entry_rcu(x, net->xfrm.state_bysrc + h, bysrc) { 1054 if (x->props.family != family || 1055 x->id.proto != proto || 1056 !xfrm_addr_equal(&x->id.daddr, daddr, family) || 1057 !xfrm_addr_equal(&x->props.saddr, saddr, family)) 1058 continue; 1059 1060 if ((mark & x->mark.m) != x->mark.v) 1061 continue; 1062 if (!xfrm_state_hold_rcu(x)) 1063 continue; 1064 return x; 1065 } 1066 1067 return NULL; 1068 } 1069 1070 static inline struct xfrm_state * 1071 __xfrm_state_locate(struct xfrm_state *x, int use_spi, int family) 1072 { 1073 struct net *net = xs_net(x); 1074 u32 mark = x->mark.v & x->mark.m; 1075 1076 if (use_spi) 1077 return __xfrm_state_lookup(net, mark, &x->id.daddr, 1078 x->id.spi, x->id.proto, family); 1079 else 1080 return __xfrm_state_lookup_byaddr(net, mark, 1081 &x->id.daddr, 1082 &x->props.saddr, 1083 x->id.proto, family); 1084 } 1085 1086 static void xfrm_hash_grow_check(struct net *net, int have_hash_collision) 1087 { 1088 if (have_hash_collision && 1089 (net->xfrm.state_hmask + 1) < xfrm_state_hashmax && 1090 net->xfrm.state_num > net->xfrm.state_hmask) 1091 schedule_work(&net->xfrm.state_hash_work); 1092 } 1093 1094 static void xfrm_state_look_at(struct xfrm_policy *pol, struct xfrm_state *x, 1095 const struct flowi *fl, unsigned short family, 1096 struct xfrm_state **best, int *acq_in_progress, 1097 int *error) 1098 { 1099 /* Resolution logic: 1100 * 1. There is a valid state with matching selector. Done. 1101 * 2. Valid state with inappropriate selector. Skip. 1102 * 1103 * Entering area of "sysdeps". 1104 * 1105 * 3. If state is not valid, selector is temporary, it selects 1106 * only session which triggered previous resolution. Key 1107 * manager will do something to install a state with proper 1108 * selector. 1109 */ 1110 if (x->km.state == XFRM_STATE_VALID) { 1111 if ((x->sel.family && 1112 (x->sel.family != family || 1113 !xfrm_selector_match(&x->sel, fl, family))) || 1114 !security_xfrm_state_pol_flow_match(x, pol, 1115 &fl->u.__fl_common)) 1116 return; 1117 1118 if (!*best || 1119 (*best)->km.dying > x->km.dying || 1120 ((*best)->km.dying == x->km.dying && 1121 (*best)->curlft.add_time < x->curlft.add_time)) 1122 *best = x; 1123 } else if (x->km.state == XFRM_STATE_ACQ) { 1124 *acq_in_progress = 1; 1125 } else if (x->km.state == XFRM_STATE_ERROR || 1126 x->km.state == XFRM_STATE_EXPIRED) { 1127 if ((!x->sel.family || 1128 (x->sel.family == family && 1129 xfrm_selector_match(&x->sel, fl, family))) && 1130 security_xfrm_state_pol_flow_match(x, pol, 1131 &fl->u.__fl_common)) 1132 *error = -ESRCH; 1133 } 1134 } 1135 1136 struct xfrm_state * 1137 xfrm_state_find(const xfrm_address_t *daddr, const xfrm_address_t *saddr, 1138 const struct flowi *fl, struct xfrm_tmpl *tmpl, 1139 struct xfrm_policy *pol, int *err, 1140 unsigned short family, u32 if_id) 1141 { 1142 static xfrm_address_t saddr_wildcard = { }; 1143 struct net *net = xp_net(pol); 1144 unsigned int h, h_wildcard; 1145 struct xfrm_state *x, *x0, *to_put; 1146 int acquire_in_progress = 0; 1147 int error = 0; 1148 struct xfrm_state *best = NULL; 1149 u32 mark = pol->mark.v & pol->mark.m; 1150 unsigned short encap_family = tmpl->encap_family; 1151 unsigned int sequence; 1152 struct km_event c; 1153 1154 to_put = NULL; 1155 1156 sequence = read_seqcount_begin(&net->xfrm.xfrm_state_hash_generation); 1157 1158 rcu_read_lock(); 1159 h = xfrm_dst_hash(net, daddr, saddr, tmpl->reqid, encap_family); 1160 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h, bydst) { 1161 #ifdef CONFIG_XFRM_OFFLOAD 1162 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) { 1163 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET) 1164 /* HW states are in the head of list, there is 1165 * no need to iterate further. 1166 */ 1167 break; 1168 1169 /* Packet offload: both policy and SA should 1170 * have same device. 1171 */ 1172 if (pol->xdo.dev != x->xso.dev) 1173 continue; 1174 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET) 1175 /* Skip HW policy for SW lookups */ 1176 continue; 1177 #endif 1178 if (x->props.family == encap_family && 1179 x->props.reqid == tmpl->reqid && 1180 (mark & x->mark.m) == x->mark.v && 1181 x->if_id == if_id && 1182 !(x->props.flags & XFRM_STATE_WILDRECV) && 1183 xfrm_state_addr_check(x, daddr, saddr, encap_family) && 1184 tmpl->mode == x->props.mode && 1185 tmpl->id.proto == x->id.proto && 1186 (tmpl->id.spi == x->id.spi || !tmpl->id.spi)) 1187 xfrm_state_look_at(pol, x, fl, family, 1188 &best, &acquire_in_progress, &error); 1189 } 1190 if (best || acquire_in_progress) 1191 goto found; 1192 1193 h_wildcard = xfrm_dst_hash(net, daddr, &saddr_wildcard, tmpl->reqid, encap_family); 1194 hlist_for_each_entry_rcu(x, net->xfrm.state_bydst + h_wildcard, bydst) { 1195 #ifdef CONFIG_XFRM_OFFLOAD 1196 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) { 1197 if (x->xso.type != XFRM_DEV_OFFLOAD_PACKET) 1198 /* HW states are in the head of list, there is 1199 * no need to iterate further. 1200 */ 1201 break; 1202 1203 /* Packet offload: both policy and SA should 1204 * have same device. 1205 */ 1206 if (pol->xdo.dev != x->xso.dev) 1207 continue; 1208 } else if (x->xso.type == XFRM_DEV_OFFLOAD_PACKET) 1209 /* Skip HW policy for SW lookups */ 1210 continue; 1211 #endif 1212 if (x->props.family == encap_family && 1213 x->props.reqid == tmpl->reqid && 1214 (mark & x->mark.m) == x->mark.v && 1215 x->if_id == if_id && 1216 !(x->props.flags & XFRM_STATE_WILDRECV) && 1217 xfrm_addr_equal(&x->id.daddr, daddr, encap_family) && 1218 tmpl->mode == x->props.mode && 1219 tmpl->id.proto == x->id.proto && 1220 (tmpl->id.spi == x->id.spi || !tmpl->id.spi)) 1221 xfrm_state_look_at(pol, x, fl, family, 1222 &best, &acquire_in_progress, &error); 1223 } 1224 1225 found: 1226 x = best; 1227 if (!x && !error && !acquire_in_progress) { 1228 if (tmpl->id.spi && 1229 (x0 = __xfrm_state_lookup_all(net, mark, daddr, 1230 tmpl->id.spi, tmpl->id.proto, 1231 encap_family, 1232 &pol->xdo)) != NULL) { 1233 to_put = x0; 1234 error = -EEXIST; 1235 goto out; 1236 } 1237 1238 c.net = net; 1239 /* If the KMs have no listeners (yet...), avoid allocating an SA 1240 * for each and every packet - garbage collection might not 1241 * handle the flood. 