xref: /openbmc/linux/net/mptcp/protocol.h (revision 10d9f8ed)
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 	int		rmem_fwd_alloc;
272 	int		snd_burst;
273 	int		old_wspace;
274 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
275 						 * recovery related fields are under data_lock
276 						 * protection
277 						 */
278 	u64		bytes_acked;
279 	u64		snd_una;
280 	u64		wnd_end;
281 	unsigned long	timer_ival;
282 	u32		token;
283 	int		rmem_released;
284 	unsigned long	flags;
285 	unsigned long	cb_flags;
286 	unsigned long	push_pending;
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 	struct work_struct work;
307 	struct sk_buff  *ooo_last_skb;
308 	struct rb_root  out_of_order_queue;
309 	struct sk_buff_head receive_queue;
310 	struct list_head conn_list;
311 	struct list_head rtx_queue;
312 	struct mptcp_data_frag *first_pending;
313 	struct list_head join_list;
314 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
315 				 * ONCE annotation, the subflow outside the socket
316 				 * lock as such sock is freed after close().
317 				 */
318 	struct mptcp_pm_data	pm;
319 	struct mptcp_sched_ops	*sched;
320 	struct {
321 		u32	space;	/* bytes copied in last measurement window */
322 		u32	copied; /* bytes copied in this measurement window */
323 		u64	time;	/* start time of measurement window */
324 		u64	rtt_us; /* last maximum rtt of subflows */
325 	} rcvq_space;
326 	u8		scaling_ratio;
327 
328 	u32		subflow_id;
329 	u32		setsockopt_seq;
330 	char		ca_name[TCP_CA_NAME_MAX];
331 };
332 
333 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
334 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
335 
336 #define mptcp_for_each_subflow(__msk, __subflow)			\
337 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
338 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
339 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
340 
341 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
342 {
343 	sock_owned_by_me((const struct sock *)msk);
344 }
345 
346 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
347 
348 /* the msk socket don't use the backlog, also account for the bulk
349  * free memory
350  */
351 static inline int __mptcp_rmem(const struct sock *sk)
352 {
353 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
354 }
355 
356 static inline int mptcp_win_from_space(const struct sock *sk, int space)
357 {
358 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
359 }
360 
361 static inline int __mptcp_space(const struct sock *sk)
362 {
363 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
364 }
365 
366 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
367 {
368 	const struct mptcp_sock *msk = mptcp_sk(sk);
369 
370 	return READ_ONCE(msk->first_pending);
371 }
372 
373 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
374 {
375 	struct mptcp_sock *msk = mptcp_sk(sk);
376 	struct mptcp_data_frag *cur;
377 
378 	cur = msk->first_pending;
379 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
380 						     list_next_entry(cur, list);
381 }
382 
383 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
384 {
385 	const struct mptcp_sock *msk = mptcp_sk(sk);
386 
387 	if (!msk->first_pending)
388 		return NULL;
389 
390 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
391 		return NULL;
392 
393 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
394 }
395 
396 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
397 {
398 	struct mptcp_sock *msk = mptcp_sk(sk);
399 
400 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
401 		return NULL;
402 
403 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
404 }
405 
406 struct csum_pseudo_header {
407 	__be64 data_seq;
408 	__be32 subflow_seq;
409 	__be16 data_len;
410 	__sum16 csum;
411 };
412 
413 struct mptcp_subflow_request_sock {
414 	struct	tcp_request_sock sk;
415 	u16	mp_capable : 1,
416 		mp_join : 1,
417 		backup : 1,
418 		csum_reqd : 1,
419 		allow_join_id0 : 1;
420 	u8	local_id;
421 	u8	remote_id;
422 	u64	local_key;
423 	u64	idsn;
424 	u32	token;
425 	u32	ssn_offset;
426 	u64	thmac;
427 	u32	local_nonce;
428 	u32	remote_nonce;
429 	struct mptcp_sock	*msk;
430 	struct hlist_nulls_node token_node;
431 };
432 
433 static inline struct mptcp_subflow_request_sock *
434 mptcp_subflow_rsk(const struct request_sock *rsk)
435 {
436 	return (struct mptcp_subflow_request_sock *)rsk;
437 }
438 
439 enum mptcp_data_avail {
440 	MPTCP_SUBFLOW_NODATA,
441 	MPTCP_SUBFLOW_DATA_AVAIL,
442 };
443 
444 struct mptcp_delegated_action {
445 	struct napi_struct napi;
446 	struct list_head head;
447 };
448 
449 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
450 
451 #define MPTCP_DELEGATE_SCHEDULED	0
452 #define MPTCP_DELEGATE_SEND		1
453 #define MPTCP_DELEGATE_ACK		2
454 #define MPTCP_DELEGATE_SNDBUF		3
455 
456 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
457 /* MPTCP subflow context */
458 struct mptcp_subflow_context {
459 	struct	list_head node;/* conn_list of subflows */
460 
461 	struct_group(reset,
462 
463 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
464 	u64	local_key;
465 	u64	remote_key;
466 	u64	idsn;
467 	u64	map_seq;
468 	u32	snd_isn;
469 	u32	token;
470 	u32	rel_write_seq;
471 	u32	map_subflow_seq;
472 	u32	ssn_offset;
473 	u32	map_data_len;
474 	__wsum	map_data_csum;
475 	u32	map_csum_len;
476 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
477 		request_join : 1,   /* send MP_JOIN */
478 		request_bkup : 1,
479 		mp_capable : 1,	    /* remote is MPTCP capable */
480 		mp_join : 1,	    /* remote is JOINing */
481 		fully_established : 1,	    /* path validated */
482 		pm_notified : 1,    /* PM hook called for established status */
483 		conn_finished : 1,
484 		map_valid : 1,
485 		map_csum_reqd : 1,
486 		map_data_fin : 1,
487 		mpc_map : 1,
488 		backup : 1,
