1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _NET_NF_TABLES_H 3 #define _NET_NF_TABLES_H 4 5 #include <linux/unaligned.h> 6 #include <linux/list.h> 7 #include <linux/netfilter.h> 8 #include <linux/netfilter/nfnetlink.h> 9 #include <linux/netfilter/x_tables.h> 10 #include <linux/netfilter/nf_tables.h> 11 #include <linux/u64_stats_sync.h> 12 #include <linux/rhashtable.h> 13 #include <net/netfilter/nf_flow_table.h> 14 #include <net/netlink.h> 15 #include <net/flow_offload.h> 16 #include <net/netns/generic.h> 17 18 #define NFT_MAX_HOOKS (NF_INET_INGRESS + 1) 19 20 struct module; 21 22 #define NFT_JUMP_STACK_SIZE 16 23 24 enum { 25 NFT_PKTINFO_L4PROTO = (1 << 0), 26 NFT_PKTINFO_INNER = (1 << 1), 27 NFT_PKTINFO_INNER_FULL = (1 << 2), 28 }; 29 30 struct nft_pktinfo { 31 struct sk_buff *skb; 32 const struct nf_hook_state *state; 33 u8 flags; 34 u8 tprot; 35 u16 fragoff; 36 u16 thoff; 37 u16 inneroff; 38 }; 39 40 static inline struct sock *nft_sk(const struct nft_pktinfo *pkt) 41 { 42 return pkt->state->sk; 43 } 44 45 static inline unsigned int nft_thoff(const struct nft_pktinfo *pkt) 46 { 47 return pkt->thoff; 48 } 49 50 static inline struct net *nft_net(const struct nft_pktinfo *pkt) 51 { 52 return pkt->state->net; 53 } 54 55 static inline unsigned int nft_hook(const struct nft_pktinfo *pkt) 56 { 57 return pkt->state->hook; 58 } 59 60 static inline u8 nft_pf(const struct nft_pktinfo *pkt) 61 { 62 return pkt->state->pf; 63 } 64 65 static inline const struct net_device *nft_in(const struct nft_pktinfo *pkt) 66 { 67 return pkt->state->in; 68 } 69 70 static inline const struct net_device *nft_out(const struct nft_pktinfo *pkt) 71 { 72 return pkt->state->out; 73 } 74 75 static inline void nft_set_pktinfo(struct nft_pktinfo *pkt, 76 struct sk_buff *skb, 77 const struct nf_hook_state *state) 78 { 79 pkt->skb = skb; 80 pkt->state = state; 81 } 82 83 static inline void nft_set_pktinfo_unspec(struct nft_pktinfo *pkt) 84 { 85 pkt->flags = 0; 86 pkt->tprot = 0; 87 pkt->thoff = 0; 88 pkt->fragoff = 0; 89 } 90 91 /** 92 * struct nft_verdict - nf_tables verdict 93 * 94 * @code: nf_tables/netfilter verdict code 95 * @chain: destination chain for NFT_JUMP/NFT_GOTO 96 */ 97 struct nft_verdict { 98 u32 code; 99 struct nft_chain *chain; 100 }; 101 102 struct nft_data { 103 union { 104 u32 data[4]; 105 struct nft_verdict verdict; 106 }; 107 } __attribute__((aligned(__alignof__(u64)))); 108 109 #define NFT_REG32_NUM 20 110 111 /** 112 * struct nft_regs - nf_tables register set 113 * 114 * @data: data registers 115 * @verdict: verdict register 116 * 117 * The first four data registers alias to the verdict register. 118 */ 119 struct nft_regs { 120 union { 121 u32 data[NFT_REG32_NUM]; 122 struct nft_verdict verdict; 123 }; 124 }; 125 126 struct nft_regs_track { 127 struct { 128 const struct nft_expr *selector; 129 const struct nft_expr *bitwise; 130 u8 num_reg; 131 } regs[NFT_REG32_NUM]; 132 133 const struct nft_expr *cur; 134 const struct nft_expr *last; 135 }; 136 137 /* Store/load an u8, u16 or u64 integer to/from the u32 data register. 138 * 139 * Note, when using concatenations, register allocation happens at 32-bit 140 * level. So for store instruction, pad the rest part with zero to avoid 141 * garbage values. 142 */ 143 144 static inline void nft_reg_store8(u32 *dreg, u8 val) 145 { 146 *dreg = 0; 147 *(u8 *)dreg = val; 148 } 149 150 static inline u8 nft_reg_load8(const u32 *sreg) 151 { 152 return *(u8 *)sreg; 153 } 154 155 static inline void nft_reg_store16(u32 *dreg, u16 val) 156 { 157 *dreg = 0; 158 *(u16 *)dreg = val; 159 } 160 161 static inline void nft_reg_store_be16(u32 *dreg, __be16 val) 162 { 163 nft_reg_store16(dreg, (__force __u16)val); 164 } 165 166 static inline u16 nft_reg_load16(const u32 *sreg) 167 { 168 return *(u16 *)sreg; 169 } 170 171 static inline __be16 nft_reg_load_be16(const u32 *sreg) 172 { 173 return (__force __be16)nft_reg_load16(sreg); 174 } 175 176 static inline __be32 nft_reg_load_be32(const u32 *sreg) 177 { 178 return *(__force __be32 *)sreg; 179 } 180 181 static inline void nft_reg_store64(u64 *dreg, u64 val) 182 { 183 put_unaligned(val, dreg); 184 } 185 186 static inline u64 nft_reg_load64(const u32 *sreg) 187 { 188 return get_unaligned((u64 *)sreg); 189 } 190 191 static inline void nft_data_copy(u32 *dst, const struct nft_data *src, 192 unsigned int len) 193 { 194 if (len % NFT_REG32_SIZE) 195 dst[len / NFT_REG32_SIZE] = 0; 196 memcpy(dst, src, len); 197 } 198 199 /** 200 * struct nft_ctx - nf_tables rule/set context 201 * 202 * @net: net namespace 203 * @table: the table the chain is contained in 204 * @chain: the chain the rule is contained in 205 * @nla: netlink attributes 206 * @portid: netlink portID of the original message 207 * @seq: netlink sequence number 208 * @flags: modifiers to new request 209 * @family: protocol family 210 * @level: depth of the chains 211 * @report: notify via unicast netlink message 212 * @reg_inited: bitmap of initialised registers 213 */ 214 struct nft_ctx { 215 struct net *net; 216 struct nft_table *table; 217 struct nft_chain *chain; 218 const struct nlattr * const *nla; 219 u32 portid; 220 u32 seq; 221 u16 flags; 222 u8 family; 223 u8 level; 224 bool report; 225 DECLARE_BITMAP(reg_inited, NFT_REG32_NUM); 226 }; 227 228 enum nft_data_desc_flags { 229 NFT_DATA_DESC_SETELEM = (1 << 0), 230 }; 231 232 struct nft_data_desc { 233 enum nft_data_types type; 234 unsigned int size; 235 unsigned int len; 236 unsigned int flags; 237 }; 238 239 int nft_data_init(const struct nft_ctx *ctx, struct nft_data *data, 240 struct nft_data_desc *desc, const struct nlattr *nla); 241 void nft_data_hold(const struct nft_data *data, enum nft_data_types type); 242 void nft_data_release(const struct nft_data *data, enum nft_data_types type); 243 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data, 244 enum nft_data_types type, unsigned int len); 245 246 static inline enum nft_data_types nft_dreg_to_type(enum nft_registers reg) 247 { 248 return reg == NFT_REG_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE; 249 } 250 251 static inline enum nft_registers nft_type_to_reg(enum nft_data_types type) 252 { 253 return type == NFT_DATA_VERDICT ? NFT_REG_VERDICT : NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE; 254 } 255 256 int nft_parse_u32_check(const struct nlattr *attr, int max, u32 *dest); 257 int nft_dump_register(struct sk_buff *skb, unsigned int attr, unsigned int reg); 258 259 int nft_parse_register_load(const struct nft_ctx *ctx, 260 const struct nlattr *attr, u8 *sreg, u32 len); 261 int nft_parse_register_store(const struct nft_ctx *ctx, 262 const struct nlattr *attr, u8 *dreg, 263 const struct nft_data *data, 264 enum nft_data_types type, unsigned int len); 265 266 /** 267 * struct nft_userdata - user defined data associated with an object 268 * 269 * @len: length of the data 270 * @data: content 271 * 272 * The presence of user data is indicated in an object specific fashion, 273 * so a length of zero can't occur and the value "len" indicates data 274 * of length len + 1. 