xref: /linux/net/mptcp/protocol.h (revision 987b741c52c7c6c68d46fbaeb95b8d1087f10b7f)
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 
30 /* MPTCP option subtypes */
31 #define MPTCPOPT_MP_CAPABLE	0
32 #define MPTCPOPT_MP_JOIN	1
33 #define MPTCPOPT_DSS		2
34 #define MPTCPOPT_ADD_ADDR	3
35 #define MPTCPOPT_RM_ADDR	4
36 #define MPTCPOPT_MP_PRIO	5
37 #define MPTCPOPT_MP_FAIL	6
38 #define MPTCPOPT_MP_FASTCLOSE	7
39 #define MPTCPOPT_RST		8
40 
41 /* MPTCP suboption lengths */
42 #define TCPOLEN_MPTCP_MPC_SYN		4
43 #define TCPOLEN_MPTCP_MPC_SYNACK	12
44 #define TCPOLEN_MPTCP_MPC_ACK		20
45 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
46 #define TCPOLEN_MPTCP_MPJ_SYN		12
47 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
48 #define TCPOLEN_MPTCP_MPJ_ACK		24
49 #define TCPOLEN_MPTCP_DSS_BASE		4
50 #define TCPOLEN_MPTCP_DSS_ACK32		4
51 #define TCPOLEN_MPTCP_DSS_ACK64		8
52 #define TCPOLEN_MPTCP_DSS_MAP32		10
53 #define TCPOLEN_MPTCP_DSS_MAP64		14
54 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
55 #define TCPOLEN_MPTCP_ADD_ADDR		16
56 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
57 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
58 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
59 #define TCPOLEN_MPTCP_ADD_ADDR6		28
60 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
61 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
62 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
63 #define TCPOLEN_MPTCP_PORT_LEN		2
64 #define TCPOLEN_MPTCP_PORT_ALIGN	2
65 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
66 #define TCPOLEN_MPTCP_PRIO		3
67 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
68 #define TCPOLEN_MPTCP_FASTCLOSE		12
69 #define TCPOLEN_MPTCP_RST		4
70 
71 /* MPTCP MP_JOIN flags */
72 #define MPTCPOPT_BACKUP		BIT(0)
73 #define MPTCPOPT_HMAC_LEN	20
74 #define MPTCPOPT_THMAC_LEN	8
75 
76 /* MPTCP MP_CAPABLE flags */
77 #define MPTCP_VERSION_MASK	(0x0F)
78 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
79 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
80 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
81 #define MPTCP_CAP_FLAG_MASK	(0x3F)
82 
83 /* MPTCP DSS flags */
84 #define MPTCP_DSS_DATA_FIN	BIT(4)
85 #define MPTCP_DSS_DSN64		BIT(3)
86 #define MPTCP_DSS_HAS_MAP	BIT(2)
87 #define MPTCP_DSS_ACK64		BIT(1)
88 #define MPTCP_DSS_HAS_ACK	BIT(0)
89 #define MPTCP_DSS_FLAG_MASK	(0x1F)
90 
91 /* MPTCP ADD_ADDR flags */
92 #define MPTCP_ADDR_ECHO		BIT(0)
93 
94 /* MPTCP MP_PRIO flags */
95 #define MPTCP_PRIO_BKUP		BIT(0)
96 
97 /* MPTCP TCPRST flags */
98 #define MPTCP_RST_TRANSIENT	BIT(0)
99 
100 /* MPTCP socket flags */
101 #define MPTCP_DATA_READY	0
102 #define MPTCP_NOSPACE		1
103 #define MPTCP_WORK_RTX		2
104 #define MPTCP_WORK_EOF		3
105 #define MPTCP_FALLBACK_DONE	4
106 #define MPTCP_WORK_CLOSE_SUBFLOW 5
107 #define MPTCP_PUSH_PENDING	6
108 #define MPTCP_CLEAN_UNA		7
109 #define MPTCP_ERROR_REPORT	8
110 #define MPTCP_RETRANSMIT	9
111 #define MPTCP_WORK_SYNC_SETSOCKOPT 10
112 
113 static inline bool before64(__u64 seq1, __u64 seq2)
114 {
115 	return (__s64)(seq1 - seq2) < 0;
116 }
117 
118 #define after64(seq2, seq1)	before64(seq1, seq2)
119 
120 struct mptcp_options_received {
121 	u64	sndr_key;
122 	u64	rcvr_key;
123 	u64	data_ack;
124 	u64	data_seq;
125 	u32	subflow_seq;
126 	u16	data_len;
127 	u16	mp_capable : 1,
128 		mp_join : 1,
129 		fastclose : 1,
130 		reset : 1,
131 		dss : 1,
132 		add_addr : 1,
133 		rm_addr : 1,
134 		mp_prio : 1,
135 		echo : 1,
136 		backup : 1;
137 	u32	token;
138 	u32	nonce;
139 	u64	thmac;
140 	u8	hmac[MPTCPOPT_HMAC_LEN];
141 	u8	join_id;
142 	u8	use_map:1,
143 		dsn64:1,
144 		data_fin:1,
145 		use_ack:1,
146 		ack64:1,
147 		mpc_map:1,
148 		__unused:2;
149 	struct mptcp_addr_info addr;
150 	struct mptcp_rm_list rm_list;
151 	u64	ahmac;
152 	u8	reset_reason:4;
153 	u8	reset_transient:1;
154 };
155 
156 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
157 {
158 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
159 		     ((nib & 0xF) << 8) | field);
160 }
161 
162 enum mptcp_pm_status {
163 	MPTCP_PM_ADD_ADDR_RECEIVED,
164 	MPTCP_PM_ADD_ADDR_SEND_ACK,
165 	MPTCP_PM_RM_ADDR_RECEIVED,
166 	MPTCP_PM_ESTABLISHED,
167 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
