xref: /openbmc/linux/net/mptcp/protocol.h (revision bffbf6e2)
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_ADD_ADDR6	BIT(7)
26 #define OPTION_MPTCP_RM_ADDR	BIT(8)
27 #define OPTION_MPTCP_FASTCLOSE	BIT(9)
28 #define OPTION_MPTCP_PRIO	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 
40 /* MPTCP suboption lengths */
41 #define TCPOLEN_MPTCP_MPC_SYN		4
42 #define TCPOLEN_MPTCP_MPC_SYNACK	12
43 #define TCPOLEN_MPTCP_MPC_ACK		20
44 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
45 #define TCPOLEN_MPTCP_MPJ_SYN		12
46 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
47 #define TCPOLEN_MPTCP_MPJ_ACK		24
48 #define TCPOLEN_MPTCP_DSS_BASE		4
49 #define TCPOLEN_MPTCP_DSS_ACK32		4
50 #define TCPOLEN_MPTCP_DSS_ACK64		8
51 #define TCPOLEN_MPTCP_DSS_MAP32		10
52 #define TCPOLEN_MPTCP_DSS_MAP64		14
53 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
54 #define TCPOLEN_MPTCP_ADD_ADDR		16
55 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
56 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
57 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
58 #define TCPOLEN_MPTCP_ADD_ADDR6		28
59 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
60 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
61 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
62 #define TCPOLEN_MPTCP_PORT_LEN		2
63 #define TCPOLEN_MPTCP_PORT_ALIGN	2
64 #define TCPOLEN_MPTCP_RM_ADDR_BASE	4
65 #define TCPOLEN_MPTCP_PRIO		3
66 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
67 #define TCPOLEN_MPTCP_FASTCLOSE		12
68 
69 /* MPTCP MP_JOIN flags */
70 #define MPTCPOPT_BACKUP		BIT(0)
71 #define MPTCPOPT_HMAC_LEN	20
72 #define MPTCPOPT_THMAC_LEN	8
73 
74 /* MPTCP MP_CAPABLE flags */
75 #define MPTCP_VERSION_MASK	(0x0F)
76 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
77 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
78 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
79 #define MPTCP_CAP_FLAG_MASK	(0x3F)
80 
81 /* MPTCP DSS flags */
82 #define MPTCP_DSS_DATA_FIN	BIT(4)
83 #define MPTCP_DSS_DSN64		BIT(3)
84 #define MPTCP_DSS_HAS_MAP	BIT(2)
85 #define MPTCP_DSS_ACK64		BIT(1)
86 #define MPTCP_DSS_HAS_ACK	BIT(0)
87 #define MPTCP_DSS_FLAG_MASK	(0x1F)
88 
89 /* MPTCP ADD_ADDR flags */
90 #define MPTCP_ADDR_ECHO		BIT(0)
91 #define MPTCP_ADDR_IPVERSION_4	4
92 #define MPTCP_ADDR_IPVERSION_6	6
93 
94 /* MPTCP MP_PRIO flags */
95 #define MPTCP_PRIO_BKUP		BIT(0)
96 
97 /* MPTCP socket flags */
98 #define MPTCP_DATA_READY	0
99 #define MPTCP_NOSPACE		1
100 #define MPTCP_WORK_RTX		2
101 #define MPTCP_WORK_EOF		3
102 #define MPTCP_FALLBACK_DONE	4
103 #define MPTCP_WORK_CLOSE_SUBFLOW 5
104 #define MPTCP_PUSH_PENDING	6
105 #define MPTCP_CLEAN_UNA		7
106 #define MPTCP_ERROR_REPORT	8
107 
108 static inline bool before64(__u64 seq1, __u64 seq2)
109 {
110 	return (__s64)(seq1 - seq2) < 0;
111 }
112 
113 #define after64(seq2, seq1)	before64(seq1, seq2)
114 
115 struct mptcp_options_received {
116 	u64	sndr_key;
117 	u64	rcvr_key;
118 	u64	data_ack;
119 	u64	data_seq;
120 	u32	subflow_seq;
121 	u16	data_len;
122 	u16	mp_capable : 1,
123 		mp_join : 1,
124 		fastclose : 1,
125 		dss : 1,
126 		add_addr : 1,
127 		rm_addr : 1,
128 		mp_prio : 1,
129 		family : 4,
130 		echo : 1,
131 		backup : 1;
132 	u32	token;
133 	u32	nonce;
134 	u64	thmac;
135 	u8	hmac[MPTCPOPT_HMAC_LEN];
136 	u8	join_id;
137 	u8	use_map:1,
138 		dsn64:1,
139 		data_fin:1,
140 		use_ack:1,
141 		ack64:1,
142 		mpc_map:1,
143 		__unused:2;
144 	u8	addr_id;
145 	u8	rm_id;
146 	union {
147 		struct in_addr	addr;
148 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
149 		struct in6_addr	addr6;
150 #endif
151 	};
152 	u64	ahmac;
153 	u16	port;
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 struct mptcp_addr_info {
163 	sa_family_t		family;
164 	__be16			port;
165 	u8			id;
166 	u8			flags;
167 	int			ifindex;
168 	union {
169 		struct in_addr addr;
170 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
171 		struct in6_addr addr6;
172 #endif
173 	};
174 };
175 
176 enum mptcp_pm_status {
177 	MPTCP_PM_ADD_ADDR_RECEIVED,
178 	MPTCP_PM_ADD_ADDR_SEND_ACK,
179 	MPTCP_PM_RM_ADDR_RECEIVED,
180 	MPTCP_PM_ESTABLISHED,
181 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
182 	MPTCP_PM_SUBFLOW_ESTABLISHED,
183 };
184 
185 enum mptcp_addr_signal_status {
186 	MPTCP_ADD_ADDR_SIGNAL,
187 	MPTCP_ADD_ADDR_ECHO,
188 	MPTCP_ADD_ADDR_IPV6,
189 	MPTCP_ADD_ADDR_PORT,
190 	MPTCP_RM_ADDR_SIGNAL,
191 };
192 
193 struct mptcp_pm_data {
194 	struct mptcp_addr_info local;
195 	struct mptcp_addr_info remote;
196 	struct list_head anno_list;
197 
198 	spinlock_t	lock;		/*protects the whole PM data */
199 
200 	u8		addr_signal;