489 		send_mp_prio : 1,
490 		send_mp_fail : 1,
491 		send_fastclose : 1,
492 		send_infinite_map : 1,
493 		remote_key_valid : 1,        /* received the peer key from */
494 		disposable : 1,	    /* ctx can be free at ulp release time */
495 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
496 		local_id_valid : 1, /* local_id is correctly initialized */
497 		valid_csum_seen : 1,        /* at least one csum validated */
498 		is_mptfo : 1,	    /* subflow is doing TFO */
499 		__unused : 9;
500 	enum mptcp_data_avail data_avail;
501 	bool	scheduled;
502 	u32	remote_nonce;
503 	u64	thmac;
504 	u32	local_nonce;
505 	u32	remote_token;
506 	union {
507 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
508 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
509 	};
510 	u8	local_id;
511 	u8	remote_id;
512 	u8	reset_seen:1;
513 	u8	reset_transient:1;
514 	u8	reset_reason:4;
515 	u8	stale_count;
516 
517 	u32	subflow_id;
518 
519 	long	delegated_status;
520 	unsigned long	fail_tout;
521 
522 	);
523 
524 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
525 
526 	u32	setsockopt_seq;
527 	u32	stale_rcv_tstamp;
528 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
529 				     * protected by the msk socket lock
530 				     */
531 
532 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
533 	struct	sock *conn;	    /* parent mptcp_sock */
534 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
535 	void	(*tcp_state_change)(struct sock *sk);
536 	void	(*tcp_error_report)(struct sock *sk);
537 
538 	struct	rcu_head rcu;
539 };
540 
541 static inline struct mptcp_subflow_context *
542 mptcp_subflow_ctx(const struct sock *sk)
543 {
544 	struct inet_connection_sock *icsk = inet_csk(sk);
545 
546 	/* Use RCU on icsk_ulp_data only for sock diag code */
547 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
548 }
549 
550 static inline struct sock *
551 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
552 {
553 	return subflow->tcp_sock;
554 }
555 
556 static inline void
557 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
558 {
559 	memset(&subflow->reset, 0, sizeof(subflow->reset));
560 	subflow->request_mptcp = 1;
561 }
562 
563 static inline u64
564 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
565 {
566 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
567 		      subflow->ssn_offset -
568 		      subflow->map_subflow_seq;
569 }
570 
571 static inline u64
572 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
573 {
574 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
575 }
576 
577 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
578 
579 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
580 {
581 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
582 	struct mptcp_delegated_action *delegated;
583 	bool schedule;
584 
585 	/* the caller held the subflow bh socket lock */
586 	lockdep_assert_in_softirq();
587 
588 	/* The implied barrier pairs with tcp_release_cb_override()
589 	 * mptcp_napi_poll(), and ensures the below list check sees list
590 	 * updates done prior to delegated status bits changes
591 	 */
592 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
593 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
594 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
595 			return;
596 
597 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
598 		schedule = list_empty(&delegated->head);
599 		list_add_tail(&subflow->delegated_node, &delegated->head);
600 		sock_hold(mptcp_subflow_tcp_sock(subflow));
601 		if (schedule)
602 			napi_schedule(&delegated->napi);
603 	}
604 }
605 
606 static inline struct mptcp_subflow_context *
607 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
608 {
609 	struct mptcp_subflow_context *ret;
610 
611 	if (list_empty(&delegated->head))
612 		return NULL;
613 
614 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
615 	list_del_init(&ret->delegated_node);
616 	return ret;
617 }
618 
619 int mptcp_is_enabled(const struct net *net);
620 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
621 int mptcp_is_checksum_enabled(const struct net *net);
622 int mptcp_allow_join_id0(const struct net *net);
623 unsigned int mptcp_stale_loss_cnt(const struct net *net);
624 int mptcp_get_pm_type(const struct net *net);
625 const char *mptcp_get_scheduler(const struct net *net);
626 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
627 				     const struct mptcp_options_received *mp_opt);
628 bool __mptcp_retransmit_pending_data(struct sock *sk);
629 void mptcp_check_and_set_pending(struct sock *sk);
630 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
631 bool mptcp_subflow_data_available(struct sock *sk);
632 void __init mptcp_subflow_init(void);
633 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
634 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
635 		     struct mptcp_subflow_context *subflow);
636 void __mptcp_subflow_send_ack(struct sock *ssk);
637 void mptcp_subflow_reset(struct sock *ssk);
638 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
639 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
640 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
641 bool __mptcp_close(struct sock *sk, long timeout);
642 void mptcp_cancel_work(struct sock *sk);
643 void __mptcp_unaccepted_force_close(struct sock *sk);
644 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
645 
646 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
647 			   const struct mptcp_addr_info *b, bool use_port);
648 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
649 
650 /* called with sk socket lock held */
651 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
652 			    const struct mptcp_addr_info *remote);
653 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
654 				struct socket **new_sock);