168 	MPTCP_PM_SUBFLOW_ESTABLISHED,
169 };
170 
171 enum mptcp_addr_signal_status {
172 	MPTCP_ADD_ADDR_SIGNAL,
173 	MPTCP_ADD_ADDR_ECHO,
174 	MPTCP_ADD_ADDR_IPV6,
175 	MPTCP_ADD_ADDR_PORT,
176 	MPTCP_RM_ADDR_SIGNAL,
177 };
178 
179 struct mptcp_pm_data {
180 	struct mptcp_addr_info local;
181 	struct mptcp_addr_info remote;
182 	struct list_head anno_list;
183 
184 	spinlock_t	lock;		/*protects the whole PM data */
185 
186 	u8		addr_signal;
187 	bool		server_side;
188 	bool		work_pending;
189 	bool		accept_addr;
190 	bool		accept_subflow;
191 	u8		add_addr_signaled;
192 	u8		add_addr_accepted;
193 	u8		local_addr_used;
194 	u8		subflows;
195 	u8		status;
196 	struct mptcp_rm_list rm_list_tx;
197 	struct mptcp_rm_list rm_list_rx;
198 };
199 
200 struct mptcp_data_frag {
201 	struct list_head list;
202 	u64 data_seq;
203 	u16 data_len;
204 	u16 offset;
205 	u16 overhead;
206 	u16 already_sent;
207 	struct page *page;
208 };
209 
210 /* MPTCP connection sock */
211 struct mptcp_sock {
212 	/* inet_connection_sock must be the first member */
213 	struct inet_connection_sock sk;
214 	u64		local_key;
215 	u64		remote_key;
216 	u64		write_seq;
217 	u64		snd_nxt;
218 	u64		ack_seq;
219 	u64		rcv_wnd_sent;
220 	u64		rcv_data_fin_seq;
221 	int		wmem_reserved;
222 	struct sock	*last_snd;
223 	int		snd_burst;
224 	int		old_wspace;
225 	u64		snd_una;
226 	u64		wnd_end;
227 	unsigned long	timer_ival;
228 	u32		token;
229 	int		rmem_released;
230 	unsigned long	flags;
231 	bool		can_ack;
232 	bool		fully_established;
233 	bool		rcv_data_fin;
234 	bool		snd_data_fin_enable;
235 	bool		rcv_fastclose;
236 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
237 	spinlock_t	join_list_lock;
238 	struct sock	*ack_hint;
239 	struct work_struct work;
240 	struct sk_buff  *ooo_last_skb;
241 	struct rb_root  out_of_order_queue;
242 	struct sk_buff_head receive_queue;
243 	struct sk_buff_head skb_tx_cache;	/* this is wmem accounted */
244 	int		tx_pending_data;
245 	int		size_goal_cache;
246 	struct list_head conn_list;
247 	struct list_head rtx_queue;
248 	struct mptcp_data_frag *first_pending;
249 	struct list_head join_list;
250 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
251 	struct sock	*first;
252 	struct mptcp_pm_data	pm;
253 	struct {
254 		u32	space;	/* bytes copied in last measurement window */
255 		u32	copied; /* bytes copied in this measurement window */
256 		u64	time;	/* start time of measurement window */
257 		u64	rtt_us; /* last maximum rtt of subflows */
258 	} rcvq_space;
259 
260 	u32 setsockopt_seq;
261 	char		ca_name[TCP_CA_NAME_MAX];
262 };
263 
264 #define mptcp_lock_sock(___sk, cb) do {					\
265 	struct sock *__sk = (___sk); /* silence macro reuse warning */	\
266 	might_sleep();							\
267 	spin_lock_bh(&__sk->sk_lock.slock);				\
268 	if (__sk->sk_lock.owned)					\
269 		__lock_sock(__sk);					\
270 	cb;								\
271 	__sk->sk_lock.owned = 1;					\
272 	spin_unlock(&__sk->sk_lock.slock);				\
273 	mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_);		\
274 	local_bh_enable();						\
275 } while (0)
276 
277 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
278 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
279 
280 #define mptcp_for_each_subflow(__msk, __subflow)			\
281 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
282 
283 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
284 {
285 	sock_owned_by_me((const struct sock *)msk);
286 }
287 
288 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
289 {
290 	return (struct mptcp_sock *)sk;
291 }
292 
293 static inline int __mptcp_space(const struct sock *sk)
294 {
295 	return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_released);
296 }
297 
298 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
299 {
300 	const struct mptcp_sock *msk = mptcp_sk(sk);
301 
302 	return READ_ONCE(msk->first_pending);
303 }
304 
305 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
306 {
307 	struct mptcp_sock *msk = mptcp_sk(sk);
308 	struct mptcp_data_frag *cur;
309 
310 	cur = msk->first_pending;
311 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
312 						     list_next_entry(cur, list);
313 }
314 
315 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
316 {
317 	struct mptcp_sock *msk = mptcp_sk(sk);
318 
319 	if (!msk->first_pending)
