xref: /linux/include/net/netfilter/nf_flow_table.h (revision 8f7aa3d3c7323f4ca2768a9e74ebbe359c4f8f88)
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 
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	l3_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 	u8				dir:2,
152 					xmit_type:3,
153 					encap_num:2,
154 					tun_num:2,
155 					in_vlan_ingress:2;
156 	u16				mtu;
157 	union {
158 		struct {
159 			struct dst_entry *dst_cache;
160 			u32		ifidx;
161 			u32		dst_cookie;
162 		};
163 		struct {
164 			u32		ifidx;
165 			u8		h_source[ETH_ALEN];
166 			u8		h_dest[ETH_ALEN];
167 		} out;
168 		struct {
169 			u32		iifidx;
170 		} tc;
171 	};
172 };
173 
174 struct flow_offload_tuple_rhash {
175 	struct rhash_head		node;
176 	struct flow_offload_tuple	tuple;
177 };
178 
179 enum nf_flow_flags {
180 	NF_FLOW_SNAT,
181 	NF_FLOW_DNAT,
182 	NF_FLOW_CLOSING,
183 	NF_FLOW_TEARDOWN,
184 	NF_FLOW_HW,
185 	NF_FLOW_HW_DYING,
186 	NF_FLOW_HW_DEAD,
187 	NF_FLOW_HW_PENDING,
188 	NF_FLOW_HW_BIDIRECTIONAL,
189 	NF_FLOW_HW_ESTABLISHED,
190 };
191 
192 enum flow_offload_type {
193 	NF_FLOW_OFFLOAD_UNSPEC	= 0,
194 	NF_FLOW_OFFLOAD_ROUTE,
195 };
196 
197 struct flow_offload {
198 	struct flow_offload_tuple_rhash		tuplehash[FLOW_OFFLOAD_DIR_MAX];
199 	struct nf_conn				*ct;
200 	unsigned long				flags;
201 	u16					type;
202 	u32					timeout;
203 	struct rcu_head				rcu_head;
204 };
205 
206 #define NF_FLOW_TIMEOUT (30 * HZ)
207 #define nf_flowtable_time_stamp	(u32)jiffies
208 
209 unsigned long flow_offload_get_timeout(struct flow_offload *flow);
210 
211 static inline __s32 nf_flow_timeout_delta(unsigned int timeout)
212 {
213 	return (__s32)(timeout - nf_flowtable_time_stamp);
214 }
215 
216 struct nf_flow_route {
217 	struct {
218 		struct dst_entry		*dst;
219 		struct {
220 			u32			ifindex;
221 			struct {
222 				u16		id;
223 				__be16		proto;
224 			} encap[NF_FLOW_TABLE_ENCAP_MAX];
225 			struct flow_offload_tunnel tun;
226 			u8			num_encaps:2,
227 						num_tuns:2,
228 						ingress_vlans:2;
229 		} in;
230 		struct {
231 			u32			ifindex;
232 			u32			hw_ifindex;
233 			u8			h_source[ETH_ALEN];
234 			u8			h_dest[ETH_ALEN];
235 		} out;
236 		enum flow_offload_xmit_type	xmit_type;
237 	} tuple[FLOW_OFFLOAD_DIR_MAX];
238 };
239 
240 struct flow_offload *flow_offload_alloc(struct nf_conn *ct);
241 void flow_offload_free(struct flow_offload *flow);
242 
243 struct nft_flowtable;
244 struct nft_pktinfo;
245 int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
246 		   struct nf_flow_route *route, enum ip_conntrack_dir dir,
247 		   struct nft_flowtable *ft);
248 
249 static inline int
250 nf_flow_table_offload_add_cb(struct nf_flowtable *flow_table,
251 			     flow_setup_cb_t *cb, void *cb_priv)
252 {
253 	struct flow_block *block = &flow_table->flow_block;
254 	struct flow_block_cb *block_cb;
255 	int err = 0;
256 
257 	down_write(&flow_table->flow_block_lock);
258 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
259 	if (block_cb) {
260 		err = -EEXIST;
261 		goto unlock;
262 	}
263 
264 	block_cb = flow_block_cb_alloc(cb, cb_priv, cb_priv, NULL);
265 	if (IS_ERR(block_cb)) {
266 		err = PTR_ERR(block_cb);
267 		goto unlock;
268 	}
269 
270 	list_add_tail(&block_cb->list, &block->cb_list);
271 	up_write(&flow_table->flow_block_lock);
272 
273 	if (flow_table->type->get)
274 		flow_table->type->get(flow_table);
275 	return 0;
276 
277 unlock:
278 	up_write(&flow_table->flow_block_lock);
279 	return err;
280 }
281 
282 static inline void
283 nf_flow_table_offload_del_cb(struct nf_flowtable *flow_table,
284 			     flow_setup_cb_t *cb, void *cb_priv)
285 {
286 	struct flow_block *block = &flow_table->flow_block;
287 	struct flow_block_cb *block_cb;
288 
289 	down_write(&flow_table->flow_block_lock);
290 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
291 	if (block_cb) {
292 		list_del(&block_cb->list);
293 		flow_block_cb_free(block_cb);
294 	} else {
295 		WARN_ON(true);
296 	}
297 	up_write(&flow_table->flow_block_lock);
298 
299 	if (flow_table->type->put)
300 		flow_table->type->put(flow_table);
301 }
302 
303 void flow_offload_route_init(struct flow_offload *flow,
304 			     struct nf_flow_route *route);
305 
306 int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow);
307 void flow_offload_refresh(struct nf_flowtable *flow_table,
