xref: /linux/net/mptcp/protocol.h (revision 1e328ed559201c4a3733b148b0afe3cfb4ad8b45)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3  *
4  * Copyright (c) 2017 - 2019, Intel Corporation.
5  */
6 
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9 
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 
14 #define MPTCP_SUPPORTED_VERSION	1
15 
16 /* MPTCP option bits */
17 #define OPTION_MPTCP_MPC_SYN	BIT(0)
18 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
19 #define OPTION_MPTCP_MPC_ACK	BIT(2)
20 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
21 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
22 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
23 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
24 #define OPTION_MPTCP_ADD_ADDR6	BIT(7)
25 #define OPTION_MPTCP_RM_ADDR	BIT(8)
26 #define OPTION_MPTCP_FASTCLOSE	BIT(9)
27 #define OPTION_MPTCP_PRIO	BIT(10)
28 
29 /* MPTCP option subtypes */
30 #define MPTCPOPT_MP_CAPABLE	0
31 #define MPTCPOPT_MP_JOIN	1
32 #define MPTCPOPT_DSS		2
33 #define MPTCPOPT_ADD_ADDR	3
34 #define MPTCPOPT_RM_ADDR	4
35 #define MPTCPOPT_MP_PRIO	5
36 #define MPTCPOPT_MP_FAIL	6
37 #define MPTCPOPT_MP_FASTCLOSE	7
38 
39 /* MPTCP suboption lengths */
40 #define TCPOLEN_MPTCP_MPC_SYN		4
41 #define TCPOLEN_MPTCP_MPC_SYNACK	12
42 #define TCPOLEN_MPTCP_MPC_ACK		20
43 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
44 #define TCPOLEN_MPTCP_MPJ_SYN		12
45 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
46 #define TCPOLEN_MPTCP_MPJ_ACK		24
47 #define TCPOLEN_MPTCP_DSS_BASE		4
48 #define TCPOLEN_MPTCP_DSS_ACK32		4
49 #define TCPOLEN_MPTCP_DSS_ACK64		8
50 #define TCPOLEN_MPTCP_DSS_MAP32		10
51 #define TCPOLEN_MPTCP_DSS_MAP64		14
52 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
53 #define TCPOLEN_MPTCP_ADD_ADDR		16
54 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	20
55 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
56 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	12
57 #define TCPOLEN_MPTCP_ADD_ADDR6		28
58 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	32
59 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
60 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	24
61 #define TCPOLEN_MPTCP_PORT_LEN		4
62 #define TCPOLEN_MPTCP_RM_ADDR_BASE	4
63 #define TCPOLEN_MPTCP_PRIO		4
64 #define TCPOLEN_MPTCP_FASTCLOSE		12
65 
66 /* MPTCP MP_JOIN flags */
67 #define MPTCPOPT_BACKUP		BIT(0)
68 #define MPTCPOPT_HMAC_LEN	20
69 #define MPTCPOPT_THMAC_LEN	8
70 
71 /* MPTCP MP_CAPABLE flags */
72 #define MPTCP_VERSION_MASK	(0x0F)
73 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
74 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
75 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
76 #define MPTCP_CAP_FLAG_MASK	(0x3F)
77 
78 /* MPTCP DSS flags */
79 #define MPTCP_DSS_DATA_FIN	BIT(4)
80 #define MPTCP_DSS_DSN64		BIT(3)
81 #define MPTCP_DSS_HAS_MAP	BIT(2)
82 #define MPTCP_DSS_ACK64		BIT(1)
83 #define MPTCP_DSS_HAS_ACK	BIT(0)
84 #define MPTCP_DSS_FLAG_MASK	(0x1F)
85 
86 /* MPTCP ADD_ADDR flags */
87 #define MPTCP_ADDR_ECHO		BIT(0)
88 #define MPTCP_ADDR_IPVERSION_4	4
89 #define MPTCP_ADDR_IPVERSION_6	6
90 
91 /* MPTCP MP_PRIO flags */
92 #define MPTCP_PRIO_BKUP		BIT(0)
93 
94 /* MPTCP socket flags */
95 #define MPTCP_DATA_READY	0
96 #define MPTCP_NOSPACE		1
97 #define MPTCP_WORK_RTX		2
98 #define MPTCP_WORK_EOF		3
99 #define MPTCP_FALLBACK_DONE	4
100 #define MPTCP_WORK_CLOSE_SUBFLOW 5
101 #define MPTCP_PUSH_PENDING	6
102 #define MPTCP_CLEAN_UNA		7
103 
104 static inline bool before64(__u64 seq1, __u64 seq2)
105 {
106 	return (__s64)(seq1 - seq2) < 0;
107 }
108 
109 #define after64(seq2, seq1)	before64(seq1, seq2)
110 
111 struct mptcp_options_received {
112 	u64	sndr_key;
113 	u64	rcvr_key;
114 	u64	data_ack;
115 	u64	data_seq;
116 	u32	subflow_seq;
117 	u16	data_len;
118 	u16	mp_capable : 1,
119 		mp_join : 1,
120 		fastclose : 1,
121 		dss : 1,
122 		add_addr : 1,
123 		rm_addr : 1,
124 		mp_prio : 1,
125 		family : 4,
126 		echo : 1,
127 		backup : 1;
128 	u32	token;
129 	u32	nonce;
130 	u64	thmac;
