1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _NET_NF_TABLES_H
3 #define _NET_NF_TABLES_H
5 #include <asm/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>
18 #define NFT_MAX_HOOKS (NF_INET_INGRESS + 1)
22 #define NFT_JUMP_STACK_SIZE 16
25 NFT_PKTINFO_L4PROTO = (1 << 0),
26 NFT_PKTINFO_INNER = (1 << 1),
27 NFT_PKTINFO_INNER_FULL = (1 << 2),
32 const struct nf_hook_state *state;
40 static inline struct sock *nft_sk(const struct nft_pktinfo *pkt)
42 return pkt->state->sk;
45 static inline unsigned int nft_thoff(const struct nft_pktinfo *pkt)
50 static inline struct net *nft_net(const struct nft_pktinfo *pkt)
52 return pkt->state->net;
55 static inline unsigned int nft_hook(const struct nft_pktinfo *pkt)
57 return pkt->state->hook;
60 static inline u8 nft_pf(const struct nft_pktinfo *pkt)
62 return pkt->state->pf;
65 static inline const struct net_device *nft_in(const struct nft_pktinfo *pkt)
67 return pkt->state->in;
70 static inline const struct net_device *nft_out(const struct nft_pktinfo *pkt)
72 return pkt->state->out;
75 static inline void nft_set_pktinfo(struct nft_pktinfo *pkt,
77 const struct nf_hook_state *state)
83 static inline void nft_set_pktinfo_unspec(struct nft_pktinfo *pkt)
92 * struct nft_verdict - nf_tables verdict
94 * @code: nf_tables/netfilter verdict code
95 * @chain: destination chain for NFT_JUMP/NFT_GOTO
99 struct nft_chain *chain;
105 struct nft_verdict verdict;
107 } __attribute__((aligned(__alignof__(u64))));
109 #define NFT_REG32_NUM 20
112 * struct nft_regs - nf_tables register set
114 * @data: data registers
115 * @verdict: verdict register
117 * The first four data registers alias to the verdict register.
121 u32 data[NFT_REG32_NUM];
122 struct nft_verdict verdict;
126 struct nft_regs_track {
128 const struct nft_expr *selector;
129 const struct nft_expr *bitwise;
131 } regs[NFT_REG32_NUM];
133 const struct nft_expr *cur;
134 const struct nft_expr *last;
137 /* Store/load an u8, u16 or u64 integer to/from the u32 data register.
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
144 static inline void nft_reg_store8(u32 *dreg, u8 val)
150 static inline u8 nft_reg_load8(const u32 *sreg)
155 static inline void nft_reg_store16(u32 *dreg, u16 val)
161 static inline void nft_reg_store_be16(u32 *dreg, __be16 val)
163 nft_reg_store16(dreg, (__force __u16)val);
166 static inline u16 nft_reg_load16(const u32 *sreg)
171 static inline __be16 nft_reg_load_be16(const u32 *sreg)
173 return (__force __be16)nft_reg_load16(sreg);
176 static inline __be32 nft_reg_load_be32(const u32 *sreg)
178 return *(__force __be32 *)sreg;
181 static inline void nft_reg_store64(u32 *dreg, u64 val)
183 put_unaligned(val, (u64 *)dreg);
186 static inline u64 nft_reg_load64(const u32 *sreg)
188 return get_unaligned((u64 *)sreg);
191 static inline void nft_data_copy(u32 *dst, const struct nft_data *src,
194 if (len % NFT_REG32_SIZE)
195 dst[len / NFT_REG32_SIZE] = 0;
196 memcpy(dst, src, len);
200 * struct nft_ctx - nf_tables rule/set context
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 * @family: protocol family
209 * @level: depth of the chains
210 * @report: notify via unicast netlink message
214 struct nft_table *table;
215 struct nft_chain *chain;
216 const struct nlattr * const *nla;
225 enum nft_data_desc_flags {
226 NFT_DATA_DESC_SETELEM = (1 << 0),
229 struct nft_data_desc {
230 enum nft_data_types type;
236 int nft_data_init(const struct nft_ctx *ctx, struct nft_data *data,
237 struct nft_data_desc *desc, const struct nlattr *nla);
238 void nft_data_hold(const struct nft_data *data, enum nft_data_types type);
239 void nft_data_release(const struct nft_data *data, enum nft_data_types type);
240 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data,
241 enum nft_data_types type, unsigned int len);
243 static inline enum nft_data_types nft_dreg_to_type(enum nft_registers reg)
245 return reg == NFT_REG_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE;
248 static inline enum nft_registers nft_type_to_reg(enum nft_data_types type)
250 return type == NFT_DATA_VERDICT ? NFT_REG_VERDICT : NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE;
253 int nft_parse_u32_check(const struct nlattr *attr, int max, u32 *dest);
254 int nft_dump_register(struct sk_buff *skb, unsigned int attr, unsigned int reg);
256 int nft_parse_register_load(const struct nlattr *attr, u8 *sreg, u32 len);
257 int nft_parse_register_store(const struct nft_ctx *ctx,
258 const struct nlattr *attr, u8 *dreg,
259 const struct nft_data *data,
260 enum nft_data_types type, unsigned int len);
263 * struct nft_userdata - user defined data associated with an object
265 * @len: length of the data
268 * The presence of user data is indicated in an object specific fashion,
269 * so a length of zero can't occur and the value "len" indicates data
272 struct nft_userdata {
274 unsigned char data[];
278 * struct nft_set_elem - generic representation of set elements
281 * @key_end: closing element key
282 * @priv: element private data and extensions
284 struct nft_set_elem {
286 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)];
290 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)];
294 u32 buf[NFT_DATA_VALUE_MAXLEN / sizeof(u32)];
301 struct nft_set_iter {
306 int (*fn)(const struct nft_ctx *ctx,
308 const struct nft_set_iter *iter,
309 struct nft_set_elem *elem);
313 * struct nft_set_desc - description of set elements
319 * @objtype: object type
321 * @size: number of set elements
322 * @policy: set policy
323 * @gc_int: garbage collector interval
324 * @field_len: length of each field in concatenation, bytes
325 * @field_count: number of concatenated fields in element
326 * @expr: set must support for expressions
328 struct nft_set_desc {
338 u8 field_len[NFT_REG32_COUNT];
344 * enum nft_set_class - performance class
346 * @NFT_LOOKUP_O_1: constant, O(1)
347 * @NFT_LOOKUP_O_LOG_N: logarithmic, O(log N)
348 * @NFT_LOOKUP_O_N: linear, O(N)
352 NFT_SET_CLASS_O_LOG_N,
357 * struct nft_set_estimate - estimation of memory and performance
360 * @size: required memory
361 * @lookup: lookup performance class
362 * @space: memory class
364 struct nft_set_estimate {
366 enum nft_set_class lookup;
367 enum nft_set_class space;
370 #define NFT_EXPR_MAXATTR 16
371 #define NFT_EXPR_SIZE(size) (sizeof(struct nft_expr) + \
372 ALIGN(size, __alignof__(struct nft_expr)))
375 * struct nft_expr - nf_tables expression
377 * @ops: expression ops
378 * @data: expression private data
