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 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 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 union {
159 struct {
160 struct dst_entry *dst_cache;
161 u32 ifidx;
162 u32 dst_cookie;
163 };
164 struct {
165 u32 ifidx;
166 u8 h_source[ETH_ALEN];
167 u8 h_dest[ETH_ALEN];
168 } out;
169 struct {
170 u32 iifidx;
171 } tc;
172 };
173 };
174
175 struct flow_offload_tuple_rhash {
176 struct rhash_head node;
177 struct flow_offload_tuple tuple;
178 };
179
180 enum nf_flow_flags {
181 NF_FLOW_SNAT,
182 NF_FLOW_DNAT,
183 NF_FLOW_CLOSING,
184 NF_FLOW_TEARDOWN,
185 NF_FLOW_HW,
186 NF_FLOW_HW_DYING,
187 NF_FLOW_HW_DEAD,
188 NF_FLOW_HW_PENDING,
189 NF_FLOW_HW_BIDIRECTIONAL,
190 NF_FLOW_HW_ESTABLISHED,
191 };
192
193 enum flow_offload_type {
194 NF_FLOW_OFFLOAD_UNSPEC = 0,
195 NF_FLOW_OFFLOAD_ROUTE,
196 };
197
198 struct flow_offload {
199 struct flow_offload_tuple_rhash tuplehash[FLOW_OFFLOAD_DIR_MAX];
200 struct nf_conn *ct;
201 unsigned long flags;
202 u16 type;
203 u32 timeout;
204 struct rcu_head rcu_head;
205 };
206
207 #define NF_FLOW_TIMEOUT (30 * HZ)
208 #define nf_flowtable_time_stamp (u32)jiffies
209
210 unsigned long flow_offload_get_timeout(struct flow_offload *flow);
211
nf_flow_timeout_delta(unsigned int timeout)212 static inline __s32 nf_flow_timeout_delta(unsigned int timeout)
213 {
214 return (__s32)(timeout - nf_flowtable_time_stamp);
215 }
216
217 struct nf_flow_route {
218 struct {
219 struct dst_entry *dst;
220 struct {
221 u32 ifindex;
222 struct {
223 u16 id;
224 __be16 proto;
225 } encap[NF_FLOW_TABLE_ENCAP_MAX];
226 struct flow_offload_tunnel tun;
227 u8 num_encaps:2,
228 num_tuns:2,
229 ingress_vlans:2;
230 } in;
231 struct {
232 u32 ifindex;
233 u32 hw_ifindex;
234 u8 h_source[ETH_ALEN];
235 u8 h_dest[ETH_ALEN];
236 u8 needs_gso_segment:1;
237 } out;
238 enum flow_offload_xmit_type xmit_type;
239 } tuple[FLOW_OFFLOAD_DIR_MAX];
240 };
241
242 struct flow_offload *flow_offload_alloc(struct nf_conn *ct);
243 void flow_offload_free(struct flow_offload *flow);
244
245 struct nft_flowtable;
246 struct nft_pktinfo;
247 int nft_flow_route(const struct nft_pktinfo *pkt, const struct nf_conn *ct,
248 struct nf_flow_route *route, enum ip_conntrack_dir dir,
249 struct nft_flowtable *ft);
250
251 static inline int
nf_flow_table_offload_add_cb(struct nf_flowtable * flow_table,flow_setup_cb_t * cb,void * cb_priv)252 nf_flow_table_offload_add_cb(struct nf_flowtable *flow_table,
253 flow_setup_cb_t *cb, void *cb_priv)
254 {
255 struct flow_block *block = &flow_table->flow_block;
256 struct flow_block_cb *block_cb;
257 int err = 0;
258
259 down_write(&flow_table->flow_block_lock);
260 block_cb = flow_block_cb_lookup(block, cb, cb_priv);
261 if (block_cb) {
262 err = -EEXIST;
263 goto unlock;
264 }
265
266 block_cb = flow_block_cb_alloc(cb, cb_priv, cb_priv, NULL);
267 if (IS_ERR(block_cb)) {
268 err = PTR_ERR(block_cb);
269 goto unlock;
270 }
271
272 list_add_tail(&block_cb->list, &block->cb_list);
273 up_write(&flow_table->flow_block_lock);
274
275 if (flow_table->type->get)
276 flow_table->type->get(flow_table);
277 return 0;
278
279 unlock:
280 up_write(&flow_table->flow_block_lock);
281 return err;
282 }
283
284 static inline void
nf_flow_table_offload_del_cb(struct nf_flowtable * flow_table,flow_setup_cb_t * cb,void * cb_priv)285 nf_flow_table_offload_del_cb(struct nf_flowtable *flow_table,
286 flow_setup_cb_t *cb, void *cb_priv)
287 {
288 struct flow_block *block = &flow_table->flow_block;
289 struct flow_block_cb *block_cb;
290
291 down_write(&flow_table->flow_block_lock);
292 block_cb = flow_block_cb_lookup(block, cb, cb_priv);
293 if (block_cb) {
294 list_del(&block_cb->list);
295 flow_block_cb_free(block_cb);
296 } else {
297 WARN_ON(true);
298 }
299 up_write(&flow_table->flow_block_lock);
300
301 if (flow_table->type->put)
302 flow_table->type->put(flow_table);
303 }
304
305 void flow_offload_route_init(struct flow_offload *flow,
306 struct nf_flow_route *route);
307
308 int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow);
309 void flow_offload_refresh(struct nf_flowtable *flow_table,
310 struct flow_offload *flow, bool force);
311
312 struct flow_offload_tuple_rhash *flow_offload_lookup(struct nf_flowtable *flow_table,
