xref: /openbmc/linux/net/mptcp/protocol.h (revision 6b6c2ebd83f2bf97e8f221479372aaca97a4a9b2)
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 #include <net/genetlink.h>
15 
16 #define MPTCP_SUPPORTED_VERSION	1
17 
18 /* MPTCP option bits */
19 #define OPTION_MPTCP_MPC_SYN	BIT(0)
20 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
21 #define OPTION_MPTCP_MPC_ACK	BIT(2)
22 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
23 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
24 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
25 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
26 #define OPTION_MPTCP_RM_ADDR	BIT(7)
27 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
28 #define OPTION_MPTCP_PRIO	BIT(9)
29 #define OPTION_MPTCP_RST	BIT(10)
30 #define OPTION_MPTCP_DSS	BIT(11)
31 #define OPTION_MPTCP_FAIL	BIT(12)
32 
33 #define OPTION_MPTCP_CSUMREQD	BIT(13)
34 
35 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
36 				 OPTION_MPTCP_MPC_ACK)
37 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
38 				 OPTION_MPTCP_MPJ_ACK)
39 
40 /* MPTCP option subtypes */
41 #define MPTCPOPT_MP_CAPABLE	0
42 #define MPTCPOPT_MP_JOIN	1
43 #define MPTCPOPT_DSS		2
44 #define MPTCPOPT_ADD_ADDR	3
45 #define MPTCPOPT_RM_ADDR	4
46 #define MPTCPOPT_MP_PRIO	5
47 #define MPTCPOPT_MP_FAIL	6
48 #define MPTCPOPT_MP_FASTCLOSE	7
49 #define MPTCPOPT_RST		8
50 
51 /* MPTCP suboption lengths */
52 #define TCPOLEN_MPTCP_MPC_SYN		4
53 #define TCPOLEN_MPTCP_MPC_SYNACK	12
54 #define TCPOLEN_MPTCP_MPC_ACK		20
55 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
56 #define TCPOLEN_MPTCP_MPJ_SYN		12
57 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
58 #define TCPOLEN_MPTCP_MPJ_ACK		24
59 #define TCPOLEN_MPTCP_DSS_BASE		4
60 #define TCPOLEN_MPTCP_DSS_ACK32		4
61 #define TCPOLEN_MPTCP_DSS_ACK64		8
62 #define TCPOLEN_MPTCP_DSS_MAP32		10
63 #define TCPOLEN_MPTCP_DSS_MAP64		14
64 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
65 #define TCPOLEN_MPTCP_ADD_ADDR		16
66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
69 #define TCPOLEN_MPTCP_ADD_ADDR6		28
70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
73 #define TCPOLEN_MPTCP_PORT_LEN		2
74 #define TCPOLEN_MPTCP_PORT_ALIGN	2
75 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
76 #define TCPOLEN_MPTCP_PRIO		3
77 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
78 #define TCPOLEN_MPTCP_FASTCLOSE		12
79 #define TCPOLEN_MPTCP_RST		4
80 #define TCPOLEN_MPTCP_FAIL		12
81 
82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
83 
84 /* MPTCP MP_JOIN flags */
85 #define MPTCPOPT_BACKUP		BIT(0)
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_FALLBACK_DONE	4
117 #define MPTCP_WORK_CLOSE_SUBFLOW 5
118 
119 /* MPTCP socket release cb flags */
120 #define MPTCP_PUSH_PENDING	1
121 #define MPTCP_CLEAN_UNA		2
122 #define MPTCP_ERROR_REPORT	3
123 #define MPTCP_RETRANSMIT	4
124 #define MPTCP_FLUSH_JOIN_LIST	5
125 #define MPTCP_SYNC_STATE	6
126 #define MPTCP_SYNC_SNDBUF	7
127 
128 struct mptcp_skb_cb {
129 	u64 map_seq;
130 	u64 end_seq;
131 	u32 offset;
132 	u8  has_rxtstamp:1;
133 };
134 
135 #define MPTCP_SKB_CB(__skb)	((struct mptcp_skb_cb *)&((__skb)->cb[0]))
136 
137 static inline bool before64(__u64 seq1, __u64 seq2)
138 {
139 	return (__s64)(seq1 - seq2) < 0;
140 }
141 
142 #define after64(seq2, seq1)	before64(seq1, seq2)
143 
144 struct mptcp_options_received {
145 	u64	sndr_key;
146 	u64	rcvr_key;
147 	u64	data_ack;
148 	u64	data_seq;
149 	u32	subflow_seq;
150 	u16	data_len;
151 	__sum16	csum;
152 	u16	suboptions;
153 	u32	token;
154 	u32	nonce;
155 	u16	use_map:1,
156 		dsn64:1,
157 		data_fin:1,
158 		use_ack:1,
159 		ack64:1,
160 		mpc_map:1,
161 		reset_reason:4,
162 		reset_transient:1,
163 		echo:1,
164 		backup:1,
165 		deny_join_id0:1,
166 		__unused:2;
167 	u8	join_id;
168 	u64	thmac;
169 	u8	hmac[MPTCPOPT_HMAC_LEN];
170 	struct mptcp_addr_info addr;
171 	struct mptcp_rm_list rm_list;
172 	u64	ahmac;
173 	u64	fail_seq;
174 };
175 
176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
177 {
178 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 		     ((nib & 0xF) << 8) | field);
180 }
181 
182 enum mptcp_pm_status {