381 const struct nft_expr_ops *ops;
383 __attribute__((aligned(__alignof__(u64))));
386 static inline void *nft_expr_priv(const struct nft_expr *expr)
388 return (void *)expr->data;
391 struct nft_expr_info;
393 int nft_expr_inner_parse(const struct nft_ctx *ctx, const struct nlattr *nla,
394 struct nft_expr_info *info);
395 int nft_expr_clone(struct nft_expr *dst, struct nft_expr *src);
396 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr);
397 int nft_expr_dump(struct sk_buff *skb, unsigned int attr,
398 const struct nft_expr *expr, bool reset);
399 bool nft_expr_reduce_bitwise(struct nft_regs_track *track,
400 const struct nft_expr *expr);
405 * struct nft_set_ops - nf_tables set operations
407 * @lookup: look up an element within the set
408 * @update: update an element if exists, add it if doesn't exist
409 * @delete: delete an element
410 * @insert: insert new element into set
411 * @activate: activate new element in the next generation
412 * @deactivate: lookup for element and deactivate it in the next generation
413 * @flush: deactivate element in the next generation
414 * @remove: remove element from set
415 * @walk: iterate over all set elements
416 * @get: get set elements
417 * @privsize: function to return size of set private data
418 * @init: initialize private data of new set instance
419 * @destroy: destroy private data of set instance
420 * @elemsize: element private size
422 * Operations lookup, update and delete have simpler interfaces, are faster
423 * and currently only used in the packet path. All the rest are slower,
424 * control plane functions.
427 bool (*lookup)(const struct net *net,
428 const struct nft_set *set,
430 const struct nft_set_ext **ext);
431 bool (*update)(struct nft_set *set,
433 void *(*new)(struct nft_set *,
434 const struct nft_expr *,
436 const struct nft_expr *expr,
437 struct nft_regs *regs,
438 const struct nft_set_ext **ext);
439 bool (*delete)(const struct nft_set *set,
442 int (*insert)(const struct net *net,
443 const struct nft_set *set,
444 const struct nft_set_elem *elem,
445 struct nft_set_ext **ext);
446 void (*activate)(const struct net *net,
447 const struct nft_set *set,
448 const struct nft_set_elem *elem);
449 void * (*deactivate)(const struct net *net,
450 const struct nft_set *set,
451 const struct nft_set_elem *elem);
452 bool (*flush)(const struct net *net,
453 const struct nft_set *set,
455 void (*remove)(const struct net *net,
456 const struct nft_set *set,
457 const struct nft_set_elem *elem);
458 void (*walk)(const struct nft_ctx *ctx,
460 struct nft_set_iter *iter);
461 void * (*get)(const struct net *net,
462 const struct nft_set *set,
463 const struct nft_set_elem *elem,
466 u64 (*privsize)(const struct nlattr * const nla[],
467 const struct nft_set_desc *desc);
468 bool (*estimate)(const struct nft_set_desc *desc,
470 struct nft_set_estimate *est);
471 int (*init)(const struct nft_set *set,
472 const struct nft_set_desc *desc,
473 const struct nlattr * const nla[]);
474 void (*destroy)(const struct nft_set *set);
475 void (*gc_init)(const struct nft_set *set);
477 unsigned int elemsize;
481 * struct nft_set_type - nf_tables set type
483 * @ops: set ops for this type
484 * @features: features supported by the implementation
486 struct nft_set_type {
487 const struct nft_set_ops ops;
490 #define to_set_type(o) container_of(o, struct nft_set_type, ops)
492 struct nft_set_elem_expr {
495 __attribute__((aligned(__alignof__(struct nft_expr))));
498 #define nft_setelem_expr_at(__elem_expr, __offset) \
499 ((struct nft_expr *)&__elem_expr->data[__offset])
501 #define nft_setelem_expr_foreach(__expr, __elem_expr, __size) \
502 for (__expr = nft_setelem_expr_at(__elem_expr, 0), __size = 0; \
503 __size < (__elem_expr)->size; \
504 __size += (__expr)->ops->size, __expr = ((void *)(__expr)) + (__expr)->ops->size)
506 #define NFT_SET_EXPR_MAX 2
509 * struct nft_set - nf_tables set instance
511 * @list: table set list node
512 * @bindings: list of set bindings
513 * @table: table this set belongs to
514 * @net: netnamespace this set belongs to
515 * @name: name of the set
516 * @handle: unique handle of the set
517 * @ktype: key type (numeric type defined by userspace, not used in the kernel)
518 * @dtype: data type (verdict or numeric type defined by userspace)
519 * @objtype: object type (see NFT_OBJECT_* definitions)
520 * @size: maximum set size
521 * @field_len: length of each field in concatenation, bytes
522 * @field_count: number of concatenated fields in element
523 * @use: number of rules references to this set
524 * @nelems: number of elements
525 * @ndeact: number of deactivated elements queued for removal
526 * @timeout: default timeout value in jiffies
527 * @gc_int: garbage collection interval in msecs
528 * @policy: set parameterization (see enum nft_set_policies)
529 * @udlen: user data length
531 * @expr: stateful expression
534 * @genmask: generation mask
537 * @data: private set data
540 struct list_head list;
541 struct list_head bindings;
542 struct nft_table *table;
550 u8 field_len[NFT_REG32_COUNT];
559 unsigned char *udata;
560 /* runtime data below here */
561 const struct nft_set_ops *ops ____cacheline_aligned;
567 struct nft_expr *exprs[NFT_SET_EXPR_MAX];
568 struct list_head catchall_list;
570 __attribute__((aligned(__alignof__(u64))));
573 static inline bool nft_set_is_anonymous(const struct nft_set *set)
575 return set->flags & NFT_SET_ANONYMOUS;
578 static inline void *nft_set_priv(const struct nft_set *set)
580 return (void *)set->data;
583 static inline struct nft_set *nft_set_container_of(const void *priv)
585 return (void *)priv - offsetof(struct nft_set, data);
588 struct nft_set *nft_set_lookup_global(const struct net *net,
589 const struct nft_table *table,
590 const struct nlattr *nla_set_name,
591 const struct nlattr *nla_set_id,
594 struct nft_set_ext *nft_set_catchall_lookup(const struct net *net,
595 const struct nft_set *set);
596 void *nft_set_catchall_gc(const struct nft_set *set);
598 static inline unsigned long nft_set_gc_interval(const struct nft_set *set)
600 u32 gc_int = READ_ONCE(set->gc_int);
602 return gc_int ? msecs_to_jiffies(gc_int) : HZ;
606 * struct nft_set_binding - nf_tables set binding
608 * @list: set bindings list node
609 * @chain: chain containing the rule bound to the set
610 * @flags: set action flags
612 * A set binding contains all information necessary for validation
613 * of new elements added to a bound set.
