xref: /linux/net/mptcp/protocol.h (revision 41fb0cf1bced59c1fe178cf6cc9f716b5da9e40e)
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 #include <uapi/linux/mptcp.h>
14 
15 #define MPTCP_SUPPORTED_VERSION	1
16 
17 /* MPTCP option bits */
18 #define OPTION_MPTCP_MPC_SYN	BIT(0)
19 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
20 #define OPTION_MPTCP_MPC_ACK	BIT(2)
21 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
22 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
23 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
24 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
25 #define OPTION_MPTCP_RM_ADDR	BIT(7)
26 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
27 #define OPTION_MPTCP_PRIO	BIT(9)
28 #define OPTION_MPTCP_RST	BIT(10)
29 #define OPTION_MPTCP_DSS	BIT(11)
30 #define OPTION_MPTCP_FAIL	BIT(12)
31 
32 #define OPTION_MPTCP_CSUMREQD	BIT(13)
33 
34 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
35 				 OPTION_MPTCP_MPC_ACK)
36 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
37 				 OPTION_MPTCP_MPJ_ACK)
38 
39 /* MPTCP option subtypes */
40 #define MPTCPOPT_MP_CAPABLE	0
41 #define MPTCPOPT_MP_JOIN	1
42 #define MPTCPOPT_DSS		2
43 #define MPTCPOPT_ADD_ADDR	3
44 #define MPTCPOPT_RM_ADDR	4
45 #define MPTCPOPT_MP_PRIO	5
46 #define MPTCPOPT_MP_FAIL	6
47 #define MPTCPOPT_MP_FASTCLOSE	7
48 #define MPTCPOPT_RST		8
49 
50 /* MPTCP suboption lengths */
51 #define TCPOLEN_MPTCP_MPC_SYN		4
52 #define TCPOLEN_MPTCP_MPC_SYNACK	12
53 #define TCPOLEN_MPTCP_MPC_ACK		20
54 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
55 #define TCPOLEN_MPTCP_MPJ_SYN		12
56 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
57 #define TCPOLEN_MPTCP_MPJ_ACK		24
58 #define TCPOLEN_MPTCP_DSS_BASE		4
59 #define TCPOLEN_MPTCP_DSS_ACK32		4
60 #define TCPOLEN_MPTCP_DSS_ACK64		8
61 #define TCPOLEN_MPTCP_DSS_MAP32		10
62 #define TCPOLEN_MPTCP_DSS_MAP64		14
63 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
64 #define TCPOLEN_MPTCP_ADD_ADDR		16
65 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
66 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
68 #define TCPOLEN_MPTCP_ADD_ADDR6		28
69 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
70 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
72 #define TCPOLEN_MPTCP_PORT_LEN		2
73 #define TCPOLEN_MPTCP_PORT_ALIGN	2
74 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
75 #define TCPOLEN_MPTCP_PRIO		3
76 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
77 #define TCPOLEN_MPTCP_FASTCLOSE		12
78 #define TCPOLEN_MPTCP_RST		4
79 #define TCPOLEN_MPTCP_FAIL		12
80 
81 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
82 
83 /* MPTCP MP_JOIN flags */
84 #define MPTCPOPT_BACKUP		BIT(0)
85 #define MPTCPOPT_HMAC_LEN	20
86 #define MPTCPOPT_THMAC_LEN	8
87 
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK	(0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
94 #define MPTCP_CAP_FLAG_MASK	(0x1F)
95 
96 /* MPTCP DSS flags */
97 #define MPTCP_DSS_DATA_FIN	BIT(4)
98 #define MPTCP_DSS_DSN64		BIT(3)
99 #define MPTCP_DSS_HAS_MAP	BIT(2)
100 #define MPTCP_DSS_ACK64		BIT(1)
101 #define MPTCP_DSS_HAS_ACK	BIT(0)
102 #define MPTCP_DSS_FLAG_MASK	(0x1F)
103 
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO		BIT(0)
106 
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP		BIT(0)
109 
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT	BIT(0)
112 
113 /* MPTCP socket flags */
114 #define MPTCP_DATA_READY	0
115 #define MPTCP_NOSPACE		1
116 #define MPTCP_WORK_RTX		2
117 #define MPTCP_WORK_EOF		3
118 #define MPTCP_FALLBACK_DONE	4
119 #define MPTCP_WORK_CLOSE_SUBFLOW 5
120 #define MPTCP_PUSH_PENDING	6
121 #define MPTCP_CLEAN_UNA		7
122 #define MPTCP_ERROR_REPORT	8
123 #define MPTCP_RETRANSMIT	9
124 #define MPTCP_WORK_SYNC_SETSOCKOPT 10
125 #define MPTCP_CONNECTED		11
126 
127 static inline bool before64(__u64 seq1, __u64 seq2)
128 {
129 	return (__s64)(seq1 - seq2) < 0;
130 }
131 
132 #define after64(seq2, seq1)	before64(seq1, seq2)
133 
134 struct mptcp_options_received {
135 	u64	sndr_key;
136 	u64	rcvr_key;
137 	u64	data_ack;
138 	u64	data_seq;
139 	u32	subflow_seq;
140 	u16	data_len;
141 	__sum16	csum;
142 	u16	suboptions;
143 	u32	token;
144 	u32	nonce;
145 	u16	use_map:1,
146 		dsn64:1,
147 		data_fin:1,
148 		use_ack:1,
149 		ack64:1,
150 		mpc_map:1,
151 		reset_reason:4,
152 		reset_transient:1,
153 		echo:1,
154 		backup:1,
155 		deny_join_id0:1,
156 		__unused:2;
157 	u8	join_id;
158 	u64	thmac;
159 	u8	hmac[MPTCPOPT_HMAC_LEN];