201 	bool		server_side;
202 	bool		work_pending;
203 	bool		accept_addr;
204 	bool		accept_subflow;
205 	u8		add_addr_signaled;
206 	u8		add_addr_accepted;
207 	u8		local_addr_used;
208 	u8		subflows;
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_released;
243 	unsigned long	flags;
244 	bool		can_ack;
245 	bool		fully_established;
246 	bool		rcv_data_fin;
247 	bool		snd_data_fin_enable;
248 	bool		rcv_fastclose;
249 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
250 	spinlock_t	join_list_lock;
251 	struct sock	*ack_hint;
252 	struct work_struct work;
253 	struct sk_buff  *ooo_last_skb;
254 	struct rb_root  out_of_order_queue;
255 	struct sk_buff_head receive_queue;
256 	struct sk_buff_head skb_tx_cache;	/* this is wmem accounted */
257 	int		tx_pending_data;
258 	int		size_goal_cache;
259 	struct list_head conn_list;
260 	struct list_head rtx_queue;
261 	struct mptcp_data_frag *first_pending;
262 	struct list_head join_list;
263 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
264 	struct sock	*first;
265 	struct mptcp_pm_data	pm;
266 	struct {
267 		u32	space;	/* bytes copied in last measurement window */
268 		u32	copied; /* bytes copied in this measurement window */
269 		u64	time;	/* start time of measurement window */
270 		u64	rtt_us; /* last maximum rtt of subflows */
271 	} rcvq_space;
272 };
273 
274 #define mptcp_lock_sock(___sk, cb) do {					\
275 	struct sock *__sk = (___sk); /* silence macro reuse warning */	\
276 	might_sleep();							\
277 	spin_lock_bh(&__sk->sk_lock.slock);				\
278 	if (__sk->sk_lock.owned)					\
279 		__lock_sock(__sk);					\
280 	cb;								\
281 	__sk->sk_lock.owned = 1;					\
282 	spin_unlock(&__sk->sk_lock.slock);				\
283 	mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_);		\
284 	local_bh_enable();						\
285 } while (0)
286 
287 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
288 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
289 
290 #define mptcp_for_each_subflow(__msk, __subflow)			\
291 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
292 
293 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
294 {
295 	sock_owned_by_me((const struct sock *)msk);
296 }
297 
298 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
299 {
300 	return (struct mptcp_sock *)sk;
301 }
302 
303 static inline int __mptcp_space(const struct sock *sk)
304 {
305 	return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_released);
306 }
307 
308 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
309 {
310 	const struct mptcp_sock *msk = mptcp_sk(sk);
311 
312 	return READ_ONCE(msk->first_pending);
313 }
314 
315 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
316 {
317 	struct mptcp_sock *msk = mptcp_sk(sk);
318 	struct mptcp_data_frag *cur;
319 
320 	cur = msk->first_pending;
321 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
322 						     list_next_entry(cur, list);
323 }
324 
325 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
326 {
327 	struct mptcp_sock *msk = mptcp_sk(sk);
328 
329 	if (!msk->first_pending)
330 		return NULL;
331 
332 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
333 		return NULL;
334 
335 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
336 }
337 
338 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
339 {
340 	struct mptcp_sock *msk = mptcp_sk(sk);
341 
342 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
343 		return NULL;
344 
345 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
346 }
347 
348 struct mptcp_subflow_request_sock {
349 	struct	tcp_request_sock sk;
350 	u16	mp_capable : 1,
351 		mp_join : 1,
352 		backup : 1;
353 	u8	local_id;
354 	u8	remote_id;
355 	u64	local_key;
356 	u64	idsn;
357 	u32	token;
358 	u32	ssn_offset;
359 	u64	thmac;
360 	u32	local_nonce;
361 	u32	remote_nonce;
362 	struct mptcp_sock	*msk;
363 	struct hlist_nulls_node token_node;
364 };
365 
366 static inline struct mptcp_subflow_request_sock *
367 mptcp_subflow_rsk(const struct request_sock *rsk)
368 {
369 	return (struct mptcp_subflow_request_sock *)rsk;
370 }
371 
372 enum mptcp_data_avail {
373 	MPTCP_SUBFLOW_NODATA,
374 	MPTCP_SUBFLOW_DATA_AVAIL,
375 	MPTCP_SUBFLOW_OOO_DATA
376 };
377 
378 struct mptcp_delegated_action {
379 	struct napi_struct napi;
380 	struct list_head head;
381 };
382 
383 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
384 
385 #define MPTCP_DELEGATE_SEND		0
386 
387 /* MPTCP subflow context */
388 struct mptcp_subflow_context {
389 	struct	list_head node;/* conn_list of subflows */
390 	u64	local_key;
391 	u64	remote_key;
392 	u64	idsn;
393 	u64	map_seq;
394 	u32	snd_isn;
395 	u32	token;
396 	u32	rel_write_seq;