615 struct nft_set_binding {
616 struct list_head list;
617 const struct nft_chain *chain;
621 enum nft_trans_phase;
622 void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set,
623 struct nft_set_binding *binding,
624 enum nft_trans_phase phase);
625 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
626 struct nft_set_binding *binding);
627 void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set);
630 * enum nft_set_extensions - set extension type IDs
632 * @NFT_SET_EXT_KEY: element key
633 * @NFT_SET_EXT_KEY_END: upper bound element key, for ranges
634 * @NFT_SET_EXT_DATA: mapping data
635 * @NFT_SET_EXT_FLAGS: element flags
636 * @NFT_SET_EXT_TIMEOUT: element timeout
637 * @NFT_SET_EXT_EXPIRATION: element expiration time
638 * @NFT_SET_EXT_USERDATA: user data associated with the element
639 * @NFT_SET_EXT_EXPRESSIONS: expressions assiciated with the element
640 * @NFT_SET_EXT_OBJREF: stateful object reference associated with element
641 * @NFT_SET_EXT_NUM: number of extension types
643 enum nft_set_extensions {
649 NFT_SET_EXT_EXPIRATION,
650 NFT_SET_EXT_USERDATA,
651 NFT_SET_EXT_EXPRESSIONS,
657 * struct nft_set_ext_type - set extension type
659 * @len: fixed part length of the extension
660 * @align: alignment requirements of the extension
662 struct nft_set_ext_type {
667 extern const struct nft_set_ext_type nft_set_ext_types[];
670 * struct nft_set_ext_tmpl - set extension template
672 * @len: length of extension area
673 * @offset: offsets of individual extension types
675 struct nft_set_ext_tmpl {
677 u8 offset[NFT_SET_EXT_NUM];
678 u8 ext_len[NFT_SET_EXT_NUM];
682 * struct nft_set_ext - set extensions
684 * @genmask: generation mask
685 * @offset: offsets of individual extension types
686 * @data: beginning of extension data
690 u8 offset[NFT_SET_EXT_NUM];
694 static inline void nft_set_ext_prepare(struct nft_set_ext_tmpl *tmpl)
696 memset(tmpl, 0, sizeof(*tmpl));
697 tmpl->len = sizeof(struct nft_set_ext);
700 static inline int nft_set_ext_add_length(struct nft_set_ext_tmpl *tmpl, u8 id,
703 tmpl->len = ALIGN(tmpl->len, nft_set_ext_types[id].align);
704 if (tmpl->len > U8_MAX)
707 tmpl->offset[id] = tmpl->len;
708 tmpl->ext_len[id] = nft_set_ext_types[id].len + len;
709 tmpl->len += tmpl->ext_len[id];
714 static inline int nft_set_ext_add(struct nft_set_ext_tmpl *tmpl, u8 id)
716 return nft_set_ext_add_length(tmpl, id, 0);
719 static inline void nft_set_ext_init(struct nft_set_ext *ext,
720 const struct nft_set_ext_tmpl *tmpl)
722 memcpy(ext->offset, tmpl->offset, sizeof(ext->offset));
725 static inline bool __nft_set_ext_exists(const struct nft_set_ext *ext, u8 id)
727 return !!ext->offset[id];
730 static inline bool nft_set_ext_exists(const struct nft_set_ext *ext, u8 id)
732 return ext && __nft_set_ext_exists(ext, id);
735 static inline void *nft_set_ext(const struct nft_set_ext *ext, u8 id)
737 return (void *)ext + ext->offset[id];
740 static inline struct nft_data *nft_set_ext_key(const struct nft_set_ext *ext)
742 return nft_set_ext(ext, NFT_SET_EXT_KEY);
745 static inline struct nft_data *nft_set_ext_key_end(const struct nft_set_ext *ext)
747 return nft_set_ext(ext, NFT_SET_EXT_KEY_END);
750 static inline struct nft_data *nft_set_ext_data(const struct nft_set_ext *ext)
752 return nft_set_ext(ext, NFT_SET_EXT_DATA);
755 static inline u8 *nft_set_ext_flags(const struct nft_set_ext *ext)
757 return nft_set_ext(ext, NFT_SET_EXT_FLAGS);
760 static inline u64 *nft_set_ext_timeout(const struct nft_set_ext *ext)
762 return nft_set_ext(ext, NFT_SET_EXT_TIMEOUT);
765 static inline u64 *nft_set_ext_expiration(const struct nft_set_ext *ext)
767 return nft_set_ext(ext, NFT_SET_EXT_EXPIRATION);
770 static inline struct nft_userdata *nft_set_ext_userdata(const struct nft_set_ext *ext)
772 return nft_set_ext(ext, NFT_SET_EXT_USERDATA);
775 static inline struct nft_set_elem_expr *nft_set_ext_expr(const struct nft_set_ext *ext)
777 return nft_set_ext(ext, NFT_SET_EXT_EXPRESSIONS);
780 static inline bool nft_set_elem_expired(const struct nft_set_ext *ext)
782 return nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION) &&
783 time_is_before_eq_jiffies64(*nft_set_ext_expiration(ext));
786 static inline struct nft_set_ext *nft_set_elem_ext(const struct nft_set *set,
789 return elem + set->ops->elemsize;
792 static inline struct nft_object **nft_set_ext_obj(const struct nft_set_ext *ext)
794 return nft_set_ext(ext, NFT_SET_EXT_OBJREF);
797 struct nft_expr *nft_set_elem_expr_alloc(const struct nft_ctx *ctx,
798 const struct nft_set *set,
799 const struct nlattr *attr);
801 void *nft_set_elem_init(const struct nft_set *set,
802 const struct nft_set_ext_tmpl *tmpl,
803 const u32 *key, const u32 *key_end, const u32 *data,
804 u64 timeout, u64 expiration, gfp_t gfp);
805 int nft_set_elem_expr_clone(const struct nft_ctx *ctx, struct nft_set *set,
806 struct nft_expr *expr_array[]);