160 	struct mptcp_addr_info addr;
161 	struct mptcp_rm_list rm_list;
162 	u64	ahmac;
163 	u64	fail_seq;
164 };
165 
166 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
167 {
168 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
169 		     ((nib & 0xF) << 8) | field);
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_RM_ADDR_SIGNAL,
185 };
186 
187 struct mptcp_pm_data {
188 	struct mptcp_addr_info local;
189 	struct mptcp_addr_info remote;
190 	struct list_head anno_list;
191 
192 	spinlock_t	lock;		/*protects the whole PM data */
193 
194 	u8		addr_signal;
195 	bool		server_side;
196 	bool		work_pending;
197 	bool		accept_addr;
198 	bool		accept_subflow;
199 	bool		remote_deny_join_id0;
200 	u8		add_addr_signaled;
201 	u8		add_addr_accepted;
202 	u8		local_addr_used;
203 	u8		subflows;
204 	u8		status;
205 	struct mptcp_rm_list rm_list_tx;
206 	struct mptcp_rm_list rm_list_rx;
207 };
208 
209 struct mptcp_data_frag {
210 	struct list_head list;
211 	u64 data_seq;
212 	u16 data_len;
213 	u16 offset;
214 	u16 overhead;
215 	u16 already_sent;
216 	struct page *page;
217 };
218 
219 /* MPTCP connection sock */
220 struct mptcp_sock {
221 	/* inet_connection_sock must be the first member */
222 	struct inet_connection_sock sk;
223 	u64		local_key;
224 	u64		remote_key;
225 	u64		write_seq;
226 	u64		snd_nxt;
227 	u64		ack_seq;
228 	u64		rcv_wnd_sent;
229 	u64		rcv_data_fin_seq;
230 	int		rmem_fwd_alloc;
231 	struct sock	*last_snd;
232 	int		snd_burst;
233 	int		old_wspace;
234 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
235 						 * recovery related fields are under data_lock
236 						 * protection
237 						 */
238 	u64		snd_una;
239 	u64		wnd_end;
240 	unsigned long	timer_ival;
241 	u32		token;
242 	int		rmem_released;
243 	unsigned long	flags;
244 	bool		recovery;		/* closing subflow write queue reinjected */
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 	bool		csum_enabled;
252 	u8		recvmsg_inq:1,
253 			cork:1,
254 			nodelay:1;
255 	spinlock_t	join_list_lock;
256 	struct work_struct work;
257 	struct sk_buff  *ooo_last_skb;
258 	struct rb_root  out_of_order_queue;
259 	struct sk_buff_head receive_queue;
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 	u32 setsockopt_seq;
275 	char		ca_name[TCP_CA_NAME_MAX];
276 };
277 
278 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
279 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
280 
281 #define mptcp_for_each_subflow(__msk, __subflow)			\
282 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
283 
284 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
285 {
286 	sock_owned_by_me((const struct sock *)msk);
287 }
288 
289 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
290 {
291 	return (struct mptcp_sock *)sk;
292 }
293 
294 /* the msk socket don't use the backlog, also account for the bulk
295  * free memory
296  */
297 static inline int __mptcp_rmem(const struct sock *sk)
298 {
299 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
300 }
301 
302 static inline int __mptcp_space(const struct sock *sk)
303 {
304 	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
305 }
306 
307 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
308 {
309 	const struct mptcp_sock *msk = mptcp_sk(sk);
310 
311 	return READ_ONCE(msk->first_pending);
312 }
313 
314 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
315 {
316 	struct mptcp_sock *msk = mptcp_sk(sk);
317 	struct mptcp_data_frag *cur;
318 
319 	cur = msk->first_pending;
320 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
321 						     list_next_entry(cur, list);
322 }
323 
324 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
325 {
326 	struct mptcp_sock *msk = mptcp_sk(sk);
327 
328 	if (!msk->first_pending)
329 		return NULL;
330 
331 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
332 		return NULL;
333 
334 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
335 }
336 
337 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
338 {
339 	struct mptcp_sock *msk = mptcp_sk(sk);
340 
341 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
342 		return NULL;
343 
344 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
345 }
346 
347 struct csum_pseudo_header {
348 	__be64 data_seq;
349 	__be32 subflow_seq;
350 	__be16 data_len;
351 	__sum16 csum;
352 };
353 
354 struct mptcp_subflow_request_sock {
355 	struct	tcp_request_sock sk;
356 	u16	mp_capable : 1,
357 		mp_join : 1,
358 		backup : 1,
359 		csum_reqd : 1,