275 */ 276 struct nft_userdata { 277 u8 len; 278 unsigned char data[]; 279 }; 280 281 /* placeholder structure for opaque set element backend representation. */ 282 struct nft_elem_priv { }; 283 284 /** 285 * struct nft_set_elem - generic representation of set elements 286 * 287 * @key: element key 288 * @key_end: closing element key 289 * @data: element data 290 * @priv: element private data and extensions 291 */ 292 struct nft_set_elem { 293 union { 294 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)]; 295 struct nft_data val; 296 } key; 297 union { 298 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)]; 299 struct nft_data val; 300 } key_end; 301 union { 302 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)]; 303 struct nft_data val; 304 } data; 305 struct nft_elem_priv *priv; 306 }; 307 308 static inline void *nft_elem_priv_cast(const struct nft_elem_priv *priv) 309 { 310 return (void *)priv; 311 } 312 313 314 /** 315 * enum nft_iter_type - nftables set iterator type 316 * 317 * @NFT_ITER_UNSPEC: unspecified, to catch errors 318 * @NFT_ITER_READ: read-only iteration over set elements 319 * @NFT_ITER_UPDATE: iteration under mutex to update set element state 320 */ 321 enum nft_iter_type { 322 NFT_ITER_UNSPEC, 323 NFT_ITER_READ, 324 NFT_ITER_UPDATE, 325 }; 326 327 struct nft_set; 328 struct nft_set_iter { 329 u8 genmask; 330 enum nft_iter_type type:8; 331 unsigned int count; 332 unsigned int skip; 333 int err; 334 int (*fn)(const struct nft_ctx *ctx, 335 struct nft_set *set, 336 const struct nft_set_iter *iter, 337 struct nft_elem_priv *elem_priv); 338 }; 339 340 /** 341 * struct nft_set_desc - description of set elements 342 * 343 * @ktype: key type 344 * @klen: key length 345 * @dtype: data type 346 * @dlen: data length 347 * @objtype: object type 348 * @size: number of set elements 349 * @policy: set policy 350 * @gc_int: garbage collector interval 351 * @timeout: element timeout 352 * @field_len: length of each field in concatenation, bytes 353 * @field_count: number of concatenated fields in element 354 * @expr: set must support for expressions 355 */ 356 struct nft_set_desc { 357 u32 ktype; 358 unsigned int klen; 359 u32 dtype; 360 unsigned int dlen; 361 u32 objtype; 362 unsigned int size; 363 u32 policy; 364 u32 gc_int; 365 u64 timeout; 366 u8 field_len[NFT_REG32_COUNT]; 367 u8 field_count; 368 bool expr; 369 }; 370 371 /** 372 * enum nft_set_class - performance class 373 * 374 * @NFT_SET_CLASS_O_1: constant, O(1) 375 * @NFT_SET_CLASS_O_LOG_N: logarithmic, O(log N) 376 * @NFT_SET_CLASS_O_N: linear, O(N) 377 */ 378 enum nft_set_class { 379 NFT_SET_CLASS_O_1, 380 NFT_SET_CLASS_O_LOG_N, 381 NFT_SET_CLASS_O_N, 382 }; 383 384 /** 385 * struct nft_set_estimate - estimation of memory and performance 386 * characteristics 387 * 388 * @size: required memory 389 * @lookup: lookup performance class 390 * @space: memory class 391 */ 392 struct nft_set_estimate { 393 u64 size; 394 enum nft_set_class lookup; 395 enum nft_set_class space; 396 }; 397 398 #define NFT_EXPR_MAXATTR 16 399 #define NFT_EXPR_SIZE(size) (sizeof(struct nft_expr) + \ 400 ALIGN(size, __alignof__(struct nft_expr))) 401 402 /** 403 * struct nft_expr - nf_tables expression 404 * 405 * @ops: expression ops 406 * @data: expression private data 407 */ 408 struct nft_expr { 409 const struct nft_expr_ops *ops; 410 unsigned char data[] 411 __attribute__((aligned(__alignof__(u64)))); 412 }; 413 414 static inline void *nft_expr_priv(const struct nft_expr *expr) 415 { 416 return (void *)expr->data; 417 } 418 419 struct nft_expr_info; 420 421 int nft_expr_inner_parse(const struct nft_ctx *ctx, const struct nlattr *nla, 422 struct nft_expr_info *info); 423 int nft_expr_clone(struct nft_expr *dst, struct nft_expr *src, gfp_t gfp); 424 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr); 425 int nft_expr_dump(struct sk_buff *skb, unsigned int attr, 426 const struct nft_expr *expr, bool reset); 427 bool nft_expr_reduce_bitwise(struct nft_regs_track *track, 428 const struct nft_expr *expr); 429 430 struct nft_set_ext; 431 432 /** 433 * struct nft_set_ops - nf_tables set operations 434 * 435 * @lookup: look up an element within the set 436 * @update: update an element if exists, add it if doesn't exist 437 * @delete: delete an element 438 * @insert: insert new element into set 439 * @activate: activate new element in the next generation 440 * @deactivate: lookup for element and deactivate it in the next generation 441 * @flush: deactivate element in the next generation 442 * @remove: remove element from set 443 * @walk: iterate over all set elements 444 * @get: get set elements 445 * @ksize: kernel set size 446 * @usize: userspace set size 447 * @adjust_maxsize: delta to adjust maximum set size 448 * @commit: commit set elements 449 * @abort: abort set elements 450 * @privsize: function to return size of set private data 451 * @estimate: estimate the required memory size and the lookup complexity class 452 * @init: initialize private data of new set instance 453 * @destroy: destroy private data of set instance 454 * @gc_init: initialize garbage collection 455 * @elemsize: element private size 456 * 457 * Operations lookup, update and delete have simpler interfaces, are faster 458 * and currently only used in the packet path. All the rest are slower, 459 * control plane functions. 460 */ 461 struct nft_set_ops { 462 bool (*lookup)(const struct net *net, 463 const struct nft_set *set, 464 const u32 *key, 465 const struct nft_set_ext **ext); 466 bool (*update)(struct nft_set *set, 467 const u32 *key, 468 struct nft_elem_priv * 469 (*new)(struct nft_set *, 470 const struct nft_expr *, 471 struct nft_regs *), 472 const struct nft_expr *expr, 473 struct nft_regs *regs, 474 const struct nft_set_ext **ext); 475 bool (*delete)(const struct nft_set *set, 476 const u32 *key); 477 478 int (*insert)(const struct net *net, 479 const struct nft_set *set, 480 const struct nft_set_elem *elem, 481 struct nft_elem_priv **priv); 482 void (*activate)(const struct net *net, 483 const struct nft_set *set, 484 struct nft_elem_priv *elem_priv); 485 struct nft_elem_priv * (*deactivate)(const struct net *net, 486 const struct nft_set *set, 487 const struct nft_set_elem *elem); 488 void (*flush)(const struct net *net, 489 const struct nft_set *set, 490 struct nft_elem_priv *priv); 491 void (*remove)(const struct net *net, 492 const struct nft_set *set, 493 struct nft_elem_priv *elem_priv); 494 void (*walk)(const struct nft_ctx *ctx, 495 struct nft_set *set, 496 struct nft_set_iter *iter); 497 struct nft_elem_priv * (*get)(const struct net *net, 498 const struct nft_set *set, 499 const struct nft_set_elem *elem, 500 unsigned int flags); 501 u32 (*ksize)(u32 size); 502 u32 (*usize)(u32 size); 503 u32 (*adjust_maxsize)(const struct nft_set *set); 504 void (*commit)(struct nft_set *set); 505 void (*abort)(const struct nft_set *set); 506 u64 (*privsize)(const struct nlattr * const nla[], 507 const struct nft_set_desc *desc); 508 bool (*estimate)(const struct nft_set_desc *desc, 509 u32 features, 510 struct nft_set_estimate *est); 511 int (*init)(const struct nft_set *set, 512 const struct nft_set_desc *desc, 513 const struct nlattr * const nla[]); 514 void (*destroy)(const struct nft_ctx *ctx, 515 const struct nft_set *set); 516 void (*gc_init)(const struct nft_set *set); 517 518 unsigned int elemsize; 519 }; 520 521 /** 522 * struct nft_set_type - nf_tables set type 523 * 524 * @ops: set ops for this type 525 * @features: features supported by the implementation 526 */ 527 struct nft_set_type { 528 const struct nft_set_ops ops; 529 u32 features; 530 }; 531 #define to_set_type(o) container_of(o, struct nft_set_type, ops) 532 533 struct nft_set_elem_expr { 534 u8 size; 535 unsigned char data[] 536 __attribute__((aligned(__alignof__(struct nft_expr)))); 537 }; 538 539 #define nft_setelem_expr_at(__elem_expr, __offset) \ 540 ((struct nft_expr *)&__elem_expr->data[__offset]) 541 542 #define nft_setelem_expr_foreach(__expr, __elem_expr, __size) \ 543 for (__expr = nft_setelem_expr_at(__elem_expr, 0), __size = 0; \ 544 __size < (__elem_expr)->size; \ 545 __size += (__expr)->ops->size, __expr = ((void *)(__expr)) + (__expr)->ops->size) 546 547 #define NFT_SET_EXPR_MAX 2 548 549 /** 550 * struct nft_set - nf_tables set instance 551 * 552 * @list: table set list node 553 * @bindings: list of set bindings 554 * @refs: internal refcounting for async set destruction 555 * @table: table this set belongs to 556 * @net: netnamespace this set belongs to 557 * @name: name of the set 558 * @handle: unique handle of the set 559 * @ktype: key type (numeric type defined by userspace, not used in the kernel) 560 * @dtype: data type (verdict or numeric type defined by userspace) 561 * @objtype: object type (see NFT_OBJECT_* definitions) 562 * @size: maximum set size 563 * @field_len: length of each field in concatenation, bytes 564 * @field_count: number of concatenated fields in element 565 * @use: number of rules references to this set 566 * @nelems: number of elements 567 * @ndeact: number of deactivated elements queued for removal 568 * @timeout: default timeout value in jiffies 569 * @gc_int: garbage collection interval in msecs 570 * @policy: set parameterization (see enum nft_set_policies) 571 * @udlen: user data length 572 * @udata: user data 573 * @pending_update: list of pending update set element 574 * @ops: set ops 575 * @flags: set flags 576 * @dead: set will be freed, never cleared 577 * @genmask: generation mask 578 * @klen: key length 579 * @dlen: data length 580 * @num_exprs: numbers of exprs 581 * @exprs: stateful expression 582 * @catchall_list: list of catch-all set element 583 * @data: private set data 584 */ 585 struct nft_set { 586 struct list_head list; 587 struct list_head bindings; 588 refcount_t refs; 589 struct nft_table *table; 590 possible_net_t net; 591 char *name; 592 u64 handle; 593 u32 ktype; 594 u32 dtype; 595 u32 objtype; 596 u32 size; 597 u8 field_len[NFT_REG32_COUNT]; 598 u8 field_count; 599 u32 use; 600 atomic_t nelems; 601 u32 ndeact; 602 u64 timeout; 603 u32 gc_int; 604 u16 policy; 605 u16 udlen; 606 unsigned char *udata; 607 struct list_head pending_update; 608 /* runtime data below here */ 609 const struct nft_set_ops *ops ____cacheline_aligned; 610 u16 flags:13, 611 dead:1, 612 genmask:2; 613 u8 klen; 614 u8 dlen; 615 u8 num_exprs; 616 struct nft_expr *exprs[NFT_SET_EXPR_MAX]; 617 struct list_head catchall_list; 618 unsigned char data[] 619 __attribute__((aligned(__alignof__(u64)))); 620 }; 621 622 static inline bool nft_set_is_anonymous(const struct nft_set *set) 623 { 624 return set->flags & NFT_SET_ANONYMOUS; 625 } 626 627 static inline void *nft_set_priv(const struct nft_set *set) 628 { 629 return (void *)set->data; 630 } 631 632 static inline enum nft_data_types nft_set_datatype(const struct nft_set *set) 633 { 634 return set->dtype == NFT_DATA_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE; 635 } 636 637 static inline bool nft_set_gc_is_pending(const struct nft_set *s) 638 { 639 return refcount_read(&s->refs) != 1; 640 } 641 642 static inline struct nft_set *nft_set_container_of(const void *priv) 643 { 644 return (void *)priv - offsetof(struct nft_set, data); 645 } 646 647 struct nft_set *nft_set_lookup_global(const struct net *net, 648 const struct nft_table *table, 649 const struct nlattr *nla_set_name, 650 const struct nlattr *nla_set_id, 651 u8 genmask); 652 653 struct nft_set_ext *nft_set_catchall_lookup(const struct net *net, 654 const struct nft_set *set); 655 656 static inline unsigned long nft_set_gc_interval(const struct nft_set *set) 657 { 658 u32 gc_int = READ_ONCE(set->gc_int); 659 660 return gc_int ? msecs_to_jiffies(gc_int) : HZ; 661 } 662 663 /** 664 * struct nft_set_binding - nf_tables set binding 665 * 666 * @list: set bindings list node 667 * @chain: chain containing the rule bound to the set 668 * @flags: set action flags 669 * 670 * A set binding contains all information necessary for validation 671 * of new elements added to a bound set. 672 */ 673 struct nft_set_binding { 674 struct list_head list; 675 const struct nft_chain *chain; 676 u32 flags; 677 }; 678 679 enum nft_trans_phase; 680 void nf_tables_activate_set(const struct nft_ctx *ctx, struct nft_set *set); 681 void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set, 682 struct nft_set_binding *binding, 683 enum nft_trans_phase phase); 684 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set, 685 struct nft_set_binding *binding); 686 void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set); 687 688 /** 689 * enum nft_set_extensions - set extension type IDs 690 * 691 * @NFT_SET_EXT_KEY: element key 692 * @NFT_SET_EXT_KEY_END: upper bound element key, for ranges 693 * @NFT_SET_EXT_DATA: mapping data 694 * @NFT_SET_EXT_FLAGS: element flags 695 * @NFT_SET_EXT_TIMEOUT: element timeout 696 * @NFT_SET_EXT_USERDATA: user data associated with the element 697 * @NFT_SET_EXT_EXPRESSIONS: expressions associated with the element 698 * @NFT_SET_EXT_OBJREF: stateful object reference associated with element 699 * @NFT_SET_EXT_NUM: number of extension types 700 */ 701 enum nft_set_extensions { 702 NFT_SET_EXT_KEY, 703 NFT_SET_EXT_KEY_END, 704 NFT_SET_EXT_DATA, 705 NFT_SET_EXT_FLAGS, 706 NFT_SET_EXT_TIMEOUT, 707 NFT_SET_EXT_USERDATA, 708 NFT_SET_EXT_EXPRESSIONS, 709 NFT_SET_EXT_OBJREF, 710 NFT_SET_EXT_NUM 711 }; 712 713 /** 714 * struct nft_set_ext_type - set extension type 715 * 716 * @len: fixed part length of the extension 717 * @align: alignment requirements of the extension 718 */ 719 struct nft_set_ext_type { 720 u8 len; 721 u8 align; 722 }; 723 724 extern const struct nft_set_ext_type nft_set_ext_types[]; 725 726 /** 727 * struct nft_set_ext_tmpl - set extension template 728 * 729 * @len: length of extension area 730 * @offset: offsets of individual extension types 731 * @ext_len: length of the expected extension(used to sanity check) 732 */ 733 struct nft_set_ext_tmpl { 734 u16 len; 735 u8 offset[NFT_SET_EXT_NUM]; 736 u8 ext_len[NFT_SET_EXT_NUM]; 737 }; 738 739 /** 740 * struct nft_set_ext - set extensions 741 * 742 * @genmask: generation mask, but also flags (see NFT_SET_ELEM_DEAD_BIT) 743 * @offset: offsets of individual extension types 744 * @data: beginning of extension data 745 * 746 * This structure must be aligned to word size, otherwise atomic bitops 747 * on genmask field can cause alignment failure on some archs. 