xref: /linux/include/net/netfilter/nf_flow_table.h (revision 91ec2035134982b98fab0609a9fd8480e8217dc1)
1 #ifndef _NF_FLOW_TABLE_H
2 #define _NF_FLOW_TABLE_H
3 
4 #include <linux/in.h>
5 #include <linux/in6.h>
6 #include <linux/netdevice.h>
7 #include <linux/rhashtable-types.h>
8 #include <linux/rcupdate.h>
9 #include <linux/netfilter.h>
10 #include <linux/netfilter/nf_conntrack_tuple_common.h>
11 #include <net/flow_offload.h>
12 #include <net/dst.h>
13 #include <linux/if_pppox.h>
14 #include <linux/ppp_defs.h>
15 
16 struct nf_flowtable;
17 struct nf_flow_rule;
18 struct flow_offload;
19 enum flow_offload_tuple_dir;
20 
21 struct nf_flow_key {
22 	struct flow_dissector_key_meta			meta;
23 	struct flow_dissector_key_control		control;
24 	struct flow_dissector_key_control		enc_control;
25 	struct flow_dissector_key_basic			basic;
26 	struct flow_dissector_key_vlan			vlan;
27 	struct flow_dissector_key_vlan			cvlan;
28 	union {
29 		struct flow_dissector_key_ipv4_addrs	ipv4;
30 		struct flow_dissector_key_ipv6_addrs	ipv6;
31 	};
32 	struct flow_dissector_key_keyid			enc_key_id;
33 	union {
34 		struct flow_dissector_key_ipv4_addrs	enc_ipv4;
35 		struct flow_dissector_key_ipv6_addrs	enc_ipv6;
36 	};
37 	struct flow_dissector_key_tcp			tcp;
38 	struct flow_dissector_key_ports			tp;
39 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
40 
41 struct nf_flow_match {
42 	struct flow_dissector	dissector;
43 	struct nf_flow_key	key;
44 	struct nf_flow_key	mask;
45 };
46 
47 struct nf_flow_rule {
48 	struct nf_flow_match	match;
49 	struct flow_rule	*rule;
50 };
51 
52 struct nf_flowtable_type {
53 	struct list_head		list;
54 	int				family;
55 	int				(*init)(struct nf_flowtable *ft);
56 	bool				(*gc)(const struct flow_offload *flow);
57 	int				(*setup)(struct nf_flowtable *ft,
58 						 struct net_device *dev,
59 						 enum flow_block_command cmd);
60 	int				(*action)(struct net *net,
61 						  struct flow_offload *flow,
62 						  enum flow_offload_tuple_dir dir,
63 						  struct nf_flow_rule *flow_rule);
64 	void				(*free)(struct nf_flowtable *ft);
65 	void				(*get)(struct nf_flowtable *ft);
66 	void				(*put)(struct nf_flowtable *ft);
67 	nf_hookfn			*hook;
68 	struct module			*owner;
69 };
70 
71 enum nf_flowtable_flags {
72 	NF_FLOWTABLE_HW_OFFLOAD		= 0x1,	/* NFT_FLOWTABLE_HW_OFFLOAD */
73 	NF_FLOWTABLE_COUNTER		= 0x2,	/* NFT_FLOWTABLE_COUNTER */
74 };
75 
76 struct nf_flowtable {
77 	unsigned int			flags;		/* readonly in datapath */
78 	int				priority;	/* control path (padding hole) */
79 	struct rhashtable		rhashtable;	/* datapath, read-mostly members come first */
80 
81 	struct list_head		list;		/* slowpath parts */
82 	const struct nf_flowtable_type	*type;
83 	struct delayed_work		gc_work;
84 	struct flow_block		flow_block;
85 	struct rw_semaphore		flow_block_lock; /* Guards flow_block */
86 	possible_net_t			net;
87 };
88 
nf_flowtable_hw_offload(struct nf_flowtable * flowtable)89 static inline bool nf_flowtable_hw_offload(struct nf_flowtable *flowtable)
90 {
91 	return flowtable->flags & NF_FLOWTABLE_HW_OFFLOAD;
92 }
93 
94 enum flow_offload_tuple_dir {
95 	FLOW_OFFLOAD_DIR_ORIGINAL = IP_CT_DIR_ORIGINAL,
96 	FLOW_OFFLOAD_DIR_REPLY = IP_CT_DIR_REPLY,
97 };
98 #define FLOW_OFFLOAD_DIR_MAX	IP_CT_DIR_MAX
99 
100 enum flow_offload_xmit_type {
101 	FLOW_OFFLOAD_XMIT_UNSPEC	= 0,
102 	FLOW_OFFLOAD_XMIT_NEIGH,
103 	FLOW_OFFLOAD_XMIT_XFRM,
