xref: /openbmc/linux/net/mptcp/protocol.h (revision 34fa67e7)
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 atomic flags */
114 #define MPTCP_NOSPACE		1
115 #define MPTCP_WORK_RTX		2
116 #define MPTCP_WORK_EOF		3
117 #define MPTCP_FALLBACK_DONE	4
118 #define MPTCP_WORK_CLOSE_SUBFLOW 5
119 
120 /* MPTCP socket release cb flags */
121 #define MPTCP_PUSH_PENDING	1
122 #define MPTCP_CLEAN_UNA		2
123 #define MPTCP_ERROR_REPORT	3
124 #define MPTCP_RETRANSMIT	4
125 #define MPTCP_FLUSH_JOIN_LIST	5
126 #define MPTCP_CONNECTED		6
127 
128 static inline bool before64(__u64 seq1, __u64 seq2)
129 {
130 	return (__s64)(seq1 - seq2) < 0;
131 }
132 
133 #define after64(seq2, seq1)	before64(seq1, seq2)
134 
135 struct mptcp_options_received {
136 	u64	sndr_key;
137 	u64	rcvr_key;
138 	u64	data_ack;
139 	u64	data_seq;
140 	u32	subflow_seq;
141 	u16	data_len;
142 	__sum16	csum;
143 	u16	suboptions;
144 	u32	token;
145 	u32	nonce;
146 	u16	use_map:1,
147 		dsn64:1,
148 		data_fin:1,
149 		use_ack:1,
150 		ack64:1,
151 		mpc_map:1,
152 		reset_reason:4,
153 		reset_transient:1,
154 		echo:1,
155 		backup:1,
156 		deny_join_id0:1,
157 		__unused:2;
158 	u8	join_id;
159 	u64	thmac;
160 	u8	hmac[MPTCPOPT_HMAC_LEN];
161 	struct mptcp_addr_info addr;
162 	struct mptcp_rm_list rm_list;
163 	u64	ahmac;
164 	u64	fail_seq;
165 };
166 
167 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
168 {
169 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
170 		     ((nib & 0xF) << 8) | field);
171 }
172 
173 enum mptcp_pm_status {
174 	MPTCP_PM_ADD_ADDR_RECEIVED,
175 	MPTCP_PM_ADD_ADDR_SEND_ACK,
176 	MPTCP_PM_RM_ADDR_RECEIVED,
177 	MPTCP_PM_ESTABLISHED,
178 	MPTCP_PM_SUBFLOW_ESTABLISHED,
179 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
180 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
181 					 * accounted int id_avail_bitmap
182 					 */
183 };
184 
185 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
186 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
187 
188 enum mptcp_addr_signal_status {
189 	MPTCP_ADD_ADDR_SIGNAL,
190 	MPTCP_ADD_ADDR_ECHO,
191 	MPTCP_RM_ADDR_SIGNAL,
192 };
193 
194 /* max value of mptcp_addr_info.id */
195 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
196 
197 struct mptcp_pm_data {
198 	struct mptcp_addr_info local;
199 	struct mptcp_addr_info remote;
200 	struct list_head anno_list;
201 
202 	spinlock_t	lock;		/*protects the whole PM data */
203 
204 	u8		addr_signal;
205 	bool		server_side;
206 	bool		work_pending;
207 	bool		accept_addr;
208 	bool		accept_subflow;
209 	bool		remote_deny_join_id0;
210 	u8		add_addr_signaled;
211 	u8		add_addr_accepted;
212 	u8		local_addr_used;
213 	u8		subflows;
214 	u8		status;
215 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
216 	struct mptcp_rm_list rm_list_tx;
217 	struct mptcp_rm_list rm_list_rx;
218 };
219 
220 struct mptcp_data_frag {
221 	struct list_head list;
222 	u64 data_seq;
223 	u16 data_len;
224 	u16 offset;
225 	u16 overhead;
226 	u16 already_sent;
227 	struct page *page;
228 };
229 
230 /* MPTCP connection sock */
231 struct mptcp_sock {
232 	/* inet_connection_sock must be the first member */
233 	struct inet_connection_sock sk;
234 	u64		local_key;
235 	u64		remote_key;
236 	u64		write_seq;
237 	u64		snd_nxt;
238 	u64		ack_seq;
239 	u64		rcv_wnd_sent;
240 	u64		rcv_data_fin_seq;
241 	int		rmem_fwd_alloc;
242 	struct sock	*last_snd;
243 	int		snd_burst;
244 	int		old_wspace;
245 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
246 						 * recovery related fields are under data_lock
247 						 * protection
248 						 */
249 	u64		snd_una;
250 	u64		wnd_end;
251 	unsigned long	timer_ival;
252 	u32		token;
253 	int		rmem_released;
254 	unsigned long	flags;
255 	unsigned long	cb_flags;
256 	unsigned long	push_pending;
257 	bool		recovery;		/* closing subflow write queue reinjected */
258 	bool		can_ack;
259 	bool		fully_established;
260 	bool		rcv_data_fin;
261 	bool		snd_data_fin_enable;
262 	bool		rcv_fastclose;
263 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
264 	bool		csum_enabled;
265 	u8		recvmsg_inq:1,
266 			cork:1,
267 			nodelay:1;
268 	struct work_struct work;
269 	struct sk_buff  *ooo_last_skb;
270 	struct rb_root  out_of_order_queue;
271 	struct sk_buff_head receive_queue;
272 	struct list_head conn_list;
273 	struct list_head rtx_queue;
274 	struct mptcp_data_frag *first_pending;
275 	struct list_head join_list;
276 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
277 	struct sock	*first;
278 	struct mptcp_pm_data	pm;
279 	struct {
280 		u32	space;	/* bytes copied in last measurement window */
281 		u32	copied; /* bytes copied in this measurement window */
282 		u64	time;	/* start time of measurement window */
283 		u64	rtt_us; /* last maximum rtt of subflows */
284 	} rcvq_space;
285 
286 	u32 setsockopt_seq;
287 	char		ca_name[TCP_CA_NAME_MAX];
288 };
289 
290 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
291 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
292 
293 #define mptcp_for_each_subflow(__msk, __subflow)			\