655 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
656 			 struct sockaddr_storage *addr,
657 			 unsigned short family);
658 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
659 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
660 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
661 void mptcp_sched_init(void);
662 int mptcp_init_sched(struct mptcp_sock *msk,
663 		     struct mptcp_sched_ops *sched);
664 void mptcp_release_sched(struct mptcp_sock *msk);
665 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
666 				 bool scheduled);
667 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
668 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
669 int mptcp_sched_get_send(struct mptcp_sock *msk);
670 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
671 
672 static inline bool __tcp_can_send(const struct sock *ssk)
673 {
674 	/* only send if our side has not closed yet */
675 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
676 }
677 
678 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
679 {
680 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
681 	if (subflow->request_join && !subflow->fully_established)
682 		return false;
683 
684 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
685 }
686 
687 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
688 
689 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
690 
691 void mptcp_subflow_drop_ctx(struct sock *ssk);
692 
693 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
694 					      struct mptcp_subflow_context *ctx)
695 {
696 	sk->sk_data_ready = sock_def_readable;
697 	sk->sk_state_change = ctx->tcp_state_change;
698 	sk->sk_write_space = sk_stream_write_space;
699 	sk->sk_error_report = ctx->tcp_error_report;
700 
701 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
702 }
703 
704 void __init mptcp_proto_init(void);
705 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
706 int __init mptcp_proto_v6_init(void);
707 #endif
708 
709 struct sock *mptcp_sk_clone_init(const struct sock *sk,
710 				 const struct mptcp_options_received *mp_opt,
711 				 struct sock *ssk,
712 				 struct request_sock *req);
713 void mptcp_get_options(const struct sk_buff *skb,
714 		       struct mptcp_options_received *mp_opt);
715 
716 void mptcp_finish_connect(struct sock *sk);
717 void __mptcp_sync_state(struct sock *sk, int state);
718 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
719 
720 static inline void mptcp_stop_tout_timer(struct sock *sk)
721 {
722 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
723 		return;
724 
725 	sk_stop_timer(sk, &sk->sk_timer);
726 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
727 }
728 
729 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
730 {
731 	/* avoid 0 timestamp, as that means no close timeout */
732 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
733 }
734 
735 static inline void mptcp_start_tout_timer(struct sock *sk)
736 {
737 	mptcp_set_close_tout(sk, tcp_jiffies32);
738 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
739 }
740 
741 static inline bool mptcp_is_fully_established(struct sock *sk)
742 {
743 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
744 	       READ_ONCE(mptcp_sk(sk)->fully_established);
745 }
746 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
747 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
748 bool mptcp_finish_join(struct sock *sk);
749 bool mptcp_schedule_work(struct sock *sk);
750 int mptcp_setsockopt(struct sock *sk, int level, int optname,
751 		     sockptr_t optval, unsigned int optlen);
752 int mptcp_getsockopt(struct sock *sk, int level, int optname,
753 		     char __user *optval, int __user *option);
754 
755 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
756 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
757 {
758 	if (use_64bit)
759 		return cur_seq;
760 
761 	return __mptcp_expand_seq(old_seq, cur_seq);
762 }
763 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
764 void __mptcp_data_acked(struct sock *sk);
765 void __mptcp_error_report(struct sock *sk);
766 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
767 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
768 {
769 	return READ_ONCE(msk->snd_data_fin_enable) &&
770 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
771 }
772 
773 static inline void __mptcp_sync_sndbuf(struct sock *sk)
774 {
775 	struct mptcp_subflow_context *subflow;
776 	int ssk_sndbuf, new_sndbuf;
777 
778 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
779 		return;
780 
781 	new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
782 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
783 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
784 
785 		subflow->cached_sndbuf = ssk_sndbuf;
786 		new_sndbuf += ssk_sndbuf;
787 	}
788 
789 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
790 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
791 }
792 
793 /* The called held both the msk socket and the subflow socket locks,
794  * possibly under BH
795  */
796 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
797 {
798 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
799 
800 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
801 		__mptcp_sync_sndbuf(sk);
802 }
803 
804 /* the caller held only the subflow socket lock, either in process or
805  * BH context. Additionally this can be called under the msk data lock,
806  * so we can't acquire such lock here: let the delegate action acquires
807  * the needed locks in suitable order.
808  */
809 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
810 {
811 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
812 
813 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
814 		return;
815 
816 	local_bh_disable();