748 */ 749 struct nft_set_ext { 750 u8 genmask; 751 u8 offset[NFT_SET_EXT_NUM]; 752 char data[]; 753 } __aligned(BITS_PER_LONG / 8); 754 755 static inline void nft_set_ext_prepare(struct nft_set_ext_tmpl *tmpl) 756 { 757 memset(tmpl, 0, sizeof(*tmpl)); 758 tmpl->len = sizeof(struct nft_set_ext); 759 } 760 761 static inline int nft_set_ext_add_length(struct nft_set_ext_tmpl *tmpl, u8 id, 762 unsigned int len) 763 { 764 tmpl->len = ALIGN(tmpl->len, nft_set_ext_types[id].align); 765 if (tmpl->len > U8_MAX) 766 return -EINVAL; 767 768 tmpl->offset[id] = tmpl->len; 769 tmpl->ext_len[id] = nft_set_ext_types[id].len + len; 770 tmpl->len += tmpl->ext_len[id]; 771 772 return 0; 773 } 774 775 static inline int nft_set_ext_add(struct nft_set_ext_tmpl *tmpl, u8 id) 776 { 777 return nft_set_ext_add_length(tmpl, id, 0); 778 } 779 780 static inline void nft_set_ext_init(struct nft_set_ext *ext, 781 const struct nft_set_ext_tmpl *tmpl) 782 { 783 memcpy(ext->offset, tmpl->offset, sizeof(ext->offset)); 784 } 785 786 static inline bool __nft_set_ext_exists(const struct nft_set_ext *ext, u8 id) 787 { 788 return !!ext->offset[id]; 789 } 790 791 static inline bool nft_set_ext_exists(const struct nft_set_ext *ext, u8 id) 792 { 793 return ext && __nft_set_ext_exists(ext, id); 794 } 795 796 static inline void *nft_set_ext(const struct nft_set_ext *ext, u8 id) 797 { 798 return (void *)ext + ext->offset[id]; 799 } 800 801 static inline struct nft_data *nft_set_ext_key(const struct nft_set_ext *ext) 802 { 803 return nft_set_ext(ext, NFT_SET_EXT_KEY); 804 } 805 806 static inline struct nft_data *nft_set_ext_key_end(const struct nft_set_ext *ext) 807 { 808 return nft_set_ext(ext, NFT_SET_EXT_KEY_END); 809 } 810 811 static inline struct nft_data *nft_set_ext_data(const struct nft_set_ext *ext) 812 { 813 return nft_set_ext(ext, NFT_SET_EXT_DATA); 814 } 815 816 static inline u8 *nft_set_ext_flags(const struct nft_set_ext *ext) 817 { 818 return nft_set_ext(ext, NFT_SET_EXT_FLAGS); 819 } 820 821 struct nft_timeout { 822 u64 timeout; 823 u64 expiration; 824 }; 825 826 static inline struct nft_timeout *nft_set_ext_timeout(const struct nft_set_ext *ext) 827 { 828 return nft_set_ext(ext, NFT_SET_EXT_TIMEOUT); 829 } 830 831 static inline struct nft_userdata *nft_set_ext_userdata(const struct nft_set_ext *ext) 832 { 833 return nft_set_ext(ext, NFT_SET_EXT_USERDATA); 834 } 835 836 static inline struct nft_set_elem_expr *nft_set_ext_expr(const struct nft_set_ext *ext) 837 { 838 return nft_set_ext(ext, NFT_SET_EXT_EXPRESSIONS); 839 } 840 841 static inline bool __nft_set_elem_expired(const struct nft_set_ext *ext, 842 u64 tstamp) 843 { 844 if (!nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT) || 845 READ_ONCE(nft_set_ext_timeout(ext)->timeout) == 0) 846 return false; 847 848 return time_after_eq64(tstamp, READ_ONCE(nft_set_ext_timeout(ext)->expiration)); 849 } 850 851 static inline bool nft_set_elem_expired(const struct nft_set_ext *ext) 852 { 853 return __nft_set_elem_expired(ext, get_jiffies_64()); 854 } 855 856 static inline struct nft_set_ext *nft_set_elem_ext(const struct nft_set *set, 857 const struct nft_elem_priv *elem_priv) 858 { 859 return (void *)elem_priv + set->ops->elemsize; 860 } 861 862 static inline struct nft_object **nft_set_ext_obj(const struct nft_set_ext *ext) 863 { 864 return nft_set_ext(ext, NFT_SET_EXT_OBJREF); 865 } 866 867 struct nft_expr *nft_set_elem_expr_alloc(const struct nft_ctx *ctx, 868 const struct nft_set *set, 869 const struct nlattr *attr); 870 871 struct nft_elem_priv *nft_set_elem_init(const struct nft_set *set, 872 const struct nft_set_ext_tmpl *tmpl, 873 const u32 *key, const u32 *key_end, 874 const u32 *data, 875 u64 timeout, u64 expiration, gfp_t gfp); 876 int nft_set_elem_expr_clone(const struct nft_ctx *ctx, struct nft_set *set, 877 struct nft_expr *expr_array[]); 878 void nft_set_elem_destroy(const struct nft_set *set, 879 const struct nft_elem_priv *elem_priv, 880 bool destroy_expr); 881 void nf_tables_set_elem_destroy(const struct nft_ctx *ctx, 882 const struct nft_set *set, 883 const struct nft_elem_priv *elem_priv); 884 885 struct nft_expr_ops; 886 /** 887 * struct nft_expr_type - nf_tables expression type 888 * 889 * @select_ops: function to select nft_expr_ops 890 * @release_ops: release nft_expr_ops 891 * @ops: default ops, used when no select_ops functions is present 892 * @inner_ops: inner ops, used for inner packet operation 893 * @list: used internally 894 * @name: Identifier 895 * @owner: module reference 896 * @policy: netlink attribute policy 897 * @maxattr: highest netlink attribute number 898 * @family: address family for AF-specific types 899 * @flags: expression type flags 900 */ 901 struct nft_expr_type { 902 const struct nft_expr_ops *(*select_ops)(const struct nft_ctx *, 903 const struct nlattr * const tb[]); 904 void (*release_ops)(const struct nft_expr_ops *ops); 905 const struct nft_expr_ops *ops; 906 const struct nft_expr_ops *inner_ops; 907 struct list_head list; 908 const char *name; 909 struct module *owner; 910 const struct nla_policy *policy; 911 unsigned int maxattr; 912 u8 family; 913 u8 flags; 914 }; 915 916 #define NFT_EXPR_STATEFUL 0x1 917 #define NFT_EXPR_GC 0x2 918 919 enum nft_trans_phase { 920 NFT_TRANS_PREPARE, 921 NFT_TRANS_PREPARE_ERROR, 922 NFT_TRANS_ABORT, 923 NFT_TRANS_COMMIT, 924 NFT_TRANS_RELEASE 925 }; 926 927 struct nft_flow_rule; 928 struct nft_offload_ctx; 929 930 /** 931 * struct nft_expr_ops - nf_tables expression operations 932 * 933 * @eval: Expression evaluation function 934 * @clone: Expression clone function 935 * @size: full expression size, including private data size 936 * @init: initialization function 937 * @activate: activate expression in the next generation 938 * @deactivate: deactivate expression in next generation 939 * @destroy: destruction function, called after synchronize_rcu 940 * @destroy_clone: destruction clone function 941 * @dump: function to dump parameters 942 * @validate: validate expression, called during loop detection 943 * @reduce: reduce expression 944 * @gc: garbage collection expression 945 * @offload: hardware offload expression 946 * @offload_action: function to report true/false to allocate one slot or not in the flow 947 * offload array 948 * @offload_stats: function to synchronize hardware stats via updating the counter expression 949 * @type: expression type 950 * @data: extra data to attach to this expression operation 951 */ 952 struct nft_expr_ops { 953 void (*eval)(const struct nft_expr *expr, 954 struct nft_regs *regs, 955 const struct nft_pktinfo *pkt); 956 int (*clone)(struct nft_expr *dst, 957 const struct nft_expr *src, gfp_t gfp); 958 unsigned int size; 959 960 int (*init)(const struct nft_ctx *ctx, 961 const struct nft_expr *expr, 962 const struct nlattr * const tb[]); 963 void (*activate)(const struct nft_ctx *ctx, 964 const struct nft_expr *expr); 965 void (*deactivate)(const struct nft_ctx *ctx, 966 const struct nft_expr *expr, 967 enum nft_trans_phase phase); 968 void (*destroy)(const struct nft_ctx *ctx, 969 const struct nft_expr *expr); 970 void (*destroy_clone)(const struct nft_ctx *ctx, 971 const struct nft_expr *expr); 972 int (*dump)(struct sk_buff *skb, 973 const struct nft_expr *expr, 974 bool reset); 975 int (*validate)(const struct nft_ctx *ctx, 976 const struct nft_expr *expr); 977 bool (*reduce)(struct nft_regs_track *track, 978 const struct nft_expr *expr); 979 bool (*gc)(struct net *net, 980 const struct nft_expr *expr); 981 int (*offload)(struct nft_offload_ctx *ctx, 982 struct nft_flow_rule *flow, 983 const struct nft_expr *expr); 984 bool (*offload_action)(const struct nft_expr *expr); 985 void (*offload_stats)(struct nft_expr *expr, 986 const struct flow_stats *stats); 987 const struct nft_expr_type *type; 988 void *data; 989 }; 990 991 /** 992 * struct nft_rule - nf_tables rule 993 * 994 * @list: used internally 995 * @handle: rule handle 996 * @genmask: generation mask 997 * @dlen: length of expression data 998 * @udata: user data is appended to the rule 999 * @data: expression data 1000 */ 1001 struct nft_rule { 1002 struct list_head list; 1003 u64 handle:42, 1004 genmask:2, 1005 dlen:12, 1006 udata:1; 1007 unsigned char data[] 1008 __attribute__((aligned(__alignof__(struct nft_expr)))); 1009 }; 1010 1011 static inline struct nft_expr *nft_expr_first(const struct nft_rule *rule) 1012 { 1013 return (struct nft_expr *)&rule->data[0]; 1014 } 1015 1016 static inline struct nft_expr *nft_expr_next(const struct nft_expr *expr) 1017 { 1018 return ((void *)expr) + expr->ops->size; 1019 } 1020 1021 static inline struct nft_expr *nft_expr_last(const struct nft_rule *rule) 1022 { 1023 