104 	FLOW_OFFLOAD_XMIT_DIRECT,
105 	FLOW_OFFLOAD_XMIT_TC,
106 };
107 
108 #define NF_FLOW_TABLE_ENCAP_MAX		2
109 
110 struct flow_offload_tunnel {
111 	union {
112 		struct in_addr	src_v4;
113 		struct in6_addr	src_v6;
114 	};
115 	union {
116 		struct in_addr	dst_v4;
117 		struct in6_addr	dst_v6;
118 	};
119 
120 	u8	inner_proto;
121 };
122 
123 struct flow_offload_tuple {
124 	union {
125 		struct in_addr		src_v4;
126 		struct in6_addr		src_v6;
127 	};
128 	union {
129 		struct in_addr		dst_v4;
130 		struct in6_addr		dst_v6;
131 	};
132 	struct {
133 		__be16			src_port;
134 		__be16			dst_port;
135 	};
136 
137 	int				iifidx;
138 
139 	u8				l3proto;
140 	u8				l4proto;
141 	struct {
142 		u16			id;
143 		__be16			proto;
144 	} encap[NF_FLOW_TABLE_ENCAP_MAX];
145 
146 	struct flow_offload_tunnel	tun;
147 
148 	/* All members above are keys for lookups, see flow_offload_hash(). */
149 	struct { }			__hash;
150 
151 	u16				dir:2,
152 					xmit_type:3,
153 					encap_num:2,
154 					needs_gso_segment:1,
155 					tun_num:2,
156 					in_vlan_ingress:2;
157 	u16				mtu;
158 	u32				dst_cookie;
159 	struct dst_entry		*dst_cache;
160 
161 	union {
162 		struct {
163 			u32		ifidx;
164 		};
165 		struct {
166 			u32		ifidx;
167 			u8		h_source[ETH_ALEN];
168 			u8		h_dest[ETH_ALEN];
169 		} out;
170 		struct {
171 			u32		iifidx;
172 		} tc;
173 	};
174 };
175 
176 struct flow_offload_tuple_rhash {
177 	struct rhash_head		node;
178 	struct flow_offload_tuple	tuple;
179 };
180 
181 enum nf_flow_flags {
182 	NF_FLOW_SNAT,
183 	NF_FLOW_DNAT,
184 	NF_FLOW_CLOSING,
185 	NF_FLOW_TEARDOWN,
186 	NF_FLOW_HW,
187 	NF_FLOW_HW_DYING,
188 	NF_FLOW_HW_DEAD,
189 	NF_FLOW_HW_PENDING,
190 	NF_FLOW_HW_BIDIRECTIONAL,
191 	NF_FLOW_HW_ESTABLISHED,
192 };
193 
194 enum flow_offload_type {
195 	NF_FLOW_OFFLOAD_UNSPEC	= 0,
196 	NF_FLOW_OFFLOAD_ROUTE,
197 };
198 
199 struct flow_offload {
200 	struct flow_offload_tuple_rhash		tuplehash[FLOW_OFFLOAD_DIR_MAX];
201 	struct nf_conn				*ct;
202 	unsigned long				flags;
203 	u16					type;
204 	u32					timeout;
205 	struct rcu_head				rcu_head;
206 };
207 
208 #define NF_FLOW_TIMEOUT (30 * HZ)
209 #define nf_flowtable_time_stamp	(u32)jiffies
210 
211 unsigned long flow_offload_get_timeout(struct flow_offload *flow);
212 
nf_flow_timeout_delta(unsigned int timeout)213 static inline __s32 nf_flow_timeout_delta(unsigned int timeout)
214 {
215 	return (__s32)(timeout - nf_flowtable_time_stamp);
216 }
217 
218 struct nf_flow_route {
219 	struct {
220 		struct dst_entry		*dst;
221 		struct {
222 			u32			ifindex;
223 			struct {
224 				u16		id;
225 				__be16		proto;
226 			} encap[NF_FLOW_TABLE_ENCAP_MAX];
227 			struct flow_offload_tunnel tun;
228 			u8			num_encaps:2,
229 						num_tuns:2,
230 						ingress_vlans:2;
231 		} in;
232 		struct {
233 			u32			ifindex;
234 			u32			hw_ifindex;
235 			u8			h_source[ETH_ALEN];
236 			u8			h_dest[ETH_ALEN];
237 			u8			needs_gso_segment:1;
238 		} out;
239 		enum flow_offload_xmit_type	xmit_type;
240 	} tuple[FLOW_OFFLOAD_DIR_MAX];
241 };
242 
243 struct flow_offload *flow_offload_alloc(struct nf_conn *ct);
244 void flow_offload_free(struct flow_offload *flow);
245 
246 struct nft_flowtable;
247 struct nft_pktinfo;
248 int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
249 		   struct nf_flow_route *route, enum ip_conntrack_dir dir,
250 		   struct nft_flowtable *ft);
251 
252 static inline int