807 void nft_set_elem_destroy(const struct nft_set *set, void *elem,
811 * struct nft_set_gc_batch_head - nf_tables set garbage collection batch
814 * @set: set the elements belong to
815 * @cnt: count of elements
817 struct nft_set_gc_batch_head {
819 const struct nft_set *set;
823 #define NFT_SET_GC_BATCH_SIZE ((PAGE_SIZE - \
824 sizeof(struct nft_set_gc_batch_head)) / \
828 * struct nft_set_gc_batch - nf_tables set garbage collection batch
830 * @head: GC batch head
831 * @elems: garbage collection elements
833 struct nft_set_gc_batch {
834 struct nft_set_gc_batch_head head;
835 void *elems[NFT_SET_GC_BATCH_SIZE];
838 struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set,
840 void nft_set_gc_batch_release(struct rcu_head *rcu);
842 static inline void nft_set_gc_batch_complete(struct nft_set_gc_batch *gcb)
845 call_rcu(&gcb->head.rcu, nft_set_gc_batch_release);
848 static inline struct nft_set_gc_batch *
849 nft_set_gc_batch_check(const struct nft_set *set, struct nft_set_gc_batch *gcb,
853 if (gcb->head.cnt + 1 < ARRAY_SIZE(gcb->elems))
855 nft_set_gc_batch_complete(gcb);
857 return nft_set_gc_batch_alloc(set, gfp);
860 static inline void nft_set_gc_batch_add(struct nft_set_gc_batch *gcb,
863 gcb->elems[gcb->head.cnt++] = elem;
868 * struct nft_expr_type - nf_tables expression type
870 * @select_ops: function to select nft_expr_ops
871 * @release_ops: release nft_expr_ops
872 * @ops: default ops, used when no select_ops functions is present
873 * @list: used internally
875 * @owner: module reference
876 * @policy: netlink attribute policy
877 * @maxattr: highest netlink attribute number
878 * @family: address family for AF-specific types
879 * @flags: expression type flags
881 struct nft_expr_type {
882 const struct nft_expr_ops *(*select_ops)(const struct nft_ctx *,
883 const struct nlattr * const tb[]);
884 void (*release_ops)(const struct nft_expr_ops *ops);
885 const struct nft_expr_ops *ops;
886 const struct nft_expr_ops *inner_ops;
887 struct list_head list;
889 struct module *owner;
890 const struct nla_policy *policy;
891 unsigned int maxattr;
896 #define NFT_EXPR_STATEFUL 0x1
897 #define NFT_EXPR_GC 0x2
899 enum nft_trans_phase {
906 struct nft_flow_rule;
907 struct nft_offload_ctx;
910 * struct nft_expr_ops - nf_tables expression operations
912 * @eval: Expression evaluation function
913 * @size: full expression size, including private data size
914 * @init: initialization function
915 * @activate: activate expression in the next generation
916 * @deactivate: deactivate expression in next generation
917 * @destroy: destruction function, called after synchronize_rcu
918 * @dump: function to dump parameters
919 * @type: expression type
920 * @validate: validate expression, called during loop detection
921 * @data: extra data to attach to this expression operation
923 struct nft_expr_ops {
924 void (*eval)(const struct nft_expr *expr,
925 struct nft_regs *regs,
926 const struct nft_pktinfo *pkt);
927 int (*clone)(struct nft_expr *dst,
928 const struct nft_expr *src);
931 int (*init)(const struct nft_ctx *ctx,
932 const struct nft_expr *expr,
933 const struct nlattr * const tb[]);
934 void (*activate)(const struct nft_ctx *ctx,
935 const struct nft_expr *expr);
936 void (*deactivate)(const struct nft_ctx *ctx,
937 const struct nft_expr *expr,
938 enum nft_trans_phase phase);
939 void (*destroy)(const struct nft_ctx *ctx,
940 const struct nft_expr *expr);
941 void (*destroy_clone)(const struct nft_ctx *ctx,
942 const struct nft_expr *expr);
943 int (*dump)(struct sk_buff *skb,
944 const struct nft_expr *expr,
946 int (*validate)(const struct nft_ctx *ctx,
947 const struct nft_expr *expr,
948 const struct nft_data **data);
949 bool (*reduce)(struct nft_regs_track *track,
950 const struct nft_expr *expr);
951 bool (*gc)(struct net *net,
952 const struct nft_expr *expr);
953 int (*offload)(struct nft_offload_ctx *ctx,
954 struct nft_flow_rule *flow,
955 const struct nft_expr *expr);
956 bool (*offload_action)(const struct nft_expr *expr);
957 void (*offload_stats)(struct nft_expr *expr,
958 const struct flow_stats *stats);
959 const struct nft_expr_type *type;
964 * struct nft_rule - nf_tables rule
966 * @list: used internally
967 * @handle: rule handle
968 * @genmask: generation mask
969 * @dlen: length of expression data
970 * @udata: user data is appended to the rule
971 * @data: expression data
974 struct list_head list;
980 __attribute__((aligned(__alignof__(struct nft_expr))));
983 static inline struct nft_expr *nft_expr_first(const struct nft_rule *rule)
985 return (struct nft_expr *)&rule->data[0];
988 static inline struct nft_expr *nft_expr_next(const struct nft_expr *expr)
990 return ((void *)expr) + expr->ops->size;
993 static inline struct nft_expr *nft_expr_last(const struct nft_rule *rule)
995 return (struct nft_expr *)&rule->data[rule->dlen];