return (struct nft_expr *)&rule->data[rule->dlen]; 1024 } 1025 1026 static inline bool nft_expr_more(const struct nft_rule *rule, 1027 const struct nft_expr *expr) 1028 { 1029 return expr != nft_expr_last(rule) && expr->ops; 1030 } 1031 1032 static inline struct nft_userdata *nft_userdata(const struct nft_rule *rule) 1033 { 1034 return (void *)&rule->data[rule->dlen]; 1035 } 1036 1037 void nft_rule_expr_activate(const struct nft_ctx *ctx, struct nft_rule *rule); 1038 void nft_rule_expr_deactivate(const struct nft_ctx *ctx, struct nft_rule *rule, 1039 enum nft_trans_phase phase); 1040 void nf_tables_rule_destroy(const struct nft_ctx *ctx, struct nft_rule *rule); 1041 1042 static inline void nft_set_elem_update_expr(const struct nft_set_ext *ext, 1043 struct nft_regs *regs, 1044 const struct nft_pktinfo *pkt) 1045 { 1046 struct nft_set_elem_expr *elem_expr; 1047 struct nft_expr *expr; 1048 u32 size; 1049 1050 if (__nft_set_ext_exists(ext, NFT_SET_EXT_EXPRESSIONS)) { 1051 elem_expr = nft_set_ext_expr(ext); 1052 nft_setelem_expr_foreach(expr, elem_expr, size) { 1053 expr->ops->eval(expr, regs, pkt); 1054 if (regs->verdict.code == NFT_BREAK) 1055 return; 1056 } 1057 } 1058 } 1059 1060 /* 1061 * The last pointer isn't really necessary, but the compiler isn't able to 1062 * determine that the result of nft_expr_last() is always the same since it 1063 * can't assume that the dlen value wasn't changed within calls in the loop. 1064 */ 1065 #define nft_rule_for_each_expr(expr, last, rule) \ 1066 for ((expr) = nft_expr_first(rule), (last) = nft_expr_last(rule); \ 1067 (expr) != (last); \ 1068 (expr) = nft_expr_next(expr)) 1069 1070 #define NFT_CHAIN_POLICY_UNSET U8_MAX 1071 1072 struct nft_rule_dp { 1073 u64 is_last:1, 1074 dlen:12, 1075 handle:42; /* for tracing */ 1076 unsigned char data[] 1077 __attribute__((aligned(__alignof__(struct nft_expr)))); 1078 }; 1079 1080 struct nft_rule_dp_last { 1081 struct nft_rule_dp end; /* end of nft_rule_blob marker */ 1082 struct rcu_head h; /* call_rcu head */ 1083 struct nft_rule_blob *blob; /* ptr to free via call_rcu */ 1084 const struct nft_chain *chain; /* for nftables tracing */ 1085 }; 1086 1087 static inline const struct nft_rule_dp *nft_rule_next(const struct nft_rule_dp *rule) 1088 { 1089 return (void *)rule + sizeof(*rule) + rule->dlen; 1090 } 1091 1092 struct nft_rule_blob { 1093 unsigned long size; 1094 unsigned char data[] 1095 __attribute__((aligned(__alignof__(struct nft_rule_dp)))); 1096 }; 1097 1098 /** 1099 * struct nft_chain - nf_tables chain 1100 * 1101 * @blob_gen_0: rule blob pointer to the current generation 1102 * @blob_gen_1: rule blob pointer to the future generation 1103 * @rules: list of rules in the chain 1104 * @list: used internally 1105 * @rhlhead: used internally 1106 * @table: table that this chain belongs to 1107 * @handle: chain handle 1108 * @use: number of jump references to this chain 1109 * @flags: bitmask of enum NFTA_CHAIN_FLAGS 1110 * @bound: bind or not 1111 * @genmask: generation mask 1112 * @name: name of the chain 1113 * @udlen: user data length 1114 * @udata: user data in the chain 1115 * @blob_next: rule blob pointer to the next in the chain 1116 */ 1117 struct nft_chain { 1118 struct nft_rule_blob __rcu *blob_gen_0; 1119 struct nft_rule_blob __rcu *blob_gen_1; 1120 struct list_head rules; 1121 struct list_head list; 1122 struct rhlist_head rhlhead; 1123 struct nft_table *table; 1124 u64 handle; 1125 u32 use; 1126 u8 flags:5, 1127 bound:1, 1128 genmask:2; 1129 char *name; 1130 u16 udlen; 1131 u8 *udata; 1132 1133 /* Only used during control plane commit phase: */ 1134 struct nft_rule_blob *blob_next; 1135 }; 1136 1137 int nft_chain_validate(const struct nft_ctx *ctx, const struct nft_chain *chain); 1138 int nft_setelem_validate(const struct nft_ctx *ctx, struct nft_set *set, 1139 const struct nft_set_iter *iter, 1140 struct nft_elem_priv *elem_priv); 1141 int nft_set_catchall_validate(const struct nft_ctx *ctx, struct nft_set *set); 1142 int nf_tables_bind_chain(const struct nft_ctx *ctx, struct nft_chain *chain); 1143 void nf_tables_unbind_chain(const struct nft_ctx *ctx, struct nft_chain *chain); 1144 1145 struct nft_hook; 1146 void nf_tables_chain_device_notify(const struct nft_chain *chain, 1147 const struct nft_hook *hook, 1148 const struct net_device *dev, int event); 1149 1150 enum nft_chain_types { 1151 NFT_CHAIN_T_DEFAULT = 0, 1152 NFT_CHAIN_T_ROUTE, 1153 NFT_CHAIN_T_NAT, 1154 NFT_CHAIN_T_MAX 1155 }; 1156 1157 /** 1158 * struct nft_chain_type - nf_tables chain type info 1159 * 1160 * @name: name of the type 1161 * @type: numeric identifier 1162 * @family: address family 1163 * @owner: module owner 1164 * @hook_mask: mask of valid hooks 1165 * @hooks: array of hook functions 1166 * @ops_register: base chain register function 1167 * @ops_unregister: base chain unregister function 1168 */ 1169 struct nft_chain_type { 1170 const char *name; 1171 enum nft_chain_types type; 1172 int family; 1173 struct module *owner; 1174 unsigned int hook_mask; 1175 nf_hookfn *hooks[NFT_MAX_HOOKS]; 1176 int (*ops_register)(struct net *net, const struct nf_hook_ops *ops); 1177 void (*ops_unregister)(struct net *net, const struct nf_hook_ops *ops); 1178 }; 1179 1180 int nft_chain_validate_dependency(const struct nft_chain *chain, 1181 enum nft_chain_types type); 1182 int nft_chain_validate_hooks(const struct nft_chain *chain, 1183 unsigned int hook_flags); 1184 1185 static inline bool nft_chain_binding(const struct nft_chain *chain) 1186 { 1187 return chain->flags & NFT_CHAIN_BINDING; 1188 } 1189 1190 static inline bool nft_chain_is_bound(struct nft_chain *chain) 1191 { 1192 return (chain->flags & NFT_CHAIN_BINDING) && chain->bound; 1193 } 1194 1195 int nft_chain_add(struct nft_table *table, struct nft_chain *chain); 1196 void nft_chain_del(struct nft_chain *chain); 1197 void nf_tables_chain_destroy(struct nft_chain *chain); 1198 1199 struct nft_stats { 1200 u64 bytes; 1201 u64 pkts; 1202 struct u64_stats_sync syncp; 1203 }; 1204 1205 struct nft_hook { 1206 struct list_head list; 1207 struct list_head ops_list; 1208 struct rcu_head rcu; 1209 char ifname[IFNAMSIZ]; 1210 u8 ifnamelen; 1211 }; 1212 1213 struct nf_hook_ops *nft_hook_find_ops(const struct nft_hook *hook, 1214 const struct net_device *dev); 1215 struct nf_hook_ops *nft_hook_find_ops_rcu(const struct nft_hook *hook, 1216 const struct net_device *dev); 1217 1218 /** 1219 * struct nft_base_chain - nf_tables base chain 1220 * 1221 * @ops: netfilter hook ops 1222 * @hook_list: list of netfilter hooks (for NFPROTO_NETDEV family) 1223 * @type: chain type 1224 * @policy: default policy 1225 * @flags: indicate the base chain disabled or not 1226 * @stats: per-cpu chain stats 1227 * @chain: the chain 1228 * @flow_block: flow block (for hardware offload) 1229 */ 1230 struct nft_base_chain { 1231 struct nf_hook_ops ops; 1232 struct list_head hook_list; 1233 const struct nft_chain_type *type; 1234 u8 policy; 1235 u8 flags; 1236 struct nft_stats __percpu *stats; 1237 struct nft_chain chain; 1238 struct flow_block flow_block; 1239 }; 1240 1241 static inline struct nft_base_chain *nft_base_chain(const struct nft_chain *chain) 1242 { 1243 return container_of(chain, struct nft_base_chain, chain); 1244 } 1245 1246 static inline bool nft_is_base_chain(const struct nft_chain *chain) 1247 { 1248 return chain->flags & NFT_CHAIN_BASE; 1249 } 1250 1251 unsigned int nft_do_chain(struct nft_pktinfo *pkt, void *priv); 1252 1253 static inline bool nft_use_inc(u32 *use) 1254 { 1255 if (*use == UINT_MAX) 1256 return