1242 */ 1243 if (!km_is_alive(&c)) { 1244 error = -ESRCH; 1245 goto out; 1246 } 1247 1248 x = xfrm_state_alloc(net); 1249 if (x == NULL) { 1250 error = -ENOMEM; 1251 goto out; 1252 } 1253 /* Initialize temporary state matching only 1254 * to current session. */ 1255 xfrm_init_tempstate(x, fl, tmpl, daddr, saddr, family); 1256 memcpy(&x->mark, &pol->mark, sizeof(x->mark)); 1257 x->if_id = if_id; 1258 1259 error = security_xfrm_state_alloc_acquire(x, pol->security, fl->flowi_secid); 1260 if (error) { 1261 x->km.state = XFRM_STATE_DEAD; 1262 to_put = x; 1263 x = NULL; 1264 goto out; 1265 } 1266 #ifdef CONFIG_XFRM_OFFLOAD 1267 if (pol->xdo.type == XFRM_DEV_OFFLOAD_PACKET) { 1268 struct xfrm_dev_offload *xdo = &pol->xdo; 1269 struct xfrm_dev_offload *xso = &x->xso; 1270 1271 xso->type = XFRM_DEV_OFFLOAD_PACKET; 1272 xso->dir = xdo->dir; 1273 xso->dev = xdo->dev; 1274 xso->real_dev = xdo->real_dev; 1275 netdev_tracker_alloc(xso->dev, &xso->dev_tracker, 1276 GFP_ATOMIC); 1277 error = xso->dev->xfrmdev_ops->xdo_dev_state_add(x, NULL); 1278 if (error) { 1279 xso->dir = 0; 1280 netdev_put(xso->dev, &xso->dev_tracker); 1281 xso->dev = NULL; 1282 xso->real_dev = NULL; 1283 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED; 1284 x->km.state = XFRM_STATE_DEAD; 1285 to_put = x; 1286 x = NULL; 1287 goto out; 1288 } 1289 } 1290 #endif 1291 if (km_query(x, tmpl, pol) == 0) { 1292 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1293 x->km.state = XFRM_STATE_ACQ; 1294 list_add(&x->km.all, &net->xfrm.state_all); 1295 XFRM_STATE_INSERT(bydst, &x->bydst, 1296 net->xfrm.state_bydst + h, 1297 x->xso.type); 1298 h = xfrm_src_hash(net, daddr, saddr, encap_family); 1299 XFRM_STATE_INSERT(bysrc, &x->bysrc, 1300 net->xfrm.state_bysrc + h, 1301 x->xso.type); 1302 if (x->id.spi) { 1303 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, encap_family); 1304 XFRM_STATE_INSERT(byspi, &x->byspi, 1305 net->xfrm.state_byspi + h, 1306 x->xso.type); 1307 } 1308 if (x->km.seq) { 1309 h = xfrm_seq_hash(net, x->km.seq); 1310 XFRM_STATE_INSERT(byseq, &x->byseq, 1311 net->xfrm.state_byseq + h, 1312 x->xso.type); 1313 } 1314 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires; 1315 hrtimer_start(&x->mtimer, 1316 ktime_set(net->xfrm.sysctl_acq_expires, 0), 1317 HRTIMER_MODE_REL_SOFT); 1318 net->xfrm.state_num++; 1319 xfrm_hash_grow_check(net, x->bydst.next != NULL); 1320 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1321 } else { 1322 #ifdef CONFIG_XFRM_OFFLOAD 1323 struct xfrm_dev_offload *xso = &x->xso; 1324 1325 if (xso->type == XFRM_DEV_OFFLOAD_PACKET) { 1326 xso->dev->xfrmdev_ops->xdo_dev_state_delete(x); 1327 xso->dir = 0; 1328 netdev_put(xso->dev, &xso->dev_tracker); 1329 xso->dev = NULL; 1330 xso->real_dev = NULL; 1331 xso->type = XFRM_DEV_OFFLOAD_UNSPECIFIED; 1332 } 1333 #endif 1334 x->km.state = XFRM_STATE_DEAD; 1335 to_put = x; 1336 x = NULL; 1337 error = -ESRCH; 1338 } 1339 } 1340 out: 1341 if (x) { 1342 if (!xfrm_state_hold_rcu(x)) { 1343 *err = -EAGAIN; 1344 x = NULL; 1345 } 1346 } else { 1347 *err = acquire_in_progress ? -EAGAIN : error; 1348 } 1349 rcu_read_unlock(); 1350 if (to_put) 1351 xfrm_state_put(to_put); 1352 1353 if (read_seqcount_retry(&net->xfrm.xfrm_state_hash_generation, sequence)) { 1354 *err = -EAGAIN; 1355 if (x) { 1356 xfrm_state_put(x); 1357 x = NULL; 1358 } 1359 } 1360 1361 return x; 1362 } 1363 1364 struct xfrm_state * 1365 xfrm_stateonly_find(struct net *net, u32 mark, u32 if_id, 1366 xfrm_address_t *daddr, xfrm_address_t *saddr, 1367 unsigned short family, u8 mode, u8 proto, u32 reqid) 1368 { 1369 unsigned int h; 1370 struct xfrm_state *rx = NULL, *x = NULL; 1371 1372 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1373 h = xfrm_dst_hash(net, daddr, saddr, reqid, family); 1374 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) { 1375 if (x->props.family == family && 1376 x->props.reqid == reqid && 1377 (mark & x->mark.m) == x->mark.v && 1378 x->if_id == if_id && 1379 !(x->props.flags & XFRM_STATE_WILDRECV) && 1380 xfrm_state_addr_check(x, daddr, saddr, family) && 1381 mode == x->props.mode && 1382 proto == x->id.proto && 1383 x->km.state == XFRM_STATE_VALID) { 1384 rx = x; 1385 break; 1386 } 1387 } 1388 1389 if (rx) 1390 xfrm_state_hold(rx); 1391 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1392 1393 1394 return rx; 1395 } 1396 EXPORT_SYMBOL(xfrm_stateonly_find); 1397 1398 struct xfrm_state *xfrm_state_lookup_byspi(struct net *net, __be32 spi, 1399 unsigned short family) 1400 { 1401 struct xfrm_state *x; 1402 struct xfrm_state_walk *w; 1403 1404 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1405 list_for_each_entry(w, &net->xfrm.state_all, all) { 1406 x = container_of(w, struct xfrm_state, km); 1407 if (x->props.family != family || 1408 x->id.spi != spi) 1409 continue; 1410 1411 xfrm_state_hold(x); 1412 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1413 return x; 1414 } 1415 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1416 return NULL; 1417 } 1418 EXPORT_SYMBOL(xfrm_state_lookup_byspi); 1419 1420 static void __xfrm_state_insert(struct xfrm_state *x) 1421 { 1422 struct net *net = xs_net(x); 1423 unsigned int h; 1424 1425 list_add(&x->km.all, &net->xfrm.state_all); 1426 1427 h = xfrm_dst_hash(net, &x->id.daddr, &x->props.saddr, 1428 x->props.reqid, x->props.family); 1429 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h, 1430 x->xso.type); 1431 1432 h = xfrm_src_hash(net, &x->id.daddr, &x->props.saddr, x->props.family); 1433 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h, 1434 x->xso.type); 1435 1436 if (x->id.spi) { 1437 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, 1438 x->props.family); 1439 1440 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h, 1441 x->xso.type); 1442 } 1443 1444 if (x->km.seq) { 1445 h = xfrm_seq_hash(net, x->km.seq); 1446 1447 XFRM_STATE_INSERT(byseq, &x->byseq, net->xfrm.state_byseq + h, 1448 x->xso.type); 1449 } 1450 1451 hrtimer_start(&x->mtimer, ktime_set(1, 0), HRTIMER_MODE_REL_SOFT); 1452 if (x->replay_maxage) 1453 mod_timer(&x->rtimer, jiffies + x->replay_maxage); 1454 1455 net->xfrm.state_num++; 1456 1457 xfrm_hash_grow_check(net, x->bydst.next != NULL); 1458 } 1459 1460 /* net->xfrm.xfrm_state_lock is held */ 1461 static void __xfrm_state_bump_genids(struct xfrm_state *xnew) 1462 { 1463 struct net *net = xs_net(xnew); 1464 unsigned short family = xnew->props.family; 1465 u32 reqid = xnew->props.reqid; 1466 struct xfrm_state *x; 1467 unsigned int h; 1468 u32 mark = xnew->mark.v & xnew->mark.m; 1469 u32 if_id = xnew->if_id; 1470 1471 h = xfrm_dst_hash(net, &xnew->id.daddr, &xnew->props.saddr, reqid, family); 1472 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) { 1473 if (x->props.family == family && 1474 x->props.reqid == reqid && 1475 x->if_id == if_id && 1476 (mark & x->mark.m) == x->mark.v && 1477 xfrm_addr_equal(&x->id.daddr, &xnew->id.daddr, family) && 1478 xfrm_addr_equal(&x->props.saddr, &xnew->props.saddr, family)) 1479 x->genid++; 1480 } 1481 } 1482 1483 void xfrm_state_insert(struct xfrm_state *x) 1484 { 1485 struct net *net = xs_net(x); 1486 1487 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1488 __xfrm_state_bump_genids(x); 1489 __xfrm_state_insert(x); 1490 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1491 } 1492 EXPORT_SYMBOL(xfrm_state_insert); 1493 1494 /* net->xfrm.xfrm_state_lock is held */ 1495 static struct xfrm_state *__find_acq_core(struct net *net, 1496 const struct xfrm_mark *m, 1497 unsigned short family, u8 mode, 1498 u32 reqid, u32 if_id, u8 proto, 1499 const xfrm_address_t *daddr, 1500 const xfrm_address_t *saddr, 1501 int create) 1502 { 1503 unsigned int h = xfrm_dst_hash(net, daddr, saddr, reqid, family); 1504 struct xfrm_state *x; 1505 u32 mark = m->v & m->m; 1506 1507 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) { 1508 if (x->props.reqid != reqid || 1509 x->props.mode != mode || 1510 x->props.family != family || 1511 x->km.state != XFRM_STATE_ACQ || 1512 x->id.spi != 0 || 1513 x->id.proto != proto || 1514 (mark & x->mark.m) != x->mark.v || 1515 !xfrm_addr_equal(&x->id.daddr, daddr, family) || 1516 !xfrm_addr_equal(&x->props.saddr, saddr, family)) 1517 continue; 1518 1519 xfrm_state_hold(x); 1520 return x; 1521 } 1522 1523 if (!create) 1524 return NULL; 1525 1526 x = xfrm_state_alloc(net); 1527 if (likely(x)) { 1528 switch (family) { 1529 case AF_INET: 1530 x->sel.daddr.a4 = daddr->a4; 1531 x->sel.saddr.a4 = saddr->a4; 1532 x->sel.prefixlen_d = 32; 1533 x->sel.prefixlen_s = 32; 1534 x->props.saddr.a4 = saddr->a4; 1535 x->id.daddr.a4 = daddr->a4; 1536 break; 1537 1538 case AF_INET6: 1539 x->sel.daddr.in6 = daddr->in6; 1540 x->sel.saddr.in6 = saddr->in6; 1541 x->sel.prefixlen_d = 128; 1542 x->sel.prefixlen_s = 128; 1543 x->props.saddr.in6 = saddr->in6; 1544 x->id.daddr.in6 = daddr->in6; 1545 break; 1546 } 1547 1548 x->km.state = XFRM_STATE_ACQ; 1549 x->id.proto = proto; 1550 x->props.family = family; 1551 x->props.mode = mode; 1552 x->props.reqid = reqid; 1553 x->if_id = if_id; 1554 x->mark.v = m->v; 1555 x->mark.m = m->m; 1556 x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires; 1557 xfrm_state_hold(x); 1558 hrtimer_start(&x->mtimer, 1559 ktime_set(net->xfrm.sysctl_acq_expires, 0), 1560 HRTIMER_MODE_REL_SOFT); 1561 list_add(&x->km.all, &net->xfrm.state_all); 1562 XFRM_STATE_INSERT(bydst, &x->bydst, net->xfrm.state_bydst + h, 1563 x->xso.type); 1564 h = xfrm_src_hash(net, daddr, saddr, family); 1565 XFRM_STATE_INSERT(bysrc, &x->bysrc, net->xfrm.state_bysrc + h, 1566 x->xso.type); 1567 1568 net->xfrm.state_num++; 1569 1570 xfrm_hash_grow_check(net, x->bydst.next != NULL); 1571 } 1572 1573 return x; 1574 } 1575 1576 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq); 1577 1578 int xfrm_state_add(struct xfrm_state *x) 1579 { 1580 struct net *net = xs_net(x); 1581 struct xfrm_state *x1, *to_put; 1582 int family; 1583 int err; 1584 u32 mark = x->mark.v & x->mark.m; 1585 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY); 1586 1587 family = x->props.family; 1588 1589 to_put = NULL; 1590 1591 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1592 1593 x1 = __xfrm_state_locate(x, use_spi, family); 1594 if (x1) { 1595 to_put = x1; 1596 x1 = NULL; 1597 err = -EEXIST; 1598 goto out; 1599 } 1600 1601 if (use_spi && x->km.seq) { 1602 x1 = __xfrm_find_acq_byseq(net, mark, x->km.seq); 1603 if (x1 && ((x1->id.proto != x->id.proto) || 1604 !xfrm_addr_equal(&x1->id.daddr, &x->id.daddr, family))) { 1605 to_put = x1; 1606 x1 = NULL; 1607 } 1608 } 1609 1610 if (use_spi && !x1) 1611 x1 = __find_acq_core(net, &x->mark, family, x->props.mode, 1612 x->props.reqid, x->if_id, x->id.proto, 1613 &x->id.daddr, &x->props.saddr, 0); 1614 1615 __xfrm_state_bump_genids(x); 1616 __xfrm_state_insert(x); 1617 err = 0; 1618 1619 out: 1620 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1621 1622 if (x1) { 1623 xfrm_state_delete(x1); 1624 xfrm_state_put(x1); 1625 } 1626 1627 if (to_put) 1628 xfrm_state_put(to_put); 1629 1630 return err; 1631 } 1632 EXPORT_SYMBOL(xfrm_state_add); 1633 1634 #ifdef CONFIG_XFRM_MIGRATE 1635 static inline int clone_security(struct xfrm_state *x, struct xfrm_sec_ctx *security) 1636 { 1637 struct xfrm_user_sec_ctx *uctx; 1638 int size = sizeof(*uctx) + security->ctx_len; 1639 int err; 1640 1641 uctx = kmalloc(size, GFP_KERNEL); 1642 if (!uctx) 1643 return -ENOMEM; 1644 1645 uctx->exttype = XFRMA_SEC_CTX; 1646 uctx->len = size; 1647 uctx->ctx_doi = security->ctx_doi; 1648 uctx->ctx_alg = security->ctx_alg; 1649 uctx->ctx_len = security->ctx_len; 1650 memcpy(uctx + 1, security->ctx_str, security->ctx_len); 1651 err = security_xfrm_state_alloc(x, uctx); 1652 kfree(uctx); 1653 if (err) 1654 return err; 1655 1656 return 0; 1657 } 1658 1659 static struct xfrm_state *xfrm_state_clone(struct xfrm_state *orig, 1660 struct xfrm_encap_tmpl *encap) 1661 { 1662 struct net *net = xs_net(orig); 1663 struct xfrm_state *x = xfrm_state_alloc(net); 1664 if (!x) 1665 goto out; 1666 1667 memcpy(&x->id, &orig->id, sizeof(x->id)); 1668 memcpy(&x->sel, &orig->sel, sizeof(x->sel)); 1669 memcpy(&x->lft, &orig->lft, sizeof(x->lft)); 1670 x->props.mode = orig->props.mode; 1671 x->props.replay_window = orig->props.replay_window; 1672 x->props.reqid = orig->props.reqid; 1673 x->props.family = orig->props.family; 1674 x->props.saddr = orig->props.saddr; 1675 1676 if (orig->aalg) { 1677 x->aalg = xfrm_algo_auth_clone(orig->aalg); 1678 if (!x->aalg) 1679 goto error; 1680 } 1681 x->props.aalgo = orig->props.aalgo; 1682 1683 if (orig->aead) { 1684 x->aead = xfrm_algo_aead_clone(orig->aead); 1685 x->geniv = orig->geniv; 1686 if (!x->aead) 1687 goto error; 1688 } 1689 if (orig->ealg) { 1690 x->ealg = xfrm_algo_clone(orig->ealg); 1691 if (!x->ealg) 1692 goto error; 1693 } 1694 x->props.ealgo = orig->props.ealgo; 1695 1696 if (orig->calg) { 1697 x->calg = xfrm_algo_clone(orig->calg); 1698 if (!x->calg) 1699 goto error; 1700 } 1701 x->props.calgo = orig->props.calgo; 1702 1703 if (encap || orig->encap) { 1704 if (encap) 1705 x->encap = kmemdup(encap, sizeof(*x->encap), 1706 GFP_KERNEL); 1707 else 1708 x->encap = kmemdup(orig->encap, sizeof(*x->encap), 1709 GFP_KERNEL); 1710 1711 if (!x->encap) 1712 goto error; 1713 } 1714 1715 if (orig->security) 1716 if (clone_security(x, orig->security)) 1717 goto error; 1718 1719 if (orig->coaddr) { 1720 x->coaddr = kmemdup(orig->coaddr, sizeof(*x->coaddr), 1721 GFP_KERNEL); 1722 if (!x->coaddr) 1723 goto error; 1724 } 1725 1726 if (orig->replay_esn) { 1727 if (xfrm_replay_clone(x, orig)) 1728 goto error; 1729 } 1730 1731 memcpy(&x->mark, &orig->mark, sizeof(x->mark)); 1732 memcpy(&x->props.smark, &orig->props.smark, sizeof(x->props.smark)); 1733 1734 x->props.flags = orig->props.flags; 1735 x->props.extra_flags = orig->props.extra_flags; 1736 1737 x->if_id = orig->if_id; 1738 x->tfcpad = orig->tfcpad; 1739 x->replay_maxdiff = orig->replay_maxdiff; 1740 x->replay_maxage = orig->replay_maxage; 1741 memcpy(&x->curlft, &orig->curlft, sizeof(x->curlft)); 1742 x->km.state = orig->km.state; 1743 x->km.seq = orig->km.seq; 1744 x->replay = orig->replay; 1745 x->preplay = orig->preplay; 1746 x->mapping_maxage = orig->mapping_maxage; 1747 x->lastused = orig->lastused; 1748 x->new_mapping = 0; 1749 x->new_mapping_sport = 0; 1750 1751 return x; 1752 1753 error: 1754 xfrm_state_put(x); 1755 out: 1756 return NULL; 1757 } 1758 1759 struct xfrm_state *xfrm_migrate_state_find(struct xfrm_migrate *m, struct net *net, 1760 u32 if_id) 1761 { 1762 unsigned int h; 1763 struct xfrm_state *x = NULL; 1764 1765 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1766 1767 if (m->reqid) { 1768 h = xfrm_dst_hash(net, &m->old_daddr, &m->old_saddr, 1769 m->reqid, m->old_family); 1770 hlist_for_each_entry(x, net->xfrm.state_bydst+h, bydst) { 1771 if (x->props.mode != m->mode || 1772 x->id.proto != m->proto) 1773 continue; 1774 if (m->reqid && x->props.reqid != m->reqid) 1775 continue; 1776 if (if_id != 0 && x->if_id != if_id) 1777 continue; 1778 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr, 1779 m->old_family) || 1780 