313 struct flow_offload_tuple *tuple);
314 void nf_flow_table_gc_run(struct nf_flowtable *flow_table);
315 void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
316 struct net_device *dev);
317 void nf_flow_table_cleanup(struct net_device *dev);
318
319 int nf_flow_table_init(struct nf_flowtable *flow_table);
320 void nf_flow_table_free(struct nf_flowtable *flow_table);
321
322 void flow_offload_teardown(struct flow_offload *flow);
323
324 void nf_flow_snat_port(const struct flow_offload *flow,
325 struct sk_buff *skb, unsigned int thoff,
326 u8 protocol, enum flow_offload_tuple_dir dir);
327 void nf_flow_dnat_port(const struct flow_offload *flow,
328 struct sk_buff *skb, unsigned int thoff,
329 u8 protocol, enum flow_offload_tuple_dir dir);
330
331 struct flow_ports {
332 __be16 source, dest;
333 };
334
335 struct nf_flowtable *nf_flowtable_by_dev(const struct net_device *dev);
336 int nf_flow_offload_xdp_setup(struct nf_flowtable *flowtable,
337 struct net_device *dev,
338 enum flow_block_command cmd);
339
340 unsigned int nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
341 const struct nf_hook_state *state);
342 unsigned int nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
343 const struct nf_hook_state *state);
344
345 #if (IS_BUILTIN(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF)) || \
346 (IS_MODULE(CONFIG_NF_FLOW_TABLE) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
347 extern int nf_flow_register_bpf(void);
348 #else
nf_flow_register_bpf(void)349 static inline int nf_flow_register_bpf(void)
350 {
351 return 0;
352 }
353 #endif
354
355 #define MODULE_ALIAS_NF_FLOWTABLE(family) \
356 MODULE_ALIAS("nf-flowtable-" __stringify(family))
357
358 void nf_flow_offload_add(struct nf_flowtable *flowtable,
359 struct flow_offload *flow);
360 void nf_flow_offload_del(struct nf_flowtable *flowtable,
361 struct flow_offload *flow);
362 void nf_flow_offload_stats(struct nf_flowtable *flowtable,
363 struct flow_offload *flow);
364
365 void nf_flow_table_offload_flush(struct nf_flowtable *flowtable);
366 void nf_flow_table_offload_flush_cleanup(struct nf_flowtable *flowtable);
367
368 int nf_flow_table_offload_setup(struct nf_flowtable *flowtable,
369 struct net_device *dev,
370 enum flow_block_command cmd);
371 int nf_flow_rule_route_ipv4(struct net *net, struct flow_offload *flow,
372 enum flow_offload_tuple_dir dir,
373 struct nf_flow_rule *flow_rule);
374 int nf_flow_rule_route_ipv6(struct net *net, struct flow_offload *flow,
375 enum flow_offload_tuple_dir dir,
376 struct nf_flow_rule *flow_rule);
377
378 int nf_flow_table_offload_init(void);
379 void nf_flow_table_offload_exit(void);
380
__nf_flow_pppoe_proto(const struct sk_buff * skb)381 static inline __be16 __nf_flow_pppoe_proto(const struct sk_buff *skb)
382 {
383 __be16 proto;
384
385 proto = *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
386 sizeof(struct pppoe_hdr)));
387 switch (proto) {
388 case htons(PPP_IP):
389 return htons(ETH_P_IP);
390 case htons(PPP_IPV6):
391 return htons(ETH_P_IPV6);
392 }
393
394 return 0;
395 }
396
nf_flow_pppoe_proto(struct sk_buff * skb,__be16 * inner_proto)397 static inline bool nf_flow_pppoe_proto(struct sk_buff *skb, __be16 *inner_proto)
398 {
399 if (!pskb_may_pull(skb, ETH_HLEN + PPPOE_SES_HLEN))
400 return false;
401
402 *inner_proto = __nf_flow_pppoe_proto(skb);
403
404 return true;
405 }
406
407 #define NF_FLOW_TABLE_STAT_INC(net, count) __this_cpu_inc((net)->ft.stat->count)
408 #define NF_FLOW_TABLE_STAT_DEC(net, count) __this_cpu_dec((net)->ft.stat->count)
409 #define NF_FLOW_TABLE_STAT_INC_ATOMIC(net, count) \
410 this_cpu_inc((net)->ft.stat->count)
411 #define NF_FLOW_TABLE_STAT_DEC_ATOMIC(net, count) \
412 this_cpu_dec((net)->ft.stat->count)
413
414 #ifdef CONFIG_NF_FLOW_TABLE_PROCFS
415 int nf_flow_table_init_proc(struct net *net);
416 void nf_flow_table_fini_proc(struct net *net);
417 #else
nf_flow_table_init_proc(struct net * net)418 static inline int nf_flow_table_init_proc(struct net *net)
419 {
420 return 0;
421 }
422
nf_flow_table_fini_proc(struct net * net)423 static inline void nf_flow_table_fini_proc(struct net *net)
424 {
425 }
426 #endif /* CONFIG_NF_FLOW_TABLE_PROCFS */
427
428 #endif /* _NF_FLOW_TABLE_H */
429