false; 1257 1258 (*use)++; 1259 1260 return true; 1261 } 1262 1263 static inline void nft_use_dec(u32 *use) 1264 { 1265 WARN_ON_ONCE((*use)-- == 0); 1266 } 1267 1268 /* For error and abort path: restore use counter to previous state. */ 1269 static inline void nft_use_inc_restore(u32 *use) 1270 { 1271 WARN_ON_ONCE(!nft_use_inc(use)); 1272 } 1273 1274 #define nft_use_dec_restore nft_use_dec 1275 1276 /** 1277 * struct nft_table - nf_tables table 1278 * 1279 * @list: used internally 1280 * @chains_ht: chains in the table 1281 * @chains: same, for stable walks 1282 * @sets: sets in the table 1283 * @objects: stateful objects in the table 1284 * @flowtables: flow tables in the table 1285 * @hgenerator: handle generator state 1286 * @handle: table handle 1287 * @use: number of chain references to this table 1288 * @family:address family 1289 * @flags: table flag (see enum nft_table_flags) 1290 * @genmask: generation mask 1291 * @nlpid: netlink port ID 1292 * @name: name of the table 1293 * @udlen: length of the user data 1294 * @udata: user data 1295 * @validate_state: internal, set when transaction adds jumps 1296 */ 1297 struct nft_table { 1298 struct list_head list; 1299 struct rhltable chains_ht; 1300 struct list_head chains; 1301 struct list_head sets; 1302 struct list_head objects; 1303 struct list_head flowtables; 1304 u64 hgenerator; 1305 u64 handle; 1306 u32 use; 1307 u16 family:6, 1308 flags:8, 1309 genmask:2; 1310 u32 nlpid; 1311 char *name; 1312 u16 udlen; 1313 u8 *udata; 1314 u8 validate_state; 1315 }; 1316 1317 static inline bool nft_table_has_owner(const struct nft_table *table) 1318 { 1319 return table->flags & NFT_TABLE_F_OWNER; 1320 } 1321 1322 static inline bool nft_table_is_orphan(const struct nft_table *table) 1323 { 1324 return (table->flags & (NFT_TABLE_F_OWNER | NFT_TABLE_F_PERSIST)) == 1325 NFT_TABLE_F_PERSIST; 1326 } 1327 1328 static inline bool nft_base_chain_netdev(int family, u32 hooknum) 1329 { 1330 return family == NFPROTO_NETDEV || 1331 (family == NFPROTO_INET && hooknum == NF_INET_INGRESS); 1332 } 1333 1334 void nft_register_chain_type(const struct nft_chain_type *); 1335 void nft_unregister_chain_type(const struct nft_chain_type *); 1336 1337 int nft_register_expr(struct nft_expr_type *); 1338 void nft_unregister_expr(struct nft_expr_type *); 1339 1340 int nft_verdict_dump(struct sk_buff *skb, int type, 1341 const struct nft_verdict *v); 1342 1343 /** 1344 * struct nft_object_hash_key - key to lookup nft_object 1345 * 1346 * @name: name of the stateful object to look up 1347 * @table: table the object belongs to 1348 */ 1349 struct nft_object_hash_key { 1350 const char *name; 1351 const struct nft_table *table; 1352 }; 1353 1354 /** 1355 * struct nft_object - nf_tables stateful object 1356 * 1357 * @list: table stateful object list node 1358 * @rhlhead: nft_objname_ht node 1359 * @key: keys that identify this object 1360 * @genmask: generation mask 1361 * @use: number of references to this stateful object 1362 * @handle: unique object handle 1363 * @udlen: length of user data 1364 * @udata: user data 1365 * @ops: object operations 1366 * @data: object data, layout depends on type 1367 */ 1368 struct nft_object { 1369 struct list_head list; 1370 struct rhlist_head rhlhead; 1371 struct nft_object_hash_key key; 1372 u32 genmask:2; 1373 u32 use; 1374 u64 handle; 1375 u16 udlen; 1376 u8 *udata; 1377 /* runtime data below here */ 1378 const struct nft_object_ops *ops ____cacheline_aligned; 1379 unsigned char data[] 1380 __attribute__((aligned(__alignof__(u64)))); 1381 }; 1382 1383 static inline void *nft_obj_data(const struct nft_object *obj) 1384 { 1385 return (void *)obj->data; 1386 } 1387 1388 #define nft_expr_obj(expr) *((struct nft_object **)nft_expr_priv(expr)) 1389 1390 struct nft_object *nft_obj_lookup(const struct net *net, 1391 const struct nft_table *table, 1392 const struct nlattr *nla, u32 objtype, 1393 u8 genmask); 1394 1395 void nft_obj_notify(struct net *net, const struct nft_table *table, 1396 struct nft_object *obj, u32 portid, u32 seq, 1397 int event, u16 flags, int family, int report, gfp_t gfp); 1398 1399 /** 1400 * struct nft_object_type - stateful object type 1401 * 1402 * @select_ops: function to select nft_object_ops 1403 * @ops: default ops, used when no select_ops functions is present 1404 * @list: list node in list of object types 1405 * @type: stateful object numeric type 1406 * @owner: module owner 1407 * @maxattr: maximum netlink attribute 1408 * @family: address family for AF-specific object types 1409 * @policy: netlink attribute policy 1410 */ 1411 struct nft_object_type { 1412 const struct nft_object_ops *(*select_ops)(const struct nft_ctx *, 1413 const struct nlattr * const tb[]); 1414 const struct nft_object_ops *ops; 1415 struct list_head list; 1416 u32 type; 1417 unsigned int maxattr; 1418 u8 family; 1419 struct module *owner; 1420 const struct nla_policy *policy; 1421 }; 1422 1423 /** 1424 * struct nft_object_ops - stateful object operations 1425 * 1426 * @eval: stateful object evaluation function 1427 * @size: stateful object size 1428 * @init: initialize object from netlink attributes 1429 * @destroy: release existing stateful object 1430 * @dump: netlink dump stateful object 1431 * @update: update stateful object 1432 * @type: pointer to object type 1433 */ 1434 struct nft_object_ops { 1435 void (*eval)(struct nft_object *obj, 1436 struct nft_regs *regs, 1437 const struct nft_pktinfo *pkt); 1438 unsigned int size; 1439 int (*init)(const struct nft_ctx *ctx, 1440 const struct nlattr *const tb[], 1441 struct nft_object *obj); 1442 void (*destroy)(const struct nft_ctx *ctx, 1443 struct nft_object *obj); 1444 int (*dump)(struct sk_buff *skb, 1445 struct nft_object *obj, 1446 bool reset); 1447 void (*update)(struct nft_object *obj, 1448 struct nft_object *newobj); 1449 const struct nft_object_type *type; 1450 }; 1451 1452 int nft_register_obj(struct nft_object_type *obj_type); 1453 void nft_unregister_obj(struct nft_object_type *obj_type); 1454 1455 #define NFT_NETDEVICE_MAX 256 1456 1457 /** 1458 * struct nft_flowtable - nf_tables flow table 1459 * 1460 * @list: flow table list node in table list 1461 * @table: the table the flow table is contained in 1462 * @name: name of this flow table 1463 * @hooknum: hook number 1464 * @ops_len: number of hooks in array 1465 * @genmask: generation mask 1466 * @use: number of references to this flow table 1467 * @handle: unique object handle 1468 * @hook_list: hook list for hooks per net_device in flowtables 1469 * @data: rhashtable and garbage collector 1470 */ 1471 struct nft_flowtable { 1472 struct list_head list; 1473 struct nft_table *table; 1474 char *name; 1475 int hooknum; 1476 int ops_len; 1477 u32 genmask:2; 1478 u32 use; 1479 u64 handle; 1480 /* runtime data below here */ 1481 struct list_head hook_list ____cacheline_aligned; 1482 struct nf_flowtable data; 1483 }; 1484 1485 struct nft_flowtable *nft_flowtable_lookup(const struct net *net, 1486 const struct nft_table *table, 1487 const struct nlattr *nla, 1488 u8 genmask); 1489 1490 void nf_tables_deactivate_flowtable(const struct nft_ctx *ctx, 1491 struct nft_flowtable *flowtable, 1492 enum nft_trans_phase phase); 1493 1494 void nft_register_flowtable_type(struct nf_flowtable_type *type); 1495 void nft_unregister_flowtable_type(struct nf_flowtable_type *type); 1496 1497 /** 1498 * struct nft_traceinfo - nft tracing information and state 1499 * 1500 * @trace: other struct members are initialised 1501 * @nf_trace: copy of skb->nf_trace before rule