998 static inline bool nft_expr_more(const struct nft_rule *rule,
999 const struct nft_expr *expr)
1001 return expr != nft_expr_last(rule) && expr->ops;
1004 static inline struct nft_userdata *nft_userdata(const struct nft_rule *rule)
1006 return (void *)&rule->data[rule->dlen];
1009 void nf_tables_rule_release(const struct nft_ctx *ctx, struct nft_rule *rule);
1011 static inline void nft_set_elem_update_expr(const struct nft_set_ext *ext,
1012 struct nft_regs *regs,
1013 const struct nft_pktinfo *pkt)
1015 struct nft_set_elem_expr *elem_expr;
1016 struct nft_expr *expr;
1019 if (__nft_set_ext_exists(ext, NFT_SET_EXT_EXPRESSIONS)) {
1020 elem_expr = nft_set_ext_expr(ext);
1021 nft_setelem_expr_foreach(expr, elem_expr, size) {
1022 expr->ops->eval(expr, regs, pkt);
1023 if (regs->verdict.code == NFT_BREAK)
1030 * The last pointer isn't really necessary, but the compiler isn't able to
1031 * determine that the result of nft_expr_last() is always the same since it
1032 * can't assume that the dlen value wasn't changed within calls in the loop.
1034 #define nft_rule_for_each_expr(expr, last, rule) \
1035 for ((expr) = nft_expr_first(rule), (last) = nft_expr_last(rule); \
1037 (expr) = nft_expr_next(expr))
1039 #define NFT_CHAIN_POLICY_UNSET U8_MAX
1041 struct nft_rule_dp {
1044 handle:42; /* for tracing */
1045 unsigned char data[]
1046 __attribute__((aligned(__alignof__(struct nft_expr))));
1049 struct nft_rule_dp_last {
1050 struct nft_rule_dp end; /* end of nft_rule_blob marker */
1051 struct rcu_head h; /* call_rcu head */
1052 struct nft_rule_blob *blob; /* ptr to free via call_rcu */
1053 const struct nft_chain *chain; /* for nftables tracing */
1056 static inline const struct nft_rule_dp *nft_rule_next(const struct nft_rule_dp *rule)
1058 return (void *)rule + sizeof(*rule) + rule->dlen;
1061 struct nft_rule_blob {
1063 unsigned char data[]
1064 __attribute__((aligned(__alignof__(struct nft_rule_dp))));
1068 * struct nft_chain - nf_tables chain
1070 * @rules: list of rules in the chain
1071 * @list: used internally
1072 * @rhlhead: used internally
1073 * @table: table that this chain belongs to
1074 * @handle: chain handle
1075 * @use: number of jump references to this chain
1076 * @flags: bitmask of enum nft_chain_flags
1077 * @name: name of the chain
1080 struct nft_rule_blob __rcu *blob_gen_0;
1081 struct nft_rule_blob __rcu *blob_gen_1;
1082 struct list_head rules;
1083 struct list_head list;
1084 struct rhlist_head rhlhead;
1085 struct nft_table *table;
1095 /* Only used during control plane commit phase: */
1096 struct nft_rule_blob *blob_next;
1099 int nft_chain_validate(const struct nft_ctx *ctx, const struct nft_chain *chain);
1100 int nft_setelem_validate(const struct nft_ctx *ctx, struct nft_set *set,
1101 const struct nft_set_iter *iter,
1102 struct nft_set_elem *elem);
1103 int nft_set_catchall_validate(const struct nft_ctx *ctx, struct nft_set *set);
1105 enum nft_chain_types {
1106 NFT_CHAIN_T_DEFAULT = 0,
1113 * struct nft_chain_type - nf_tables chain type info
1115 * @name: name of the type
1116 * @type: numeric identifier
1117 * @family: address family
1118 * @owner: module owner
1119 * @hook_mask: mask of valid hooks
1120 * @hooks: array of hook functions
1121 * @ops_register: base chain register function
1122 * @ops_unregister: base chain unregister function
1124 struct nft_chain_type {
1126 enum nft_chain_types type;
1128 struct module *owner;
1129 unsigned int hook_mask;
1130 nf_hookfn *hooks[NFT_MAX_HOOKS];
1131 int (*ops_register)(struct net *net, const struct nf_hook_ops *ops);
1132 void (*ops_unregister)(struct net *net, const struct nf_hook_ops *ops);
1135 int nft_chain_validate_dependency(const struct nft_chain *chain,
1136 enum nft_chain_types type);
1137 int nft_chain_validate_hooks(const struct nft_chain *chain,
1138 unsigned int hook_flags);
1140 static inline bool nft_chain_is_bound(struct nft_chain *chain)
1142 return (chain->flags & NFT_CHAIN_BINDING) && chain->bound;
1145 void nft_chain_del(struct nft_chain *chain);
1146 void nf_tables_chain_destroy(struct nft_ctx *ctx);
1151 struct u64_stats_sync syncp;
1155 struct list_head list;
1156 struct nf_hook_ops ops;
1157 struct rcu_head rcu;
1161 * struct nft_base_chain - nf_tables base chain
1163 * @ops: netfilter hook ops
1164 * @hook_list: list of netfilter hooks (for NFPROTO_NETDEV family)
1166 * @policy: default policy
1167 * @stats: per-cpu chain stats
1169 * @flow_block: flow block (for hardware offload)
1171 struct nft_base_chain {
1172 struct nf_hook_ops ops;
1173 struct list_head hook_list;
1174 const struct nft_chain_type *type;
1177 struct nft_stats __percpu *stats;
1178 struct nft_chain chain;
1179 struct flow_block flow_block;
1182 static inline struct nft_base_chain *nft_base_chain(const struct nft_chain *chain)
1184 return container_of(chain, struct nft_base_chain, chain);
