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_tuple {
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 	struct {
120 		__be16			src_port;
121 		__be16			dst_port;
122 	};
123 
124 	int				iifidx;
125 
126 	u8				l3proto;
127 	u8				l4proto;
128 	struct {
129 		u16			id;
130 		__be16			proto;
131 	} encap[NF_FLOW_TABLE_ENCAP_MAX];
132 
133 	/* All members above are keys for lookups, see flow_offload_hash(). */
134 	struct { }			__hash;
135 
136 	u8				dir:2,
137 					xmit_type:3,
138 					encap_num:2,
139 					in_vlan_ingress:2;
140 	u16				mtu;
141 	union {
142 		struct {
143 			struct dst_entry *dst_cache;
144 			u32		dst_cookie;
145 		};
146 		struct {
147 			u32		ifidx;
148 			u32		hw_ifidx;
149 			u8		h_source[ETH_ALEN];
150 			u8		h_dest[ETH_ALEN];
151 		} out;
152 		struct {
153 			u32		iifidx;
154 		} tc;
155 	};
156 };
157 
158 struct flow_offload_tuple_rhash {
159 	struct rhash_head		node;
160 	struct flow_offload_tuple	tuple;
161 };
162 
163 enum nf_flow_flags {
164 	NF_FLOW_SNAT,
165 	NF_FLOW_DNAT,
166 	NF_FLOW_CLOSING,
167 	NF_FLOW_TEARDOWN,
168 	NF_FLOW_HW,
169 	NF_FLOW_HW_DYING,
170 	NF_FLOW_HW_DEAD,
171 	NF_FLOW_HW_PENDING,
172 	NF_FLOW_HW_BIDIRECTIONAL,
173 	NF_FLOW_HW_ESTABLISHED,
174 };
175 
176 enum flow_offload_type {
177 	NF_FLOW_OFFLOAD_UNSPEC	= 0,
178 	NF_FLOW_OFFLOAD_ROUTE,
179 };
180 
181 struct flow_offload {
182 	struct flow_offload_tuple_rhash		tuplehash[FLOW_OFFLOAD_DIR_MAX];
183 	struct nf_conn				*ct;
184 	unsigned long				flags;
185 	u16					type;
186 	u32					timeout;
187 	struct rcu_head				rcu_head;
188 };
189 
190 #define NF_FLOW_TIMEOUT (30 * HZ)
191 #define nf_flowtable_time_stamp	(u32)jiffies
192 
193 unsigned long flow_offload_get_timeout(struct flow_offload *flow);
194 
195 static inline __s32 nf_flow_timeout_delta(unsigned int timeout)
196 {
197 	return (__s32)(timeout - nf_flowtable_time_stamp);
198 }
199 
200 struct nf_flow_route {
201 	struct {
202 		struct dst_entry		*dst;
203 		struct {
204 			u32			ifindex;
205 			struct {
206 				u16		id;
207 				__be16		proto;
208 			} encap[NF_FLOW_TABLE_ENCAP_MAX];
209 			u8			num_encaps:2,
210 						ingress_vlans:2;
211 		} in;
212 		struct {
213 			u32			ifindex;
214 			u32			hw_ifindex;
215 			u8			h_source[ETH_ALEN];
216 			u8			h_dest[ETH_ALEN];
217 		} out;
218 		enum flow_offload_xmit_type	xmit_type;
219 	} tuple[FLOW_OFFLOAD_DIR_MAX];
220 };
221 
222 struct flow_offload *flow_offload_alloc(struct nf_conn *ct);
223 void flow_offload_free(struct flow_offload *flow);
224 
225 static inline int
226 nf_flow_table_offload_add_cb(struct nf_flowtable *flow_table,
227 			     flow_setup_cb_t *cb, void *cb_priv)
228 {
229 	struct flow_block *block = &flow_table->flow_block;
230 	struct flow_block_cb *block_cb;
231 	int err = 0;
232 
233 	down_write(&flow_table->flow_block_lock);
234 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
235 	if (block_cb) {
236 		err = -EEXIST;
237 		goto unlock;
238 	}
239 
240 	block_cb = flow_block_cb_alloc(cb, cb_priv, cb_priv, NULL);
241 	if (IS_ERR(block_cb)) {
242 		err = PTR_ERR(block_cb);
243 		goto unlock;
244 	}
245 
246 	list_add_tail(&block_cb->list, &block->cb_list);
247 	up_write(&flow_table->flow_block_lock);
248 
249 	if (flow_table->type->get)
250 		flow_table->type->get(flow_table);
251 	return 0;
252 
253 unlock:
254 	up_write(&flow_table->flow_block_lock);
255 	return err;
256 }
257 
258 static inline void
259 nf_flow_table_offload_del_cb(struct nf_flowtable *flow_table,
260 			     flow_setup_cb_t *cb, void *cb_priv)
261 {
262 	struct flow_block *block = &flow_table->flow_block;
263 	struct flow_block_cb *block_cb;
264 
265 	down_write(&flow_table->flow_block_lock);
266 	block_cb = flow_block_cb_lookup(block, cb, cb_priv);
267 	if (block_cb) {
268 		list_del(&block_cb->list);
269 		flow_block_cb_free(block_cb);
270 	} else {
271 		WARN_ON(true);
272 	}
273 	up_write(&flow_table->flow_block_lock);
274 
275 	if (flow_table->type->put)
276 		flow_table->type->put(flow_table);
277 }
278 
279 void flow_offload_route_init(struct flow_offload *flow,
280 			     struct nf_flow_route *route);
281 
282 int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow);
283 void flow_offload_refresh(struct nf_flowtable *flow_table,
284 			  struct flow_offload *flow, bool force);