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr, 1781 m->old_family)) 1782 continue; 1783 xfrm_state_hold(x); 1784 break; 1785 } 1786 } else { 1787 h = xfrm_src_hash(net, &m->old_daddr, &m->old_saddr, 1788 m->old_family); 1789 hlist_for_each_entry(x, net->xfrm.state_bysrc+h, bysrc) { 1790 if (x->props.mode != m->mode || 1791 x->id.proto != m->proto) 1792 continue; 1793 if (if_id != 0 && x->if_id != if_id) 1794 continue; 1795 if (!xfrm_addr_equal(&x->id.daddr, &m->old_daddr, 1796 m->old_family) || 1797 !xfrm_addr_equal(&x->props.saddr, &m->old_saddr, 1798 m->old_family)) 1799 continue; 1800 xfrm_state_hold(x); 1801 break; 1802 } 1803 } 1804 1805 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1806 1807 return x; 1808 } 1809 EXPORT_SYMBOL(xfrm_migrate_state_find); 1810 1811 struct xfrm_state *xfrm_state_migrate(struct xfrm_state *x, 1812 struct xfrm_migrate *m, 1813 struct xfrm_encap_tmpl *encap) 1814 { 1815 struct xfrm_state *xc; 1816 1817 xc = xfrm_state_clone(x, encap); 1818 if (!xc) 1819 return NULL; 1820 1821 xc->props.family = m->new_family; 1822 1823 if (xfrm_init_state(xc) < 0) 1824 goto error; 1825 1826 memcpy(&xc->id.daddr, &m->new_daddr, sizeof(xc->id.daddr)); 1827 memcpy(&xc->props.saddr, &m->new_saddr, sizeof(xc->props.saddr)); 1828 1829 /* add state */ 1830 if (xfrm_addr_equal(&x->id.daddr, &m->new_daddr, m->new_family)) { 1831 /* a care is needed when the destination address of the 1832 state is to be updated as it is a part of triplet */ 1833 xfrm_state_insert(xc); 1834 } else { 1835 if (xfrm_state_add(xc) < 0) 1836 goto error; 1837 } 1838 1839 return xc; 1840 error: 1841 xfrm_state_put(xc); 1842 return NULL; 1843 } 1844 EXPORT_SYMBOL(xfrm_state_migrate); 1845 #endif 1846 1847 int xfrm_state_update(struct xfrm_state *x) 1848 { 1849 struct xfrm_state *x1, *to_put; 1850 int err; 1851 int use_spi = xfrm_id_proto_match(x->id.proto, IPSEC_PROTO_ANY); 1852 struct net *net = xs_net(x); 1853 1854 to_put = NULL; 1855 1856 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1857 x1 = __xfrm_state_locate(x, use_spi, x->props.family); 1858 1859 err = -ESRCH; 1860 if (!x1) 1861 goto out; 1862 1863 if (xfrm_state_kern(x1)) { 1864 to_put = x1; 1865 err = -EEXIST; 1866 goto out; 1867 } 1868 1869 if (x1->km.state == XFRM_STATE_ACQ) { 1870 __xfrm_state_insert(x); 1871 x = NULL; 1872 } 1873 err = 0; 1874 1875 out: 1876 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1877 1878 if (to_put) 1879 xfrm_state_put(to_put); 1880 1881 if (err) 1882 return err; 1883 1884 if (!x) { 1885 xfrm_state_delete(x1); 1886 xfrm_state_put(x1); 1887 return 0; 1888 } 1889 1890 err = -EINVAL; 1891 spin_lock_bh(&x1->lock); 1892 if (likely(x1->km.state == XFRM_STATE_VALID)) { 1893 if (x->encap && x1->encap && 1894 x->encap->encap_type == x1->encap->encap_type) 1895 memcpy(x1->encap, x->encap, sizeof(*x1->encap)); 1896 else if (x->encap || x1->encap) 1897 goto fail; 1898 1899 if (x->coaddr && x1->coaddr) { 1900 memcpy(x1->coaddr, x->coaddr, sizeof(*x1->coaddr)); 1901 } 1902 if (!use_spi && memcmp(&x1->sel, &x->sel, sizeof(x1->sel))) 1903 memcpy(&x1->sel, &x->sel, sizeof(x1->sel)); 1904 memcpy(&x1->lft, &x->lft, sizeof(x1->lft)); 1905 x1->km.dying = 0; 1906 1907 hrtimer_start(&x1->mtimer, ktime_set(1, 0), 1908 HRTIMER_MODE_REL_SOFT); 1909 if (READ_ONCE(x1->curlft.use_time)) 1910 xfrm_state_check_expire(x1); 1911 1912 if (x->props.smark.m || x->props.smark.v || x->if_id) { 1913 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1914 1915 if (x->props.smark.m || x->props.smark.v) 1916 x1->props.smark = x->props.smark; 1917 1918 if (x->if_id) 1919 x1->if_id = x->if_id; 1920 1921 __xfrm_state_bump_genids(x1); 1922 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1923 } 1924 1925 err = 0; 1926 x->km.state = XFRM_STATE_DEAD; 1927 __xfrm_state_put(x); 1928 } 1929 1930 fail: 1931 spin_unlock_bh(&x1->lock); 1932 1933 xfrm_state_put(x1); 1934 1935 return err; 1936 } 1937 EXPORT_SYMBOL(xfrm_state_update); 1938 1939 int xfrm_state_check_expire(struct xfrm_state *x) 1940 { 1941 xfrm_dev_state_update_curlft(x); 1942 1943 if (!READ_ONCE(x->curlft.use_time)) 1944 WRITE_ONCE(x->curlft.use_time, ktime_get_real_seconds()); 1945 1946 if (x->curlft.bytes >= x->lft.hard_byte_limit || 1947 x->curlft.packets >= x->lft.hard_packet_limit) { 1948 x->km.state = XFRM_STATE_EXPIRED; 1949 hrtimer_start(&x->mtimer, 0, HRTIMER_MODE_REL_SOFT); 1950 return -EINVAL; 1951 } 1952 1953 if (!x->km.dying && 1954 (x->curlft.bytes >= x->lft.soft_byte_limit || 1955 x->curlft.packets >= x->lft.soft_packet_limit)) { 1956 x->km.dying = 1; 1957 km_state_expired(x, 0, 0); 1958 } 1959 return 0; 1960 } 1961 EXPORT_SYMBOL(xfrm_state_check_expire); 1962 1963 struct xfrm_state * 1964 xfrm_state_lookup(struct net *net, u32 mark, const xfrm_address_t *daddr, __be32 spi, 1965 u8 proto, unsigned short family) 1966 { 1967 struct xfrm_state *x; 1968 1969 rcu_read_lock(); 1970 x = __xfrm_state_lookup(net, mark, daddr, spi, proto, family); 1971 rcu_read_unlock(); 1972 return x; 1973 } 1974 EXPORT_SYMBOL(xfrm_state_lookup); 1975 1976 struct xfrm_state * 1977 xfrm_state_lookup_byaddr(struct net *net, u32 mark, 1978 const xfrm_address_t *daddr, const xfrm_address_t *saddr, 1979 u8 proto, unsigned short family) 1980 { 1981 struct xfrm_state *x; 1982 1983 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1984 x = __xfrm_state_lookup_byaddr(net, mark, daddr, saddr, proto, family); 1985 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 1986 return x; 1987 } 1988 EXPORT_SYMBOL(xfrm_state_lookup_byaddr); 1989 1990 struct xfrm_state * 1991 xfrm_find_acq(struct net *net, const struct xfrm_mark *mark, u8 mode, u32 reqid, 1992 u32 if_id, u8 proto, const xfrm_address_t *daddr, 1993 const xfrm_address_t *saddr, int create, unsigned short family) 1994 { 1995 struct xfrm_state *x; 1996 1997 spin_lock_bh(&net->xfrm.xfrm_state_lock); 1998 x = __find_acq_core(net, mark, family, mode, reqid, if_id, proto, daddr, saddr, create); 1999 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 2000 2001 return x; 2002 } 2003 EXPORT_SYMBOL(xfrm_find_acq); 2004 2005 #ifdef CONFIG_XFRM_SUB_POLICY 2006 #if IS_ENABLED(CONFIG_IPV6) 2007 /* distribution counting sort function for xfrm_state and xfrm_tmpl */ 2008 static void 2009 __xfrm6_sort(void **dst, void **src, int n, 2010 int (*cmp)(const void *p), int maxclass) 2011 { 2012 int count[XFRM_MAX_DEPTH] = { }; 2013 int class[XFRM_MAX_DEPTH]; 2014 int i; 2015 2016 for (i = 0; i < n; i++) { 2017 int c = cmp(src[i]); 2018 2019 class[i] = c; 2020 count[c]++; 2021 } 2022 2023 for (i = 2; i < maxclass; i++) 2024 count[i] += count[i - 1]; 2025 2026 for (i = 0; i < n; i++) { 2027 dst[count[class[i] - 1]++] = src[i]; 2028 src[i] = NULL; 2029 } 2030 } 2031 2032 /* Rule for xfrm_state: 2033 * 2034 * rule 1: select IPsec transport except AH 2035 * rule 2: select MIPv6 RO or inbound trigger 2036 * rule 3: select IPsec transport AH 2037 * rule 4: select IPsec tunnel 2038 * rule 5: others 2039 */ 2040 static int __xfrm6_state_sort_cmp(const void *p) 2041 { 2042 const struct xfrm_state *v = p; 2043 2044 switch (v->props.mode) { 2045 case