1187 static inline bool nft_is_base_chain(const struct nft_chain *chain)
1189 return chain->flags & NFT_CHAIN_BASE;
1192 int __nft_release_basechain(struct nft_ctx *ctx);
1194 unsigned int nft_do_chain(struct nft_pktinfo *pkt, void *priv);
1197 * struct nft_table - nf_tables table
1199 * @list: used internally
1200 * @chains_ht: chains in the table
1201 * @chains: same, for stable walks
1202 * @sets: sets in the table
1203 * @objects: stateful objects in the table
1204 * @flowtables: flow tables in the table
1205 * @hgenerator: handle generator state
1206 * @handle: table handle
1207 * @use: number of chain references to this table
1208 * @flags: table flag (see enum nft_table_flags)
1209 * @genmask: generation mask
1210 * @afinfo: address family info
1211 * @name: name of the table
1212 * @validate_state: internal, set when transaction adds jumps
1215 struct list_head list;
1216 struct rhltable chains_ht;
1217 struct list_head chains;
1218 struct list_head sets;
1219 struct list_head objects;
1220 struct list_head flowtables;
1234 static inline bool nft_table_has_owner(const struct nft_table *table)
1236 return table->flags & NFT_TABLE_F_OWNER;
1239 static inline bool nft_base_chain_netdev(int family, u32 hooknum)
1241 return family == NFPROTO_NETDEV ||
1242 (family == NFPROTO_INET && hooknum == NF_INET_INGRESS);
1245 void nft_register_chain_type(const struct nft_chain_type *);
1246 void nft_unregister_chain_type(const struct nft_chain_type *);
1248 int nft_register_expr(struct nft_expr_type *);
1249 void nft_unregister_expr(struct nft_expr_type *);
1251 int nft_verdict_dump(struct sk_buff *skb, int type,
1252 const struct nft_verdict *v);
1255 * struct nft_object_hash_key - key to lookup nft_object
1257 * @name: name of the stateful object to look up
1258 * @table: table the object belongs to
1260 struct nft_object_hash_key {
1262 const struct nft_table *table;
1266 * struct nft_object - nf_tables stateful object
1268 * @list: table stateful object list node
1269 * @key: keys that identify this object
1270 * @rhlhead: nft_objname_ht node
1271 * @genmask: generation mask
1272 * @use: number of references to this stateful object
1273 * @handle: unique object handle
1274 * @ops: object operations
1275 * @data: object data, layout depends on type
1278 struct list_head list;
1279 struct rhlist_head rhlhead;
1280 struct nft_object_hash_key key;
1286 /* runtime data below here */
1287 const struct nft_object_ops *ops ____cacheline_aligned;
1288 unsigned char data[]
1289 __attribute__((aligned(__alignof__(u64))));
1292 static inline void *nft_obj_data(const struct nft_object *obj)
1294 return (void *)obj->data;
1297 #define nft_expr_obj(expr) *((struct nft_object **)nft_expr_priv(expr))
1299 struct nft_object *nft_obj_lookup(const struct net *net,
1300 const struct nft_table *table,
1301 const struct nlattr *nla, u32 objtype,
1304 void nft_obj_notify(struct net *net, const struct nft_table *table,
1305 struct nft_object *obj, u32 portid, u32 seq,
1306 int event, u16 flags, int family, int report, gfp_t gfp);
1309 * struct nft_object_type - stateful object type
1311 * @select_ops: function to select nft_object_ops
1312 * @ops: default ops, used when no select_ops functions is present
1313 * @list: list node in list of object types
1314 * @type: stateful object numeric type
1315 * @owner: module owner
1316 * @maxattr: maximum netlink attribute
1317 * @policy: netlink attribute policy
1319 struct nft_object_type {
1320 const struct nft_object_ops *(*select_ops)(const struct nft_ctx *,
1321 const struct nlattr * const tb[]);
1322 const struct nft_object_ops *ops;
1323 struct list_head list;
1325 unsigned int maxattr;
1326 struct module *owner;
1327 const struct nla_policy *policy;
1331 * struct nft_object_ops - stateful object operations
1333 * @eval: stateful object evaluation function
1334 * @size: stateful object size
1335 * @init: initialize object from netlink attributes
1336 * @destroy: release existing stateful object
1337 * @dump: netlink dump stateful object
1338 * @update: update stateful object
1340 struct nft_object_ops {
1341 void (*eval)(struct nft_object *obj,
1342 struct nft_regs *regs,
1343 const struct nft_pktinfo *pkt);
1345 int (*init)(const struct nft_ctx *ctx,
1346 const struct nlattr *const tb[],
1347 struct nft_object *obj);
1348 void (*destroy)(const struct nft_ctx *ctx,
1349 struct nft_object *obj);
1350 int (*dump)(struct sk_buff *skb,
1351 struct nft_object *obj,
1353 void (*update)(struct nft_object *obj,
1354 struct nft_object *newobj);
1355 const struct nft_object_type *type;
1358 int nft_register_obj(struct nft_object_type *obj_type);
1359 void nft_unregister_obj(struct nft_object_type *obj_type);
1361 #define NFT_NETDEVICE_MAX 256
1364 * struct nft_flowtable - nf_tables flow table
1366 * @list: flow table list node in table list
1367 * @table: the table the flow table is contained in