360 		allow_join_id0 : 1;
361 	u8	local_id;
362 	u8	remote_id;
363 	u64	local_key;
364 	u64	idsn;
365 	u32	token;
366 	u32	ssn_offset;
367 	u64	thmac;
368 	u32	local_nonce;
369 	u32	remote_nonce;
370 	struct mptcp_sock	*msk;
371 	struct hlist_nulls_node token_node;
372 };
373 
374 static inline struct mptcp_subflow_request_sock *
375 mptcp_subflow_rsk(const struct request_sock *rsk)
376 {
377 	return (struct mptcp_subflow_request_sock *)rsk;
378 }
379 
380 enum mptcp_data_avail {
381 	MPTCP_SUBFLOW_NODATA,
382 	MPTCP_SUBFLOW_DATA_AVAIL,
383 };
384 
385 struct mptcp_delegated_action {
386 	struct napi_struct napi;
387 	struct list_head head;
388 };
389 
390 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
391 
392 #define MPTCP_DELEGATE_SEND		0
393 #define MPTCP_DELEGATE_ACK		1
394 
395 /* MPTCP subflow context */
396 struct mptcp_subflow_context {
397 	struct	list_head node;/* conn_list of subflows */
398 	u64	local_key;
399 	u64	remote_key;
400 	u64	idsn;
401 	u64	map_seq;
402 	u32	snd_isn;
403 	u32	token;
404 	u32	rel_write_seq;
405 	u32	map_subflow_seq;
406 	u32	ssn_offset;
407 	u32	map_data_len;
408 	__wsum	map_data_csum;
409 	u32	map_csum_len;
410 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
411 		request_join : 1,   /* send MP_JOIN */
412 		request_bkup : 1,
413 		mp_capable : 1,	    /* remote is MPTCP capable */
414 		mp_join : 1,	    /* remote is JOINing */
415 		fully_established : 1,	    /* path validated */
416 		pm_notified : 1,    /* PM hook called for established status */
417 		conn_finished : 1,
418 		map_valid : 1,
419 		map_csum_reqd : 1,
420 		map_data_fin : 1,
421 		mpc_map : 1,
422 		backup : 1,
423 		send_mp_prio : 1,
424 		send_mp_fail : 1,
425 		rx_eof : 1,
426 		can_ack : 1,        /* only after processing the remote a key */
427 		disposable : 1,	    /* ctx can be free at ulp release time */
428 		stale : 1;	    /* unable to snd/rcv data, do not use for xmit */
429 	enum mptcp_data_avail data_avail;
430 	u32	remote_nonce;
431 	u64	thmac;
432 	u32	local_nonce;
433 	u32	remote_token;
434 	u8	hmac[MPTCPOPT_HMAC_LEN];
435 	u8	local_id;
436 	u8	remote_id;
437 	u8	reset_seen:1;
438 	u8	reset_transient:1;
439 	u8	reset_reason:4;
440 	u8	stale_count;
441 
442 	long	delegated_status;
443 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
444 
445 	u32	setsockopt_seq;
446 	u32	stale_rcv_tstamp;
447 
448 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
449 	struct	sock *conn;	    /* parent mptcp_sock */
450 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
451 	void	(*tcp_data_ready)(struct sock *sk);
452 	void	(*tcp_state_change)(struct sock *sk);
453 	void	(*tcp_write_space)(struct sock *sk);
454 	void	(*tcp_error_report)(struct sock *sk);
455 
456 	struct	rcu_head rcu;
457 };
458 
459 static inline struct mptcp_subflow_context *
460 mptcp_subflow_ctx(const struct sock *sk)
461 {
462 	struct inet_connection_sock *icsk = inet_csk(sk);
463 
464 	/* Use RCU on icsk_ulp_data only for sock diag code */
465 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
466 }
467 
468 static inline struct sock *
469 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
470 {
471 	return subflow->tcp_sock;
472 }
473 
474 static inline u64
475 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
476 {
477 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
478 		      subflow->ssn_offset -
479 		      subflow->map_subflow_seq;
480 }
481 
482 static inline u64
483 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
484 {
485 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
486 }
487 
488 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
489 					     struct mptcp_subflow_context *subflow)
490 {
491 	sock_hold(mptcp_subflow_tcp_sock(subflow));
492 	spin_lock_bh(&msk->join_list_lock);
493 	list_add_tail(&subflow->node, &msk->join_list);
494 	spin_unlock_bh(&msk->join_list_lock);
495 }
496 
497 void mptcp_subflow_process_delegated(struct sock *ssk);
498 
499 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
500 {
501 	struct mptcp_delegated_action *delegated;
502 	bool schedule;
503 
504 	/* the caller held the subflow bh socket lock */
505 	lockdep_assert_in_softirq();
506 
507 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
508 	 * ensures the below list check sees list updates done prior to status
509 	 * bit changes
510 	 */
511 	if (!test_and_set_bit(action, &subflow->delegated_status)) {
512 		/* still on delegated list from previous scheduling */