320 		return NULL;
321 
322 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
323 		return NULL;
324 
325 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
326 }
327 
328 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
329 {
330 	struct mptcp_sock *msk = mptcp_sk(sk);
331 
332 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
333 		return NULL;
334 
335 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
336 }
337 
338 struct mptcp_subflow_request_sock {
339 	struct	tcp_request_sock sk;
340 	u16	mp_capable : 1,
341 		mp_join : 1,
342 		backup : 1;
343 	u8	local_id;
344 	u8	remote_id;
345 	u64	local_key;
346 	u64	idsn;
347 	u32	token;
348 	u32	ssn_offset;
349 	u64	thmac;
350 	u32	local_nonce;
351 	u32	remote_nonce;
352 	struct mptcp_sock	*msk;
353 	struct hlist_nulls_node token_node;
354 };
355 
356 static inline struct mptcp_subflow_request_sock *
357 mptcp_subflow_rsk(const struct request_sock *rsk)
358 {
359 	return (struct mptcp_subflow_request_sock *)rsk;
360 }
361 
362 enum mptcp_data_avail {
363 	MPTCP_SUBFLOW_NODATA,
364 	MPTCP_SUBFLOW_DATA_AVAIL,
365 	MPTCP_SUBFLOW_OOO_DATA
366 };
367 
368 struct mptcp_delegated_action {
369 	struct napi_struct napi;
370 	struct list_head head;
371 };
372 
373 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
374 
375 #define MPTCP_DELEGATE_SEND		0
376 
377 /* MPTCP subflow context */
378 struct mptcp_subflow_context {
379 	struct	list_head node;/* conn_list of subflows */
380 	u64	local_key;
381 	u64	remote_key;
382 	u64	idsn;
383 	u64	map_seq;
384 	u32	snd_isn;
385 	u32	token;
386 	u32	rel_write_seq;
387 	u32	map_subflow_seq;
388 	u32	ssn_offset;
389 	u32	map_data_len;
390 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
391 		request_join : 1,   /* send MP_JOIN */
392 		request_bkup : 1,
393 		mp_capable : 1,	    /* remote is MPTCP capable */
394 		mp_join : 1,	    /* remote is JOINing */
395 		fully_established : 1,	    /* path validated */
396 		pm_notified : 1,    /* PM hook called for established status */
397 		conn_finished : 1,
398 		map_valid : 1,
399 		mpc_map : 1,
400 		backup : 1,
401 		send_mp_prio : 1,
402 		rx_eof : 1,
403 		can_ack : 1,        /* only after processing the remote a key */
404 		disposable : 1;	    /* ctx can be free at ulp release time */
405 	enum mptcp_data_avail data_avail;
406 	u32	remote_nonce;
407 	u64	thmac;
408 	u32	local_nonce;
409 	u32	remote_token;
410 	u8	hmac[MPTCPOPT_HMAC_LEN];
411 	u8	local_id;
412 	u8	remote_id;
413 	u8	reset_seen:1;
414 	u8	reset_transient:1;
415 	u8	reset_reason:4;
416 
417 	long	delegated_status;
418 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
419 
420 	u32 setsockopt_seq;
421 
422 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
423 	struct	sock *conn;	    /* parent mptcp_sock */
424 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
425 	void	(*tcp_data_ready)(struct sock *sk);
426 	void	(*tcp_state_change)(struct sock *sk);
427 	void	(*tcp_write_space)(struct sock *sk);
428 	void	(*tcp_error_report)(struct sock *sk);
429 
430 	struct	rcu_head rcu;
431 };
432 
433 static inline struct mptcp_subflow_context *
434 mptcp_subflow_ctx(const struct sock *sk)
435 {
436 	struct inet_connection_sock *icsk = inet_csk(sk);
437 
438 	/* Use RCU on icsk_ulp_data only for sock diag code */
439 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
440 }
441 
442 static inline struct sock *
443 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
444 {
445 	return subflow->tcp_sock;
446 }
447 
448 static inline u64
449 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
450 {
451 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
452 		      subflow->ssn_offset -
453 		      subflow->map_subflow_seq;
454 }
455 
456 static inline u64
457 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
458 {
459 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
460 }
461 
462 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
463 					     struct mptcp_subflow_context *subflow)
464 {
465 	sock_hold(mptcp_subflow_tcp_sock(subflow));
466 	spin_lock_bh(&msk->join_list_lock);
467 	list_add_tail(&subflow->node, &msk->join_list);
468 	spin_unlock_bh(&msk->join_list_lock);
469 }
470 
471 void mptcp_subflow_process_delegated(struct sock *ssk);
472 