1368 * @name: name of this flow table
1369 * @hooknum: hook number
1370 * @ops_len: number of hooks in array
1371 * @genmask: generation mask
1372 * @use: number of references to this flow table
1373 * @handle: unique object handle
1374 * @dev_name: array of device names
1375 * @data: rhashtable and garbage collector
1376 * @ops: array of hooks
1378 struct nft_flowtable {
1379 struct list_head list;
1380 struct nft_table *table;
1387 /* runtime data below here */
1388 struct list_head hook_list ____cacheline_aligned;
1389 struct nf_flowtable data;
1392 struct nft_flowtable *nft_flowtable_lookup(const struct nft_table *table,
1393 const struct nlattr *nla,
1396 void nf_tables_deactivate_flowtable(const struct nft_ctx *ctx,
1397 struct nft_flowtable *flowtable,
1398 enum nft_trans_phase phase);
1400 void nft_register_flowtable_type(struct nf_flowtable_type *type);
1401 void nft_unregister_flowtable_type(struct nf_flowtable_type *type);
1404 * struct nft_traceinfo - nft tracing information and state
1406 * @trace: other struct members are initialised
1407 * @nf_trace: copy of skb->nf_trace before rule evaluation
1408 * @type: event type (enum nft_trace_types)
1409 * @skbid: hash of skb to be used as trace id
1410 * @packet_dumped: packet headers sent in a previous traceinfo message
1411 * @pkt: pktinfo currently processed
1412 * @basechain: base chain currently processed
1413 * @rule: rule that was evaluated
1414 * @verdict: verdict given by rule
1416 struct nft_traceinfo {
1420 enum nft_trace_types type:8;
1422 const struct nft_pktinfo *pkt;
1423 const struct nft_base_chain *basechain;
1424 const struct nft_rule_dp *rule;
1425 const struct nft_verdict *verdict;
1428 void nft_trace_init(struct nft_traceinfo *info, const struct nft_pktinfo *pkt,
1429 const struct nft_verdict *verdict,
1430 const struct nft_chain *basechain);
1432 void nft_trace_notify(struct nft_traceinfo *info);
1434 #define MODULE_ALIAS_NFT_CHAIN(family, name) \
1435 MODULE_ALIAS("nft-chain-" __stringify(family) "-" name)
1437 #define MODULE_ALIAS_NFT_AF_EXPR(family, name) \
1438 MODULE_ALIAS("nft-expr-" __stringify(family) "-" name)
1440 #define MODULE_ALIAS_NFT_EXPR(name) \
1441 MODULE_ALIAS("nft-expr-" name)
1443 #define MODULE_ALIAS_NFT_OBJ(type) \
1444 MODULE_ALIAS("nft-obj-" __stringify(type))
1446 #if IS_ENABLED(CONFIG_NF_TABLES)
1449 * The gencursor defines two generations, the currently active and the
1450 * next one. Objects contain a bitmask of 2 bits specifying the generations
1451 * they're active in. A set bit means they're inactive in the generation
1452 * represented by that bit.
1454 * New objects start out as inactive in the current and active in the
1455 * next generation. When committing the ruleset the bitmask is cleared,
1456 * meaning they're active in all generations. When removing an object,
1457 * it is set inactive in the next generation. After committing the ruleset,
1458 * the objects are removed.
1460 static inline unsigned int nft_gencursor_next(const struct net *net)
1462 return net->nft.gencursor + 1 == 1 ? 1 : 0;
1465 static inline u8 nft_genmask_next(const struct net *net)
1467 return 1 << nft_gencursor_next(net);
1470 static inline u8 nft_genmask_cur(const struct net *net)
1472 /* Use READ_ONCE() to prevent refetching the value for atomicity */
1473 return 1 << READ_ONCE(net->nft.gencursor);
1476 #define NFT_GENMASK_ANY ((1 << 0) | (1 << 1))
1479 * Generic transaction helpers
1482 /* Check if this object is currently active. */
1483 #define nft_is_active(__net, __obj) \
1484 (((__obj)->genmask & nft_genmask_cur(__net)) == 0)
1486 /* Check if this object is active in the next generation. */
1487 #define nft_is_active_next(__net, __obj) \
1488 (((__obj)->genmask & nft_genmask_next(__net)) == 0)
1490 /* This object becomes active in the next generation. */
1491 #define nft_activate_next(__net, __obj) \
1492 (__obj)->genmask = nft_genmask_cur(__net)
1494 /* This object becomes inactive in the next generation. */
1495 #define nft_deactivate_next(__net, __obj) \
1496 (__obj)->genmask = nft_genmask_next(__net)
1498 /* After committing the ruleset, clear the stale generation bit. */
1499 #define nft_clear(__net, __obj) \
1500 (__obj)->genmask &= ~nft_genmask_next(__net)
1501 #define nft_active_genmask(__obj, __genmask) \
1502 !((__obj)->genmask & __genmask)
1505 * Set element transaction helpers
1508 static inline bool nft_set_elem_active(const struct nft_set_ext *ext,
1511 return !(ext->genmask & genmask);
1514 static inline void nft_set_elem_change_active(const struct net *net,
1515 const struct nft_set *set,
1516 struct nft_set_ext *ext)
1518 ext->genmask ^= nft_genmask_next(net);
1521 #endif /* IS_ENABLED(CONFIG_NF_TABLES) */
1524 * We use a free bit in the genmask field to indicate the element
1525 * is busy, meaning it is currently being processed either by
1526 * the netlink API or GC.