397 	u32	map_subflow_seq;
398 	u32	ssn_offset;
399 	u32	map_data_len;
400 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
401 		request_join : 1,   /* send MP_JOIN */
402 		request_bkup : 1,
403 		mp_capable : 1,	    /* remote is MPTCP capable */
404 		mp_join : 1,	    /* remote is JOINing */
405 		fully_established : 1,	    /* path validated */
406 		pm_notified : 1,    /* PM hook called for established status */
407 		conn_finished : 1,
408 		map_valid : 1,
409 		mpc_map : 1,
410 		backup : 1,
411 		send_mp_prio : 1,
412 		rx_eof : 1,
413 		can_ack : 1,        /* only after processing the remote a key */
414 		disposable : 1;	    /* ctx can be free at ulp release time */
415 	enum mptcp_data_avail data_avail;
416 	u32	remote_nonce;
417 	u64	thmac;
418 	u32	local_nonce;
419 	u32	remote_token;
420 	u8	hmac[MPTCPOPT_HMAC_LEN];
421 	u8	local_id;
422 	u8	remote_id;
423 
424 	long	delegated_status;
425 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
426 
427 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
428 	struct	sock *conn;	    /* parent mptcp_sock */
429 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
430 	void	(*tcp_data_ready)(struct sock *sk);
431 	void	(*tcp_state_change)(struct sock *sk);
432 	void	(*tcp_write_space)(struct sock *sk);
433 	void	(*tcp_error_report)(struct sock *sk);
434 
435 	struct	rcu_head rcu;
436 };
437 
438 static inline struct mptcp_subflow_context *
439 mptcp_subflow_ctx(const struct sock *sk)
440 {
441 	struct inet_connection_sock *icsk = inet_csk(sk);
442 
443 	/* Use RCU on icsk_ulp_data only for sock diag code */
444 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
445 }
446 
447 static inline struct sock *
448 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
449 {
450 	return subflow->tcp_sock;
451 }
452 
453 static inline u64
454 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
455 {
456 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
457 		      subflow->ssn_offset -
458 		      subflow->map_subflow_seq;
459 }
460 
461 static inline u64
462 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
463 {
464 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
465 }
466 
467 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
468 					     struct mptcp_subflow_context *subflow)
469 {
470 	sock_hold(mptcp_subflow_tcp_sock(subflow));
471 	spin_lock_bh(&msk->join_list_lock);
472 	list_add_tail(&subflow->node, &msk->join_list);
473 	spin_unlock_bh(&msk->join_list_lock);
474 }
475 
476 void mptcp_subflow_process_delegated(struct sock *ssk);
477 
478 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
479 {
480 	struct mptcp_delegated_action *delegated;
481 	bool schedule;
482 
483 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
484 	 * ensures the below list check sees list updates done prior to status
485 	 * bit changes
486 	 */
487 	if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
488 		/* still on delegated list from previous scheduling */
489 		if (!list_empty(&subflow->delegated_node))
490 			return;
491 
492 		/* the caller held the subflow bh socket lock */
493 		lockdep_assert_in_softirq();
494 
495 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
496 		schedule = list_empty(&delegated->head);
497 		list_add_tail(&subflow->delegated_node, &delegated->head);
498 		sock_hold(mptcp_subflow_tcp_sock(subflow));
499 		if (schedule)
500 			napi_schedule(&delegated->napi);
501 	}
502 }
503 
504 static inline struct mptcp_subflow_context *
505 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
506 {
507 	struct mptcp_subflow_context *ret;
508 
509 	if (list_empty(&delegated->head))
510 		return NULL;
511 
512 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
513 	list_del_init(&ret->delegated_node);
514 	return ret;
515 }
516 
517 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
518 {
519 	return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
520 }
521 
522 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
523 {
524 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
525 	 * touching the status bit
526 	 */
527 	smp_wmb();
528 	clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
529 }
530 
531 int mptcp_is_enabled(struct net *net);
532 unsigned int mptcp_get_add_addr_timeout(struct net *net);
533 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
534 				     struct mptcp_options_received *mp_opt);
535 bool mptcp_subflow_data_available(struct sock *sk);
536 void __init mptcp_subflow_init(void);
537 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
538 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
539 		     struct mptcp_subflow_context *subflow);