XFRM_MODE_TRANSPORT: 2046 if (v->id.proto != IPPROTO_AH) 2047 return 1; 2048 else 2049 return 3; 2050 #if IS_ENABLED(CONFIG_IPV6_MIP6) 2051 case XFRM_MODE_ROUTEOPTIMIZATION: 2052 case XFRM_MODE_IN_TRIGGER: 2053 return 2; 2054 #endif 2055 case XFRM_MODE_TUNNEL: 2056 case XFRM_MODE_BEET: 2057 return 4; 2058 } 2059 return 5; 2060 } 2061 2062 /* Rule for xfrm_tmpl: 2063 * 2064 * rule 1: select IPsec transport 2065 * rule 2: select MIPv6 RO or inbound trigger 2066 * rule 3: select IPsec tunnel 2067 * rule 4: others 2068 */ 2069 static int __xfrm6_tmpl_sort_cmp(const void *p) 2070 { 2071 const struct xfrm_tmpl *v = p; 2072 2073 switch (v->mode) { 2074 case XFRM_MODE_TRANSPORT: 2075 return 1; 2076 #if IS_ENABLED(CONFIG_IPV6_MIP6) 2077 case XFRM_MODE_ROUTEOPTIMIZATION: 2078 case XFRM_MODE_IN_TRIGGER: 2079 return 2; 2080 #endif 2081 case XFRM_MODE_TUNNEL: 2082 case XFRM_MODE_BEET: 2083 return 3; 2084 } 2085 return 4; 2086 } 2087 #else 2088 static inline int __xfrm6_state_sort_cmp(const void *p) { return 5; } 2089 static inline int __xfrm6_tmpl_sort_cmp(const void *p) { return 4; } 2090 2091 static inline void 2092 __xfrm6_sort(void **dst, void **src, int n, 2093 int (*cmp)(const void *p), int maxclass) 2094 { 2095 int i; 2096 2097 for (i = 0; i < n; i++) 2098 dst[i] = src[i]; 2099 } 2100 #endif /* CONFIG_IPV6 */ 2101 2102 void 2103 xfrm_tmpl_sort(struct xfrm_tmpl **dst, struct xfrm_tmpl **src, int n, 2104 unsigned short family) 2105 { 2106 int i; 2107 2108 if (family == AF_INET6) 2109 __xfrm6_sort((void **)dst, (void **)src, n, 2110 __xfrm6_tmpl_sort_cmp, 5); 2111 else 2112 for (i = 0; i < n; i++) 2113 dst[i] = src[i]; 2114 } 2115 2116 void 2117 xfrm_state_sort(struct xfrm_state **dst, struct xfrm_state **src, int n, 2118 unsigned short family) 2119 { 2120 int i; 2121 2122 if (family == AF_INET6) 2123 __xfrm6_sort((void **)dst, (void **)src, n, 2124 __xfrm6_state_sort_cmp, 6); 2125 else 2126 for (i = 0; i < n; i++) 2127 dst[i] = src[i]; 2128 } 2129 #endif 2130 2131 /* Silly enough, but I'm lazy to build resolution list */ 2132 2133 static struct xfrm_state *__xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq) 2134 { 2135 unsigned int h = xfrm_seq_hash(net, seq); 2136 struct xfrm_state *x; 2137 2138 hlist_for_each_entry_rcu(x, net->xfrm.state_byseq + h, byseq) { 2139 if (x->km.seq == seq && 2140 (mark & x->mark.m) == x->mark.v && 2141 x->km.state == XFRM_STATE_ACQ) { 2142 xfrm_state_hold(x); 2143 return x; 2144 } 2145 } 2146 2147 return NULL; 2148 } 2149 2150 struct xfrm_state *xfrm_find_acq_byseq(struct net *net, u32 mark, u32 seq) 2151 { 2152 struct xfrm_state *x; 2153 2154 spin_lock_bh(&net->xfrm.xfrm_state_lock); 2155 x = __xfrm_find_acq_byseq(net, mark, seq); 2156 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 2157 return x; 2158 } 2159 EXPORT_SYMBOL(xfrm_find_acq_byseq); 2160 2161 u32 xfrm_get_acqseq(void) 2162 { 2163 u32 res; 2164 static atomic_t acqseq; 2165 2166 do { 2167 res = atomic_inc_return(&acqseq); 2168 } while (!res); 2169 2170 return res; 2171 } 2172 EXPORT_SYMBOL(xfrm_get_acqseq); 2173 2174 int verify_spi_info(u8 proto, u32 min, u32 max, struct netlink_ext_ack *extack) 2175 { 2176 switch (proto) { 2177 case IPPROTO_AH: 2178 case IPPROTO_ESP: 2179 break; 2180 2181 case IPPROTO_COMP: 2182 /* IPCOMP spi is 16-bits. */ 2183 if (max >= 0x10000) { 2184 NL_SET_ERR_MSG(extack, "IPCOMP SPI must be <= 65535"); 2185 return -EINVAL; 2186 } 2187 break; 2188 2189 default: 2190 NL_SET_ERR_MSG(extack, "Invalid protocol, must be one of AH, ESP, IPCOMP"); 2191 return -EINVAL; 2192 } 2193 2194 if (min > max) { 2195 NL_SET_ERR_MSG(extack, "Invalid SPI range: min > max"); 2196 return -EINVAL; 2197 } 2198 2199 return 0; 2200 } 2201 EXPORT_SYMBOL(verify_spi_info); 2202 2203 int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high, 2204 struct netlink_ext_ack *extack) 2205 { 2206 struct net *net = xs_net(x); 2207 unsigned int h; 2208 struct xfrm_state *x0; 2209 int err = -ENOENT; 2210 __be32 minspi = htonl(low); 2211 __be32 maxspi = htonl(high); 2212 __be32 newspi = 0; 2213 u32 mark = x->mark.v & x->mark.m; 2214 2215 spin_lock_bh(&x->lock); 2216 if (x->km.state == XFRM_STATE_DEAD) { 2217 NL_SET_ERR_MSG(extack, "Target ACQUIRE is in DEAD state"); 2218 goto unlock; 2219 } 2220 2221 err = 0; 2222 if (x->id.spi) 2223 goto unlock; 2224 2225 err = -ENOENT; 2226 2227 if (minspi == maxspi) { 2228 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, minspi, x->id.proto, x->props.family); 2229 if (x0) { 2230 NL_SET_ERR_MSG(extack, "Requested SPI is already in use"); 2231 xfrm_state_put(x0); 2232 goto unlock; 2233 } 2234 newspi = minspi; 2235 } else { 2236 u32 spi = 0; 2237 for (h = 0; h < high-low+1; h++) { 2238 spi = get_random_u32_inclusive(low, high); 2239 x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family); 2240 if (x0 == NULL) { 2241 newspi = htonl(spi); 2242 break; 2243 } 2244 xfrm_state_put(x0); 2245 } 2246 } 2247 if (newspi) { 2248 spin_lock_bh(&net->xfrm.xfrm_state_lock); 2249 x->id.spi = newspi; 2250 h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family); 2251 XFRM_STATE_INSERT(byspi, &x->byspi, net->xfrm.state_byspi + h, 2252 x->xso.type); 2253 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 2254 2255 err = 0; 2256 } else { 2257 NL_SET_ERR_MSG(extack, "No SPI available in the requested range"); 2258 } 2259 2260 unlock: 2261 spin_unlock_bh(&x->lock); 2262 2263 return err; 2264 } 2265 EXPORT_SYMBOL(xfrm_alloc_spi); 2266 2267 static bool __xfrm_state_filter_match(struct xfrm_state *x, 2268 struct xfrm_address_filter *filter) 2269 { 2270 if (filter) { 2271 if ((filter->family == AF_INET || 2272 filter->family == AF_INET6) && 2273 x->props.family != filter->family) 2274 return false; 2275 2276 return addr_match(&x->props.saddr, &filter->saddr, 2277 filter->splen) && 2278 addr_match(&x->id.daddr, &filter->daddr, 2279 filter->dplen); 2280 } 2281 return true; 2282 } 2283 2284 int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk, 2285 int (*func)(struct xfrm_state *, int, void*), 2286 void *data) 2287 { 2288 struct xfrm_state *state; 2289 struct xfrm_state_walk *x; 2290 int err = 0; 2291 2292 if (walk->seq != 0 && list_empty(&walk->all)) 2293 return 0; 2294 2295 spin_lock_bh(&net->xfrm.xfrm_state_lock); 2296 if (list_empty(&walk->all)) 2297 x = list_first_entry(&net->xfrm.state_all, struct xfrm_state_walk, all); 2298 else 2299 x = list_first_entry(&walk->all, struct xfrm_state_walk, all); 2300 list_for_each_entry_from(x, &net->xfrm.state_all, all) { 2301 if (x->state == XFRM_STATE_DEAD) 2302 continue; 2303 state = container_of(x, struct xfrm_state, km); 2304 if (!xfrm_id_proto_match(state->id.proto, walk->proto)) 2305 continue; 2306 if (!