513 		if (!list_empty(&subflow->delegated_node))
514 			return;
515 
516 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
517 		schedule = list_empty(&delegated->head);
518 		list_add_tail(&subflow->delegated_node, &delegated->head);
519 		sock_hold(mptcp_subflow_tcp_sock(subflow));
520 		if (schedule)
521 			napi_schedule(&delegated->napi);
522 	}
523 }
524 
525 static inline struct mptcp_subflow_context *
526 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
527 {
528 	struct mptcp_subflow_context *ret;
529 
530 	if (list_empty(&delegated->head))
531 		return NULL;
532 
533 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
534 	list_del_init(&ret->delegated_node);
535 	return ret;
536 }
537 
538 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
539 {
540 	return !!READ_ONCE(subflow->delegated_status);
541 }
542 
543 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
544 {
545 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
546 	 * touching the status bit
547 	 */
548 	smp_wmb();
549 	clear_bit(action, &subflow->delegated_status);
550 }
551 
552 int mptcp_is_enabled(const struct net *net);
553 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
554 int mptcp_is_checksum_enabled(const struct net *net);
555 int mptcp_allow_join_id0(const struct net *net);
556 unsigned int mptcp_stale_loss_cnt(const struct net *net);
557 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
558 				     struct mptcp_options_received *mp_opt);
559 bool __mptcp_retransmit_pending_data(struct sock *sk);
560 void mptcp_check_and_set_pending(struct sock *sk);
561 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
562 bool mptcp_subflow_data_available(struct sock *sk);
563 void __init mptcp_subflow_init(void);
564 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
565 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
566 		     struct mptcp_subflow_context *subflow);
567 void mptcp_subflow_send_ack(struct sock *ssk);
568 void mptcp_subflow_reset(struct sock *ssk);
569 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
570 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
571 
572 /* called with sk socket lock held */
573 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
574 			    const struct mptcp_addr_info *remote);
575 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
576 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
577 			 struct sockaddr_storage *addr,
578 			 unsigned short family);
579 
580 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
581 {
582 	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
583 
584 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
585 	if (subflow->request_join && !subflow->fully_established)
586 		return false;
587 
588 	/* only send if our side has not closed yet */
589 	return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
590 }
591 
592 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
593 
594 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
595 
596 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
597 					      struct mptcp_subflow_context *ctx)
598 {
599 	sk->sk_data_ready = ctx->tcp_data_ready;
600 	sk->sk_state_change = ctx->tcp_state_change;
601 	sk->sk_write_space = ctx->tcp_write_space;
602 	sk->sk_error_report = ctx->tcp_error_report;
603 
604 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
605 }
606 
607 static inline bool mptcp_has_another_subflow(struct sock *ssk)
608 {
609 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp;
610 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
611 
612 	mptcp_for_each_subflow(msk, tmp) {
613 		if (tmp != subflow)
614 			return true;
615 	}
616 
617 	return false;
618 }
619 
620 void __init mptcp_proto_init(void);
621 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
622 int __init mptcp_proto_v6_init(void);
623 #endif
624 
625 struct sock *mptcp_sk_clone(const struct sock *sk,
626 			    const struct mptcp_options_received *mp_opt,
627 			    struct request_sock *req);
628 void mptcp_get_options(const struct sock *sk,
629 		       const struct sk_buff *skb,
630 		       struct mptcp_options_received *mp_opt);
631 
632 void mptcp_finish_connect(struct sock *sk);
633 void __mptcp_set_connected(struct sock *sk);
634 static inline bool mptcp_is_fully_established(struct sock *sk)
635 {
636 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
637 	       READ_ONCE(mptcp_sk(sk)->fully_established);
638 }
639 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
