1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Copyright (c) 2007-2009 Patrick McHardy <kaber@trash.net> 4 * 5 * Development of this code funded by Astaro AG (http://www.astaro.com/) 6 */ 7 8 #include <linux/module.h> 9 #include <linux/init.h> 10 #include <linux/list.h> 11 #include <linux/skbuff.h> 12 #include <linux/netlink.h> 13 #include <linux/vmalloc.h> 14 #include <linux/rhashtable.h> 15 #include <linux/audit.h> 16 #include <linux/netfilter.h> 17 #include <linux/netfilter/nfnetlink.h> 18 #include <linux/netfilter/nf_tables.h> 19 #include <net/netfilter/nf_flow_table.h> 20 #include <net/netfilter/nf_tables_core.h> 21 #include <net/netfilter/nf_tables.h> 22 #include <net/netfilter/nf_tables_offload.h> 23 #include <net/net_namespace.h> 24 #include <net/sock.h> 25 26 #define NFT_MODULE_AUTOLOAD_LIMIT (MODULE_NAME_LEN - sizeof("nft-expr-255-")) 27 28 unsigned int nf_tables_net_id __read_mostly; 29 30 static LIST_HEAD(nf_tables_expressions); 31 static LIST_HEAD(nf_tables_objects); 32 static LIST_HEAD(nf_tables_flowtables); 33 static LIST_HEAD(nf_tables_destroy_list); 34 static DEFINE_SPINLOCK(nf_tables_destroy_list_lock); 35 36 enum { 37 NFT_VALIDATE_SKIP = 0, 38 NFT_VALIDATE_NEED, 39 NFT_VALIDATE_DO, 40 }; 41 42 static struct rhltable nft_objname_ht; 43 44 static u32 nft_chain_hash(const void *data, u32 len, u32 seed); 45 static u32 nft_chain_hash_obj(const void *data, u32 len, u32 seed); 46 static int nft_chain_hash_cmp(struct rhashtable_compare_arg *, const void *); 47 48 static u32 nft_objname_hash(const void *data, u32 len, u32 seed); 49 static u32 nft_objname_hash_obj(const void *data, u32 len, u32 seed); 50 static int nft_objname_hash_cmp(struct rhashtable_compare_arg *, const void *); 51 52 static const struct rhashtable_params nft_chain_ht_params = { 53 .head_offset = offsetof(struct nft_chain, rhlhead), 54 .key_offset = offsetof(struct nft_chain, name), 55 .hashfn = nft_chain_hash, 56 .obj_hashfn = nft_chain_hash_obj, 57 .obj_cmpfn = nft_chain_hash_cmp, 58 .automatic_shrinking = true, 59 }; 60 61 static const struct rhashtable_params nft_objname_ht_params = { 62 .head_offset = offsetof(struct nft_object, rhlhead), 63 .key_offset = offsetof(struct nft_object, key), 64 .hashfn = nft_objname_hash, 65 .obj_hashfn = nft_objname_hash_obj, 66 .obj_cmpfn = nft_objname_hash_cmp, 67 .automatic_shrinking = true, 68 }; 69 70 struct nft_audit_data { 71 struct nft_table *table; 72 int entries; 73 int op; 74 struct list_head list; 75 }; 76 77 static const u8 nft2audit_op[NFT_MSG_MAX] = { // enum nf_tables_msg_types 78 [NFT_MSG_NEWTABLE] = AUDIT_NFT_OP_TABLE_REGISTER, 79 [NFT_MSG_GETTABLE] = AUDIT_NFT_OP_INVALID, 80 [NFT_MSG_DELTABLE] = AUDIT_NFT_OP_TABLE_UNREGISTER, 81 [NFT_MSG_NEWCHAIN] = AUDIT_NFT_OP_CHAIN_REGISTER, 82 [NFT_MSG_GETCHAIN] = AUDIT_NFT_OP_INVALID, 83 [NFT_MSG_DELCHAIN] = AUDIT_NFT_OP_CHAIN_UNREGISTER, 84 [NFT_MSG_NEWRULE] = AUDIT_NFT_OP_RULE_REGISTER, 85 [NFT_MSG_GETRULE] = AUDIT_NFT_OP_INVALID, 86 [NFT_MSG_DELRULE] = AUDIT_NFT_OP_RULE_UNREGISTER, 87 [NFT_MSG_NEWSET] = AUDIT_NFT_OP_SET_REGISTER, 88 [NFT_MSG_GETSET] = AUDIT_NFT_OP_INVALID, 89 [NFT_MSG_DELSET] = AUDIT_NFT_OP_SET_UNREGISTER, 90 [NFT_MSG_NEWSETELEM] = AUDIT_NFT_OP_SETELEM_REGISTER, 91 [NFT_MSG_GETSETELEM] = AUDIT_NFT_OP_INVALID, 92 [NFT_MSG_DELSETELEM] = AUDIT_NFT_OP_SETELEM_UNREGISTER, 93 [NFT_MSG_NEWGEN] = AUDIT_NFT_OP_GEN_REGISTER, 94 [NFT_MSG_GETGEN] = AUDIT_NFT_OP_INVALID, 95 [NFT_MSG_TRACE] = AUDIT_NFT_OP_INVALID, 96 [NFT_MSG_NEWOBJ] = AUDIT_NFT_OP_OBJ_REGISTER, 97 [NFT_MSG_GETOBJ] = AUDIT_NFT_OP_INVALID, 98 [NFT_MSG_DELOBJ] = AUDIT_NFT_OP_OBJ_UNREGISTER, 99 [NFT_MSG_GETOBJ_RESET] = AUDIT_NFT_OP_OBJ_RESET, 100 [NFT_MSG_NEWFLOWTABLE] = AUDIT_NFT_OP_FLOWTABLE_REGISTER, 101 [NFT_MSG_GETFLOWTABLE] = AUDIT_NFT_OP_INVALID, 102 [NFT_MSG_DELFLOWTABLE] = AUDIT_NFT_OP_FLOWTABLE_UNREGISTER, 103 }; 104 105 static void nft_validate_state_update(struct net *net, u8 new_validate_state) 106 { 107 struct nftables_pernet *nft_net = nft_pernet(net); 108 109 switch (nft_net->validate_state) { 110 case NFT_VALIDATE_SKIP: 111 WARN_ON_ONCE(new_validate_state == NFT_VALIDATE_DO); 112 break; 113 case NFT_VALIDATE_NEED: 114 break; 115 case NFT_VALIDATE_DO: 116 if (new_validate_state == NFT_VALIDATE_NEED) 117 return; 118 } 119 120 nft_net->validate_state = new_validate_state; 121 } 122 static void nf_tables_trans_destroy_work(struct work_struct *w); 123 static DECLARE_WORK(trans_destroy_work, nf_tables_trans_destroy_work); 124 125 static void nft_ctx_init(struct nft_ctx *ctx, 126 struct net *net, 127 const struct sk_buff *skb, 128 const struct nlmsghdr *nlh, 129 u8 family, 130 struct nft_table *table, 131 struct nft_chain *chain, 132 const struct nlattr * const *nla) 133 { 134 ctx->net = net; 135 ctx->family = family; 136 ctx->level = 0; 137 ctx->table = table; 138 ctx->chain = chain; 139 ctx->nla = nla; 140 ctx->portid = NETLINK_CB(skb).portid; 141 ctx->report = nlmsg_report(nlh); 142 ctx->flags = nlh->nlmsg_flags; 143 ctx->seq = nlh->nlmsg_seq; 144 } 145 146 static struct nft_trans *nft_trans_alloc_gfp(const struct nft_ctx *ctx, 147 int msg_type, u32 size, gfp_t gfp) 148 { 149 struct nft_trans *trans; 150 151 trans = kzalloc(sizeof(struct nft_trans) + size, gfp); 152 if (trans == NULL) 153 return NULL; 154 155 INIT_LIST_HEAD(&trans->list); 156 trans->msg_type = msg_type; 157 trans->ctx = *ctx; 158 159 return trans; 160 } 161 162 static struct nft_trans *nft_trans_alloc(const struct nft_ctx *ctx, 163 int msg_type, u32 size) 164 { 165 return nft_trans_alloc_gfp(ctx, msg_type, size, GFP_KERNEL); 166 } 167 168 static void nft_trans_destroy(struct nft_trans *trans) 169 { 170 list_del(&trans->list); 171 kfree(trans); 172 } 173 174 static void nft_set_trans_bind(const struct nft_ctx *ctx, struct nft_set *set) 175 { 176 struct nftables_pernet *nft_net; 177 struct net *net = ctx->net; 178 struct nft_trans *trans; 179 180 if (!nft_set_is_anonymous(set)) 181 return; 182 183 nft_net = nft_pernet(net); 184 list_for_each_entry_reverse(trans, &nft_net->commit_list, list) { 185 switch (trans->msg_type) { 186 case NFT_MSG_NEWSET: 187 if (nft_trans_set(trans) == set) 188 nft_trans_set_bound(trans) = true; 189 break; 190 case NFT_MSG_NEWSETELEM: 191 if (nft_trans_elem_set(trans) == set) 192 nft_trans_elem_set_bound(trans) = true; 193 break; 194 } 195 } 196 } 197 198 static int nft_netdev_register_hooks(struct net *net, 199 struct list_head *hook_list) 200 { 201 struct nft_hook *hook; 202 int err, j; 203 204 j = 0; 205 list_for_each_entry(hook, hook_list, list) { 206 err = nf_register_net_hook(net, &hook->ops); 207 if (err < 0) 208 goto err_register; 209 210 j++; 211 } 212 return 0; 213 214 err_register: 215 list_for_each_entry(hook, hook_list, list) { 216 if (j-- <= 0) 217 break; 218 219 nf_unregister_net_hook(net, &hook->ops); 220 } 221 return err; 222 } 223 224 static void nft_netdev_unregister_hooks(struct net *net, 225 struct list_head *hook_list, 226 bool release_netdev) 227 { 228 struct nft_hook *hook, *next; 229 230 list_for_each_entry_safe(hook, next, hook_list, list) { 231 nf_unregister_net_hook(net, &hook->ops); 232 if (release_netdev) { 233 list_del(&hook->list); 234 kfree_rcu(hook, rcu); 235 } 236 } 237 } 238 239 static int nf_tables_register_hook(struct net *net, 240 const struct nft_table *table, 241 struct nft_chain *chain) 242 { 243 struct nft_base_chain *basechain; 244 const struct nf_hook_ops *ops; 245 246 if (table->flags & NFT_TABLE_F_DORMANT || 247 !nft_is_base_chain(chain)) 248 return 0; 249 250 basechain = nft_base_chain(chain); 251 ops = &basechain->ops; 252 253 if (basechain->type->ops_register) 254 return basechain->type->ops_register(net, ops); 255 256 if (nft_base_chain_netdev(table->family, basechain->ops.hooknum)) 257 return nft_netdev_register_hooks(net, &basechain->hook_list); 258 259 return nf_register_net_hook(net, &basechain->ops); 260 } 261 262 static void __nf_tables_unregister_hook(struct net *net, 263 const struct nft_table *table, 264 struct nft_chain *chain, 265 bool release_netdev) 266 { 267 struct nft_base_chain *basechain; 268 const struct nf_hook_ops *ops; 269 270 if (table->flags & NFT_TABLE_F_DORMANT || 271 !nft_is_base_chain(chain)) 272 return; 273 basechain = nft_base_chain(chain); 274 ops = &basechain->ops; 275 276 if (basechain->type->ops_unregister) 277 return basechain->type->ops_unregister(net, ops); 278 279 if (nft_base_chain_netdev(table->family, basechain->ops.hooknum)) 280 nft_netdev_unregister_hooks(net, &basechain->hook_list, 281 release_netdev); 282 else 283 nf_unregister_net_hook(net, &basechain->ops); 284 } 285 286 static void nf_tables_unregister_hook(struct net *net, 287 const struct nft_table *table, 288 struct nft_chain *chain) 289 { 290 return __nf_tables_unregister_hook(net, table, chain, false); 291 } 292 293 static void nft_trans_commit_list_add_tail(struct net *net, struct nft_trans *trans) 294 { 295 struct nftables_pernet *nft_net = nft_pernet(net); 296 297 list_add_tail(&trans->list, &nft_net->commit_list); 298 } 299 300 static int nft_trans_table_add(struct nft_ctx *ctx, int msg_type) 301 { 302 struct nft_trans *trans; 303 304 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_table)); 305 if (trans == NULL) 306 return -ENOMEM; 307 308 if (msg_type == NFT_MSG_NEWTABLE) 309 nft_activate_next(ctx->net, ctx->table); 310 311 nft_trans_commit_list_add_tail(ctx->net, trans); 312 return 0; 313 } 314 315 static int nft_deltable(struct nft_ctx *ctx) 316 { 317 int err; 318 319 err = nft_trans_table_add(ctx, NFT_MSG_DELTABLE); 320 if (err < 0) 321 return err; 322 323 nft_deactivate_next(ctx->net, ctx->table); 324 return err; 325 } 326 327 static struct nft_trans *nft_trans_chain_add(struct nft_ctx *ctx, int msg_type) 328 { 329 struct nft_trans *trans; 330 331 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_chain)); 332 if (trans == NULL) 333 return ERR_PTR(-ENOMEM); 334 335 if (msg_type == NFT_MSG_NEWCHAIN) { 336 nft_activate_next(ctx->net, ctx->chain); 337 338 if (ctx->nla[NFTA_CHAIN_ID]) { 339 nft_trans_chain_id(trans) = 340 ntohl(nla_get_be32(ctx->nla[NFTA_CHAIN_ID])); 341 } 342 } 343 344 nft_trans_commit_list_add_tail(ctx->net, trans); 345 return trans; 346 } 347 348 static int nft_delchain(struct nft_ctx *ctx) 349 { 350 struct nft_trans *trans; 351 352 trans = nft_trans_chain_add(ctx, NFT_MSG_DELCHAIN); 353 if (IS_ERR(trans)) 354 return PTR_ERR(trans); 355 356 ctx->table->use--; 357 nft_deactivate_next(ctx->net, ctx->chain); 358 359 return 0; 360 } 361 362 static void nft_rule_expr_activate(const struct nft_ctx *ctx, 363 struct nft_rule *rule) 364 { 365 struct nft_expr *expr; 366 367 expr = nft_expr_first(rule); 368 while (nft_expr_more(rule, expr)) { 369 if (expr->ops->activate) 370 expr->ops->activate(ctx, expr); 371 372 expr = nft_expr_next(expr); 373 } 374 } 375 376 static void nft_rule_expr_deactivate(const struct nft_ctx *ctx, 377 struct nft_rule *rule, 378 enum nft_trans_phase phase) 379 { 380 struct nft_expr *expr; 381 382 expr = nft_expr_first(rule); 383 while (nft_expr_more(rule, expr)) { 384 if (expr->ops->deactivate) 385 expr->ops->deactivate(ctx, expr, phase); 386 387 expr = nft_expr_next(expr); 388 } 389 } 390 391 static int 392 nf_tables_delrule_deactivate(struct nft_ctx *ctx, struct nft_rule *rule) 393 { 394 /* You cannot delete the same rule twice */ 395 if (nft_is_active_next(ctx->net, rule)) { 396 nft_deactivate_next(ctx->net, rule); 397 ctx->chain->use--; 398 return 0; 399 } 400 return -ENOENT; 401 } 402 403 static struct nft_trans *nft_trans_rule_add(struct nft_ctx *ctx, int msg_type, 404 struct nft_rule *rule) 405 { 406 struct nft_trans *trans; 407 408 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_rule)); 409 if (trans == NULL) 410 return NULL; 411 412 if (msg_type == NFT_MSG_NEWRULE && ctx->nla[NFTA_RULE_ID] != NULL) { 413 nft_trans_rule_id(trans) = 414 ntohl(nla_get_be32(ctx->nla[NFTA_RULE_ID])); 415 } 416 nft_trans_rule(trans) = rule; 417 nft_trans_commit_list_add_tail(ctx->net, trans); 418 419 return trans; 420 } 421 422 static int nft_delrule(struct nft_ctx *ctx, struct nft_rule *rule) 423 { 424 struct nft_flow_rule *flow; 425 struct nft_trans *trans; 426 int err; 427 428 trans = nft_trans_rule_add(ctx, NFT_MSG_DELRULE, rule); 429 if (trans == NULL) 430 return -ENOMEM; 431 432 if (ctx->chain->flags & NFT_CHAIN_HW_OFFLOAD) { 433 flow = nft_flow_rule_create(ctx->net, rule); 434 if (IS_ERR(flow)) { 435 nft_trans_destroy(trans); 436 return PTR_ERR(flow); 437 } 438 439 nft_trans_flow_rule(trans) = flow; 440 } 441 442 err = nf_tables_delrule_deactivate(ctx, rule); 443 if (err < 0) { 444 nft_trans_destroy(trans); 445 return err; 446 } 447 nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_PREPARE); 448 449 return 0; 450 } 451 452 static int nft_delrule_by_chain(struct nft_ctx *ctx) 453 { 454 struct nft_rule *rule; 455 int err; 456 457 list_for_each_entry(rule, &ctx->chain->rules, list) { 458 if (!nft_is_active_next(ctx->net, rule)) 459 continue; 460 461 err = nft_delrule(ctx, rule); 462 if (err < 0) 463 return err; 464 } 465 return 0; 466 } 467 468 static int nft_trans_set_add(const struct nft_ctx *ctx, int msg_type, 469 struct nft_set *set) 470 { 471 struct nft_trans *trans; 472 473 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_set)); 474 if (trans == NULL) 475 return -ENOMEM; 476 477 if (msg_type == NFT_MSG_NEWSET && ctx->nla[NFTA_SET_ID] != NULL) { 478 nft_trans_set_id(trans) = 479 ntohl(nla_get_be32(ctx->nla[NFTA_SET_ID])); 480 nft_activate_next(ctx->net, set); 481 } 482 nft_trans_set(trans) = set; 483 nft_trans_commit_list_add_tail(ctx->net, trans); 484 485 return 0; 486 } 487 488 static int nft_delset(const struct nft_ctx *ctx, struct nft_set *set) 489 { 490 int err; 491 492 err = nft_trans_set_add(ctx, NFT_MSG_DELSET, set); 493 if (err < 0) 494 return err; 495 496 nft_deactivate_next(ctx->net, set); 497 ctx->table->use--; 498 499 return err; 500 } 501 502 static int nft_trans_obj_add(struct nft_ctx *ctx, int msg_type, 503 struct nft_object *obj) 504 { 505 struct nft_trans *trans; 506 507 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_obj)); 508 if (trans == NULL) 509 return -ENOMEM; 510 511 if (msg_type == NFT_MSG_NEWOBJ) 512 nft_activate_next(ctx->net, obj); 513 514 nft_trans_obj(trans) = obj; 515 nft_trans_commit_list_add_tail(ctx->net, trans); 516 517 return 0; 518 } 519 520 static int nft_delobj(struct nft_ctx *ctx, struct nft_object *obj) 521 { 522 int err; 523 524 err = nft_trans_obj_add(ctx, NFT_MSG_DELOBJ, obj); 525 if (err < 0) 526 return err; 527 528 nft_deactivate_next(ctx->net, obj); 529 ctx->table->use--; 530 531 return err; 532 } 533 534 static int nft_trans_flowtable_add(struct nft_ctx *ctx, int msg_type, 535 struct nft_flowtable *flowtable) 536 { 537 struct nft_trans *trans; 538 539 trans = nft_trans_alloc(ctx, msg_type, 540 sizeof(struct nft_trans_flowtable)); 541 if (trans == NULL) 542 return -ENOMEM; 543 544 if (msg_type == NFT_MSG_NEWFLOWTABLE) 545 nft_activate_next(ctx->net, flowtable); 546 547 INIT_LIST_HEAD(&nft_trans_flowtable_hooks(trans)); 548 nft_trans_flowtable(trans) = flowtable; 549 nft_trans_commit_list_add_tail(ctx->net, trans); 550 551 return 0; 552 } 553 554 static int nft_delflowtable(struct nft_ctx *ctx, 555 struct nft_flowtable *flowtable) 556 { 557 int err; 558 559 err = nft_trans_flowtable_add(ctx, NFT_MSG_DELFLOWTABLE, flowtable); 560 if (err < 0) 561 return err; 562 563 nft_deactivate_next(ctx->net, flowtable); 564 ctx->table->use--; 565 566 return err; 567 } 568 569 static void __nft_reg_track_clobber(struct nft_regs_track *track, u8 dreg) 570 { 571 int i; 572 573 for (i = track->regs[dreg].num_reg; i > 0; i--) 574 __nft_reg_track_cancel(track, dreg - i); 575 } 576 577 static void __nft_reg_track_update(struct nft_regs_track *track, 578 const struct nft_expr *expr, 579 u8 dreg, u8 num_reg) 580 { 581 track->regs[dreg].selector = expr; 582 track->regs[dreg].bitwise = NULL; 583 track->regs[dreg].num_reg = num_reg; 584 } 585 586 void nft_reg_track_update(struct nft_regs_track *track, 587 const struct nft_expr *expr, u8 dreg, u8 len) 588 { 589 unsigned int regcount; 590 int i; 591 592 __nft_reg_track_clobber(track, dreg); 593 594 regcount = DIV_ROUND_UP(len, NFT_REG32_SIZE); 595 for (i = 0; i < regcount; i++, dreg++) 596 __nft_reg_track_update(track, expr, dreg, i); 597 } 598 EXPORT_SYMBOL_GPL(nft_reg_track_update); 599 600 void nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg, u8 len) 601 { 602 unsigned int regcount; 603 int i; 604 605 __nft_reg_track_clobber(track, dreg); 606 607 regcount = DIV_ROUND_UP(len, NFT_REG32_SIZE); 608 for (i = 0; i < regcount; i++, dreg++) 609 __nft_reg_track_cancel(track, dreg); 610 } 611 EXPORT_SYMBOL_GPL(nft_reg_track_cancel); 612 613 void __nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg) 614 { 615 track->regs[dreg].selector = NULL; 616 track->regs[dreg].bitwise = NULL; 617 track->regs[dreg].num_reg = 0; 618 } 619 EXPORT_SYMBOL_GPL(__nft_reg_track_cancel); 620 621 /* 622 * Tables 623 */ 624 625 static struct nft_table *nft_table_lookup(const struct net *net, 626 const struct nlattr *nla, 627 u8 family, u8 genmask, u32 nlpid) 628 { 629 struct nftables_pernet *nft_net; 630 struct nft_table *table; 631 632 if (nla == NULL) 633 return ERR_PTR(-EINVAL); 634 635 nft_net = nft_pernet(net); 636 list_for_each_entry_rcu(table, &nft_net->tables, list, 637 lockdep_is_held(&nft_net->commit_mutex)) { 638 if (!nla_strcmp(nla, table->name) && 639 table->family == family && 640 nft_active_genmask(table, genmask)) { 641 if (nft_table_has_owner(table) && 642 nlpid && table->nlpid != nlpid) 643 return ERR_PTR(-EPERM); 644 645 return table; 646 } 647 } 648 649 return ERR_PTR(-ENOENT); 650 } 651 652 static struct nft_table *nft_table_lookup_byhandle(const struct net *net, 653 const struct nlattr *nla, 654 u8 genmask, u32 nlpid) 655 { 656 struct nftables_pernet *nft_net; 657 struct nft_table *table; 658 659 nft_net = nft_pernet(net); 660 list_for_each_entry(table, &nft_net->tables, list) { 661 if (be64_to_cpu(nla_get_be64(nla)) == table->handle && 662 nft_active_genmask(table, genmask)) { 663 if (nft_table_has_owner(table) && 664 nlpid && table->nlpid != nlpid) 665 return ERR_PTR(-EPERM); 666 667 return table; 668 } 669 } 670 671 return ERR_PTR(-ENOENT); 672 } 673 674 static inline u64 nf_tables_alloc_handle(struct nft_table *table) 675 { 676 return ++table->hgenerator; 677 } 678 679 static const struct nft_chain_type *chain_type[NFPROTO_NUMPROTO][NFT_CHAIN_T_MAX]; 680 681 static const struct nft_chain_type * 682 __nft_chain_type_get(u8 family, enum nft_chain_types type) 683 { 684 if (family >= NFPROTO_NUMPROTO || 685 type >= NFT_CHAIN_T_MAX) 686 return NULL; 687 688 return chain_type[family][type]; 689 } 690 691 static const struct nft_chain_type * 692 __nf_tables_chain_type_lookup(const struct nlattr *nla, u8 family) 693 { 694 const struct nft_chain_type *type; 695 int i; 696 697 for (i = 0; i < NFT_CHAIN_T_MAX; i++) { 698 type = __nft_chain_type_get(family, i); 699 if (!type) 700 continue; 701 if (!nla_strcmp(nla, type->name)) 702 return type; 703 } 704 return NULL; 705 } 706 707 struct nft_module_request { 708 struct list_head list; 709 char module[MODULE_NAME_LEN]; 710 bool done; 711 }; 712 713 #ifdef CONFIG_MODULES 714 __printf(2, 3) int nft_request_module(struct net *net, const char *fmt, 715 ...) 716 { 717 char module_name[MODULE_NAME_LEN]; 718 struct nftables_pernet *nft_net; 719 struct nft_module_request *req; 720 va_list args; 721 int ret; 722 723 va_start(args, fmt); 724 ret = vsnprintf(module_name, MODULE_NAME_LEN, fmt, args); 725 va_end(args); 726 if (ret >= MODULE_NAME_LEN) 727 return 0; 728 729 nft_net = nft_pernet(net); 730 list_for_each_entry(req, &nft_net->module_list, list) { 731 if (!strcmp(req->module, module_name)) { 732 if (req->done) 733 return 0; 734 735 /* A request to load this module already exists. */ 736 return -EAGAIN; 737 } 738 } 739 740 req = kmalloc(sizeof(*req), GFP_KERNEL); 741 if (!req) 742 return -ENOMEM; 743 744 req->done = false; 745 strscpy(req->module, module_name, MODULE_NAME_LEN); 746 list_add_tail(&req->list, &nft_net->module_list); 747 748 return -EAGAIN; 749 } 750 EXPORT_SYMBOL_GPL(nft_request_module); 751 #endif 752 753 static void lockdep_nfnl_nft_mutex_not_held(void) 754 { 755 #ifdef CONFIG_PROVE_LOCKING 756 if (debug_locks) 757 WARN_ON_ONCE(lockdep_nfnl_is_held(NFNL_SUBSYS_NFTABLES)); 758 #endif 759 } 760 761 static const struct nft_chain_type * 762 nf_tables_chain_type_lookup(struct net *net, const struct nlattr *nla, 763 u8 family, bool autoload) 764 { 765 const struct nft_chain_type *type; 766 767 type = __nf_tables_chain_type_lookup(nla, family); 768 if (type != NULL) 769 return type; 770 771 lockdep_nfnl_nft_mutex_not_held(); 772 #ifdef CONFIG_MODULES 773 if (autoload) { 774 if (nft_request_module(net, "nft-chain-%u-%.*s", family, 775 nla_len(nla), 776 (const char *)nla_data(nla)) == -EAGAIN) 777 return ERR_PTR(-EAGAIN); 778 } 779 #endif 780 return ERR_PTR(-ENOENT); 781 } 782 783 static __be16 nft_base_seq(const struct net *net) 784 { 785 struct nftables_pernet *nft_net = nft_pernet(net); 786 787 return htons(nft_net->base_seq & 0xffff); 788 } 789 790 static const struct nla_policy nft_table_policy[NFTA_TABLE_MAX + 1] = { 791 [NFTA_TABLE_NAME] = { .type = NLA_STRING, 792 .len = NFT_TABLE_MAXNAMELEN - 1 }, 793 [NFTA_TABLE_FLAGS] = { .type = NLA_U32 }, 794 [NFTA_TABLE_HANDLE] = { .type = NLA_U64 }, 795 [NFTA_TABLE_USERDATA] = { .type = NLA_BINARY, 796 .len = NFT_USERDATA_MAXLEN } 797 }; 798 799 static int nf_tables_fill_table_info(struct sk_buff *skb, struct net *net, 800 u32 portid, u32 seq, int event, u32 flags, 801 int family, const struct nft_table *table) 802 { 803 struct nlmsghdr *nlh; 804 805 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 806 nlh = nfnl_msg_put(skb, portid, seq, event, flags, family, 807 NFNETLINK_V0, nft_base_seq(net)); 808 if (!nlh) 809 goto nla_put_failure; 810 811 if (nla_put_string(skb, NFTA_TABLE_NAME, table->name) || 812 nla_put_be32(skb, NFTA_TABLE_FLAGS, 813 htonl(table->flags & NFT_TABLE_F_MASK)) || 814 nla_put_be32(skb, NFTA_TABLE_USE, htonl(table->use)) || 815 nla_put_be64(skb, NFTA_TABLE_HANDLE, cpu_to_be64(table->handle), 816 NFTA_TABLE_PAD)) 817 goto nla_put_failure; 818 if (nft_table_has_owner(table) && 819 nla_put_be32(skb, NFTA_TABLE_OWNER, htonl(table->nlpid))) 820 goto nla_put_failure; 821 822 if (table->udata) { 823 if (nla_put(skb, NFTA_TABLE_USERDATA, table->udlen, table->udata)) 824 goto nla_put_failure; 825 } 826 827 nlmsg_end(skb, nlh); 828 return 0; 829 830 nla_put_failure: 831 nlmsg_trim(skb, nlh); 832 return -1; 833 } 834 835 struct nftnl_skb_parms { 836 bool report; 837 }; 838 #define NFT_CB(skb) (*(struct nftnl_skb_parms*)&((skb)->cb)) 839 840 static void nft_notify_enqueue(struct sk_buff *skb, bool report, 841 struct list_head *notify_list) 842 { 843 NFT_CB(skb).report = report; 844 list_add_tail(&skb->list, notify_list); 845 } 846 847 static void nf_tables_table_notify(const struct nft_ctx *ctx, int event) 848 { 849 struct nftables_pernet *nft_net; 850 struct sk_buff *skb; 851 u16 flags = 0; 852 int err; 853 854 if (!ctx->report && 855 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES)) 856 return; 857 858 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 859 if (skb == NULL) 860 goto err; 861 862 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 863 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 864 865 err = nf_tables_fill_table_info(skb, ctx->net, ctx->portid, ctx->seq, 866 event, flags, ctx->family, ctx->table); 867 if (err < 0) { 868 kfree_skb(skb); 869 goto err; 870 } 871 872 nft_net = nft_pernet(ctx->net); 873 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 874 return; 875 err: 876 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS); 877 } 878 879 static int nf_tables_dump_tables(struct sk_buff *skb, 880 struct netlink_callback *cb) 881 { 882 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 883 struct nftables_pernet *nft_net; 884 const struct nft_table *table; 885 unsigned int idx = 0, s_idx = cb->args[0]; 886 struct net *net = sock_net(skb->sk); 887 int family = nfmsg->nfgen_family; 888 889 rcu_read_lock(); 890 nft_net = nft_pernet(net); 891 cb->seq = READ_ONCE(nft_net->base_seq); 892 893 list_for_each_entry_rcu(table, &nft_net->tables, list) { 894 if (family != NFPROTO_UNSPEC && family != table->family) 895 continue; 896 897 if (idx < s_idx) 898 goto cont; 899 if (idx > s_idx) 900 memset(&cb->args[1], 0, 901 sizeof(cb->args) - sizeof(cb->args[0])); 902 if (!nft_is_active(net, table)) 903 continue; 904 if (nf_tables_fill_table_info(skb, net, 905 NETLINK_CB(cb->skb).portid, 906 cb->nlh->nlmsg_seq, 907 NFT_MSG_NEWTABLE, NLM_F_MULTI, 908 table->family, table) < 0) 909 goto done; 910 911 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 912 cont: 913 idx++; 914 } 915 done: 916 rcu_read_unlock(); 917 cb->args[0] = idx; 918 return skb->len; 919 } 920 921 static int nft_netlink_dump_start_rcu(struct sock *nlsk, struct sk_buff *skb, 922 const struct nlmsghdr *nlh, 923 struct netlink_dump_control *c) 924 { 925 int err; 926 927 if (!try_module_get(THIS_MODULE)) 928 return -EINVAL; 929 930 rcu_read_unlock(); 931 err = netlink_dump_start(nlsk, skb, nlh, c); 932 rcu_read_lock(); 933 module_put(THIS_MODULE); 934 935 return err; 936 } 937 938 /* called with rcu_read_lock held */ 939 static int nf_tables_gettable(struct sk_buff *skb, const struct nfnl_info *info, 940 const struct nlattr * const nla[]) 941 { 942 struct netlink_ext_ack *extack = info->extack; 943 u8 genmask = nft_genmask_cur(info->net); 944 u8 family = info->nfmsg->nfgen_family; 945 const struct nft_table *table; 946 struct net *net = info->net; 947 struct sk_buff *skb2; 948 int err; 949 950 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 951 struct netlink_dump_control c = { 952 .dump = nf_tables_dump_tables, 953 .module = THIS_MODULE, 954 }; 955 956 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 957 } 958 959 table = nft_table_lookup(net, nla[NFTA_TABLE_NAME], family, genmask, 0); 960 if (IS_ERR(table)) { 961 NL_SET_BAD_ATTR(extack, nla[NFTA_TABLE_NAME]); 962 return PTR_ERR(table); 963 } 964 965 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 966 if (!skb2) 967 return -ENOMEM; 968 969 err = nf_tables_fill_table_info(skb2, net, NETLINK_CB(skb).portid, 970 info->nlh->nlmsg_seq, NFT_MSG_NEWTABLE, 971 0, family, table); 972 if (err < 0) 973 goto err_fill_table_info; 974 975 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 976 977 err_fill_table_info: 978 kfree_skb(skb2); 979 return err; 980 } 981 982 static void nft_table_disable(struct net *net, struct nft_table *table, u32 cnt) 983 { 984 struct nft_chain *chain; 985 u32 i = 0; 986 987 list_for_each_entry(chain, &table->chains, list) { 988 if (!nft_is_active_next(net, chain)) 989 continue; 990 if (!nft_is_base_chain(chain)) 991 continue; 992 993 if (cnt && i++ == cnt) 994 break; 995 996 nf_tables_unregister_hook(net, table, chain); 997 } 998 } 999 1000 static int nf_tables_table_enable(struct net *net, struct nft_table *table) 1001 { 1002 struct nft_chain *chain; 1003 int err, i = 0; 1004 1005 list_for_each_entry(chain, &table->chains, list) { 1006 if (!nft_is_active_next(net, chain)) 1007 continue; 1008 if (!nft_is_base_chain(chain)) 1009 continue; 1010 1011 err = nf_tables_register_hook(net, table, chain); 1012 if (err < 0) 1013 goto err_register_hooks; 1014 1015 i++; 1016 } 1017 return 0; 1018 1019 err_register_hooks: 1020 if (i) 1021 nft_table_disable(net, table, i); 1022 return err; 1023 } 1024 1025 static void nf_tables_table_disable(struct net *net, struct nft_table *table) 1026 { 1027 table->flags &= ~NFT_TABLE_F_DORMANT; 1028 nft_table_disable(net, table, 0); 1029 table->flags |= NFT_TABLE_F_DORMANT; 1030 } 1031 1032 #define __NFT_TABLE_F_INTERNAL (NFT_TABLE_F_MASK + 1) 1033 #define __NFT_TABLE_F_WAS_DORMANT (__NFT_TABLE_F_INTERNAL << 0) 1034 #define __NFT_TABLE_F_WAS_AWAKEN (__NFT_TABLE_F_INTERNAL << 1) 1035 #define __NFT_TABLE_F_UPDATE (__NFT_TABLE_F_WAS_DORMANT | \ 1036 __NFT_TABLE_F_WAS_AWAKEN) 1037 1038 static int nf_tables_updtable(struct nft_ctx *ctx) 1039 { 1040 struct nft_trans *trans; 1041 u32 flags; 1042 int ret; 1043 1044 if (!ctx->nla[NFTA_TABLE_FLAGS]) 1045 return 0; 1046 1047 flags = ntohl(nla_get_be32(ctx->nla[NFTA_TABLE_FLAGS])); 1048 if (flags & ~NFT_TABLE_F_MASK) 1049 return -EOPNOTSUPP; 1050 1051 if (flags == ctx->table->flags) 1052 return 0; 1053 1054 if ((nft_table_has_owner(ctx->table) && 1055 !(flags & NFT_TABLE_F_OWNER)) || 1056 (!nft_table_has_owner(ctx->table) && 1057 flags & NFT_TABLE_F_OWNER)) 1058 return -EOPNOTSUPP; 1059 1060 trans = nft_trans_alloc(ctx, NFT_MSG_NEWTABLE, 1061 sizeof(struct nft_trans_table)); 1062 if (trans == NULL) 1063 return -ENOMEM; 1064 1065 if ((flags & NFT_TABLE_F_DORMANT) && 1066 !(ctx->table->flags & NFT_TABLE_F_DORMANT)) { 1067 ctx->table->flags |= NFT_TABLE_F_DORMANT; 1068 if (!(ctx->table->flags & __NFT_TABLE_F_UPDATE)) 1069 ctx->table->flags |= __NFT_TABLE_F_WAS_AWAKEN; 1070 } else if (!(flags & NFT_TABLE_F_DORMANT) && 1071 ctx->table->flags & NFT_TABLE_F_DORMANT) { 1072 ctx->table->flags &= ~NFT_TABLE_F_DORMANT; 1073 if (!(ctx->table->flags & __NFT_TABLE_F_UPDATE)) { 1074 ret = nf_tables_table_enable(ctx->net, ctx->table); 1075 if (ret < 0) 1076 goto err_register_hooks; 1077 1078 ctx->table->flags |= __NFT_TABLE_F_WAS_DORMANT; 1079 } 1080 } 1081 1082 nft_trans_table_update(trans) = true; 1083 nft_trans_commit_list_add_tail(ctx->net, trans); 1084 1085 return 0; 1086 1087 err_register_hooks: 1088 nft_trans_destroy(trans); 1089 return ret; 1090 } 1091 1092 static u32 nft_chain_hash(const void *data, u32 len, u32 seed) 1093 { 1094 const char *name = data; 1095 1096 return jhash(name, strlen(name), seed); 1097 } 1098 1099 static u32 nft_chain_hash_obj(const void *data, u32 len, u32 seed) 1100 { 1101 const struct nft_chain *chain = data; 1102 1103 return nft_chain_hash(chain->name, 0, seed); 1104 } 1105 1106 static int nft_chain_hash_cmp(struct rhashtable_compare_arg *arg, 1107 const void *ptr) 1108 { 1109 const struct nft_chain *chain = ptr; 1110 const char *name = arg->key; 1111 1112 return strcmp(chain->name, name); 1113 } 1114 1115 static u32 nft_objname_hash(const void *data, u32 len, u32 seed) 1116 { 1117 const struct nft_object_hash_key *k = data; 1118 1119 seed ^= hash_ptr(k->table, 32); 1120 1121 return jhash(k->name, strlen(k->name), seed); 1122 } 1123 1124 static u32 nft_objname_hash_obj(const void *data, u32 len, u32 seed) 1125 { 1126 const struct nft_object *obj = data; 1127 1128 return nft_objname_hash(&obj->key, 0, seed); 1129 } 1130 1131 static int nft_objname_hash_cmp(struct rhashtable_compare_arg *arg, 1132 const void *ptr) 1133 { 1134 const struct nft_object_hash_key *k = arg->key; 1135 const struct nft_object *obj = ptr; 1136 1137 if (obj->key.table != k->table) 1138 return -1; 1139 1140 return strcmp(obj->key.name, k->name); 1141 } 1142 1143 static bool nft_supported_family(u8 family) 1144 { 1145 return false 1146 #ifdef CONFIG_NF_TABLES_INET 1147 || family == NFPROTO_INET 1148 #endif 1149 #ifdef CONFIG_NF_TABLES_IPV4 1150 || family == NFPROTO_IPV4 1151 #endif 1152 #ifdef CONFIG_NF_TABLES_ARP 1153 || family == NFPROTO_ARP 1154 #endif 1155 #ifdef CONFIG_NF_TABLES_NETDEV 1156 || family == NFPROTO_NETDEV 1157 #endif 1158 #if IS_ENABLED(CONFIG_NF_TABLES_BRIDGE) 1159 || family == NFPROTO_BRIDGE 1160 #endif 1161 #ifdef CONFIG_NF_TABLES_IPV6 1162 || family == NFPROTO_IPV6 1163 #endif 1164 ; 1165 } 1166 1167 static int nf_tables_newtable(struct sk_buff *skb, const struct nfnl_info *info, 1168 const struct nlattr * const nla[]) 1169 { 1170 struct nftables_pernet *nft_net = nft_pernet(info->net); 1171 struct netlink_ext_ack *extack = info->extack; 1172 u8 genmask = nft_genmask_next(info->net); 1173 u8 family = info->nfmsg->nfgen_family; 1174 struct net *net = info->net; 1175 const struct nlattr *attr; 1176 struct nft_table *table; 1177 struct nft_ctx ctx; 1178 u32 flags = 0; 1179 int err; 1180 1181 if (!nft_supported_family(family)) 1182 return -EOPNOTSUPP; 1183 1184 lockdep_assert_held(&nft_net->commit_mutex); 1185 attr = nla[NFTA_TABLE_NAME]; 1186 table = nft_table_lookup(net, attr, family, genmask, 1187 NETLINK_CB(skb).portid); 1188 if (IS_ERR(table)) { 1189 if (PTR_ERR(table) != -ENOENT) 1190 return PTR_ERR(table); 1191 } else { 1192 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 1193 NL_SET_BAD_ATTR(extack, attr); 1194 return -EEXIST; 1195 } 1196 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 1197 return -EOPNOTSUPP; 1198 1199 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 1200 1201 return nf_tables_updtable(&ctx); 1202 } 1203 1204 if (nla[NFTA_TABLE_FLAGS]) { 1205 flags = ntohl(nla_get_be32(nla[NFTA_TABLE_FLAGS])); 1206 if (flags & ~NFT_TABLE_F_MASK) 1207 return -EOPNOTSUPP; 1208 } 1209 1210 err = -ENOMEM; 1211 table = kzalloc(sizeof(*table), GFP_KERNEL_ACCOUNT); 1212 if (table == NULL) 1213 goto err_kzalloc; 1214 1215 table->name = nla_strdup(attr, GFP_KERNEL_ACCOUNT); 1216 if (table->name == NULL) 1217 goto err_strdup; 1218 1219 if (nla[NFTA_TABLE_USERDATA]) { 1220 table->udata = nla_memdup(nla[NFTA_TABLE_USERDATA], GFP_KERNEL_ACCOUNT); 1221 if (table->udata == NULL) 1222 goto err_table_udata; 1223 1224 table->udlen = nla_len(nla[NFTA_TABLE_USERDATA]); 1225 } 1226 1227 err = rhltable_init(&table->chains_ht, &nft_chain_ht_params); 1228 if (err) 1229 goto err_chain_ht; 1230 1231 INIT_LIST_HEAD(&table->chains); 1232 INIT_LIST_HEAD(&table->sets); 1233 INIT_LIST_HEAD(&table->objects); 1234 INIT_LIST_HEAD(&table->flowtables); 1235 table->family = family; 1236 table->flags = flags; 1237 table->handle = ++nft_net->table_handle; 1238 if (table->flags & NFT_TABLE_F_OWNER) 1239 table->nlpid = NETLINK_CB(skb).portid; 1240 1241 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 1242 err = nft_trans_table_add(&ctx, NFT_MSG_NEWTABLE); 1243 if (err < 0) 1244 goto err_trans; 1245 1246 list_add_tail_rcu(&table->list, &nft_net->tables); 1247 return 0; 1248 err_trans: 1249 rhltable_destroy(&table->chains_ht); 1250 err_chain_ht: 1251 kfree(table->udata); 1252 err_table_udata: 1253 kfree(table->name); 1254 err_strdup: 1255 kfree(table); 1256 err_kzalloc: 1257 return err; 1258 } 1259 1260 static int nft_flush_table(struct nft_ctx *ctx) 1261 { 1262 struct nft_flowtable *flowtable, *nft; 1263 struct nft_chain *chain, *nc; 1264 struct nft_object *obj, *ne; 1265 struct nft_set *set, *ns; 1266 int err; 1267 1268 list_for_each_entry(chain, &ctx->table->chains, list) { 1269 if (!nft_is_active_next(ctx->net, chain)) 1270 continue; 1271 1272 if (nft_chain_is_bound(chain)) 1273 continue; 1274 1275 ctx->chain = chain; 1276 1277 err = nft_delrule_by_chain(ctx); 1278 if (err < 0) 1279 goto out; 1280 } 1281 1282 list_for_each_entry_safe(set, ns, &ctx->table->sets, list) { 1283 if (!nft_is_active_next(ctx->net, set)) 1284 continue; 1285 1286 if (nft_set_is_anonymous(set) && 1287 !list_empty(&set->bindings)) 1288 continue; 1289 1290 err = nft_delset(ctx, set); 1291 if (err < 0) 1292 goto out; 1293 } 1294 1295 list_for_each_entry_safe(flowtable, nft, &ctx->table->flowtables, list) { 1296 if (!nft_is_active_next(ctx->net, flowtable)) 1297 continue; 1298 1299 err = nft_delflowtable(ctx, flowtable); 1300 if (err < 0) 1301 goto out; 1302 } 1303 1304 list_for_each_entry_safe(obj, ne, &ctx->table->objects, list) { 1305 if (!nft_is_active_next(ctx->net, obj)) 1306 continue; 1307 1308 err = nft_delobj(ctx, obj); 1309 if (err < 0) 1310 goto out; 1311 } 1312 1313 list_for_each_entry_safe(chain, nc, &ctx->table->chains, list) { 1314 if (!nft_is_active_next(ctx->net, chain)) 1315 continue; 1316 1317 if (nft_chain_is_bound(chain)) 1318 continue; 1319 1320 