294 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
295 
296 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
297 {
298 	sock_owned_by_me((const struct sock *)msk);
299 }
300 
301 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
302 {
303 	return (struct mptcp_sock *)sk;
304 }
305 
306 /* the msk socket don't use the backlog, also account for the bulk
307  * free memory
308  */
309 static inline int __mptcp_rmem(const struct sock *sk)
310 {
311 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
312 }
313 
314 static inline int __mptcp_space(const struct sock *sk)
315 {
316 	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
317 }
318 
319 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
320 {
321 	const struct mptcp_sock *msk = mptcp_sk(sk);
322 
323 	return READ_ONCE(msk->first_pending);
324 }
325 
326 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
327 {
328 	struct mptcp_sock *msk = mptcp_sk(sk);
329 	struct mptcp_data_frag *cur;
330 
331 	cur = msk->first_pending;
332 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
333 						     list_next_entry(cur, list);
334 }
335 
336 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
337 {
338 	struct mptcp_sock *msk = mptcp_sk(sk);
339 
340 	if (!msk->first_pending)
341 		return NULL;
342 
343 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
344 		return NULL;
345 
346 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
347 }
348 
349 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
350 {
351 	struct mptcp_sock *msk = mptcp_sk(sk);
352 
353 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
354 		return NULL;
355 
356 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
357 }
358 
359 struct csum_pseudo_header {
360 	__be64 data_seq;
361 	__be32 subflow_seq;
362 	__be16 data_len;
363 	__sum16 csum;
364 };
365 
366 struct mptcp_subflow_request_sock {
367 	struct	tcp_request_sock sk;
368 	u16	mp_capable : 1,
369 		mp_join : 1,
370 		backup : 1,
371 		csum_reqd : 1,
372 		allow_join_id0 : 1;
373 	u8	local_id;
374 	u8	remote_id;
375 	u64	local_key;
376 	u64	idsn;
377 	u32	token;
378 	u32	ssn_offset;
379 	u64	thmac;
380 	u32	local_nonce;
381 	u32	remote_nonce;
382 	struct mptcp_sock	*msk;
383 	struct hlist_nulls_node token_node;
384 };
385 
386 static inline struct mptcp_subflow_request_sock *
387 mptcp_subflow_rsk(const struct request_sock *rsk)
388 {
389 	return (struct mptcp_subflow_request_sock *)rsk;
390 }
391 
392 enum mptcp_data_avail {
393 	MPTCP_SUBFLOW_NODATA,
394 	MPTCP_SUBFLOW_DATA_AVAIL,
395 };
396 
397 struct mptcp_delegated_action {
398 	struct napi_struct napi;
399 	struct list_head head;
400 };
401 
402 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
403 
404 #define MPTCP_DELEGATE_SEND		0
405 #define MPTCP_DELEGATE_ACK		1
406 
407 /* MPTCP subflow context */
408 struct mptcp_subflow_context {
409 	struct	list_head node;/* conn_list of subflows */
410 
411 	char	reset_start[0];
412 
413 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
414 	u64	local_key;
415 	u64	remote_key;
416 	u64	idsn;
417 	u64	map_seq;
418 	u32	snd_isn;
419 	u32	token;
420 	u32	rel_write_seq;
421 	u32	map_subflow_seq;
422 	u32	ssn_offset;
423 	u32	map_data_len;
424 	__wsum	map_data_csum;
425 	u32	map_csum_len;
426 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
427 		request_join : 1,   /* send MP_JOIN */
428 		request_bkup : 1,
429 		mp_capable : 1,	    /* remote is MPTCP capable */
430 		mp_join : 1,	    /* remote is JOINing */
431 		fully_established : 1,	    /* path validated */
432 		pm_notified : 1,    /* PM hook called for established status */
433 		conn_finished : 1,
434 		map_valid : 1,
435 		map_csum_reqd : 1,
436 		map_data_fin : 1,
437 		mpc_map : 1,
438 		backup : 1,
439 		send_mp_prio : 1,
440 		send_mp_fail : 1,
441 		send_fastclose : 1,
442 		rx_eof : 1,
443 		can_ack : 1,        /* only after processing the remote a key */
444 		disposable : 1,	    /* ctx can be free at ulp release time */
445 		stale : 1;	    /* unable to snd/rcv data, do not use for xmit */
446 	enum mptcp_data_avail data_avail;
447 	u32	remote_nonce;
448 	u64	thmac;
449 	u32	local_nonce;
450 	u32	remote_token;
451 	u8	hmac[MPTCPOPT_HMAC_LEN];
452 	u8	local_id;
453 	u8	remote_id;
454 	u8	reset_seen:1;
455 	u8	reset_transient:1;
456 	u8	reset_reason:4;
457 	u8	stale_count;
458 
459 	long	delegated_status;
460 
461 	char	reset_end[0];
462 
463 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
464 
465 	u32	setsockopt_seq;
466 	u32	stale_rcv_tstamp;
467 
468 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
469 	struct	sock *conn;	    /* parent mptcp_sock */
470 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
471 	void	(*tcp_data_ready)(struct sock *sk);
472 	void	(*tcp_state_change)(struct sock *sk);
473 	void	(*tcp_write_space)(struct sock *sk);
474 	void	(*tcp_error_report)(struct sock *sk);
475 
476 	struct	rcu_head rcu;
477 };
478 
479 static inline struct mptcp_subflow_context *
480 mptcp_subflow_ctx(const struct sock *sk)
481 {