640 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
641 bool mptcp_finish_join(struct sock *sk);
642 bool mptcp_schedule_work(struct sock *sk);
643 int mptcp_setsockopt(struct sock *sk, int level, int optname,
644 		     sockptr_t optval, unsigned int optlen);
645 int mptcp_getsockopt(struct sock *sk, int level, int optname,
646 		     char __user *optval, int __user *option);
647 
648 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
649 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
650 {
651 	if (use_64bit)
652 		return cur_seq;
653 
654 	return __mptcp_expand_seq(old_seq, cur_seq);
655 }
656 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
657 void __mptcp_data_acked(struct sock *sk);
658 void __mptcp_error_report(struct sock *sk);
659 void mptcp_subflow_eof(struct sock *sk);
660 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
661 void __mptcp_flush_join_list(struct mptcp_sock *msk);
662 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
663 {
664 	return READ_ONCE(msk->snd_data_fin_enable) &&
665 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
666 }
667 
668 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
669 {
670 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
671 		return false;
672 
673 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
674 	return true;
675 }
676 
677 static inline void mptcp_write_space(struct sock *sk)
678 {
679 	if (sk_stream_is_writeable(sk)) {
680 		/* pairs with memory barrier in mptcp_poll */
681 		smp_mb();
682 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
683 			sk_stream_write_space(sk);
684 	}
685 }
686 
687 void mptcp_destroy_common(struct mptcp_sock *msk);
688 
689 #define MPTCP_TOKEN_MAX_RETRIES	4
690 
691 void __init mptcp_token_init(void);
692 static inline void mptcp_token_init_request(struct request_sock *req)
693 {
694 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
695 }
696 
697 int mptcp_token_new_request(struct request_sock *req);
698 void mptcp_token_destroy_request(struct request_sock *req);
699 int mptcp_token_new_connect(struct sock *sk);
700 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
701 			struct mptcp_sock *msk);
702 bool mptcp_token_exists(u32 token);
703 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
704 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
705 					 long *s_num);
706 void mptcp_token_destroy(struct mptcp_sock *msk);
707 
708 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
709 
710 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
711 
712 void __init mptcp_pm_init(void);
713 void mptcp_pm_data_init(struct mptcp_sock *msk);
714 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
715 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
716 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
717 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
718 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
719 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
720 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
721 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
722 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
723 				const struct mptcp_addr_info *addr);
724 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
725 			      struct mptcp_addr_info *addr);
726 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
727 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
728 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
729 			       const struct mptcp_rm_list *rm_list);
730 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
731 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
732 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
733 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
734 struct mptcp_pm_add_entry *
735 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
736 		       struct mptcp_addr_info *addr, bool check_id);
737 struct mptcp_pm_add_entry *
738 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
739 				struct mptcp_addr_info *addr);
740 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id,
741 					 u8 *flags, int *ifindex);
742 
743 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
744 			   const struct mptcp_addr_info *addr,
745 			   bool echo);
746 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
747 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
748 
749 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
750 		 const struct sock *ssk, gfp_t gfp);