131 	u8	hmac[MPTCPOPT_HMAC_LEN];
132 	u8	join_id;
133 	u8	use_map:1,
134 		dsn64:1,
135 		data_fin:1,
136 		use_ack:1,
137 		ack64:1,
138 		mpc_map:1,
139 		__unused:2;
140 	u8	addr_id;
141 	u8	rm_id;
142 	union {
143 		struct in_addr	addr;
144 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
145 		struct in6_addr	addr6;
146 #endif
147 	};
148 	u64	ahmac;
149 	u16	port;
150 };
151 
152 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
153 {
154 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
155 		     ((nib & 0xF) << 8) | field);
156 }
157 
158 struct mptcp_addr_info {
159 	sa_family_t		family;
160 	__be16			port;
161 	u8			id;
162 	u8			flags;
163 	int			ifindex;
164 	union {
165 		struct in_addr addr;
166 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
167 		struct in6_addr addr6;
168 #endif
169 	};
170 };
171 
172 enum mptcp_pm_status {
173 	MPTCP_PM_ADD_ADDR_RECEIVED,
174 	MPTCP_PM_ADD_ADDR_SEND_ACK,
175 	MPTCP_PM_RM_ADDR_RECEIVED,
176 	MPTCP_PM_ESTABLISHED,
177 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
178 	MPTCP_PM_SUBFLOW_ESTABLISHED,
179 };
180 
181 enum mptcp_addr_signal_status {
182 	MPTCP_ADD_ADDR_SIGNAL,
183 	MPTCP_ADD_ADDR_ECHO,
184 	MPTCP_ADD_ADDR_IPV6,
185 	MPTCP_ADD_ADDR_PORT,
186 	MPTCP_RM_ADDR_SIGNAL,
187 };
188 
189 struct mptcp_pm_data {
190 	struct mptcp_addr_info local;
191 	struct mptcp_addr_info remote;
192 	struct list_head anno_list;
193 
194 	spinlock_t	lock;		/*protects the whole PM data */
195 
196 	u8		addr_signal;
197 	bool		server_side;
198 	bool		work_pending;
199 	bool		accept_addr;
200 	bool		accept_subflow;
201 	u8		add_addr_signaled;
202 	u8		add_addr_accepted;
203 	u8		local_addr_used;
204 	u8		subflows;
205 	u8		add_addr_signal_max;
206 	u8		add_addr_accept_max;
207 	u8		local_addr_max;
208 	u8		subflows_max;
209 	u8		status;
210 	u8		rm_id;
211 };
212 
213 struct mptcp_data_frag {
214 	struct list_head list;
215 	u64 data_seq;
216 	u16 data_len;
217 	u16 offset;
218 	u16 overhead;
219 	u16 already_sent;
220 	struct page *page;
221 };
222 
223 /* MPTCP connection sock */
224 struct mptcp_sock {
225 	/* inet_connection_sock must be the first member */
226 	struct inet_connection_sock sk;
227 	u64		local_key;
228 	u64		remote_key;
229 	u64		write_seq;
230 	u64		snd_nxt;
231 	u64		ack_seq;
232 	u64		rcv_wnd_sent;
233 	u64		rcv_data_fin_seq;
234 	int		wmem_reserved;
235 	struct sock	*last_snd;
236 	int		snd_burst;
237 	int		old_wspace;
238 	u64		snd_una;
239 	u64		wnd_end;
240 	unsigned long	timer_ival;
241 	u32		token;
242 	int		rmem_pending;
243 	int		rmem_released;
244 	unsigned long	flags;
245 	bool		can_ack;
246 	bool		fully_established;
247 	bool		rcv_data_fin;
248 	bool		snd_data_fin_enable;
249 	bool		rcv_fastclose;
250 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
251 	spinlock_t	join_list_lock;
252 	struct sock	*ack_hint;
253 	struct work_struct work;
254 	struct sk_buff  *ooo_last_skb;
255 	struct rb_root  out_of_order_queue;
256 	struct sk_buff_head receive_queue;
257 	struct sk_buff_head skb_tx_cache;	/* this is wmem accounted */
258 	int		tx_pending_data;
259 	int		size_goal_cache;
260 	struct list_head conn_list;
261 	struct list_head rtx_queue;
262 	struct mptcp_data_frag *first_pending;
263 	struct list_head join_list;
264 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
265 	struct sock	*first;
266 	struct mptcp_pm_data	pm;
267 	struct {
268 		u32	space;	/* bytes copied in last measurement window */
269 		u32	copied; /* bytes copied in this measurement window */
270 		u64	time;	/* start time of measurement window */
271 		u64	rtt_us; /* last maximum rtt of subflows */
272 	} rcvq_space;
273 };
274 
275 #define mptcp_lock_sock(___sk, cb) do {					\
276 	struct sock *__sk = (___sk); /* silence macro reuse warning */	\
277 	might_sleep();							\
278 	spin_lock_bh(&__sk->sk_lock.slock);				\
279 	if (__sk->sk_lock.owned)					\
280 		__lock_sock(__sk);					\
281 	cb;								\
282 	__sk->sk_lock.owned = 1;					\
283 	spin_unlock(&__sk->sk_lock.slock);				\
284 	mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_);		\