__xfrm_state_filter_match(state, walk->filter)) 2307 continue; 2308 err = func(state, walk->seq, data); 2309 if (err) { 2310 list_move_tail(&walk->all, &x->all); 2311 goto out; 2312 } 2313 walk->seq++; 2314 } 2315 if (walk->seq == 0) { 2316 err = -ENOENT; 2317 goto out; 2318 } 2319 list_del_init(&walk->all); 2320 out: 2321 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 2322 return err; 2323 } 2324 EXPORT_SYMBOL(xfrm_state_walk); 2325 2326 void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto, 2327 struct xfrm_address_filter *filter) 2328 { 2329 INIT_LIST_HEAD(&walk->all); 2330 walk->proto = proto; 2331 walk->state = XFRM_STATE_DEAD; 2332 walk->seq = 0; 2333 walk->filter = filter; 2334 } 2335 EXPORT_SYMBOL(xfrm_state_walk_init); 2336 2337 void xfrm_state_walk_done(struct xfrm_state_walk *walk, struct net *net) 2338 { 2339 kfree(walk->filter); 2340 2341 if (list_empty(&walk->all)) 2342 return; 2343 2344 spin_lock_bh(&net->xfrm.xfrm_state_lock); 2345 list_del(&walk->all); 2346 spin_unlock_bh(&net->xfrm.xfrm_state_lock); 2347 } 2348 EXPORT_SYMBOL(xfrm_state_walk_done); 2349 2350 static void xfrm_replay_timer_handler(struct timer_list *t) 2351 { 2352 struct xfrm_state *x = from_timer(x, t, rtimer); 2353 2354 spin_lock(&x->lock); 2355 2356 if (x->km.state == XFRM_STATE_VALID) { 2357 if (xfrm_aevent_is_on(xs_net(x))) 2358 xfrm_replay_notify(x, XFRM_REPLAY_TIMEOUT); 2359 else 2360 x->xflags |= XFRM_TIME_DEFER; 2361 } 2362 2363 spin_unlock(&x->lock); 2364 } 2365 2366 static LIST_HEAD(xfrm_km_list); 2367 2368 void km_policy_notify(struct xfrm_policy *xp, int dir, const struct km_event *c) 2369 { 2370 struct xfrm_mgr *km; 2371 2372 rcu_read_lock(); 2373 list_for_each_entry_rcu(km, &xfrm_km_list, list) 2374 if (km->notify_policy) 2375 km->notify_policy(xp, dir, c); 2376 rcu_read_unlock(); 2377 } 2378 2379 void km_state_notify(struct xfrm_state *x, const struct km_event *c) 2380 { 2381 struct xfrm_mgr *km; 2382 rcu_read_lock(); 2383 list_for_each_entry_rcu(km, &xfrm_km_list, list) 2384 if (km->notify) 2385 km->notify(x, c); 2386 rcu_read_unlock(); 2387 } 2388 2389 EXPORT_SYMBOL(km_policy_notify); 2390 EXPORT_SYMBOL(km_state_notify); 2391 2392 void km_state_expired(struct xfrm_state *x, int hard, u32 portid) 2393 { 2394 struct km_event c; 2395 2396 c.data.hard = hard; 2397 c.portid = portid; 2398 c.event = XFRM_MSG_EXPIRE; 2399 km_state_notify(x, &c); 2400 } 2401 2402 EXPORT_SYMBOL(km_state_expired); 2403 /* 2404 * We send to all registered managers regardless of failure 2405 * We are happy with one success 2406 */ 2407 int km_query(struct xfrm_state *x, struct xfrm_tmpl *t, struct xfrm_policy *pol) 2408 { 2409 int err = -EINVAL, acqret; 2410 struct xfrm_mgr *km; 2411 2412 rcu_read_lock(); 2413 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2414 acqret = km->acquire(x, t, pol); 2415 if (!acqret) 2416 err = acqret; 2417 } 2418 rcu_read_unlock(); 2419 return err; 2420 } 2421 EXPORT_SYMBOL(km_query); 2422 2423 static int __km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport) 2424 { 2425 int err = -EINVAL; 2426 struct xfrm_mgr *km; 2427 2428 rcu_read_lock(); 2429 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2430 if (km->new_mapping) 2431 err = km->new_mapping(x, ipaddr, sport); 2432 if (!err) 2433 break; 2434 } 2435 rcu_read_unlock(); 2436 return err; 2437 } 2438 2439 int km_new_mapping(struct xfrm_state *x, xfrm_address_t *ipaddr, __be16 sport) 2440 { 2441 int ret = 0; 2442 2443 if (x->mapping_maxage) { 2444 if ((jiffies / HZ - x->new_mapping) > x->mapping_maxage || 2445 x->new_mapping_sport != sport) { 2446 x->new_mapping_sport = sport; 2447 x->new_mapping = jiffies / HZ; 2448 ret = __km_new_mapping(x, ipaddr, sport); 2449 } 2450 } else { 2451 ret = __km_new_mapping(x, ipaddr, sport); 2452 } 2453 2454 return ret; 2455 } 2456 EXPORT_SYMBOL(km_new_mapping); 2457 2458 void km_policy_expired(struct xfrm_policy *pol, int dir, int hard, u32 portid) 2459 { 2460 struct km_event c; 2461 2462 c.data.hard = hard; 2463 c.portid = portid; 2464 c.event = XFRM_MSG_POLEXPIRE; 2465 km_policy_notify(pol, dir, &c); 2466 } 2467 EXPORT_SYMBOL(km_policy_expired); 2468 2469 #ifdef CONFIG_XFRM_MIGRATE 2470 int km_migrate(const struct xfrm_selector *sel, u8 dir, u8 type, 2471 const struct xfrm_migrate *m, int num_migrate, 2472 const struct xfrm_kmaddress *k, 2473 const struct xfrm_encap_tmpl *encap) 2474 { 2475 int err = -EINVAL; 2476 int ret; 2477 struct xfrm_mgr *km; 2478 2479 rcu_read_lock(); 2480 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2481 if (km->migrate) { 2482 ret = km->migrate(sel, dir, type, m, num_migrate, k, 2483 encap); 2484 if (!ret) 2485 err = ret; 2486 } 2487 } 2488 rcu_read_unlock(); 2489 return err; 2490 } 2491 EXPORT_SYMBOL(km_migrate); 2492 #endif 2493 2494 int km_report(struct net *net, u8 proto, struct xfrm_selector *sel, xfrm_address_t *addr) 2495 { 2496 int err = -EINVAL; 2497 int ret; 2498 struct xfrm_mgr *km; 2499 2500 rcu_read_lock(); 2501 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2502 if (km->report) { 2503 ret = km->report(net, proto, sel, addr); 2504 if (!ret) 2505 err = ret; 2506 } 2507 } 2508 rcu_read_unlock(); 2509 return err; 2510 } 2511 EXPORT_SYMBOL(km_report); 2512 2513 static bool km_is_alive(const struct km_event *c) 2514 { 2515 struct xfrm_mgr *km; 2516 bool is_alive = false; 2517 2518 rcu_read_lock(); 2519 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2520 if (km->is_alive && km->is_alive(c)) { 2521 is_alive = true; 2522 break; 2523 } 2524 } 2525 rcu_read_unlock(); 2526 2527 return is_alive; 2528 } 2529 2530 #if IS_ENABLED(CONFIG_XFRM_USER_COMPAT) 2531 static DEFINE_SPINLOCK(xfrm_translator_lock); 2532 static struct xfrm_translator __rcu *xfrm_translator; 2533 2534 struct xfrm_translator *xfrm_get_translator(void) 2535 { 2536 struct xfrm_translator *xtr; 2537 2538 rcu_read_lock(); 2539 xtr = rcu_dereference(xfrm_translator); 2540 if (unlikely(!xtr)) 2541 goto out; 2542 if (!try_module_get(xtr->owner)) 2543 xtr = NULL; 2544 out: 2545 rcu_read_unlock(); 2546 return xtr; 2547 } 2548 EXPORT_SYMBOL_GPL(xfrm_get_translator); 2549 2550 void xfrm_put_translator(struct xfrm_translator *xtr) 2551 { 2552 module_put(xtr->owner); 2553 } 2554 EXPORT_SYMBOL_GPL(xfrm_put_translator); 2555 2556 int xfrm_register_translator(struct xfrm_translator *xtr) 2557 { 2558 int err = 0; 2559 2560 spin_lock_bh(&xfrm_translator_lock); 2561 if (unlikely(xfrm_translator != NULL)) 2562 err = -EEXIST; 2563 else 2564 rcu_assign_pointer(xfrm_translator, xtr); 2565 spin_unlock_bh(&xfrm_translator_lock); 2566 2567 return err; 2568 } 2569 EXPORT_SYMBOL_GPL(xfrm_register_translator); 2570 2571 int xfrm_unregister_translator(struct xfrm_translator *xtr) 2572 { 2573 int err = 0; 2574 2575 spin_lock_bh(&xfrm_translator_lock); 2576 if (likely(xfrm_translator != NULL)) { 2577 if (rcu_access_pointer(xfrm_translator) != xtr) 2578 err = -EINVAL; 2579 else 2580 RCU_INIT_POINTER(xfrm_translator, NULL); 2581 } 2582 spin_unlock_bh(&xfrm_translator_lock); 2583 synchronize_rcu(); 2584 2585 return err; 2586 } 2587 EXPORT_SYMBOL_GPL(xfrm_unregister_translator); 2588 #endif 2589 2590 int xfrm_user_policy(struct sock *sk, int optname, sockptr_t optval, int optlen) 2591 { 2592 int err; 2593 u8 *data; 2594 struct xfrm_mgr *km; 2595 struct xfrm_policy *pol = NULL; 2596 2597 if (sockptr_is_null(optval) && !optlen) { 2598 xfrm_sk_policy_insert(sk, XFRM_POLICY_IN, NULL); 2599 xfrm_sk_policy_insert(sk, XFRM_POLICY_OUT, NULL); 2600 __sk_dst_reset(sk); 2601 return 0; 2602 } 2603 