482 	struct inet_connection_sock *icsk = inet_csk(sk);
483 
484 	/* Use RCU on icsk_ulp_data only for sock diag code */
485 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
486 }
487 
488 static inline struct sock *
489 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
490 {
491 	return subflow->tcp_sock;
492 }
493 
494 static inline void
495 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
496 {
497 	memset(subflow->reset_start, 0, subflow->reset_end - subflow->reset_start);
498 	subflow->request_mptcp = 1;
499 }
500 
501 static inline u64
502 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
503 {
504 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
505 		      subflow->ssn_offset -
506 		      subflow->map_subflow_seq;
507 }
508 
509 static inline u64
510 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
511 {
512 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
513 }
514 
515 void mptcp_subflow_process_delegated(struct sock *ssk);
516 
517 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
518 {
519 	struct mptcp_delegated_action *delegated;
520 	bool schedule;
521 
522 	/* the caller held the subflow bh socket lock */
523 	lockdep_assert_in_softirq();
524 
525 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
526 	 * ensures the below list check sees list updates done prior to status
527 	 * bit changes
528 	 */
529 	if (!test_and_set_bit(action, &subflow->delegated_status)) {
530 		/* still on delegated list from previous scheduling */
531 		if (!list_empty(&subflow->delegated_node))
532 			return;
533 
534 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
535 		schedule = list_empty(&delegated->head);
536 		list_add_tail(&subflow->delegated_node, &delegated->head);
537 		sock_hold(mptcp_subflow_tcp_sock(subflow));
538 		if (schedule)
539 			napi_schedule(&delegated->napi);
540 	}
541 }
542 
543 static inline struct mptcp_subflow_context *
544 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
545 {
546 	struct mptcp_subflow_context *ret;
547 
548 	if (list_empty(&delegated->head))
549 		return NULL;
550 
551 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
552 	list_del_init(&ret->delegated_node);
553 	return ret;
554 }
555 
556 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
557 {
558 	return !!READ_ONCE(subflow->delegated_status);
559 }
560 
561 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
562 {
563 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
564 	 * touching the status bit
565 	 */
566 	smp_wmb();
567 	clear_bit(action, &subflow->delegated_status);
568 }
569 
570 int mptcp_is_enabled(const struct net *net);
571 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
572 int mptcp_is_checksum_enabled(const struct net *net);
573 int mptcp_allow_join_id0(const struct net *net);
574 unsigned int mptcp_stale_loss_cnt(const struct net *net);
575 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
576 				     struct mptcp_options_received *mp_opt);
577 bool __mptcp_retransmit_pending_data(struct sock *sk);
578 void mptcp_check_and_set_pending(struct sock *sk);
579 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
580 bool mptcp_subflow_data_available(struct sock *sk);
581 void __init mptcp_subflow_init(void);
582 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
583 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
584 		     struct mptcp_subflow_context *subflow);
585 void mptcp_subflow_send_ack(struct sock *ssk);
586 void mptcp_subflow_reset(struct sock *ssk);
587 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
588 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
589 
590 /* called with sk socket lock held */
591 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
592 			    const struct mptcp_addr_info *remote);
593 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
594 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
595 			 struct sockaddr_storage *addr,
596 			 unsigned short family);
597 
598 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
599 {
600 	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
601 
602 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
603 	if (subflow->request_join && !subflow->fully_established)
604 		return false;
605 
606 	/* only send if our side has not closed yet */
607 	return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
608 }
609 
610 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
611 
612 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
613 
614 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
615 					      struct mptcp_subflow_context *ctx)
616 {
617 	sk->sk_data_ready = ctx->tcp_data_ready;
618 	sk->sk_state_change = ctx->tcp_state_change;
619 	sk->sk_write_space = ctx->tcp_write_space;
620 	sk->sk_error_report = ctx->tcp_error_report;
621 
622 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
623 }
624 
625 static inline bool mptcp_has_another_subflow(struct sock *ssk)
626 {
627 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp;
628 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