540 void mptcp_subflow_reset(struct sock *ssk);
541 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
542 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
543 
544 /* called with sk socket lock held */
545 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
546 			    const struct mptcp_addr_info *remote);
547 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
548 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
549 			 struct sockaddr_storage *addr,
550 			 unsigned short family);
551 
552 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
553 					      struct mptcp_subflow_context *ctx)
554 {
555 	sk->sk_data_ready = ctx->tcp_data_ready;
556 	sk->sk_state_change = ctx->tcp_state_change;
557 	sk->sk_write_space = ctx->tcp_write_space;
558 	sk->sk_error_report = ctx->tcp_error_report;
559 
560 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
561 }
562 
563 void __init mptcp_proto_init(void);
564 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
565 int __init mptcp_proto_v6_init(void);
566 #endif
567 
568 struct sock *mptcp_sk_clone(const struct sock *sk,
569 			    const struct mptcp_options_received *mp_opt,
570 			    struct request_sock *req);
571 void mptcp_get_options(const struct sk_buff *skb,
572 		       struct mptcp_options_received *mp_opt);
573 
574 void mptcp_finish_connect(struct sock *sk);
575 static inline bool mptcp_is_fully_established(struct sock *sk)
576 {
577 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
578 	       READ_ONCE(mptcp_sk(sk)->fully_established);
579 }
580 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
581 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
582 bool mptcp_finish_join(struct sock *sk);
583 bool mptcp_schedule_work(struct sock *sk);
584 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
585 void __mptcp_data_acked(struct sock *sk);
586 void __mptcp_error_report(struct sock *sk);
587 void mptcp_subflow_eof(struct sock *sk);
588 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
589 void __mptcp_flush_join_list(struct mptcp_sock *msk);
590 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
591 {
592 	return READ_ONCE(msk->snd_data_fin_enable) &&
593 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
594 }
595 
596 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
597 {
598 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
599 		return false;
600 
601 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
602 	return true;
603 }
604 
605 static inline void mptcp_write_space(struct sock *sk)
606 {
607 	if (sk_stream_is_writeable(sk)) {
608 		/* pairs with memory barrier in mptcp_poll */
609 		smp_mb();
610 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
611 			sk_stream_write_space(sk);
612 	}
613 }
614 
615 void mptcp_destroy_common(struct mptcp_sock *msk);
616 
617 void __init mptcp_token_init(void);
618 static inline void mptcp_token_init_request(struct request_sock *req)
619 {
620 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
621 }
622 
623 int mptcp_token_new_request(struct request_sock *req);
624 void mptcp_token_destroy_request(struct request_sock *req);
625 int mptcp_token_new_connect(struct sock *sk);
626 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
627 			struct mptcp_sock *msk);
628 bool mptcp_token_exists(u32 token);
629 struct mptcp_sock *mptcp_token_get_sock(u32 token);
630 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
631 					 long *s_num);
632 void mptcp_token_destroy(struct mptcp_sock *msk);
633 
634 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
635 
636 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
637 
638 void __init mptcp_pm_init(void);
639 void mptcp_pm_data_init(struct mptcp_sock *msk);
640 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
641 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
642 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
643 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
644 void mptcp_pm_subflow_established(struct mptcp_sock *msk,
645 				  struct mptcp_subflow_context *subflow);
646 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
647 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
648 				const struct mptcp_addr_info *addr);
649 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
650 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, u8 rm_id);
651 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
652 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
653 				 struct mptcp_addr_info *addr,
654 				 u8 bkup);
655 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
656 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