285 	local_bh_enable();						\
286 } while (0)
287 
288 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
289 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
290 
291 #define mptcp_for_each_subflow(__msk, __subflow)			\
292 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
293 
294 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
295 {
296 	return (struct mptcp_sock *)sk;
297 }
298 
299 static inline int __mptcp_space(const struct sock *sk)
300 {
301 	return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_pending);
302 }
303 
304 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
305 {
306 	const struct mptcp_sock *msk = mptcp_sk(sk);
307 
308 	return READ_ONCE(msk->first_pending);
309 }
310 
311 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
312 {
313 	struct mptcp_sock *msk = mptcp_sk(sk);
314 	struct mptcp_data_frag *cur;
315 
316 	cur = msk->first_pending;
317 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
318 						     list_next_entry(cur, list);
319 }
320 
321 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
322 {
323 	struct mptcp_sock *msk = mptcp_sk(sk);
324 
325 	if (!msk->first_pending)
326 		return NULL;
327 
328 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
329 		return NULL;
330 
331 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
332 }
333 
334 static inline struct mptcp_data_frag *mptcp_rtx_tail(const struct sock *sk)
335 {
336 	struct mptcp_sock *msk = mptcp_sk(sk);
337 
338 	if (!before64(msk->snd_nxt, READ_ONCE(msk->snd_una)))
339 		return NULL;
340 
341 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
342 }
343 
344 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
345 {
346 	struct mptcp_sock *msk = mptcp_sk(sk);
347 
348 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
349 }
350 
351 struct mptcp_subflow_request_sock {
352 	struct	tcp_request_sock sk;
353 	u16	mp_capable : 1,
354 		mp_join : 1,
355 		backup : 1;
356 	u8	local_id;
357 	u8	remote_id;
358 	u64	local_key;
359 	u64	idsn;
360 	u32	token;
361 	u32	ssn_offset;
362 	u64	thmac;
363 	u32	local_nonce;
364 	u32	remote_nonce;
365 	struct mptcp_sock	*msk;
366 	struct hlist_nulls_node token_node;
367 };
368 
369 static inline struct mptcp_subflow_request_sock *
370 mptcp_subflow_rsk(const struct request_sock *rsk)
371 {
372 	return (struct mptcp_subflow_request_sock *)rsk;
373 }
374 
375 enum mptcp_data_avail {
376 	MPTCP_SUBFLOW_NODATA,
377 	MPTCP_SUBFLOW_DATA_AVAIL,
378 	MPTCP_SUBFLOW_OOO_DATA
379 };
380 
381 struct mptcp_delegated_action {
382 	struct napi_struct napi;
383 	struct list_head head;
384 };
385 
386 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
387 
388 #define MPTCP_DELEGATE_SEND		0
389 
390 /* MPTCP subflow context */
391 struct mptcp_subflow_context {
392 	struct	list_head node;/* conn_list of subflows */
393 	u64	local_key;
394 	u64	remote_key;
395 	u64	idsn;
396 	u64	map_seq;
397 	u32	snd_isn;
398 	u32	token;
399 	u32	rel_write_seq;
400 	u32	map_subflow_seq;
401 	u32	ssn_offset;
402 	u32	map_data_len;
403 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
404 		request_join : 1,   /* send MP_JOIN */
405 		request_bkup : 1,
406 		mp_capable : 1,	    /* remote is MPTCP capable */
407 		mp_join : 1,	    /* remote is JOINing */
408 		fully_established : 1,	    /* path validated */
409 		pm_notified : 1,    /* PM hook called for established status */
410 		conn_finished : 1,
411 		map_valid : 1,
412 		mpc_map : 1,
413 		backup : 1,
414 		send_mp_prio : 1,
415 		rx_eof : 1,
416 		can_ack : 1,        /* only after processing the remote a key */
417 		disposable : 1;	    /* ctx can be free at ulp release time */
418 	enum mptcp_data_avail data_avail;
419 	u32	remote_nonce;
420 	u64	thmac;
421 	u32	local_nonce;
422 	u32	remote_token;
423 	u8	hmac[MPTCPOPT_HMAC_LEN];
424 	u8	local_id;
425 	u8	remote_id;
426 
427 	long	delegated_status;
428 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
429 
430 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
431 	struct	sock *conn;	    /* parent mptcp_sock */