ctx->chain = chain; 1321 1322 err = nft_delchain(ctx); 1323 if (err < 0) 1324 goto out; 1325 } 1326 1327 err = nft_deltable(ctx); 1328 out: 1329 return err; 1330 } 1331 1332 static int nft_flush(struct nft_ctx *ctx, int family) 1333 { 1334 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 1335 const struct nlattr * const *nla = ctx->nla; 1336 struct nft_table *table, *nt; 1337 int err = 0; 1338 1339 list_for_each_entry_safe(table, nt, &nft_net->tables, list) { 1340 if (family != AF_UNSPEC && table->family != family) 1341 continue; 1342 1343 ctx->family = table->family; 1344 1345 if (!nft_is_active_next(ctx->net, table)) 1346 continue; 1347 1348 if (nft_table_has_owner(table) && table->nlpid != ctx->portid) 1349 continue; 1350 1351 if (nla[NFTA_TABLE_NAME] && 1352 nla_strcmp(nla[NFTA_TABLE_NAME], table->name) != 0) 1353 continue; 1354 1355 ctx->table = table; 1356 1357 err = nft_flush_table(ctx); 1358 if (err < 0) 1359 goto out; 1360 } 1361 out: 1362 return err; 1363 } 1364 1365 static int nf_tables_deltable(struct sk_buff *skb, const struct nfnl_info *info, 1366 const struct nlattr * const nla[]) 1367 { 1368 struct netlink_ext_ack *extack = info->extack; 1369 u8 genmask = nft_genmask_next(info->net); 1370 u8 family = info->nfmsg->nfgen_family; 1371 struct net *net = info->net; 1372 const struct nlattr *attr; 1373 struct nft_table *table; 1374 struct nft_ctx ctx; 1375 1376 nft_ctx_init(&ctx, net, skb, info->nlh, 0, NULL, NULL, nla); 1377 if (family == AF_UNSPEC || 1378 (!nla[NFTA_TABLE_NAME] && !nla[NFTA_TABLE_HANDLE])) 1379 return nft_flush(&ctx, family); 1380 1381 if (nla[NFTA_TABLE_HANDLE]) { 1382 attr = nla[NFTA_TABLE_HANDLE]; 1383 table = nft_table_lookup_byhandle(net, attr, genmask, 1384 NETLINK_CB(skb).portid); 1385 } else { 1386 attr = nla[NFTA_TABLE_NAME]; 1387 table = nft_table_lookup(net, attr, family, genmask, 1388 NETLINK_CB(skb).portid); 1389 } 1390 1391 if (IS_ERR(table)) { 1392 NL_SET_BAD_ATTR(extack, attr); 1393 return PTR_ERR(table); 1394 } 1395 1396 if (info->nlh->nlmsg_flags & NLM_F_NONREC && 1397 table->use > 0) 1398 return -EBUSY; 1399 1400 ctx.family = family; 1401 ctx.table = table; 1402 1403 return nft_flush_table(&ctx); 1404 } 1405 1406 static void nf_tables_table_destroy(struct nft_ctx *ctx) 1407 { 1408 if (WARN_ON(ctx->table->use > 0)) 1409 return; 1410 1411 rhltable_destroy(&ctx->table->chains_ht); 1412 kfree(ctx->table->name); 1413 kfree(ctx->table->udata); 1414 kfree(ctx->table); 1415 } 1416 1417 void nft_register_chain_type(const struct nft_chain_type *ctype) 1418 { 1419 nfnl_lock(NFNL_SUBSYS_NFTABLES); 1420 if (WARN_ON(__nft_chain_type_get(ctype->family, ctype->type))) { 1421 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 1422 return; 1423 } 1424 chain_type[ctype->family][ctype->type] = ctype; 1425 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 1426 } 1427 EXPORT_SYMBOL_GPL(nft_register_chain_type); 1428 1429 void nft_unregister_chain_type(const struct nft_chain_type *ctype) 1430 { 1431 nfnl_lock(NFNL_SUBSYS_NFTABLES); 1432 chain_type[ctype->family][ctype->type] = NULL; 1433 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 1434 } 1435 EXPORT_SYMBOL_GPL(nft_unregister_chain_type); 1436 1437 /* 1438 * Chains 1439 */ 1440 1441 static struct nft_chain * 1442 nft_chain_lookup_byhandle(const struct nft_table *table, u64 handle, u8 genmask) 1443 { 1444 struct nft_chain *chain; 1445 1446 list_for_each_entry(chain, &table->chains, list) { 1447 if (chain->handle == handle && 1448 nft_active_genmask(chain, genmask)) 1449 return chain; 1450 } 1451 1452 return ERR_PTR(-ENOENT); 1453 } 1454 1455 static bool lockdep_commit_lock_is_held(const struct net *net) 1456 { 1457 #ifdef CONFIG_PROVE_LOCKING 1458 struct nftables_pernet *nft_net = nft_pernet(net); 1459 1460 return lockdep_is_held(&nft_net->commit_mutex); 1461 #else 1462 return true; 1463 #endif 1464 } 1465 1466 static struct nft_chain *nft_chain_lookup(struct net *net, 1467 struct nft_table *table, 1468 const struct nlattr *nla, u8 genmask) 1469 { 1470 char search[NFT_CHAIN_MAXNAMELEN + 1]; 1471 struct rhlist_head *tmp, *list; 1472 struct nft_chain *chain; 1473 1474 if (nla == NULL) 1475 return ERR_PTR(-EINVAL); 1476 1477 nla_strscpy(search, nla, sizeof(search)); 1478 1479 WARN_ON(!rcu_read_lock_held() && 1480 !lockdep_commit_lock_is_held(net)); 1481 1482 chain = ERR_PTR(-ENOENT); 1483 rcu_read_lock(); 1484 list = rhltable_lookup(&table->chains_ht, search, nft_chain_ht_params); 1485 if (!list) 1486 goto out_unlock; 1487 1488 rhl_for_each_entry_rcu(chain, tmp, list, rhlhead) { 1489 if (nft_active_genmask(chain, genmask)) 1490 goto out_unlock; 1491 } 1492 chain = ERR_PTR(-ENOENT); 1493 out_unlock: 1494 rcu_read_unlock(); 1495 return chain; 1496 } 1497 1498 static const struct nla_policy nft_chain_policy[NFTA_CHAIN_MAX + 1] = { 1499 [NFTA_CHAIN_TABLE] = { .type = NLA_STRING, 1500 .len = NFT_TABLE_MAXNAMELEN - 1 }, 1501 [NFTA_CHAIN_HANDLE] = { .type = NLA_U64 }, 1502 [NFTA_CHAIN_NAME] = { .type = NLA_STRING, 1503 .len = NFT_CHAIN_MAXNAMELEN - 1 }, 1504 [NFTA_CHAIN_HOOK] = { .type = NLA_NESTED }, 1505 [NFTA_CHAIN_POLICY] = { .type = NLA_U32 }, 1506 [NFTA_CHAIN_TYPE] = { .type = NLA_STRING, 1507 .len = NFT_MODULE_AUTOLOAD_LIMIT }, 1508 [NFTA_CHAIN_COUNTERS] = { .type = NLA_NESTED }, 1509 [NFTA_CHAIN_FLAGS] = { .type = NLA_U32 }, 1510 [NFTA_CHAIN_ID] = { .type = NLA_U32 }, 1511 [NFTA_CHAIN_USERDATA] = { .type = NLA_BINARY, 1512 .len = NFT_USERDATA_MAXLEN }, 1513 }; 1514 1515 static const struct nla_policy nft_hook_policy[NFTA_HOOK_MAX + 1] = { 1516 [NFTA_HOOK_HOOKNUM] = { .type = NLA_U32 }, 1517 [NFTA_HOOK_PRIORITY] = { .type = NLA_U32 }, 1518 [NFTA_HOOK_DEV] = { .type = NLA_STRING, 1519 .len = IFNAMSIZ - 1 }, 1520 }; 1521 1522 static int nft_dump_stats(struct sk_buff *skb, struct nft_stats __percpu *stats) 1523 { 1524 struct nft_stats *cpu_stats, total; 1525 struct nlattr *nest; 1526 unsigned int seq; 1527 u64 pkts, bytes; 1528 int cpu; 1529 1530 if (!stats) 1531 return 0; 1532 1533 memset(&total, 0, sizeof(total)); 1534 for_each_possible_cpu(cpu) { 1535 cpu_stats = per_cpu_ptr(stats, cpu); 1536 do { 1537 seq = u64_stats_fetch_begin_irq(&cpu_stats->syncp); 1538 pkts = cpu_stats->pkts; 1539 bytes = cpu_stats->bytes; 1540 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, seq)); 1541 total.pkts += pkts; 1542 total.bytes += bytes; 1543 } 1544 nest = nla_nest_start_noflag(skb, NFTA_CHAIN_COUNTERS); 1545 if (nest == NULL) 1546 goto nla_put_failure; 1547 1548 if (nla_put_be64(skb, NFTA_COUNTER_PACKETS, cpu_to_be64(total.pkts), 1549 NFTA_COUNTER_PAD) || 1550 nla_put_be64(skb, NFTA_COUNTER_BYTES, cpu_to_be64(total.bytes), 1551 NFTA_COUNTER_PAD)) 1552 goto nla_put_failure; 1553 1554 nla_nest_end(skb, nest); 1555 return 0; 1556 1557 nla_put_failure: 1558 return -ENOSPC; 1559 } 1560 1561 static int nft_dump_basechain_hook(struct sk_buff *skb, int family, 1562 const struct nft_base_chain *basechain) 1563 { 1564 const struct nf_hook_ops *ops = &basechain->ops; 1565 struct nft_hook *hook, *first = NULL; 1566 struct nlattr *nest, *nest_devs; 1567 int n = 0; 1568 1569 nest = nla_nest_start_noflag(skb, NFTA_CHAIN_HOOK); 1570 if (nest == NULL) 1571 goto nla_put_failure; 1572 if (nla_put_be32(skb, NFTA_HOOK_HOOKNUM, htonl(ops->hooknum))) 1573 goto nla_put_failure; 1574 if (nla_put_be32(skb, NFTA_HOOK_PRIORITY, htonl(ops->priority))) 1575 goto nla_put_failure; 1576 1577 if (nft_base_chain_netdev(family, ops->hooknum)) { 1578 nest_devs = nla_nest_start_noflag(skb, NFTA_HOOK_DEVS); 1579 list_for_each_entry(hook, &basechain->hook_list, list) { 1580 if (!first) 1581 first = hook; 1582 1583 if (nla_put_string(skb, NFTA_DEVICE_NAME, 1584 hook->ops.dev->name)) 1585 goto nla_put_failure; 1586 n++; 1587 } 1588 nla_nest_end(skb, nest_devs); 1589 1590 if (n == 1 && 1591 nla_put_string(skb, NFTA_HOOK_DEV, first->ops.dev->name)) 1592 goto nla_put_failure; 1593 } 1594 nla_nest_end(skb, nest); 1595 1596 return 0; 1597 nla_put_failure: 1598 return -1; 1599 } 1600 1601 static int nf_tables_fill_chain_info(struct sk_buff *skb, struct net *net, 1602 u32 portid, u32 seq, int event, u32 flags, 1603 int family, const struct nft_table *table, 1604 const struct nft_chain *chain) 1605 { 1606 struct nlmsghdr *nlh; 1607 1608 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 1609 nlh = nfnl_msg_put(skb, portid, seq, event, flags, family, 1610 NFNETLINK_V0, nft_base_seq(net)); 1611 if (!nlh) 1612 goto nla_put_failure; 1613 1614 if (nla_put_string(skb, NFTA_CHAIN_TABLE, table->name)) 1615 goto nla_put_failure; 1616 if (nla_put_be64(skb, NFTA_CHAIN_HANDLE, cpu_to_be64(chain->handle), 1617 NFTA_CHAIN_PAD)) 1618 goto nla_put_failure; 1619 if (nla_put_string(skb, NFTA_CHAIN_NAME, chain->name)) 1620 goto nla_put_failure; 1621 1622 if (nft_is_base_chain(chain)) { 1623 const struct nft_base_chain *basechain = nft_base_chain(chain); 1624 struct nft_stats __percpu *stats; 1625 1626 if (nft_dump_basechain_hook(skb, family, basechain)) 1627 goto nla_put_failure; 1628 1629 if (nla_put_be32(skb, NFTA_CHAIN_POLICY, 1630 htonl(basechain->policy))) 1631 goto nla_put_failure; 1632 1633 if (nla_put_string(skb, NFTA_CHAIN_TYPE, basechain->type->name)) 1634 goto nla_put_failure; 1635 1636 stats = rcu_dereference_check(basechain->stats, 1637 lockdep_commit_lock_is_held(net)); 1638 if (nft_dump_stats(skb, stats)) 1639 goto nla_put_failure; 1640 } 1641 1642 if (chain->flags && 1643 nla_put_be32(skb, NFTA_CHAIN_FLAGS, htonl(chain->flags))) 1644 goto nla_put_failure; 1645 1646 if (nla_put_be32(skb, NFTA_CHAIN_USE, htonl(chain->use))) 1647 goto nla_put_failure; 1648 1649 if (chain->udata && 1650 nla_put(skb, NFTA_CHAIN_USERDATA, chain->udlen, chain->udata)) 1651 goto nla_put_failure; 1652 1653 nlmsg_end(skb, nlh); 1654 return 0; 1655 1656 nla_put_failure: 1657 nlmsg_trim(skb, nlh); 1658 return -1; 1659 } 1660 1661 static void nf_tables_chain_notify(const struct nft_ctx *ctx, int event) 1662 { 1663 struct nftables_pernet *nft_net; 1664 struct sk_buff *skb; 1665 u16 flags = 0; 1666 int err; 1667 1668 if (!ctx->report && 1669 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES)) 1670 return; 1671 1672 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 1673 if (skb == NULL) 1674 goto err; 1675 1676 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 1677 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 1678 1679 err = nf_tables_fill_chain_info(skb, ctx->net, ctx->portid, ctx->seq, 1680 event, flags, ctx->family, ctx->table, 1681 ctx->chain); 1682 if (err < 0) { 1683 kfree_skb(skb); 1684 goto err; 1685 } 1686 1687 nft_net = nft_pernet(ctx->net); 1688 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 1689 return; 1690 err: 1691 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS); 1692 } 1693 1694 static int nf_tables_dump_chains(struct sk_buff *skb, 1695 struct netlink_callback *cb) 1696 { 1697 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 1698 unsigned int idx = 0, s_idx = cb->args[0]; 1699 struct net *net = sock_net(skb->sk); 1700 int family = nfmsg->nfgen_family; 1701 struct nftables_pernet *nft_net; 1702 const struct nft_table *table; 1703 const struct nft_chain *chain; 1704 1705 rcu_read_lock(); 1706 nft_net = nft_pernet(net); 1707 cb->seq = READ_ONCE(nft_net->base_seq); 1708 1709 list_for_each_entry_rcu(table, &nft_net->tables, list) { 1710 if (family != NFPROTO_UNSPEC && family != table->family) 1711 continue; 1712 1713 list_for_each_entry_rcu(chain, &table->chains, list) { 1714 if (idx < s_idx) 1715 goto cont; 1716 if (idx > s_idx) 1717 memset(&cb->args[1], 0, 1718 sizeof(cb->args) - sizeof(cb->args[0])); 1719 if (!nft_is_active(net, chain)) 1720 continue; 1721 if (nf_tables_fill_chain_info(skb, net, 1722 NETLINK_CB(cb->skb).portid, 1723 cb->nlh->nlmsg_seq, 1724 NFT_MSG_NEWCHAIN, 1725 NLM_F_MULTI, 1726 table->family, table, 1727 chain) < 0) 1728 goto done; 1729 1730 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1731 cont: 1732 idx++; 1733 } 1734 } 1735 done: 1736 rcu_read_unlock(); 1737 cb->args[0] = idx; 1738 return skb->len; 1739 } 1740 1741 /* called with rcu_read_lock held */ 1742 static int nf_tables_getchain(struct sk_buff *skb, const struct nfnl_info *info, 1743 const struct nlattr * const nla[]) 1744 { 1745 struct netlink_ext_ack *extack = info->extack; 1746 u8 genmask = nft_genmask_cur(info->net); 1747 u8 family = info->nfmsg->nfgen_family; 1748 const struct nft_chain *chain; 1749 struct net *net = info->net; 1750 struct nft_table *table; 1751 struct sk_buff *skb2; 1752 int err; 1753 1754 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 1755 struct netlink_dump_control c = { 1756 .dump = nf_tables_dump_chains, 1757 .module = THIS_MODULE, 1758 }; 1759 1760 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 1761 } 1762 1763 table = nft_table_lookup(net, nla[NFTA_CHAIN_TABLE], family, genmask, 0); 1764 if (IS_ERR(table)) { 1765 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_TABLE]); 1766 return PTR_ERR(table); 1767 } 1768 1769 chain = nft_chain_lookup(net, table, nla[NFTA_CHAIN_NAME], genmask); 1770 if (IS_ERR(chain)) { 1771 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_NAME]); 1772 return PTR_ERR(chain); 1773 } 1774 1775 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 1776 if (!skb2) 1777 return -ENOMEM; 1778 1779 err = nf_tables_fill_chain_info(skb2, net, NETLINK_CB(skb).portid, 1780 info->nlh->nlmsg_seq, NFT_MSG_NEWCHAIN, 1781 0, family, table, chain); 1782 if (err < 0) 1783 goto err_fill_chain_info; 1784 1785 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 1786 1787 err_fill_chain_info: 1788 kfree_skb(skb2); 1789 return err; 1790 } 1791 1792 static const struct nla_policy nft_counter_policy[NFTA_COUNTER_MAX + 1] = { 1793 [NFTA_COUNTER_PACKETS] = { .type = NLA_U64 }, 1794 [NFTA_COUNTER_BYTES] = { .type = NLA_U64 }, 1795 }; 1796 1797 static struct nft_stats __percpu *nft_stats_alloc(const struct nlattr *attr) 1798 { 1799 struct nlattr *tb[NFTA_COUNTER_MAX+1]; 1800 struct nft_stats __percpu *newstats; 1801 struct nft_stats *stats; 1802 int err; 1803 1804 err = nla_parse_nested_deprecated(tb, NFTA_COUNTER_MAX, attr, 1805 nft_counter_policy, NULL); 1806 if (err < 0) 1807 return ERR_PTR(err); 1808 1809 if (!tb[NFTA_COUNTER_BYTES] || !tb[NFTA_COUNTER_PACKETS]) 1810 return ERR_PTR(-EINVAL); 1811 1812 newstats = netdev_alloc_pcpu_stats(struct nft_stats); 1813 if (newstats == NULL) 1814 return ERR_PTR(-ENOMEM); 1815 1816 /* Restore old counters on this cpu, no problem. Per-cpu statistics 1817 * are not exposed to userspace. 1818 */ 1819 preempt_disable(); 1820 stats = this_cpu_ptr(newstats); 1821 stats->bytes = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_BYTES])); 1822 stats->pkts = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_PACKETS])); 1823 preempt_enable(); 1824 1825 return newstats; 1826 } 1827 1828 static void nft_chain_stats_replace(struct nft_trans *trans) 1829 { 1830 struct nft_base_chain *chain = nft_base_chain(trans->ctx.chain); 1831 1832 if (!nft_trans_chain_stats(trans)) 1833 return; 1834 1835 nft_trans_chain_stats(trans) = 1836 rcu_replace_pointer(chain->stats, nft_trans_chain_stats(trans), 1837 lockdep_commit_lock_is_held(trans->ctx.net)); 1838 1839 if (!nft_trans_chain_stats(trans)) 1840 static_branch_inc(&nft_counters_enabled); 1841 } 1842 1843 static void nf_tables_chain_free_chain_rules(struct nft_chain *chain) 1844 { 1845 struct nft_rule_blob *g0 = rcu_dereference_raw(chain->blob_gen_0); 1846 struct nft_rule_blob *g1 = rcu_dereference_raw(chain->blob_gen_1); 1847 1848 if (g0 != g1) 1849 kvfree(g1); 1850 kvfree(g0); 1851 1852 /* should be NULL either via abort or via successful commit */ 1853 WARN_ON_ONCE(chain->blob_next); 1854 kvfree(chain->blob_next); 1855 } 1856 1857 void nf_tables_chain_destroy(struct nft_ctx *ctx) 1858 { 1859 struct nft_chain *chain = ctx->chain; 1860 struct nft_hook *hook, *next; 1861 1862 if (WARN_ON(chain->use > 0)) 1863 return; 1864 1865 /* no concurrent access possible anymore */ 1866 nf_tables_chain_free_chain_rules(chain); 1867 1868 if (nft_is_base_chain(chain)) { 1869 struct nft_base_chain *basechain = nft_base_chain(chain); 1870 1871 if (nft_base_chain_netdev(ctx->family, basechain->ops.hooknum)) { 1872 list_for_each_entry_safe(hook, next, 1873 &basechain->hook_list, list) { 1874 list_del_rcu(&hook->list); 1875 kfree_rcu(hook, rcu); 1876 } 1877 } 1878 module_put(basechain->type->owner); 1879 if (rcu_access_pointer(basechain->stats)) { 1880 static_branch_dec(&nft_counters_enabled); 1881 free_percpu(rcu_dereference_raw(basechain->stats)); 1882 } 1883 kfree(chain->name); 1884 kfree(chain->udata); 1885 kfree(basechain); 1886 } else { 1887 kfree(chain->name); 1888 kfree(chain->udata); 1889 kfree(chain); 1890 } 1891 } 1892 1893 static struct nft_hook *nft_netdev_hook_alloc(struct net *net, 1894 const struct nlattr *attr) 1895 { 1896 struct net_device *dev; 1897 char ifname[IFNAMSIZ]; 1898 struct nft_hook *hook; 1899 int err; 1900 1901 hook = kmalloc(sizeof(struct nft_hook), GFP_KERNEL_ACCOUNT); 1902 if (!hook) { 1903 err = -ENOMEM; 1904 goto err_hook_alloc; 1905 } 1906 1907 nla_strscpy(ifname, attr, IFNAMSIZ); 1908 /* nf_tables_netdev_event() is called under rtnl_mutex, this is 1909 * indirectly serializing all the other holders of the commit_mutex with 1910 * the rtnl_mutex. 1911 */ 1912 dev = __dev_get_by_name(net, ifname); 1913 if (!dev) { 1914 err = -ENOENT; 1915 goto err_hook_dev; 1916 } 1917 hook->ops.dev = dev; 1918 1919 return hook; 1920 1921 err_hook_dev: 1922 kfree(hook); 1923 err_hook_alloc: 1924 return ERR_PTR(err); 1925 } 1926 1927 static struct nft_hook *nft_hook_list_find(struct list_head *hook_list, 1928 const struct nft_hook *this) 1929 { 1930 struct nft_hook *hook; 1931 1932 list_for_each_entry(hook, hook_list, list) { 1933 if (this->ops.dev == hook->ops.dev) 1934 return hook; 1935 } 1936 1937 return NULL; 1938 } 1939 1940 static int nf_tables_parse_netdev_hooks(struct net *net, 1941 const struct nlattr *attr, 1942 struct list_head *hook_list) 1943 { 1944 struct nft_hook *hook, *next; 1945 const struct nlattr *tmp; 1946 int rem, n = 0, err; 1947 1948 nla_for_each_nested(tmp, attr, rem) { 1949 if (nla_type(tmp) != NFTA_DEVICE_NAME) { 1950 err = -EINVAL; 1951 goto err_hook; 1952 } 1953 1954 hook = nft_netdev_hook_alloc(net, tmp); 1955 if (IS_ERR(hook)) { 1956 err = PTR_ERR(hook); 1957 goto err_hook; 1958 } 1959 if (nft_hook_list_find(hook_list, hook)) { 1960 kfree(hook); 1961 err = -EEXIST; 1962 goto err_hook; 1963 } 1964 list_add_tail(&hook->list, hook_list); 1965 n++; 1966 1967 if (n == NFT_NETDEVICE_MAX) { 1968 err = -EFBIG; 1969 goto err_hook; 1970 } 1971 } 1972 1973 return 0; 1974 1975 err_hook: 1976 list_for_each_entry_safe(hook, next, hook_list, list) { 1977 list_del(&hook->list); 1978 kfree(hook); 1979 } 1980 return err; 1981 } 1982 1983 struct nft_chain_hook { 1984 u32 num; 1985 s32 priority; 1986 const struct nft_chain_type *type; 1987 struct list_head list; 1988 }; 1989 1990 static int nft_chain_parse_netdev(struct net *net, 1991 struct nlattr *tb[], 1992 struct list_head *hook_list) 1993 { 1994 struct nft_hook *hook; 1995 int err; 1996 1997 if (tb[NFTA_HOOK_DEV]) { 1998 hook = nft_netdev_hook_alloc(net, tb[NFTA_HOOK_DEV]); 1999 if (IS_ERR(hook)) 2000 return PTR_ERR(hook); 2001 2002 list_add_tail(&hook->list, hook_list); 2003 } else if (tb[NFTA_HOOK_DEVS]) { 2004 err = nf_tables_parse_netdev_hooks(net, tb[NFTA_HOOK_DEVS], 2005 hook_list); 2006 if (err < 0) 2007 return err; 2008 2009 if (list_empty(hook_list)) 2010 return -EINVAL; 2011 } else { 2012 return -EINVAL; 2013 } 2014 2015 return 0; 2016 } 2017 2018 static int nft_chain_parse_hook(struct net *net, 2019 const struct nlattr * const nla[], 2020 struct nft_chain_hook *hook, u8 family, 2021 struct netlink_ext_ack *extack, bool autoload) 2022 { 2023 struct nftables_pernet *nft_net = nft_pernet(net); 2024 struct nlattr *ha[NFTA_HOOK_MAX + 1]; 2025 const struct nft_chain_type *type; 2026 int err; 2027 2028 lockdep_assert_held(&nft_net->commit_mutex); 2029 lockdep_nfnl_nft_mutex_not_held(); 2030 2031 err = nla_parse_nested_deprecated(ha, NFTA_HOOK_MAX, 2032 nla[NFTA_CHAIN_HOOK], 2033 nft_hook_policy, NULL); 2034 if (err < 0) 2035 return err; 2036 2037 if (ha[NFTA_HOOK_HOOKNUM] == NULL || 2038 ha[NFTA_HOOK_PRIORITY] == NULL) 2039 return -EINVAL; 2040 2041 hook->num = ntohl(nla_get_be32(ha[NFTA_HOOK_HOOKNUM])); 2042 hook->priority = ntohl(nla_get_be32(ha[NFTA_HOOK_PRIORITY])); 2043 2044 type = __nft_chain_type_get(family, NFT_CHAIN_T_DEFAULT); 2045 if (!type) 2046 return -EOPNOTSUPP; 2047 2048 if (nla[NFTA_CHAIN_TYPE]) { 2049 type = nf_tables_chain_type_lookup(net, nla[NFTA_CHAIN_TYPE], 2050 family, autoload); 2051 if (IS_ERR(type)) { 2052 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_TYPE]); 2053 return PTR_ERR(type); 2054 } 2055 } 2056 if (hook->num >= NFT_MAX_HOOKS || !(type->hook_mask & (1 << hook->num))) 2057 return -EOPNOTSUPP; 2058 2059 if (type->type == NFT_CHAIN_T_NAT && 2060 hook->priority <= NF_IP_PRI_CONNTRACK) 2061 return -EOPNOTSUPP; 2062 2063 if (!try_module_get(type->owner)) { 2064 if (nla[NFTA_CHAIN_TYPE]) 2065 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_TYPE]); 2066 return -ENOENT; 2067 } 2068 2069 hook->type = type; 2070 2071 INIT_LIST_HEAD(&hook->list); 2072 if (nft_base_chain_netdev(family, hook->num)) { 2073 err = nft_chain_parse_netdev(net, ha, &hook->list); 2074 if (err < 0) { 2075 module_put(type->owner); 2076 return err; 2077 } 2078 } else if (ha[NFTA_HOOK_DEV] || ha[NFTA_HOOK_DEVS]) { 2079 module_put(type->owner); 2080 return -EOPNOTSUPP; 2081 } 2082 2083 return 0; 2084 } 2085 2086 static void nft_chain_release_hook(struct nft_chain_hook *hook) 2087 { 2088 struct nft_hook *h, *next; 2089 2090 list_for_each_entry_safe(h, next, &hook->list, list) { 2091 list_del(&h->list); 2092 kfree(h); 2093 } 2094 module_put(hook->type->owner); 2095 } 2096 2097 struct nft_rules_old { 2098 struct rcu_head h; 2099 struct nft_rule_blob *blob; 2100 }; 2101 2102 static void nft_last_rule(struct nft_rule_blob *blob, const void *ptr) 2103 { 2104 struct nft_rule_dp *prule; 2105 2106 prule = (struct nft_rule_dp *)ptr; 2107 prule->is_last = 1; 2108 /* blob size does not include the trailer rule */ 2109 } 2110 2111 static struct nft_rule_blob *nf_tables_chain_alloc_rules(unsigned int size) 2112 { 2113 struct nft_rule_blob *blob; 2114 2115 /* size must include room for the last rule */ 2116 if (size < offsetof(struct nft_rule_dp, data)) 2117 return NULL; 2118 2119 size += sizeof(struct nft_rule_blob) + sizeof(struct nft_rules_old); 2120 if (size > INT_MAX) 2121 return NULL; 2122 2123 blob = kvmalloc(size, GFP_KERNEL_ACCOUNT); 2124 if (!blob) 2125 return NULL; 2126 2127 blob->size = 0; 2128 nft_last_rule(blob, blob->data); 2129 2130 return blob; 2131 } 2132 2133 static void nft_basechain_hook_init(struct nf_hook_ops *ops, u8 family, 2134 const struct nft_chain_hook *hook, 2135 struct nft_chain *chain) 2136 { 2137 ops->pf = family; 2138 ops->hooknum = hook->num; 2139 ops->priority = hook->priority; 2140 ops->priv = chain; 2141 ops->hook = hook->type->hooks[ops->hooknum]; 2142 ops->hook_ops_type = NF_HOOK_OP_NF_TABLES; 2143 } 2144 2145 static int nft_basechain_init(struct nft_base_chain *basechain, u8 family, 2146 struct nft_chain_hook *hook, u32 flags) 2147 { 2148 struct nft_chain *chain; 2149 struct nft_hook *h; 2150 2151 basechain->type = hook->type; 2152 INIT_LIST_HEAD(&basechain->hook_list); 2153 chain = &basechain->chain; 2154 2155 if (nft_base_chain_netdev(family, hook->num)) { 2156 list_splice_init(&hook->list, &basechain->hook_list); 2157 list_for_each_entry(h, &basechain->hook_list, list) 2158 nft_basechain_hook_init(&h->ops, family, hook, chain); 2159 2160 basechain->ops.hooknum = hook->num; 2161 basechain->ops.priority = hook->priority; 2162 } else { 2163 nft_basechain_hook_init(&basechain->ops, family, hook, chain); 2164 } 2165 2166 chain->flags |= NFT_CHAIN_BASE | flags; 2167 basechain->policy = NF_ACCEPT; 2168 if (chain->flags & NFT_CHAIN_HW_OFFLOAD && 2169 !nft_chain_offload_support(basechain)) { 2170 list_splice_init(&basechain->hook_list, &hook->list); 2171 return -EOPNOTSUPP; 2172 } 2173 2174 flow_block_init(&basechain->flow_block); 2175 2176 return 0; 2177 } 2178 2179 static int nft_chain_add(struct nft_table *table, struct nft_chain *chain) 2180 { 2181 int err; 2182 2183 err = rhltable_insert_key(&table->chains_ht, chain->name, 2184 &chain->rhlhead, nft_chain_ht_params); 2185 if (err) 2186 return err; 2187 2188 list_add_tail_rcu(&chain->list, &table->chains); 2189 2190 return 0; 2191 } 2192 2193 static u64 chain_id; 2194 2195 static int nf_tables_addchain(struct nft_ctx *ctx, u8 family, u8 genmask, 2196 u8 policy, u32 flags, 2197 struct netlink_ext_ack *extack) 2198 { 2199 const struct nlattr * const *nla = ctx->nla; 2200 struct nft_table *table = ctx->table; 2201 struct nft_base_chain *basechain; 2202 struct net *net = ctx->net; 2203 char name[NFT_NAME_MAXLEN]; 2204 struct nft_rule_blob *blob; 2205 struct nft_trans *trans; 2206 struct nft_chain *chain; 2207 unsigned int data_size; 2208 int err; 2209 2210 if (table->use == UINT_MAX) 2211 return -EOVERFLOW; 2212 2213 if (nla[NFTA_CHAIN_HOOK]) { 2214 struct nft_stats __percpu *stats = NULL; 2215 struct nft_chain_hook hook; 2216 2217 if (flags & NFT_CHAIN_BINDING) 2218 return -EOPNOTSUPP; 2219 2220 err = nft_chain_parse_hook(net, nla, &hook, family, extack, 2221 true); 2222 if (err < 0) 2223 return err; 2224 2225 basechain = kzalloc(sizeof(*basechain), GFP_KERNEL_ACCOUNT); 2226 if (basechain == NULL) { 2227 nft_chain_release_hook(&hook); 2228 return -ENOMEM; 2229 } 2230 chain = &basechain->chain; 2231 2232 if (nla[NFTA_CHAIN_COUNTERS]) { 2233 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]); 2234 if (IS_ERR(stats)) { 2235 nft_chain_release_hook(&hook); 2236 kfree(basechain); 2237 return PTR_ERR(stats); 2238 } 2239 rcu_assign_pointer(basechain->stats, stats); 2240 } 2241 2242 err = nft_basechain_init(basechain, family, &hook, flags); 2243 if (err < 0) { 2244 nft_chain_release_hook(&hook); 2245 kfree(basechain); 2246 free_percpu(stats); 2247 return err; 2248 } 2249 if (stats) 2250 static_branch_inc(&nft_counters_enabled); 2251 } else { 2252 if (flags & NFT_CHAIN_BASE) 2253 return -EINVAL; 2254 if (flags & NFT_CHAIN_HW_OFFLOAD) 2255 return -EOPNOTSUPP; 2256 2257 chain = kzalloc(sizeof(*chain), GFP_KERNEL_ACCOUNT); 2258 if (chain == NULL) 2259 return -ENOMEM; 2260 2261 chain->flags = flags; 2262 } 2263 ctx->chain = chain; 2264 2265 INIT_LIST_HEAD(&chain->rules); 2266 chain->handle = nf_tables_alloc_handle(table); 2267 chain->table = table; 2268 2269 if (nla[NFTA_CHAIN_NAME]) { 2270 chain->name = nla_strdup(nla[NFTA_CHAIN_NAME], GFP_KERNEL_ACCOUNT); 2271 } else { 2272 if (!(flags & NFT_CHAIN_BINDING)) { 2273 err = -EINVAL; 2274 goto err_destroy_chain; 2275 } 2276 2277 snprintf(name, sizeof(name), "__chain%llu", ++chain_id); 2278 chain->name = kstrdup(name, GFP_KERNEL_ACCOUNT); 2279 } 2280 2281 if (!chain->name) { 2282 err = -ENOMEM; 2283 goto err_destroy_chain; 2284 } 2285 2286 if (nla[NFTA_CHAIN_USERDATA]) { 2287 chain->udata = nla_memdup(nla[NFTA_CHAIN_USERDATA], GFP_KERNEL_ACCOUNT); 2288 if (chain->udata == NULL) { 2289 err = -ENOMEM; 2290 goto err_destroy_chain; 2291 } 2292 chain->udlen = nla_len(nla[NFTA_CHAIN_USERDATA]); 2293 } 2294 2295 data_size = offsetof(struct nft_rule_dp, data); /* last rule */ 2296 blob = nf_tables_chain_alloc_rules(data_size); 2297 if (!blob) { 2298 err = -ENOMEM; 2299 goto err_destroy_chain; 2300 } 2301 2302 RCU_INIT_POINTER(chain->blob_gen_0, blob); 2303 RCU_INIT_POINTER(chain->blob_gen_1, blob); 2304 2305 err = nf_tables_register_hook(net, table, chain); 2306 if (err < 0) 2307 goto err_destroy_chain; 2308 2309 trans = nft_trans_chain_add(ctx, NFT_MSG_NEWCHAIN); 2310 if (IS_ERR(trans)) { 2311 err = PTR_ERR(trans); 2312 goto err_unregister_hook; 2313 } 2314 2315 nft_trans_chain_policy(trans) = NFT_CHAIN_POLICY_UNSET; 2316 if (nft_is_base_chain(chain)) 2317 nft_trans_chain_policy(trans) = policy; 2318 2319 err = nft_chain_add(table, chain); 2320 if (err < 0) { 2321 nft_trans_destroy(trans); 2322 goto err_unregister_hook; 2323 } 2324 2325 table->use++; 2326 2327 return 0; 2328 err_unregister_hook: 2329 nf_tables_unregister_hook(net, table, chain); 2330 err_destroy_chain: 2331 nf_tables_chain_destroy(ctx); 2332 2333 return err; 2334 } 2335 2336 static bool nft_hook_list_equal(struct list_head *hook_list1, 2337 struct list_head *hook_list2) 2338 { 2339 struct nft_hook *hook; 2340 int n = 0, m = 0; 2341 2342 n = 0; 2343 list_for_each_entry(hook, hook_list2, list) { 2344 if (!nft_hook_list_find(hook_list1, hook)) 2345 return false; 2346 2347 n++; 2348 } 2349 list_for_each_entry(hook, hook_list1, list) 2350 m++; 2351 2352 return n == m; 2353 } 2354 2355 static int nf_tables_updchain(struct nft_ctx *ctx, u8 genmask, u8 policy, 2356 u32 flags, const struct nlattr *attr, 2357 struct netlink_ext_ack *extack) 2358 { 2359 const struct nlattr * const *nla = ctx->nla; 2360 struct nft_table *table = ctx->table; 2361 struct nft_chain *chain = ctx->chain; 2362 struct nft_base_chain *basechain; 2363 struct nft_stats *stats = NULL; 2364 struct nft_chain_hook hook; 2365 struct nf_hook_ops *ops; 2366 struct nft_trans *trans; 2367 int err; 2368 2369 if (chain->flags ^ flags) 2370 return -EOPNOTSUPP; 2371 2372 if (nla[NFTA_CHAIN_HOOK]) { 2373 if (!nft_is_base_chain(chain)) { 2374 NL_SET_BAD_ATTR(extack, attr); 2375 return -EEXIST; 2376 } 2377 err = nft_chain_parse_hook(ctx->net, nla, &hook, ctx->family, 2378 extack, false); 2379 if (err < 0) 2380 return err; 2381 2382 basechain = nft_base_chain(chain); 2383 if (basechain->type != hook.type) { 2384 nft_chain_release_hook(&hook); 2385 NL_SET_BAD_ATTR(extack, attr); 2386 return -EEXIST; 2387 } 2388 2389 if (nft_base_chain_netdev(ctx->family, hook.num)) { 2390 if (!nft_hook_list_equal(&basechain->hook_list, 2391 &hook.list)) { 2392 nft_chain_release_hook(&hook); 2393 NL_SET_BAD_ATTR(extack, attr); 2394 return -EEXIST; 2395 } 2396 } else { 2397 ops = &basechain->ops; 2398 if (ops->hooknum != hook.num || 2399 ops->priority != hook.priority) { 2400 nft_chain_release_hook(&hook); 2401 NL_SET_BAD_ATTR(extack, attr); 2402 return -EEXIST; 2403 } 2404 } 2405 nft_chain_release_hook(&hook); 2406 } 2407 2408 if (nla[NFTA_CHAIN_HANDLE] && 2409 nla[NFTA_CHAIN_NAME]) { 2410 struct nft_chain *chain2; 2411 2412 chain2 = nft_chain_lookup(ctx->net, table, 2413 nla[NFTA_CHAIN_NAME], genmask); 2414 if (!IS_ERR(chain2)) { 2415 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_NAME]); 2416 return -EEXIST; 2417 } 2418 } 2419 2420 if (nla[NFTA_CHAIN_COUNTERS]) { 2421 if (!nft_is_base_chain(chain)) 2422 return -EOPNOTSUPP; 2423 2424 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]); 2425 if (IS_ERR(stats)) 2426 return PTR_ERR(stats); 2427 } 2428 2429 err = -ENOMEM; 2430 trans = nft_trans_alloc(ctx, NFT_MSG_NEWCHAIN, 2431 sizeof(struct nft_trans_chain)); 2432 if (trans == NULL) 2433 goto err; 2434 2435 nft_trans_chain_stats(trans) = stats; 2436 nft_trans_chain_update(trans) = true; 2437 2438 if (nla[NFTA_CHAIN_POLICY]) 2439 nft_trans_chain_policy(trans) = policy; 2440 else 2441 nft_trans_chain_policy(trans) = -1; 2442 2443 if (nla[NFTA_CHAIN_HANDLE] && 2444 nla[NFTA_CHAIN_NAME]) { 2445 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 2446 struct nft_trans *tmp; 2447 char *name; 2448 2449 err = -ENOMEM; 2450 name = nla_strdup(nla[NFTA_CHAIN_NAME], GFP_KERNEL_ACCOUNT); 2451 if (!name) 2452 goto err; 2453 2454 err = -EEXIST; 2455 list_for_each_entry(tmp, &nft_net->commit_list, list) { 2456 if (tmp->msg_type == NFT_MSG_NEWCHAIN && 2457 tmp->ctx.table == table && 2458 nft_trans_chain_update(tmp) && 2459 nft_trans_chain_name(tmp) && 2460 strcmp(name, nft_trans_chain_name(tmp)) == 0) { 2461 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_NAME]); 2462 kfree(name); 2463 goto err; 2464 } 2465 } 2466 2467 nft_trans_chain_name(trans) = name; 2468 } 2469 nft_trans_commit_list_add_tail(ctx->net, trans); 2470 2471 return 0; 2472 err: 2473 free_percpu(stats); 2474 kfree(trans); 2475 return err; 2476 } 2477 2478 static struct nft_chain *nft_chain_lookup_byid(const struct net *net, 2479 const struct nft_table *table, 2480 const struct nlattr *nla) 2481 { 2482 struct nftables_pernet *nft_net = nft_pernet(net); 2483 u32 id = ntohl(nla_get_be32(nla)); 2484 struct nft_trans *trans; 2485 2486 list_for_each_entry(trans, &nft_net->commit_list, list) { 2487 struct nft_chain *chain = trans->ctx.chain; 2488 2489 if (trans->msg_type == NFT_MSG_NEWCHAIN && 2490 chain->table == table && 2491 id == nft_trans_chain_id(trans)) 2492 return chain; 2493 } 2494 return ERR_PTR(-ENOENT); 2495 } 2496 2497 static int nf_tables_newchain(struct sk_buff *skb, const struct nfnl_info *info, 2498 const struct nlattr * const nla[]) 2499 { 2500 struct nftables_pernet *nft_net = nft_pernet(info->net); 2501 struct netlink_ext_ack *extack = info->extack; 2502 u8 genmask = nft_genmask_next(info->net); 2503 u8 family = info->nfmsg->nfgen_family; 2504 struct nft_chain *chain = NULL; 2505 struct net *net = info->net; 2506 const struct nlattr *attr; 2507 struct nft_table *table; 2508 u8 policy = NF_ACCEPT; 2509 struct nft_ctx ctx; 2510 u64 handle = 0; 2511 u32 flags = 0; 2512 2513 lockdep_assert_held(&nft_net->commit_mutex); 2514 2515 table = nft_table_lookup(net, nla[NFTA_CHAIN_TABLE], family, genmask, 2516 NETLINK_CB(skb).portid); 2517 if (IS_ERR(table)) { 2518 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_TABLE]); 2519 return PTR_ERR(table); 2520 } 2521 2522 chain = NULL; 2523 attr = nla[NFTA_CHAIN_NAME]; 2524 2525 if (nla[NFTA_CHAIN_HANDLE]) { 2526 handle = be64_to_cpu(nla_get_be64(nla[NFTA_CHAIN_HANDLE])); 2527 chain = nft_chain_lookup_byhandle(table, handle, genmask); 2528 if (IS_ERR(chain)) { 2529 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_HANDLE]); 2530 return PTR_ERR(chain); 2531 } 2532 attr = nla[NFTA_CHAIN_HANDLE]; 2533 } else if (nla[NFTA_CHAIN_NAME]) { 2534 chain = nft_chain_lookup(net, table, attr, genmask); 2535 if (IS_ERR(chain)) { 2536 if (PTR_ERR(chain) != -ENOENT) { 2537 NL_SET_BAD_ATTR(extack, attr); 2538 return PTR_ERR(chain); 2539 } 2540 chain = NULL; 2541 } 2542 } else if (!nla[NFTA_CHAIN_ID]) { 2543 return -EINVAL; 2544 } 2545 2546 if (nla[NFTA_CHAIN_POLICY]) { 2547 if (chain != NULL && 2548 !nft_is_base_chain(chain)) { 2549 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_POLICY]); 2550 return -EOPNOTSUPP; 2551 } 2552 2553 if (chain == NULL && 2554 nla[NFTA_CHAIN_HOOK] == NULL) { 2555 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_POLICY]); 2556 return -EOPNOTSUPP; 2557 } 2558 2559 policy = ntohl(nla_get_be32(nla[NFTA_CHAIN_POLICY])); 2560 switch (policy) { 2561 case NF_DROP: 2562 case NF_ACCEPT: 2563 break; 2564 default: 2565 return -EINVAL; 2566 } 2567 } 2568 2569 if (nla[NFTA_CHAIN_FLAGS]) 2570 flags = ntohl(nla_get_be32(nla[NFTA_CHAIN_FLAGS])); 2571 else if (chain) 2572 flags = chain->flags; 2573 2574 if (flags & ~NFT_CHAIN_FLAGS) 2575 return -EOPNOTSUPP; 2576 2577 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, chain, nla); 2578 2579 if (chain != NULL) { 2580 if (chain->flags & NFT_CHAIN_BINDING) 2581 return -EINVAL; 2582 2583 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 2584 NL_SET_BAD_ATTR(extack, attr); 2585 return -EEXIST; 2586 } 2587 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 2588 return -EOPNOTSUPP; 2589 2590 flags |= chain->flags & NFT_CHAIN_BASE; 2591 return nf_tables_updchain(&ctx, genmask, policy, flags, attr, 2592 extack); 2593 } 2594 2595 return nf_tables_addchain(&ctx, family, genmask, policy, flags, extack); 2596 } 2597 2598 static int nf_tables_delchain(struct sk_buff *skb, const struct nfnl_info *info, 2599 const struct nlattr * const nla[]) 2600 { 2601 struct netlink_ext_ack *extack = info->extack; 2602 u8 genmask = nft_genmask_next(info->net); 2603 u8 family = info->nfmsg->nfgen_family; 2604 struct net *net = info->net; 2605 const struct nlattr *attr; 2606 struct nft_table *table; 2607 struct nft_chain *chain; 2608 struct nft_rule *rule; 2609 struct nft_ctx ctx; 2610 u64 handle; 2611 u32 use; 2612 int err; 2613 2614 table = nft_table_lookup(net, nla[NFTA_CHAIN_TABLE], family, genmask, 2615 NETLINK_CB(skb).portid); 2616 if (IS_ERR(table)) { 2617 NL_SET_BAD_ATTR(extack, nla[NFTA_CHAIN_TABLE]); 2618 return PTR_ERR(table); 2619 } 2620 2621 if (nla[NFTA_CHAIN_HANDLE]) { 2622 attr = nla[NFTA_CHAIN_HANDLE]; 2623 handle = be64_to_cpu(nla_get_be64(attr)); 2624 chain = nft_chain_lookup_byhandle(table, handle, genmask); 2625 } else { 2626 attr = nla[NFTA_CHAIN_NAME]; 2627 chain = nft_chain_lookup(net, table, attr, genmask); 2628 } 2629 if (IS_ERR(chain)) { 2630 NL_SET_BAD_ATTR(extack, attr); 2631 return PTR_ERR(chain); 2632 } 2633 2634 if (info->nlh->nlmsg_flags & NLM_F_NONREC && 2635 chain->use > 0) 2636 return -EBUSY; 2637 2638 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, chain, nla); 2639 2640 use = chain->use; 2641 list_for_each_entry(rule, &chain->rules, list) { 2642 if (!nft_is_active_next(net, rule)) 2643 continue; 2644 use--; 2645 2646 err = nft_delrule(&ctx, rule); 2647 if (err < 0) 2648 return err; 2649 } 2650 2651 /* There are rules and elements that are still holding references to us, 2652 * we cannot do a recursive removal in this case. 2653 */ 2654 if (use > 0) { 2655 NL_SET_BAD_ATTR(extack, attr); 2656 return -EBUSY; 2657 } 2658 2659 return nft_delchain(&ctx); 2660 } 2661 2662 /* 2663 * Expressions 2664 */ 2665 2666 /** 2667 * nft_register_expr - register nf_tables expr type 2668 * @type: expr type 2669 * 2670 * Registers the expr type for use with nf_tables. Returns zero on 2671 * success or a negative errno code otherwise. 