751 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
752 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
753 
754 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
755 {
756 	return READ_ONCE(msk->pm.addr_signal) &
757 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
758 }
759 
760 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
761 {
762 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
763 }
764 
765 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
766 {
767 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
768 }
769 
770 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
771 {
772 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
773 }
774 
775 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
776 {
777 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
778 
779 	if (family == AF_INET6)
780 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
781 	if (!echo)
782 		len += MPTCPOPT_THMAC_LEN;
783 	/* account for 2 trailing 'nop' options */
784 	if (port)
785 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
786 
787 	return len;
788 }
789 
790 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
791 {
792 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
793 		return -EINVAL;
794 
795 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
796 }
797 
798 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
799 			      unsigned int opt_size, unsigned int remaining,
800 			      struct mptcp_addr_info *addr, bool *echo,
801 			      bool *port, bool *drop_other_suboptions);
802 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
803 			     struct mptcp_rm_list *rm_list);
804 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
805 
806 void __init mptcp_pm_nl_init(void);
807 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
808 void mptcp_pm_nl_work(struct mptcp_sock *msk);
809 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
810 				     const struct mptcp_rm_list *rm_list);
811 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
812 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
813 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
814 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
815 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
816 
817 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
818 void mptcp_sockopt_sync_all(struct mptcp_sock *msk);
819 
820 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
821 {
822 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
823 }
824 
825 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
826 
827 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
828 {
829 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
830 }
831 
832 static inline bool mptcp_check_fallback(const struct sock *sk)
833 {
834 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
835 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
836 
837 	return __mptcp_check_fallback(msk);
838 }
839 
840 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
841 {
842 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
843 		pr_debug("TCP fallback already done (msk=%p)", msk);
844 		return;
845 	}
846 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
847 }
848 
849 static inline void mptcp_do_fallback(struct sock *sk)
850 {
851 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
852 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
853 
854 	__mptcp_do_fallback(msk);
855 }
856 
857 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
858 
859 static inline bool subflow_simultaneous_connect(struct sock *sk)
860 {
861 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
862 	struct sock *parent = subflow->conn;
863 
864 	return sk->sk_state == TCP_ESTABLISHED &&
865 	       !mptcp_sk(parent)->pm.server_side &&
866 	       !subflow->conn_finished;
867 }
868 
869 #ifdef CONFIG_SYN_COOKIES
870 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
871 				       struct sk_buff *skb);
872 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
873 					struct sk_buff *skb);
874 void __init mptcp_join_cookie_init(void);
875 #else
876 static inline void
877 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
878 				  struct sk_buff *skb) {}
879 static inline bool
880 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
881 				   struct sk_buff *skb)
882 {
883 	return false;
884 }
885 
886 static inline void mptcp_join_cookie_init(void) {}
887 #endif
888 
889 #endif /* __MPTCP_PROTOCOL_H */
890