432 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
433 	void	(*tcp_data_ready)(struct sock *sk);
434 	void	(*tcp_state_change)(struct sock *sk);
435 	void	(*tcp_write_space)(struct sock *sk);
436 
437 	struct	rcu_head rcu;
438 };
439 
440 static inline struct mptcp_subflow_context *
441 mptcp_subflow_ctx(const struct sock *sk)
442 {
443 	struct inet_connection_sock *icsk = inet_csk(sk);
444 
445 	/* Use RCU on icsk_ulp_data only for sock diag code */
446 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
447 }
448 
449 static inline struct sock *
450 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
451 {
452 	return subflow->tcp_sock;
453 }
454 
455 static inline u64
456 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
457 {
458 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
459 		      subflow->ssn_offset -
460 		      subflow->map_subflow_seq;
461 }
462 
463 static inline u64
464 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
465 {
466 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
467 }
468 
469 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
470 					     struct mptcp_subflow_context *subflow)
471 {
472 	sock_hold(mptcp_subflow_tcp_sock(subflow));
473 	spin_lock_bh(&msk->join_list_lock);
474 	list_add_tail(&subflow->node, &msk->join_list);
475 	spin_unlock_bh(&msk->join_list_lock);
476 }
477 
478 void mptcp_subflow_process_delegated(struct sock *ssk);
479 
480 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
481 {
482 	struct mptcp_delegated_action *delegated;
483 	bool schedule;
484 
485 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
486 	 * ensures the below list check sees list updates done prior to status
487 	 * bit changes
488 	 */
489 	if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
490 		/* still on delegated list from previous scheduling */
491 		if (!list_empty(&subflow->delegated_node))
492 			return;
493 
494 		/* the caller held the subflow bh socket lock */
495 		lockdep_assert_in_softirq();
496 
497 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
498 		schedule = list_empty(&delegated->head);
499 		list_add_tail(&subflow->delegated_node, &delegated->head);
500 		sock_hold(mptcp_subflow_tcp_sock(subflow));
501 		if (schedule)
502 			napi_schedule(&delegated->napi);
503 	}
504 }
505 
506 static inline struct mptcp_subflow_context *
507 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
508 {
509 	struct mptcp_subflow_context *ret;
510 
511 	if (list_empty(&delegated->head))
512 		return NULL;
513 
514 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
515 	list_del_init(&ret->delegated_node);
516 	return ret;
517 }
518 
519 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
520 {
521 	return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
522 }
523 
524 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
525 {
526 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
527 	 * touching the status bit
528 	 */
529 	smp_wmb();
530 	clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
531 }
532 
533 int mptcp_is_enabled(struct net *net);
534 unsigned int mptcp_get_add_addr_timeout(struct net *net);
535 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
536 				     struct mptcp_options_received *mp_opt);
537 bool mptcp_subflow_data_available(struct sock *sk);
538 void __init mptcp_subflow_init(void);
539 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
540 void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
541 		       struct mptcp_subflow_context *subflow);
542 void mptcp_subflow_reset(struct sock *ssk);
543 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
544 
545 /* called with sk socket lock held */
546 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
547 			    const struct mptcp_addr_info *remote);
548 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
549 
550 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
551 					      struct mptcp_subflow_context *ctx)
552 {
553 	sk->sk_data_ready = ctx->tcp_data_ready;
554 	sk->sk_state_change = ctx->tcp_state_change;
555 	sk->sk_write_space = ctx->tcp_write_space;