473 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
474 {
475 	struct mptcp_delegated_action *delegated;
476 	bool schedule;
477 
478 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
479 	 * ensures the below list check sees list updates done prior to status
480 	 * bit changes
481 	 */
482 	if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
483 		/* still on delegated list from previous scheduling */
484 		if (!list_empty(&subflow->delegated_node))
485 			return;
486 
487 		/* the caller held the subflow bh socket lock */
488 		lockdep_assert_in_softirq();
489 
490 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
491 		schedule = list_empty(&delegated->head);
492 		list_add_tail(&subflow->delegated_node, &delegated->head);
493 		sock_hold(mptcp_subflow_tcp_sock(subflow));
494 		if (schedule)
495 			napi_schedule(&delegated->napi);
496 	}
497 }
498 
499 static inline struct mptcp_subflow_context *
500 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
501 {
502 	struct mptcp_subflow_context *ret;
503 
504 	if (list_empty(&delegated->head))
505 		return NULL;
506 
507 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
508 	list_del_init(&ret->delegated_node);
509 	return ret;
510 }
511 
512 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
513 {
514 	return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
515 }
516 
517 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
518 {
519 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
520 	 * touching the status bit
521 	 */
522 	smp_wmb();
523 	clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
524 }
525 
526 int mptcp_is_enabled(struct net *net);
527 unsigned int mptcp_get_add_addr_timeout(struct net *net);
528 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
529 				     struct mptcp_options_received *mp_opt);
530 bool mptcp_subflow_data_available(struct sock *sk);
531 void __init mptcp_subflow_init(void);
532 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
533 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
534 		     struct mptcp_subflow_context *subflow);
535 void mptcp_subflow_reset(struct sock *ssk);
536 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
537 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
538 
539 /* called with sk socket lock held */
540 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
541 			    const struct mptcp_addr_info *remote,
542 			    u8 flags, int ifindex);
543 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
544 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
545 			 struct sockaddr_storage *addr,
546 			 unsigned short family);
547 
548 static inline bool mptcp_subflow_active(struct mptcp_subflow_context *subflow)
549 {
550 	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
551 
552 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
553 	if (subflow->request_join && !subflow->fully_established)
554 		return false;
555 
556 	/* only send if our side has not closed yet */
557 	return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
558 }
559 
560 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
561 					      struct mptcp_subflow_context *ctx)
562 {
563 	sk->sk_data_ready = ctx->tcp_data_ready;
564 	sk->sk_state_change = ctx->tcp_state_change;
565 	sk->sk_write_space = ctx->tcp_write_space;
566 	sk->sk_error_report = ctx->tcp_error_report;
567 
568 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
569 }
570 
571 void __init mptcp_proto_init(void);
572 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
573 int __init mptcp_proto_v6_init(void);
574 #endif
575 
576 struct sock *mptcp_sk_clone(const struct sock *sk,
577 			    const struct mptcp_options_received *mp_opt,
578 			    struct request_sock *req);
579 void mptcp_get_options(const struct sk_buff *skb,
580 		       struct mptcp_options_received *mp_opt);
581 
582 void mptcp_finish_connect(struct sock *sk);
583 static inline bool mptcp_is_fully_established(struct sock *sk)
584 {
585 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
586 	       READ_ONCE(mptcp_sk(sk)->fully_established);
587 }
588 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
589 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
590 bool mptcp_finish_join(struct sock *sk);
