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