657 struct mptcp_pm_add_entry *
658 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
659 		       struct mptcp_addr_info *addr);
660 
661 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
662 			   const struct mptcp_addr_info *addr,
663 			   bool echo, bool port);
664 int mptcp_pm_remove_addr(struct mptcp_sock *msk, u8 local_id);
665 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, u8 local_id);
666 
667 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
668 		 const struct sock *ssk, gfp_t gfp);
669 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
670 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
671 
672 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
673 {
674 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
675 }
676 
677 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
678 {
679 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
680 }
681 
682 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk)
683 {
684 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6);
685 }
686 
687 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk)
688 {
689 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT);
690 }
691 
692 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
693 {
694 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
695 }
696 
697 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
698 {
699 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
700 
701 	if (family == AF_INET6)
702 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
703 	if (!echo)
704 		len += MPTCPOPT_THMAC_LEN;
705 	/* account for 2 trailing 'nop' options */
706 	if (port)
707 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
708 
709 	return len;
710 }
711 
712 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
713 			      struct mptcp_addr_info *saddr, bool *echo, bool *port);
714 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
715 			     u8 *rm_id);
716 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
717 
718 void __init mptcp_pm_nl_init(void);
719 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
720 void mptcp_pm_nl_work(struct mptcp_sock *msk);
721 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, u8 rm_id);
722 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
723 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
724 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
725 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
726 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
727 
728 static inline struct mptcp_ext *mptcp_get_ext(struct sk_buff *skb)
729 {
730 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
731 }
732 
733 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
734 
735 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
736 {
737 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
738 }
739 
740 static inline bool mptcp_check_fallback(const struct sock *sk)
741 {
742 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
743 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
744 
745 	return __mptcp_check_fallback(msk);
746 }
747 
748 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
749 {
750 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
751 		pr_debug("TCP fallback already done (msk=%p)", msk);
752 		return;
753 	}
754 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
755 }
756 
757 static inline void mptcp_do_fallback(struct sock *sk)
758 {
759 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
760 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
761 
762 	__mptcp_do_fallback(msk);
763 }
764 
765 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
766 
767 static inline bool subflow_simultaneous_connect(struct sock *sk)
768 {
769 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
770 	struct sock *parent = subflow->conn;
771 
772 	return sk->sk_state == TCP_ESTABLISHED &&
773 	       !mptcp_sk(parent)->pm.server_side &&
774 	       !subflow->conn_finished;
775 }
776 
777 #ifdef CONFIG_SYN_COOKIES
778 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
779 				       struct sk_buff *skb);
780 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
781 					struct sk_buff *skb);
782 void __init mptcp_join_cookie_init(void);
783 #else
784 static inline void
785 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
786 				  struct sk_buff *skb) {}
787 static inline bool
788 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
789 				   struct sk_buff *skb)
790 {
791 	return false;
792 }
793 
794 static inline void mptcp_join_cookie_init(void) {}
795 #endif
796 
797 #endif /* __MPTCP_PROTOCOL_H */
798