183 	MPTCP_PM_ADD_ADDR_RECEIVED,
184 	MPTCP_PM_ADD_ADDR_SEND_ACK,
185 	MPTCP_PM_RM_ADDR_RECEIVED,
186 	MPTCP_PM_ESTABLISHED,
187 	MPTCP_PM_SUBFLOW_ESTABLISHED,
188 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
189 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 					 * accounted int id_avail_bitmap
191 					 */
192 };
193 
194 enum mptcp_pm_type {
195 	MPTCP_PM_TYPE_KERNEL = 0,
196 	MPTCP_PM_TYPE_USERSPACE,
197 
198 	__MPTCP_PM_TYPE_NR,
199 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
200 };
201 
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
204 
205 enum mptcp_addr_signal_status {
206 	MPTCP_ADD_ADDR_SIGNAL,
207 	MPTCP_ADD_ADDR_ECHO,
208 	MPTCP_RM_ADDR_SIGNAL,
209 };
210 
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
213 
214 struct mptcp_pm_data {
215 	struct mptcp_addr_info local;
216 	struct mptcp_addr_info remote;
217 	struct list_head anno_list;
218 	struct list_head userspace_pm_local_addr_list;
219 
220 	spinlock_t	lock;		/*protects the whole PM data */
221 
222 	u8		addr_signal;
223 	bool		server_side;
224 	bool		work_pending;
225 	bool		accept_addr;
226 	bool		accept_subflow;
227 	bool		remote_deny_join_id0;
228 	u8		add_addr_signaled;
229 	u8		add_addr_accepted;
230 	u8		local_addr_used;
231 	u8		pm_type;
232 	u8		subflows;
233 	u8		status;
234 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 	struct mptcp_rm_list rm_list_tx;
236 	struct mptcp_rm_list rm_list_rx;
237 };
238 
239 struct mptcp_pm_addr_entry {
240 	struct list_head	list;
241 	struct mptcp_addr_info	addr;
242 	u8			flags;
243 	int			ifindex;
244 	struct socket		*lsk;
245 };
246 
247 struct mptcp_data_frag {
248 	struct list_head list;
249 	u64 data_seq;
250 	u16 data_len;
251 	u16 offset;
252 	u16 overhead;
253 	u16 already_sent;
254 	struct page *page;
255 };
256 
257 /* MPTCP connection sock */
258 struct mptcp_sock {
259 	/* inet_connection_sock must be the first member */
260 	struct inet_connection_sock sk;
261 	u64		local_key;
262 	u64		remote_key;
263 	u64		write_seq;
264 	u64		bytes_sent;
265 	u64		snd_nxt;
266 	u64		bytes_received;
267 	u64		ack_seq;
268 	atomic64_t	rcv_wnd_sent;
269 	u64		rcv_data_fin_seq;
270 	u64		bytes_retrans;
271 	u64		bytes_consumed;
272 	int		rmem_fwd_alloc;
273 	int		snd_burst;
274 	int		old_wspace;
275 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
276 						 * recovery related fields are under data_lock
277 						 * protection
278 						 */
279 	u64		bytes_acked;
280 	u64		snd_una;
281 	u64		wnd_end;
282 	unsigned long	timer_ival;
283 	u32		token;
284 	int		rmem_released;
285 	unsigned long	flags;
286 	unsigned long	cb_flags;
287 	bool		recovery;		/* closing subflow write queue reinjected */
288 	bool		can_ack;
289 	bool		fully_established;
290 	bool		rcv_data_fin;
291 	bool		snd_data_fin_enable;
292 	bool		rcv_fastclose;
293 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
294 	bool		csum_enabled;
295 	bool		allow_infinite_fallback;
296 	u8		pending_state; /* A subflow asked to set this sk_state,
297 					* protected by the msk data lock
298 					*/
299 	u8		mpc_endpoint_id;
300 	u8		recvmsg_inq:1,
301 			cork:1,
302 			nodelay:1,
303 			fastopening:1,
304 			in_accept_queue:1,
305 			free_first:1,
306 			rcvspace_init:1;
307 	int		keepalive_cnt;
308 	int		keepalive_idle;
309 	int		keepalive_intvl;
310 	struct work_struct work;
311 	struct sk_buff  *ooo_last_skb;
312 	struct rb_root  out_of_order_queue;
313 	struct sk_buff_head receive_queue;
314 	struct list_head conn_list;
315 	struct list_head rtx_queue;
316 	struct mptcp_data_frag *first_pending;
317 	struct list_head join_list;
318 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
319 				 * ONCE annotation, the subflow outside the socket
320 				 * lock as such sock is freed after close().
321 				 */
322 	struct mptcp_pm_data	pm;
323 	struct mptcp_sched_ops	*sched;
324 	struct {
325 		u32	space;	/* bytes copied in last measurement window */
326 		u32	copied; /* bytes copied in this measurement window */
327 		u64	time;	/* start time of measurement window */
328 		u64	rtt_us; /* last maximum rtt of subflows */
329 	} rcvq_space;
330 	u8		scaling_ratio;
331 
332 	u32		subflow_id;
333 	u32		setsockopt_seq;