308 			  struct flow_offload *flow, bool force);
309 
310 struct flow_offload_tuple_rhash *flow_offload_lookup(struct nf_flowtable *flow_table,
311 						     struct flow_offload_tuple *tuple);
312 void nf_flow_table_gc_run(struct nf_flowtable *flow_table);
313 void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
314 			      struct net_device *dev);
315 void nf_flow_table_cleanup(struct net_device *dev);
316 
317 int nf_flow_table_init(struct nf_flowtable *flow_table);
318 void nf_flow_table_free(struct nf_flowtable *flow_table);
319 
320 void flow_offload_teardown(struct flow_offload *flow);
321 
322 void nf_flow_snat_port(const struct flow_offload *flow,
323 		       struct sk_buff *skb, unsigned int thoff,
324 		       u8 protocol, enum flow_offload_tuple_dir dir);
325 void nf_flow_dnat_port(const struct flow_offload *flow,
326 		       struct sk_buff *skb, unsigned int thoff,
327 		       u8 protocol, enum flow_offload_tuple_dir dir);
328 
329 struct flow_ports {
330 	__be16 source, dest;
331 };
332 
333 struct nf_flowtable *nf_flowtable_by_dev(const struct net_device *dev);
334 int nf_flow_offload_xdp_setup(struct nf_flowtable *flowtable,
335 			      struct net_device *dev,
336 			      enum flow_block_command cmd);
337 
338 unsigned int nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
339 				     const struct nf_hook_state *state);
340 unsigned int nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
341 				       const struct nf_hook_state *state);
342 
343 #if (IS_BUILTIN(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) || \
344     (IS_MODULE(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
345 extern int nf_flow_register_bpf(void);
346 #else
347 static inline int nf_flow_register_bpf(void)
348 {
349 	return 0;
350 }
351 #endif
352 
353 #define MODULE_ALIAS_NF_FLOWTABLE(family)	\
354 	MODULE_ALIAS("nf-flowtable-" __stringify(family))
355 
356 void nf_flow_offload_add(struct nf_flowtable *flowtable,
357 			 struct flow_offload *flow);
358 void nf_flow_offload_del(struct nf_flowtable *flowtable,
359 			 struct flow_offload *flow);
360 void nf_flow_offload_stats(struct nf_flowtable *flowtable,
361 			   struct flow_offload *flow);
362 
363 void nf_flow_table_offload_flush(struct nf_flowtable *flowtable);
364 void nf_flow_table_offload_flush_cleanup(struct nf_flowtable *flowtable);
365 
366 int nf_flow_table_offload_setup(struct nf_flowtable *flowtable,
367 				struct net_device *dev,
368 				enum flow_block_command cmd);
369 int nf_flow_rule_route_ipv4(struct net *net, struct flow_offload *flow,
370 			    enum flow_offload_tuple_dir dir,
371 			    struct nf_flow_rule *flow_rule);
372 int nf_flow_rule_route_ipv6(struct net *net, struct flow_offload *flow,
373 			    enum flow_offload_tuple_dir dir,
374 			    struct nf_flow_rule *flow_rule);
375 
376 int nf_flow_table_offload_init(void);
377 void nf_flow_table_offload_exit(void);
378 
379 static inline __be16 __nf_flow_pppoe_proto(const struct sk_buff *skb)
380 {
381 	__be16 proto;
382 
383 	proto = *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
384 			     sizeof(struct pppoe_hdr)));
385 	switch (proto) {
386 	case htons(PPP_IP):
387 		return htons(ETH_P_IP);
388 	case htons(PPP_IPV6):
389 		return htons(ETH_P_IPV6);
390 	}
391 
392 	return 0;
393 }
394 
395 static inline bool nf_flow_pppoe_proto(struct sk_buff *skb, __be16 *inner_proto)
396 {
397 	if (!pskb_may_pull(skb, ETH_HLEN + PPPOE_SES_HLEN))
398 		return false;
399 
400 	*inner_proto = __nf_flow_pppoe_proto(skb);
401 
402 	return true;
403 }
404 
405 #define NF_FLOW_TABLE_STAT_INC(net, count) __this_cpu_inc((net)->ft.stat->count)
406 #define NF_FLOW_TABLE_STAT_DEC(net, count) __this_cpu_dec((net)->ft.stat->count)
407 #define NF_FLOW_TABLE_STAT_INC_ATOMIC(net, count)	\
408 	this_cpu_inc((net)->ft.stat->count)
409 #define NF_FLOW_TABLE_STAT_DEC_ATOMIC(net, count)	\
410 	this_cpu_dec((net)->ft.stat->count)
411 
412 #ifdef CONFIG_NF_FLOW_TABLE_PROCFS
413 int nf_flow_table_init_proc(struct net *net);
414 void nf_flow_table_fini_proc(struct net *net);
415 #else
416 static inline int nf_flow_table_init_proc(struct net *net)
417 {
418 	return 0;
419 }
420 
421 static inline void nf_flow_table_fini_proc(struct net *net)
422 {
423 }
424 #endif /* CONFIG_NF_FLOW_TABLE_PROCFS */
425 
426 #endif /* _NF_FLOW_TABLE_H */
427