2672 */ 2673 int nft_register_expr(struct nft_expr_type *type) 2674 { 2675 nfnl_lock(NFNL_SUBSYS_NFTABLES); 2676 if (type->family == NFPROTO_UNSPEC) 2677 list_add_tail_rcu(&type->list, &nf_tables_expressions); 2678 else 2679 list_add_rcu(&type->list, &nf_tables_expressions); 2680 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 2681 return 0; 2682 } 2683 EXPORT_SYMBOL_GPL(nft_register_expr); 2684 2685 /** 2686 * nft_unregister_expr - unregister nf_tables expr type 2687 * @type: expr type 2688 * 2689 * Unregisters the expr typefor use with nf_tables. 2690 */ 2691 void nft_unregister_expr(struct nft_expr_type *type) 2692 { 2693 nfnl_lock(NFNL_SUBSYS_NFTABLES); 2694 list_del_rcu(&type->list); 2695 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 2696 } 2697 EXPORT_SYMBOL_GPL(nft_unregister_expr); 2698 2699 static const struct nft_expr_type *__nft_expr_type_get(u8 family, 2700 struct nlattr *nla) 2701 { 2702 const struct nft_expr_type *type, *candidate = NULL; 2703 2704 list_for_each_entry(type, &nf_tables_expressions, list) { 2705 if (!nla_strcmp(nla, type->name)) { 2706 if (!type->family && !candidate) 2707 candidate = type; 2708 else if (type->family == family) 2709 candidate = type; 2710 } 2711 } 2712 return candidate; 2713 } 2714 2715 #ifdef CONFIG_MODULES 2716 static int nft_expr_type_request_module(struct net *net, u8 family, 2717 struct nlattr *nla) 2718 { 2719 if (nft_request_module(net, "nft-expr-%u-%.*s", family, 2720 nla_len(nla), (char *)nla_data(nla)) == -EAGAIN) 2721 return -EAGAIN; 2722 2723 return 0; 2724 } 2725 #endif 2726 2727 static const struct nft_expr_type *nft_expr_type_get(struct net *net, 2728 u8 family, 2729 struct nlattr *nla) 2730 { 2731 const struct nft_expr_type *type; 2732 2733 if (nla == NULL) 2734 return ERR_PTR(-EINVAL); 2735 2736 type = __nft_expr_type_get(family, nla); 2737 if (type != NULL && try_module_get(type->owner)) 2738 return type; 2739 2740 lockdep_nfnl_nft_mutex_not_held(); 2741 #ifdef CONFIG_MODULES 2742 if (type == NULL) { 2743 if (nft_expr_type_request_module(net, family, nla) == -EAGAIN) 2744 return ERR_PTR(-EAGAIN); 2745 2746 if (nft_request_module(net, "nft-expr-%.*s", 2747 nla_len(nla), 2748 (char *)nla_data(nla)) == -EAGAIN) 2749 return ERR_PTR(-EAGAIN); 2750 } 2751 #endif 2752 return ERR_PTR(-ENOENT); 2753 } 2754 2755 static const struct nla_policy nft_expr_policy[NFTA_EXPR_MAX + 1] = { 2756 [NFTA_EXPR_NAME] = { .type = NLA_STRING, 2757 .len = NFT_MODULE_AUTOLOAD_LIMIT }, 2758 [NFTA_EXPR_DATA] = { .type = NLA_NESTED }, 2759 }; 2760 2761 static int nf_tables_fill_expr_info(struct sk_buff *skb, 2762 const struct nft_expr *expr) 2763 { 2764 if (nla_put_string(skb, NFTA_EXPR_NAME, expr->ops->type->name)) 2765 goto nla_put_failure; 2766 2767 if (expr->ops->dump) { 2768 struct nlattr *data = nla_nest_start_noflag(skb, 2769 NFTA_EXPR_DATA); 2770 if (data == NULL) 2771 goto nla_put_failure; 2772 if (expr->ops->dump(skb, expr) < 0) 2773 goto nla_put_failure; 2774 nla_nest_end(skb, data); 2775 } 2776 2777 return skb->len; 2778 2779 nla_put_failure: 2780 return -1; 2781 }; 2782 2783 int nft_expr_dump(struct sk_buff *skb, unsigned int attr, 2784 const struct nft_expr *expr) 2785 { 2786 struct nlattr *nest; 2787 2788 nest = nla_nest_start_noflag(skb, attr); 2789 if (!nest) 2790 goto nla_put_failure; 2791 if (nf_tables_fill_expr_info(skb, expr) < 0) 2792 goto nla_put_failure; 2793 nla_nest_end(skb, nest); 2794 return 0; 2795 2796 nla_put_failure: 2797 return -1; 2798 } 2799 2800 struct nft_expr_info { 2801 const struct nft_expr_ops *ops; 2802 const struct nlattr *attr; 2803 struct nlattr *tb[NFT_EXPR_MAXATTR + 1]; 2804 }; 2805 2806 static int nf_tables_expr_parse(const struct nft_ctx *ctx, 2807 const struct nlattr *nla, 2808 struct nft_expr_info *info) 2809 { 2810 const struct nft_expr_type *type; 2811 const struct nft_expr_ops *ops; 2812 struct nlattr *tb[NFTA_EXPR_MAX + 1]; 2813 int err; 2814 2815 err = nla_parse_nested_deprecated(tb, NFTA_EXPR_MAX, nla, 2816 nft_expr_policy, NULL); 2817 if (err < 0) 2818 return err; 2819 2820 type = nft_expr_type_get(ctx->net, ctx->family, tb[NFTA_EXPR_NAME]); 2821 if (IS_ERR(type)) 2822 return PTR_ERR(type); 2823 2824 if (tb[NFTA_EXPR_DATA]) { 2825 err = nla_parse_nested_deprecated(info->tb, type->maxattr, 2826 tb[NFTA_EXPR_DATA], 2827 type->policy, NULL); 2828 if (err < 0) 2829 goto err1; 2830 } else 2831 memset(info->tb, 0, sizeof(info->tb[0]) * (type->maxattr + 1)); 2832 2833 if (type->select_ops != NULL) { 2834 ops = type->select_ops(ctx, 2835 (const struct nlattr * const *)info->tb); 2836 if (IS_ERR(ops)) { 2837 err = PTR_ERR(ops); 2838 #ifdef CONFIG_MODULES 2839 if (err == -EAGAIN) 2840 if (nft_expr_type_request_module(ctx->net, 2841 ctx->family, 2842 tb[NFTA_EXPR_NAME]) != -EAGAIN) 2843 err = -ENOENT; 2844 #endif 2845 goto err1; 2846 } 2847 } else 2848 ops = type->ops; 2849 2850 info->attr = nla; 2851 info->ops = ops; 2852 2853 return 0; 2854 2855 err1: 2856 module_put(type->owner); 2857 return err; 2858 } 2859 2860 static int nf_tables_newexpr(const struct nft_ctx *ctx, 2861 const struct nft_expr_info *expr_info, 2862 struct nft_expr *expr) 2863 { 2864 const struct nft_expr_ops *ops = expr_info->ops; 2865 int err; 2866 2867 expr->ops = ops; 2868 if (ops->init) { 2869 err = ops->init(ctx, expr, (const struct nlattr **)expr_info->tb); 2870 if (err < 0) 2871 goto err1; 2872 } 2873 2874 return 0; 2875 err1: 2876 expr->ops = NULL; 2877 return err; 2878 } 2879 2880 static void nf_tables_expr_destroy(const struct nft_ctx *ctx, 2881 struct nft_expr *expr) 2882 { 2883 const struct nft_expr_type *type = expr->ops->type; 2884 2885 if (expr->ops->destroy) 2886 expr->ops->destroy(ctx, expr); 2887 module_put(type->owner); 2888 } 2889 2890 static struct nft_expr *nft_expr_init(const struct nft_ctx *ctx, 2891 const struct nlattr *nla) 2892 { 2893 struct nft_expr_info expr_info; 2894 struct nft_expr *expr; 2895 struct module *owner; 2896 int err; 2897 2898 err = nf_tables_expr_parse(ctx, nla, &expr_info); 2899 if (err < 0) 2900 goto err_expr_parse; 2901 2902 err = -EOPNOTSUPP; 2903 if (!(expr_info.ops->type->flags & NFT_EXPR_STATEFUL)) 2904 goto err_expr_stateful; 2905 2906 err = -ENOMEM; 2907 expr = kzalloc(expr_info.ops->size, GFP_KERNEL_ACCOUNT); 2908 if (expr == NULL) 2909 goto err_expr_stateful; 2910 2911 err = nf_tables_newexpr(ctx, &expr_info, expr); 2912 if (err < 0) 2913 goto err_expr_new; 2914 2915 return expr; 2916 err_expr_new: 2917 kfree(expr); 2918 err_expr_stateful: 2919 owner = expr_info.ops->type->owner; 2920 if (expr_info.ops->type->release_ops) 2921 expr_info.ops->type->release_ops(expr_info.ops); 2922 2923 module_put(owner); 2924 err_expr_parse: 2925 return ERR_PTR(err); 2926 } 2927 2928 int nft_expr_clone(struct nft_expr *dst, struct nft_expr *src) 2929 { 2930 int err; 2931 2932 if (src->ops->clone) { 2933 dst->ops = src->ops; 2934 err = src->ops->clone(dst, src); 2935 if (err < 0) 2936 return err; 2937 } else { 2938 memcpy(dst, src, src->ops->size); 2939 } 2940 2941 __module_get(src->ops->type->owner); 2942 2943 return 0; 2944 } 2945 2946 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr) 2947 { 2948 nf_tables_expr_destroy(ctx, expr); 2949 kfree(expr); 2950 } 2951 2952 /* 2953 * Rules 2954 */ 2955 2956 static struct nft_rule *__nft_rule_lookup(const struct nft_chain *chain, 2957 u64 handle) 2958 { 2959 struct nft_rule *rule; 2960 2961 // FIXME: this sucks 2962 list_for_each_entry_rcu(rule, &chain->rules, list) { 2963 if (handle == rule->handle) 2964 return rule; 2965 } 2966 2967 return ERR_PTR(-ENOENT); 2968 } 2969 2970 static struct nft_rule *nft_rule_lookup(const struct nft_chain *chain, 2971 const struct nlattr *nla) 2972 { 2973 if (nla == NULL) 2974 return ERR_PTR(-EINVAL); 2975 2976 return __nft_rule_lookup(chain, be64_to_cpu(nla_get_be64(nla))); 2977 } 2978 2979 static const struct nla_policy nft_rule_policy[NFTA_RULE_MAX + 1] = { 2980 [NFTA_RULE_TABLE] = { .type = NLA_STRING, 2981 .len = NFT_TABLE_MAXNAMELEN - 1 }, 2982 [NFTA_RULE_CHAIN] = { .type = NLA_STRING, 2983 .len = NFT_CHAIN_MAXNAMELEN - 1 }, 2984 [NFTA_RULE_HANDLE] = { .type = NLA_U64 }, 2985 [NFTA_RULE_EXPRESSIONS] = { .type = NLA_NESTED }, 2986 [NFTA_RULE_COMPAT] = { .type = NLA_NESTED }, 2987 [NFTA_RULE_POSITION] = { .type = NLA_U64 }, 2988 [NFTA_RULE_USERDATA] = { .type = NLA_BINARY, 2989 .len = NFT_USERDATA_MAXLEN }, 2990 [NFTA_RULE_ID] = { .type = NLA_U32 }, 2991 [NFTA_RULE_POSITION_ID] = { .type = NLA_U32 }, 2992 [NFTA_RULE_CHAIN_ID] = { .type = NLA_U32 }, 2993 }; 2994 2995 static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net, 2996 u32 portid, u32 seq, int event, 2997 u32 flags, int family, 2998 const struct nft_table *table, 2999 const struct nft_chain *chain, 3000 const struct nft_rule *rule, u64 handle) 3001 { 3002 struct nlmsghdr *nlh; 3003 const struct nft_expr *expr, *next; 3004 struct nlattr *list; 3005 u16 type = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 3006 3007 nlh = nfnl_msg_put(skb, portid, seq, type, flags, family, NFNETLINK_V0, 3008 nft_base_seq(net)); 3009 if (!nlh) 3010 goto nla_put_failure; 3011 3012 if (nla_put_string(skb, NFTA_RULE_TABLE, table->name)) 3013 goto nla_put_failure; 3014 if (nla_put_string(skb, NFTA_RULE_CHAIN, chain->name)) 3015 goto nla_put_failure; 3016 if (nla_put_be64(skb, NFTA_RULE_HANDLE, cpu_to_be64(rule->handle), 3017 NFTA_RULE_PAD)) 3018 goto nla_put_failure; 3019 3020 if (event != NFT_MSG_DELRULE && handle) { 3021 if (nla_put_be64(skb, NFTA_RULE_POSITION, cpu_to_be64(handle), 3022 NFTA_RULE_PAD)) 3023 goto nla_put_failure; 3024 } 3025 3026 if (chain->flags & NFT_CHAIN_HW_OFFLOAD) 3027 nft_flow_rule_stats(chain, rule); 3028 3029 list = nla_nest_start_noflag(skb, NFTA_RULE_EXPRESSIONS); 3030 if (list == NULL) 3031 goto nla_put_failure; 3032 nft_rule_for_each_expr(expr, next, rule) { 3033 if (nft_expr_dump(skb, NFTA_LIST_ELEM, expr) < 0) 3034 goto nla_put_failure; 3035 } 3036 nla_nest_end(skb, list); 3037 3038 if (rule->udata) { 3039 struct nft_userdata *udata = nft_userdata(rule); 3040 if (nla_put(skb, NFTA_RULE_USERDATA, udata->len + 1, 3041 udata->data) < 0) 3042 goto nla_put_failure; 3043 } 3044 3045 nlmsg_end(skb, nlh); 3046 return 0; 3047 3048 nla_put_failure: 3049 nlmsg_trim(skb, nlh); 3050 return -1; 3051 } 3052 3053 static void nf_tables_rule_notify(const struct nft_ctx *ctx, 3054 const struct nft_rule *rule, int event) 3055 { 3056 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 3057 const struct nft_rule *prule; 3058 struct sk_buff *skb; 3059 u64 handle = 0; 3060 u16 flags = 0; 3061 int err; 3062 3063 if (!ctx->report && 3064 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES)) 3065 return; 3066 3067 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 3068 if (skb == NULL) 3069 goto err; 3070 3071 if (event == NFT_MSG_NEWRULE && 3072 !list_is_first(&rule->list, &ctx->chain->rules) && 3073 !list_is_last(&rule->list, &ctx->chain->rules)) { 3074 prule = list_prev_entry(rule, list); 3075 handle = prule->handle; 3076 } 3077 if (ctx->flags & (NLM_F_APPEND | NLM_F_REPLACE)) 3078 flags |= NLM_F_APPEND; 3079 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 3080 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 3081 3082 err = nf_tables_fill_rule_info(skb, ctx->net, ctx->portid, ctx->seq, 3083 event, flags, ctx->family, ctx->table, 3084 ctx->chain, rule, handle); 3085 if (err < 0) { 3086 kfree_skb(skb); 3087 goto err; 3088 } 3089 3090 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 3091 return; 3092 err: 3093 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS); 3094 } 3095 3096 struct nft_rule_dump_ctx { 3097 char *table; 3098 char *chain; 3099 }; 3100 3101 static int __nf_tables_dump_rules(struct sk_buff *skb, 3102 unsigned int *idx, 3103 struct netlink_callback *cb, 3104 const struct nft_table *table, 3105 const struct nft_chain *chain) 3106 { 3107 struct net *net = sock_net(skb->sk); 3108 const struct nft_rule *rule, *prule; 3109 unsigned int s_idx = cb->args[0]; 3110 u64 handle; 3111 3112 prule = NULL; 3113 list_for_each_entry_rcu(rule, &chain->rules, list) { 3114 if (!nft_is_active(net, rule)) 3115 goto cont_skip; 3116 if (*idx < s_idx) 3117 goto cont; 3118 if (*idx > s_idx) { 3119 memset(&cb->args[1], 0, 3120 sizeof(cb->args) - sizeof(cb->args[0])); 3121 } 3122 if (prule) 3123 handle = prule->handle; 3124 else 3125 handle = 0; 3126 3127 if (nf_tables_fill_rule_info(skb, net, NETLINK_CB(cb->skb).portid, 3128 cb->nlh->nlmsg_seq, 3129 NFT_MSG_NEWRULE, 3130 NLM_F_MULTI | NLM_F_APPEND, 3131 table->family, 3132 table, chain, rule, handle) < 0) 3133 return 1; 3134 3135 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 3136 cont: 3137 prule = rule; 3138 cont_skip: 3139 (*idx)++; 3140 } 3141 return 0; 3142 } 3143 3144 static int nf_tables_dump_rules(struct sk_buff *skb, 3145 struct netlink_callback *cb) 3146 { 3147 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 3148 const struct nft_rule_dump_ctx *ctx = cb->data; 3149 struct nft_table *table; 3150 const struct nft_chain *chain; 3151 unsigned int idx = 0; 3152 struct net *net = sock_net(skb->sk); 3153 int family = nfmsg->nfgen_family; 3154 struct nftables_pernet *nft_net; 3155 3156 rcu_read_lock(); 3157 nft_net = nft_pernet(net); 3158 cb->seq = READ_ONCE(nft_net->base_seq); 3159 3160 list_for_each_entry_rcu(table, &nft_net->tables, list) { 3161 if (family != NFPROTO_UNSPEC && family != table->family) 3162 continue; 3163 3164 if (ctx && ctx->table && strcmp(ctx->table, table->name) != 0) 3165 continue; 3166 3167 if (ctx && ctx->table && ctx->chain) { 3168 struct rhlist_head *list, *tmp; 3169 3170 list = rhltable_lookup(&table->chains_ht, ctx->chain, 3171 nft_chain_ht_params); 3172 if (!list) 3173 goto done; 3174 3175 rhl_for_each_entry_rcu(chain, tmp, list, rhlhead) { 3176 if (!nft_is_active(net, chain)) 3177 continue; 3178 __nf_tables_dump_rules(skb, &idx, 3179 cb, table, chain); 3180 break; 3181 } 3182 goto done; 3183 } 3184 3185 list_for_each_entry_rcu(chain, &table->chains, list) { 3186 if (__nf_tables_dump_rules(skb, &idx, cb, table, chain)) 3187 goto done; 3188 } 3189 3190 if (ctx && ctx->table) 3191 break; 3192 } 3193 done: 3194 rcu_read_unlock(); 3195 3196 cb->args[0] = idx; 3197 return skb->len; 3198 } 3199 3200 static int nf_tables_dump_rules_start(struct netlink_callback *cb) 3201 { 3202 const struct nlattr * const *nla = cb->data; 3203 struct nft_rule_dump_ctx *ctx = NULL; 3204 3205 if (nla[NFTA_RULE_TABLE] || nla[NFTA_RULE_CHAIN]) { 3206 ctx = kzalloc(sizeof(*ctx), GFP_ATOMIC); 3207 if (!ctx) 3208 return -ENOMEM; 3209 3210 if (nla[NFTA_RULE_TABLE]) { 3211 ctx->table = nla_strdup(nla[NFTA_RULE_TABLE], 3212 GFP_ATOMIC); 3213 if (!ctx->table) { 3214 kfree(ctx); 3215 return -ENOMEM; 3216 } 3217 } 3218 if (nla[NFTA_RULE_CHAIN]) { 3219 ctx->chain = nla_strdup(nla[NFTA_RULE_CHAIN], 3220 GFP_ATOMIC); 3221 if (!ctx->chain) { 3222 kfree(ctx->table); 3223 kfree(ctx); 3224 return -ENOMEM; 3225 } 3226 } 3227 } 3228 3229 cb->data = ctx; 3230 return 0; 3231 } 3232 3233 static int nf_tables_dump_rules_done(struct netlink_callback *cb) 3234 { 3235 struct nft_rule_dump_ctx *ctx = cb->data; 3236 3237 if (ctx) { 3238 kfree(ctx->table); 3239 kfree(ctx->chain); 3240 kfree(ctx); 3241 } 3242 return 0; 3243 } 3244 3245 /* called with rcu_read_lock held */ 3246 static int nf_tables_getrule(struct sk_buff *skb, const struct nfnl_info *info, 3247 const struct nlattr * const nla[]) 3248 { 3249 struct netlink_ext_ack *extack = info->extack; 3250 u8 genmask = nft_genmask_cur(info->net); 3251 u8 family = info->nfmsg->nfgen_family; 3252 const struct nft_chain *chain; 3253 const struct nft_rule *rule; 3254 struct net *net = info->net; 3255 struct nft_table *table; 3256 struct sk_buff *skb2; 3257 int err; 3258 3259 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 3260 struct netlink_dump_control c = { 3261 .start= nf_tables_dump_rules_start, 3262 .dump = nf_tables_dump_rules, 3263 .done = nf_tables_dump_rules_done, 3264 .module = THIS_MODULE, 3265 .data = (void *)nla, 3266 }; 3267 3268 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 3269 } 3270 3271 table = nft_table_lookup(net, nla[NFTA_RULE_TABLE], family, genmask, 0); 3272 if (IS_ERR(table)) { 3273 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_TABLE]); 3274 return PTR_ERR(table); 3275 } 3276 3277 chain = nft_chain_lookup(net, table, nla[NFTA_RULE_CHAIN], genmask); 3278 if (IS_ERR(chain)) { 3279 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_CHAIN]); 3280 return PTR_ERR(chain); 3281 } 3282 3283 rule = nft_rule_lookup(chain, nla[NFTA_RULE_HANDLE]); 3284 if (IS_ERR(rule)) { 3285 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_HANDLE]); 3286 return PTR_ERR(rule); 3287 } 3288 3289 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 3290 if (!skb2) 3291 return -ENOMEM; 3292 3293 err = nf_tables_fill_rule_info(skb2, net, NETLINK_CB(skb).portid, 3294 info->nlh->nlmsg_seq, NFT_MSG_NEWRULE, 0, 3295 family, table, chain, rule, 0); 3296 if (err < 0) 3297 goto err_fill_rule_info; 3298 3299 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 3300 3301 err_fill_rule_info: 3302 kfree_skb(skb2); 3303 return err; 3304 } 3305 3306 static void nf_tables_rule_destroy(const struct nft_ctx *ctx, 3307 struct nft_rule *rule) 3308 { 3309 struct nft_expr *expr, *next; 3310 3311 /* 3312 * Careful: some expressions might not be initialized in case this 3313 * is called on error from nf_tables_newrule(). 3314 */ 3315 expr = nft_expr_first(rule); 3316 while (nft_expr_more(rule, expr)) { 3317 next = nft_expr_next(expr); 3318 nf_tables_expr_destroy(ctx, expr); 3319 expr = next; 3320 } 3321 kfree(rule); 3322 } 3323 3324 void nf_tables_rule_release(const struct nft_ctx *ctx, struct nft_rule *rule) 3325 { 3326 nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_RELEASE); 3327 nf_tables_rule_destroy(ctx, rule); 3328 } 3329 3330 int nft_chain_validate(const struct nft_ctx *ctx, const struct nft_chain *chain) 3331 { 3332 struct nft_expr *expr, *last; 3333 const struct nft_data *data; 3334 struct nft_rule *rule; 3335 int err; 3336 3337 if (ctx->level == NFT_JUMP_STACK_SIZE) 3338 return -EMLINK; 3339 3340 list_for_each_entry(rule, &chain->rules, list) { 3341 if (!nft_is_active_next(ctx->net, rule)) 3342 continue; 3343 3344 nft_rule_for_each_expr(expr, last, rule) { 3345 if (!expr->ops->validate) 3346 continue; 3347 3348 err = expr->ops->validate(ctx, expr, &data); 3349 if (err < 0) 3350 return err; 3351 } 3352 3353 cond_resched(); 3354 } 3355 3356 return 0; 3357 } 3358 EXPORT_SYMBOL_GPL(nft_chain_validate); 3359 3360 static int nft_table_validate(struct net *net, const struct nft_table *table) 3361 { 3362 struct nft_chain *chain; 3363 struct nft_ctx ctx = { 3364 .net = net, 3365 .family = table->family, 3366 }; 3367 int err; 3368 3369 list_for_each_entry(chain, &table->chains, list) { 3370 if (!nft_is_base_chain(chain)) 3371 continue; 3372 3373 ctx.chain = chain; 3374 err = nft_chain_validate(&ctx, chain); 3375 if (err < 0) 3376 return err; 3377 } 3378 3379 return 0; 3380 } 3381 3382 static struct nft_rule *nft_rule_lookup_byid(const struct net *net, 3383 const struct nft_chain *chain, 3384 const struct nlattr *nla); 3385 3386 #define NFT_RULE_MAXEXPRS 128 3387 3388 static int nf_tables_newrule(struct sk_buff *skb, const struct nfnl_info *info, 3389 const struct nlattr * const nla[]) 3390 { 3391 struct nftables_pernet *nft_net = nft_pernet(info->net); 3392 struct netlink_ext_ack *extack = info->extack; 3393 unsigned int size, i, n, ulen = 0, usize = 0; 3394 u8 genmask = nft_genmask_next(info->net); 3395 struct nft_rule *rule, *old_rule = NULL; 3396 struct nft_expr_info *expr_info = NULL; 3397 u8 family = info->nfmsg->nfgen_family; 3398 struct nft_flow_rule *flow = NULL; 3399 struct net *net = info->net; 3400 struct nft_userdata *udata; 3401 struct nft_table *table; 3402 struct nft_chain *chain; 3403 struct nft_trans *trans; 3404 u64 handle, pos_handle; 3405 struct nft_expr *expr; 3406 struct nft_ctx ctx; 3407 struct nlattr *tmp; 3408 int err, rem; 3409 3410 lockdep_assert_held(&nft_net->commit_mutex); 3411 3412 table = nft_table_lookup(net, nla[NFTA_RULE_TABLE], family, genmask, 3413 NETLINK_CB(skb).portid); 3414 if (IS_ERR(table)) { 3415 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_TABLE]); 3416 return PTR_ERR(table); 3417 } 3418 3419 if (nla[NFTA_RULE_CHAIN]) { 3420 chain = nft_chain_lookup(net, table, nla[NFTA_RULE_CHAIN], 3421 genmask); 3422 if (IS_ERR(chain)) { 3423 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_CHAIN]); 3424 return PTR_ERR(chain); 3425 } 3426 if (nft_chain_is_bound(chain)) 3427 return -EOPNOTSUPP; 3428 3429 } else if (nla[NFTA_RULE_CHAIN_ID]) { 3430 chain = nft_chain_lookup_byid(net, table, nla[NFTA_RULE_CHAIN_ID]); 3431 if (IS_ERR(chain)) { 3432 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_CHAIN_ID]); 3433 return PTR_ERR(chain); 3434 } 3435 } else { 3436 return -EINVAL; 3437 } 3438 3439 if (nla[NFTA_RULE_HANDLE]) { 3440 handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_HANDLE])); 3441 rule = __nft_rule_lookup(chain, handle); 3442 if (IS_ERR(rule)) { 3443 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_HANDLE]); 3444 return PTR_ERR(rule); 3445 } 3446 3447 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 3448 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_HANDLE]); 3449 return -EEXIST; 3450 } 3451 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 3452 old_rule = rule; 3453 else 3454 return -EOPNOTSUPP; 3455 } else { 3456 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE) || 3457 info->nlh->nlmsg_flags & NLM_F_REPLACE) 3458 return -EINVAL; 3459 handle = nf_tables_alloc_handle(table); 3460 3461 if (chain->use == UINT_MAX) 3462 return -EOVERFLOW; 3463 3464 if (nla[NFTA_RULE_POSITION]) { 3465 pos_handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_POSITION])); 3466 old_rule = __nft_rule_lookup(chain, pos_handle); 3467 if (IS_ERR(old_rule)) { 3468 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_POSITION]); 3469 return PTR_ERR(old_rule); 3470 } 3471 } else if (nla[NFTA_RULE_POSITION_ID]) { 3472 old_rule = nft_rule_lookup_byid(net, chain, nla[NFTA_RULE_POSITION_ID]); 3473 if (IS_ERR(old_rule)) { 3474 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_POSITION_ID]); 3475 return PTR_ERR(old_rule); 3476 } 3477 } 3478 } 3479 3480 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, chain, nla); 3481 3482 n = 0; 3483 size = 0; 3484 if (nla[NFTA_RULE_EXPRESSIONS]) { 3485 expr_info = kvmalloc_array(NFT_RULE_MAXEXPRS, 3486 sizeof(struct nft_expr_info), 3487 GFP_KERNEL); 3488 if (!expr_info) 3489 return -ENOMEM; 3490 3491 nla_for_each_nested(tmp, nla[NFTA_RULE_EXPRESSIONS], rem) { 3492 err = -EINVAL; 3493 if (nla_type(tmp) != NFTA_LIST_ELEM) 3494 goto err_release_expr; 3495 if (n == NFT_RULE_MAXEXPRS) 3496 goto err_release_expr; 3497 err = nf_tables_expr_parse(&ctx, tmp, &expr_info[n]); 3498 if (err < 0) { 3499 NL_SET_BAD_ATTR(extack, tmp); 3500 goto err_release_expr; 3501 } 3502 size += expr_info[n].ops->size; 3503 n++; 3504 } 3505 } 3506 /* Check for overflow of dlen field */ 3507 err = -EFBIG; 3508 if (size >= 1 << 12) 3509 goto err_release_expr; 3510 3511 if (nla[NFTA_RULE_USERDATA]) { 3512 ulen = nla_len(nla[NFTA_RULE_USERDATA]); 3513 if (ulen > 0) 3514 usize = sizeof(struct nft_userdata) + ulen; 3515 } 3516 3517 err = -ENOMEM; 3518 rule = kzalloc(sizeof(*rule) + size + usize, GFP_KERNEL_ACCOUNT); 3519 if (rule == NULL) 3520 goto err_release_expr; 3521 3522 nft_activate_next(net, rule); 3523 3524 rule->handle = handle; 3525 rule->dlen = size; 3526 rule->udata = ulen ? 1 : 0; 3527 3528 if (ulen) { 3529 udata = nft_userdata(rule); 3530 udata->len = ulen - 1; 3531 nla_memcpy(udata->data, nla[NFTA_RULE_USERDATA], ulen); 3532 } 3533 3534 expr = nft_expr_first(rule); 3535 for (i = 0; i < n; i++) { 3536 err = nf_tables_newexpr(&ctx, &expr_info[i], expr); 3537 if (err < 0) { 3538 NL_SET_BAD_ATTR(extack, expr_info[i].attr); 3539 goto err_release_rule; 3540 } 3541 3542 if (expr_info[i].ops->validate) 3543 nft_validate_state_update(net, NFT_VALIDATE_NEED); 3544 3545 expr_info[i].ops = NULL; 3546 expr = nft_expr_next(expr); 3547 } 3548 3549 if (chain->flags & NFT_CHAIN_HW_OFFLOAD) { 3550 flow = nft_flow_rule_create(net, rule); 3551 if (IS_ERR(flow)) { 3552 err = PTR_ERR(flow); 3553 goto err_release_rule; 3554 } 3555 } 3556 3557 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) { 3558 err = nft_delrule(&ctx, old_rule); 3559 if (err < 0) 3560 goto err_destroy_flow_rule; 3561 3562 trans = nft_trans_rule_add(&ctx, NFT_MSG_NEWRULE, rule); 3563 if (trans == NULL) { 3564 err = -ENOMEM; 3565 goto err_destroy_flow_rule; 3566 } 3567 list_add_tail_rcu(&rule->list, &old_rule->list); 3568 } else { 3569 trans = nft_trans_rule_add(&ctx, NFT_MSG_NEWRULE, rule); 3570 if (!trans) { 3571 err = -ENOMEM; 3572 goto err_destroy_flow_rule; 3573 } 3574 3575 if (info->nlh->nlmsg_flags & NLM_F_APPEND) { 3576 if (old_rule) 3577 list_add_rcu(&rule->list, &old_rule->list); 3578 else 3579 list_add_tail_rcu(&rule->list, &chain->rules); 3580 } else { 3581 if (old_rule) 3582 list_add_tail_rcu(&rule->list, &old_rule->list); 3583 else 3584 list_add_rcu(&rule->list, &chain->rules); 3585 } 3586 } 3587 kvfree(expr_info); 3588 chain->use++; 3589 3590 if (flow) 3591 nft_trans_flow_rule(trans) = flow; 3592 3593 if (nft_net->validate_state == NFT_VALIDATE_DO) 3594 return nft_table_validate(net, table); 3595 3596 return 0; 3597 3598 err_destroy_flow_rule: 3599 if (flow) 3600 nft_flow_rule_destroy(flow); 3601 err_release_rule: 3602 nf_tables_rule_release(&ctx, rule); 3603 err_release_expr: 3604 for (i = 0; i < n; i++) { 3605 if (expr_info[i].ops) { 3606 module_put(expr_info[i].ops->type->owner); 3607 if (expr_info[i].ops->type->release_ops) 3608 expr_info[i].ops->type->release_ops(expr_info[i].ops); 3609 } 3610 } 3611 kvfree(expr_info); 3612 3613 return err; 3614 } 3615 3616 static struct nft_rule *nft_rule_lookup_byid(const struct net *net, 3617 const struct nft_chain *chain, 3618 const struct nlattr *nla) 3619 { 3620 struct nftables_pernet *nft_net = nft_pernet(net); 3621 u32 id = ntohl(nla_get_be32(nla)); 3622 struct nft_trans *trans; 3623 3624 list_for_each_entry(trans, &nft_net->commit_list, list) { 3625 struct nft_rule *rule = nft_trans_rule(trans); 3626 3627 if (trans->msg_type == NFT_MSG_NEWRULE && 3628 trans->ctx.chain == chain && 3629 id == nft_trans_rule_id(trans)) 3630 return rule; 3631 } 3632 return ERR_PTR(-ENOENT); 3633 } 3634 3635 static int nf_tables_delrule(struct sk_buff *skb, const struct nfnl_info *info, 3636 const struct nlattr * const nla[]) 3637 { 3638 struct netlink_ext_ack *extack = info->extack; 3639 u8 genmask = nft_genmask_next(info->net); 3640 u8 family = info->nfmsg->nfgen_family; 3641 struct nft_chain *chain = NULL; 3642 struct net *net = info->net; 3643 struct nft_table *table; 3644 struct nft_rule *rule; 3645 struct nft_ctx ctx; 3646 int err = 0; 3647 3648 table = nft_table_lookup(net, nla[NFTA_RULE_TABLE], family, genmask, 3649 NETLINK_CB(skb).portid); 3650 if (IS_ERR(table)) { 3651 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_TABLE]); 3652 return PTR_ERR(table); 3653 } 3654 3655 if (nla[NFTA_RULE_CHAIN]) { 3656 chain = nft_chain_lookup(net, table, nla[NFTA_RULE_CHAIN], 3657 genmask); 3658 if (IS_ERR(chain)) { 3659 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_CHAIN]); 3660 return PTR_ERR(chain); 3661 } 3662 if (nft_chain_is_bound(chain)) 3663 return -EOPNOTSUPP; 3664 } 3665 3666 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, chain, nla); 3667 3668 if (chain) { 3669 if (nla[NFTA_RULE_HANDLE]) { 3670 rule = nft_rule_lookup(chain, nla[NFTA_RULE_HANDLE]); 3671 if (IS_ERR(rule)) { 3672 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_HANDLE]); 3673 return PTR_ERR(rule); 3674 } 3675 3676 err = nft_delrule(&ctx, rule); 3677 } else if (nla[NFTA_RULE_ID]) { 3678 rule = nft_rule_lookup_byid(net, chain, nla[NFTA_RULE_ID]); 3679 if (IS_ERR(rule)) { 3680 NL_SET_BAD_ATTR(extack, nla[NFTA_RULE_ID]); 3681 return PTR_ERR(rule); 3682 } 3683 3684 err = nft_delrule(&ctx, rule); 3685 } else { 3686 err = nft_delrule_by_chain(&ctx); 3687 } 3688 } else { 3689 list_for_each_entry(chain, &table->chains, list) { 3690 if (!nft_is_active_next(net, chain)) 3691 continue; 3692 3693 ctx.chain = chain; 3694 err = nft_delrule_by_chain(&ctx); 3695 if (err < 0) 3696 break; 3697 } 3698 } 3699 3700 return err; 3701 } 3702 3703 /* 3704 * Sets 3705 */ 3706 static const struct nft_set_type *nft_set_types[] = { 3707 &nft_set_hash_fast_type, 3708 &nft_set_hash_type, 3709 &nft_set_rhash_type, 3710 &nft_set_bitmap_type, 3711 &nft_set_rbtree_type, 3712 #if defined(CONFIG_X86_64) && !defined(CONFIG_UML) 3713 &nft_set_pipapo_avx2_type, 3714 #endif 3715 &nft_set_pipapo_type, 3716 }; 3717 3718 #define NFT_SET_FEATURES (NFT_SET_INTERVAL | NFT_SET_MAP | \ 3719 NFT_SET_TIMEOUT | NFT_SET_OBJECT | \ 3720 NFT_SET_EVAL) 3721 3722 static bool nft_set_ops_candidate(const struct nft_set_type *type, u32 flags) 3723 { 3724 return (flags & type->features) == (flags & NFT_SET_FEATURES); 3725 } 3726 3727 /* 3728 * Select a set implementation based on the data characteristics and the 3729 * given policy. The total memory use might not be known if no size is 3730 * given, in that case the amount of memory per element is used. 3731 */ 3732 static const struct nft_set_ops * 3733 nft_select_set_ops(const struct nft_ctx *ctx, 3734 const struct nlattr * const nla[], 3735 const struct nft_set_desc *desc, 3736 enum nft_set_policies policy) 3737 { 3738 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 3739 const struct nft_set_ops *ops, *bops; 3740 struct nft_set_estimate est, best; 3741 const struct nft_set_type *type; 3742 u32 flags = 0; 3743 int i; 3744 3745 lockdep_assert_held(&nft_net->commit_mutex); 3746 lockdep_nfnl_nft_mutex_not_held(); 3747 3748 if (nla[NFTA_SET_FLAGS] != NULL) 3749 flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS])); 3750 3751 bops = NULL; 3752 best.size = ~0; 3753 best.lookup = ~0; 3754 best.space = ~0; 3755 3756 for (i = 0; i < ARRAY_SIZE(nft_set_types); i++) { 3757 type = nft_set_types[i]; 3758 ops = &type->ops; 3759 3760 if (!nft_set_ops_candidate(type, flags)) 3761 continue; 3762 if (!ops->estimate(desc, flags, &est)) 3763 continue; 3764 3765 switch (policy) { 3766 case NFT_SET_POL_PERFORMANCE: 3767 if (est.lookup < best.lookup) 3768 break; 3769 if (est.lookup == best.lookup && 3770 est.space < best.space) 3771 break; 3772 continue; 3773 case NFT_SET_POL_MEMORY: 3774 if (!desc->size) { 3775 if (est.space < best.space) 3776 break; 3777 if (est.space == best.space && 3778 est.lookup < best.lookup) 3779 break; 3780 } else if (est.size < best.size || !bops) { 3781 break; 3782 } 3783 continue; 3784 default: 3785 break; 3786 } 3787 3788 bops = ops; 3789 best = est; 3790 } 3791 3792 if (bops != NULL) 3793 return bops; 3794 3795 return ERR_PTR(-EOPNOTSUPP); 3796 } 3797 3798 static const struct nla_policy nft_set_policy[NFTA_SET_MAX + 1] = { 3799 [NFTA_SET_TABLE] = { .type = NLA_STRING, 3800 .len = NFT_TABLE_MAXNAMELEN - 1 }, 3801 [NFTA_SET_NAME] = { .type = NLA_STRING, 3802 .len = NFT_SET_MAXNAMELEN - 1 }, 3803 [NFTA_SET_FLAGS] = { .type = NLA_U32 }, 3804 [NFTA_SET_KEY_TYPE] = { .type = NLA_U32 }, 3805 [NFTA_SET_KEY_LEN] = { .type = NLA_U32 }, 3806 [NFTA_SET_DATA_TYPE] = { .type = NLA_U32 }, 3807 [NFTA_SET_DATA_LEN] = { .type = NLA_U32 }, 3808 [NFTA_SET_POLICY] = { .type = NLA_U32 }, 3809 [NFTA_SET_DESC] = { .type = NLA_NESTED }, 3810 [NFTA_SET_ID] = { .type = NLA_U32 }, 3811 [NFTA_SET_TIMEOUT] = { .type = NLA_U64 }, 3812 [NFTA_SET_GC_INTERVAL] = { .type = NLA_U32 }, 3813 [NFTA_SET_USERDATA] = { .type = NLA_BINARY, 3814 .len = NFT_USERDATA_MAXLEN }, 3815 [NFTA_SET_OBJ_TYPE] = { .type = NLA_U32 }, 3816 [NFTA_SET_HANDLE] = { .type = NLA_U64 }, 3817 [NFTA_SET_EXPR] = { .type = NLA_NESTED }, 3818 [NFTA_SET_EXPRESSIONS] = { .type = NLA_NESTED }, 3819 }; 3820 3821 static const struct nla_policy nft_set_desc_policy[NFTA_SET_DESC_MAX + 1] = { 3822 [NFTA_SET_DESC_SIZE] = { .type = NLA_U32 }, 3823 [NFTA_SET_DESC_CONCAT] = { .type = NLA_NESTED }, 3824 }; 3825 3826 static struct nft_set *nft_set_lookup(const struct nft_table *table, 3827 const struct nlattr *nla, u8 genmask) 3828 { 3829 struct nft_set *set; 3830 3831 if (nla == NULL) 3832 return ERR_PTR(-EINVAL); 3833 3834 list_for_each_entry_rcu(set, &table->sets, list) { 3835 if (!nla_strcmp(nla, set->name) && 3836 nft_active_genmask(set, genmask)) 3837 return set; 3838 } 3839 return ERR_PTR(-ENOENT); 3840 } 3841 3842 static struct nft_set *nft_set_lookup_byhandle(const struct nft_table *table, 3843 const struct nlattr *nla, 3844 u8 genmask) 3845 { 3846 struct nft_set *set; 3847 3848 list_for_each_entry(set, &table->sets, list) { 3849 if (be64_to_cpu(nla_get_be64(nla)) == set->handle && 3850 nft_active_genmask(set, genmask)) 3851 return set; 3852 } 3853 return ERR_PTR(-ENOENT); 3854 } 3855 3856 static struct nft_set *nft_set_lookup_byid(const struct net *net, 3857 const struct nft_table *table, 3858 const struct nlattr *nla, u8 genmask) 3859 { 3860 struct nftables_pernet *nft_net = nft_pernet(net); 3861 u32 id = ntohl(nla_get_be32(nla)); 3862 struct nft_trans *trans; 3863 3864 list_for_each_entry(trans, &nft_net->commit_list, list) { 3865 if (trans->msg_type == NFT_MSG_NEWSET) { 3866 struct nft_set *set = nft_trans_set(trans); 3867 3868 if (id == nft_trans_set_id(trans) && 3869 set->table == table && 3870 nft_active_genmask(set, genmask)) 3871 return set; 3872 } 3873 } 3874 return ERR_PTR(-ENOENT); 3875 } 3876 3877 struct nft_set *nft_set_lookup_global(const struct net *net, 3878 const struct nft_table *table, 3879 const struct nlattr *nla_set_name, 3880 const struct nlattr *nla_set_id, 3881 u8 genmask) 3882 { 3883 struct nft_set *set; 3884 3885 set = nft_set_lookup(table, nla_set_name, genmask); 3886 if (IS_ERR(set)) { 3887 if (!nla_set_id) 3888 return set; 3889 3890 set = nft_set_lookup_byid(net, table, nla_set_id, genmask); 3891 } 3892 return set; 3893 } 3894 EXPORT_SYMBOL_GPL(nft_set_lookup_global); 3895 3896 static int nf_tables_set_alloc_name(struct nft_ctx *ctx, struct nft_set *set, 3897 const char *name) 3898 { 3899 const struct nft_set *i; 3900 const char *p; 3901 unsigned long *inuse; 3902 unsigned int n = 0, min = 0; 3903 3904 p = strchr(name, '%'); 3905 if (p != NULL) { 3906 if (p[1] != 'd' || strchr(p + 2, '%')) 3907 return -EINVAL; 3908 3909 inuse = (unsigned long *)get_zeroed_page(GFP_KERNEL); 3910 if (inuse == NULL) 3911 return -ENOMEM; 3912 cont: 3913 list_for_each_entry(i, &ctx->table->sets, list) { 3914 int tmp; 3915 3916 if (!nft_is_active_next(ctx->net, i)) 3917 continue; 3918 if (!sscanf(i->name, name, &tmp)) 3919 continue; 3920 if (tmp < min || tmp >= min + BITS_PER_BYTE * PAGE_SIZE) 3921 continue; 3922 3923 set_bit(tmp - min, inuse); 3924 } 3925 3926 n = find_first_zero_bit(inuse, BITS_PER_BYTE * PAGE_SIZE); 3927 if (n >= BITS_PER_BYTE * PAGE_SIZE) { 3928 min += BITS_PER_BYTE * PAGE_SIZE; 3929 memset(inuse, 0, PAGE_SIZE); 3930 goto cont; 3931 } 3932 free_page((unsigned long)inuse); 3933 } 3934 3935 set->name = kasprintf(GFP_KERNEL_ACCOUNT, name, min + n); 3936 if (!set->name) 3937 return -ENOMEM; 3938 3939 list_for_each_entry(i, &ctx->table->sets, list) { 3940 if (!nft_is_active_next(ctx->net, i)) 3941 continue; 3942 if (!strcmp(set->name, i->name)) { 3943 kfree(set->name); 3944 set->name = NULL; 3945 return -ENFILE; 3946 } 3947 } 3948 return 0; 3949 } 3950 3951 int nf_msecs_to_jiffies64(const struct nlattr *nla, u64 *result) 3952 { 3953 u64 ms = be64_to_cpu(nla_get_be64(nla)); 3954 u64 max = (u64)(~((u64)0)); 3955 3956 max = div_u64(max, NSEC_PER_MSEC); 3957 if (ms >= max) 3958 return -ERANGE; 3959 3960 ms *= NSEC_PER_MSEC; 3961 *result = nsecs_to_jiffies64(ms); 3962 return 0; 3963 } 3964 3965 __be64 nf_jiffies64_to_msecs(u64 input) 3966 { 3967 return cpu_to_be64(jiffies64_to_msecs(input)); 3968 } 3969 3970 static int nf_tables_fill_set_concat(struct sk_buff *skb, 3971 const struct nft_set *set) 3972 { 3973 struct nlattr *concat, *field; 3974 int i; 3975 3976 concat = nla_nest_start_noflag(skb, NFTA_SET_DESC_CONCAT); 3977 if (!concat) 3978 return -ENOMEM; 3979 3980 for (i = 0; i < set->field_count; i++) { 3981 field = nla_nest_start_noflag(skb, NFTA_LIST_ELEM); 3982 if (!field) 3983 return -ENOMEM; 3984 3985 if (nla_put_be32(skb, NFTA_SET_FIELD_LEN, 3986 htonl(set->field_len[i]))) 3987 return -ENOMEM; 3988 3989 nla_nest_end(skb, field); 3990 } 3991 3992 nla_nest_end(skb, concat); 3993 3994 return 0; 3995 } 3996 3997 static int nf_tables_fill_set(struct sk_buff *skb, const struct nft_ctx *ctx, 3998 const struct nft_set *set, u16 event, u16 flags) 3999 { 4000 struct nlmsghdr *nlh; 4001 u32 portid = ctx->portid; 4002 struct nlattr *nest; 4003 u32 seq = ctx->seq; 4004 int i; 4005 4006 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 4007 nlh = nfnl_msg_put(skb, portid, seq, event, flags, ctx->family, 4008 NFNETLINK_V0, nft_base_seq(ctx->net)); 4009 if (!nlh) 4010 goto nla_put_failure; 4011 4012 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name)) 4013 goto nla_put_failure; 4014 if (nla_put_string(skb, NFTA_SET_NAME, set->name)) 4015 goto nla_put_failure; 4016 if (nla_put_be64(skb, NFTA_SET_HANDLE, cpu_to_be64(set->handle), 4017 NFTA_SET_PAD)) 4018 goto nla_put_failure; 4019 if (set->flags != 0) 4020 if (nla_put_be32(skb, NFTA_SET_FLAGS, htonl(set->flags))) 4021 goto nla_put_failure; 4022 4023 if (nla_put_be32(skb, NFTA_SET_KEY_TYPE, htonl(set->ktype))) 4024 goto nla_put_failure; 4025 if (nla_put_be32(skb, NFTA_SET_KEY_LEN, htonl(set->klen))) 4026 goto nla_put_failure; 4027 if (set->flags & NFT_SET_MAP) { 4028 if (nla_put_be32(skb, NFTA_SET_DATA_TYPE, htonl(set->dtype))) 4029 goto nla_put_failure; 4030 if (nla_put_be32(skb, NFTA_SET_DATA_LEN, htonl(set->dlen))) 4031 goto nla_put_failure; 4032 } 4033 if (set->flags & NFT_SET_OBJECT && 4034 nla_put_be32(skb, NFTA_SET_OBJ_TYPE, htonl(set->objtype))) 4035 goto nla_put_failure; 4036 4037 if (set->timeout && 4038 nla_put_be64(skb, NFTA_SET_TIMEOUT, 4039 nf_jiffies64_to_msecs(set->timeout), 4040 NFTA_SET_PAD)) 4041 goto nla_put_failure; 4042 if (set->gc_int && 4043 nla_put_be32(skb, NFTA_SET_GC_INTERVAL, htonl(set->gc_int))) 4044 goto nla_put_failure; 4045 4046 if (set->policy != NFT_SET_POL_PERFORMANCE) { 4047 if (nla_put_be32(skb, NFTA_SET_POLICY, htonl(set->policy))) 4048 goto nla_put_failure; 4049 } 4050 4051 if (set->udata && 4052 nla_put(skb, NFTA_SET_USERDATA, set->udlen, set->udata)) 4053 goto nla_put_failure; 4054 4055 nest = nla_nest_start_noflag(skb, NFTA_SET_DESC); 4056 if (!nest) 4057 goto nla_put_failure; 4058 if (set->size && 4059 nla_put_be32(skb, NFTA_SET_DESC_SIZE, htonl(set->size))) 4060 goto nla_put_failure; 4061 4062 if (set->field_count > 1 && 4063 nf_tables_fill_set_concat(skb, set)) 4064 goto nla_put_failure; 4065 4066 nla_nest_end(skb, nest); 4067 4068 if (set->num_exprs == 1) { 4069 nest = nla_nest_start_noflag(skb, NFTA_SET_EXPR); 4070 if (nf_tables_fill_expr_info(skb, set->exprs[0]) < 0) 4071 goto nla_put_failure; 4072 4073 nla_nest_end(skb, nest); 4074 } else if (set->num_exprs > 1) { 4075 nest = nla_nest_start_noflag(skb, NFTA_SET_EXPRESSIONS); 4076 if (nest == NULL) 4077 goto nla_put_failure; 4078 4079 for (i = 0; i < set->num_exprs; i++) { 4080 if (nft_expr_dump(skb, NFTA_LIST_ELEM, 4081 set->exprs[i]) < 0) 4082 goto nla_put_failure; 4083 } 4084 nla_nest_end(skb, nest); 4085 } 4086 4087 nlmsg_end(skb, nlh); 4088 return 0; 4089 4090 nla_put_failure: 4091 nlmsg_trim(skb, nlh); 4092 return -1; 4093 } 4094 4095 static void nf_tables_set_notify(const struct nft_ctx *ctx, 4096 const struct nft_set *set, int event, 4097 gfp_t gfp_flags) 4098 { 4099 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 4100 u32 portid = ctx->portid; 4101 struct sk_buff *skb; 4102 u16 flags = 0; 4103 int err; 4104 4105 if (!ctx->report && 4106 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES)) 4107 return; 4108 4109 skb = nlmsg_new(NLMSG_GOODSIZE, gfp_flags); 4110 if (skb == NULL) 4111 goto err; 4112 4113 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 4114 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 4115 4116 err = nf_tables_fill_set(skb, ctx, set, event, flags); 4117 if (err < 0) { 4118 kfree_skb(skb); 4119 goto err; 4120 } 4121 4122 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 4123 return; 4124 err: 4125 nfnetlink_set_err(ctx->net, portid, NFNLGRP_NFTABLES, -ENOBUFS); 4126 } 4127 4128 static int nf_tables_dump_sets(struct sk_buff *skb, struct netlink_callback *cb) 4129 { 4130 const struct nft_set *set; 4131 unsigned int idx, s_idx = cb->args[0]; 4132 struct nft_table *table, *cur_table = (struct nft_table *)cb->args[2]; 4133 struct net *net = sock_net(skb->sk); 4134 struct nft_ctx *ctx = cb->data, ctx_set; 4135 struct nftables_pernet *nft_net; 4136 4137 if (cb->args[1]) 4138 return skb->len; 4139 4140 rcu_read_lock(); 4141 nft_net = nft_pernet(net); 4142 cb->seq = READ_ONCE(nft_net->base_seq); 4143 4144 list_for_each_entry_rcu(table, &nft_net->tables, list) { 4145 if (ctx->family != NFPROTO_UNSPEC && 4146 ctx->family != table->family) 4147 continue; 4148 4149 if (ctx->table && ctx->table != table) 4150 continue; 4151 4152 if (cur_table) { 4153 if (cur_table != table) 4154 continue; 4155 4156 cur_table = NULL; 4157 } 4158 idx = 0; 4159 list_for_each_entry_rcu(set, &table->sets, list) { 4160 if (idx < s_idx) 4161 goto cont; 4162 if (!nft_is_active(net, set)) 4163 goto cont; 4164 4165 ctx_set = *ctx; 4166 ctx_set.table = table; 4167 ctx_set.family = table->family; 4168 4169 if (nf_tables_fill_set(skb, &ctx_set, set, 4170 NFT_MSG_NEWSET, 4171 NLM_F_MULTI) < 0) { 4172 cb->args[0] = idx; 4173 cb->args[2] = (unsigned long) table; 4174 goto done; 4175 } 4176 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4177 cont: 4178 idx++; 4179 } 4180 if (s_idx) 4181 s_idx = 0; 4182 } 4183 cb->args[1] = 1; 4184 done: 4185 rcu_read_unlock(); 4186 return skb->len; 4187 } 4188 4189 static int nf_tables_dump_sets_start(struct netlink_callback *cb) 4190 { 4191 struct nft_ctx *ctx_dump = NULL; 4192 4193 ctx_dump = kmemdup(cb->data, sizeof(*ctx_dump), GFP_ATOMIC); 4194 if (ctx_dump == NULL) 4195 return -ENOMEM; 4196 4197 cb->data = ctx_dump; 4198 return 0; 4199 } 4200 4201 static int nf_tables_dump_sets_done(struct netlink_callback *cb) 4202 { 4203 kfree(cb->data); 4204 return 0; 4205 } 4206 4207 /* called with rcu_read_lock held */ 4208 static int nf_tables_getset(struct sk_buff *skb, const struct nfnl_info *info, 4209 const struct nlattr * const nla[]) 4210 { 4211 struct netlink_ext_ack *extack = info->extack; 4212 u8 genmask = nft_genmask_cur(info->net); 4213 u8 family = info->nfmsg->nfgen_family; 4214 struct nft_table *table = NULL; 4215 struct net *net = info->net; 4216 const struct nft_set *set; 4217 struct sk_buff *skb2; 4218 struct nft_ctx ctx; 4219 int err; 4220 4221 if (nla[NFTA_SET_TABLE]) { 4222 table = nft_table_lookup(net, nla[NFTA_SET_TABLE], family, 4223 genmask, 0); 4224 if (IS_ERR(table)) { 4225 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_TABLE]); 4226 return PTR_ERR(table); 4227 } 4228 } 4229 4230 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 4231 4232 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 4233 struct netlink_dump_control c = { 4234 .start = nf_tables_dump_sets_start, 4235 .dump = nf_tables_dump_sets, 4236 .done = nf_tables_dump_sets_done, 4237 .data = &ctx, 4238 .module = THIS_MODULE, 4239 }; 4240 4241 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 4242 } 4243 4244 /* Only accept unspec with dump */ 4245 if (info->nfmsg->nfgen_family == NFPROTO_UNSPEC) 4246 return -EAFNOSUPPORT; 4247 if (!nla[NFTA_SET_TABLE]) 4248 return -EINVAL; 4249 4250 set = nft_set_lookup(table, nla[NFTA_SET_NAME], genmask); 4251 if (IS_ERR(set)) 4252 return PTR_ERR(set); 4253 4254 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 4255 if (skb2 == NULL) 4256 return -ENOMEM; 4257 4258 err = nf_tables_fill_set(skb2, &ctx, set, NFT_MSG_NEWSET, 0); 4259 if (err < 0) 4260 goto err_fill_set_info; 4261 4262 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 4263 4264 err_fill_set_info: 4265 kfree_skb(skb2); 4266 return err; 4267 } 4268 4269 static const struct nla_policy nft_concat_policy[NFTA_SET_FIELD_MAX + 1] = { 4270 [NFTA_SET_FIELD_LEN] = { .type = NLA_U32 }, 4271 }; 4272 4273 static int nft_set_desc_concat_parse(const struct nlattr *attr, 4274 struct nft_set_desc *desc) 4275 { 4276 struct nlattr *tb[NFTA_SET_FIELD_MAX + 1]; 4277 u32 len; 4278 int err; 4279 4280 if (desc->field_count >= ARRAY_SIZE(desc->field_len)) 4281 return -E2BIG; 4282 4283 err = nla_parse_nested_deprecated(tb, NFTA_SET_FIELD_MAX, attr, 4284 nft_concat_policy, NULL); 4285 if (err < 0) 4286 return err; 4287 4288 if (!tb[NFTA_SET_FIELD_LEN]) 4289 return -EINVAL; 4290 4291 len = ntohl(nla_get_be32(tb[NFTA_SET_FIELD_LEN])); 4292 if (!len || len > U8_MAX) 4293 return -EINVAL; 4294 4295 desc->field_len[desc->field_count++] = len; 4296 4297 return 0; 4298 } 4299 4300 static int nft_set_desc_concat(struct nft_set_desc *desc, 4301 const struct nlattr *nla) 4302 { 4303 struct nlattr *attr; 4304 u32 num_regs = 0; 4305 int rem, err, i; 4306 4307 nla_for_each_nested(attr, nla, rem) { 4308 if (nla_type(attr) != NFTA_LIST_ELEM) 4309 return -EINVAL; 4310 4311 err = nft_set_desc_concat_parse(attr, desc); 4312 if (err < 0) 4313 return err; 4314 } 4315 4316 for (i = 0; i < desc->field_count; i++) 4317 num_regs += DIV_ROUND_UP(desc->field_len[i], sizeof(u32)); 4318 4319 if (num_regs > NFT_REG32_COUNT) 4320 return -E2BIG; 4321 4322 return 0; 4323 } 4324 4325 static int nf_tables_set_desc_parse(struct nft_set_desc *desc, 4326 const struct nlattr *nla) 4327 { 4328 struct nlattr *da[NFTA_SET_DESC_MAX + 1]; 4329 int err; 4330 4331 err = nla_parse_nested_deprecated(da, NFTA_SET_DESC_MAX, nla, 4332 nft_set_desc_policy, NULL); 4333 if (err < 0) 4334 return err; 4335 4336 if (da[NFTA_SET_DESC_SIZE] != NULL) 4337 desc->size = ntohl(nla_get_be32(da[NFTA_SET_DESC_SIZE])); 4338 if (da[NFTA_SET_DESC_CONCAT]) 4339 err = nft_set_desc_concat(desc, da[NFTA_SET_DESC_CONCAT]); 4340 4341 return err; 4342 } 4343 4344 static int nf_tables_newset(struct sk_buff *skb, const struct nfnl_info *info, 4345 const struct nlattr * const nla[]) 4346 { 4347 u32 ktype, dtype, flags, policy, gc_int, objtype; 4348 struct netlink_ext_ack *extack = info->extack; 4349 u8 genmask = nft_genmask_next(info->net); 4350 u8 family = info->nfmsg->nfgen_family; 4351 const struct nft_set_ops *ops; 4352 struct nft_expr *expr = NULL; 4353 struct net *net = info->net; 4354 struct nft_set_desc desc; 4355 struct nft_table *table; 4356 unsigned char *udata; 4357 struct nft_set *set; 4358 struct nft_ctx ctx; 4359 size_t alloc_size; 4360 u64 timeout; 4361 char *name; 4362 int err, i; 4363 u16 udlen; 4364 u64 size; 4365 4366 if (nla[NFTA_SET_TABLE] == NULL || 4367 nla[NFTA_SET_NAME] == NULL || 4368 nla[NFTA_SET_KEY_LEN] == NULL || 4369 nla[NFTA_SET_ID] == NULL) 4370 return -EINVAL; 4371 4372 memset(&desc, 0, sizeof(desc)); 4373 4374 ktype = NFT_DATA_VALUE; 4375 if (nla[NFTA_SET_KEY_TYPE] != NULL) { 4376 ktype = ntohl(nla_get_be32(nla[NFTA_SET_KEY_TYPE])); 4377 if ((ktype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK) 4378 return -EINVAL; 4379 } 4380 4381 desc.klen = ntohl(nla_get_be32(nla[NFTA_SET_KEY_LEN])); 4382 if (desc.klen == 0 || desc.klen > NFT_DATA_VALUE_MAXLEN) 4383 return -EINVAL; 4384 4385 flags = 0; 4386 if (nla[NFTA_SET_FLAGS] != NULL) { 4387 flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS])); 4388 if (flags & ~(NFT_SET_ANONYMOUS | NFT_SET_CONSTANT | 4389 NFT_SET_INTERVAL | NFT_SET_TIMEOUT | 4390 NFT_SET_MAP | NFT_SET_EVAL | 4391 NFT_SET_OBJECT | NFT_SET_CONCAT | NFT_SET_EXPR)) 4392 return -EOPNOTSUPP; 4393 /* Only one of these operations is supported */ 4394 if ((flags & (NFT_SET_MAP | NFT_SET_OBJECT)) == 4395 (NFT_SET_MAP | NFT_SET_OBJECT)) 4396 return -EOPNOTSUPP; 4397 if ((flags & (NFT_SET_EVAL | NFT_SET_OBJECT)) == 4398 (NFT_SET_EVAL | NFT_SET_OBJECT)) 4399 return -EOPNOTSUPP; 4400 } 4401 4402 dtype = 0; 4403 if (nla[NFTA_SET_DATA_TYPE] != NULL) { 4404 if (!(flags & NFT_SET_MAP)) 4405 return -EINVAL; 4406 4407 dtype = ntohl(nla_get_be32(nla[NFTA_SET_DATA_TYPE])); 4408 if ((dtype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK && 4409 dtype != NFT_DATA_VERDICT) 4410 return -EINVAL; 4411 4412 if (dtype != NFT_DATA_VERDICT) { 4413 if (nla[NFTA_SET_DATA_LEN] == NULL) 4414 return -EINVAL; 4415 desc.dlen = ntohl(nla_get_be32(nla[NFTA_SET_DATA_LEN])); 4416 if (desc.dlen == 0 || desc.dlen > NFT_DATA_VALUE_MAXLEN) 4417 return -EINVAL; 4418 } else 4419 desc.dlen = sizeof(struct nft_verdict); 4420 } else if (flags & NFT_SET_MAP) 4421 return -EINVAL; 4422 4423 if (nla[NFTA_SET_OBJ_TYPE] != NULL) { 4424 if (!(flags & NFT_SET_OBJECT)) 4425 return -EINVAL; 4426 4427 objtype = ntohl(nla_get_be32(nla[NFTA_SET_OBJ_TYPE])); 4428 if (objtype == NFT_OBJECT_UNSPEC || 4429 objtype > NFT_OBJECT_MAX) 4430 return -EOPNOTSUPP; 4431 } else if (flags & NFT_SET_OBJECT) 4432 return -EINVAL; 4433 else 4434 objtype = NFT_OBJECT_UNSPEC; 4435 4436 timeout = 0; 4437 if (nla[NFTA_SET_TIMEOUT] != NULL) { 4438 if (!(flags & NFT_SET_TIMEOUT)) 4439 return -EINVAL; 4440 4441 err = nf_msecs_to_jiffies64(nla[NFTA_SET_TIMEOUT], &timeout); 4442 if (err) 4443 return err; 4444 } 4445 gc_int = 0; 4446 if (nla[NFTA_SET_GC_INTERVAL] != NULL) { 4447 if (!(flags & NFT_SET_TIMEOUT)) 4448 return -EINVAL; 4449 gc_int = ntohl(nla_get_be32(nla[NFTA_SET_GC_INTERVAL])); 4450 } 4451 4452 policy = NFT_SET_POL_PERFORMANCE; 4453 if (nla[NFTA_SET_POLICY] != NULL) 4454 policy = ntohl(nla_get_be32(nla[NFTA_SET_POLICY])); 4455 4456 if (nla[NFTA_SET_DESC] != NULL) { 4457 err = nf_tables_set_desc_parse(&desc, nla[NFTA_SET_DESC]); 4458 if (err < 0) 4459 return err; 4460 4461 if (desc.field_count > 1 && !(flags & NFT_SET_CONCAT)) 4462 return -EINVAL; 4463 } else if (flags & NFT_SET_CONCAT) { 4464 return -EINVAL; 4465 } 4466 4467 if (nla[NFTA_SET_EXPR] || nla[NFTA_SET_EXPRESSIONS]) 4468 desc.expr = true; 4469 4470 table = nft_table_lookup(net, nla[NFTA_SET_TABLE], family, genmask, 4471 NETLINK_CB(skb).portid); 4472 if (IS_ERR(table)) { 4473 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_TABLE]); 4474 return PTR_ERR(table); 4475 } 4476 4477 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 4478 4479 set = nft_set_lookup(table, nla[NFTA_SET_NAME], genmask); 4480 if (IS_ERR(set)) { 4481 if (PTR_ERR(set) != -ENOENT) { 4482 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_NAME]); 4483 return PTR_ERR(set); 4484 } 4485 } else { 4486 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 4487 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_NAME]); 4488 return -EEXIST; 4489 } 4490 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 4491 return -EOPNOTSUPP; 4492 4493 return 0; 4494 } 4495 4496 if (!(info->nlh->nlmsg_flags & NLM_F_CREATE)) 4497 return -ENOENT; 4498 4499 ops = nft_select_set_ops(&ctx, nla, &desc, policy); 4500 if (IS_ERR(ops)) 4501 return PTR_ERR(ops); 4502 4503 udlen = 0; 4504 if (nla[NFTA_SET_USERDATA]) 4505 udlen = nla_len(nla[NFTA_SET_USERDATA]); 4506 4507 size = 0; 4508 if (ops->privsize != NULL) 4509 size = ops->privsize(nla, &desc); 4510 alloc_size = sizeof(*set) + size + udlen; 4511 if (alloc_size < size || alloc_size > INT_MAX) 4512 return -ENOMEM; 4513 set = kvzalloc(alloc_size, GFP_KERNEL_ACCOUNT); 4514 if (!set) 4515 return -ENOMEM; 4516 4517 name = nla_strdup(nla[NFTA_SET_NAME], GFP_KERNEL_ACCOUNT); 4518 if (!name) { 4519 err = -ENOMEM; 4520 goto err_set_name; 4521 } 4522 4523 err = nf_tables_set_alloc_name(&ctx, set, name); 4524 kfree(name); 4525 if (err < 0) 4526 goto err_set_name; 4527 4528 udata = NULL; 4529 if (udlen) { 4530 udata = set->data + size; 4531 nla_memcpy(udata, nla[NFTA_SET_USERDATA], udlen); 4532 } 4533 4534 INIT_LIST_HEAD(&set->bindings); 4535 INIT_LIST_HEAD(&set->catchall_list); 4536 set->table = table; 4537 write_pnet(&set->net, net); 4538 set->ops = ops; 4539 set->ktype = ktype; 4540 set->klen = desc.klen; 4541 set->dtype = dtype; 4542 set->objtype = objtype; 4543 set->dlen = desc.dlen; 4544 set->flags = flags; 4545 set->size = desc.size; 4546 set->policy = policy; 4547 set->udlen = udlen; 4548 set->udata = udata; 4549 set->timeout = timeout; 4550 set->gc_int = gc_int; 4551 4552 set->field_count = desc.field_count; 4553 for (i = 0; i < desc.field_count; i++) 4554 set->field_len[i] = desc.field_len[i]; 4555 4556 err = ops->init(set, &desc, nla); 4557 if (err < 0) 4558 goto err_set_init; 4559 4560 if (nla[NFTA_SET_EXPR]) { 4561 expr = nft_set_elem_expr_alloc(&ctx, set, nla[NFTA_SET_EXPR]); 4562 if (IS_ERR(expr)) { 4563 err = PTR_ERR(expr); 4564 goto err_set_expr_alloc; 4565 } 4566 set->exprs[0] = expr; 4567 set->num_exprs++; 4568 } else if (nla[NFTA_SET_EXPRESSIONS]) { 4569 struct nft_expr *expr; 4570 struct nlattr *tmp; 4571 int left; 4572 4573 if (!(flags & NFT_SET_EXPR)) { 4574 err = -EINVAL; 4575 goto err_set_expr_alloc; 4576 } 4577 i = 0; 4578 nla_for_each_nested(tmp, nla[NFTA_SET_EXPRESSIONS], left) { 4579 if (i == NFT_SET_EXPR_MAX) { 4580 err = -E2BIG; 4581 goto err_set_expr_alloc; 4582 } 4583 if (nla_type(tmp) != NFTA_LIST_ELEM) { 4584 err = -EINVAL; 4585 goto err_set_expr_alloc; 4586 } 4587 expr = nft_set_elem_expr_alloc(&ctx, set, tmp); 4588 if (IS_ERR(expr)) { 4589 err = PTR_ERR(expr); 4590 goto err_set_expr_alloc; 4591 } 4592 set->exprs[i++] = expr; 4593 set->num_exprs++; 4594 } 4595 } 4596 4597 set->handle = nf_tables_alloc_handle(table); 4598 4599 err = nft_trans_set_add(&ctx, NFT_MSG_NEWSET, set); 4600 if (err < 0) 4601 goto err_set_expr_alloc; 4602 4603 list_add_tail_rcu(&set->list, &table->sets); 4604 table->use++; 4605 return 0; 4606 4607 err_set_expr_alloc: 4608 for (i = 0; i < set->num_exprs; i++) 4609 nft_expr_destroy(&ctx, set->exprs[i]); 4610 4611 ops->destroy(set); 4612 err_set_init: 4613 kfree(set->name); 4614 err_set_name: 4615 kvfree(set); 4616 return err; 4617 } 4618 4619 struct nft_set_elem_catchall { 4620 struct list_head list; 4621 struct rcu_head rcu; 4622 void *elem; 4623 }; 4624 4625 static void nft_set_catchall_destroy(const struct nft_ctx *ctx, 4626 struct nft_set *set) 4627 { 4628 struct nft_set_elem_catchall *next, *catchall; 4629 4630 list_for_each_entry_safe(catchall, next, &set->catchall_list, list) { 4631 list_del_rcu(&catchall->list); 4632 nft_set_elem_destroy(set, catchall->elem, true); 4633 kfree_rcu(catchall, rcu); 4634 } 4635 } 4636 4637 static void nft_set_destroy(const struct nft_ctx *ctx, struct nft_set *set) 4638 { 4639 int i; 4640 4641 if (WARN_ON(set->use > 0)) 4642 return; 4643 4644 for (i = 0; i < set->num_exprs; i++) 4645 nft_expr_destroy(ctx, set->exprs[i]); 4646 4647 set->ops->destroy(set); 4648 nft_set_catchall_destroy(ctx, set); 4649 kfree(set->name); 4650 kvfree(set); 4651 } 4652 4653 static int nf_tables_delset(struct sk_buff *skb, const struct nfnl_info *info, 4654 const struct nlattr * const nla[]) 4655 { 4656 struct netlink_ext_ack *extack = info->extack; 4657 u8 genmask = nft_genmask_next(info->net); 4658 u8 family = info->nfmsg->nfgen_family; 4659 struct net *net = info->net; 4660 const struct nlattr *attr; 4661 struct nft_table *table; 4662 struct nft_set *set; 4663 struct nft_ctx ctx; 4664 4665 if (info->nfmsg->nfgen_family == NFPROTO_UNSPEC) 4666 return -EAFNOSUPPORT; 4667 4668 table = nft_table_lookup(net, nla[NFTA_SET_TABLE], family, 4669 genmask, NETLINK_CB(skb).portid); 4670 if (IS_ERR(table)) { 4671 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_TABLE]); 4672 return PTR_ERR(table); 4673 } 4674 4675 if (nla[NFTA_SET_HANDLE]) { 4676 attr = nla[NFTA_SET_HANDLE]; 4677 set = nft_set_lookup_byhandle(table, attr, genmask); 4678 } else { 4679 attr = nla[NFTA_SET_NAME]; 4680 set = nft_set_lookup(table, attr, genmask); 4681 } 4682 4683 if (IS_ERR(set)) { 4684 NL_SET_BAD_ATTR(extack, attr); 4685 return PTR_ERR(set); 4686 } 4687 if (set->use || 4688 (info->nlh->nlmsg_flags & NLM_F_NONREC && 4689 atomic_read(&set->nelems) > 0)) { 4690 NL_SET_BAD_ATTR(extack, attr); 4691 return -EBUSY; 4692 } 4693 4694 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 4695 4696 return nft_delset(&ctx, set); 4697 } 4698 4699 static int nft_validate_register_store(const struct nft_ctx *ctx, 4700 enum nft_registers reg, 4701 const struct nft_data *data, 4702 enum nft_data_types type, 4703 unsigned int len); 4704 4705 static int nft_setelem_data_validate(const struct nft_ctx *ctx, 4706 struct nft_set *set, 4707 struct nft_set_elem *elem) 4708 { 4709 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 4710 enum nft_registers dreg; 4711 4712 dreg = nft_type_to_reg(set->dtype); 4713 return nft_validate_register_store(ctx, dreg, nft_set_ext_data(ext), 4714 set->dtype == NFT_DATA_VERDICT ? 4715 NFT_DATA_VERDICT : NFT_DATA_VALUE, 4716 set->dlen); 4717 } 4718 4719 static int nf_tables_bind_check_setelem(const struct nft_ctx *ctx, 4720 struct nft_set *set, 4721 const struct nft_set_iter *iter, 4722 struct nft_set_elem *elem) 4723 { 4724 return nft_setelem_data_validate(ctx, set, elem); 4725 } 4726 4727 static int nft_set_catchall_bind_check(const struct nft_ctx *ctx, 4728 struct nft_set *set) 4729 { 4730 u8 genmask = nft_genmask_next(ctx->net); 4731 struct nft_set_elem_catchall *catchall; 4732 struct nft_set_elem elem; 4733 struct nft_set_ext *ext; 4734 int ret = 0; 4735 4736 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 4737 ext = nft_set_elem_ext(set, catchall->elem); 4738 if (!nft_set_elem_active(ext, genmask)) 4739 continue; 4740 4741 elem.priv = catchall->elem; 4742 ret = nft_setelem_data_validate(ctx, set, &elem); 4743 if (ret < 0) 4744 break; 4745 } 4746 4747 return ret; 4748 } 4749 4750 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set, 4751 struct nft_set_binding *binding) 4752 { 4753 struct nft_set_binding *i; 4754 struct nft_set_iter iter; 4755 4756 if (set->use == UINT_MAX) 4757 return -EOVERFLOW; 4758 4759 if (!list_empty(&set->bindings) && nft_set_is_anonymous(set)) 4760 return -EBUSY; 4761 4762 if (binding->flags & NFT_SET_MAP) { 4763 /* If the set is already bound to the same chain all 4764 * jumps are already validated for that chain. 4765 */ 4766 list_for_each_entry(i, &set->bindings, list) { 4767 if (i->flags & NFT_SET_MAP && 4768 i->chain == binding->chain) 4769 goto bind; 4770 } 4771 4772 iter.genmask = nft_genmask_next(ctx->net); 4773 iter.skip = 0; 4774 iter.count = 0; 4775 iter.err = 0; 4776 iter.fn = nf_tables_bind_check_setelem; 4777 4778 set->ops->walk(ctx, set, &iter); 4779 if (!iter.err) 4780 iter.err = nft_set_catchall_bind_check(ctx, set); 4781 4782 if (iter.err < 0) 4783 return iter.err; 4784 } 4785 bind: 4786 binding->chain = ctx->chain; 4787 list_add_tail_rcu(&binding->list, &set->bindings); 4788 nft_set_trans_bind(ctx, set); 4789 set->use++; 4790 4791 return 0; 4792 } 4793 EXPORT_SYMBOL_GPL(nf_tables_bind_set); 4794 4795 static void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set, 4796 struct nft_set_binding *binding, bool event) 4797 { 4798 list_del_rcu(&binding->list); 4799 4800 if (list_empty(&set->bindings) && nft_set_is_anonymous(set)) { 4801 list_del_rcu(&set->list); 4802 if (event) 4803 nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, 4804 GFP_KERNEL); 4805 } 4806 } 4807 4808 void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set, 4809 struct nft_set_binding *binding, 4810 enum nft_trans_phase phase) 4811 { 4812 switch (phase) { 4813 case NFT_TRANS_PREPARE: 4814 set->use--; 4815 return; 4816 case NFT_TRANS_ABORT: 4817 case NFT_TRANS_RELEASE: 4818 set->use--; 4819 fallthrough; 4820 default: 4821 nf_tables_unbind_set(ctx, set, binding, 4822 phase == NFT_TRANS_COMMIT); 4823 } 4824 } 4825 EXPORT_SYMBOL_GPL(nf_tables_deactivate_set); 4826 4827 void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set) 4828 { 4829 if (list_empty(&set->bindings) && nft_set_is_anonymous(set)) 4830 nft_set_destroy(ctx, set); 4831 } 4832 EXPORT_SYMBOL_GPL(nf_tables_destroy_set); 4833 4834 const struct nft_set_ext_type nft_set_ext_types[] = { 4835 [NFT_SET_EXT_KEY] = { 4836 .align = __alignof__(u32), 4837 }, 4838 [NFT_SET_EXT_DATA] = { 4839 .align = __alignof__(u32), 4840 }, 4841 [NFT_SET_EXT_EXPRESSIONS] = { 4842 .align = __alignof__(struct nft_set_elem_expr), 4843 }, 4844 [NFT_SET_EXT_OBJREF] = { 4845 .len = sizeof(struct nft_object *), 4846 .align = __alignof__(struct nft_object *), 4847 }, 4848 [NFT_SET_EXT_FLAGS] = { 4849 .len = sizeof(u8), 4850 .align = __alignof__(u8), 4851 }, 4852 [NFT_SET_EXT_TIMEOUT] = { 4853 .len = sizeof(u64), 4854 .align = __alignof__(u64), 4855 }, 4856 [NFT_SET_EXT_EXPIRATION] = { 4857 .len = sizeof(u64), 4858 .align = __alignof__(u64), 4859 }, 4860 [NFT_SET_EXT_USERDATA] = { 4861 .len = sizeof(struct nft_userdata), 4862 .align = __alignof__(struct nft_userdata), 4863 }, 4864 [NFT_SET_EXT_KEY_END] = { 4865 .align = __alignof__(u32), 4866 }, 4867 }; 4868 4869 /* 4870 * Set elements 4871 */ 4872 4873 static const struct nla_policy nft_set_elem_policy[NFTA_SET_ELEM_MAX + 1] = { 4874 [NFTA_SET_ELEM_KEY] = { .type = NLA_NESTED }, 4875 [NFTA_SET_ELEM_DATA] = { .type = NLA_NESTED }, 4876 [NFTA_SET_ELEM_FLAGS] = { .type = NLA_U32 }, 4877 [NFTA_SET_ELEM_TIMEOUT] = { .type = NLA_U64 }, 4878 [NFTA_SET_ELEM_EXPIRATION] = { .type = NLA_U64 }, 4879 [NFTA_SET_ELEM_USERDATA] = { .type = NLA_BINARY, 4880 .len = NFT_USERDATA_MAXLEN }, 4881 [NFTA_SET_ELEM_EXPR] = { .type = NLA_NESTED }, 4882 [NFTA_SET_ELEM_OBJREF] = { .type = NLA_STRING, 4883 .len = NFT_OBJ_MAXNAMELEN - 1 }, 4884 [NFTA_SET_ELEM_KEY_END] = { .type = NLA_NESTED }, 4885 [NFTA_SET_ELEM_EXPRESSIONS] = { .type = NLA_NESTED }, 4886 }; 4887 4888 static const struct nla_policy nft_set_elem_list_policy[NFTA_SET_ELEM_LIST_MAX + 1] = { 4889 [NFTA_SET_ELEM_LIST_TABLE] = { .type = NLA_STRING, 4890 .len = NFT_TABLE_MAXNAMELEN - 1 }, 4891 [NFTA_SET_ELEM_LIST_SET] = { .type = NLA_STRING, 4892 .len = NFT_SET_MAXNAMELEN - 1 }, 4893 [NFTA_SET_ELEM_LIST_ELEMENTS] = { .type = NLA_NESTED }, 4894 [NFTA_SET_ELEM_LIST_SET_ID] = { .type = NLA_U32 }, 4895 }; 4896 4897 static int nft_set_elem_expr_dump(struct sk_buff *skb, 4898 const struct nft_set *set, 4899 const struct nft_set_ext *ext) 4900 { 4901 struct nft_set_elem_expr *elem_expr; 4902 u32 size, num_exprs = 0; 4903 struct nft_expr *expr; 4904 struct nlattr *nest; 4905 4906 elem_expr = nft_set_ext_expr(ext); 4907 nft_setelem_expr_foreach(expr, elem_expr, size) 4908 num_exprs++; 4909 4910 if (num_exprs == 1) { 4911 expr = nft_setelem_expr_at(elem_expr, 0); 4912 if (nft_expr_dump(skb, NFTA_SET_ELEM_EXPR, expr) < 0) 4913 return -1; 4914 4915 return 0; 4916 } else if (num_exprs > 1) { 4917 nest = nla_nest_start_noflag(skb, NFTA_SET_ELEM_EXPRESSIONS); 4918 if (nest == NULL) 4919 goto nla_put_failure; 4920 4921 nft_setelem_expr_foreach(expr, elem_expr, size) { 4922 expr = nft_setelem_expr_at(elem_expr, size); 4923 if (nft_expr_dump(skb, NFTA_LIST_ELEM, expr) < 0) 4924 goto nla_put_failure; 4925 } 4926 nla_nest_end(skb, nest); 4927 } 4928 return 0; 4929 4930 nla_put_failure: 4931 return -1; 4932 } 4933 4934 static int nf_tables_fill_setelem(struct sk_buff *skb, 4935 const struct nft_set *set, 4936 const struct nft_set_elem *elem) 4937 { 4938 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 4939 unsigned char *b = skb_tail_pointer(skb); 4940 struct nlattr *nest; 4941 4942 nest = nla_nest_start_noflag(skb, NFTA_LIST_ELEM); 4943 if (nest == NULL) 4944 goto nla_put_failure; 4945 4946 if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY) && 4947 nft_data_dump(skb, NFTA_SET_ELEM_KEY, nft_set_ext_key(ext), 4948 NFT_DATA_VALUE, set->klen) < 0) 4949 goto nla_put_failure; 4950 4951 if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY_END) && 4952 nft_data_dump(skb, NFTA_SET_ELEM_KEY_END, nft_set_ext_key_end(ext), 4953 NFT_DATA_VALUE, set->klen) < 0) 4954 goto nla_put_failure; 4955 4956 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) && 4957 nft_data_dump(skb, NFTA_SET_ELEM_DATA, nft_set_ext_data(ext), 4958 set->dtype == NFT_DATA_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE, 4959 set->dlen) < 0) 4960 goto nla_put_failure; 4961 4962 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPRESSIONS) && 4963 nft_set_elem_expr_dump(skb, set, ext)) 4964 goto nla_put_failure; 4965 4966 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) && 4967 nla_put_string(skb, NFTA_SET_ELEM_OBJREF, 4968 (*nft_set_ext_obj(ext))->key.name) < 0) 4969 goto nla_put_failure; 4970 4971 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) && 4972 nla_put_be32(skb, NFTA_SET_ELEM_FLAGS, 4973 htonl(*nft_set_ext_flags(ext)))) 4974 goto nla_put_failure; 4975 4976 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT) && 4977 nla_put_be64(skb, NFTA_SET_ELEM_TIMEOUT, 4978 nf_jiffies64_to_msecs(*nft_set_ext_timeout(ext)), 4979 NFTA_SET_ELEM_PAD)) 4980 goto nla_put_failure; 4981 4982 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION)) { 4983 u64 expires, now = get_jiffies_64(); 4984 4985 expires = *nft_set_ext_expiration(ext); 4986 if (time_before64(now, expires)) 4987 expires -= now; 4988 else 4989 expires = 0; 4990 4991 if (nla_put_be64(skb, NFTA_SET_ELEM_EXPIRATION, 4992 nf_jiffies64_to_msecs(expires), 4993 NFTA_SET_ELEM_PAD)) 4994 goto nla_put_failure; 4995 } 4996 4997 if (nft_set_ext_exists(ext, NFT_SET_EXT_USERDATA)) { 4998 struct nft_userdata *udata; 4999 5000 udata = nft_set_ext_userdata(ext); 5001 if (nla_put(skb, NFTA_SET_ELEM_USERDATA, 5002 udata->len + 1, udata->data)) 5003 goto nla_put_failure; 5004 } 5005 5006 nla_nest_end(skb, nest); 5007 return 0; 5008 5009 nla_put_failure: 5010 nlmsg_trim(skb, b); 5011 return -EMSGSIZE; 5012 } 5013 5014 struct nft_set_dump_args { 5015 const struct netlink_callback *cb; 5016 struct nft_set_iter iter; 5017 struct sk_buff *skb; 5018 }; 5019 5020 static int nf_tables_dump_setelem(const struct nft_ctx *ctx, 5021 struct nft_set *set, 5022 const struct nft_set_iter *iter, 5023 struct nft_set_elem *elem) 5024 { 5025 struct nft_set_dump_args *args; 5026 5027 args = container_of(iter, struct nft_set_dump_args, iter); 5028 return nf_tables_fill_setelem(args->skb, set, elem); 5029 } 5030 5031 struct nft_set_dump_ctx { 5032 const struct nft_set *set; 5033 struct nft_ctx ctx; 5034 }; 5035 5036 static int nft_set_catchall_dump(struct net *net, struct sk_buff *skb, 5037 const struct nft_set *set) 5038 { 5039 struct nft_set_elem_catchall *catchall; 5040 u8 genmask = nft_genmask_cur(net); 5041 struct nft_set_elem elem; 5042 struct nft_set_ext *ext; 5043 int ret = 0; 5044 5045 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 5046 ext = nft_set_elem_ext(set, catchall->elem); 5047 if (!nft_set_elem_active(ext, genmask) || 5048 nft_set_elem_expired(ext)) 5049 continue; 5050 5051 elem.priv = catchall->elem; 5052 ret = nf_tables_fill_setelem(skb, set, &elem); 5053 break; 5054 } 5055 5056 return ret; 5057 } 5058 5059 static int nf_tables_dump_set(struct sk_buff *skb, struct netlink_callback *cb) 5060 { 5061 struct nft_set_dump_ctx *dump_ctx = cb->data; 5062 struct net *net = sock_net(skb->sk); 5063 struct nftables_pernet *nft_net; 5064 struct nft_table *table; 5065 struct nft_set *set; 5066 struct nft_set_dump_args args; 5067 bool set_found = false; 5068 struct nlmsghdr *nlh; 5069 struct nlattr *nest; 5070 u32 portid, seq; 5071 int event; 5072 5073 rcu_read_lock(); 5074 nft_net = nft_pernet(net); 5075 cb->seq = READ_ONCE(nft_net->base_seq); 5076 5077 list_for_each_entry_rcu(table, &nft_net->tables, list) { 5078 if (dump_ctx->ctx.family != NFPROTO_UNSPEC && 5079 dump_ctx->ctx.family != table->family) 5080 continue; 5081 5082 if (table != dump_ctx->ctx.table) 5083 continue; 5084 5085 list_for_each_entry_rcu(set, &table->sets, list) { 5086 if (set == dump_ctx->set) { 5087 set_found = true; 5088 break; 5089 } 5090 } 5091 break; 5092 } 5093 5094 if (!set_found) { 5095 rcu_read_unlock(); 5096 return -ENOENT; 5097 } 5098 5099 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, NFT_MSG_NEWSETELEM); 5100 portid = NETLINK_CB(cb->skb).portid; 5101 seq = cb->nlh->nlmsg_seq; 5102 5103 nlh = nfnl_msg_put(skb, portid, seq, event, NLM_F_MULTI, 5104 table->family, NFNETLINK_V0, nft_base_seq(net)); 5105 if (!nlh) 5106 goto nla_put_failure; 5107 5108 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_TABLE, table->name)) 5109 goto nla_put_failure; 5110 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_SET, set->name)) 5111 goto nla_put_failure; 5112 5113 nest = nla_nest_start_noflag(skb, NFTA_SET_ELEM_LIST_ELEMENTS); 5114 if (nest == NULL) 5115 goto nla_put_failure; 5116 5117 args.cb = cb; 5118 args.skb = skb; 5119 args.iter.genmask = nft_genmask_cur(net); 5120 args.iter.skip = cb->args[0]; 5121 args.iter.count = 0; 5122 args.iter.err = 0; 5123 args.iter.fn = nf_tables_dump_setelem; 5124 set->ops->walk(&dump_ctx->ctx, set, &args.iter); 5125 5126 if (!args.iter.err && args.iter.count == cb->args[0]) 5127 args.iter.err = nft_set_catchall_dump(net, skb, set); 5128 rcu_read_unlock(); 5129 5130 nla_nest_end(skb, nest); 5131 nlmsg_end(skb, nlh); 5132 5133 if (args.iter.err && args.iter.err != -EMSGSIZE) 5134 return args.iter.err; 5135 if (args.iter.count == cb->args[0]) 5136 return 0; 5137 5138 cb->args[0] = args.iter.count; 5139 return skb->len; 5140 5141 nla_put_failure: 5142 rcu_read_unlock(); 5143 return -ENOSPC; 5144 } 5145 5146 static int nf_tables_dump_set_start(struct netlink_callback *cb) 5147 { 5148 struct nft_set_dump_ctx *dump_ctx = cb->data; 5149 5150 cb->data = kmemdup(dump_ctx, sizeof(*dump_ctx), GFP_ATOMIC); 5151 5152 return cb->data ? 