334 	char		ca_name[TCP_CA_NAME_MAX];
335 };
336 
337 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
338 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
339 
340 #define mptcp_for_each_subflow(__msk, __subflow)			\
341 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
342 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
343 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
344 
345 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
346 {
347 	sock_owned_by_me((const struct sock *)msk);
348 }
349 
350 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
351 
352 /* the msk socket don't use the backlog, also account for the bulk
353  * free memory
354  */
355 static inline int __mptcp_rmem(const struct sock *sk)
356 {
357 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
358 }
359 
360 static inline int mptcp_win_from_space(const struct sock *sk, int space)
361 {
362 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
363 }
364 
365 static inline int __mptcp_space(const struct sock *sk)
366 {
367 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
368 }
369 
370 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
371 {
372 	const struct mptcp_sock *msk = mptcp_sk(sk);
373 
374 	return READ_ONCE(msk->first_pending);
375 }
376 
377 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
378 {
379 	struct mptcp_sock *msk = mptcp_sk(sk);
380 	struct mptcp_data_frag *cur;
381 
382 	cur = msk->first_pending;
383 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
384 						     list_next_entry(cur, list);
385 }
386 
387 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
388 {
389 	const struct mptcp_sock *msk = mptcp_sk(sk);
390 
391 	if (!msk->first_pending)
392 		return NULL;
393 
394 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
395 		return NULL;
396 
397 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
398 }
399 
400 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
401 {
402 	struct mptcp_sock *msk = mptcp_sk(sk);
403 
404 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
405 		return NULL;
406 
407 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
408 }
409 
410 struct csum_pseudo_header {
411 	__be64 data_seq;
412 	__be32 subflow_seq;
413 	__be16 data_len;
414 	__sum16 csum;
415 };
416 
417 struct mptcp_subflow_request_sock {
418 	struct	tcp_request_sock sk;
419 	u16	mp_capable : 1,
420 		mp_join : 1,
421 		backup : 1,
422 		csum_reqd : 1,
423 		allow_join_id0 : 1;
424 	u8	local_id;
425 	u8	remote_id;
426 	u64	local_key;
427 	u64	idsn;
428 	u32	token;
429 	u32	ssn_offset;
430 	u64	thmac;
431 	u32	local_nonce;
432 	u32	remote_nonce;
433 	struct mptcp_sock	*msk;
434 	struct hlist_nulls_node token_node;
435 };
436 
437 static inline struct mptcp_subflow_request_sock *
438 mptcp_subflow_rsk(const struct request_sock *rsk)
439 {
440 	return (struct mptcp_subflow_request_sock *)rsk;
441 }
442 
443 enum mptcp_data_avail {
444 	MPTCP_SUBFLOW_NODATA,
445 	MPTCP_SUBFLOW_DATA_AVAIL,
446 };
447 
448 struct mptcp_delegated_action {
449 	struct napi_struct napi;
450 	struct list_head head;
451 };
452 
453 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
454 
455 #define MPTCP_DELEGATE_SCHEDULED	0
456 #define MPTCP_DELEGATE_SEND		1
457 #define MPTCP_DELEGATE_ACK		2
458 #define MPTCP_DELEGATE_SNDBUF		3
459 
460 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
461 /* MPTCP subflow context */
462 struct mptcp_subflow_context {
463 	struct	list_head node;/* conn_list of subflows */
464 
465 	struct_group(reset,
466 
467 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
468 	u64	local_key;
469 	u64	remote_key;
470 	u64	idsn;
471 	u64	map_seq;
472 	u32	snd_isn;
473 	u32	token;
474 	u32	rel_write_seq;
475 	u32	map_subflow_seq;
476 	u32	ssn_offset;
477 	u32	map_data_len;
478 	__wsum	map_data_csum;
479 	u32	map_csum_len;
480 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
481 		request_join : 1,   /* send MP_JOIN */
482 		request_bkup : 1,
483 		mp_capable : 1,	    /* remote is MPTCP capable */
484 		mp_join : 1,	    /* remote is JOINing */
485 		fully_established : 1,	    /* path validated */
486 		pm_notified : 1,    /* PM hook called for established status */