1528 * Even though the genmask is only a single byte wide, this works
1529 * because the extension structure if fully constant once initialized,
1530 * so there are no non-atomic write accesses unless it is already
1533 #define NFT_SET_ELEM_BUSY_MASK (1 << 2)
1535 #if defined(__LITTLE_ENDIAN_BITFIELD)
1536 #define NFT_SET_ELEM_BUSY_BIT 2
1537 #elif defined(__BIG_ENDIAN_BITFIELD)
1538 #define NFT_SET_ELEM_BUSY_BIT (BITS_PER_LONG - BITS_PER_BYTE + 2)
1543 static inline int nft_set_elem_mark_busy(struct nft_set_ext *ext)
1545 unsigned long *word = (unsigned long *)ext;
1547 BUILD_BUG_ON(offsetof(struct nft_set_ext, genmask) != 0);
1548 return test_and_set_bit(NFT_SET_ELEM_BUSY_BIT, word);
1551 static inline void nft_set_elem_clear_busy(struct nft_set_ext *ext)
1553 unsigned long *word = (unsigned long *)ext;
1555 clear_bit(NFT_SET_ELEM_BUSY_BIT, word);
1559 * struct nft_trans - nf_tables object update in transaction
1561 * @list: used internally
1562 * @msg_type: message type
1563 * @put_net: ctx->net needs to be put
1564 * @ctx: transaction context
1565 * @data: internal information related to the transaction
1568 struct list_head list;
1575 struct nft_trans_rule {
1576 struct nft_rule *rule;
1577 struct nft_flow_rule *flow;
1581 #define nft_trans_rule(trans) \
1582 (((struct nft_trans_rule *)trans->data)->rule)
1583 #define nft_trans_flow_rule(trans) \
1584 (((struct nft_trans_rule *)trans->data)->flow)
1585 #define nft_trans_rule_id(trans) \
1586 (((struct nft_trans_rule *)trans->data)->rule_id)
1588 struct nft_trans_set {
1589 struct nft_set *set;
1597 #define nft_trans_set(trans) \
1598 (((struct nft_trans_set *)trans->data)->set)
1599 #define nft_trans_set_id(trans) \
1600 (((struct nft_trans_set *)trans->data)->set_id)
1601 #define nft_trans_set_bound(trans) \
1602 (((struct nft_trans_set *)trans->data)->bound)
1603 #define nft_trans_set_update(trans) \
1604 (((struct nft_trans_set *)trans->data)->update)
1605 #define nft_trans_set_timeout(trans) \
1606 (((struct nft_trans_set *)trans->data)->timeout)
1607 #define nft_trans_set_gc_int(trans) \
1608 (((struct nft_trans_set *)trans->data)->gc_int)
1610 struct nft_trans_chain {
1613 struct nft_stats __percpu *stats;
1618 #define nft_trans_chain_update(trans) \
1619 (((struct nft_trans_chain *)trans->data)->update)
1620 #define nft_trans_chain_name(trans) \
1621 (((struct nft_trans_chain *)trans->data)->name)
1622 #define nft_trans_chain_stats(trans) \
1623 (((struct nft_trans_chain *)trans->data)->stats)
1624 #define nft_trans_chain_policy(trans) \
1625 (((struct nft_trans_chain *)trans->data)->policy)
1626 #define nft_trans_chain_id(trans) \
1627 (((struct nft_trans_chain *)trans->data)->chain_id)
1629 struct nft_trans_table {
1633 #define nft_trans_table_update(trans) \
1634 (((struct nft_trans_table *)trans->data)->update)
1636 struct nft_trans_elem {
1637 struct nft_set *set;
1638 struct nft_set_elem elem;
1642 #define nft_trans_elem_set(trans) \
1643 (((struct nft_trans_elem *)trans->data)->set)
1644 #define nft_trans_elem(trans) \
1645 (((struct nft_trans_elem *)trans->data)->elem)
1646 #define nft_trans_elem_set_bound(trans) \
1647 (((struct nft_trans_elem *)trans->data)->bound)
1649 struct nft_trans_obj {
1650 struct nft_object *obj;
1651 struct nft_object *newobj;
1655 #define nft_trans_obj(trans) \
1656 (((struct nft_trans_obj *)trans->data)->obj)
1657 #define nft_trans_obj_newobj(trans) \
1658 (((struct nft_trans_obj *)trans->data)->newobj)
1659 #define nft_trans_obj_update(trans) \
1660 (((struct nft_trans_obj *)trans->data)->update)
1662 struct nft_trans_flowtable {
1663 struct nft_flowtable *flowtable;
1665 struct list_head hook_list;
1669 #define nft_trans_flowtable(trans) \
1670 (((struct nft_trans_flowtable *)trans->data)->flowtable)
1671 #define nft_trans_flowtable_update(trans) \
1672 (((struct nft_trans_flowtable *)trans->data)->update)
1673 #define nft_trans_flowtable_hooks(trans) \
1674 (((struct nft_trans_flowtable *)trans->data)->hook_list)
1675 #define nft_trans_flowtable_flags(trans) \
1676 (((struct nft_trans_flowtable *)trans->data)->flags)
1678 int __init nft_chain_filter_init(void);
1679 void nft_chain_filter_fini(void);
1681 void __init nft_chain_route_init(void);
1682 void nft_chain_route_fini(void);
1684 void nf_tables_trans_destroy_flush_work(void);
1686 int nf_msecs_to_jiffies64(const struct nlattr *nla, u64 *result);
1687 __be64 nf_jiffies64_to_msecs(u64 input);
1689 #ifdef CONFIG_MODULES
1690 __printf(2, 3) int nft_request_module(struct net *net, const char *fmt, ...);
1692 static inline int nft_request_module(struct net *net, const char *fmt, ...) { return -ENOENT; }
1695 struct nftables_pernet {
1696 struct list_head tables;
1697 struct list_head commit_list;
1698 struct list_head module_list;
1699 struct list_head notify_list;
1700 struct mutex commit_mutex;
1702 unsigned int base_seq;
1705 extern unsigned int nf_tables_net_id;
1707 static inline struct nftables_pernet *nft_pernet(const struct net *net)
1709 return net_generic(net, nf_tables_net_id);
1712 #define __NFT_REDUCE_READONLY 1UL
1713 #define NFT_REDUCE_READONLY (void *)__NFT_REDUCE_READONLY
1715 static inline bool nft_reduce_is_readonly(const struct nft_expr *expr)
1717 return expr->ops->reduce == NFT_REDUCE_READONLY;
1720 void nft_reg_track_update(struct nft_regs_track *track,
1721 const struct nft_expr *expr, u8 dreg, u8 len);
1722 void nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg, u8 len);
1723 void __nft_reg_track_cancel(struct nft_regs_track *track, u8 dreg);
1725 static inline bool nft_reg_track_cmp(struct nft_regs_track *track,
1726 const struct nft_expr *expr, u8 dreg)
1728 return track->regs[dreg].selector &&
1729 track->regs[dreg].selector->ops == expr->ops &&
1730 track->regs[dreg].num_reg == 0;
1733 #endif /* _NET_NF_TABLES_H */