evaluation 1502 * @type: event type (enum nft_trace_types) 1503 * @skbid: hash of skb to be used as trace id 1504 * @packet_dumped: packet headers sent in a previous traceinfo message 1505 * @basechain: base chain currently processed 1506 */ 1507 struct nft_traceinfo { 1508 bool trace; 1509 bool nf_trace; 1510 bool packet_dumped; 1511 enum nft_trace_types type:8; 1512 u32 skbid; 1513 const struct nft_base_chain *basechain; 1514 }; 1515 1516 void nft_trace_init(struct nft_traceinfo *info, const struct nft_pktinfo *pkt, 1517 const struct nft_chain *basechain); 1518 1519 void nft_trace_notify(const struct nft_pktinfo *pkt, 1520 const struct nft_verdict *verdict, 1521 const struct nft_rule_dp *rule, 1522 struct nft_traceinfo *info); 1523 1524 #define MODULE_ALIAS_NFT_CHAIN(family, name) \ 1525 MODULE_ALIAS("nft-chain-" __stringify(family) "-" name) 1526 1527 #define MODULE_ALIAS_NFT_AF_EXPR(family, name) \ 1528 MODULE_ALIAS("nft-expr-" __stringify(family) "-" name) 1529 1530 #define MODULE_ALIAS_NFT_EXPR(name) \ 1531 MODULE_ALIAS("nft-expr-" name) 1532 1533 #define MODULE_ALIAS_NFT_OBJ(type) \ 1534 MODULE_ALIAS("nft-obj-" __stringify(type)) 1535 1536 #if IS_ENABLED(CONFIG_NF_TABLES) 1537 1538 /* 1539 * The gencursor defines two generations, the currently active and the 1540 * next one. Objects contain a bitmask of 2 bits specifying the generations 1541 * they're active in. A set bit means they're inactive in the generation 1542 * represented by that bit. 1543 * 1544 * New objects start out as inactive in the current and active in the 1545 * next generation. When committing the ruleset the bitmask is cleared, 1546 * meaning they're active in all generations. When removing an object, 1547 * it is set inactive in the next generation. After committing the ruleset, 1548 * the objects are removed. 1549 */ 1550 static inline unsigned int nft_gencursor_next(const struct net *net) 1551 { 1552 return net->nft.gencursor + 1 == 1 ? 1 : 0; 1553 } 1554 1555 static inline u8 nft_genmask_next(const struct net *net) 1556 { 1557 return 1 << nft_gencursor_next(net); 1558 } 1559 1560 static inline u8 nft_genmask_cur(const struct net *net) 1561 { 1562 /* Use READ_ONCE() to prevent refetching the value for atomicity */ 1563 return 1 << READ_ONCE(net->nft.gencursor); 1564 } 1565 1566 #define NFT_GENMASK_ANY ((1 << 0) | (1 << 1)) 1567 1568 /* 1569 * Generic transaction helpers 1570 */ 1571 1572 /* Check if this object is currently active. */ 1573 #define nft_is_active(__net, __obj) \ 1574 (((__obj)->genmask & nft_genmask_cur(__net)) == 0) 1575 1576 /* Check if this object is active in the next generation. */ 1577 #define nft_is_active_next(__net, __obj) \ 1578 (((__obj)->genmask & nft_genmask_next(__net)) == 0) 1579 1580 /* This object becomes active in the next generation. */ 1581 #define nft_activate_next(__net, __obj) \ 1582 (__obj)->genmask = nft_genmask_cur(__net) 1583 1584 /* This object becomes inactive in the next generation. */ 1585 #define nft_deactivate_next(__net, __obj) \ 1586 (__obj)->genmask = nft_genmask_next(__net) 1587 1588 /* After committing the ruleset, clear the stale generation bit. */ 1589 #define nft_clear(__net, __obj) \ 1590 (__obj)->genmask &= ~nft_genmask_next(__net) 1591 #define nft_active_genmask(__obj, __genmask) \ 1592 !((__obj)->genmask & __genmask) 1593 1594 /* 1595 * Set element transaction helpers 1596 */ 1597 1598 static inline bool nft_set_elem_active(const struct nft_set_ext *ext, 1599 u8 genmask) 1600 { 1601 return !(ext->genmask & genmask); 1602 } 1603 1604 static inline void nft_set_elem_change_active(const struct net *net, 1605 const struct nft_set *set, 1606 struct nft_set_ext *ext) 1607 { 1608 ext->genmask ^= nft_genmask_next(net); 1609 } 1610 1611 #endif /* IS_ENABLED(CONFIG_NF_TABLES) */ 1612 1613 #define NFT_SET_ELEM_DEAD_MASK (1 << 2) 1614 1615 #if defined(__LITTLE_ENDIAN_BITFIELD) 1616 #define NFT_SET_ELEM_DEAD_BIT 2 1617 #elif defined(__BIG_ENDIAN_BITFIELD) 1618 #define NFT_SET_ELEM_DEAD_BIT (BITS_PER_LONG - BITS_PER_BYTE + 2) 1619 #else 1620 #error 1621 #endif 1622 1623 static inline void nft_set_elem_dead(struct nft_set_ext *ext) 1624 { 1625 unsigned long *word = (unsigned long *)ext; 1626 1627 BUILD_BUG_ON(offsetof(struct nft_set_ext, genmask) != 0); 1628 set_bit(NFT_SET_ELEM_DEAD_BIT, word); 1629 } 1630 1631 static inline int nft_set_elem_is_dead(const struct nft_set_ext *ext) 1632 { 1633 unsigned long *word = (unsigned long *)ext; 1634 1635 BUILD_BUG_ON(offsetof(struct nft_set_ext, genmask) != 0); 1636 return test_bit(NFT_SET_ELEM_DEAD_BIT, word); 1637 } 1638 1639 /** 1640 * struct nft_trans - nf_tables object update in transaction 1641 * 1642 * @list: used internally 1643 * @net: struct net 1644 * @table: struct nft_table the object resides in 1645 * @msg_type: message type 1646 * @seq: netlink sequence number 1647 * @flags: modifiers to new request 1648 * @report: notify via unicast netlink message 1649 * @put_net: net needs to be put 1650 * 1651 * This is the information common to all objects in the transaction, 1652 * this must always be the first member of derived sub-types. 1653 */ 1654 struct nft_trans { 1655 struct list_head list; 1656 struct net *net; 1657 struct nft_table *table; 1658 int msg_type; 1659 u32 seq; 1660 u16 flags; 1661 u8 report:1; 1662 u8 put_net:1; 1663 }; 1664 1665 /** 1666 * struct nft_trans_binding - nf_tables object with binding support in transaction 1667 * @nft_trans: base structure, MUST be first member 1668 * @binding_list: list of objects with possible bindings 1669 * 1670 * This is the base type used by objects that can be bound to a chain. 1671 */ 1672 struct nft_trans_binding { 1673 struct nft_trans nft_trans; 1674 struct list_head binding_list; 1675 }; 1676 1677 struct nft_trans_rule { 1678 struct nft_trans nft_trans; 1679 struct nft_rule *rule; 1680 struct nft_chain *chain; 1681 struct nft_flow_rule *flow; 1682 u32 rule_id; 1683 bool bound; 1684 }; 1685 1686 #define nft_trans_container_rule(trans) \ 1687 container_of(trans, struct nft_trans_rule, nft_trans) 1688 #define nft_trans_rule(trans) \ 1689 nft_trans_container_rule(trans)->rule 1690 #define nft_trans_flow_rule(trans) \ 1691 nft_trans_container_rule(trans)->flow 1692 #define nft_trans_rule_id(trans) \ 1693 nft_trans_container_rule(trans)->rule_id 1694 #define nft_trans_rule_bound(trans) \ 1695 nft_trans_container_rule(trans)->bound 1696 #define nft_trans_rule_chain(trans) \ 1697 nft_trans_container_rule(trans)->chain 1698 1699 struct nft_trans_set { 1700 struct nft_trans_binding nft_trans_binding; 1701 struct list_head list_trans_newset; 1702 struct nft_set *set; 1703 u32 set_id; 1704 u32 gc_int; 1705 u64 timeout; 1706 bool update; 1707 bool bound; 1708 u32 size; 1709 }; 1710 1711 #define nft_trans_container_set(t) \ 1712 container_of(t, struct nft_trans_set, nft_trans_binding.nft_trans) 1713 #define nft_trans_set(trans) \ 1714 nft_trans_container_set(trans)->set 1715 #define nft_trans_set_id(trans) \ 1716 nft_trans_container_set(trans)->set_id 1717 #define nft_trans_set_bound(trans) \ 1718 nft_trans_container_set(trans)->bound 1719 #define nft_trans_set_update(trans) \ 1720 nft_trans_container_set(trans)->update 1721 #define nft_trans_set_timeout(trans) \ 1722 nft_trans_container_set(trans)->timeout 1723 #define nft_trans_set_gc_int(trans) \ 1724 nft_trans_container_set(trans)->gc_int 1725 #define nft_trans_set_size(trans) \ 1726 nft_trans_container_set(trans)->size 1727 1728 struct nft_trans_chain { 1729 struct nft_trans_binding nft_trans_binding; 1730 struct nft_chain *chain; 