487 		conn_finished : 1,
488 		map_valid : 1,
489 		map_csum_reqd : 1,
490 		map_data_fin : 1,
491 		mpc_map : 1,
492 		backup : 1,
493 		send_mp_prio : 1,
494 		send_mp_fail : 1,
495 		send_fastclose : 1,
496 		send_infinite_map : 1,
497 		remote_key_valid : 1,        /* received the peer key from */
498 		disposable : 1,	    /* ctx can be free at ulp release time */
499 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
500 		valid_csum_seen : 1,        /* at least one csum validated */
501 		is_mptfo : 1,	    /* subflow is doing TFO */
502 		__unused : 10;
503 	enum mptcp_data_avail data_avail;
504 	bool	scheduled;
505 	u32	remote_nonce;
506 	u64	thmac;
507 	u32	local_nonce;
508 	u32	remote_token;
509 	union {
510 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
511 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
512 	};
513 	s16	local_id;	    /* if negative not initialized yet */
514 	u8	remote_id;
515 	u8	reset_seen:1;
516 	u8	reset_transient:1;
517 	u8	reset_reason:4;
518 	u8	stale_count;
519 
520 	u32	subflow_id;
521 
522 	long	delegated_status;
523 	unsigned long	fail_tout;
524 
525 	);
526 
527 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
528 
529 	u32	setsockopt_seq;
530 	u32	stale_rcv_tstamp;
531 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
532 				     * protected by the msk socket lock
533 				     */
534 
535 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
536 	struct	sock *conn;	    /* parent mptcp_sock */
537 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
538 	void	(*tcp_state_change)(struct sock *sk);
539 	void	(*tcp_error_report)(struct sock *sk);
540 
541 	struct	rcu_head rcu;
542 };
543 
544 static inline struct mptcp_subflow_context *
545 mptcp_subflow_ctx(const struct sock *sk)
546 {
547 	struct inet_connection_sock *icsk = inet_csk(sk);
548 
549 	/* Use RCU on icsk_ulp_data only for sock diag code */
550 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
551 }
552 
553 static inline struct sock *
554 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
555 {
556 	return subflow->tcp_sock;
557 }
558 
559 static inline void
560 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
561 {
562 	memset(&subflow->reset, 0, sizeof(subflow->reset));
563 	subflow->request_mptcp = 1;
564 	WRITE_ONCE(subflow->local_id, -1);
565 }
566 
567 static inline u64
568 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
569 {
570 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
571 		      subflow->ssn_offset -
572 		      subflow->map_subflow_seq;
573 }
574 
575 static inline u64
576 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
577 {
578 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
579 }
580 
581 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
582 
583 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
584 {
585 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
586 	struct mptcp_delegated_action *delegated;
587 	bool schedule;
588 
589 	/* the caller held the subflow bh socket lock */
590 	lockdep_assert_in_softirq();
591 
592 	/* The implied barrier pairs with tcp_release_cb_override()
593 	 * mptcp_napi_poll(), and ensures the below list check sees list
594 	 * updates done prior to delegated status bits changes
595 	 */
596 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
597 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
598 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
599 			return;
600 
601 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
602 		schedule = list_empty(&delegated->head);
603 		list_add_tail(&subflow->delegated_node, &delegated->head);
604 		sock_hold(mptcp_subflow_tcp_sock(subflow));
605 		if (schedule)
606 			napi_schedule(&delegated->napi);
607 	}
608 }
609 
610 static inline struct mptcp_subflow_context *
611 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
612 {
613 	struct mptcp_subflow_context *ret;
614 
615 	if (list_empty(&delegated->head))
616 		return NULL;
617 
618 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
619 	list_del_init(&ret->delegated_node);
620 	return ret;
621 }
622 
623 int mptcp_is_enabled(const struct net *net);