2604 if (optlen <= 0 || optlen > PAGE_SIZE) 2605 return -EMSGSIZE; 2606 2607 data = memdup_sockptr(optval, optlen); 2608 if (IS_ERR(data)) 2609 return PTR_ERR(data); 2610 2611 if (in_compat_syscall()) { 2612 struct xfrm_translator *xtr = xfrm_get_translator(); 2613 2614 if (!xtr) { 2615 kfree(data); 2616 return -EOPNOTSUPP; 2617 } 2618 2619 err = xtr->xlate_user_policy_sockptr(&data, optlen); 2620 xfrm_put_translator(xtr); 2621 if (err) { 2622 kfree(data); 2623 return err; 2624 } 2625 } 2626 2627 err = -EINVAL; 2628 rcu_read_lock(); 2629 list_for_each_entry_rcu(km, &xfrm_km_list, list) { 2630 pol = km->compile_policy(sk, optname, data, 2631 optlen, &err); 2632 if (err >= 0) 2633 break; 2634 } 2635 rcu_read_unlock(); 2636 2637 if (err >= 0) { 2638 xfrm_sk_policy_insert(sk, err, pol); 2639 xfrm_pol_put(pol); 2640 __sk_dst_reset(sk); 2641 err = 0; 2642 } 2643 2644 kfree(data); 2645 return err; 2646 } 2647 EXPORT_SYMBOL(xfrm_user_policy); 2648 2649 static DEFINE_SPINLOCK(xfrm_km_lock); 2650 2651 void xfrm_register_km(struct xfrm_mgr *km) 2652 { 2653 spin_lock_bh(&xfrm_km_lock); 2654 list_add_tail_rcu(&km->list, &xfrm_km_list); 2655 spin_unlock_bh(&xfrm_km_lock); 2656 } 2657 EXPORT_SYMBOL(xfrm_register_km); 2658 2659 void xfrm_unregister_km(struct xfrm_mgr *km) 2660 { 2661 spin_lock_bh(&xfrm_km_lock); 2662 list_del_rcu(&km->list); 2663 spin_unlock_bh(&xfrm_km_lock); 2664 synchronize_rcu(); 2665 } 2666 EXPORT_SYMBOL(xfrm_unregister_km); 2667 2668 int xfrm_state_register_afinfo(struct xfrm_state_afinfo *afinfo) 2669 { 2670 int err = 0; 2671 2672 if (WARN_ON(afinfo->family >= NPROTO)) 2673 return -EAFNOSUPPORT; 2674 2675 spin_lock_bh(&xfrm_state_afinfo_lock); 2676 if (unlikely(xfrm_state_afinfo[afinfo->family] != NULL)) 2677 err = -EEXIST; 2678 else 2679 rcu_assign_pointer(xfrm_state_afinfo[afinfo->family], afinfo); 2680 spin_unlock_bh(&xfrm_state_afinfo_lock); 2681 return err; 2682 } 2683 EXPORT_SYMBOL(xfrm_state_register_afinfo); 2684 2685 int xfrm_state_unregister_afinfo(struct xfrm_state_afinfo *afinfo) 2686 { 2687 int err = 0, family = afinfo->family; 2688 2689 if (WARN_ON(family >= NPROTO)) 2690 return -EAFNOSUPPORT; 2691 2692 spin_lock_bh(&xfrm_state_afinfo_lock); 2693 if (likely(xfrm_state_afinfo[afinfo->family] != NULL)) { 2694 if (rcu_access_pointer(xfrm_state_afinfo[family]) != afinfo) 2695 err = -EINVAL; 2696 else 2697 RCU_INIT_POINTER(xfrm_state_afinfo[afinfo->family], NULL); 2698 } 2699 spin_unlock_bh(&xfrm_state_afinfo_lock); 2700 synchronize_rcu(); 2701 return err; 2702 } 2703 EXPORT_SYMBOL(xfrm_state_unregister_afinfo); 2704 2705 struct xfrm_state_afinfo *xfrm_state_afinfo_get_rcu(unsigned int family) 2706 { 2707 if (unlikely(family >= NPROTO)) 2708 return NULL; 2709 2710 return rcu_dereference(xfrm_state_afinfo[family]); 2711 } 2712 EXPORT_SYMBOL_GPL(xfrm_state_afinfo_get_rcu); 2713 2714 struct xfrm_state_afinfo *xfrm_state_get_afinfo(unsigned int family) 2715 { 2716 struct xfrm_state_afinfo *afinfo; 2717 if (unlikely(family >= NPROTO)) 2718 return NULL; 2719 rcu_read_lock(); 2720 afinfo = rcu_dereference(xfrm_state_afinfo[family]); 2721 if (unlikely(!afinfo)) 2722 rcu_read_unlock(); 2723 return afinfo; 2724 } 2725 2726 void xfrm_flush_gc(void) 2727 { 2728 flush_work(&xfrm_state_gc_work); 2729 } 2730 EXPORT_SYMBOL(xfrm_flush_gc); 2731 2732 /* Temporarily located here until net/xfrm/xfrm_tunnel.c is created */ 2733 void xfrm_state_delete_tunnel(struct xfrm_state *x) 2734 { 2735 if (x->tunnel) { 2736 struct xfrm_state *t = x->tunnel; 2737 2738 if (atomic_read(&t->tunnel_users) == 2) 2739 xfrm_state_delete(t); 2740 atomic_dec(&t->tunnel_users); 2741 xfrm_state_put_sync(t); 2742 x->tunnel = NULL; 2743 } 2744 } 2745 EXPORT_SYMBOL(xfrm_state_delete_tunnel); 2746 2747 u32 xfrm_state_mtu(struct xfrm_state *x, int mtu) 2748 { 2749 const struct xfrm_type *type = READ_ONCE(x->type); 2750 struct crypto_aead *aead; 2751 u32 blksize, net_adj = 0; 2752 2753 if (x->km.state != XFRM_STATE_VALID || 2754 !type || type->proto != IPPROTO_ESP) 2755 return mtu - x->props.header_len; 2756 2757 aead = x->data; 2758 blksize = ALIGN(crypto_aead_blocksize(aead), 4); 2759 2760 switch (x->props.mode) { 2761 case XFRM_MODE_TRANSPORT: 2762 case XFRM_MODE_BEET: 2763 if (x->props.family == AF_INET) 2764 net_adj = sizeof(struct iphdr); 2765 else if (x->props.family == AF_INET6) 2766 net_adj = sizeof(struct ipv6hdr); 2767 break; 2768 case XFRM_MODE_TUNNEL: 2769 break; 2770 default: 2771 WARN_ON_ONCE(1); 2772 break; 2773 } 2774 2775 return ((mtu - x->props.header_len - crypto_aead_authsize(aead) - 2776 net_adj) & ~(blksize - 1)) + net_adj - 2; 2777 } 2778 EXPORT_SYMBOL_GPL(xfrm_state_mtu); 2779 2780 int __xfrm_init_state(struct xfrm_state *x, bool init_replay, bool offload, 2781 struct netlink_ext_ack *extack) 2782 { 2783 const struct xfrm_mode *inner_mode; 2784 const struct xfrm_mode *outer_mode; 2785 int family = x->props.family; 2786 int err; 2787 2788 if (family == AF_INET && 2789 READ_ONCE(xs_net(x)->ipv4.sysctl_ip_no_pmtu_disc)) 2790 x->props.flags |= XFRM_STATE_NOPMTUDISC; 2791 2792 err = -EPROTONOSUPPORT; 2793 2794 if (x->sel.family != AF_UNSPEC) { 2795 inner_mode = xfrm_get_mode(x->props.mode, x->sel.family); 2796 if (inner_mode == NULL) { 2797 NL_SET_ERR_MSG(extack, "Requested mode not found"); 2798 goto error; 2799 } 2800 2801 if (!(inner_mode->flags & XFRM_MODE_FLAG_TUNNEL) && 2802 family != x->sel.family) { 2803 NL_SET_ERR_MSG(extack, "Only tunnel modes can accommodate a change of family"); 2804 goto error; 2805 } 2806 2807 x->inner_mode = *inner_mode; 2808 } else { 2809 const struct xfrm_mode *inner_mode_iaf; 2810 int iafamily = AF_INET; 2811 2812 inner_mode = xfrm_get_mode(x->props.mode, x->props.family); 2813 if (inner_mode == NULL) { 2814 NL_SET_ERR_MSG(extack, "Requested mode not found"); 2815 goto error; 2816 } 2817 2818 x->inner_mode = *inner_mode; 2819 2820 if (x->props.family == AF_INET) 2821 iafamily = AF_INET6; 2822 2823 inner_mode_iaf = xfrm_get_mode(x->props.mode, iafamily); 2824 if (inner_mode_iaf) { 2825 if (inner_mode_iaf->flags & XFRM_MODE_FLAG_TUNNEL) 2826 x->inner_mode_iaf = *inner_mode_iaf; 2827 } 2828 } 2829 2830 x->type = xfrm_get_type(x->id.proto, family); 2831 if (x->type == NULL) { 2832 NL_SET_ERR_MSG(extack, "Requested type not found"); 2833 goto error; 2834 } 2835 2836 x->type_offload = xfrm_get_type_offload(x->id.proto, family, offload); 2837 2838 err = x->type->init_state(x, extack); 2839 if (err) 2840 goto error; 2841 2842 outer_mode = xfrm_get_mode(x->props.mode, family); 2843 if (!outer_mode) { 2844 NL_SET_ERR_MSG(extack, "Requested mode not found"); 2845 err = -EPROTONOSUPPORT; 2846 goto error; 2847 } 2848 2849 x->outer_mode = *outer_mode; 2850 if (init_replay) { 2851 err = xfrm_init_replay(x, extack); 2852 if (err) 2853 goto error; 2854 } 2855 2856 error: 2857 return err; 2858 } 2859 2860 EXPORT_SYMBOL(__xfrm_init_state); 2861 2862 int xfrm_init_state(struct xfrm_state *x) 2863 { 2864 int err; 2865 2866 err = __xfrm_init_state(x, true, false, NULL); 2867 if (!err) 2868 x->km.state = XFRM_STATE_VALID; 2869 2870 return err; 2871 } 2872 2873 