591 bool mptcp_schedule_work(struct sock *sk);
592 int mptcp_setsockopt(struct sock *sk, int level, int optname,
593 		     sockptr_t optval, unsigned int optlen);
594 int mptcp_getsockopt(struct sock *sk, int level, int optname,
595 		     char __user *optval, int __user *option);
596 
597 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
598 void __mptcp_data_acked(struct sock *sk);
599 void __mptcp_error_report(struct sock *sk);
600 void mptcp_subflow_eof(struct sock *sk);
601 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
602 void __mptcp_flush_join_list(struct mptcp_sock *msk);
603 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
604 {
605 	return READ_ONCE(msk->snd_data_fin_enable) &&
606 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
607 }
608 
609 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
610 {
611 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
612 		return false;
613 
614 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
615 	return true;
616 }
617 
618 static inline void mptcp_write_space(struct sock *sk)
619 {
620 	if (sk_stream_is_writeable(sk)) {
621 		/* pairs with memory barrier in mptcp_poll */
622 		smp_mb();
623 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
624 			sk_stream_write_space(sk);
625 	}
626 }
627 
628 void mptcp_destroy_common(struct mptcp_sock *msk);
629 
630 void __init mptcp_token_init(void);
631 static inline void mptcp_token_init_request(struct request_sock *req)
632 {
633 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
634 }
635 
636 int mptcp_token_new_request(struct request_sock *req);
637 void mptcp_token_destroy_request(struct request_sock *req);
638 int mptcp_token_new_connect(struct sock *sk);
639 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
640 			struct mptcp_sock *msk);
641 bool mptcp_token_exists(u32 token);
642 struct mptcp_sock *mptcp_token_get_sock(u32 token);
643 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
644 					 long *s_num);
645 void mptcp_token_destroy(struct mptcp_sock *msk);
646 
647 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
648 
649 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
650 
651 void __init mptcp_pm_init(void);
652 void mptcp_pm_data_init(struct mptcp_sock *msk);
653 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
654 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
655 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
656 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
657 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
658 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
659 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
660 				const struct mptcp_addr_info *addr);
661 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
662 			      struct mptcp_addr_info *addr);
663 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
664 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
665 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
666 			       const struct mptcp_rm_list *rm_list);
667 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
668 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
669 				 struct mptcp_addr_info *addr,
670 				 u8 bkup);
671 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
672 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
673 struct mptcp_pm_add_entry *
674 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
675 		       struct mptcp_addr_info *addr, bool check_id);
676 struct mptcp_pm_add_entry *
677 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
678 				struct mptcp_addr_info *addr);
679 
680 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
681 			   const struct mptcp_addr_info *addr,
682 			   bool echo);
683 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
684 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
685 
686 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
687 		 const struct sock *ssk, gfp_t gfp);
688 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
689 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
690 
691 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
692 {
693 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
694 }
695 
696 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
697 {