0 : -ENOMEM; 5153 } 5154 5155 static int nf_tables_dump_set_done(struct netlink_callback *cb) 5156 { 5157 kfree(cb->data); 5158 return 0; 5159 } 5160 5161 static int nf_tables_fill_setelem_info(struct sk_buff *skb, 5162 const struct nft_ctx *ctx, u32 seq, 5163 u32 portid, int event, u16 flags, 5164 const struct nft_set *set, 5165 const struct nft_set_elem *elem) 5166 { 5167 struct nlmsghdr *nlh; 5168 struct nlattr *nest; 5169 int err; 5170 5171 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 5172 nlh = nfnl_msg_put(skb, portid, seq, event, flags, ctx->family, 5173 NFNETLINK_V0, nft_base_seq(ctx->net)); 5174 if (!nlh) 5175 goto nla_put_failure; 5176 5177 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name)) 5178 goto nla_put_failure; 5179 if (nla_put_string(skb, NFTA_SET_NAME, set->name)) 5180 goto nla_put_failure; 5181 5182 nest = nla_nest_start_noflag(skb, NFTA_SET_ELEM_LIST_ELEMENTS); 5183 if (nest == NULL) 5184 goto nla_put_failure; 5185 5186 err = nf_tables_fill_setelem(skb, set, elem); 5187 if (err < 0) 5188 goto nla_put_failure; 5189 5190 nla_nest_end(skb, nest); 5191 5192 nlmsg_end(skb, nlh); 5193 return 0; 5194 5195 nla_put_failure: 5196 nlmsg_trim(skb, nlh); 5197 return -1; 5198 } 5199 5200 static int nft_setelem_parse_flags(const struct nft_set *set, 5201 const struct nlattr *attr, u32 *flags) 5202 { 5203 if (attr == NULL) 5204 return 0; 5205 5206 *flags = ntohl(nla_get_be32(attr)); 5207 if (*flags & ~(NFT_SET_ELEM_INTERVAL_END | NFT_SET_ELEM_CATCHALL)) 5208 return -EOPNOTSUPP; 5209 if (!(set->flags & NFT_SET_INTERVAL) && 5210 *flags & NFT_SET_ELEM_INTERVAL_END) 5211 return -EINVAL; 5212 if ((*flags & (NFT_SET_ELEM_INTERVAL_END | NFT_SET_ELEM_CATCHALL)) == 5213 (NFT_SET_ELEM_INTERVAL_END | NFT_SET_ELEM_CATCHALL)) 5214 return -EINVAL; 5215 5216 return 0; 5217 } 5218 5219 static int nft_setelem_parse_key(struct nft_ctx *ctx, struct nft_set *set, 5220 struct nft_data *key, struct nlattr *attr) 5221 { 5222 struct nft_data_desc desc = { 5223 .type = NFT_DATA_VALUE, 5224 .size = NFT_DATA_VALUE_MAXLEN, 5225 .len = set->klen, 5226 }; 5227 5228 return nft_data_init(ctx, key, &desc, attr); 5229 } 5230 5231 static int nft_setelem_parse_data(struct nft_ctx *ctx, struct nft_set *set, 5232 struct nft_data_desc *desc, 5233 struct nft_data *data, 5234 struct nlattr *attr) 5235 { 5236 u32 dtype; 5237 5238 if (set->dtype == NFT_DATA_VERDICT) 5239 dtype = NFT_DATA_VERDICT; 5240 else 5241 dtype = NFT_DATA_VALUE; 5242 5243 desc->type = dtype; 5244 desc->size = NFT_DATA_VALUE_MAXLEN; 5245 desc->len = set->dlen; 5246 desc->flags = NFT_DATA_DESC_SETELEM; 5247 5248 return nft_data_init(ctx, data, desc, attr); 5249 } 5250 5251 static void *nft_setelem_catchall_get(const struct net *net, 5252 const struct nft_set *set) 5253 { 5254 struct nft_set_elem_catchall *catchall; 5255 u8 genmask = nft_genmask_cur(net); 5256 struct nft_set_ext *ext; 5257 void *priv = NULL; 5258 5259 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 5260 ext = nft_set_elem_ext(set, catchall->elem); 5261 if (!nft_set_elem_active(ext, genmask) || 5262 nft_set_elem_expired(ext)) 5263 continue; 5264 5265 priv = catchall->elem; 5266 break; 5267 } 5268 5269 return priv; 5270 } 5271 5272 static int nft_setelem_get(struct nft_ctx *ctx, struct nft_set *set, 5273 struct nft_set_elem *elem, u32 flags) 5274 { 5275 void *priv; 5276 5277 if (!(flags & NFT_SET_ELEM_CATCHALL)) { 5278 priv = set->ops->get(ctx->net, set, elem, flags); 5279 if (IS_ERR(priv)) 5280 return PTR_ERR(priv); 5281 } else { 5282 priv = nft_setelem_catchall_get(ctx->net, set); 5283 if (!priv) 5284 return -ENOENT; 5285 } 5286 elem->priv = priv; 5287 5288 return 0; 5289 } 5290 5291 static int nft_get_set_elem(struct nft_ctx *ctx, struct nft_set *set, 5292 const struct nlattr *attr) 5293 { 5294 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1]; 5295 struct nft_set_elem elem; 5296 struct sk_buff *skb; 5297 uint32_t flags = 0; 5298 int err; 5299 5300 err = nla_parse_nested_deprecated(nla, NFTA_SET_ELEM_MAX, attr, 5301 nft_set_elem_policy, NULL); 5302 if (err < 0) 5303 return err; 5304 5305 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags); 5306 if (err < 0) 5307 return err; 5308 5309 if (!nla[NFTA_SET_ELEM_KEY] && !(flags & NFT_SET_ELEM_CATCHALL)) 5310 return -EINVAL; 5311 5312 if (nla[NFTA_SET_ELEM_KEY]) { 5313 err = nft_setelem_parse_key(ctx, set, &elem.key.val, 5314 nla[NFTA_SET_ELEM_KEY]); 5315 if (err < 0) 5316 return err; 5317 } 5318 5319 if (nla[NFTA_SET_ELEM_KEY_END]) { 5320 err = nft_setelem_parse_key(ctx, set, &elem.key_end.val, 5321 nla[NFTA_SET_ELEM_KEY_END]); 5322 if (err < 0) 5323 return err; 5324 } 5325 5326 err = nft_setelem_get(ctx, set, &elem, flags); 5327 if (err < 0) 5328 return err; 5329 5330 err = -ENOMEM; 5331 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_ATOMIC); 5332 if (skb == NULL) 5333 return err; 5334 5335 err = nf_tables_fill_setelem_info(skb, ctx, ctx->seq, ctx->portid, 5336 NFT_MSG_NEWSETELEM, 0, set, &elem); 5337 if (err < 0) 5338 goto err_fill_setelem; 5339 5340 return nfnetlink_unicast(skb, ctx->net, ctx->portid); 5341 5342 err_fill_setelem: 5343 kfree_skb(skb); 5344 return err; 5345 } 5346 5347 /* called with rcu_read_lock held */ 5348 static int nf_tables_getsetelem(struct sk_buff *skb, 5349 const struct nfnl_info *info, 5350 const struct nlattr * const nla[]) 5351 { 5352 struct netlink_ext_ack *extack = info->extack; 5353 u8 genmask = nft_genmask_cur(info->net); 5354 u8 family = info->nfmsg->nfgen_family; 5355 struct net *net = info->net; 5356 struct nft_table *table; 5357 struct nft_set *set; 5358 struct nlattr *attr; 5359 struct nft_ctx ctx; 5360 int rem, err = 0; 5361 5362 table = nft_table_lookup(net, nla[NFTA_SET_ELEM_LIST_TABLE], family, 5363 genmask, NETLINK_CB(skb).portid); 5364 if (IS_ERR(table)) { 5365 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_ELEM_LIST_TABLE]); 5366 return PTR_ERR(table); 5367 } 5368 5369 set = nft_set_lookup(table, nla[NFTA_SET_ELEM_LIST_SET], genmask); 5370 if (IS_ERR(set)) 5371 return PTR_ERR(set); 5372 5373 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 5374 5375 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 5376 struct netlink_dump_control c = { 5377 .start = nf_tables_dump_set_start, 5378 .dump = nf_tables_dump_set, 5379 .done = nf_tables_dump_set_done, 5380 .module = THIS_MODULE, 5381 }; 5382 struct nft_set_dump_ctx dump_ctx = { 5383 .set = set, 5384 .ctx = ctx, 5385 }; 5386 5387 c.data = &dump_ctx; 5388 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 5389 } 5390 5391 if (!nla[NFTA_SET_ELEM_LIST_ELEMENTS]) 5392 return -EINVAL; 5393 5394 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) { 5395 err = nft_get_set_elem(&ctx, set, attr); 5396 if (err < 0) { 5397 NL_SET_BAD_ATTR(extack, attr); 5398 break; 5399 } 5400 } 5401 5402 return err; 5403 } 5404 5405 static void nf_tables_setelem_notify(const struct nft_ctx *ctx, 5406 const struct nft_set *set, 5407 const struct nft_set_elem *elem, 5408 int event) 5409 { 5410 struct nftables_pernet *nft_net; 5411 struct net *net = ctx->net; 5412 u32 portid = ctx->portid; 5413 struct sk_buff *skb; 5414 u16 flags = 0; 5415 int err; 5416 5417 if (!ctx->report && !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES)) 5418 return; 5419 5420 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 5421 if (skb == NULL) 5422 goto err; 5423 5424 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 5425 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 5426 5427 err = nf_tables_fill_setelem_info(skb, ctx, 0, portid, event, flags, 5428 set, elem); 5429 if (err < 0) { 5430 kfree_skb(skb); 5431 goto err; 5432 } 5433 5434 nft_net = nft_pernet(net); 5435 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 5436 return; 5437 err: 5438 nfnetlink_set_err(net, portid, NFNLGRP_NFTABLES, -ENOBUFS); 5439 } 5440 5441 static struct nft_trans *nft_trans_elem_alloc(struct nft_ctx *ctx, 5442 int msg_type, 5443 struct nft_set *set) 5444 { 5445 struct nft_trans *trans; 5446 5447 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_elem)); 5448 if (trans == NULL) 5449 return NULL; 5450 5451 nft_trans_elem_set(trans) = set; 5452 return trans; 5453 } 5454 5455 struct nft_expr *nft_set_elem_expr_alloc(const struct nft_ctx *ctx, 5456 const struct nft_set *set, 5457 const struct nlattr *attr) 5458 { 5459 struct nft_expr *expr; 5460 int err; 5461 5462 expr = nft_expr_init(ctx, attr); 5463 if (IS_ERR(expr)) 5464 return expr; 5465 5466 err = -EOPNOTSUPP; 5467 if (expr->ops->type->flags & NFT_EXPR_GC) { 5468 if (set->flags & NFT_SET_TIMEOUT) 5469 goto err_set_elem_expr; 5470 if (!set->ops->gc_init) 5471 goto err_set_elem_expr; 5472 set->ops->gc_init(set); 5473 } 5474 5475 return expr; 5476 5477 err_set_elem_expr: 5478 nft_expr_destroy(ctx, expr); 5479 return ERR_PTR(err); 5480 } 5481 5482 static int nft_set_ext_check(const struct nft_set_ext_tmpl *tmpl, u8 id, u32 len) 5483 { 5484 len += nft_set_ext_types[id].len; 5485 if (len > tmpl->ext_len[id] || 5486 len > U8_MAX) 5487 return -1; 5488 5489 return 0; 5490 } 5491 5492 static int nft_set_ext_memcpy(const struct nft_set_ext_tmpl *tmpl, u8 id, 5493 void *to, const void *from, u32 len) 5494 { 5495 if (nft_set_ext_check(tmpl, id, len) < 0) 5496 return -1; 5497 5498 memcpy(to, from, len); 5499 5500 return 0; 5501 } 5502 5503 void *nft_set_elem_init(const struct nft_set *set, 5504 const struct nft_set_ext_tmpl *tmpl, 5505 const u32 *key, const u32 *key_end, 5506 const u32 *data, u64 timeout, u64 expiration, gfp_t gfp) 5507 { 5508 struct nft_set_ext *ext; 5509 void *elem; 5510 5511 elem = kzalloc(set->ops->elemsize + tmpl->len, gfp); 5512 if (elem == NULL) 5513 return ERR_PTR(-ENOMEM); 5514 5515 ext = nft_set_elem_ext(set, elem); 5516 nft_set_ext_init(ext, tmpl); 5517 5518 if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY) && 5519 nft_set_ext_memcpy(tmpl, NFT_SET_EXT_KEY, 5520 nft_set_ext_key(ext), key, set->klen) < 0) 5521 goto err_ext_check; 5522 5523 if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY_END) && 5524 nft_set_ext_memcpy(tmpl, NFT_SET_EXT_KEY_END, 5525 nft_set_ext_key_end(ext), key_end, set->klen) < 0) 5526 goto err_ext_check; 5527 5528 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) && 5529 nft_set_ext_memcpy(tmpl, NFT_SET_EXT_DATA, 5530 nft_set_ext_data(ext), data, set->dlen) < 0) 5531 goto err_ext_check; 5532 5533 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION)) { 5534 *nft_set_ext_expiration(ext) = get_jiffies_64() + expiration; 5535 if (expiration == 0) 5536 *nft_set_ext_expiration(ext) += timeout; 5537 } 5538 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT)) 5539 *nft_set_ext_timeout(ext) = timeout; 5540 5541 return elem; 5542 5543 err_ext_check: 5544 kfree(elem); 5545 5546 return ERR_PTR(-EINVAL); 5547 } 5548 5549 static void __nft_set_elem_expr_destroy(const struct nft_ctx *ctx, 5550 struct nft_expr *expr) 5551 { 5552 if (expr->ops->destroy_clone) { 5553 expr->ops->destroy_clone(ctx, expr); 5554 module_put(expr->ops->type->owner); 5555 } else { 5556 nf_tables_expr_destroy(ctx, expr); 5557 } 5558 } 5559 5560 static void nft_set_elem_expr_destroy(const struct nft_ctx *ctx, 5561 struct nft_set_elem_expr *elem_expr) 5562 { 5563 struct nft_expr *expr; 5564 u32 size; 5565 5566 nft_setelem_expr_foreach(expr, elem_expr, size) 5567 __nft_set_elem_expr_destroy(ctx, expr); 5568 } 5569 5570 void nft_set_elem_destroy(const struct nft_set *set, void *elem, 5571 bool destroy_expr) 5572 { 5573 struct nft_set_ext *ext = nft_set_elem_ext(set, elem); 5574 struct nft_ctx ctx = { 5575 .net = read_pnet(&set->net), 5576 .family = set->table->family, 5577 }; 5578 5579 nft_data_release(nft_set_ext_key(ext), NFT_DATA_VALUE); 5580 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA)) 5581 nft_data_release(nft_set_ext_data(ext), set->dtype); 5582 if (destroy_expr && nft_set_ext_exists(ext, NFT_SET_EXT_EXPRESSIONS)) 5583 nft_set_elem_expr_destroy(&ctx, nft_set_ext_expr(ext)); 5584 5585 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF)) 5586 (*nft_set_ext_obj(ext))->use--; 5587 kfree(elem); 5588 } 5589 EXPORT_SYMBOL_GPL(nft_set_elem_destroy); 5590 5591 /* Only called from commit path, nft_setelem_data_deactivate() already deals 5592 * with the refcounting from the preparation phase. 5593 */ 5594 static void nf_tables_set_elem_destroy(const struct nft_ctx *ctx, 5595 const struct nft_set *set, void *elem) 5596 { 5597 struct nft_set_ext *ext = nft_set_elem_ext(set, elem); 5598 5599 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPRESSIONS)) 5600 nft_set_elem_expr_destroy(ctx, nft_set_ext_expr(ext)); 5601 5602 kfree(elem); 5603 } 5604 5605 int nft_set_elem_expr_clone(const struct nft_ctx *ctx, struct nft_set *set, 5606 struct nft_expr *expr_array[]) 5607 { 5608 struct nft_expr *expr; 5609 int err, i, k; 5610 5611 for (i = 0; i < set->num_exprs; i++) { 5612 expr = kzalloc(set->exprs[i]->ops->size, GFP_KERNEL_ACCOUNT); 5613 if (!expr) 5614 goto err_expr; 5615 5616 err = nft_expr_clone(expr, set->exprs[i]); 5617 if (err < 0) { 5618 kfree(expr); 5619 goto err_expr; 5620 } 5621 expr_array[i] = expr; 5622 } 5623 5624 return 0; 5625 5626 err_expr: 5627 for (k = i - 1; k >= 0; k--) 5628 nft_expr_destroy(ctx, expr_array[k]); 5629 5630 return -ENOMEM; 5631 } 5632 5633 static int nft_set_elem_expr_setup(struct nft_ctx *ctx, 5634 const struct nft_set_ext_tmpl *tmpl, 5635 const struct nft_set_ext *ext, 5636 struct nft_expr *expr_array[], 5637 u32 num_exprs) 5638 { 5639 struct nft_set_elem_expr *elem_expr = nft_set_ext_expr(ext); 5640 u32 len = sizeof(struct nft_set_elem_expr); 5641 struct nft_expr *expr; 5642 int i, err; 5643 5644 if (num_exprs == 0) 5645 return 0; 5646 5647 for (i = 0; i < num_exprs; i++) 5648 len += expr_array[i]->ops->size; 5649 5650 if (nft_set_ext_check(tmpl, NFT_SET_EXT_EXPRESSIONS, len) < 0) 5651 return -EINVAL; 5652 5653 for (i = 0; i < num_exprs; i++) { 5654 expr = nft_setelem_expr_at(elem_expr, elem_expr->size); 5655 err = nft_expr_clone(expr, expr_array[i]); 5656 if (err < 0) 5657 goto err_elem_expr_setup; 5658 5659 elem_expr->size += expr_array[i]->ops->size; 5660 nft_expr_destroy(ctx, expr_array[i]); 5661 expr_array[i] = NULL; 5662 } 5663 5664 return 0; 5665 5666 err_elem_expr_setup: 5667 for (; i < num_exprs; i++) { 5668 nft_expr_destroy(ctx, expr_array[i]); 5669 expr_array[i] = NULL; 5670 } 5671 5672 return -ENOMEM; 5673 } 5674 5675 struct nft_set_ext *nft_set_catchall_lookup(const struct net *net, 5676 const struct nft_set *set) 5677 { 5678 struct nft_set_elem_catchall *catchall; 5679 u8 genmask = nft_genmask_cur(net); 5680 struct nft_set_ext *ext; 5681 5682 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 5683 ext = nft_set_elem_ext(set, catchall->elem); 5684 if (nft_set_elem_active(ext, genmask) && 5685 !nft_set_elem_expired(ext)) 5686 return ext; 5687 } 5688 5689 return NULL; 5690 } 5691 EXPORT_SYMBOL_GPL(nft_set_catchall_lookup); 5692 5693 void *nft_set_catchall_gc(const struct nft_set *set) 5694 { 5695 struct nft_set_elem_catchall *catchall, *next; 5696 struct nft_set_ext *ext; 5697 void *elem = NULL; 5698 5699 list_for_each_entry_safe(catchall, next, &set->catchall_list, list) { 5700 ext = nft_set_elem_ext(set, catchall->elem); 5701 5702 if (!nft_set_elem_expired(ext) || 5703 nft_set_elem_mark_busy(ext)) 5704 continue; 5705 5706 elem = catchall->elem; 5707 list_del_rcu(&catchall->list); 5708 kfree_rcu(catchall, rcu); 5709 break; 5710 } 5711 5712 return elem; 5713 } 5714 EXPORT_SYMBOL_GPL(nft_set_catchall_gc); 5715 5716 static int nft_setelem_catchall_insert(const struct net *net, 5717 struct nft_set *set, 5718 const struct nft_set_elem *elem, 5719 struct nft_set_ext **pext) 5720 { 5721 struct nft_set_elem_catchall *catchall; 5722 u8 genmask = nft_genmask_next(net); 5723 struct nft_set_ext *ext; 5724 5725 list_for_each_entry(catchall, &set->catchall_list, list) { 5726 ext = nft_set_elem_ext(set, catchall->elem); 5727 if (nft_set_elem_active(ext, genmask)) { 5728 *pext = ext; 5729 return -EEXIST; 5730 } 5731 } 5732 5733 catchall = kmalloc(sizeof(*catchall), GFP_KERNEL); 5734 if (!catchall) 5735 return -ENOMEM; 5736 5737 catchall->elem = elem->priv; 5738 list_add_tail_rcu(&catchall->list, &set->catchall_list); 5739 5740 return 0; 5741 } 5742 5743 static int nft_setelem_insert(const struct net *net, 5744 struct nft_set *set, 5745 const struct nft_set_elem *elem, 5746 struct nft_set_ext **ext, unsigned int flags) 5747 { 5748 int ret; 5749 5750 if (flags & NFT_SET_ELEM_CATCHALL) 5751 ret = nft_setelem_catchall_insert(net, set, elem, ext); 5752 else 5753 ret = set->ops->insert(net, set, elem, ext); 5754 5755 return ret; 5756 } 5757 5758 static bool nft_setelem_is_catchall(const struct nft_set *set, 5759 const struct nft_set_elem *elem) 5760 { 5761 struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 5762 5763 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) && 5764 *nft_set_ext_flags(ext) & NFT_SET_ELEM_CATCHALL) 5765 return true; 5766 5767 return false; 5768 } 5769 5770 static void nft_setelem_activate(struct net *net, struct nft_set *set, 5771 struct nft_set_elem *elem) 5772 { 5773 struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 5774 5775 if (nft_setelem_is_catchall(set, elem)) { 5776 nft_set_elem_change_active(net, set, ext); 5777 nft_set_elem_clear_busy(ext); 5778 } else { 5779 set->ops->activate(net, set, elem); 5780 } 5781 } 5782 5783 static int nft_setelem_catchall_deactivate(const struct net *net, 5784 struct nft_set *set, 5785 struct nft_set_elem *elem) 5786 { 5787 struct nft_set_elem_catchall *catchall; 5788 struct nft_set_ext *ext; 5789 5790 list_for_each_entry(catchall, &set->catchall_list, list) { 5791 ext = nft_set_elem_ext(set, catchall->elem); 5792 if (!nft_is_active(net, ext) || 5793 nft_set_elem_mark_busy(ext)) 5794 continue; 5795 5796 kfree(elem->priv); 5797 elem->priv = catchall->elem; 5798 nft_set_elem_change_active(net, set, ext); 5799 return 0; 5800 } 5801 5802 return -ENOENT; 5803 } 5804 5805 static int __nft_setelem_deactivate(const struct net *net, 5806 struct nft_set *set, 5807 struct nft_set_elem *elem) 5808 { 5809 void *priv; 5810 5811 priv = set->ops->deactivate(net, set, elem); 5812 if (!priv) 5813 return -ENOENT; 5814 5815 kfree(elem->priv); 5816 elem->priv = priv; 5817 set->ndeact++; 5818 5819 return 0; 5820 } 5821 5822 static int nft_setelem_deactivate(const struct net *net, 5823 struct nft_set *set, 5824 struct nft_set_elem *elem, u32 flags) 5825 { 5826 int ret; 5827 5828 if (flags & NFT_SET_ELEM_CATCHALL) 5829 ret = nft_setelem_catchall_deactivate(net, set, elem); 5830 else 5831 ret = __nft_setelem_deactivate(net, set, elem); 5832 5833 return ret; 5834 } 5835 5836 static void nft_setelem_catchall_remove(const struct net *net, 5837 const struct nft_set *set, 5838 const struct nft_set_elem *elem) 5839 { 5840 struct nft_set_elem_catchall *catchall, *next; 5841 5842 list_for_each_entry_safe(catchall, next, &set->catchall_list, list) { 5843 if (catchall->elem == elem->priv) { 5844 list_del_rcu(&catchall->list); 5845 kfree_rcu(catchall, rcu); 5846 break; 5847 } 5848 } 5849 } 5850 5851 static void nft_setelem_remove(const struct net *net, 5852 const struct nft_set *set, 5853 const struct nft_set_elem *elem) 5854 { 5855 if (nft_setelem_is_catchall(set, elem)) 5856 nft_setelem_catchall_remove(net, set, elem); 5857 else 5858 set->ops->remove(net, set, elem); 5859 } 5860 5861 static bool nft_setelem_valid_key_end(const struct nft_set *set, 5862 struct nlattr **nla, u32 flags) 5863 { 5864 if ((set->flags & (NFT_SET_CONCAT | NFT_SET_INTERVAL)) == 5865 (NFT_SET_CONCAT | NFT_SET_INTERVAL)) { 5866 if (flags & NFT_SET_ELEM_INTERVAL_END) 5867 return false; 5868 5869 if (nla[NFTA_SET_ELEM_KEY_END] && 5870 flags & NFT_SET_ELEM_CATCHALL) 5871 return false; 5872 } else { 5873 if (nla[NFTA_SET_ELEM_KEY_END]) 5874 return false; 5875 } 5876 5877 return true; 5878 } 5879 5880 static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set, 5881 const struct nlattr *attr, u32 nlmsg_flags) 5882 { 5883 struct nft_expr *expr_array[NFT_SET_EXPR_MAX] = {}; 5884 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1]; 5885 u8 genmask = nft_genmask_next(ctx->net); 5886 u32 flags = 0, size = 0, num_exprs = 0; 5887 struct nft_set_ext_tmpl tmpl; 5888 struct nft_set_ext *ext, *ext2; 5889 struct nft_set_elem elem; 5890 struct nft_set_binding *binding; 5891 struct nft_object *obj = NULL; 5892 struct nft_userdata *udata; 5893 struct nft_data_desc desc; 5894 enum nft_registers dreg; 5895 struct nft_trans *trans; 5896 u64 timeout; 5897 u64 expiration; 5898 int err, i; 5899 u8 ulen; 5900 5901 err = nla_parse_nested_deprecated(nla, NFTA_SET_ELEM_MAX, attr, 5902 nft_set_elem_policy, NULL); 5903 if (err < 0) 5904 return err; 5905 5906 nft_set_ext_prepare(&tmpl); 5907 5908 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags); 5909 if (err < 0) 5910 return err; 5911 5912 if (!nla[NFTA_SET_ELEM_KEY] && !(flags & NFT_SET_ELEM_CATCHALL)) 5913 return -EINVAL; 5914 5915 if (flags != 0) { 5916 err = nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS); 5917 if (err < 0) 5918 return err; 5919 } 5920 5921 if (set->flags & NFT_SET_MAP) { 5922 if (nla[NFTA_SET_ELEM_DATA] == NULL && 5923 !(flags & NFT_SET_ELEM_INTERVAL_END)) 5924 return -EINVAL; 5925 } else { 5926 if (nla[NFTA_SET_ELEM_DATA] != NULL) 5927 return -EINVAL; 5928 } 5929 5930 if (set->flags & NFT_SET_OBJECT) { 5931 if (!nla[NFTA_SET_ELEM_OBJREF] && 5932 !(flags & NFT_SET_ELEM_INTERVAL_END)) 5933 return -EINVAL; 5934 } else { 5935 if (nla[NFTA_SET_ELEM_OBJREF]) 5936 return -EINVAL; 5937 } 5938 5939 if (!nft_setelem_valid_key_end(set, nla, flags)) 5940 return -EINVAL; 5941 5942 if ((flags & NFT_SET_ELEM_INTERVAL_END) && 5943 (nla[NFTA_SET_ELEM_DATA] || 5944 nla[NFTA_SET_ELEM_OBJREF] || 5945 nla[NFTA_SET_ELEM_TIMEOUT] || 5946 nla[NFTA_SET_ELEM_EXPIRATION] || 5947 nla[NFTA_SET_ELEM_USERDATA] || 5948 nla[NFTA_SET_ELEM_EXPR] || 5949 nla[NFTA_SET_ELEM_KEY_END] || 5950 nla[NFTA_SET_ELEM_EXPRESSIONS])) 5951 return -EINVAL; 5952 5953 timeout = 0; 5954 if (nla[NFTA_SET_ELEM_TIMEOUT] != NULL) { 5955 if (!(set->flags & NFT_SET_TIMEOUT)) 5956 return -EINVAL; 5957 err = nf_msecs_to_jiffies64(nla[NFTA_SET_ELEM_TIMEOUT], 5958 &timeout); 5959 if (err) 5960 return err; 5961 } else if (set->flags & NFT_SET_TIMEOUT) { 5962 timeout = set->timeout; 5963 } 5964 5965 expiration = 0; 5966 if (nla[NFTA_SET_ELEM_EXPIRATION] != NULL) { 5967 if (!(set->flags & NFT_SET_TIMEOUT)) 5968 return -EINVAL; 5969 err = nf_msecs_to_jiffies64(nla[NFTA_SET_ELEM_EXPIRATION], 5970 &expiration); 5971 if (err) 5972 return err; 5973 } 5974 5975 if (nla[NFTA_SET_ELEM_EXPR]) { 5976 struct nft_expr *expr; 5977 5978 if (set->num_exprs && set->num_exprs != 1) 5979 return -EOPNOTSUPP; 5980 5981 expr = nft_set_elem_expr_alloc(ctx, set, 5982 nla[NFTA_SET_ELEM_EXPR]); 5983 if (IS_ERR(expr)) 5984 return PTR_ERR(expr); 5985 5986 expr_array[0] = expr; 5987 num_exprs = 1; 5988 5989 if (set->num_exprs && set->exprs[0]->ops != expr->ops) { 5990 err = -EOPNOTSUPP; 5991 goto err_set_elem_expr; 5992 } 5993 } else if (nla[NFTA_SET_ELEM_EXPRESSIONS]) { 5994 struct nft_expr *expr; 5995 struct nlattr *tmp; 5996 int left; 5997 5998 i = 0; 5999 nla_for_each_nested(tmp, nla[NFTA_SET_ELEM_EXPRESSIONS], left) { 6000 if (i == NFT_SET_EXPR_MAX || 6001 (set->num_exprs && set->num_exprs == i)) { 6002 err = -E2BIG; 6003 goto err_set_elem_expr; 6004 } 6005 if (nla_type(tmp) != NFTA_LIST_ELEM) { 6006 err = -EINVAL; 6007 goto err_set_elem_expr; 6008 } 6009 expr = nft_set_elem_expr_alloc(ctx, set, tmp); 6010 if (IS_ERR(expr)) { 6011 err = PTR_ERR(expr); 6012 goto err_set_elem_expr; 6013 } 6014 expr_array[i] = expr; 6015 num_exprs++; 6016 6017 if (set->num_exprs && expr->ops != set->exprs[i]->ops) { 6018 err = -EOPNOTSUPP; 6019 goto err_set_elem_expr; 6020 } 6021 i++; 6022 } 6023 if (set->num_exprs && set->num_exprs != i) { 6024 err = -EOPNOTSUPP; 6025 goto err_set_elem_expr; 6026 } 6027 } else if (set->num_exprs > 0) { 6028 err = nft_set_elem_expr_clone(ctx, set, expr_array); 6029 if (err < 0) 6030 goto err_set_elem_expr_clone; 6031 6032 num_exprs = set->num_exprs; 6033 } 6034 6035 if (nla[NFTA_SET_ELEM_KEY]) { 6036 err = nft_setelem_parse_key(ctx, set, &elem.key.val, 6037 nla[NFTA_SET_ELEM_KEY]); 6038 if (err < 0) 6039 goto err_set_elem_expr; 6040 6041 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, set->klen); 6042 if (err < 0) 6043 goto err_parse_key; 6044 } 6045 6046 if (nla[NFTA_SET_ELEM_KEY_END]) { 6047 err = nft_setelem_parse_key(ctx, set, &elem.key_end.val, 6048 nla[NFTA_SET_ELEM_KEY_END]); 6049 if (err < 0) 6050 goto err_parse_key; 6051 6052 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY_END, set->klen); 6053 if (err < 0) 6054 goto err_parse_key_end; 6055 } 6056 6057 if (timeout > 0) { 6058 err = nft_set_ext_add(&tmpl, NFT_SET_EXT_EXPIRATION); 6059 if (err < 0) 6060 goto err_parse_key_end; 6061 6062 if (timeout != set->timeout) { 6063 err = nft_set_ext_add(&tmpl, NFT_SET_EXT_TIMEOUT); 6064 if (err < 0) 6065 goto err_parse_key_end; 6066 } 6067 } 6068 6069 if (num_exprs) { 6070 for (i = 0; i < num_exprs; i++) 6071 size += expr_array[i]->ops->size; 6072 6073 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_EXPRESSIONS, 6074 sizeof(struct nft_set_elem_expr) + size); 6075 if (err < 0) 6076 goto err_parse_key_end; 6077 } 6078 6079 if (nla[NFTA_SET_ELEM_OBJREF] != NULL) { 6080 obj = nft_obj_lookup(ctx->net, ctx->table, 6081 nla[NFTA_SET_ELEM_OBJREF], 6082 set->objtype, genmask); 6083 if (IS_ERR(obj)) { 6084 err = PTR_ERR(obj); 6085 goto err_parse_key_end; 6086 } 6087 err = nft_set_ext_add(&tmpl, NFT_SET_EXT_OBJREF); 6088 if (err < 0) 6089 goto err_parse_key_end; 6090 } 6091 6092 if (nla[NFTA_SET_ELEM_DATA] != NULL) { 6093 err = nft_setelem_parse_data(ctx, set, &desc, &elem.data.val, 6094 nla[NFTA_SET_ELEM_DATA]); 6095 if (err < 0) 6096 goto err_parse_key_end; 6097 6098 dreg = nft_type_to_reg(set->dtype); 6099 list_for_each_entry(binding, &set->bindings, list) { 6100 struct nft_ctx bind_ctx = { 6101 .net = ctx->net, 6102 .family = ctx->family, 6103 .table = ctx->table, 6104 .chain = (struct nft_chain *)binding->chain, 6105 }; 6106 6107 if (!(binding->flags & NFT_SET_MAP)) 6108 continue; 6109 6110 err = nft_validate_register_store(&bind_ctx, dreg, 6111 &elem.data.val, 6112 desc.type, desc.len); 6113 if (err < 0) 6114 goto err_parse_data; 6115 6116 if (desc.type == NFT_DATA_VERDICT && 6117 (elem.data.val.verdict.code == NFT_GOTO || 6118 elem.data.val.verdict.code == NFT_JUMP)) 6119 nft_validate_state_update(ctx->net, 6120 NFT_VALIDATE_NEED); 6121 } 6122 6123 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_DATA, desc.len); 6124 if (err < 0) 6125 goto err_parse_data; 6126 } 6127 6128 /* The full maximum length of userdata can exceed the maximum 6129 * offset value (U8_MAX) for following extensions, therefor it 6130 * must be the last extension added. 6131 */ 6132 ulen = 0; 6133 if (nla[NFTA_SET_ELEM_USERDATA] != NULL) { 6134 ulen = nla_len(nla[NFTA_SET_ELEM_USERDATA]); 6135 if (ulen > 0) { 6136 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_USERDATA, 6137 ulen); 6138 if (err < 0) 6139 goto err_parse_data; 6140 } 6141 } 6142 6143 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, 6144 elem.key_end.val.data, elem.data.val.data, 6145 timeout, expiration, GFP_KERNEL_ACCOUNT); 6146 if (IS_ERR(elem.priv)) { 6147 err = PTR_ERR(elem.priv); 6148 goto err_parse_data; 6149 } 6150 6151 ext = nft_set_elem_ext(set, elem.priv); 6152 if (flags) 6153 *nft_set_ext_flags(ext) = flags; 6154 6155 if (ulen > 0) { 6156 if (nft_set_ext_check(&tmpl, NFT_SET_EXT_USERDATA, ulen) < 0) { 6157 err = -EINVAL; 6158 goto err_elem_userdata; 6159 } 6160 udata = nft_set_ext_userdata(ext); 6161 udata->len = ulen - 1; 6162 nla_memcpy(&udata->data, nla[NFTA_SET_ELEM_USERDATA], ulen); 6163 } 6164 if (obj) { 6165 *nft_set_ext_obj(ext) = obj; 6166 obj->use++; 6167 } 6168 err = nft_set_elem_expr_setup(ctx, &tmpl, ext, expr_array, num_exprs); 6169 if (err < 0) 6170 goto err_elem_free; 6171 6172 trans = nft_trans_elem_alloc(ctx, NFT_MSG_NEWSETELEM, set); 6173 if (trans == NULL) { 6174 err = -ENOMEM; 6175 goto err_elem_free; 6176 } 6177 6178 ext->genmask = nft_genmask_cur(ctx->net) | NFT_SET_ELEM_BUSY_MASK; 6179 6180 err = nft_setelem_insert(ctx->net, set, &elem, &ext2, flags); 6181 if (err) { 6182 if (err == -EEXIST) { 6183 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) ^ 6184 nft_set_ext_exists(ext2, NFT_SET_EXT_DATA) || 6185 nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) ^ 6186 nft_set_ext_exists(ext2, NFT_SET_EXT_OBJREF)) 6187 goto err_element_clash; 6188 if ((nft_set_ext_exists(ext, NFT_SET_EXT_DATA) && 6189 nft_set_ext_exists(ext2, NFT_SET_EXT_DATA) && 6190 memcmp(nft_set_ext_data(ext), 6191 nft_set_ext_data(ext2), set->dlen) != 0) || 6192 (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) && 6193 nft_set_ext_exists(ext2, NFT_SET_EXT_OBJREF) && 6194 *nft_set_ext_obj(ext) != *nft_set_ext_obj(ext2))) 6195 goto err_element_clash; 6196 else if (!(nlmsg_flags & NLM_F_EXCL)) 6197 err = 0; 6198 } else if (err == -ENOTEMPTY) { 6199 /* ENOTEMPTY reports overlapping between this element 6200 * and an existing one. 6201 */ 6202 err = -EEXIST; 6203 } 6204 goto err_element_clash; 6205 } 6206 6207 if (!(flags & NFT_SET_ELEM_CATCHALL) && set->size && 6208 !atomic_add_unless(&set->nelems, 1, set->size + set->ndeact)) { 6209 err = -ENFILE; 6210 goto err_set_full; 6211 } 6212 6213 nft_trans_elem(trans) = elem; 6214 nft_trans_commit_list_add_tail(ctx->net, trans); 6215 return 0; 6216 6217 err_set_full: 6218 nft_setelem_remove(ctx->net, set, &elem); 6219 err_element_clash: 6220 kfree(trans); 6221 err_elem_free: 6222 if (obj) 6223 obj->use--; 6224 err_elem_userdata: 6225 nf_tables_set_elem_destroy(ctx, set, elem.priv); 6226 err_parse_data: 6227 if (nla[NFTA_SET_ELEM_DATA] != NULL) 6228 nft_data_release(&elem.data.val, desc.type); 6229 err_parse_key_end: 6230 nft_data_release(&elem.key_end.val, NFT_DATA_VALUE); 6231 err_parse_key: 6232 nft_data_release(&elem.key.val, NFT_DATA_VALUE); 6233 err_set_elem_expr: 6234 for (i = 0; i < num_exprs && expr_array[i]; i++) 6235 nft_expr_destroy(ctx, expr_array[i]); 6236 err_set_elem_expr_clone: 6237 return err; 6238 } 6239 6240 static int nf_tables_newsetelem(struct sk_buff *skb, 6241 const struct nfnl_info *info, 6242 const struct nlattr * const nla[]) 6243 { 6244 struct nftables_pernet *nft_net = nft_pernet(info->net); 6245 struct netlink_ext_ack *extack = info->extack; 6246 u8 genmask = nft_genmask_next(info->net); 6247 u8 family = info->nfmsg->nfgen_family; 6248 struct net *net = info->net; 6249 const struct nlattr *attr; 6250 struct nft_table *table; 6251 struct nft_set *set; 6252 struct nft_ctx ctx; 6253 int rem, err; 6254 6255 if (nla[NFTA_SET_ELEM_LIST_ELEMENTS] == NULL) 6256 return -EINVAL; 6257 6258 table = nft_table_lookup(net, nla[NFTA_SET_ELEM_LIST_TABLE], family, 6259 genmask, NETLINK_CB(skb).portid); 6260 if (IS_ERR(table)) { 6261 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_ELEM_LIST_TABLE]); 6262 return PTR_ERR(table); 6263 } 6264 6265 set = nft_set_lookup_global(net, table, nla[NFTA_SET_ELEM_LIST_SET], 6266 nla[NFTA_SET_ELEM_LIST_SET_ID], genmask); 6267 if (IS_ERR(set)) 6268 return PTR_ERR(set); 6269 6270 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT) 6271 return -EBUSY; 6272 6273 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 6274 6275 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) { 6276 err = nft_add_set_elem(&ctx, set, attr, info->nlh->nlmsg_flags); 6277 if (err < 0) { 6278 NL_SET_BAD_ATTR(extack, attr); 6279 return err; 6280 } 6281 } 6282 6283 if (nft_net->validate_state == NFT_VALIDATE_DO) 6284 return nft_table_validate(net, table); 6285 6286 return 0; 6287 } 6288 6289 /** 6290 * nft_data_hold - hold a nft_data item 6291 * 6292 * @data: struct nft_data to release 6293 * @type: type of data 6294 * 6295 * Hold a nft_data item. NFT_DATA_VALUE types can be silently discarded, 6296 * NFT_DATA_VERDICT bumps the reference to chains in case of NFT_JUMP and 6297 * NFT_GOTO verdicts. This function must be called on active data objects 6298 * from the second phase of the commit protocol. 6299 */ 6300 void nft_data_hold(const struct nft_data *data, enum nft_data_types type) 6301 { 6302 struct nft_chain *chain; 6303 struct nft_rule *rule; 6304 6305 if (type == NFT_DATA_VERDICT) { 6306 switch (data->verdict.code) { 6307 case NFT_JUMP: 6308 case NFT_GOTO: 6309 chain = data->verdict.chain; 6310 chain->use++; 6311 6312 if (!nft_chain_is_bound(chain)) 6313 break; 6314 6315 chain->table->use++; 6316 list_for_each_entry(rule, &chain->rules, list) 6317 chain->use++; 6318 6319 nft_chain_add(chain->table, chain); 6320 break; 6321 } 6322 } 6323 } 6324 6325 static void nft_setelem_data_activate(const struct net *net, 6326 const struct nft_set *set, 6327 struct nft_set_elem *elem) 6328 { 6329 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 6330 6331 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA)) 6332 nft_data_hold(nft_set_ext_data(ext), set->dtype); 6333 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF)) 6334 (*nft_set_ext_obj(ext))->use++; 6335 } 6336 6337 static void nft_setelem_data_deactivate(const struct net *net, 6338 const struct nft_set *set, 6339 struct nft_set_elem *elem) 6340 { 6341 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 6342 6343 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA)) 6344 nft_data_release(nft_set_ext_data(ext), set->dtype); 6345 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF)) 6346 (*nft_set_ext_obj(ext))->use--; 6347 } 6348 6349 static int nft_del_setelem(struct nft_ctx *ctx, struct nft_set *set, 6350 const struct nlattr *attr) 6351 { 6352 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1]; 6353 struct nft_set_ext_tmpl tmpl; 6354 struct nft_set_elem elem; 6355 struct nft_set_ext *ext; 6356 struct nft_trans *trans; 6357 u32 flags = 0; 6358 int err; 6359 6360 err = nla_parse_nested_deprecated(nla, NFTA_SET_ELEM_MAX, attr, 6361 nft_set_elem_policy, NULL); 6362 if (err < 0) 6363 return err; 6364 6365 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags); 6366 if (err < 0) 6367 return err; 6368 6369 if (!nla[NFTA_SET_ELEM_KEY] && !