EXPORT_SYMBOL(xfrm_init_state); 2874 2875 int __net_init xfrm_state_init(struct net *net) 2876 { 2877 unsigned int sz; 2878 2879 if (net_eq(net, &init_net)) 2880 xfrm_state_cache = KMEM_CACHE(xfrm_state, 2881 SLAB_HWCACHE_ALIGN | SLAB_PANIC); 2882 2883 INIT_LIST_HEAD(&net->xfrm.state_all); 2884 2885 sz = sizeof(struct hlist_head) * 8; 2886 2887 net->xfrm.state_bydst = xfrm_hash_alloc(sz); 2888 if (!net->xfrm.state_bydst) 2889 goto out_bydst; 2890 net->xfrm.state_bysrc = xfrm_hash_alloc(sz); 2891 if (!net->xfrm.state_bysrc) 2892 goto out_bysrc; 2893 net->xfrm.state_byspi = xfrm_hash_alloc(sz); 2894 if (!net->xfrm.state_byspi) 2895 goto out_byspi; 2896 net->xfrm.state_byseq = xfrm_hash_alloc(sz); 2897 if (!net->xfrm.state_byseq) 2898 goto out_byseq; 2899 net->xfrm.state_hmask = ((sz / sizeof(struct hlist_head)) - 1); 2900 2901 net->xfrm.state_num = 0; 2902 INIT_WORK(&net->xfrm.state_hash_work, xfrm_hash_resize); 2903 spin_lock_init(&net->xfrm.xfrm_state_lock); 2904 seqcount_spinlock_init(&net->xfrm.xfrm_state_hash_generation, 2905 &net->xfrm.xfrm_state_lock); 2906 return 0; 2907 2908 out_byseq: 2909 xfrm_hash_free(net->xfrm.state_byspi, sz); 2910 out_byspi: 2911 xfrm_hash_free(net->xfrm.state_bysrc, sz); 2912 out_bysrc: 2913 xfrm_hash_free(net->xfrm.state_bydst, sz); 2914 out_bydst: 2915 return -ENOMEM; 2916 } 2917 2918 void xfrm_state_fini(struct net *net) 2919 { 2920 unsigned int sz; 2921 2922 flush_work(&net->xfrm.state_hash_work); 2923 flush_work(&xfrm_state_gc_work); 2924 xfrm_state_flush(net, 0, false, true); 2925 2926 WARN_ON(!list_empty(&net->xfrm.state_all)); 2927 2928 sz = (net->xfrm.state_hmask + 1) * sizeof(struct hlist_head); 2929 WARN_ON(!hlist_empty(net->xfrm.state_byseq)); 2930 xfrm_hash_free(net->xfrm.state_byseq, sz); 2931 WARN_ON(!hlist_empty(net->xfrm.state_byspi)); 2932 xfrm_hash_free(net->xfrm.state_byspi, sz); 2933 WARN_ON(!hlist_empty(net->xfrm.state_bysrc)); 2934 xfrm_hash_free(net->xfrm.state_bysrc, sz); 2935 WARN_ON(!hlist_empty(net->xfrm.state_bydst)); 2936 xfrm_hash_free(net->xfrm.state_bydst, sz); 2937 } 2938 2939 #ifdef CONFIG_AUDITSYSCALL 2940 static void xfrm_audit_helper_sainfo(struct xfrm_state *x, 2941 struct audit_buffer *audit_buf) 2942 { 2943 struct xfrm_sec_ctx *ctx = x->security; 2944 u32 spi = ntohl(x->id.spi); 2945 2946 if (ctx) 2947 audit_log_format(audit_buf, " sec_alg=%u sec_doi=%u sec_obj=%s", 2948 ctx->ctx_alg, ctx->ctx_doi, ctx->ctx_str); 2949 2950 switch (x->props.family) { 2951 case AF_INET: 2952 audit_log_format(audit_buf, " src=%pI4 dst=%pI4", 2953 &x->props.saddr.a4, &x->id.daddr.a4); 2954 break; 2955 case AF_INET6: 2956 audit_log_format(audit_buf, " src=%pI6 dst=%pI6", 2957 x->props.saddr.a6, x->id.daddr.a6); 2958 break; 2959 } 2960 2961 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi); 2962 } 2963 2964 static void xfrm_audit_helper_pktinfo(struct sk_buff *skb, u16 family, 2965 struct audit_buffer *audit_buf) 2966 { 2967 const struct iphdr *iph4; 2968 const struct ipv6hdr *iph6; 2969 2970 switch (family) { 2971 case AF_INET: 2972 iph4 = ip_hdr(skb); 2973 audit_log_format(audit_buf, " src=%pI4 dst=%pI4", 2974 &iph4->saddr, &iph4->daddr); 2975 break; 2976 case AF_INET6: 2977 iph6 = ipv6_hdr(skb); 2978 audit_log_format(audit_buf, 2979 " src=%pI6 dst=%pI6 flowlbl=0x%x%02x%02x", 2980 &iph6->saddr, &iph6->daddr, 2981 iph6->flow_lbl[0] & 0x0f, 2982 iph6->flow_lbl[1], 2983 iph6->flow_lbl[2]); 2984 break; 2985 } 2986 } 2987 2988 void xfrm_audit_state_add(struct xfrm_state *x, int result, bool task_valid) 2989 { 2990 struct audit_buffer *audit_buf; 2991 2992 audit_buf = xfrm_audit_start("SAD-add"); 2993 if (audit_buf == NULL) 2994 return; 2995 xfrm_audit_helper_usrinfo(task_valid, audit_buf); 2996 xfrm_audit_helper_sainfo(x, audit_buf); 2997 audit_log_format(audit_buf, " res=%u", result); 2998 audit_log_end(audit_buf); 2999 } 3000 EXPORT_SYMBOL_GPL(xfrm_audit_state_add); 3001 3002 void xfrm_audit_state_delete(struct xfrm_state *x, int result, bool task_valid) 3003 { 3004 struct audit_buffer *audit_buf; 3005 3006 audit_buf = xfrm_audit_start("SAD-delete"); 3007 if (audit_buf == NULL) 3008 return; 3009 xfrm_audit_helper_usrinfo(task_valid, audit_buf); 3010 xfrm_audit_helper_sainfo(x, audit_buf); 3011 audit_log_format(audit_buf, " res=%u", result); 3012 audit_log_end(audit_buf); 3013 } 3014 EXPORT_SYMBOL_GPL(xfrm_audit_state_delete); 3015 3016 void xfrm_audit_state_replay_overflow(struct xfrm_state *x, 3017 struct sk_buff *skb) 3018 { 3019 struct audit_buffer *audit_buf; 3020 u32 spi; 3021 3022 audit_buf = xfrm_audit_start("SA-replay-overflow"); 3023 if (audit_buf == NULL) 3024 return; 3025 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf); 3026 /* don't record the sequence number because it's inherent in this kind 3027 * of audit message */ 3028 spi = ntohl(x->id.spi); 3029 audit_log_format(audit_buf, " spi=%u(0x%x)", spi, spi); 3030 audit_log_end(audit_buf); 3031 } 3032 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay_overflow); 3033 3034 void xfrm_audit_state_replay(struct xfrm_state *x, 3035 struct sk_buff *skb, __be32 net_seq) 3036 { 3037 struct audit_buffer *audit_buf; 3038 u32 spi; 3039 3040 audit_buf = xfrm_audit_start("SA-replayed-pkt"); 3041 if (audit_buf == NULL) 3042 return; 3043 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf); 3044 spi = ntohl(x->id.spi); 3045 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u", 3046 spi, spi, ntohl(net_seq)); 3047 audit_log_end(audit_buf); 3048 } 3049 EXPORT_SYMBOL_GPL(xfrm_audit_state_replay); 3050 3051 void xfrm_audit_state_notfound_simple(struct sk_buff *skb, u16 family) 3052 { 3053 struct audit_buffer *audit_buf; 3054 3055 audit_buf = xfrm_audit_start("SA-notfound"); 3056 if (audit_buf == NULL) 3057 return; 3058 xfrm_audit_helper_pktinfo(skb, family, audit_buf); 3059 audit_log_end(audit_buf); 3060 } 3061 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound_simple); 3062 3063 void xfrm_audit_state_notfound(struct sk_buff *skb, u16 family, 3064 __be32 net_spi, __be32 net_seq) 3065 { 3066 struct audit_buffer *audit_buf; 3067 u32 spi; 3068 3069 audit_buf = xfrm_audit_start("SA-notfound"); 3070 if (audit_buf == NULL) 3071 return; 3072 xfrm_audit_helper_pktinfo(skb, family, audit_buf); 3073 spi = ntohl(net_spi); 3074 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u", 3075 spi, spi, ntohl(net_seq)); 3076 audit_log_end(audit_buf); 3077 } 3078 EXPORT_SYMBOL_GPL(xfrm_audit_state_notfound); 3079 3080 void xfrm_audit_state_icvfail(struct xfrm_state *x, 3081 struct sk_buff *skb, u8 proto) 3082 { 3083 struct audit_buffer *audit_buf; 3084 __be32 net_spi; 3085 __be32 net_seq; 3086 3087 audit_buf = xfrm_audit_start("SA-icv-failure"); 3088 if (audit_buf == NULL) 3089 return; 3090 xfrm_audit_helper_pktinfo(skb, x->props.family, audit_buf); 3091 if (xfrm_parse_spi(skb, proto, &net_spi, &net_seq) == 0) { 3092 u32 spi = ntohl(net_spi); 3093 audit_log_format(audit_buf, " spi=%u(0x%x) seqno=%u", 3094 spi, spi, ntohl(net_seq)); 3095 } 3096 audit_log_end(audit_buf); 3097 } 3098 EXPORT_SYMBOL_GPL(xfrm_audit_state_icvfail); 3099 #endif /* CONFIG_AUDITSYSCALL */ 3100