556 
557 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
558 }
559 
560 void __init mptcp_proto_init(void);
561 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
562 int __init mptcp_proto_v6_init(void);
563 #endif
564 
565 struct sock *mptcp_sk_clone(const struct sock *sk,
566 			    const struct mptcp_options_received *mp_opt,
567 			    struct request_sock *req);
568 void mptcp_get_options(const struct sk_buff *skb,
569 		       struct mptcp_options_received *mp_opt);
570 
571 void mptcp_finish_connect(struct sock *sk);
572 static inline bool mptcp_is_fully_established(struct sock *sk)
573 {
574 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
575 	       READ_ONCE(mptcp_sk(sk)->fully_established);
576 }
577 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
578 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
579 bool mptcp_finish_join(struct sock *sk);
580 bool mptcp_schedule_work(struct sock *sk);
581 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
582 void __mptcp_data_acked(struct sock *sk);
583 void mptcp_subflow_eof(struct sock *sk);
584 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
585 void __mptcp_flush_join_list(struct mptcp_sock *msk);
586 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
587 {
588 	return READ_ONCE(msk->snd_data_fin_enable) &&
589 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
590 }
591 
592 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
593 {
594 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
595 		return false;
596 
597 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
598 	return true;
599 }
600 
601 static inline void mptcp_write_space(struct sock *sk)
602 {
603 	if (sk_stream_is_writeable(sk)) {
604 		/* pairs with memory barrier in mptcp_poll */
605 		smp_mb();
606 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
607 			sk_stream_write_space(sk);
608 	}
609 }
610 
611 void mptcp_destroy_common(struct mptcp_sock *msk);
612 
613 void __init mptcp_token_init(void);
614 static inline void mptcp_token_init_request(struct request_sock *req)
615 {
616 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
617 }
618 
619 int mptcp_token_new_request(struct request_sock *req);
620 void mptcp_token_destroy_request(struct request_sock *req);
621 int mptcp_token_new_connect(struct sock *sk);
622 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
623 			struct mptcp_sock *msk);
624 bool mptcp_token_exists(u32 token);
625 struct mptcp_sock *mptcp_token_get_sock(u32 token);
626 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
627 					 long *s_num);
628 void mptcp_token_destroy(struct mptcp_sock *msk);
629 
630 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
631 
632 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
633 
634 void __init mptcp_pm_init(void);
635 void mptcp_pm_data_init(struct mptcp_sock *msk);
636 void mptcp_pm_new_connection(struct mptcp_sock *msk, int server_side);
637 void mptcp_pm_fully_established(struct mptcp_sock *msk);
638 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
639 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
640 void mptcp_pm_subflow_established(struct mptcp_sock *msk,
641 				  struct mptcp_subflow_context *subflow);
642 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
643 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
644 				const struct mptcp_addr_info *addr);
645 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
646 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, u8 rm_id);
647 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
648 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
649 				 struct mptcp_addr_info *addr,
650 				 u8 bkup);
651 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
652 struct mptcp_pm_add_entry *
653 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
654 		       struct mptcp_addr_info *addr);
655 
656 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
657 			   const struct mptcp_addr_info *addr,
658 			   bool echo, bool port);