624 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
625 int mptcp_is_checksum_enabled(const struct net *net);
626 int mptcp_allow_join_id0(const struct net *net);
627 unsigned int mptcp_stale_loss_cnt(const struct net *net);
628 int mptcp_get_pm_type(const struct net *net);
629 const char *mptcp_get_scheduler(const struct net *net);
630 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
631 				       struct mptcp_subflow_context *subflow,
632 				       const struct mptcp_options_received *mp_opt);
633 bool __mptcp_retransmit_pending_data(struct sock *sk);
634 void mptcp_check_and_set_pending(struct sock *sk);
635 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
636 bool mptcp_subflow_data_available(struct sock *sk);
637 void __init mptcp_subflow_init(void);
638 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
639 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
640 		     struct mptcp_subflow_context *subflow);
641 void __mptcp_subflow_send_ack(struct sock *ssk);
642 void mptcp_subflow_reset(struct sock *ssk);
643 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
644 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
645 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
646 bool __mptcp_close(struct sock *sk, long timeout);
647 void mptcp_cancel_work(struct sock *sk);
648 void __mptcp_unaccepted_force_close(struct sock *sk);
649 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
650 void mptcp_set_state(struct sock *sk, int state);
651 
652 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
653 			   const struct mptcp_addr_info *b, bool use_port);
654 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
655 
656 /* called with sk socket lock held */
657 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
658 			    const struct mptcp_addr_info *remote);
659 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
660 				struct socket **new_sock);
661 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
662 			 struct sockaddr_storage *addr,
663 			 unsigned short family);
664 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
665 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
666 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
667 void mptcp_sched_init(void);
668 int mptcp_init_sched(struct mptcp_sock *msk,
669 		     struct mptcp_sched_ops *sched);
670 void mptcp_release_sched(struct mptcp_sock *msk);
671 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
672 				 bool scheduled);
673 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
674 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
675 int mptcp_sched_get_send(struct mptcp_sock *msk);
676 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
677 
678 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
679 {
680 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
681 }
682 
683 static inline bool mptcp_epollin_ready(const struct sock *sk)
684 {
685 	/* mptcp doesn't have to deal with small skbs in the receive queue,
686 	 * at it can always coalesce them
687 	 */
688 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
689 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
690 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
691 	       READ_ONCE(tcp_memory_pressure);
692 }
693 
694 int mptcp_set_rcvlowat(struct sock *sk, int val);
695 
696 static inline bool __tcp_can_send(const struct sock *ssk)
697 {
698 	/* only send if our side has not closed yet */
699 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
700 }
701 
702 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
703 {
704 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
705 	if (subflow->request_join && !subflow->fully_established)
706 		return false;
707 
708 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
709 }
710 
711 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
712 
713 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
714 
715 void mptcp_subflow_drop_ctx(struct sock *ssk);
716 
717 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
718 					      struct mptcp_subflow_context *ctx)
719 {
720 	sk->sk_data_ready = sock_def_readable;
721 	sk->sk_state_change = ctx->tcp_state_change;