817 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
818 	local_bh_enable();
819 }
820 
821 static inline void mptcp_write_space(struct sock *sk)
822 {
823 	if (sk_stream_is_writeable(sk)) {
824 		/* pairs with memory barrier in mptcp_poll */
825 		smp_mb();
826 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
827 			sk_stream_write_space(sk);
828 	}
829 }
830 
831 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
832 
833 #define MPTCP_TOKEN_MAX_RETRIES	4
834 
835 void __init mptcp_token_init(void);
836 static inline void mptcp_token_init_request(struct request_sock *req)
837 {
838 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
839 }
840 
841 int mptcp_token_new_request(struct request_sock *req);
842 void mptcp_token_destroy_request(struct request_sock *req);
843 int mptcp_token_new_connect(struct sock *ssk);
844 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
845 			struct mptcp_sock *msk);
846 bool mptcp_token_exists(u32 token);
847 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
848 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
849 					 long *s_num);
850 void mptcp_token_destroy(struct mptcp_sock *msk);
851 
852 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
853 
854 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
855 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
856 
857 void __init mptcp_pm_init(void);
858 void mptcp_pm_data_init(struct mptcp_sock *msk);
859 void mptcp_pm_data_reset(struct mptcp_sock *msk);
860 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
861 			struct mptcp_addr_info *addr);
862 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
863 			 bool require_family,
864 			 struct mptcp_pm_addr_entry *entry);
865 bool mptcp_pm_addr_families_match(const struct sock *sk,
866 				  const struct mptcp_addr_info *loc,
867 				  const struct mptcp_addr_info *rem);
868 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
869 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
870 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
871 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
872 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
873 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
874 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
875 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
876 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
877 				 const struct mptcp_subflow_context *subflow);
878 void mptcp_pm_add_addr_received(const struct sock *ssk,
879 				const struct mptcp_addr_info *addr);
880 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
881 			      const struct mptcp_addr_info *addr);
882 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
883 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
884 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
885 			       const struct mptcp_rm_list *rm_list);
886 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
887 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
888 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
889 				 struct mptcp_addr_info *addr,
890 				 struct mptcp_addr_info *rem,
891 				 u8 bkup);
892 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
893 			      const struct mptcp_addr_info *addr);
894 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
895 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
896 struct mptcp_pm_add_entry *
897 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
898 		       const struct mptcp_addr_info *addr, bool check_id);
899 struct mptcp_pm_add_entry *
900 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
901 				const struct mptcp_addr_info *addr);
902 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
903 					 unsigned int id,
904 					 u8 *flags, int *ifindex);
905 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
906 					    u8 *flags, int *ifindex);
907 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
908 						   unsigned int id,
909 						   u8 *flags, int *ifindex);
910 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
911 		       struct mptcp_pm_addr_entry *loc,
912 		       struct mptcp_pm_addr_entry *rem, u8 bkup);
913 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
914 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
915 				 struct mptcp_pm_addr_entry *loc,
916 				 struct mptcp_pm_addr_entry *rem, u8 bkup);
917 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
918 			   const struct mptcp_addr_info *addr,
919 			   bool echo);
920 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
921 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
922 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
923 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
924 					struct list_head *rm_list);
925 
926 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
927 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
928 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
929 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
930 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
931 
932 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
933 		 const struct sock *ssk, gfp_t gfp);
934 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
935 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
936 void mptcp_event_pm_listener(const struct sock *ssk,
937 			     enum mptcp_event_type event);
938 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
939 
940 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
941 				   const struct mptcp_options_received *mp_opt);
942 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
943 					      struct request_sock *req);
944 
945 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
946 {
947 	return READ_ONCE(msk->pm.addr_signal) &
948 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