nf_flow_table_offload_add_cb(struct nf_flowtable * flow_table,flow_setup_cb_t * cb,void * cb_priv)253 nf_flow_table_offload_add_cb(struct nf_flowtable *flow_table,
254 			     flow_setup_cb_t *cb, void *cb_priv)
255 {
256 	struct flow_block *block = &flow_table->flow_block;
257 	struct flow_block_cb *block_cb;
258 	int err = 0;
259 
260 	down_write(&flow_table->flow_block_lock);
261 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
262 	if (block_cb) {
263 		err = -EEXIST;
264 		goto unlock;
265 	}
266 
267 	block_cb = flow_block_cb_alloc(cb, cb_priv, cb_priv, NULL);
268 	if (IS_ERR(block_cb)) {
269 		err = PTR_ERR(block_cb);
270 		goto unlock;
271 	}
272 
273 	list_add_tail(&block_cb->list, &block->cb_list);
274 	up_write(&flow_table->flow_block_lock);
275 
276 	if (flow_table->type->get)
277 		flow_table->type->get(flow_table);
278 	return 0;
279 
280 unlock:
281 	up_write(&flow_table->flow_block_lock);
282 	return err;
283 }
284 
285 static inline void
nf_flow_table_offload_del_cb(struct nf_flowtable * flow_table,flow_setup_cb_t * cb,void * cb_priv)286 nf_flow_table_offload_del_cb(struct nf_flowtable *flow_table,
287 			     flow_setup_cb_t *cb, void *cb_priv)
288 {
289 	struct flow_block *block = &flow_table->flow_block;
290 	struct flow_block_cb *block_cb;
291 
292 	down_write(&flow_table->flow_block_lock);
293 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
294 	if (block_cb) {
295 		list_del(&block_cb->list);
296 		flow_block_cb_free(block_cb);
297 	} else {
298 		WARN_ON(true);
299 	}
300 	up_write(&flow_table->flow_block_lock);
301 
302 	if (flow_table->type->put)
303 		flow_table->type->put(flow_table);
304 }
305 
306 void flow_offload_route_init(struct flow_offload *flow,
307 			     struct nf_flow_route *route);
308 
309 int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow);
310 void flow_offload_refresh(struct nf_flowtable *flow_table,
311 			  struct flow_offload *flow, bool force);
312 
nf_flow_dst_check(struct flow_offload_tuple * tuple)313 static inline bool nf_flow_dst_check(struct flow_offload_tuple *tuple)
314 {
315 	if (!tuple->dst_cache)
316 		return true;
317 
318 	return dst_check(tuple->dst_cache, tuple->dst_cookie);
319 }
320 
321 struct flow_offload_tuple_rhash *flow_offload_lookup(struct nf_flowtable *flow_table,
322 						     struct flow_offload_tuple *tuple);
323 void nf_flow_table_gc_run(struct nf_flowtable *flow_table);
324 void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
325 			      struct net_device *dev);
326 void nf_flow_table_cleanup(struct net_device *dev);
327 
328 int nf_flow_table_init(struct nf_flowtable *flow_table);
329 void nf_flow_table_free(struct nf_flowtable *flow_table);
330 
331 void flow_offload_teardown(struct flow_offload *flow);
332 
333 void nf_flow_snat_port(const struct flow_offload *flow,
334 		       struct sk_buff *skb, unsigned int thoff,
335 		       u8 protocol, enum flow_offload_tuple_dir dir);
336 void nf_flow_dnat_port(const struct flow_offload *flow,
337 		       struct sk_buff *skb, unsigned int thoff,
338 		       u8 protocol, enum flow_offload_tuple_dir dir);
339 
340 struct flow_ports {
341 	__be16 source, dest;
342 };
343 
344 struct nf_flowtable *nf_flowtable_by_dev(const struct net_device *dev);
345 int nf_flow_offload_xdp_setup(struct nf_flowtable *flowtable,
346 			      struct net_device *dev,
347 			      enum flow_block_command cmd);
348 