722 	sk->sk_write_space = sk_stream_write_space;
723 	sk->sk_error_report = ctx->tcp_error_report;
724 
725 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
726 }
727 
728 void __init mptcp_proto_init(void);
729 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
730 int __init mptcp_proto_v6_init(void);
731 #endif
732 
733 struct sock *mptcp_sk_clone_init(const struct sock *sk,
734 				 const struct mptcp_options_received *mp_opt,
735 				 struct sock *ssk,
736 				 struct request_sock *req);
737 void mptcp_get_options(const struct sk_buff *skb,
738 		       struct mptcp_options_received *mp_opt);
739 
740 void mptcp_finish_connect(struct sock *sk);
741 void __mptcp_sync_state(struct sock *sk, int state);
742 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
743 
744 static inline void mptcp_stop_tout_timer(struct sock *sk)
745 {
746 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
747 		return;
748 
749 	sk_stop_timer(sk, &sk->sk_timer);
750 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
751 }
752 
753 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
754 {
755 	/* avoid 0 timestamp, as that means no close timeout */
756 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
757 }
758 
759 static inline void mptcp_start_tout_timer(struct sock *sk)
760 {
761 	mptcp_set_close_tout(sk, tcp_jiffies32);
762 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
763 }
764 
765 static inline bool mptcp_is_fully_established(struct sock *sk)
766 {
767 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
768 	       READ_ONCE(mptcp_sk(sk)->fully_established);
769 }
770 
771 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
772 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
773 bool mptcp_finish_join(struct sock *sk);
774 bool mptcp_schedule_work(struct sock *sk);
775 int mptcp_setsockopt(struct sock *sk, int level, int optname,
776 		     sockptr_t optval, unsigned int optlen);
777 int mptcp_getsockopt(struct sock *sk, int level, int optname,
778 		     char __user *optval, int __user *option);
779 
780 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
781 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
782 {
783 	if (use_64bit)
784 		return cur_seq;
785 
786 	return __mptcp_expand_seq(old_seq, cur_seq);
787 }
788 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
789 void __mptcp_data_acked(struct sock *sk);
790 void __mptcp_error_report(struct sock *sk);
791 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
792 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
793 {
794 	return READ_ONCE(msk->snd_data_fin_enable) &&
795 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
796 }
797 
798 static inline void mptcp_write_space(struct sock *sk)
799 {
800 	if (sk_stream_is_writeable(sk)) {
801 		/* pairs with memory barrier in mptcp_poll */
802 		smp_mb();
803 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
804 			sk_stream_write_space(sk);
805 	}
806 }
807 
808 static inline void __mptcp_sync_sndbuf(struct sock *sk)
809 {
810 	struct mptcp_subflow_context *subflow;
811 	int ssk_sndbuf, new_sndbuf;
812 
813 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
814 		return;
815 
816 	new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
817 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
818 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
819 
820 		subflow->cached_sndbuf = ssk_sndbuf;
821 		new_sndbuf += ssk_sndbuf;
822 	}
823 
824 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
825 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
826 	mptcp_write_space(sk);
827 }
828 
829 /* The called held both the msk socket and the subflow socket locks,
830  * possibly under BH
831  */
832 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
833 {
834 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
835 
836 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
837 		__mptcp_sync_sndbuf(sk);
838 }
839 
840 /* the caller held only the subflow socket lock, either in process or
841  * BH context. Additionally this can be called under the msk data lock,
842  * so we can't acquire such lock here: let the delegate action acquires
843  * the needed locks in suitable order.