1731 char *name; 1732 struct nft_stats __percpu *stats; 1733 u8 policy; 1734 bool update; 1735 bool bound; 1736 u32 chain_id; 1737 struct nft_base_chain *basechain; 1738 struct list_head hook_list; 1739 }; 1740 1741 #define nft_trans_container_chain(t) \ 1742 container_of(t, struct nft_trans_chain, nft_trans_binding.nft_trans) 1743 #define nft_trans_chain(trans) \ 1744 nft_trans_container_chain(trans)->chain 1745 #define nft_trans_chain_update(trans) \ 1746 nft_trans_container_chain(trans)->update 1747 #define nft_trans_chain_name(trans) \ 1748 nft_trans_container_chain(trans)->name 1749 #define nft_trans_chain_stats(trans) \ 1750 nft_trans_container_chain(trans)->stats 1751 #define nft_trans_chain_policy(trans) \ 1752 nft_trans_container_chain(trans)->policy 1753 #define nft_trans_chain_bound(trans) \ 1754 nft_trans_container_chain(trans)->bound 1755 #define nft_trans_chain_id(trans) \ 1756 nft_trans_container_chain(trans)->chain_id 1757 #define nft_trans_basechain(trans) \ 1758 nft_trans_container_chain(trans)->basechain 1759 #define nft_trans_chain_hooks(trans) \ 1760 nft_trans_container_chain(trans)->hook_list 1761 1762 struct nft_trans_table { 1763 struct nft_trans nft_trans; 1764 bool update; 1765 }; 1766 1767 #define nft_trans_container_table(trans) \ 1768 container_of(trans, struct nft_trans_table, nft_trans) 1769 #define nft_trans_table_update(trans) \ 1770 nft_trans_container_table(trans)->update 1771 1772 enum nft_trans_elem_flags { 1773 NFT_TRANS_UPD_TIMEOUT = (1 << 0), 1774 NFT_TRANS_UPD_EXPIRATION = (1 << 1), 1775 }; 1776 1777 struct nft_elem_update { 1778 u64 timeout; 1779 u64 expiration; 1780 u8 flags; 1781 }; 1782 1783 struct nft_trans_one_elem { 1784 struct nft_elem_priv *priv; 1785 struct nft_elem_update *update; 1786 }; 1787 1788 struct nft_trans_elem { 1789 struct nft_trans nft_trans; 1790 struct nft_set *set; 1791 bool bound; 1792 unsigned int nelems; 1793 struct nft_trans_one_elem elems[] __counted_by(nelems); 1794 }; 1795 1796 #define nft_trans_container_elem(t) \ 1797 container_of(t, struct nft_trans_elem, nft_trans) 1798 #define nft_trans_elem_set(trans) \ 1799 nft_trans_container_elem(trans)->set 1800 #define nft_trans_elem_set_bound(trans) \ 1801 nft_trans_container_elem(trans)->bound 1802 1803 struct nft_trans_obj { 1804 struct nft_trans nft_trans; 1805 struct nft_object *obj; 1806 struct nft_object *newobj; 1807 bool update; 1808 }; 1809 1810 #define nft_trans_container_obj(t) \ 1811 container_of(t, struct nft_trans_obj, nft_trans) 1812 #define nft_trans_obj(trans) \ 1813 nft_trans_container_obj(trans)->obj 1814 #define nft_trans_obj_newobj(trans) \ 1815 nft_trans_container_obj(trans)->newobj 1816 #define nft_trans_obj_update(trans) \ 1817 nft_trans_container_obj(trans)->update 1818 1819 struct nft_trans_flowtable { 1820 struct nft_trans nft_trans; 1821 struct nft_flowtable *flowtable; 1822 struct list_head hook_list; 1823 u32 flags; 1824 bool update; 1825 }; 1826 1827 #define nft_trans_container_flowtable(t) \ 1828 container_of(t, struct nft_trans_flowtable, nft_trans) 1829 #define nft_trans_flowtable(trans) \ 1830 nft_trans_container_flowtable(trans)->flowtable 1831 #define nft_trans_flowtable_update(trans) \ 1832 nft_trans_container_flowtable(trans)->update 1833 #define nft_trans_flowtable_hooks(trans) \ 1834 nft_trans_container_flowtable(trans)->hook_list 1835 #define nft_trans_flowtable_flags(trans) \ 1836 nft_trans_container_flowtable(trans)->flags 1837 1838 #define NFT_TRANS_GC_BATCHCOUNT 256 1839 1840 struct nft_trans_gc { 1841 struct list_head list; 1842 struct net *net; 1843 struct nft_set *set; 1844 u32 seq; 1845 u16 count; 1846 struct nft_elem_priv *priv[NFT_TRANS_GC_BATCHCOUNT]; 1847 struct rcu_head rcu; 1848 }; 1849 1850 static inline void nft_ctx_update(struct nft_ctx *ctx, 1851 const struct nft_trans *trans) 1852 { 1853 switch (trans->msg_type) { 1854 case NFT_MSG_NEWRULE: 1855 case NFT_MSG_DELRULE: 1856 case NFT_MSG_DESTROYRULE: 1857 ctx->chain = nft_trans_rule_chain(trans); 1858 break; 1859 case NFT_MSG_NEWCHAIN: 1860 case NFT_MSG_DELCHAIN: 1861 case NFT_MSG_DESTROYCHAIN: 1862 ctx->chain = nft_trans_chain(trans); 1863 break; 1864 default: 1865 ctx->chain = NULL; 1866 break; 1867 } 1868 1869 ctx->net = trans->net; 1870 ctx->table = trans->table; 1871 ctx->family = trans->table->family; 1872 ctx->report = trans->report; 1873 ctx->flags = trans->flags; 1874 ctx->seq = trans->seq; 1875 } 1876 1877 struct nft_trans_gc *nft_trans_gc_alloc(struct nft_set *set, 1878 unsigned int gc_seq, gfp_t gfp); 1879 void nft_trans_gc_destroy(struct nft_trans_gc *trans); 1880 1881 struct nft_trans_gc *nft_trans_gc_queue_async(struct nft_trans_gc *gc, 1882 unsigned int gc_seq, gfp_t gfp); 1883 void nft_trans_gc_queue_async_done(struct nft_trans_gc *gc); 1884 1885 struct nft_trans_gc *nft_trans_gc_queue_sync(struct nft_trans_gc *gc, gfp_t gfp); 1886 void nft_trans_gc_queue_sync_done(struct nft_trans_gc *trans); 1887 1888 void nft_trans_gc_elem_add(struct nft_trans_gc *gc, void *priv); 1889 1890 struct nft_trans_gc *nft_trans_gc_catchall_async(struct nft_trans_gc *gc, 1891 unsigned int gc_seq); 1892 struct nft_trans_gc *nft_trans_gc_catchall_sync(struct nft_trans_gc *gc); 1893 1894 void nft_setelem_data_deactivate(const struct net *net, 1895 const struct nft_set *set, 1896 struct nft_elem_priv *elem_priv); 1897 1898 int __init nft_chain_filter_init(void); 1899 void nft_chain_filter_fini(void); 1900 1901 void __init nft_chain_route_init(void); 1902 void nft_chain_route_fini(void); 1903 1904 void nf_tables_trans_destroy_flush_work(struct net *net); 1905 1906 int nf_msecs_to_jiffies64(const struct nlattr *nla, u64 *result); 1907 __be64 nf_jiffies64_to_msecs(u64 input); 1908 1909 #ifdef CONFIG_MODULES 1910 __printf(2, 3) int nft_request_module(struct net *net, const char *fmt, ...); 1911 #else 1912 static inline int nft_request_module(struct net *net, const char *fmt, ...) { return -ENOENT; } 1913 #endif 1914 1915 struct nftables_pernet { 1916 struct list_head tables; 1917 struct list_head commit_list; 1918 struct list_head destroy_list; 1919 struct list_head commit_set_list; 1920 struct list_head binding_list; 1921 struct list_head module_list; 1922 struct list_head notify_list; 1923 struct mutex commit_mutex; 1924 u64 table_handle; 1925 u64 tstamp; 1926 unsigned int base_seq; 1927 unsigned int gc_seq; 1928 u8 validate_state; 1929 struct work_struct destroy_work; 1930 }; 1931 1932 extern unsigned int nf_tables_net_id; 1933 1934 static inline struct nftables_pernet *nft_pernet(const struct net *net) 1935 { 1936 return net_generic(net, nf_tables_net_id); 1937 } 1938 1939 static inline u64 nft_net_tstamp(const struct net *net) 1940 { 1941 return nft_pernet(net)->tstamp; 1942 } 1943 1944 #define __NFT_REDUCE_READONLY 1UL 1945 #define NFT_REDUCE_READONLY (void *)__NFT_REDUCE_READONLY 1946 1947 static inline bool nft_reduce_is_readonly(const struct nft_expr *expr) 1948 { 1949 return expr->ops->reduce == NFT_REDUCE_READONLY; 1950 } 1951 1952 void nft_reg_track_update(struct nft_regs_track *track, 1953 const struct nft_expr *expr, u8 dreg, u8 len); 1954 void nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg, u8 len); 1955 void __nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg); 1956 1957 static inline bool nft_reg_track_cmp(struct nft_regs_track *track, 1958 const struct nft_expr *expr, u8 dreg) 1959 { 1960 return track->regs[dreg].selector && 1961 track->regs[dreg].selector->ops == expr->ops && 1962 track->regs[dreg].num_reg == 0; 1963 } 1964 1965 #endif /* _NET_NF_TABLES_H */ 1966