349 unsigned int nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
350 				     const struct nf_hook_state *state);
351 unsigned int nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
352 				       const struct nf_hook_state *state);
353 
354 #if (IS_BUILTIN(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) || \
355     (IS_MODULE(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
356 extern int nf_flow_register_bpf(void);
357 #else
nf_flow_register_bpf(void)358 static inline int nf_flow_register_bpf(void)
359 {
360 	return 0;
361 }
362 #endif
363 
364 #define MODULE_ALIAS_NF_FLOWTABLE(family)	\
365 	MODULE_ALIAS("nf-flowtable-" __stringify(family))
366 
367 void nf_flow_offload_add(struct nf_flowtable *flowtable,
368 			 struct flow_offload *flow);
369 void nf_flow_offload_refresh(struct nf_flowtable *flowtable,
370 			     struct flow_offload *flow);
371 void nf_flow_offload_del(struct nf_flowtable *flowtable,
372 			 struct flow_offload *flow);
373 void nf_flow_offload_stats(struct nf_flowtable *flowtable,
374 			   struct flow_offload *flow);
375 
376 void nf_flow_table_offload_flush(struct nf_flowtable *flowtable);
377 void nf_flow_table_offload_flush_cleanup(struct nf_flowtable *flowtable);
378 
379 int nf_flow_table_offload_setup(struct nf_flowtable *flowtable,
380 				struct net_device *dev,
381 				enum flow_block_command cmd);
382 int nf_flow_rule_route_ipv4(struct net *net, struct flow_offload *flow,
383 			    enum flow_offload_tuple_dir dir,
384 			    struct nf_flow_rule *flow_rule);
385 int nf_flow_rule_route_ipv6(struct net *net, struct flow_offload *flow,
386 			    enum flow_offload_tuple_dir dir,
387 			    struct nf_flow_rule *flow_rule);
388 
389 int nf_flow_table_offload_init(void);
390 void nf_flow_table_offload_exit(void);
391 
__nf_flow_pppoe_proto(const struct sk_buff * skb)392 static inline __be16 __nf_flow_pppoe_proto(const struct sk_buff *skb)
393 {
394 	__be16 proto;
395 
396 	proto = *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
397 			     sizeof(struct pppoe_hdr)));
398 	switch (proto) {
399 	case htons(PPP_IP):
400 		return htons(ETH_P_IP);
401 	case htons(PPP_IPV6):
402 		return htons(ETH_P_IPV6);
403 	}
404 
405 	return 0;
406 }
407 
nf_flow_pppoe_proto(struct sk_buff * skb,__be16 * inner_proto)408 static inline bool nf_flow_pppoe_proto(struct sk_buff *skb, __be16 *inner_proto)
409 {
410 	if (!pskb_may_pull(skb, ETH_HLEN + PPPOE_SES_HLEN))
411 		return false;
412 
413 	*inner_proto = __nf_flow_pppoe_proto(skb);
414 
415 	return true;
416 }
417 
418 #define NF_FLOW_TABLE_STAT_INC(net, count) __this_cpu_inc((net)->ft.stat->count)
419 #define NF_FLOW_TABLE_STAT_DEC(net, count) __this_cpu_dec((net)->ft.stat->count)
420 #define NF_FLOW_TABLE_STAT_INC_ATOMIC(net, count)	\
421 	this_cpu_inc((net)->ft.stat->count)
422 #define NF_FLOW_TABLE_STAT_DEC_ATOMIC(net, count)	\
423 	this_cpu_dec((net)->ft.stat->count)
424 
425 #ifdef CONFIG_NF_FLOW_TABLE_PROCFS
426 int nf_flow_table_init_proc(struct net *net);
427 void nf_flow_table_fini_proc(struct net *net);
428 #else
nf_flow_table_init_proc(struct net * net)429 static inline int nf_flow_table_init_proc(struct net *net)
430 {
431 	return 0;
432 }
433 
nf_flow_table_fini_proc(struct net * net)434 static inline void nf_flow_table_fini_proc(struct net *net)
435 {
436 }
437 #endif /* CONFIG_NF_FLOW_TABLE_PROCFS */
438 
439 #endif /* _NF_FLOW_TABLE_H */
440