659 int mptcp_pm_remove_addr(struct mptcp_sock *msk, u8 local_id);
660 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, u8 local_id);
661 
662 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
663 {
664 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
665 }
666 
667 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
668 {
669 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
670 }
671 
672 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk)
673 {
674 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6);
675 }
676 
677 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk)
678 {
679 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT);
680 }
681 
682 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
683 {
684 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
685 }
686 
687 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
688 {
689 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
690 
691 	if (family == AF_INET6)
692 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
693 	if (!echo)
694 		len += MPTCPOPT_THMAC_LEN;
695 	if (port)
696 		len += TCPOLEN_MPTCP_PORT_LEN;
697 
698 	return len;
699 }
700 
701 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
702 			      struct mptcp_addr_info *saddr, bool *echo, bool *port);
703 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
704 			     u8 *rm_id);
705 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
706 
707 void __init mptcp_pm_nl_init(void);
708 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
709 void mptcp_pm_nl_fully_established(struct mptcp_sock *msk);
710 void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk);
711 void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk);
712 void mptcp_pm_nl_add_addr_send_ack(struct mptcp_sock *msk);
713 void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk);
714 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, u8 rm_id);
715 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
716 
717 static inline struct mptcp_ext *mptcp_get_ext(struct sk_buff *skb)
718 {
719 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
720 }
721 
722 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
723 
724 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
725 {
726 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
727 }
728 
729 static inline bool mptcp_check_fallback(const struct sock *sk)
730 {
731 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
732 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
733 
734 	return __mptcp_check_fallback(msk);
735 }
736 
737 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
738 {
739 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
740 		pr_debug("TCP fallback already done (msk=%p)", msk);
741 		return;
742 	}
743 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
744 }
745 
746 static inline void mptcp_do_fallback(struct sock *sk)
747 {
748 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
749 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
750 
751 	__mptcp_do_fallback(msk);
752 }
753 
754 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
755 
756 static inline bool subflow_simultaneous_connect(struct sock *sk)
757 {
758 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
759 	struct sock *parent = subflow->conn;
760 
761 	return sk->sk_state == TCP_ESTABLISHED &&
762 	       !mptcp_sk(parent)->pm.server_side &&
763 	       !subflow->conn_finished;
764 }
765 
766 #ifdef CONFIG_SYN_COOKIES
767 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
768 				       struct sk_buff *skb);
769 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
770 					struct sk_buff *skb);
771 void __init mptcp_join_cookie_init(void);
772 #else
773 static inline void
774 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
775 				  struct sk_buff *skb) {}
776 static inline bool
777 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
778 				   struct sk_buff *skb)
779 {
780 	return false;
781 }
782 
783 static inline void mptcp_join_cookie_init(void) {}
784 #endif
785 
786 #endif /* __MPTCP_PROTOCOL_H */
787