(flags & NFT_SET_ELEM_CATCHALL)) 6370 return -EINVAL; 6371 6372 if (!nft_setelem_valid_key_end(set, nla, flags)) 6373 return -EINVAL; 6374 6375 nft_set_ext_prepare(&tmpl); 6376 6377 if (flags != 0) { 6378 err = nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS); 6379 if (err < 0) 6380 return err; 6381 } 6382 6383 if (nla[NFTA_SET_ELEM_KEY]) { 6384 err = nft_setelem_parse_key(ctx, set, &elem.key.val, 6385 nla[NFTA_SET_ELEM_KEY]); 6386 if (err < 0) 6387 return err; 6388 6389 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, set->klen); 6390 if (err < 0) 6391 goto fail_elem; 6392 } 6393 6394 if (nla[NFTA_SET_ELEM_KEY_END]) { 6395 err = nft_setelem_parse_key(ctx, set, &elem.key_end.val, 6396 nla[NFTA_SET_ELEM_KEY_END]); 6397 if (err < 0) 6398 goto fail_elem; 6399 6400 err = nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY_END, set->klen); 6401 if (err < 0) 6402 goto fail_elem_key_end; 6403 } 6404 6405 err = -ENOMEM; 6406 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, 6407 elem.key_end.val.data, NULL, 0, 0, 6408 GFP_KERNEL_ACCOUNT); 6409 if (IS_ERR(elem.priv)) { 6410 err = PTR_ERR(elem.priv); 6411 goto fail_elem_key_end; 6412 } 6413 6414 ext = nft_set_elem_ext(set, elem.priv); 6415 if (flags) 6416 *nft_set_ext_flags(ext) = flags; 6417 6418 trans = nft_trans_elem_alloc(ctx, NFT_MSG_DELSETELEM, set); 6419 if (trans == NULL) 6420 goto fail_trans; 6421 6422 err = nft_setelem_deactivate(ctx->net, set, &elem, flags); 6423 if (err < 0) 6424 goto fail_ops; 6425 6426 nft_setelem_data_deactivate(ctx->net, set, &elem); 6427 6428 nft_trans_elem(trans) = elem; 6429 nft_trans_commit_list_add_tail(ctx->net, trans); 6430 return 0; 6431 6432 fail_ops: 6433 kfree(trans); 6434 fail_trans: 6435 kfree(elem.priv); 6436 fail_elem_key_end: 6437 nft_data_release(&elem.key_end.val, NFT_DATA_VALUE); 6438 fail_elem: 6439 nft_data_release(&elem.key.val, NFT_DATA_VALUE); 6440 return err; 6441 } 6442 6443 static int nft_setelem_flush(const struct nft_ctx *ctx, 6444 struct nft_set *set, 6445 const struct nft_set_iter *iter, 6446 struct nft_set_elem *elem) 6447 { 6448 struct nft_trans *trans; 6449 int err; 6450 6451 trans = nft_trans_alloc_gfp(ctx, NFT_MSG_DELSETELEM, 6452 sizeof(struct nft_trans_elem), GFP_ATOMIC); 6453 if (!trans) 6454 return -ENOMEM; 6455 6456 if (!set->ops->flush(ctx->net, set, elem->priv)) { 6457 err = -ENOENT; 6458 goto err1; 6459 } 6460 set->ndeact++; 6461 6462 nft_setelem_data_deactivate(ctx->net, set, elem); 6463 nft_trans_elem_set(trans) = set; 6464 nft_trans_elem(trans) = *elem; 6465 nft_trans_commit_list_add_tail(ctx->net, trans); 6466 6467 return 0; 6468 err1: 6469 kfree(trans); 6470 return err; 6471 } 6472 6473 static int __nft_set_catchall_flush(const struct nft_ctx *ctx, 6474 struct nft_set *set, 6475 struct nft_set_elem *elem) 6476 { 6477 struct nft_trans *trans; 6478 6479 trans = nft_trans_alloc_gfp(ctx, NFT_MSG_DELSETELEM, 6480 sizeof(struct nft_trans_elem), GFP_KERNEL); 6481 if (!trans) 6482 return -ENOMEM; 6483 6484 nft_setelem_data_deactivate(ctx->net, set, elem); 6485 nft_trans_elem_set(trans) = set; 6486 nft_trans_elem(trans) = *elem; 6487 nft_trans_commit_list_add_tail(ctx->net, trans); 6488 6489 return 0; 6490 } 6491 6492 static int nft_set_catchall_flush(const struct nft_ctx *ctx, 6493 struct nft_set *set) 6494 { 6495 u8 genmask = nft_genmask_next(ctx->net); 6496 struct nft_set_elem_catchall *catchall; 6497 struct nft_set_elem elem; 6498 struct nft_set_ext *ext; 6499 int ret = 0; 6500 6501 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 6502 ext = nft_set_elem_ext(set, catchall->elem); 6503 if (!nft_set_elem_active(ext, genmask) || 6504 nft_set_elem_mark_busy(ext)) 6505 continue; 6506 6507 elem.priv = catchall->elem; 6508 ret = __nft_set_catchall_flush(ctx, set, &elem); 6509 if (ret < 0) 6510 break; 6511 } 6512 6513 return ret; 6514 } 6515 6516 static int nft_set_flush(struct nft_ctx *ctx, struct nft_set *set, u8 genmask) 6517 { 6518 struct nft_set_iter iter = { 6519 .genmask = genmask, 6520 .fn = nft_setelem_flush, 6521 }; 6522 6523 set->ops->walk(ctx, set, &iter); 6524 if (!iter.err) 6525 iter.err = nft_set_catchall_flush(ctx, set); 6526 6527 return iter.err; 6528 } 6529 6530 static int nf_tables_delsetelem(struct sk_buff *skb, 6531 const struct nfnl_info *info, 6532 const struct nlattr * const nla[]) 6533 { 6534 struct netlink_ext_ack *extack = info->extack; 6535 u8 genmask = nft_genmask_next(info->net); 6536 u8 family = info->nfmsg->nfgen_family; 6537 struct net *net = info->net; 6538 const struct nlattr *attr; 6539 struct nft_table *table; 6540 struct nft_set *set; 6541 struct nft_ctx ctx; 6542 int rem, err = 0; 6543 6544 table = nft_table_lookup(net, nla[NFTA_SET_ELEM_LIST_TABLE], family, 6545 genmask, NETLINK_CB(skb).portid); 6546 if (IS_ERR(table)) { 6547 NL_SET_BAD_ATTR(extack, nla[NFTA_SET_ELEM_LIST_TABLE]); 6548 return PTR_ERR(table); 6549 } 6550 6551 set = nft_set_lookup(table, nla[NFTA_SET_ELEM_LIST_SET], genmask); 6552 if (IS_ERR(set)) 6553 return PTR_ERR(set); 6554 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT) 6555 return -EBUSY; 6556 6557 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 6558 6559 if (!nla[NFTA_SET_ELEM_LIST_ELEMENTS]) 6560 return nft_set_flush(&ctx, set, genmask); 6561 6562 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) { 6563 err = nft_del_setelem(&ctx, set, attr); 6564 if (err < 0) { 6565 NL_SET_BAD_ATTR(extack, attr); 6566 break; 6567 } 6568 } 6569 return err; 6570 } 6571 6572 void nft_set_gc_batch_release(struct rcu_head *rcu) 6573 { 6574 struct nft_set_gc_batch *gcb; 6575 unsigned int i; 6576 6577 gcb = container_of(rcu, struct nft_set_gc_batch, head.rcu); 6578 for (i = 0; i < gcb->head.cnt; i++) 6579 nft_set_elem_destroy(gcb->head.set, gcb->elems[i], true); 6580 kfree(gcb); 6581 } 6582 6583 struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set, 6584 gfp_t gfp) 6585 { 6586 struct nft_set_gc_batch *gcb; 6587 6588 gcb = kzalloc(sizeof(*gcb), gfp); 6589 if (gcb == NULL) 6590 return gcb; 6591 gcb->head.set = set; 6592 return gcb; 6593 } 6594 6595 /* 6596 * Stateful objects 6597 */ 6598 6599 /** 6600 * nft_register_obj- register nf_tables stateful object type 6601 * @obj_type: object type 6602 * 6603 * Registers the object type for use with nf_tables. Returns zero on 6604 * success or a negative errno code otherwise. 6605 */ 6606 int nft_register_obj(struct nft_object_type *obj_type) 6607 { 6608 if (obj_type->type == NFT_OBJECT_UNSPEC) 6609 return -EINVAL; 6610 6611 nfnl_lock(NFNL_SUBSYS_NFTABLES); 6612 list_add_rcu(&obj_type->list, &nf_tables_objects); 6613 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 6614 return 0; 6615 } 6616 EXPORT_SYMBOL_GPL(nft_register_obj); 6617 6618 /** 6619 * nft_unregister_obj - unregister nf_tables object type 6620 * @obj_type: object type 6621 * 6622 * Unregisters the object type for use with nf_tables. 6623 */ 6624 void nft_unregister_obj(struct nft_object_type *obj_type) 6625 { 6626 nfnl_lock(NFNL_SUBSYS_NFTABLES); 6627 list_del_rcu(&obj_type->list); 6628 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 6629 } 6630 EXPORT_SYMBOL_GPL(nft_unregister_obj); 6631 6632 struct nft_object *nft_obj_lookup(const struct net *net, 6633 const struct nft_table *table, 6634 const struct nlattr *nla, u32 objtype, 6635 u8 genmask) 6636 { 6637 struct nft_object_hash_key k = { .table = table }; 6638 char search[NFT_OBJ_MAXNAMELEN]; 6639 struct rhlist_head *tmp, *list; 6640 struct nft_object *obj; 6641 6642 nla_strscpy(search, nla, sizeof(search)); 6643 k.name = search; 6644 6645 WARN_ON_ONCE(!rcu_read_lock_held() && 6646 !lockdep_commit_lock_is_held(net)); 6647 6648 rcu_read_lock(); 6649 list = rhltable_lookup(&nft_objname_ht, &k, nft_objname_ht_params); 6650 if (!list) 6651 goto out; 6652 6653 rhl_for_each_entry_rcu(obj, tmp, list, rhlhead) { 6654 if (objtype == obj->ops->type->type && 6655 nft_active_genmask(obj, genmask)) { 6656 rcu_read_unlock(); 6657 return obj; 6658 } 6659 } 6660 out: 6661 rcu_read_unlock(); 6662 return ERR_PTR(-ENOENT); 6663 } 6664 EXPORT_SYMBOL_GPL(nft_obj_lookup); 6665 6666 static struct nft_object *nft_obj_lookup_byhandle(const struct nft_table *table, 6667 const struct nlattr *nla, 6668 u32 objtype, u8 genmask) 6669 { 6670 struct nft_object *obj; 6671 6672 list_for_each_entry(obj, &table->objects, list) { 6673 if (be64_to_cpu(nla_get_be64(nla)) == obj->handle && 6674 objtype == obj->ops->type->type && 6675 nft_active_genmask(obj, genmask)) 6676 return obj; 6677 } 6678 return ERR_PTR(-ENOENT); 6679 } 6680 6681 static const struct nla_policy nft_obj_policy[NFTA_OBJ_MAX + 1] = { 6682 [NFTA_OBJ_TABLE] = { .type = NLA_STRING, 6683 .len = NFT_TABLE_MAXNAMELEN - 1 }, 6684 [NFTA_OBJ_NAME] = { .type = NLA_STRING, 6685 .len = NFT_OBJ_MAXNAMELEN - 1 }, 6686 [NFTA_OBJ_TYPE] = { .type = NLA_U32 }, 6687 [NFTA_OBJ_DATA] = { .type = NLA_NESTED }, 6688 [NFTA_OBJ_HANDLE] = { .type = NLA_U64}, 6689 [NFTA_OBJ_USERDATA] = { .type = NLA_BINARY, 6690 .len = NFT_USERDATA_MAXLEN }, 6691 }; 6692 6693 static struct nft_object *nft_obj_init(const struct nft_ctx *ctx, 6694 const struct nft_object_type *type, 6695 const struct nlattr *attr) 6696 { 6697 struct nlattr **tb; 6698 const struct nft_object_ops *ops; 6699 struct nft_object *obj; 6700 int err = -ENOMEM; 6701 6702 tb = kmalloc_array(type->maxattr + 1, sizeof(*tb), GFP_KERNEL); 6703 if (!tb) 6704 goto err1; 6705 6706 if (attr) { 6707 err = nla_parse_nested_deprecated(tb, type->maxattr, attr, 6708 type->policy, NULL); 6709 if (err < 0) 6710 goto err2; 6711 } else { 6712 memset(tb, 0, sizeof(tb[0]) * (type->maxattr + 1)); 6713 } 6714 6715 if (type->select_ops) { 6716 ops = type->select_ops(ctx, (const struct nlattr * const *)tb); 6717 if (IS_ERR(ops)) { 6718 err = PTR_ERR(ops); 6719 goto err2; 6720 } 6721 } else { 6722 ops = type->ops; 6723 } 6724 6725 err = -ENOMEM; 6726 obj = kzalloc(sizeof(*obj) + ops->size, GFP_KERNEL_ACCOUNT); 6727 if (!obj) 6728 goto err2; 6729 6730 err = ops->init(ctx, (const struct nlattr * const *)tb, obj); 6731 if (err < 0) 6732 goto err3; 6733 6734 obj->ops = ops; 6735 6736 kfree(tb); 6737 return obj; 6738 err3: 6739 kfree(obj); 6740 err2: 6741 kfree(tb); 6742 err1: 6743 return ERR_PTR(err); 6744 } 6745 6746 static int nft_object_dump(struct sk_buff *skb, unsigned int attr, 6747 struct nft_object *obj, bool reset) 6748 { 6749 struct nlattr *nest; 6750 6751 nest = nla_nest_start_noflag(skb, attr); 6752 if (!nest) 6753 goto nla_put_failure; 6754 if (obj->ops->dump(skb, obj, reset) < 0) 6755 goto nla_put_failure; 6756 nla_nest_end(skb, nest); 6757 return 0; 6758 6759 nla_put_failure: 6760 return -1; 6761 } 6762 6763 static const struct nft_object_type *__nft_obj_type_get(u32 objtype) 6764 { 6765 const struct nft_object_type *type; 6766 6767 list_for_each_entry(type, &nf_tables_objects, list) { 6768 if (objtype == type->type) 6769 return type; 6770 } 6771 return NULL; 6772 } 6773 6774 static const struct nft_object_type * 6775 nft_obj_type_get(struct net *net, u32 objtype) 6776 { 6777 const struct nft_object_type *type; 6778 6779 type = __nft_obj_type_get(objtype); 6780 if (type != NULL && try_module_get(type->owner)) 6781 return type; 6782 6783 lockdep_nfnl_nft_mutex_not_held(); 6784 #ifdef CONFIG_MODULES 6785 if (type == NULL) { 6786 if (nft_request_module(net, "nft-obj-%u", objtype) == -EAGAIN) 6787 return ERR_PTR(-EAGAIN); 6788 } 6789 #endif 6790 return ERR_PTR(-ENOENT); 6791 } 6792 6793 static int nf_tables_updobj(const struct nft_ctx *ctx, 6794 const struct nft_object_type *type, 6795 const struct nlattr *attr, 6796 struct nft_object *obj) 6797 { 6798 struct nft_object *newobj; 6799 struct nft_trans *trans; 6800 int err = -ENOMEM; 6801 6802 if (!try_module_get(type->owner)) 6803 return -ENOENT; 6804 6805 trans = nft_trans_alloc(ctx, NFT_MSG_NEWOBJ, 6806 sizeof(struct nft_trans_obj)); 6807 if (!trans) 6808 goto err_trans; 6809 6810 newobj = nft_obj_init(ctx, type, attr); 6811 if (IS_ERR(newobj)) { 6812 err = PTR_ERR(newobj); 6813 goto err_free_trans; 6814 } 6815 6816 nft_trans_obj(trans) = obj; 6817 nft_trans_obj_update(trans) = true; 6818 nft_trans_obj_newobj(trans) = newobj; 6819 nft_trans_commit_list_add_tail(ctx->net, trans); 6820 6821 return 0; 6822 6823 err_free_trans: 6824 kfree(trans); 6825 err_trans: 6826 module_put(type->owner); 6827 return err; 6828 } 6829 6830 static int nf_tables_newobj(struct sk_buff *skb, const struct nfnl_info *info, 6831 const struct nlattr * const nla[]) 6832 { 6833 struct netlink_ext_ack *extack = info->extack; 6834 u8 genmask = nft_genmask_next(info->net); 6835 u8 family = info->nfmsg->nfgen_family; 6836 const struct nft_object_type *type; 6837 struct net *net = info->net; 6838 struct nft_table *table; 6839 struct nft_object *obj; 6840 struct nft_ctx ctx; 6841 u32 objtype; 6842 int err; 6843 6844 if (!nla[NFTA_OBJ_TYPE] || 6845 !nla[NFTA_OBJ_NAME] || 6846 !nla[NFTA_OBJ_DATA]) 6847 return -EINVAL; 6848 6849 table = nft_table_lookup(net, nla[NFTA_OBJ_TABLE], family, genmask, 6850 NETLINK_CB(skb).portid); 6851 if (IS_ERR(table)) { 6852 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_TABLE]); 6853 return PTR_ERR(table); 6854 } 6855 6856 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE])); 6857 obj = nft_obj_lookup(net, table, nla[NFTA_OBJ_NAME], objtype, genmask); 6858 if (IS_ERR(obj)) { 6859 err = PTR_ERR(obj); 6860 if (err != -ENOENT) { 6861 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_NAME]); 6862 return err; 6863 } 6864 } else { 6865 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 6866 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_NAME]); 6867 return -EEXIST; 6868 } 6869 if (info->nlh->nlmsg_flags & NLM_F_REPLACE) 6870 return -EOPNOTSUPP; 6871 6872 type = __nft_obj_type_get(objtype); 6873 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 6874 6875 return nf_tables_updobj(&ctx, type, nla[NFTA_OBJ_DATA], obj); 6876 } 6877 6878 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 6879 6880 type = nft_obj_type_get(net, objtype); 6881 if (IS_ERR(type)) 6882 return PTR_ERR(type); 6883 6884 obj = nft_obj_init(&ctx, type, nla[NFTA_OBJ_DATA]); 6885 if (IS_ERR(obj)) { 6886 err = PTR_ERR(obj); 6887 goto err_init; 6888 } 6889 obj->key.table = table; 6890 obj->handle = nf_tables_alloc_handle(table); 6891 6892 obj->key.name = nla_strdup(nla[NFTA_OBJ_NAME], GFP_KERNEL_ACCOUNT); 6893 if (!obj->key.name) { 6894 err = -ENOMEM; 6895 goto err_strdup; 6896 } 6897 6898 if (nla[NFTA_OBJ_USERDATA]) { 6899 obj->udata = nla_memdup(nla[NFTA_OBJ_USERDATA], GFP_KERNEL); 6900 if (obj->udata == NULL) 6901 goto err_userdata; 6902 6903 obj->udlen = nla_len(nla[NFTA_OBJ_USERDATA]); 6904 } 6905 6906 err = nft_trans_obj_add(&ctx, NFT_MSG_NEWOBJ, obj); 6907 if (err < 0) 6908 goto err_trans; 6909 6910 err = rhltable_insert(&nft_objname_ht, &obj->rhlhead, 6911 nft_objname_ht_params); 6912 if (err < 0) 6913 goto err_obj_ht; 6914 6915 list_add_tail_rcu(&obj->list, &table->objects); 6916 table->use++; 6917 return 0; 6918 err_obj_ht: 6919 /* queued in transaction log */ 6920 INIT_LIST_HEAD(&obj->list); 6921 return err; 6922 err_trans: 6923 kfree(obj->udata); 6924 err_userdata: 6925 kfree(obj->key.name); 6926 err_strdup: 6927 if (obj->ops->destroy) 6928 obj->ops->destroy(&ctx, obj); 6929 kfree(obj); 6930 err_init: 6931 module_put(type->owner); 6932 return err; 6933 } 6934 6935 static int nf_tables_fill_obj_info(struct sk_buff *skb, struct net *net, 6936 u32 portid, u32 seq, int event, u32 flags, 6937 int family, const struct nft_table *table, 6938 struct nft_object *obj, bool reset) 6939 { 6940 struct nlmsghdr *nlh; 6941 6942 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 6943 nlh = nfnl_msg_put(skb, portid, seq, event, flags, family, 6944 NFNETLINK_V0, nft_base_seq(net)); 6945 if (!nlh) 6946 goto nla_put_failure; 6947 6948 if (nla_put_string(skb, NFTA_OBJ_TABLE, table->name) || 6949 nla_put_string(skb, NFTA_OBJ_NAME, obj->key.name) || 6950 nla_put_be32(skb, NFTA_OBJ_TYPE, htonl(obj->ops->type->type)) || 6951 nla_put_be32(skb, NFTA_OBJ_USE, htonl(obj->use)) || 6952 nft_object_dump(skb, NFTA_OBJ_DATA, obj, reset) || 6953 nla_put_be64(skb, NFTA_OBJ_HANDLE, cpu_to_be64(obj->handle), 6954 NFTA_OBJ_PAD)) 6955 goto nla_put_failure; 6956 6957 if (obj->udata && 6958 nla_put(skb, NFTA_OBJ_USERDATA, obj->udlen, obj->udata)) 6959 goto nla_put_failure; 6960 6961 nlmsg_end(skb, nlh); 6962 return 0; 6963 6964 nla_put_failure: 6965 nlmsg_trim(skb, nlh); 6966 return -1; 6967 } 6968 6969 struct nft_obj_filter { 6970 char *table; 6971 u32 type; 6972 }; 6973 6974 static int nf_tables_dump_obj(struct sk_buff *skb, struct netlink_callback *cb) 6975 { 6976 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 6977 const struct nft_table *table; 6978 unsigned int idx = 0, s_idx = cb->args[0]; 6979 struct nft_obj_filter *filter = cb->data; 6980 struct net *net = sock_net(skb->sk); 6981 int family = nfmsg->nfgen_family; 6982 struct nftables_pernet *nft_net; 6983 struct nft_object *obj; 6984 bool reset = false; 6985 6986 if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) == NFT_MSG_GETOBJ_RESET) 6987 reset = true; 6988 6989 rcu_read_lock(); 6990 nft_net = nft_pernet(net); 6991 cb->seq = READ_ONCE(nft_net->base_seq); 6992 6993 list_for_each_entry_rcu(table, &nft_net->tables, list) { 6994 if (family != NFPROTO_UNSPEC && family != table->family) 6995 continue; 6996 6997 list_for_each_entry_rcu(obj, &table->objects, list) { 6998 if (!nft_is_active(net, obj)) 6999 goto cont; 7000 if (idx < s_idx) 7001 goto cont; 7002 if (idx > s_idx) 7003 memset(&cb->args[1], 0, 7004 sizeof(cb->args) - sizeof(cb->args[0])); 7005 if (filter && filter->table && 7006 strcmp(filter->table, table->name)) 7007 goto cont; 7008 if (filter && 7009 filter->type != NFT_OBJECT_UNSPEC && 7010 obj->ops->type->type != filter->type) 7011 goto cont; 7012 if (reset) { 7013 char *buf = kasprintf(GFP_ATOMIC, 7014 "%s:%u", 7015 table->name, 7016 nft_net->base_seq); 7017 7018 audit_log_nfcfg(buf, 7019 family, 7020 obj->handle, 7021 AUDIT_NFT_OP_OBJ_RESET, 7022 GFP_ATOMIC); 7023 kfree(buf); 7024 } 7025 7026 if (nf_tables_fill_obj_info(skb, net, NETLINK_CB(cb->skb).portid, 7027 cb->nlh->nlmsg_seq, 7028 NFT_MSG_NEWOBJ, 7029 NLM_F_MULTI | NLM_F_APPEND, 7030 table->family, table, 7031 obj, reset) < 0) 7032 goto done; 7033 7034 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 7035 cont: 7036 idx++; 7037 } 7038 } 7039 done: 7040 rcu_read_unlock(); 7041 7042 cb->args[0] = idx; 7043 return skb->len; 7044 } 7045 7046 static int nf_tables_dump_obj_start(struct netlink_callback *cb) 7047 { 7048 const struct nlattr * const *nla = cb->data; 7049 struct nft_obj_filter *filter = NULL; 7050 7051 if (nla[NFTA_OBJ_TABLE] || nla[NFTA_OBJ_TYPE]) { 7052 filter = kzalloc(sizeof(*filter), GFP_ATOMIC); 7053 if (!filter) 7054 return -ENOMEM; 7055 7056 if (nla[NFTA_OBJ_TABLE]) { 7057 filter->table = nla_strdup(nla[NFTA_OBJ_TABLE], GFP_ATOMIC); 7058 if (!filter->table) { 7059 kfree(filter); 7060 return -ENOMEM; 7061 } 7062 } 7063 7064 if (nla[NFTA_OBJ_TYPE]) 7065 filter->type = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE])); 7066 } 7067 7068 cb->data = filter; 7069 return 0; 7070 } 7071 7072 static int nf_tables_dump_obj_done(struct netlink_callback *cb) 7073 { 7074 struct nft_obj_filter *filter = cb->data; 7075 7076 if (filter) { 7077 kfree(filter->table); 7078 kfree(filter); 7079 } 7080 7081 return 0; 7082 } 7083 7084 /* called with rcu_read_lock held */ 7085 static int nf_tables_getobj(struct sk_buff *skb, const struct nfnl_info *info, 7086 const struct nlattr * const nla[]) 7087 { 7088 struct netlink_ext_ack *extack = info->extack; 7089 u8 genmask = nft_genmask_cur(info->net); 7090 u8 family = info->nfmsg->nfgen_family; 7091 const struct nft_table *table; 7092 struct net *net = info->net; 7093 struct nft_object *obj; 7094 struct sk_buff *skb2; 7095 bool reset = false; 7096 u32 objtype; 7097 int err; 7098 7099 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 7100 struct netlink_dump_control c = { 7101 .start = nf_tables_dump_obj_start, 7102 .dump = nf_tables_dump_obj, 7103 .done = nf_tables_dump_obj_done, 7104 .module = THIS_MODULE, 7105 .data = (void *)nla, 7106 }; 7107 7108 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 7109 } 7110 7111 if (!nla[NFTA_OBJ_NAME] || 7112 !nla[NFTA_OBJ_TYPE]) 7113 return -EINVAL; 7114 7115 table = nft_table_lookup(net, nla[NFTA_OBJ_TABLE], family, genmask, 0); 7116 if (IS_ERR(table)) { 7117 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_TABLE]); 7118 return PTR_ERR(table); 7119 } 7120 7121 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE])); 7122 obj = nft_obj_lookup(net, table, nla[NFTA_OBJ_NAME], objtype, genmask); 7123 if (IS_ERR(obj)) { 7124 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_NAME]); 7125 return PTR_ERR(obj); 7126 } 7127 7128 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 7129 if (!skb2) 7130 return -ENOMEM; 7131 7132 if (NFNL_MSG_TYPE(info->nlh->nlmsg_type) == NFT_MSG_GETOBJ_RESET) 7133 reset = true; 7134 7135 if (reset) { 7136 const struct nftables_pernet *nft_net; 7137 char *buf; 7138 7139 nft_net = nft_pernet(net); 7140 buf = kasprintf(GFP_ATOMIC, "%s:%u", table->name, nft_net->base_seq); 7141 7142 audit_log_nfcfg(buf, 7143 family, 7144 obj->handle, 7145 AUDIT_NFT_OP_OBJ_RESET, 7146 GFP_ATOMIC); 7147 kfree(buf); 7148 } 7149 7150 err = nf_tables_fill_obj_info(skb2, net, NETLINK_CB(skb).portid, 7151 info->nlh->nlmsg_seq, NFT_MSG_NEWOBJ, 0, 7152 family, table, obj, reset); 7153 if (err < 0) 7154 goto err_fill_obj_info; 7155 7156 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 7157 7158 err_fill_obj_info: 7159 kfree_skb(skb2); 7160 return err; 7161 } 7162 7163 static void nft_obj_destroy(const struct nft_ctx *ctx, struct nft_object *obj) 7164 { 7165 if (obj->ops->destroy) 7166 obj->ops->destroy(ctx, obj); 7167 7168 module_put(obj->ops->type->owner); 7169 kfree(obj->key.name); 7170 kfree(obj->udata); 7171 kfree(obj); 7172 } 7173 7174 static int nf_tables_delobj(struct sk_buff *skb, const struct nfnl_info *info, 7175 const struct nlattr * const nla[]) 7176 { 7177 struct netlink_ext_ack *extack = info->extack; 7178 u8 genmask = nft_genmask_next(info->net); 7179 u8 family = info->nfmsg->nfgen_family; 7180 struct net *net = info->net; 7181 const struct nlattr *attr; 7182 struct nft_table *table; 7183 struct nft_object *obj; 7184 struct nft_ctx ctx; 7185 u32 objtype; 7186 7187 if (!nla[NFTA_OBJ_TYPE] || 7188 (!nla[NFTA_OBJ_NAME] && !nla[NFTA_OBJ_HANDLE])) 7189 return -EINVAL; 7190 7191 table = nft_table_lookup(net, nla[NFTA_OBJ_TABLE], family, genmask, 7192 NETLINK_CB(skb).portid); 7193 if (IS_ERR(table)) { 7194 NL_SET_BAD_ATTR(extack, nla[NFTA_OBJ_TABLE]); 7195 return PTR_ERR(table); 7196 } 7197 7198 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE])); 7199 if (nla[NFTA_OBJ_HANDLE]) { 7200 attr = nla[NFTA_OBJ_HANDLE]; 7201 obj = nft_obj_lookup_byhandle(table, attr, objtype, genmask); 7202 } else { 7203 attr = nla[NFTA_OBJ_NAME]; 7204 obj = nft_obj_lookup(net, table, attr, objtype, genmask); 7205 } 7206 7207 if (IS_ERR(obj)) { 7208 NL_SET_BAD_ATTR(extack, attr); 7209 return PTR_ERR(obj); 7210 } 7211 if (obj->use > 0) { 7212 NL_SET_BAD_ATTR(extack, attr); 7213 return -EBUSY; 7214 } 7215 7216 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 7217 7218 return nft_delobj(&ctx, obj); 7219 } 7220 7221 void nft_obj_notify(struct net *net, const struct nft_table *table, 7222 struct nft_object *obj, u32 portid, u32 seq, int event, 7223 u16 flags, int family, int report, gfp_t gfp) 7224 { 7225 struct nftables_pernet *nft_net = nft_pernet(net); 7226 struct sk_buff *skb; 7227 int err; 7228 char *buf = kasprintf(gfp, "%s:%u", 7229 table->name, nft_net->base_seq); 7230 7231 audit_log_nfcfg(buf, 7232 family, 7233 obj->handle, 7234 event == NFT_MSG_NEWOBJ ? 7235 AUDIT_NFT_OP_OBJ_REGISTER : 7236 AUDIT_NFT_OP_OBJ_UNREGISTER, 7237 gfp); 7238 kfree(buf); 7239 7240 if (!report && 7241 !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES)) 7242 return; 7243 7244 skb = nlmsg_new(NLMSG_GOODSIZE, gfp); 7245 if (skb == NULL) 7246 goto err; 7247 7248 err = nf_tables_fill_obj_info(skb, net, portid, seq, event, 7249 flags & (NLM_F_CREATE | NLM_F_EXCL), 7250 family, table, obj, false); 7251 if (err < 0) { 7252 kfree_skb(skb); 7253 goto err; 7254 } 7255 7256 nft_notify_enqueue(skb, report, &nft_net->notify_list); 7257 return; 7258 err: 7259 nfnetlink_set_err(net, portid, NFNLGRP_NFTABLES, -ENOBUFS); 7260 } 7261 EXPORT_SYMBOL_GPL(nft_obj_notify); 7262 7263 static void nf_tables_obj_notify(const struct nft_ctx *ctx, 7264 struct nft_object *obj, int event) 7265 { 7266 nft_obj_notify(ctx->net, ctx->table, obj, ctx->portid, ctx->seq, event, 7267 ctx->flags, ctx->family, ctx->report, GFP_KERNEL); 7268 } 7269 7270 /* 7271 * Flow tables 7272 */ 7273 void nft_register_flowtable_type(struct nf_flowtable_type *type) 7274 { 7275 nfnl_lock(NFNL_SUBSYS_NFTABLES); 7276 list_add_tail_rcu(&type->list, &nf_tables_flowtables); 7277 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 7278 } 7279 EXPORT_SYMBOL_GPL(nft_register_flowtable_type); 7280 7281 void nft_unregister_flowtable_type(struct nf_flowtable_type *type) 7282 { 7283 nfnl_lock(NFNL_SUBSYS_NFTABLES); 7284 list_del_rcu(&type->list); 7285 nfnl_unlock(NFNL_SUBSYS_NFTABLES); 7286 } 7287 EXPORT_SYMBOL_GPL(nft_unregister_flowtable_type); 7288 7289 static const struct nla_policy nft_flowtable_policy[NFTA_FLOWTABLE_MAX + 1] = { 7290 [NFTA_FLOWTABLE_TABLE] = { .type = NLA_STRING, 7291 .len = NFT_NAME_MAXLEN - 1 }, 7292 [NFTA_FLOWTABLE_NAME] = { .type = NLA_STRING, 7293 .len = NFT_NAME_MAXLEN - 1 }, 7294 [NFTA_FLOWTABLE_HOOK] = { .type = NLA_NESTED }, 7295 [NFTA_FLOWTABLE_HANDLE] = { .type = NLA_U64 }, 7296 [NFTA_FLOWTABLE_FLAGS] = { .type = NLA_U32 }, 7297 }; 7298 7299 struct nft_flowtable *nft_flowtable_lookup(const struct nft_table *table, 7300 const struct nlattr *nla, u8 genmask) 7301 { 7302 struct nft_flowtable *flowtable; 7303 7304 list_for_each_entry_rcu(flowtable, &table->flowtables, list) { 7305 if (!nla_strcmp(nla, flowtable->name) && 7306 nft_active_genmask(flowtable, genmask)) 7307 return flowtable; 7308 } 7309 return ERR_PTR(-ENOENT); 7310 } 7311 EXPORT_SYMBOL_GPL(nft_flowtable_lookup); 7312 7313 void nf_tables_deactivate_flowtable(const struct nft_ctx *ctx, 7314 struct nft_flowtable *flowtable, 7315 enum nft_trans_phase phase) 7316 { 7317 switch (phase) { 7318 case NFT_TRANS_PREPARE: 7319 case NFT_TRANS_ABORT: 7320 case NFT_TRANS_RELEASE: 7321 flowtable->use--; 7322 fallthrough; 7323 default: 7324 return; 7325 } 7326 } 7327 EXPORT_SYMBOL_GPL(nf_tables_deactivate_flowtable); 7328 7329 static struct nft_flowtable * 7330 nft_flowtable_lookup_byhandle(const struct nft_table *table, 7331 const struct nlattr *nla, u8 genmask) 7332 { 7333 struct nft_flowtable *flowtable; 7334 7335 list_for_each_entry(flowtable, &table->flowtables, list) { 7336 if (be64_to_cpu(nla_get_be64(nla)) == flowtable->handle && 7337 nft_active_genmask(flowtable, genmask)) 7338 return flowtable; 7339 } 7340 return ERR_PTR(-ENOENT); 7341 } 7342 7343 struct nft_flowtable_hook { 7344 u32 num; 7345 int priority; 7346 struct list_head list; 7347 }; 7348 7349 static const struct nla_policy nft_flowtable_hook_policy[NFTA_FLOWTABLE_HOOK_MAX + 1] = { 7350 [NFTA_FLOWTABLE_HOOK_NUM] = { .type = NLA_U32 }, 7351 [NFTA_FLOWTABLE_HOOK_PRIORITY] = { .type = NLA_U32 }, 7352 [NFTA_FLOWTABLE_HOOK_DEVS] = { .type = NLA_NESTED }, 7353 }; 7354 7355 static int nft_flowtable_parse_hook(const struct nft_ctx *ctx, 7356 const struct nlattr *attr, 7357 struct nft_flowtable_hook *flowtable_hook, 7358 struct nft_flowtable *flowtable, bool add) 7359 { 7360 struct nlattr *tb[NFTA_FLOWTABLE_HOOK_MAX + 1]; 7361 struct nft_hook *hook; 7362 int hooknum, priority; 7363 int err; 7364 7365 INIT_LIST_HEAD(&flowtable_hook->list); 7366 7367 err = nla_parse_nested_deprecated(tb, NFTA_FLOWTABLE_HOOK_MAX, attr, 7368 nft_flowtable_hook_policy, NULL); 7369 if (err < 0) 7370 return err; 7371 7372 if (add) { 7373 if (!tb[NFTA_FLOWTABLE_HOOK_NUM] || 7374 !tb[NFTA_FLOWTABLE_HOOK_PRIORITY]) 7375 return -EINVAL; 7376 7377 hooknum = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_NUM])); 7378 if (hooknum != NF_NETDEV_INGRESS) 7379 return -EOPNOTSUPP; 7380 7381 priority = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_PRIORITY])); 7382 7383 flowtable_hook->priority = priority; 7384 flowtable_hook->num = hooknum; 7385 } else { 7386 if (tb[NFTA_FLOWTABLE_HOOK_NUM]) { 7387 hooknum = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_NUM])); 7388 if (hooknum != flowtable->hooknum) 7389 return -EOPNOTSUPP; 7390 } 7391 7392 if (tb[NFTA_FLOWTABLE_HOOK_PRIORITY]) { 7393 priority = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_PRIORITY])); 7394 if (priority != flowtable->data.priority) 7395 return -EOPNOTSUPP; 7396 } 7397 7398 flowtable_hook->priority = flowtable->data.priority; 7399 flowtable_hook->num = flowtable->hooknum; 7400 } 7401 7402 if (tb[NFTA_FLOWTABLE_HOOK_DEVS]) { 7403 err = nf_tables_parse_netdev_hooks(ctx->net, 7404 tb[NFTA_FLOWTABLE_HOOK_DEVS], 7405 &flowtable_hook->list); 7406 if (err < 0) 7407 return err; 7408 } 7409 7410 list_for_each_entry(hook, &flowtable_hook->list, list) { 7411 hook->ops.pf = NFPROTO_NETDEV; 7412 hook->ops.hooknum = flowtable_hook->num; 7413 hook->ops.priority = flowtable_hook->priority; 7414 hook->ops.priv = &flowtable->data; 7415 hook->ops.hook = flowtable->data.type->hook; 7416 } 7417 7418 return err; 7419 } 7420 7421 static const struct nf_flowtable_type *__nft_flowtable_type_get(u8 family) 7422 { 7423 const struct nf_flowtable_type *type; 7424 7425 list_for_each_entry(type, &nf_tables_flowtables, list) { 7426 if (family == type->family) 7427 return type; 7428 } 7429 return NULL; 7430 } 7431 7432 static const struct nf_flowtable_type * 7433 nft_flowtable_type_get(struct net *net, u8 family) 7434 { 7435 const struct nf_flowtable_type *type; 7436 7437 type = __nft_flowtable_type_get(family); 7438 if (type != NULL && try_module_get(type->owner)) 7439 return type; 7440 7441 lockdep_nfnl_nft_mutex_not_held(); 7442 #ifdef CONFIG_MODULES 7443 if (type == NULL) { 7444 if (nft_request_module(net, "nf-flowtable-%u", family) == -EAGAIN) 7445 return ERR_PTR(-EAGAIN); 7446 } 7447 #endif 7448 return ERR_PTR(-ENOENT); 7449 } 7450 7451 /* Only called from error and netdev event paths. */ 7452 static void nft_unregister_flowtable_hook(struct net *net, 7453 struct nft_flowtable *flowtable, 7454 struct nft_hook *hook) 7455 { 7456 nf_unregister_net_hook(net, &hook->ops); 7457 flowtable->data.type->setup(&flowtable->data, hook->ops.dev, 7458 FLOW_BLOCK_UNBIND); 7459 } 7460 7461 static void __nft_unregister_flowtable_net_hooks(struct net *net, 7462 struct list_head *hook_list, 7463 bool release_netdev) 7464 { 7465 struct nft_hook *hook, *next; 7466 7467 list_for_each_entry_safe(hook, next, hook_list, list) { 7468 nf_unregister_net_hook(net, &hook->ops); 7469 if (release_netdev) { 7470 list_del(&hook->list); 7471 kfree_rcu(hook, rcu); 7472 } 7473 } 7474 } 7475 7476 static void nft_unregister_flowtable_net_hooks(struct net *net, 7477 struct list_head *hook_list) 7478 { 7479 __nft_unregister_flowtable_net_hooks(net, hook_list, false); 7480 } 7481 7482 static int nft_register_flowtable_net_hooks(struct net *net, 7483 struct nft_table *table, 7484 struct list_head *hook_list, 7485 struct nft_flowtable *flowtable) 7486 { 7487 struct nft_hook *hook, *hook2, *next; 7488 struct nft_flowtable *ft; 7489 int err, i = 0; 7490 7491 list_for_each_entry(hook, hook_list, list) { 7492 list_for_each_entry(ft, &table->flowtables, list) { 7493 if (!nft_is_active_next(net, ft)) 7494 continue; 7495 7496 list_for_each_entry(hook2, &ft->hook_list, list) { 7497 if (hook->ops.dev == hook2->ops.dev && 7498 hook->ops.pf == hook2->ops.pf) { 7499 err = -EEXIST; 7500 goto err_unregister_net_hooks; 7501 } 7502 } 7503 } 7504 7505 err = flowtable->data.type->setup(&flowtable->data, 7506 hook->ops.dev, 7507 FLOW_BLOCK_BIND); 7508 if (err < 0) 7509 goto err_unregister_net_hooks; 7510 7511 err = nf_register_net_hook(net, &hook->ops); 7512 if (err < 0) { 7513 flowtable->data.type->setup(&flowtable->data, 7514 hook->ops.dev, 7515 FLOW_BLOCK_UNBIND); 7516 goto err_unregister_net_hooks; 7517 } 7518 7519 i++; 7520 } 7521 7522 return 0; 7523 7524 err_unregister_net_hooks: 7525 list_for_each_entry_safe(hook, next, hook_list, list) { 7526 if (i-- <= 0) 7527 break; 7528 7529 nft_unregister_flowtable_hook(net, flowtable, hook); 7530 list_del_rcu(&hook->list); 7531 kfree_rcu(hook, rcu); 7532 } 7533 7534 return err; 7535 } 7536 7537 static void nft_flowtable_hooks_destroy(struct list_head *hook_list) 7538 { 7539 struct nft_hook *hook, *next; 7540 7541 list_for_each_entry_safe(hook, next, hook_list, list) { 7542 list_del_rcu(&hook->list); 7543 kfree_rcu(hook, rcu); 7544 } 7545 } 7546 7547 static int nft_flowtable_update(struct nft_ctx *ctx, const struct nlmsghdr *nlh, 7548 struct nft_flowtable *flowtable) 7549 { 7550 const struct nlattr * const *nla = ctx->nla; 7551 struct nft_flowtable_hook flowtable_hook; 7552 struct nft_hook *hook, *next; 7553 struct nft_trans *trans; 7554 bool unregister = false; 7555 u32 flags; 7556 int err; 7557 7558 err = nft_flowtable_parse_hook(ctx, nla[NFTA_FLOWTABLE_HOOK], 7559 &flowtable_hook, flowtable, false); 7560 if (err < 0) 7561 return err; 7562 7563 list_for_each_entry_safe(hook, next, &flowtable_hook.list, list) { 7564 if (nft_hook_list_find(&flowtable->hook_list, hook)) { 7565 list_del(&hook->list); 7566 kfree(hook); 7567 } 7568 } 7569 7570 if (nla[NFTA_FLOWTABLE_FLAGS]) { 7571 flags = ntohl(nla_get_be32(nla[NFTA_FLOWTABLE_FLAGS])); 7572 if (flags & ~NFT_FLOWTABLE_MASK) { 7573 err = -EOPNOTSUPP; 7574 goto err_flowtable_update_hook; 7575 } 7576 if ((flowtable->data.flags & NFT_FLOWTABLE_HW_OFFLOAD) ^ 7577 (flags & NFT_FLOWTABLE_HW_OFFLOAD)) { 7578 err = -EOPNOTSUPP; 7579 goto err_flowtable_update_hook; 7580 } 7581 } else { 7582 flags = flowtable->data.flags; 7583 } 7584 7585 err = nft_register_flowtable_net_hooks(ctx->net, ctx->table, 7586 &flowtable_hook.list, flowtable); 7587 if (err < 0) 7588 goto err_flowtable_update_hook; 7589 7590 trans = nft_trans_alloc(ctx, NFT_MSG_NEWFLOWTABLE, 7591 sizeof(struct nft_trans_flowtable)); 7592 if (!trans) { 7593 unregister = true; 7594 err = -ENOMEM; 7595 goto err_flowtable_update_hook; 7596 } 7597 7598 nft_trans_flowtable_flags(trans) = flags; 7599 nft_trans_flowtable(trans) = flowtable; 7600 nft_trans_flowtable_update(trans) = true; 7601 INIT_LIST_HEAD(&nft_trans_flowtable_hooks(trans)); 7602 list_splice(&flowtable_hook.list, &nft_trans_flowtable_hooks(trans)); 7603 7604 nft_trans_commit_list_add_tail(ctx->net, trans); 7605 7606 return 0; 7607 7608 err_flowtable_update_hook: 7609 list_for_each_entry_safe(hook, next, &flowtable_hook.list, list) { 7610 if (unregister) 7611 nft_unregister_flowtable_hook(ctx->net, flowtable, hook); 7612 list_del_rcu(&hook->list); 7613 kfree_rcu(hook, rcu); 7614 } 7615 7616 return err; 7617 7618 } 7619 7620 static int nf_tables_newflowtable(struct sk_buff *skb, 7621 const struct nfnl_info *info, 7622 const struct nlattr * const nla[]) 7623 { 7624 struct netlink_ext_ack *extack = info->extack; 7625 struct nft_flowtable_hook flowtable_hook; 7626 u8 genmask = nft_genmask_next(info->net); 7627 u8 family = info->nfmsg->nfgen_family; 7628 const struct nf_flowtable_type *type; 7629 struct nft_flowtable *flowtable; 7630 struct nft_hook *hook, *next; 7631 struct net *net = info->net; 7632 struct nft_table *table; 7633 struct nft_ctx ctx; 7634 int err; 7635 7636 if (!nla[NFTA_FLOWTABLE_TABLE] || 7637 !nla[NFTA_FLOWTABLE_NAME] || 7638 !nla[NFTA_FLOWTABLE_HOOK]) 7639 return -EINVAL; 7640 7641 table = nft_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE], family, 7642 genmask, NETLINK_CB(skb).portid); 7643 if (IS_ERR(table)) { 7644 NL_SET_BAD_ATTR(extack, nla[NFTA_FLOWTABLE_TABLE]); 7645 return PTR_ERR(table); 7646 } 7647 7648 flowtable = nft_flowtable_lookup(table, nla[NFTA_FLOWTABLE_NAME], 7649 genmask); 7650 if (IS_ERR(flowtable)) { 7651 err = PTR_ERR(flowtable); 7652 if (err != -ENOENT) { 7653 NL_SET_BAD_ATTR(extack, nla[NFTA_FLOWTABLE_NAME]); 7654 return err; 7655 } 7656 } else { 7657 if (info->nlh->nlmsg_flags & NLM_F_EXCL) { 7658 NL_SET_BAD_ATTR(extack, nla[NFTA_FLOWTABLE_NAME]); 7659 return -EEXIST; 7660 } 7661 7662 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 7663 7664 return nft_flowtable_update(&ctx, info->nlh, flowtable); 7665 } 7666 7667 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 7668 7669 flowtable = kzalloc(sizeof(*flowtable), GFP_KERNEL_ACCOUNT); 7670 if (!flowtable) 7671 return -ENOMEM; 7672 7673 flowtable->table = table; 7674 flowtable->handle = nf_tables_alloc_handle(table); 7675 INIT_LIST_HEAD(&flowtable->hook_list); 7676 7677 flowtable->name = nla_strdup(nla[NFTA_FLOWTABLE_NAME], GFP_KERNEL_ACCOUNT); 7678 if (!flowtable->name) { 7679 err = -ENOMEM; 7680 goto err1; 7681 } 7682 7683 type = nft_flowtable_type_get(net, family); 7684 if (IS_ERR(type)) { 7685 err = PTR_ERR(type); 7686 goto err2; 7687 } 7688 7689 if (nla[NFTA_FLOWTABLE_FLAGS]) { 7690 flowtable->data.flags = 7691 ntohl(nla_get_be32(nla[NFTA_FLOWTABLE_FLAGS])); 7692 if (flowtable->data.flags & ~NFT_FLOWTABLE_MASK) { 7693 err = -EOPNOTSUPP; 7694 goto err3; 7695 } 7696 } 7697 7698 write_pnet(&flowtable->data.net, net); 7699 flowtable->data.type = type; 7700 err = type->init(&flowtable->data); 7701 if (err < 0) 7702 goto err3; 7703 7704 err = nft_flowtable_parse_hook(&ctx, nla[NFTA_FLOWTABLE_HOOK], 7705 &flowtable_hook, flowtable, true); 7706 if (err < 0) 7707 goto err4; 7708 7709 list_splice(&flowtable_hook.list, &flowtable->hook_list); 7710 flowtable->data.priority = flowtable_hook.priority; 7711 flowtable->hooknum = flowtable_hook.num; 7712 7713 err = nft_register_flowtable_net_hooks(ctx.net, table, 7714 &flowtable->hook_list, 7715 flowtable); 7716 if (err < 0) { 7717 nft_flowtable_hooks_destroy(&flowtable->hook_list); 7718 goto err4; 7719 } 7720 7721 err = nft_trans_flowtable_add(&ctx, NFT_MSG_NEWFLOWTABLE, flowtable); 7722 if (err < 0) 7723 goto err5; 7724 7725 list_add_tail_rcu(&flowtable->list, &table->flowtables); 7726 table->use++; 7727 7728 return 0; 7729 err5: 7730 list_for_each_entry_safe(hook, next, &flowtable->hook_list, list) { 7731 nft_unregister_flowtable_hook(net, flowtable, hook); 7732 list_del_rcu(&hook->list); 7733 kfree_rcu(hook, rcu); 7734 } 7735 err4: 7736 flowtable->data.type->free(&flowtable->data); 7737 err3: 7738 module_put(type->owner); 7739 err2: 7740 kfree(flowtable->name); 7741 err1: 7742 kfree(flowtable); 7743 return err; 7744 } 7745 7746 static void nft_flowtable_hook_release(struct nft_flowtable_hook *flowtable_hook) 7747 { 7748 struct nft_hook *this, *next; 7749 7750 list_for_each_entry_safe(this, next, &flowtable_hook->list, list) { 7751 list_del(&this->list); 7752 kfree(this); 7753 } 7754 } 7755 7756 static int nft_delflowtable_hook(struct nft_ctx *ctx, 7757 struct nft_flowtable *flowtable) 7758 { 7759 const struct nlattr * const *nla = ctx->nla; 7760 struct nft_flowtable_hook flowtable_hook; 7761 LIST_HEAD(flowtable_del_list); 7762 struct nft_hook *this, *hook; 7763 struct nft_trans *trans; 7764 int err; 7765 7766 err = nft_flowtable_parse_hook(ctx, nla[NFTA_FLOWTABLE_HOOK], 7767 &flowtable_hook, flowtable, false); 7768 if (err < 0) 7769 return err; 7770 7771 list_for_each_entry(this, &flowtable_hook.list, list) { 7772 hook = nft_hook_list_find(&flowtable->hook_list, this); 7773 if (!hook) { 7774 err = -ENOENT; 7775 goto err_flowtable_del_hook; 7776 } 7777 list_move(&hook->list, &flowtable_del_list); 7778 } 7779 7780 trans = nft_trans_alloc(ctx, NFT_MSG_DELFLOWTABLE, 7781 sizeof(struct nft_trans_flowtable)); 7782 if (!trans) { 7783 err = -ENOMEM; 7784 goto err_flowtable_del_hook; 7785 } 7786 7787 nft_trans_flowtable(trans) = flowtable; 7788 nft_trans_flowtable_update(trans) = true; 7789 INIT_LIST_HEAD(&nft_trans_flowtable_hooks(trans)); 7790 list_splice(&flowtable_del_list, &nft_trans_flowtable_hooks(trans)); 7791 nft_flowtable_hook_release(&flowtable_hook); 7792 7793 nft_trans_commit_list_add_tail(ctx->net, trans); 7794 7795 return 0; 7796 7797 err_flowtable_del_hook: 7798 list_splice(&flowtable_del_list, &flowtable->hook_list); 7799 nft_flowtable_hook_release(&flowtable_hook); 7800 7801 return err; 7802 } 7803 7804 static int nf_tables_delflowtable(struct sk_buff *skb, 7805 const struct nfnl_info *info, 7806 const struct nlattr * const nla[]) 7807 { 7808 struct netlink_ext_ack *extack = info->extack; 7809 u8 genmask = nft_genmask_next(info->net); 7810 u8 family = info->nfmsg->nfgen_family; 7811 struct nft_flowtable *flowtable; 7812 struct net *net = info->net; 7813 const struct nlattr *attr; 7814 struct nft_table *table; 7815 struct nft_ctx ctx; 7816 7817 if (!nla[NFTA_FLOWTABLE_TABLE] || 7818 (!nla[NFTA_FLOWTABLE_NAME] && 7819 !nla[NFTA_FLOWTABLE_HANDLE])) 7820 return -EINVAL; 7821 7822 table = nft_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE], family, 7823 genmask, NETLINK_CB(skb).portid); 7824 if (IS_ERR(table)) { 7825 NL_SET_BAD_ATTR(extack, nla[NFTA_FLOWTABLE_TABLE]); 7826 return PTR_ERR(table); 7827 } 7828 7829 if (nla[NFTA_FLOWTABLE_HANDLE]) { 7830 attr = nla[NFTA_FLOWTABLE_HANDLE]; 7831 flowtable = nft_flowtable_lookup_byhandle(table, attr, genmask); 7832 } else { 7833 attr = nla[NFTA_FLOWTABLE_NAME]; 7834 flowtable = nft_flowtable_lookup(table, attr, genmask); 7835 } 7836 7837 if (IS_ERR(flowtable)) { 7838 NL_SET_BAD_ATTR(extack, attr); 7839 return PTR_ERR(flowtable); 7840 } 7841 7842 nft_ctx_init(&ctx, net, skb, info->nlh, family, table, NULL, nla); 7843 7844 if (nla[NFTA_FLOWTABLE_HOOK]) 7845 return nft_delflowtable_hook(&ctx, flowtable); 7846 7847 if (flowtable->use > 0) { 7848 NL_SET_BAD_ATTR(extack, attr); 7849 return -EBUSY; 7850 } 7851 7852 return nft_delflowtable(&ctx, flowtable); 7853 } 7854 7855 static int nf_tables_fill_flowtable_info(struct sk_buff *skb, struct net *net, 7856 u32 portid, u32 seq, int event, 7857 u32 flags, int family, 7858 struct nft_flowtable *flowtable, 7859 struct list_head *hook_list) 7860 { 7861 struct nlattr *nest, *nest_devs; 7862 struct nft_hook *hook; 7863 struct nlmsghdr *nlh; 7864 7865 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event); 7866 nlh = nfnl_msg_put(skb, portid, seq, event, flags, family, 7867 NFNETLINK_V0, nft_base_seq(net)); 7868 if (!nlh) 7869 goto nla_put_failure; 7870 7871 if (nla_put_string(skb, NFTA_FLOWTABLE_TABLE, flowtable->table->name) || 7872 nla_put_string(skb, NFTA_FLOWTABLE_NAME, flowtable->name) || 7873 nla_put_be32(skb, NFTA_FLOWTABLE_USE, htonl(flowtable->use)) || 7874 nla_put_be64(skb, NFTA_FLOWTABLE_HANDLE, cpu_to_be64(flowtable->handle), 7875 NFTA_FLOWTABLE_PAD) || 7876 nla_put_be32(skb, NFTA_FLOWTABLE_FLAGS, htonl(flowtable->data.flags))) 7877 goto nla_put_failure; 7878 7879 nest = nla_nest_start_noflag(skb, NFTA_FLOWTABLE_HOOK); 7880 if (!nest) 7881 goto nla_put_failure; 7882 if (nla_put_be32(skb, NFTA_FLOWTABLE_HOOK_NUM, htonl(flowtable->hooknum)) || 7883 nla_put_be32(skb, NFTA_FLOWTABLE_HOOK_PRIORITY, htonl(flowtable->data.priority))) 7884 goto nla_put_failure; 7885 7886 nest_devs = nla_nest_start_noflag(skb, NFTA_FLOWTABLE_HOOK_DEVS); 7887 if (!nest_devs) 7888 goto nla_put_failure; 7889 7890 list_for_each_entry_rcu(hook, hook_list, list) { 7891 if (nla_put_string(skb, NFTA_DEVICE_NAME, hook->ops.dev->name)) 7892 goto nla_put_failure; 7893 } 7894 nla_nest_end(skb, nest_devs); 7895 nla_nest_end(skb, nest); 7896 7897 nlmsg_end(skb, nlh); 7898 return 0; 7899 7900 nla_put_failure: 7901 nlmsg_trim(skb, nlh); 7902 return -1; 7903 } 7904 7905 struct nft_flowtable_filter { 7906 char *table; 7907 }; 7908 7909 static int nf_tables_dump_flowtable(struct sk_buff *skb, 7910 struct netlink_callback *cb) 7911 { 7912 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 7913 struct nft_flowtable_filter *filter = cb->data; 7914 unsigned int idx = 0, s_idx = cb->args[0]; 7915 struct net *net = sock_net(skb->sk); 7916 int family = nfmsg->nfgen_family; 7917 struct nft_flowtable *flowtable; 7918 struct nftables_pernet *nft_net; 7919 const struct nft_table *table; 7920 7921 rcu_read_lock(); 7922 nft_net = nft_pernet(net); 7923 cb->seq = READ_ONCE(nft_net->base_seq); 7924 7925 list_for_each_entry_rcu(table, &nft_net->tables, list) { 7926 if (family != NFPROTO_UNSPEC && family != table->family) 7927 continue; 7928 7929 list_for_each_entry_rcu(flowtable, &table->flowtables, list) { 7930 if (!nft_is_active(net, flowtable)) 7931 goto cont; 7932 if (idx < s_idx) 7933 goto cont; 7934 if (idx > s_idx) 7935 memset(&cb->args[1], 0, 7936 sizeof(cb->args) - sizeof(cb->args[0])); 7937 if (filter && filter->table && 7938 strcmp(filter->table, table->name)) 7939 goto cont; 7940 7941 if (nf_tables_fill_flowtable_info(skb, net, NETLINK_CB(cb->skb).portid, 7942 cb->nlh->nlmsg_seq, 7943 NFT_MSG_NEWFLOWTABLE, 7944 NLM_F_MULTI | NLM_F_APPEND, 7945 table->family, 7946 flowtable, 7947 &flowtable->hook_list) < 0) 7948 goto done; 7949 7950 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 7951 cont: 7952 idx++; 7953 } 7954 } 7955 done: 7956 rcu_read_unlock(); 7957 7958 cb->args[0] = idx; 7959 return skb->len; 7960 } 7961 7962 static int nf_tables_dump_flowtable_start(struct netlink_callback *cb) 7963 { 7964 const struct nlattr * const *nla = cb->data; 7965 struct nft_flowtable_filter *filter = NULL; 7966 7967 if (nla[NFTA_FLOWTABLE_TABLE]) { 7968 filter = kzalloc(sizeof(*filter), GFP_ATOMIC); 7969 if (!filter) 7970 return -ENOMEM; 7971 7972 filter->table = nla_strdup(nla[NFTA_FLOWTABLE_TABLE], 7973 GFP_ATOMIC); 7974 if (!filter->table) { 7975 kfree(filter); 7976 return -ENOMEM; 7977 } 7978 } 7979 7980 cb->data = filter; 7981 return 0; 7982 } 7983 7984 static int nf_tables_dump_flowtable_done(struct netlink_callback *cb) 7985 { 7986 struct nft_flowtable_filter *filter = cb->data; 7987 7988 if (!filter) 7989 return 0; 7990 7991 kfree(filter->table); 7992 kfree(filter); 7993 7994 return 0; 7995 } 7996 7997 /* called with rcu_read_lock held */ 7998 static int nf_tables_getflowtable(struct sk_buff *skb, 7999 const struct nfnl_info *info, 8000 const struct nlattr * const nla[]) 8001 { 8002 u8 genmask = nft_genmask_cur(info->net); 8003 u8 family = info->nfmsg->nfgen_family; 8004 struct nft_flowtable *flowtable; 8005 const struct nft_table *table; 8006 struct net *net = info->net; 8007 struct sk_buff *skb2; 8008 int err; 8009 8010 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 8011 struct netlink_dump_control c = { 8012 .start = nf_tables_dump_flowtable_start, 8013 .dump = nf_tables_dump_flowtable, 8014 .done = nf_tables_dump_flowtable_done, 8015 .module = THIS_MODULE, 8016 .data = (void *)nla, 8017 }; 8018 8019 return nft_netlink_dump_start_rcu(info->sk, skb, info->nlh, &c); 8020 } 8021 8022 if (!nla[NFTA_FLOWTABLE_NAME]) 8023 return -EINVAL; 8024 8025 table = nft_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE], family, 8026 genmask, 0); 8027 if (IS_ERR(table)) 8028 return PTR_ERR(table); 8029 8030 flowtable = nft_flowtable_lookup(table, nla[NFTA_FLOWTABLE_NAME], 8031 genmask); 8032 if (IS_ERR(flowtable)) 8033 return PTR_ERR(flowtable); 8034 8035 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 8036 if (!skb2) 8037 return -ENOMEM; 8038 8039 err = nf_tables_fill_flowtable_info(skb2, net, NETLINK_CB(skb).portid, 8040 info->nlh->nlmsg_seq, 8041 NFT_MSG_NEWFLOWTABLE, 0, family, 8042 flowtable, &flowtable->hook_list); 8043 if (err < 0) 8044 goto err_fill_flowtable_info; 8045 8046 return nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid); 8047 8048 err_fill_flowtable_info: 8049 kfree_skb(skb2); 8050 return err; 8051 } 8052 8053 static void nf_tables_flowtable_notify(struct nft_ctx *ctx, 8054 struct nft_flowtable *flowtable, 8055 struct list_head *hook_list, 8056 int event) 8057 { 8058 struct nftables_pernet *nft_net = nft_pernet(ctx->net); 8059 struct sk_buff *skb; 8060 u16 flags = 0; 8061 int err; 8062 8063 if (!ctx->report && 8064 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES)) 8065 return; 8066 8067 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 8068 if (skb == NULL) 8069 goto err; 8070 8071 if (ctx->flags & (NLM_F_CREATE | NLM_F_EXCL)) 8072 flags |= ctx->flags & (NLM_F_CREATE | NLM_F_EXCL); 8073 8074 err = nf_tables_fill_flowtable_info(skb, ctx->net, ctx->portid, 8075 ctx->seq, event, flags, 8076 ctx->family, flowtable, hook_list); 8077 if (err < 0) { 8078 kfree_skb(skb); 8079 goto err; 8080 } 8081 8082 nft_notify_enqueue(skb, ctx->report, &nft_net->notify_list); 8083 return; 8084 err: 8085 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS); 8086 } 8087 8088 static void nf_tables_flowtable_destroy(struct nft_flowtable *flowtable) 8089 { 8090 struct nft_hook *hook, *next; 8091 8092 flowtable->data.type->free(&flowtable->data); 8093 list_for_each_entry_safe(hook, next, &flowtable->hook_list, list) { 8094 flowtable->data.type->setup(&flowtable->data, hook->ops.dev, 8095 FLOW_BLOCK_UNBIND); 8096 list_del_rcu(&hook->list); 8097 kfree(hook); 8098 } 8099 kfree(flowtable->name); 8100 module_put(flowtable->data.type->owner); 8101 kfree(flowtable); 8102 } 8103 8104 static int nf_tables_fill_gen_info(struct sk_buff *skb, struct net *net, 8105 u32 portid, u32 seq) 8106 { 8107 struct nftables_pernet *nft_net = nft_pernet(net); 8108 struct nlmsghdr *nlh; 8109 char buf[TASK_COMM_LEN]; 8110 int event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, NFT_MSG_NEWGEN); 8111 8112 nlh = nfnl_msg_put(skb, portid, seq, event, 0, AF_UNSPEC, 8113 NFNETLINK_V0, nft_base_seq(net)); 8114 if (!nlh) 8115 goto nla_put_failure; 8116 8117 if (nla_put_be32(skb, NFTA_GEN_ID, htonl(nft_net->base_seq)) || 8118 nla_put_be32(skb, NFTA_GEN_PROC_PID, htonl(task_pid_nr(current))) || 8119 nla_put_string(skb, NFTA_GEN_PROC_NAME, get_task_comm(buf, current))) 8120 goto nla_put_failure; 8121 8122 nlmsg_end(skb, nlh); 8123 return 0; 8124 8125 nla_put_failure: 8126 nlmsg_trim(skb, nlh); 8127 return -EMSGSIZE; 8128 } 8129 8130 static void nft_flowtable_event(unsigned long event, struct net_device *dev, 8131 struct nft_flowtable *flowtable) 8132 { 8133 struct nft_hook *hook; 8134 8135 list_for_each_entry(hook, &flowtable->hook_list, list) { 8136 if (hook->ops.dev != dev) 8137 continue; 8138 8139 /* flow_offload_netdev_event() cleans up entries for us. */ 8140 nft_unregister_flowtable_hook(dev_net(dev), flowtable, hook); 8141 list_del_rcu(&hook->list); 8142 kfree_rcu(hook, rcu); 8143 break; 8144 } 8145 } 8146 8147 static int nf_tables_flowtable_event(struct notifier_block *this, 8148 unsigned long event, void *ptr) 8149 { 8150 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 8151 struct nft_flowtable *flowtable; 8152 struct nftables_pernet *nft_net; 8153 struct nft_table *table; 8154 struct net *net; 8155 8156 if (event != NETDEV_UNREGISTER) 8157 return 0; 8158 8159 net = dev_net(dev); 8160 nft_net = nft_pernet(net); 8161 mutex_lock(&nft_net->commit_mutex); 8162 list_for_each_entry(table, &nft_net->tables, list) { 8163 list_for_each_entry(flowtable, &table->flowtables, list) { 8164 nft_flowtable_event(event, dev, flowtable); 8165 } 8166 } 8167 mutex_unlock(&nft_net->commit_mutex); 8168 8169 return NOTIFY_DONE; 8170 } 8171 8172 static struct notifier_block nf_tables_flowtable_notifier = { 8173 .notifier_call = nf_tables_flowtable_event, 8174 }; 8175 8176 static void nf_tables_gen_notify(struct net *net, struct sk_buff *skb, 8177 int event) 8178 { 8179 struct nlmsghdr *nlh = nlmsg_hdr(skb); 8180 struct sk_buff *skb2; 8181 int err; 8182 8183 if (!nlmsg_report(nlh) && 8184 !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES)) 8185 return; 8186 8187 skb2 = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 8188 if (skb2 == NULL) 8189 goto err; 8190 8191 err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid, 8192 nlh->nlmsg_seq); 8193 if (err < 0) { 8194 kfree_skb(skb2); 8195 goto err; 8196 } 8197 8198 nfnetlink_send(skb2, net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES, 8199 nlmsg_report(nlh), GFP_KERNEL); 8200 return; 8201 err: 8202 nfnetlink_set_err(net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES, 8203 -ENOBUFS); 8204 } 8205 8206 static int nf_tables_getgen(struct sk_buff *skb, const struct nfnl_info *info, 8207 const struct nlattr * const nla[]) 8208 { 8209 struct sk_buff *skb2; 8210 int err; 8211 8212 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC); 8213 if (skb2 == NULL) 8214 return -ENOMEM; 8215 8216 err = nf_tables_fill_gen_info(skb2, info->net, NETLINK_CB(skb).portid, 8217 info->nlh->nlmsg_seq); 8218 if (err < 0) 8219 goto err_fill_gen_info; 8220 8221 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid); 8222 8223 err_fill_gen_info: 8224 kfree_skb(skb2); 8225 return err; 8226 } 8227 8228 static const struct nfnl_callback nf_tables_cb[NFT_MSG_MAX] = { 8229 [NFT_MSG_NEWTABLE] = { 8230 .call = nf_tables_newtable, 8231 .type = NFNL_CB_BATCH, 8232 .attr_count = NFTA_TABLE_MAX, 8233 .policy = nft_table_policy, 8234 }, 8235 [NFT_MSG_GETTABLE] = { 8236 .call = nf_tables_gettable, 8237 .type = NFNL_CB_RCU, 8238 .attr_count = NFTA_TABLE_MAX, 8239 .policy = nft_table_policy, 8240 }, 8241 [NFT_MSG_DELTABLE] = { 8242 .call = nf_tables_deltable, 8243 .type = NFNL_CB_BATCH, 8244 .attr_count = NFTA_TABLE_MAX, 8245 .policy = nft_table_policy, 8246 }, 8247 [NFT_MSG_NEWCHAIN] = { 8248 .call = nf_tables_newchain, 8249 .type = NFNL_CB_BATCH, 8250 .attr_count = NFTA_CHAIN_MAX, 8251 .policy = nft_chain_policy, 8252 }, 8253 [NFT_MSG_GETCHAIN] = { 8254 .call = nf_tables_getchain, 8255 .type = NFNL_CB_RCU, 8256 .attr_count = NFTA_CHAIN_MAX, 8257 .policy = nft_chain_policy, 8258 }, 8259 [NFT_MSG_DELCHAIN] = { 8260 .call = nf_tables_delchain, 8261 .type = NFNL_CB_BATCH, 8262 .attr_count = NFTA_CHAIN_MAX, 8263 .policy = nft_chain_policy, 8264 }, 8265 [NFT_MSG_NEWRULE] = { 8266 .call = nf_tables_newrule, 8267 .type = NFNL_CB_BATCH, 8268 .attr_count = NFTA_RULE_MAX, 8269 .policy = nft_rule_policy, 8270 }, 8271 [NFT_MSG_GETRULE] = { 8272 .call = nf_tables_getrule, 8273 .type = NFNL_CB_RCU, 8274 .attr_count = NFTA_RULE_MAX, 8275 .policy = nft_rule_policy, 8276 }, 8277 [NFT_MSG_DELRULE] = { 8278 .call = nf_tables_delrule, 8279 .type = NFNL_CB_BATCH, 8280 .attr_count = NFTA_RULE_MAX, 8281 .policy = nft_rule_policy, 8282 }, 8283 [NFT_MSG_NEWSET] = { 8284 .call = nf_tables_newset, 8285 .type = NFNL_CB_BATCH, 8286 .attr_count = NFTA_SET_MAX, 8287 .policy = nft_set_policy, 8288 }, 8289 [NFT_MSG_GETSET] = { 8290 .call = nf_tables_getset, 8291 .type = NFNL_CB_RCU, 8292 .attr_count = NFTA_SET_MAX, 8293 .policy = nft_set_policy, 8294 }, 8295 [NFT_MSG_DELSET] = { 8296 .call = nf_tables_delset, 8297 .type = NFNL_CB_BATCH, 8298 .attr_count = NFTA_SET_MAX, 8299 .policy = nft_set_policy, 8300 }, 8301 [NFT_MSG_NEWSETELEM] = { 8302 .call = nf_tables_newsetelem, 8303 .type = NFNL_CB_BATCH, 8304 .attr_count = NFTA_SET_ELEM_LIST_MAX, 8305 .policy = nft_set_elem_list_policy, 8306 }, 8307 [NFT_MSG_GETSETELEM] = { 8308 .call = nf_tables_getsetelem, 8309 .type = NFNL_CB_RCU, 8310 .attr_count = NFTA_SET_ELEM_LIST_MAX, 8311 .policy = nft_set_elem_list_policy, 8312 }, 8313 [NFT_MSG_DELSETELEM] = { 8314 .call = nf_tables_delsetelem, 8315 .type = NFNL_CB_BATCH, 8316 .attr_count = NFTA_SET_ELEM_LIST_MAX, 8317 .policy = nft_set_elem_list_policy, 8318 }, 8319 [NFT_MSG_GETGEN] = { 8320 .call = nf_tables_getgen, 8321 .type = NFNL_CB_RCU, 8322 }, 8323 [NFT_MSG_NEWOBJ] = { 8324 .call = nf_tables_newobj, 8325 .type = NFNL_CB_BATCH, 8326 .attr_count = NFTA_OBJ_MAX, 8327 .policy = nft_obj_policy, 8328 }, 8329 [NFT_MSG_GETOBJ] = { 8330 .call = nf_tables_getobj, 8331 .type = NFNL_CB_RCU, 8332 .attr_count = NFTA_OBJ_MAX, 8333 .policy = nft_obj_policy, 8334 }, 8335 [NFT_MSG_DELOBJ] = { 8336 .call = nf_tables_delobj, 8337 .type = NFNL_CB_BATCH, 8338 .attr_count = NFTA_OBJ_MAX, 8339 .policy = nft_obj_policy, 8340 }, 8341 [NFT_MSG_GETOBJ_RESET] = { 8342 .call = nf_tables_getobj, 8343 .type = NFNL_CB_RCU, 8344 .attr_count = NFTA_OBJ_MAX, 8345 .policy = nft_obj_policy, 8346 }, 8347 [NFT_MSG_NEWFLOWTABLE] = { 8348 .call = nf_tables_newflowtable, 8349 .type = NFNL_CB_BATCH, 8350 .attr_count = NFTA_FLOWTABLE_MAX, 8351 .policy = nft_flowtable_policy, 8352 }, 8353 [NFT_MSG_GETFLOWTABLE] = { 8354 .call = nf_tables_getflowtable, 8355 .type = NFNL_CB_RCU, 8356 .attr_count = NFTA_FLOWTABLE_MAX, 8357 .policy = nft_flowtable_policy, 8358 }, 8359 [NFT_MSG_DELFLOWTABLE] = { 8360 .call = nf_tables_delflowtable, 8361 .type = NFNL_CB_BATCH, 8362 .attr_count = NFTA_FLOWTABLE_MAX, 8363 .policy = nft_flowtable_policy, 8364 }, 8365 }; 8366 8367 static int nf_tables_validate(struct net *net) 8368 { 8369 struct nftables_pernet *nft_net = nft_pernet(net); 8370 struct nft_table *table; 8371 8372 switch (nft_net->validate_state) { 8373 case NFT_VALIDATE_SKIP: 8374 break; 8375 case NFT_VALIDATE_NEED: 8376 nft_validate_state_update(net, NFT_VALIDATE_DO); 8377 fallthrough; 8378 case NFT_VALIDATE_DO: 8379 list_for_each_entry(table, &nft_net->tables, list) { 8380 if (nft_table_validate(net, table) < 0) 8381 return -EAGAIN; 8382 } 8383 break; 8384 } 8385 8386 return 0; 8387 } 8388 8389 /* a drop policy has to be deferred until all rules have been activated, 8390 * otherwise a large ruleset that contains a drop-policy base chain will 8391 * cause all packets to get dropped until the full transaction has been 8392 * processed. 8393 * 8394 * We defer the drop policy until the transaction has been finalized. 8395 */ 8396 static void nft_chain_commit_drop_policy(struct nft_trans *trans) 8397 { 8398 struct nft_base_chain *basechain; 8399 8400 if (nft_trans_chain_policy(trans) != NF_DROP) 8401 return; 8402 8403 if (!nft_is_base_chain(trans->ctx.chain)) 8404 return; 8405 8406 basechain = nft_base_chain(trans->ctx.chain); 8407 basechain->policy = NF_DROP; 8408 } 8409 8410 static void nft_chain_commit_update(struct nft_trans *trans) 8411 { 8412 struct nft_base_chain *basechain; 8413 8414 if (nft_trans_chain_name(trans)) { 8415 rhltable_remove(&trans->ctx.table->chains_ht, 8416 &trans->ctx.chain->rhlhead, 8417 nft_chain_ht_params); 8418 swap(trans->ctx.chain->name, nft_trans_chain_name(trans)); 8419 rhltable_insert_key(&trans->ctx.table->chains_ht, 8420 trans->ctx.chain->name, 8421 &trans->ctx.chain->rhlhead, 8422 nft_chain_ht_params); 8423 } 8424 8425 if (!nft_is_base_chain(trans->ctx.chain)) 8426 return; 8427 8428 nft_chain_stats_replace(trans); 8429 8430 basechain = nft_base_chain(trans->ctx.chain); 8431 8432 switch (nft_trans_chain_policy(trans)) { 8433 case NF_DROP: 8434 case NF_ACCEPT: 8435 basechain->policy = nft_trans_chain_policy(trans); 8436 break; 8437 } 8438 } 8439 8440 static void nft_obj_commit_update(struct nft_trans *trans) 8441 { 8442 struct nft_object *newobj; 8443 struct nft_object *obj; 8444 8445 obj = nft_trans_obj(trans); 8446 newobj = nft_trans_obj_newobj(trans); 8447 8448 if (obj->ops->update) 8449 obj->ops->update(obj, newobj); 8450 8451 nft_obj_destroy(&trans->ctx, newobj); 8452 } 8453 8454 static void nft_commit_release(struct nft_trans *trans) 8455 { 8456 switch (trans->msg_type) { 8457 case NFT_MSG_DELTABLE: 8458 nf_tables_table_destroy(&trans->ctx); 8459 break; 8460 case NFT_MSG_NEWCHAIN: 8461 free_percpu(nft_trans_chain_stats(trans)); 8462 kfree(nft_trans_chain_name(trans)); 8463 break; 8464 case NFT_MSG_DELCHAIN: 8465 nf_tables_chain_destroy(&trans->ctx); 8466 break; 8467 case NFT_MSG_DELRULE: 8468 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans)); 8469 break; 8470 case NFT_MSG_DELSET: 8471 nft_set_destroy(&trans->ctx, nft_trans_set(trans)); 8472 break; 8473 case NFT_MSG_DELSETELEM: 8474 nf_tables_set_elem_destroy(&trans->ctx, 8475 nft_trans_elem_set(trans), 8476 nft_trans_elem(trans).priv); 8477 break; 8478 case NFT_MSG_DELOBJ: 8479 nft_obj_destroy(&trans->ctx, nft_trans_obj(trans)); 8480 break; 8481 case NFT_MSG_DELFLOWTABLE: 8482 if (nft_trans_flowtable_update(trans)) 8483 nft_flowtable_hooks_destroy(&nft_trans_flowtable_hooks(trans)); 8484 else 8485 nf_tables_flowtable_destroy(nft_trans_flowtable(trans)); 8486 break; 8487 } 8488 8489 if (trans->put_net) 8490 put_net(trans->ctx.net); 8491 8492 kfree(trans); 8493 } 8494 8495 static void nf_tables_trans_destroy_work(struct work_struct *w) 8496 { 8497 struct nft_trans *trans, *next; 8498 LIST_HEAD(head); 8499 8500 spin_lock(&nf_tables_destroy_list_lock); 8501 list_splice_init(&nf_tables_destroy_list, &head); 8502 spin_unlock(&nf_tables_destroy_list_lock); 8503 8504 if (list_empty(&head)) 8505 return; 8506 8507 synchronize_rcu(); 8508 8509 list_for_each_entry_safe(trans, next, &head, list) { 8510 list_del(&trans->list); 8511 nft_commit_release(trans); 8512 } 8513 } 8514 8515 void nf_tables_trans_destroy_flush_work(void) 8516 { 8517 flush_work(&trans_destroy_work); 8518 } 8519 EXPORT_SYMBOL_GPL(nf_tables_trans_destroy_flush_work); 8520 8521 static bool nft_expr_reduce(struct nft_regs_track *track, 8522 const struct nft_expr *expr) 8523 { 8524 return false; 8525 } 8526 8527 static int nf_tables_commit_chain_prepare(struct net *net, struct nft_chain *chain) 8528 { 8529 const struct nft_expr *expr, *last; 8530 struct nft_regs_track track = {}; 8531 unsigned int size, data_size; 8532 void *data, *data_boundary; 8533 struct nft_rule_dp *prule; 8534 struct nft_rule *rule; 8535 8536 /* already handled or inactive chain? */ 8537 if (chain->blob_next || !nft_is_active_next(net, chain)) 8538 return 0; 8539 8540 data_size = 0; 8541 list_for_each_entry(rule, &chain->rules, list) { 8542 if (nft_is_active_next(net, rule)) { 8543 data_size += sizeof(*prule) + rule->dlen; 8544 if (data_size > INT_MAX) 8545 return -ENOMEM; 8546 } 8547 } 8548 data_size += offsetof(struct nft_rule_dp, data); /* last rule */ 8549 8550 chain->blob_next = nf_tables_chain_alloc_rules(data_size); 8551 if (!chain->blob_next) 8552 return -ENOMEM; 8553 8554 data = (void *)chain->blob_next->data; 8555 data_boundary = data + data_size; 8556 size = 0; 8557 8558 list_for_each_entry(rule, &chain->rules, list) { 8559 if (!nft_is_active_next(net, rule)) 8560 continue; 8561 8562 prule = (struct nft_rule_dp *)data; 8563 data += offsetof(struct nft_rule_dp, data); 8564 if (WARN_ON_ONCE(data > data_boundary)) 8565 return -ENOMEM; 8566 8567 size = 0; 8568 track.last = nft_expr_last(rule); 8569 nft_rule_for_each_expr(expr, last, rule) { 8570 track.cur = expr; 8571 8572 if (nft_expr_reduce(&track, expr)) { 8573 expr = track.cur; 8574 continue; 8575 } 8576 8577 if (WARN_ON_ONCE(data + expr->ops->size > data_boundary)) 8578 return -ENOMEM; 8579 8580 memcpy(data + size, expr, expr->ops->size); 8581 size += expr->ops->size; 8582 } 8583 if (WARN_ON_ONCE(size >= 1 << 12)) 8584 return -ENOMEM; 8585 8586 prule->handle = rule->handle; 8587 prule->dlen = size; 8588 prule->is_last = 0; 8589 8590 data += size; 8591 size = 0; 8592 chain->blob_next->size += (unsigned long)(data - (void *)prule); 8593 } 8594 8595 prule = (struct nft_rule_dp *)data; 8596 data += offsetof(struct nft_rule_dp, data); 8597 if (WARN_ON_ONCE(data > data_boundary)) 8598 return -ENOMEM; 8599 8600 nft_last_rule(chain->blob_next, prule); 8601 8602 return 0; 8603 } 8604 8605 static void nf_tables_commit_chain_prepare_cancel(struct net *net) 8606 { 8607 struct nftables_pernet *nft_net = nft_pernet(net); 8608 struct nft_trans *trans, *next; 8609 8610 list_for_each_entry_safe(trans, next, &nft_net->commit_list, list) { 8611 struct nft_chain *chain = trans->ctx.chain; 8612 8613 if (trans->msg_type == NFT_MSG_NEWRULE || 8614 trans->msg_type == NFT_MSG_DELRULE) { 8615 kvfree(chain->blob_next); 8616 chain->blob_next = NULL; 8617 } 8618 } 8619 } 8620 8621 static void __nf_tables_commit_chain_free_rules_old(struct rcu_head *h) 8622 { 8623 struct nft_rules_old *o = container_of(h, struct nft_rules_old, h); 8624 8625 kvfree(o->blob); 8626 } 8627 8628 static void nf_tables_commit_chain_free_rules_old(struct nft_rule_blob *blob) 8629 { 8630 struct nft_rules_old *old; 8631 8632 /* rcu_head is after end marker */ 8633 old = (void *)blob + sizeof(*blob) + blob->size; 8634 old->blob = blob; 8635 8636 call_rcu(&old->h, __nf_tables_commit_chain_free_rules_old); 8637 } 8638 8639 static void nf_tables_commit_chain(struct net *net, struct nft_chain *chain) 8640 { 8641 struct nft_rule_blob *g0, *g1; 8642 bool next_genbit; 8643 8644 next_genbit = nft_gencursor_next(net); 8645 8646 g0 = rcu_dereference_protected(chain->blob_gen_0, 8647 lockdep_commit_lock_is_held(net)); 8648 g1 = rcu_dereference_protected(chain->blob_gen_1, 8649 lockdep_commit_lock_is_held(net)); 8650 8651 /* No changes to this chain? */ 8652 if (chain->blob_next == NULL) { 8653 /* chain had no change in last or next generation */ 8654 if (g0 == g1) 8655 return; 8656 /* 8657 * chain had no change in this generation; make sure next 8658 * one uses same rules as current generation. 8659 */ 8660 if (next_genbit) { 8661 rcu_assign_pointer(chain->blob_gen_1, g0); 8662 nf_tables_commit_chain_free_rules_old(g1); 8663 } else { 8664 rcu_assign_pointer(chain->blob_gen_0, g1); 8665 nf_tables_commit_chain_free_rules_old(g0); 8666 } 8667 8668 return; 8669 } 8670 8671 if (next_genbit) 8672 rcu_assign_pointer(chain->blob_gen_1, chain->blob_next); 8673 else 8674 rcu_assign_pointer(chain->blob_gen_0, chain->blob_next); 8675 8676 chain->blob_next = NULL; 8677 8678 if (g0 == g1) 8679 return; 8680 8681 if (next_genbit) 8682 nf_tables_commit_chain_free_rules_old(g1); 8683 else 8684 nf_tables_commit_chain_free_rules_old(g0); 8685 } 8686 8687 static void nft_obj_del(struct nft_object *obj) 8688 { 8689 rhltable_remove(&nft_objname_ht, &obj->rhlhead, nft_objname_ht_params); 8690 list_del_rcu(&obj->list); 8691 } 8692 8693 void nft_chain_del(struct nft_chain *chain) 8694 { 8695 struct nft_table *table = chain->table; 8696 8697 WARN_ON_ONCE(rhltable_remove(&table->chains_ht, &chain->rhlhead, 8698 nft_chain_ht_params)); 8699 list_del_rcu(&chain->list); 8700 } 8701 8702 static void nf_tables_module_autoload_cleanup(struct net *net) 8703 { 8704 struct nftables_pernet *nft_net = nft_pernet(net); 8705 struct nft_module_request *req, *next; 8706 8707 WARN_ON_ONCE(!list_empty(&nft_net->commit_list)); 8708 list_for_each_entry_safe(req, next, &nft_net->module_list, list) { 8709 WARN_ON_ONCE(!req->done); 8710 list_del(&req->list); 8711 kfree(req); 8712 } 8713 } 8714 8715 static void nf_tables_commit_release(struct net *net) 8716 { 8717 struct nftables_pernet *nft_net = nft_pernet(net); 8718 struct nft_trans *trans; 8719 8720 /* all side effects have to be made visible. 8721 * For example, if a chain named 'foo' has been deleted, a 8722 * new transaction must not find it anymore. 8723 * 8724 * Memory reclaim happens asynchronously from work queue 8725 * to prevent expensive synchronize_rcu() in commit phase. 8726 */ 8727 if (list_empty(&nft_net->commit_list)) { 8728 nf_tables_module_autoload_cleanup(net); 8729 mutex_unlock(&nft_net->commit_mutex); 8730 return; 8731 } 8732 8733 trans = list_last_entry(&nft_net->commit_list, 8734 struct nft_trans, list); 8735 get_net(trans->ctx.net); 8736 WARN_ON_ONCE(trans->put_net); 8737 8738 trans->put_net = true; 8739 spin_lock(&nf_tables_destroy_list_lock); 8740 list_splice_tail_init(&nft_net->commit_list, &nf_tables_destroy_list); 8741 spin_unlock(&nf_tables_destroy_list_lock); 8742 8743 nf_tables_module_autoload_cleanup(net); 8744 schedule_work(&trans_destroy_work); 8745 8746 mutex_unlock(&nft_net->commit_mutex); 8747 } 8748 8749 static void nft_commit_notify(struct net *net, u32 portid) 8750 { 8751 struct nftables_pernet *nft_net = nft_pernet(net); 8752 struct sk_buff *batch_skb = NULL, *nskb, *skb; 8753 unsigned char *data; 8754 int len; 8755 8756 list_for_each_entry_safe(skb, nskb, &nft_net->notify_list, list) { 8757 if (!batch_skb) { 8758 new_batch: 8759 batch_skb = skb; 8760 len = NLMSG_GOODSIZE - skb->len; 8761 list_del(&skb->list); 8762 continue; 8763 } 8764 len -= skb->len; 8765 if (len > 0 && NFT_CB(skb).report == NFT_CB(batch_skb).report) { 8766 data = skb_put(batch_skb, skb->len); 8767 memcpy(data, skb->data, skb->len); 8768 list_del(&skb->list); 8769 kfree_skb(skb); 8770 continue; 8771 } 8772 nfnetlink_send(batch_skb, net, portid, NFNLGRP_NFTABLES, 8773 NFT_CB(batch_skb).report, GFP_KERNEL); 8774 goto new_batch; 8775 } 8776 8777 if (batch_skb) { 8778 nfnetlink_send(batch_skb, net, portid, NFNLGRP_NFTABLES, 8779 NFT_CB(batch_skb).report, GFP_KERNEL); 8780 } 8781 8782 WARN_ON_ONCE(!list_empty(&nft_net->notify_list)); 8783 } 8784 8785 static int nf_tables_commit_audit_alloc(struct list_head *adl, 8786 struct nft_table *table) 8787 { 8788 struct nft_audit_data *adp; 8789 8790 list_for_each_entry(adp, adl, list) { 8791 if (adp->table == table) 8792 return 0; 8793 } 8794 adp = kzalloc(sizeof(*adp), GFP_KERNEL); 8795 if (!adp) 8796 return -ENOMEM; 8797 adp->table = table; 8798 list_add(&adp->list, adl); 8799 return 0; 8800 } 8801 8802 static void nf_tables_commit_audit_free(struct list_head *adl) 8803 { 8804 struct nft_audit_data *adp, *adn; 8805 8806 list_for_each_entry_safe(adp, adn, adl, list) { 8807 list_del(&adp->list); 8808 kfree(adp); 8809 } 8810 } 8811 8812 static void nf_tables_commit_audit_collect(struct list_head *adl, 8813 struct nft_table *table, u32 op) 8814 { 8815 struct nft_audit_data *adp; 8816 8817 list_for_each_entry(adp, adl, list) { 8818 if (adp->table == table) 8819 goto found; 8820 } 8821 WARN_ONCE(1, "table=%s not expected in commit list", table->name); 8822 return; 8823 found: 8824 adp->entries++; 8825 if (!adp->op || adp->op > op) 8826 adp->op = op; 8827 } 8828 8829 #define AUNFTABLENAMELEN (NFT_TABLE_MAXNAMELEN + 22) 8830 8831 static void nf_tables_commit_audit_log(struct list_head *adl, u32 generation) 8832 { 8833 struct nft_audit_data *adp, *adn; 8834 char aubuf[AUNFTABLENAMELEN]; 8835 8836 list_for_each_entry_safe(adp, adn, adl, list) { 8837 snprintf(aubuf, AUNFTABLENAMELEN, "%s:%u", adp->table->name, 8838 generation); 8839 audit_log_nfcfg(aubuf, adp->table->family, adp->entries, 8840 nft2audit_op[adp->op], GFP_KERNEL); 8841 list_del(&adp->list); 8842 kfree(adp); 8843 } 8844 } 8845 8846 static int nf_tables_commit(struct net *net, struct sk_buff *skb) 8847 { 8848 struct nftables_pernet *nft_net = nft_pernet(net); 8849 struct nft_trans *trans, *next; 8850 struct nft_trans_elem *te; 8851 struct nft_chain *chain; 8852 struct nft_table *table; 8853 unsigned int base_seq; 8854 LIST_HEAD(adl); 8855 int err; 8856 8857 if (list_empty(&nft_net->commit_list)) { 8858 mutex_unlock(&nft_net->commit_mutex); 8859 return 0; 8860 } 8861 8862 /* 0. Validate ruleset, otherwise roll back for error reporting. */ 8863 if (nf_tables_validate(net) < 0) 8864 return -EAGAIN; 8865 8866 err = nft_flow_rule_offload_commit(net); 8867 if (err < 0) 8868 return err; 8869 8870 /* 1. Allocate space for next generation rules_gen_X[] */ 8871 list_for_each_entry_safe(trans, next, &nft_net->commit_list, list) { 8872 int ret; 8873 8874 ret = nf_tables_commit_audit_alloc(&adl, trans->ctx.table); 8875 if (ret) { 8876 nf_tables_commit_chain_prepare_cancel(net); 8877 nf_tables_commit_audit_free(&adl); 8878 return ret; 8879 } 8880 if (trans->msg_type == NFT_MSG_NEWRULE || 8881 trans->msg_type == NFT_MSG_DELRULE) { 8882 chain = trans->ctx.chain; 8883 8884 ret = nf_tables_commit_chain_prepare(net, chain); 8885 if (ret < 0) { 8886 nf_tables_commit_chain_prepare_cancel(net); 8887 nf_tables_commit_audit_free(&adl); 8888 return ret; 8889 } 8890 } 8891 } 8892 8893 /* step 2. Make rules_gen_X visible to packet path */ 8894 list_for_each_entry(table, &nft_net->tables, list) { 8895 list_for_each_entry(chain, &table->chains, list) 8896 nf_tables_commit_chain(net, chain); 8897 } 8898 8899 /* 8900 * Bump generation counter, invalidate any dump in progress. 8901 * Cannot fail after this point. 8902 */ 8903 base_seq = READ_ONCE(nft_net->base_seq); 8904 while (++base_seq == 0) 8905 ; 8906 8907 WRITE_ONCE(nft_net->base_seq, base_seq); 8908 8909 /* step 3. Start new generation, rules_gen_X now in use. */ 8910 net->nft.gencursor = nft_gencursor_next(net); 8911 8912 list_for_each_entry_safe(trans, next, &nft_net->commit_list, list) { 8913 nf_tables_commit_audit_collect(&adl, trans->ctx.table, 8914 trans->msg_type); 8915 switch (trans->msg_type) { 8916 case NFT_MSG_NEWTABLE: 8917 if (nft_trans_table_update(trans)) { 8918 if (!