629 
630 	mptcp_for_each_subflow(msk, tmp) {
631 		if (tmp != subflow)
632 			return true;
633 	}
634 
635 	return false;
636 }
637 
638 void __init mptcp_proto_init(void);
639 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
640 int __init mptcp_proto_v6_init(void);
641 #endif
642 
643 struct sock *mptcp_sk_clone(const struct sock *sk,
644 			    const struct mptcp_options_received *mp_opt,
645 			    struct request_sock *req);
646 void mptcp_get_options(const struct sock *sk,
647 		       const struct sk_buff *skb,
648 		       struct mptcp_options_received *mp_opt);
649 
650 void mptcp_finish_connect(struct sock *sk);
651 void __mptcp_set_connected(struct sock *sk);
652 static inline bool mptcp_is_fully_established(struct sock *sk)
653 {
654 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
655 	       READ_ONCE(mptcp_sk(sk)->fully_established);
656 }
657 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
658 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
659 bool mptcp_finish_join(struct sock *sk);
660 bool mptcp_schedule_work(struct sock *sk);
661 int mptcp_setsockopt(struct sock *sk, int level, int optname,
662 		     sockptr_t optval, unsigned int optlen);
663 int mptcp_getsockopt(struct sock *sk, int level, int optname,
664 		     char __user *optval, int __user *option);
665 
666 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
667 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
668 {
669 	if (use_64bit)
670 		return cur_seq;
671 
672 	return __mptcp_expand_seq(old_seq, cur_seq);
673 }
674 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
675 void __mptcp_data_acked(struct sock *sk);
676 void __mptcp_error_report(struct sock *sk);
677 void mptcp_subflow_eof(struct sock *sk);
678 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
679 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
680 {
681 	return READ_ONCE(msk->snd_data_fin_enable) &&
682 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
683 }
684 
685 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
686 {
687 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
688 		return false;
689 
690 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
691 	return true;
692 }
693 
694 static inline void mptcp_write_space(struct sock *sk)
695 {
696 	if (sk_stream_is_writeable(sk)) {
697 		/* pairs with memory barrier in mptcp_poll */
698 		smp_mb();
699 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
700 			sk_stream_write_space(sk);
701 	}
702 }
703 
704 void mptcp_destroy_common(struct mptcp_sock *msk);
705 
706 #define MPTCP_TOKEN_MAX_RETRIES	4
707 
708 void __init mptcp_token_init(void);
709 static inline void mptcp_token_init_request(struct request_sock *req)
710 {
711 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
712 }
713 
714 int mptcp_token_new_request(struct request_sock *req);
715 void mptcp_token_destroy_request(struct request_sock *req);
716 int mptcp_token_new_connect(struct sock *sk);
717 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
718 			struct mptcp_sock *msk);
719 bool mptcp_token_exists(u32 token);
720 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
721 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
722 					 long *s_num);
723 void mptcp_token_destroy(struct mptcp_sock *msk);
724 
725 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
726 
727 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
728 u16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
729 
730 void __init mptcp_pm_init(void);
731 void mptcp_pm_data_init(struct mptcp_sock *msk);
732 void mptcp_pm_data_reset(struct mptcp_sock *msk);
733 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
734 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
735 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
736 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
737 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
738 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
739 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
740 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
741 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
742 				 const struct mptcp_subflow_context *subflow);
743 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
744 				const struct mptcp_addr_info *addr);
745 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
746 			      struct mptcp_addr_info *addr);
747 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
748 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
749 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
750 			       const struct mptcp_rm_list *rm_list);
751 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
752 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
753 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
754 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
755 struct mptcp_pm_add_entry *
756 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
757 		       struct mptcp_addr_info *addr, bool check_id);
758 struct mptcp_pm_add_entry *