844  */
845 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
846 {
847 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
848 
849 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
850 		return;
851 
852 	local_bh_disable();
853 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
854 	local_bh_enable();
855 }
856 
857 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
858 
859 #define MPTCP_TOKEN_MAX_RETRIES	4
860 
861 void __init mptcp_token_init(void);
862 static inline void mptcp_token_init_request(struct request_sock *req)
863 {
864 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
865 }
866 
867 int mptcp_token_new_request(struct request_sock *req);
868 void mptcp_token_destroy_request(struct request_sock *req);
869 int mptcp_token_new_connect(struct sock *ssk);
870 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
871 			struct mptcp_sock *msk);
872 bool mptcp_token_exists(u32 token);
873 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
874 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
875 					 long *s_num);
876 void mptcp_token_destroy(struct mptcp_sock *msk);
877 
878 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
879 
880 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
881 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
882 
883 void __init mptcp_pm_init(void);
884 void mptcp_pm_data_init(struct mptcp_sock *msk);
885 void mptcp_pm_data_reset(struct mptcp_sock *msk);
886 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
887 			struct mptcp_addr_info *addr);
888 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
889 			 bool require_family,
890 			 struct mptcp_pm_addr_entry *entry);
891 bool mptcp_pm_addr_families_match(const struct sock *sk,
892 				  const struct mptcp_addr_info *loc,
893 				  const struct mptcp_addr_info *rem);
894 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
895 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
896 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
897 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
898 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
899 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
900 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
901 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
902 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
903 				 const struct mptcp_subflow_context *subflow);
904 void mptcp_pm_add_addr_received(const struct sock *ssk,
905 				const struct mptcp_addr_info *addr);
906 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
907 			      const struct mptcp_addr_info *addr);
908 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
909 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
910 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
911 			       const struct mptcp_rm_list *rm_list);
912 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
913 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
914 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
915 				 struct mptcp_addr_info *addr,
916 				 struct mptcp_addr_info *rem,
917 				 u8 bkup);
918 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
919 			      const struct mptcp_addr_info *addr);
920 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
921 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
922 struct mptcp_pm_add_entry *
923 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
924 		       const struct mptcp_addr_info *addr, bool check_id);
925 struct mptcp_pm_add_entry *
926 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
927 				const struct mptcp_addr_info *addr);
928 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
929 					 unsigned int id,
930 					 u8 *flags, int *ifindex);
931 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
932 					    u8 *flags, int *ifindex);
933 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
934 						   unsigned int id,
935 						   u8 *flags, int *ifindex);
936 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
937 		       struct mptcp_pm_addr_entry *loc,
938 		       struct mptcp_pm_addr_entry *rem, u8 bkup);
939 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
940 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
941 				 struct mptcp_pm_addr_entry *loc,
942 				 struct mptcp_pm_addr_entry *rem, u8 bkup);
943 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
944 			   const struct mptcp_addr_info *addr,
945 			   bool echo);
946 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
947 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
948 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
949 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
950 					struct list_head *rm_list);
951 
952 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
953 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
954 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
955 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
956 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
957 
958 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
959 		 const struct sock *ssk, gfp_t gfp);
960 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
961 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
962 void mptcp_event_pm_listener(const struct sock *ssk,
963 			     enum mptcp_event_type event);
964 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
965 
966 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
967 				     const struct mptcp_options_received *mp_opt);
968 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
969 					      struct request_sock *req);
970 
971 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
972 {
973 	return READ_ONCE(msk->pm.addr_signal) &
974 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
975 }
976 
977 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
978 {
979 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