698 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
699 }
700 
701 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk)
702 {
703 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6);
704 }
705 
706 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk)
707 {
708 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT);
709 }
710 
711 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
712 {
713 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
714 }
715 
716 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
717 {
718 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
719 
720 	if (family == AF_INET6)
721 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
722 	if (!echo)
723 		len += MPTCPOPT_THMAC_LEN;
724 	/* account for 2 trailing 'nop' options */
725 	if (port)
726 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
727 
728 	return len;
729 }
730 
731 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
732 {
733 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
734 		return -EINVAL;
735 
736 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
737 }
738 
739 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
740 			      struct mptcp_addr_info *saddr, bool *echo, bool *port);
741 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
742 			     struct mptcp_rm_list *rm_list);
743 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
744 
745 void __init mptcp_pm_nl_init(void);
746 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
747 void mptcp_pm_nl_work(struct mptcp_sock *msk);
748 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
749 				     const struct mptcp_rm_list *rm_list);
750 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
751 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
752 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
753 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
754 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
755 
756 int mptcp_setsockopt(struct sock *sk, int level, int optname,
757 		     sockptr_t optval, unsigned int optlen);
758 
759 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
760 void mptcp_sockopt_sync_all(struct mptcp_sock *msk);
761 
762 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
763 {
764 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
765 }
766 
767 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
768 
769 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
770 {
771 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
772 }
773 
774 static inline bool mptcp_check_fallback(const struct sock *sk)
775 {
776 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
777 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
778 
779 	return __mptcp_check_fallback(msk);
780 }
781 
782 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
783 {
784 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
785 		pr_debug("TCP fallback already done (msk=%p)", msk);
786 		return;
787 	}
788 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
789 }
790 
791 static inline void mptcp_do_fallback(struct sock *sk)
792 {
793 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
794 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
795 
796 	__mptcp_do_fallback(msk);
797 }
798 
799 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
800 
801 static inline bool subflow_simultaneous_connect(struct sock *sk)
802 {
803 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
804 	struct sock *parent = subflow->conn;
805 
806 	return sk->sk_state == TCP_ESTABLISHED &&
807 	       !mptcp_sk(parent)->pm.server_side &&
808 	       !subflow->conn_finished;
809 }
810 
811 #ifdef CONFIG_SYN_COOKIES
812 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
813 				       struct sk_buff *skb);
814 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
815 					struct sk_buff *skb);
816 void __init mptcp_join_cookie_init(void);
817 #else
818 static inline void
819 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
820 				  struct sk_buff *skb) {}
821 static inline bool
822 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
823 				   struct sk_buff *skb)
824 {
825 	return false;
826 }
827 
828 static inline void mptcp_join_cookie_init(void) {}
829 #endif
830 
831 #endif /* __MPTCP_PROTOCOL_H */
832