759 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
760 				struct mptcp_addr_info *addr);
761 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id,
762 					 u8 *flags, int *ifindex);
763 
764 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
765 			   const struct mptcp_addr_info *addr,
766 			   bool echo);
767 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
768 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
769 
770 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
771 		 const struct sock *ssk, gfp_t gfp);
772 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
773 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
774 
775 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
776 {
777 	return READ_ONCE(msk->pm.addr_signal) &
778 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
779 }
780 
781 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
782 {
783 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
784 }
785 
786 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
787 {
788 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
789 }
790 
791 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
792 {
793 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
794 }
795 
796 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
797 {
798 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
799 
800 	if (family == AF_INET6)
801 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
802 	if (!echo)
803 		len += MPTCPOPT_THMAC_LEN;
804 	/* account for 2 trailing 'nop' options */
805 	if (port)
806 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
807 
808 	return len;
809 }
810 
811 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
812 {
813 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
814 		return -EINVAL;
815 
816 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
817 }
818 
819 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
820 			      unsigned int opt_size, unsigned int remaining,
821 			      struct mptcp_addr_info *addr, bool *echo,
822 			      bool *port, bool *drop_other_suboptions);
823 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
824 			     struct mptcp_rm_list *rm_list);
825 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
826 
827 void __init mptcp_pm_nl_init(void);
828 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
829 void mptcp_pm_nl_work(struct mptcp_sock *msk);
830 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
831 				     const struct mptcp_rm_list *rm_list);
832 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
833 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
834 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
835 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
836 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
837 
838 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
839 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
840 
841 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
842 {
843 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
844 }
845 
846 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
847 
848 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
849 {
850 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
851 }
852 
853 static inline bool mptcp_check_fallback(const struct sock *sk)
854 {
855 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
856 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
857 
858 	return __mptcp_check_fallback(msk);
859 }
860 
861 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
862 {
863 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
864 		pr_debug("TCP fallback already done (msk=%p)", msk);
865 		return;
866 	}
867 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
868 }
869 
870 static inline void mptcp_do_fallback(struct sock *sk)
871 {
872 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
873 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
874 
875 	__mptcp_do_fallback(msk);
876 }
877 
878 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
879 
880 static inline bool subflow_simultaneous_connect(struct sock *sk)
881 {
882 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
883 	struct sock *parent = subflow->conn;
884 
885 	return sk->sk_state == TCP_ESTABLISHED &&
886 	       !mptcp_sk(parent)->pm.server_side &&
887 	       !subflow->conn_finished;
888 }
889 
890 #ifdef CONFIG_SYN_COOKIES
891 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
892 				       struct sk_buff *skb);
893 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
894 					struct sk_buff *skb);
895 void __init mptcp_join_cookie_init(void);
896 #else
897 static inline void
898 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
899 				  struct sk_buff *skb) {}
900 static inline bool
901 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
902 				   struct sk_buff *skb)
903 {
904 	return false;
905 }
906 
907 static inline void mptcp_join_cookie_init(void) {}
908 #endif
909 
910 #endif /* __MPTCP_PROTOCOL_H */
911