(trans->ctx.table->flags & __NFT_TABLE_F_UPDATE)) { 8919 nft_trans_destroy(trans); 8920 break; 8921 } 8922 if (trans->ctx.table->flags & NFT_TABLE_F_DORMANT) 8923 nf_tables_table_disable(net, trans->ctx.table); 8924 8925 trans->ctx.table->flags &= ~__NFT_TABLE_F_UPDATE; 8926 } else { 8927 nft_clear(net, trans->ctx.table); 8928 } 8929 nf_tables_table_notify(&trans->ctx, NFT_MSG_NEWTABLE); 8930 nft_trans_destroy(trans); 8931 break; 8932 case NFT_MSG_DELTABLE: 8933 list_del_rcu(&trans->ctx.table->list); 8934 nf_tables_table_notify(&trans->ctx, NFT_MSG_DELTABLE); 8935 break; 8936 case NFT_MSG_NEWCHAIN: 8937 if (nft_trans_chain_update(trans)) { 8938 nft_chain_commit_update(trans); 8939 nf_tables_chain_notify(&trans->ctx, NFT_MSG_NEWCHAIN); 8940 /* trans destroyed after rcu grace period */ 8941 } else { 8942 nft_chain_commit_drop_policy(trans); 8943 nft_clear(net, trans->ctx.chain); 8944 nf_tables_chain_notify(&trans->ctx, NFT_MSG_NEWCHAIN); 8945 nft_trans_destroy(trans); 8946 } 8947 break; 8948 case NFT_MSG_DELCHAIN: 8949 nft_chain_del(trans->ctx.chain); 8950 nf_tables_chain_notify(&trans->ctx, NFT_MSG_DELCHAIN); 8951 nf_tables_unregister_hook(trans->ctx.net, 8952 trans->ctx.table, 8953 trans->ctx.chain); 8954 break; 8955 case NFT_MSG_NEWRULE: 8956 nft_clear(trans->ctx.net, nft_trans_rule(trans)); 8957 nf_tables_rule_notify(&trans->ctx, 8958 nft_trans_rule(trans), 8959 NFT_MSG_NEWRULE); 8960 if (trans->ctx.chain->flags & NFT_CHAIN_HW_OFFLOAD) 8961 nft_flow_rule_destroy(nft_trans_flow_rule(trans)); 8962 8963 nft_trans_destroy(trans); 8964 break; 8965 case NFT_MSG_DELRULE: 8966 list_del_rcu(&nft_trans_rule(trans)->list); 8967 nf_tables_rule_notify(&trans->ctx, 8968 nft_trans_rule(trans), 8969 NFT_MSG_DELRULE); 8970 nft_rule_expr_deactivate(&trans->ctx, 8971 nft_trans_rule(trans), 8972 NFT_TRANS_COMMIT); 8973 8974 if (trans->ctx.chain->flags & NFT_CHAIN_HW_OFFLOAD) 8975 nft_flow_rule_destroy(nft_trans_flow_rule(trans)); 8976 break; 8977 case NFT_MSG_NEWSET: 8978 nft_clear(net, nft_trans_set(trans)); 8979 /* This avoids hitting -EBUSY when deleting the table 8980 * from the transaction. 8981 */ 8982 if (nft_set_is_anonymous(nft_trans_set(trans)) && 8983 !list_empty(&nft_trans_set(trans)->bindings)) 8984 trans->ctx.table->use--; 8985 8986 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans), 8987 NFT_MSG_NEWSET, GFP_KERNEL); 8988 nft_trans_destroy(trans); 8989 break; 8990 case NFT_MSG_DELSET: 8991 list_del_rcu(&nft_trans_set(trans)->list); 8992 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans), 8993 NFT_MSG_DELSET, GFP_KERNEL); 8994 break; 8995 case NFT_MSG_NEWSETELEM: 8996 te = (struct nft_trans_elem *)trans->data; 8997 8998 nft_setelem_activate(net, te->set, &te->elem); 8999 nf_tables_setelem_notify(&trans->ctx, te->set, 9000 &te->elem, 9001 NFT_MSG_NEWSETELEM); 9002 nft_trans_destroy(trans); 9003 break; 9004 case NFT_MSG_DELSETELEM: 9005 te = (struct nft_trans_elem *)trans->data; 9006 9007 nf_tables_setelem_notify(&trans->ctx, te->set, 9008 &te->elem, 9009 NFT_MSG_DELSETELEM); 9010 nft_setelem_remove(net, te->set, &te->elem); 9011 if (!nft_setelem_is_catchall(te->set, &te->elem)) { 9012 atomic_dec(&te->set->nelems); 9013 te->set->ndeact--; 9014 } 9015 break; 9016 case NFT_MSG_NEWOBJ: 9017 if (nft_trans_obj_update(trans)) { 9018 nft_obj_commit_update(trans); 9019 nf_tables_obj_notify(&trans->ctx, 9020 nft_trans_obj(trans), 9021 NFT_MSG_NEWOBJ); 9022 } else { 9023 nft_clear(net, nft_trans_obj(trans)); 9024 nf_tables_obj_notify(&trans->ctx, 9025 nft_trans_obj(trans), 9026 NFT_MSG_NEWOBJ); 9027 nft_trans_destroy(trans); 9028 } 9029 break; 9030 case NFT_MSG_DELOBJ: 9031 nft_obj_del(nft_trans_obj(trans)); 9032 nf_tables_obj_notify(&trans->ctx, nft_trans_obj(trans), 9033 NFT_MSG_DELOBJ); 9034 break; 9035 case NFT_MSG_NEWFLOWTABLE: 9036 if (nft_trans_flowtable_update(trans)) { 9037 nft_trans_flowtable(trans)->data.flags = 9038 nft_trans_flowtable_flags(trans); 9039 nf_tables_flowtable_notify(&trans->ctx, 9040 nft_trans_flowtable(trans), 9041 &nft_trans_flowtable_hooks(trans), 9042 NFT_MSG_NEWFLOWTABLE); 9043 list_splice(&nft_trans_flowtable_hooks(trans), 9044 &nft_trans_flowtable(trans)->hook_list); 9045 } else { 9046 nft_clear(net, nft_trans_flowtable(trans)); 9047 nf_tables_flowtable_notify(&trans->ctx, 9048 nft_trans_flowtable(trans), 9049 &nft_trans_flowtable(trans)->hook_list, 9050 NFT_MSG_NEWFLOWTABLE); 9051 } 9052 nft_trans_destroy(trans); 9053 break; 9054 case NFT_MSG_DELFLOWTABLE: 9055 if (nft_trans_flowtable_update(trans)) { 9056 nf_tables_flowtable_notify(&trans->ctx, 9057 nft_trans_flowtable(trans), 9058 &nft_trans_flowtable_hooks(trans), 9059 NFT_MSG_DELFLOWTABLE); 9060 nft_unregister_flowtable_net_hooks(net, 9061 &nft_trans_flowtable_hooks(trans)); 9062 } else { 9063 list_del_rcu(&nft_trans_flowtable(trans)->list); 9064 nf_tables_flowtable_notify(&trans->ctx, 9065 nft_trans_flowtable(trans), 9066 &nft_trans_flowtable(trans)->hook_list, 9067 NFT_MSG_DELFLOWTABLE); 9068 nft_unregister_flowtable_net_hooks(net, 9069 &nft_trans_flowtable(trans)->hook_list); 9070 } 9071 break; 9072 } 9073 } 9074 9075 nft_commit_notify(net, NETLINK_CB(skb).portid); 9076 nf_tables_gen_notify(net, skb, NFT_MSG_NEWGEN); 9077 nf_tables_commit_audit_log(&adl, nft_net->base_seq); 9078 nf_tables_commit_release(net); 9079 9080 return 0; 9081 } 9082 9083 static void nf_tables_module_autoload(struct net *net) 9084 { 9085 struct nftables_pernet *nft_net = nft_pernet(net); 9086 struct nft_module_request *req, *next; 9087 LIST_HEAD(module_list); 9088 9089 list_splice_init(&nft_net->module_list, &module_list); 9090 mutex_unlock(&nft_net->commit_mutex); 9091 list_for_each_entry_safe(req, next, &module_list, list) { 9092 request_module("%s", req->module); 9093 req->done = true; 9094 } 9095 mutex_lock(&nft_net->commit_mutex); 9096 list_splice(&module_list, &nft_net->module_list); 9097 } 9098 9099 static void nf_tables_abort_release(struct nft_trans *trans) 9100 { 9101 switch (trans->msg_type) { 9102 case NFT_MSG_NEWTABLE: 9103 nf_tables_table_destroy(&trans->ctx); 9104 break; 9105 case NFT_MSG_NEWCHAIN: 9106 nf_tables_chain_destroy(&trans->ctx); 9107 break; 9108 case NFT_MSG_NEWRULE: 9109 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans)); 9110 break; 9111 case NFT_MSG_NEWSET: 9112 nft_set_destroy(&trans->ctx, nft_trans_set(trans)); 9113 break; 9114 case NFT_MSG_NEWSETELEM: 9115 nft_set_elem_destroy(nft_trans_elem_set(trans), 9116 nft_trans_elem(trans).priv, true); 9117 break; 9118 case NFT_MSG_NEWOBJ: 9119 nft_obj_destroy(&trans->ctx, nft_trans_obj(trans)); 9120 break; 9121 case NFT_MSG_NEWFLOWTABLE: 9122 if (nft_trans_flowtable_update(trans)) 9123 nft_flowtable_hooks_destroy(&nft_trans_flowtable_hooks(trans)); 9124 else 9125 nf_tables_flowtable_destroy(nft_trans_flowtable(trans)); 9126 break; 9127 } 9128 kfree(trans); 9129 } 9130 9131 static int __nf_tables_abort(struct net *net, enum nfnl_abort_action action) 9132 { 9133 struct nftables_pernet *nft_net = nft_pernet(net); 9134 struct nft_trans *trans, *next; 9135 struct nft_trans_elem *te; 9136 9137 if (action == NFNL_ABORT_VALIDATE && 9138 nf_tables_validate(net) < 0) 9139 return -EAGAIN; 9140 9141 list_for_each_entry_safe_reverse(trans, next, &nft_net->commit_list, 9142 list) { 9143 switch (trans->msg_type) { 9144 case NFT_MSG_NEWTABLE: 9145 if (nft_trans_table_update(trans)) { 9146 if (!(trans->ctx.table->flags & __NFT_TABLE_F_UPDATE)) { 9147 nft_trans_destroy(trans); 9148 break; 9149 } 9150 if (trans->ctx.table->flags & __NFT_TABLE_F_WAS_DORMANT) { 9151 nf_tables_table_disable(net, trans->ctx.table); 9152 trans->ctx.table->flags |= NFT_TABLE_F_DORMANT; 9153 } else if (trans->ctx.table->flags & __NFT_TABLE_F_WAS_AWAKEN) { 9154 trans->ctx.table->flags &= ~NFT_TABLE_F_DORMANT; 9155 } 9156 trans->ctx.table->flags &= ~__NFT_TABLE_F_UPDATE; 9157 nft_trans_destroy(trans); 9158 } else { 9159 list_del_rcu(&trans->ctx.table->list); 9160 } 9161 break; 9162 case NFT_MSG_DELTABLE: 9163 nft_clear(trans->ctx.net, trans->ctx.table); 9164 nft_trans_destroy(trans); 9165 break; 9166 case NFT_MSG_NEWCHAIN: 9167 if (nft_trans_chain_update(trans)) { 9168 free_percpu(nft_trans_chain_stats(trans)); 9169 kfree(nft_trans_chain_name(trans)); 9170 nft_trans_destroy(trans); 9171 } else { 9172 if (nft_chain_is_bound(trans->ctx.chain)) { 9173 nft_trans_destroy(trans); 9174 break; 9175 } 9176 trans->ctx.table->use--; 9177 nft_chain_del(trans->ctx.chain); 9178 nf_tables_unregister_hook(trans->ctx.net, 9179 trans->ctx.table, 9180 trans->ctx.chain); 9181 } 9182 break; 9183 case NFT_MSG_DELCHAIN: 9184 trans->ctx.table->use++; 9185 nft_clear(trans->ctx.net, trans->ctx.chain); 9186 nft_trans_destroy(trans); 9187 break; 9188 case NFT_MSG_NEWRULE: 9189 trans->ctx.chain->use--; 9190 list_del_rcu(&nft_trans_rule(trans)->list); 9191 nft_rule_expr_deactivate(&trans->ctx, 9192 nft_trans_rule(trans), 9193 NFT_TRANS_ABORT); 9194 if (trans->ctx.chain->flags & NFT_CHAIN_HW_OFFLOAD) 9195 nft_flow_rule_destroy(nft_trans_flow_rule(trans)); 9196 break; 9197 case NFT_MSG_DELRULE: 9198 trans->ctx.chain->use++; 9199 nft_clear(trans->ctx.net, nft_trans_rule(trans)); 9200 nft_rule_expr_activate(&trans->ctx, nft_trans_rule(trans)); 9201 if (trans->ctx.chain->flags & NFT_CHAIN_HW_OFFLOAD) 9202 nft_flow_rule_destroy(nft_trans_flow_rule(trans)); 9203 9204 nft_trans_destroy(trans); 9205 break; 9206 case NFT_MSG_NEWSET: 9207 trans->ctx.table->use--; 9208 if (nft_trans_set_bound(trans)) { 9209 nft_trans_destroy(trans); 9210 break; 9211 } 9212 list_del_rcu(&nft_trans_set(trans)->list); 9213 break; 9214 case NFT_MSG_DELSET: 9215 trans->ctx.table->use++; 9216 nft_clear(trans->ctx.net, nft_trans_set(trans)); 9217 nft_trans_destroy(trans); 9218 break; 9219 case NFT_MSG_NEWSETELEM: 9220 if (nft_trans_elem_set_bound(trans)) { 9221 nft_trans_destroy(trans); 9222 break; 9223 } 9224 te = (struct nft_trans_elem *)trans->data; 9225 nft_setelem_remove(net, te->set, &te->elem); 9226 if (!nft_setelem_is_catchall(te->set, &te->elem)) 9227 atomic_dec(&te->set->nelems); 9228 break; 9229 case NFT_MSG_DELSETELEM: 9230 te = (struct nft_trans_elem *)trans->data; 9231 9232 nft_setelem_data_activate(net, te->set, &te->elem); 9233 nft_setelem_activate(net, te->set, &te->elem); 9234 if (!nft_setelem_is_catchall(te->set, &te->elem)) 9235 te->set->ndeact--; 9236 9237 nft_trans_destroy(trans); 9238 break; 9239 case NFT_MSG_NEWOBJ: 9240 if (nft_trans_obj_update(trans)) { 9241 nft_obj_destroy(&trans->ctx, nft_trans_obj_newobj(trans)); 9242 nft_trans_destroy(trans); 9243 } else { 9244 trans->ctx.table->use--; 9245 nft_obj_del(nft_trans_obj(trans)); 9246 } 9247 break; 9248 case NFT_MSG_DELOBJ: 9249 trans->ctx.table->use++; 9250 nft_clear(trans->ctx.net, nft_trans_obj(trans)); 9251 nft_trans_destroy(trans); 9252 break; 9253 case NFT_MSG_NEWFLOWTABLE: 9254 if (nft_trans_flowtable_update(trans)) { 9255 nft_unregister_flowtable_net_hooks(net, 9256 &nft_trans_flowtable_hooks(trans)); 9257 } else { 9258 trans->ctx.table->use--; 9259 list_del_rcu(&nft_trans_flowtable(trans)->list); 9260 nft_unregister_flowtable_net_hooks(net, 9261 &nft_trans_flowtable(trans)->hook_list); 9262 } 9263 break; 9264 case NFT_MSG_DELFLOWTABLE: 9265 if (nft_trans_flowtable_update(trans)) { 9266 list_splice(&nft_trans_flowtable_hooks(trans), 9267 &nft_trans_flowtable(trans)->hook_list); 9268 } else { 9269 trans->ctx.table->use++; 9270 nft_clear(trans->ctx.net, nft_trans_flowtable(trans)); 9271 } 9272 nft_trans_destroy(trans); 9273 break; 9274 } 9275 } 9276 9277 synchronize_rcu(); 9278 9279 list_for_each_entry_safe_reverse(trans, next, 9280 &nft_net->commit_list, list) { 9281 list_del(&trans->list); 9282 nf_tables_abort_release(trans); 9283 } 9284 9285 if (action == NFNL_ABORT_AUTOLOAD) 9286 nf_tables_module_autoload(net); 9287 else 9288 nf_tables_module_autoload_cleanup(net); 9289 9290 return 0; 9291 } 9292 9293 static void nf_tables_cleanup(struct net *net) 9294 { 9295 nft_validate_state_update(net, NFT_VALIDATE_SKIP); 9296 } 9297 9298 static int nf_tables_abort(struct net *net, struct sk_buff *skb, 9299 enum nfnl_abort_action action) 9300 { 9301 struct nftables_pernet *nft_net = nft_pernet(net); 9302 int ret = __nf_tables_abort(net, action); 9303 9304 mutex_unlock(&nft_net->commit_mutex); 9305 9306 return ret; 9307 } 9308 9309 static bool nf_tables_valid_genid(struct net *net, u32 genid) 9310 { 9311 struct nftables_pernet *nft_net = nft_pernet(net); 9312 bool genid_ok; 9313 9314 mutex_lock(&nft_net->commit_mutex); 9315 9316 genid_ok = genid == 0 || nft_net->base_seq == genid; 9317 if (!genid_ok) 9318 mutex_unlock(&nft_net->commit_mutex); 9319 9320 /* else, commit mutex has to be released by commit or abort function */ 9321 return genid_ok; 9322 } 9323 9324 static const struct nfnetlink_subsystem nf_tables_subsys = { 9325 .name = "nf_tables", 9326 .subsys_id = NFNL_SUBSYS_NFTABLES, 9327 .cb_count = NFT_MSG_MAX, 9328 .cb = nf_tables_cb, 9329 .commit = nf_tables_commit, 9330 .abort = nf_tables_abort, 9331 .cleanup = nf_tables_cleanup, 9332 .valid_genid = nf_tables_valid_genid, 9333 .owner = THIS_MODULE, 9334 }; 9335 9336 int nft_chain_validate_dependency(const struct nft_chain *chain, 9337 enum nft_chain_types type) 9338 { 9339 const struct nft_base_chain *basechain; 9340 9341 if (nft_is_base_chain(chain)) { 9342 basechain = nft_base_chain(chain); 9343 if (basechain->type->type != type) 9344 return -EOPNOTSUPP; 9345 } 9346 return 0; 9347 } 9348 EXPORT_SYMBOL_GPL(nft_chain_validate_dependency); 9349 9350 int nft_chain_validate_hooks(const struct nft_chain *chain, 9351 unsigned int hook_flags) 9352 { 9353 struct nft_base_chain *basechain; 9354 9355 if (nft_is_base_chain(chain)) { 9356 basechain = nft_base_chain(chain); 9357 9358 if ((1 << basechain->ops.hooknum) & hook_flags) 9359 return 0; 9360 9361 return -EOPNOTSUPP; 9362 } 9363 9364 return 0; 9365 } 9366 EXPORT_SYMBOL_GPL(nft_chain_validate_hooks); 9367 9368 /* 9369 * Loop detection - walk through the ruleset beginning at the destination chain 9370 * of a new jump until either the source chain is reached (loop) or all 9371 * reachable chains have been traversed. 9372 * 9373 * The loop check is performed whenever a new jump verdict is added to an 9374 * expression or verdict map or a verdict map is bound to a new chain. 9375 */ 9376 9377 static int nf_tables_check_loops(const struct nft_ctx *ctx, 9378 const struct nft_chain *chain); 9379 9380 static int nft_check_loops(const struct nft_ctx *ctx, 9381 const struct nft_set_ext *ext) 9382 { 9383 const struct nft_data *data; 9384 int ret; 9385 9386 data = nft_set_ext_data(ext); 9387 switch (data->verdict.code) { 9388 case NFT_JUMP: 9389 case NFT_GOTO: 9390 ret = nf_tables_check_loops(ctx, data->verdict.chain); 9391 break; 9392 default: 9393 ret = 0; 9394 break; 9395 } 9396 9397 return ret; 9398 } 9399 9400 static int nf_tables_loop_check_setelem(const struct nft_ctx *ctx, 9401 struct nft_set *set, 9402 const struct nft_set_iter *iter, 9403 struct nft_set_elem *elem) 9404 { 9405 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv); 9406 9407 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) && 9408 *nft_set_ext_flags(ext) & NFT_SET_ELEM_INTERVAL_END) 9409 return 0; 9410 9411 return nft_check_loops(ctx, ext); 9412 } 9413 9414 static int nft_set_catchall_loops(const struct nft_ctx *ctx, 9415 struct nft_set *set) 9416 { 9417 u8 genmask = nft_genmask_next(ctx->net); 9418 struct nft_set_elem_catchall *catchall; 9419 struct nft_set_ext *ext; 9420 int ret = 0; 9421 9422 list_for_each_entry_rcu(catchall, &set->catchall_list, list) { 9423 ext = nft_set_elem_ext(set, catchall->elem); 9424 if (!nft_set_elem_active(ext, genmask)) 9425 continue; 9426 9427 ret = nft_check_loops(ctx, ext); 9428 if (ret < 0) 9429 return ret; 9430 } 9431 9432 return ret; 9433 } 9434 9435 static int nf_tables_check_loops(const struct nft_ctx *ctx, 9436 const struct nft_chain *chain) 9437 { 9438 const struct nft_rule *rule; 9439 const struct nft_expr *expr, *last; 9440 struct nft_set *set; 9441 struct nft_set_binding *binding; 9442 struct nft_set_iter iter; 9443 9444 if (ctx->chain == chain) 9445 return -ELOOP; 9446 9447 list_for_each_entry(rule, &chain->rules, list) { 9448 nft_rule_for_each_expr(expr, last, rule) { 9449 struct nft_immediate_expr *priv; 9450 const struct nft_data *data; 9451 int err; 9452 9453 if (strcmp(expr->ops->type->name, "immediate")) 9454 continue; 9455 9456 priv = nft_expr_priv(expr); 9457 if (priv->dreg != NFT_REG_VERDICT) 9458 continue; 9459 9460 data = &priv->data; 9461 switch (data->verdict.code) { 9462 case NFT_JUMP: 9463 case NFT_GOTO: 9464 err = nf_tables_check_loops(ctx, 9465 data->verdict.chain); 9466 if (err < 0) 9467 return err; 9468 break; 9469 default: 9470 break; 9471 } 9472 } 9473 } 9474 9475 list_for_each_entry(set, &ctx->table->sets, list) { 9476 if (!nft_is_active_next(ctx->net, set)) 9477 continue; 9478 if (!(set->flags & NFT_SET_MAP) || 9479 set->dtype != NFT_DATA_VERDICT) 9480 continue; 9481 9482 list_for_each_entry(binding, &set->bindings, list) { 9483 if (!(binding->flags & NFT_SET_MAP) || 9484 binding->chain != chain) 9485 continue; 9486 9487 iter.genmask = nft_genmask_next(ctx->net); 9488 iter.skip = 0; 9489 iter.count = 0; 9490 iter.err = 0; 9491 iter.fn = nf_tables_loop_check_setelem; 9492 9493 set->ops->walk(ctx, set, &iter); 9494 if (!iter.err) 9495 iter.err = nft_set_catchall_loops(ctx, set); 9496 9497 if (iter.err < 0) 9498 return iter.err; 9499 } 9500 } 9501 9502 return 0; 9503 } 9504 9505 /** 9506 * nft_parse_u32_check - fetch u32 attribute and check for maximum value 9507 * 9508 * @attr: netlink attribute to fetch value from 9509 * @max: maximum value to be stored in dest 9510 * @dest: pointer to the variable 9511 * 9512 * Parse, check and store a given u32 netlink attribute into variable. 9513 * This function returns -ERANGE if the value goes over maximum value. 9514 * Otherwise a 0 is returned and the attribute value is stored in the 9515 * destination variable. 9516 */ 9517 int nft_parse_u32_check(const struct nlattr *attr, int max, u32 *dest) 9518 { 9519 u32 val; 9520 9521 val = ntohl(nla_get_be32(attr)); 9522 if (val > max) 9523 return -ERANGE; 9524 9525 *dest = val; 9526 return 0; 9527 } 9528 EXPORT_SYMBOL_GPL(nft_parse_u32_check); 9529 9530 static int nft_parse_register(const struct nlattr *attr, u32 *preg) 9531 { 9532 unsigned int reg; 9533 9534 reg = ntohl(nla_get_be32(attr)); 9535 switch (reg) { 9536 case NFT_REG_VERDICT...NFT_REG_4: 9537 *preg = reg * NFT_REG_SIZE / NFT_REG32_SIZE; 9538 break; 9539 case NFT_REG32_00...NFT_REG32_15: 9540 *preg = reg + NFT_REG_SIZE / NFT_REG32_SIZE - NFT_REG32_00; 9541 break; 9542 default: 9543 return -ERANGE; 9544 } 9545 9546 return 0; 9547 } 9548 9549 /** 9550 * nft_dump_register - dump a register value to a netlink attribute 9551 * 9552 * @skb: socket buffer 9553 * @attr: attribute number 9554 * @reg: register number 9555 * 9556 * Construct a netlink attribute containing the register number. For 9557 * compatibility reasons, register numbers being a multiple of 4 are 9558 * translated to the corresponding 128 bit register numbers. 9559 */ 9560 int nft_dump_register(struct sk_buff *skb, unsigned int attr, unsigned int reg) 9561 { 9562 if (reg % (NFT_REG_SIZE / NFT_REG32_SIZE) == 0) 9563 reg = reg / (NFT_REG_SIZE / NFT_REG32_SIZE); 9564 else 9565 reg = reg - NFT_REG_SIZE / NFT_REG32_SIZE + NFT_REG32_00; 9566 9567 return nla_put_be32(skb, attr, htonl(reg)); 9568 } 9569 EXPORT_SYMBOL_GPL(nft_dump_register); 9570 9571 static int nft_validate_register_load(enum nft_registers reg, unsigned int len) 9572 { 9573 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE) 9574 return -EINVAL; 9575 if (len == 0) 9576 return -EINVAL; 9577 if (reg * NFT_REG32_SIZE + len > sizeof_field(struct nft_regs, data)) 9578 return -ERANGE; 9579 9580 return 0; 9581 } 9582 9583 int nft_parse_register_load(const struct nlattr *attr, u8 *sreg, u32 len) 9584 { 9585 u32 reg; 9586 int err; 9587 9588 err = nft_parse_register(attr, ®); 9589 if (err < 0) 9590 return err; 9591 9592 err = nft_validate_register_load(reg, len); 9593 if (err < 0) 9594 return err; 9595 9596 *sreg = reg; 9597 return 0; 9598 } 9599 EXPORT_SYMBOL_GPL(nft_parse_register_load); 9600 9601 static int nft_validate_register_store(const struct nft_ctx *ctx, 9602 enum nft_registers reg, 9603 const struct nft_data *data, 9604 enum nft_data_types type, 9605 unsigned int len) 9606 { 9607 int err; 9608 9609 switch (reg) { 9610 case NFT_REG_VERDICT: 9611 if (type != NFT_DATA_VERDICT) 9612 return -EINVAL; 9613 9614 if (data != NULL && 9615 (data->verdict.code == NFT_GOTO || 9616 data->verdict.code == NFT_JUMP)) { 9617 err = nf_tables_check_loops(ctx, data->verdict.chain); 9618 if (err < 0) 9619 return err; 9620 } 9621 9622 return 0; 9623 default: 9624 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE) 9625 return -EINVAL; 9626 if (len == 0) 9627 return -EINVAL; 9628 if (reg * NFT_REG32_SIZE + len > 9629 sizeof_field(struct nft_regs, data)) 9630 return -ERANGE; 9631 9632 if (data != NULL && type != NFT_DATA_VALUE) 9633 return -EINVAL; 9634 return 0; 9635 } 9636 } 9637 9638 int nft_parse_register_store(const struct nft_ctx *ctx, 9639 const struct nlattr *attr, u8 *dreg, 9640 const struct nft_data *data, 9641 enum nft_data_types type, unsigned int len) 9642 { 9643 int err; 9644 u32 reg; 9645 9646 err = nft_parse_register(attr, ®); 9647 if (err < 0) 9648 return err; 9649 9650 err = nft_validate_register_store(ctx, reg, data, type, len); 9651 if (err < 0) 9652 return err; 9653 9654 *dreg = reg; 9655 return 0; 9656 } 9657 EXPORT_SYMBOL_GPL(nft_parse_register_store); 9658 9659 static const struct nla_policy nft_verdict_policy[NFTA_VERDICT_MAX + 1] = { 9660 [NFTA_VERDICT_CODE] = { .type = NLA_U32 }, 9661 [NFTA_VERDICT_CHAIN] = { .type = NLA_STRING, 9662 .len = NFT_CHAIN_MAXNAMELEN - 1 }, 9663 [NFTA_VERDICT_CHAIN_ID] = { .type = NLA_U32 }, 9664 }; 9665 9666 static int nft_verdict_init(const struct nft_ctx *ctx, struct nft_data *data, 9667 struct nft_data_desc *desc, const struct nlattr *nla) 9668 { 9669 u8 genmask = nft_genmask_next(ctx->net); 9670 struct nlattr *tb[NFTA_VERDICT_MAX + 1]; 9671 struct nft_chain *chain; 9672 int err; 9673 9674 err = nla_parse_nested_deprecated(tb, NFTA_VERDICT_MAX, nla, 9675 nft_verdict_policy, NULL); 9676 if (err < 0) 9677 return err; 9678 9679 if (!tb[NFTA_VERDICT_CODE]) 9680 return -EINVAL; 9681 data->verdict.code = ntohl(nla_get_be32(tb[NFTA_VERDICT_CODE])); 9682 9683 switch (data->verdict.code) { 9684 default: 9685 switch (data->verdict.code & NF_VERDICT_MASK) { 9686 case NF_ACCEPT: 9687 case NF_DROP: 9688 case NF_QUEUE: 9689 break; 9690 default: 9691 return -EINVAL; 9692 } 9693 fallthrough; 9694 case NFT_CONTINUE: 9695 case NFT_BREAK: 9696 case NFT_RETURN: 9697 break; 9698 case NFT_JUMP: 9699 case NFT_GOTO: 9700 if (tb[NFTA_VERDICT_CHAIN]) { 9701 chain = nft_chain_lookup(ctx->net, ctx->table, 9702 tb[NFTA_VERDICT_CHAIN], 9703 genmask); 9704 } else if (tb[NFTA_VERDICT_CHAIN_ID]) { 9705 chain = nft_chain_lookup_byid(ctx->net, ctx->table, 9706 tb[NFTA_VERDICT_CHAIN_ID]); 9707 if (IS_ERR(chain)) 9708 return PTR_ERR(chain); 9709 } else { 9710 return -EINVAL; 9711 } 9712 9713 if (IS_ERR(chain)) 9714 return PTR_ERR(chain); 9715 if (nft_is_base_chain(chain)) 9716 return -EOPNOTSUPP; 9717 if (nft_chain_is_bound(chain)) 9718 return -EINVAL; 9719 if (desc->flags & NFT_DATA_DESC_SETELEM && 9720 chain->flags & NFT_CHAIN_BINDING) 9721 return -EINVAL; 9722 9723 chain->use++; 9724 data->verdict.chain = chain; 9725 break; 9726 } 9727 9728 desc->len = sizeof(data->verdict); 9729 9730 return 0; 9731 } 9732 9733 static void nft_verdict_uninit(const struct nft_data *data) 9734 { 9735 struct nft_chain *chain; 9736 struct nft_rule *rule; 9737 9738 switch (data->verdict.code) { 9739 case NFT_JUMP: 9740 case NFT_GOTO: 9741 chain = data->verdict.chain; 9742 chain->use--; 9743 9744 if (!nft_chain_is_bound(chain)) 9745 break; 9746 9747 chain->table->use--; 9748 list_for_each_entry(rule, &chain->rules, list) 9749 chain->use--; 9750 9751 nft_chain_del(chain); 9752 break; 9753 } 9754 } 9755 9756 int nft_verdict_dump(struct sk_buff *skb, int type, const struct nft_verdict *v) 9757 { 9758 struct nlattr *nest; 9759 9760 nest = nla_nest_start_noflag(skb, type); 9761 if (!nest) 9762 goto nla_put_failure; 9763 9764 if (nla_put_be32(skb, NFTA_VERDICT_CODE, htonl(v->code))) 9765 goto nla_put_failure; 9766 9767 switch (v->code) { 9768 case NFT_JUMP: 9769 case NFT_GOTO: 9770 if (nla_put_string(skb, NFTA_VERDICT_CHAIN, 9771 v->chain->name)) 9772 goto nla_put_failure; 9773 } 9774 nla_nest_end(skb, nest); 9775 return 0; 9776 9777 nla_put_failure: 9778 return -1; 9779 } 9780 9781 static int nft_value_init(const struct nft_ctx *ctx, 9782 struct nft_data *data, struct nft_data_desc *desc, 9783 const struct nlattr *nla) 9784 { 9785 unsigned int len; 9786 9787 len = nla_len(nla); 9788 if (len == 0) 9789 return -EINVAL; 9790 if (len > desc->size) 9791 return -EOVERFLOW; 9792 if (desc->len) { 9793 if (len != desc->len) 9794 return -EINVAL; 9795 } else { 9796 desc->len = len; 9797 } 9798 9799 nla_memcpy(data->data, nla, len); 9800 9801 return 0; 9802 } 9803 9804 static int nft_value_dump(struct sk_buff *skb, const struct nft_data *data, 9805 unsigned int len) 9806 { 9807 return nla_put(skb, NFTA_DATA_VALUE, len, data->data); 9808 } 9809 9810 static const struct nla_policy nft_data_policy[NFTA_DATA_MAX + 1] = { 9811 [NFTA_DATA_VALUE] = { .type = NLA_BINARY }, 9812 [NFTA_DATA_VERDICT] = { .type = NLA_NESTED }, 9813 }; 9814 9815 /** 9816 * nft_data_init - parse nf_tables data netlink attributes 9817 * 9818 * @ctx: context of the expression using the data 9819 * @data: destination struct nft_data 9820 * @desc: data description 9821 * @nla: netlink attribute containing data 9822 * 9823 * Parse the netlink data attributes and initialize a struct nft_data. 9824 * The type and length of data are returned in the data description. 9825 * 9826 * The caller can indicate that it only wants to accept data of type 9827 * NFT_DATA_VALUE by passing NULL for the ctx argument. 9828 */ 9829 int nft_data_init(const struct nft_ctx *ctx, struct nft_data *data, 9830 struct nft_data_desc *desc, const struct nlattr *nla) 9831 { 9832 struct nlattr *tb[NFTA_DATA_MAX + 1]; 9833 int err; 9834 9835 if (WARN_ON_ONCE(!desc->size)) 9836 return -EINVAL; 9837 9838 err = nla_parse_nested_deprecated(tb, NFTA_DATA_MAX, nla, 9839 nft_data_policy, NULL); 9840 if (err < 0) 9841 return err; 9842 9843 if (tb[NFTA_DATA_VALUE]) { 9844 if (desc->type != NFT_DATA_VALUE) 9845 return -EINVAL; 9846 9847 err = nft_value_init(ctx, data, desc, tb[NFTA_DATA_VALUE]); 9848 } else if (tb[NFTA_DATA_VERDICT] && ctx != NULL) { 9849 if (desc->type != NFT_DATA_VERDICT) 9850 return -EINVAL; 9851 9852 err = nft_verdict_init(ctx, data, desc, tb[NFTA_DATA_VERDICT]); 9853 } else { 9854 err = -EINVAL; 9855 } 9856 9857 return err; 9858 } 9859 EXPORT_SYMBOL_GPL(nft_data_init); 9860 9861 /** 9862 * nft_data_release - release a nft_data item 9863 * 9864 * @data: struct nft_data to release 9865 * @type: type of data 9866 * 9867 * Release a nft_data item. NFT_DATA_VALUE types can be silently discarded, 9868 * all others need to be released by calling this function. 9869 */ 9870 void nft_data_release(const struct nft_data *data, enum nft_data_types type) 9871 { 9872 if (type < NFT_DATA_VERDICT) 9873 return; 9874 switch (type) { 9875 case NFT_DATA_VERDICT: 9876 return nft_verdict_uninit(data); 9877 default: 9878 WARN_ON(1); 9879 } 9880 } 9881 EXPORT_SYMBOL_GPL(nft_data_release); 9882 9883 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data, 9884 enum nft_data_types type, unsigned int len) 9885 { 9886 struct nlattr *nest; 9887 int err; 9888 9889 nest = nla_nest_start_noflag(skb, attr); 9890 if (nest == NULL) 9891 return -1; 9892 9893 switch (type) { 9894 case NFT_DATA_VALUE: 9895 err = nft_value_dump(skb, data, len); 9896 break; 9897 case NFT_DATA_VERDICT: 9898 err = nft_verdict_dump(skb, NFTA_DATA_VERDICT, &data->verdict); 9899 break; 9900 default: 9901 err = -EINVAL; 9902 WARN_ON(1); 9903 } 9904 9905 nla_nest_end(skb, nest); 9906 return err; 9907 } 9908 EXPORT_SYMBOL_GPL(nft_data_dump); 9909 9910 int __nft_release_basechain(struct nft_ctx *ctx) 9911 { 9912 struct nft_rule *rule, *nr; 9913 9914 if (WARN_ON(!nft_is_base_chain(ctx->chain))) 9915 return 0; 9916 9917 nf_tables_unregister_hook(ctx->net, ctx->chain->table, ctx->chain); 9918 list_for_each_entry_safe(rule, nr, &ctx->chain->rules, list) { 9919 list_del(&rule->list); 9920 ctx->chain->use--; 9921 nf_tables_rule_release(ctx, rule); 9922 } 9923 nft_chain_del(ctx->chain); 9924 ctx->table->use--; 9925 nf_tables_chain_destroy(ctx); 9926 9927 return 0; 9928 } 9929 EXPORT_SYMBOL_GPL(__nft_release_basechain); 9930 9931 static void __nft_release_hook(struct net *net, struct nft_table *table) 9932 { 9933 struct nft_flowtable *flowtable; 9934 struct nft_chain *chain; 9935 9936 list_for_each_entry(chain, &table->chains, list) 9937 __nf_tables_unregister_hook(net, table, chain, true); 9938 list_for_each_entry(flowtable, &table->flowtables, list) 9939 __nft_unregister_flowtable_net_hooks(net, &flowtable->hook_list, 9940 true); 9941 } 9942 9943 static void __nft_release_hooks(struct net *net) 9944 { 9945 struct nftables_pernet *nft_net = nft_pernet(net); 9946 struct nft_table *table; 9947 9948 list_for_each_entry(table, &nft_net->tables, list) { 9949 if (nft_table_has_owner(table)) 9950 continue; 9951 9952 __nft_release_hook(net, table); 9953 } 9954 } 9955 9956 static void __nft_release_table(struct net *net, struct nft_table *table) 9957 { 9958 struct nft_flowtable *flowtable, *nf; 9959 struct nft_chain *chain, *nc; 9960 struct nft_object *obj, *ne; 9961 struct nft_rule *rule, *nr; 9962 struct nft_set *set, *ns; 9963 struct nft_ctx ctx = { 9964 .net = net, 9965 .family = NFPROTO_NETDEV, 9966 }; 9967 9968 ctx.family = table->family; 9969 ctx.table = table; 9970 list_for_each_entry(chain, &table->chains, list) { 9971 ctx.chain = chain; 9972 list_for_each_entry_safe(rule, nr, &chain->rules, list) { 9973 list_del(&rule->list); 9974 chain->use--; 9975 nf_tables_rule_release(&ctx, rule); 9976 } 9977 } 9978 list_for_each_entry_safe(flowtable, nf, &table->flowtables, list) { 9979 list_del(&flowtable->list); 9980 table->use--; 9981 nf_tables_flowtable_destroy(flowtable); 9982 } 9983 list_for_each_entry_safe(set, ns, &table->sets, list) { 9984 list_del(&set->list); 9985 table->use--; 9986 nft_set_destroy(&ctx, set); 9987 } 9988 list_for_each_entry_safe(obj, ne, &table->objects, list) { 9989 nft_obj_del(obj); 9990 table->use--; 9991 nft_obj_destroy(&ctx, obj); 9992 } 9993 list_for_each_entry_safe(chain, nc, &table->chains, list) { 9994 ctx.chain = chain; 9995 nft_chain_del(chain); 9996 table->use--; 9997 nf_tables_chain_destroy(&ctx); 9998 } 9999 nf_tables_table_destroy(&ctx); 10000 } 10001 10002 static void __nft_release_tables(struct net *net) 10003 { 10004 struct nftables_pernet *nft_net = nft_pernet(net); 10005 struct nft_table *table, *nt; 10006 10007 list_for_each_entry_safe(table, nt, &nft_net->tables, list) { 10008 if (nft_table_has_owner(table)) 10009 continue; 10010 10011 list_del(&table->list); 10012 10013 __nft_release_table(net, table); 10014 } 10015 } 10016 10017 static int nft_rcv_nl_event(struct notifier_block *this, unsigned long event, 10018 void *ptr) 10019 { 10020 struct nft_table *table, *to_delete[8]; 10021 struct nftables_pernet *nft_net; 10022 struct netlink_notify *n = ptr; 10023 struct net *net = n->net; 10024 unsigned int deleted; 10025 bool restart = false; 10026 10027 if (event != NETLINK_URELEASE || n->protocol != NETLINK_NETFILTER) 10028 return NOTIFY_DONE; 10029 10030 nft_net = nft_pernet(net); 10031 deleted = 0; 10032 mutex_lock(&nft_net->commit_mutex); 10033 if (!list_empty(&nf_tables_destroy_list)) 10034 rcu_barrier(); 10035 again: 10036 list_for_each_entry(table, &nft_net->tables, list) { 10037 if (nft_table_has_owner(table) && 10038 n->portid == table->nlpid) { 10039 __nft_release_hook(net, table); 10040 list_del_rcu(&table->list); 10041 to_delete[deleted++] = table; 10042 if (deleted >= ARRAY_SIZE(to_delete)) 10043 break; 10044 } 10045 } 10046 if (deleted) { 10047 restart = deleted >= ARRAY_SIZE(to_delete); 10048 synchronize_rcu(); 10049 while (deleted) 10050 __nft_release_table(net, to_delete[--deleted]); 10051 10052 if (restart) 10053 goto again; 10054 } 10055 mutex_unlock(&nft_net->commit_mutex); 10056 10057 return NOTIFY_DONE; 10058 } 10059 10060 static struct notifier_block nft_nl_notifier = { 10061 .notifier_call = nft_rcv_nl_event, 10062 }; 10063 10064 static int __net_init nf_tables_init_net(struct net *net) 10065 { 10066 struct nftables_pernet *nft_net = nft_pernet(net); 10067 10068 INIT_LIST_HEAD(&nft_net->tables); 10069 INIT_LIST_HEAD(&nft_net->commit_list); 10070 INIT_LIST_HEAD(&nft_net->module_list); 10071 INIT_LIST_HEAD(&nft_net->notify_list); 10072 mutex_init(&nft_net->commit_mutex); 10073 nft_net->base_seq = 1; 10074 nft_net->validate_state = NFT_VALIDATE_SKIP; 10075 10076 return 0; 10077 } 10078 10079 static void __net_exit nf_tables_pre_exit_net(struct net *net) 10080 { 10081 struct nftables_pernet *nft_net = nft_pernet(net); 10082 10083 mutex_lock(&nft_net->commit_mutex); 10084 __nft_release_hooks(net); 10085 mutex_unlock(&nft_net->commit_mutex); 10086 } 10087 10088 static void __net_exit nf_tables_exit_net(struct net *net) 10089 { 10090 struct nftables_pernet *nft_net = nft_pernet(net); 10091 10092 mutex_lock(&nft_net->commit_mutex); 10093 if (!list_empty(&nft_net->commit_list) || 10094 !list_empty(&nft_net->module_list)) 10095 __nf_tables_abort(net, NFNL_ABORT_NONE); 10096 __nft_release_tables(net); 10097 mutex_unlock(&nft_net->commit_mutex); 10098 WARN_ON_ONCE(!list_empty(&nft_net->tables)); 10099 WARN_ON_ONCE(!list_empty(&nft_net->module_list)); 10100 WARN_ON_ONCE(!list_empty(&nft_net->notify_list)); 10101 } 10102 10103 static struct pernet_operations nf_tables_net_ops = { 10104 .init = nf_tables_init_net, 10105 .pre_exit = nf_tables_pre_exit_net, 10106 .exit = nf_tables_exit_net, 10107 .id = &nf_tables_net_id, 10108 .size = sizeof(struct nftables_pernet), 10109 }; 10110 10111 static int __init nf_tables_module_init(void) 10112 { 10113 int err; 10114 10115 err = register_pernet_subsys(&nf_tables_net_ops); 10116 if (err < 0) 10117 return err; 10118 10119 err = nft_chain_filter_init(); 10120 if (err < 0) 10121 goto err_chain_filter; 10122 10123 err = nf_tables_core_module_init(); 10124 if (err < 0) 10125 goto err_core_module; 10126 10127 err = register_netdevice_notifier(&nf_tables_flowtable_notifier); 10128 if (err < 0) 10129 goto err_netdev_notifier; 10130 10131 err = rhltable_init(&nft_objname_ht, &nft_objname_ht_params); 10132 if (err < 0) 10133 goto err_rht_objname; 10134 10135 err = nft_offload_init(); 10136 if (err < 0) 10137 goto err_offload; 10138 10139 err = netlink_register_notifier(&nft_nl_notifier); 10140 if (err < 0) 10141 goto err_netlink_notifier; 10142 10143 /* must be last */ 10144 err = nfnetlink_subsys_register(&nf_tables_subsys); 10145 if (err < 0) 10146 goto err_nfnl_subsys; 10147 10148 nft_chain_route_init(); 10149 10150 return err; 10151 10152 err_nfnl_subsys: 10153 netlink_unregister_notifier(&nft_nl_notifier); 10154 err_netlink_notifier: 10155 nft_offload_exit(); 10156 err_offload: 10157 rhltable_destroy(&nft_objname_ht); 10158 err_rht_objname: 10159 unregister_netdevice_notifier(&nf_tables_flowtable_notifier); 10160 err_netdev_notifier: 10161 nf_tables_core_module_exit(); 10162 err_core_module: 10163 nft_chain_filter_fini(); 10164 err_chain_filter: 10165 unregister_pernet_subsys(&nf_tables_net_ops); 10166 return err; 10167 } 10168 10169 static void __exit nf_tables_module_exit(void) 10170 { 10171 nfnetlink_subsys_unregister(&nf_tables_subsys); 10172 netlink_unregister_notifier(&nft_nl_notifier); 10173 nft_offload_exit(); 10174 unregister_netdevice_notifier(&nf_tables_flowtable_notifier); 10175 nft_chain_filter_fini(); 10176 nft_chain_route_fini(); 10177 unregister_pernet_subsys(&nf_tables_net_ops); 10178 cancel_work_sync(&trans_destroy_work); 10179 rcu_barrier(); 10180 rhltable_destroy(&nft_objname_ht); 10181 nf_tables_core_module_exit(); 10182 } 10183 10184 module_init(nf_tables_module_init); 10185 module_exit(nf_tables_module_exit); 10186 10187 MODULE_LICENSE("GPL"); 10188 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 10189 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_NFTABLES); 10190