949 }
950 
951 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
952 {
953 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
954 }
955 
956 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
957 {
958 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
959 }
960 
961 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
962 {
963 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
964 }
965 
966 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
967 {
968 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
969 }
970 
971 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
972 {
973 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
974 }
975 
976 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
977 {
978 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
979 
980 	if (family == AF_INET6)
981 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
982 	if (!echo)
983 		len += MPTCPOPT_THMAC_LEN;
984 	/* account for 2 trailing 'nop' options */
985 	if (port)
986 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
987 
988 	return len;
989 }
990 
991 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
992 {
993 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
994 		return -EINVAL;
995 
996 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
997 }
998 
999 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1000 			      unsigned int opt_size, unsigned int remaining,
1001 			      struct mptcp_addr_info *addr, bool *echo,
1002 			      bool *drop_other_suboptions);
1003 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1004 			     struct mptcp_rm_list *rm_list);
1005 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1006 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1007 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1008 
1009 void __init mptcp_pm_nl_init(void);
1010 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1011 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1012 				     const struct mptcp_rm_list *rm_list);
1013 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1014 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1015 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1016 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1017 
1018 /* called under PM lock */
1019 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1020 {
1021 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1022 		WRITE_ONCE(msk->pm.accept_subflow, true);
1023 }
1024 
1025 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1026 {
1027 	spin_lock_bh(&msk->pm.lock);
1028 	__mptcp_pm_close_subflow(msk);
1029 	spin_unlock_bh(&msk->pm.lock);
1030 }
1031 
1032 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1033 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1034 
1035 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1036 {
1037 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1038 }
1039 
1040 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1041 
1042 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1043 {
1044 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1045 }
1046 
1047 static inline bool mptcp_check_fallback(const struct sock *sk)
1048 {
1049 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1050 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1051 
1052 	return __mptcp_check_fallback(msk);
1053 }
1054 
1055 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1056 {
1057 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1058 		pr_debug("TCP fallback already done (msk=%p)", msk);
1059 		return;
1060 	}
1061 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1062 }
1063 
1064 static inline void mptcp_do_fallback(struct sock *ssk)
1065 {
1066 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1067 	struct sock *sk = subflow->conn;
1068 	struct mptcp_sock *msk;
1069 
1070 	msk = mptcp_sk(sk);
1071 	__mptcp_do_fallback(msk);
1072 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1073 		gfp_t saved_allocation = ssk->sk_allocation;
1074 
1075 		/* we are in a atomic (BH) scope, override ssk default for data
1076 		 * fin allocation
1077 		 */
1078 		ssk->sk_allocation = GFP_ATOMIC;
1079 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1080 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1081 		ssk->sk_allocation = saved_allocation;
1082 	}
1083 }
1084 
1085 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1086 
1087 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1088 {
1089 	struct mptcp_ext *mpext;
1090 
1091 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1092 	if (mpext && mpext->infinite_map)
1093 		return true;
1094 
1095 	return false;
1096 }
1097 
1098 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1099 {
1100 	return (subflow->request_mptcp || subflow->request_join);
1101 }
1102 
1103 static inline bool subflow_simultaneous_connect(struct sock *sk)
1104 {
1105 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1106 
1107 	return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_FIN_WAIT1) &&
1108 	       is_active_ssk(subflow) &&
1109 	       !subflow->conn_finished;
1110 }
1111 
1112 #ifdef CONFIG_SYN_COOKIES
1113 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1114 				       struct sk_buff *skb);
1115 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1116 					struct sk_buff *skb);
1117 void __init mptcp_join_cookie_init(void);
1118 #else
1119 static inline void
1120 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1121 				  struct sk_buff *skb) {}
1122 static inline bool
1123 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1124 				   struct sk_buff *skb)
1125 {
1126 	return false;
1127 }
1128 
1129 static inline void mptcp_join_cookie_init(void) {}
1130 #endif
1131 
1132 #endif /* __MPTCP_PROTOCOL_H */
1133