980 }
981 
982 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
983 {
984 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
985 }
986 
987 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
988 {
989 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
990 }
991 
992 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
993 {
994 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
995 }
996 
997 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
998 {
999 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1000 }
1001 
1002 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1003 {
1004 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1005 
1006 	if (family == AF_INET6)
1007 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1008 	if (!echo)
1009 		len += MPTCPOPT_THMAC_LEN;
1010 	/* account for 2 trailing 'nop' options */
1011 	if (port)
1012 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1013 
1014 	return len;
1015 }
1016 
1017 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1018 {
1019 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1020 		return -EINVAL;
1021 
1022 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1023 }
1024 
1025 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1026 			      unsigned int opt_size, unsigned int remaining,
1027 			      struct mptcp_addr_info *addr, bool *echo,
1028 			      bool *drop_other_suboptions);
1029 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1030 			     struct mptcp_rm_list *rm_list);
1031 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1032 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1033 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1034 
1035 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1036 {
1037 	int local_id = READ_ONCE(subflow->local_id);
1038 
1039 	if (local_id < 0)
1040 		return 0;
1041 	return local_id;
1042 }
1043 
1044 void __init mptcp_pm_nl_init(void);
1045 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1046 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1047 				     const struct mptcp_rm_list *rm_list);
1048 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1049 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1050 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1051 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1052 
1053 /* called under PM lock */
1054 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1055 {
1056 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1057 		WRITE_ONCE(msk->pm.accept_subflow, true);
1058 }
1059 
1060 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1061 {
1062 	spin_lock_bh(&msk->pm.lock);
1063 	__mptcp_pm_close_subflow(msk);
1064 	spin_unlock_bh(&msk->pm.lock);
1065 }
1066 
1067 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1068 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1069 
1070 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1071 {
1072 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1073 }
1074 
1075 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1076 
1077 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1078 {
1079 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1080 }
1081 
1082 static inline bool mptcp_check_fallback(const struct sock *sk)
1083 {
1084 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1085 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1086 
1087 	return __mptcp_check_fallback(msk);
1088 }
1089 
1090 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1091 {
1092 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1093 		pr_debug("TCP fallback already done (msk=%p)", msk);
1094 		return;
1095 	}
1096 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1097 }
1098 
1099 static inline void mptcp_do_fallback(struct sock *ssk)
1100 {
1101 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1102 	struct sock *sk = subflow->conn;
1103 	struct mptcp_sock *msk;
1104 
1105 	msk = mptcp_sk(sk);
1106 	__mptcp_do_fallback(msk);
1107 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1108 		gfp_t saved_allocation = ssk->sk_allocation;
1109 
1110 		/* we are in a atomic (BH) scope, override ssk default for data
1111 		 * fin allocation
1112 		 */
1113 		ssk->sk_allocation = GFP_ATOMIC;
1114 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1115 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1116 		ssk->sk_allocation = saved_allocation;
1117 	}
1118 }
1119 
1120 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1121 
1122 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1123 {
1124 	struct mptcp_ext *mpext;
1125 
1126 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1127 	if (mpext && mpext->infinite_map)
1128 		return true;
1129 
1130 	return false;
1131 }
1132 
1133 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1134 {
1135 	return (subflow->request_mptcp || subflow->request_join);
1136 }
1137 
1138 static inline bool subflow_simultaneous_connect(struct sock *sk)
1139 {
1140 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1141 
1142 	return (1 << sk->sk_state) &
1143 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1144 	       is_active_ssk(subflow) &&
1145 	       !subflow->conn_finished;
1146 }
1147 
1148 #ifdef CONFIG_SYN_COOKIES
1149 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1150 				       struct sk_buff *skb);
1151 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1152 					struct sk_buff *skb);
1153 void __init mptcp_join_cookie_init(void);
1154 #else
1155 static inline void
1156 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1157 				  struct sk_buff *skb) {}
1158 static inline bool
1159 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1160 				   struct sk_buff *skb)
1161 {
1162 	return false;
1163 }
1164 
1165 static inline void mptcp_join_cookie_init(void) {}
1166 #endif
1167 
1168 #endif /* __MPTCP_PROTOCOL_H */
1169