285 
286 struct flow_offload_tuple_rhash *flow_offload_lookup(struct nf_flowtable *flow_table,
287 						     struct flow_offload_tuple *tuple);
288 void nf_flow_table_gc_run(struct nf_flowtable *flow_table);
289 void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
290 			      struct net_device *dev);
291 void nf_flow_table_cleanup(struct net_device *dev);
292 
293 int nf_flow_table_init(struct nf_flowtable *flow_table);
294 void nf_flow_table_free(struct nf_flowtable *flow_table);
295 
296 void flow_offload_teardown(struct flow_offload *flow);
297 
298 void nf_flow_snat_port(const struct flow_offload *flow,
299 		       struct sk_buff *skb, unsigned int thoff,
300 		       u8 protocol, enum flow_offload_tuple_dir dir);
301 void nf_flow_dnat_port(const struct flow_offload *flow,
302 		       struct sk_buff *skb, unsigned int thoff,
303 		       u8 protocol, enum flow_offload_tuple_dir dir);
304 
305 struct flow_ports {
306 	__be16 source, dest;
307 };
308 
309 struct nf_flowtable *nf_flowtable_by_dev(const struct net_device *dev);
310 int nf_flow_offload_xdp_setup(struct nf_flowtable *flowtable,
311 			      struct net_device *dev,
312 			      enum flow_block_command cmd);
313 
314 unsigned int nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
315 				     const struct nf_hook_state *state);
316 unsigned int nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
317 				       const struct nf_hook_state *state);
318 
319 #if (IS_BUILTIN(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) || \
320     (IS_MODULE(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
321 extern int nf_flow_register_bpf(void);
322 #else
323 static inline int nf_flow_register_bpf(void)
324 {
325 	return 0;
326 }
327 #endif
328 
329 #define MODULE_ALIAS_NF_FLOWTABLE(family)	\
330 	MODULE_ALIAS("nf-flowtable-" __stringify(family))
331 
332 void nf_flow_offload_add(struct nf_flowtable *flowtable,
333 			 struct flow_offload *flow);
334 void nf_flow_offload_del(struct nf_flowtable *flowtable,
335 			 struct flow_offload *flow);
336 void nf_flow_offload_stats(struct nf_flowtable *flowtable,
337 			   struct flow_offload *flow);
338 
339 void nf_flow_table_offload_flush(struct nf_flowtable *flowtable);
340 void nf_flow_table_offload_flush_cleanup(struct nf_flowtable *flowtable);
341 
342 int nf_flow_table_offload_setup(struct nf_flowtable *flowtable,
343 				struct net_device *dev,
344 				enum flow_block_command cmd);
345 int nf_flow_rule_route_ipv4(struct net *net, struct flow_offload *flow,
346 			    enum flow_offload_tuple_dir dir,
347 			    struct nf_flow_rule *flow_rule);
348 int nf_flow_rule_route_ipv6(struct net *net, struct flow_offload *flow,
349 			    enum flow_offload_tuple_dir dir,
350 			    struct nf_flow_rule *flow_rule);
351 
352 int nf_flow_table_offload_init(void);
353 void nf_flow_table_offload_exit(void);
354 
355 static inline __be16 __nf_flow_pppoe_proto(const struct sk_buff *skb)
356 {
357 	__be16 proto;
358 
359 	proto = *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
360 			     sizeof(struct pppoe_hdr)));
361 	switch (proto) {
362 	case htons(PPP_IP):
363 		return htons(ETH_P_IP);
364 	case htons(PPP_IPV6):
365 		return htons(ETH_P_IPV6);
366 	}
367 
368 	return 0;
369 }
370 
371 static inline bool nf_flow_pppoe_proto(struct sk_buff *skb, __be16 *inner_proto)
372 {
373 	if (!pskb_may_pull(skb, PPPOE_SES_HLEN))
374 		return false;
375 
376 	*inner_proto = __nf_flow_pppoe_proto(skb);
377 
378 	return true;
379 }
380 
381 #define NF_FLOW_TABLE_STAT_INC(net, count) __this_cpu_inc((net)->ft.stat->count)
382 #define NF_FLOW_TABLE_STAT_DEC(net, count) __this_cpu_dec((net)->ft.stat->count)
383 #define NF_FLOW_TABLE_STAT_INC_ATOMIC(net, count)	\
384 	this_cpu_inc((net)->ft.stat->count)
385 #define NF_FLOW_TABLE_STAT_DEC_ATOMIC(net, count)	\
386 	this_cpu_dec((net)->ft.stat->count)
387 
388 #ifdef CONFIG_NF_FLOW_TABLE_PROCFS
389 int nf_flow_table_init_proc(struct net *net);
390 void nf_flow_table_fini_proc(struct net *net);
391 #else
392 static inline int nf_flow_table_init_proc(struct net *net)
393 {
394 	return 0;
395 }
396 
397 static inline void nf_flow_table_fini_proc(struct net *net)
398 {
399 }
400 #endif /* CONFIG_NF_FLOW_TABLE_PROCFS */
401 
402 #endif /* _NF_FLOW_TABLE_H */
403