xref: /linux/net/mptcp/protocol.h (revision e7d759f31ca295d589f7420719c311870bb3166f)
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 #include "mptcp_pm_gen.h"
17 
18 #define MPTCP_SUPPORTED_VERSION	1
19 
20 /* MPTCP option bits */
21 #define OPTION_MPTCP_MPC_SYN	BIT(0)
22 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
23 #define OPTION_MPTCP_MPC_ACK	BIT(2)
24 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
25 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
26 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
27 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
28 #define OPTION_MPTCP_RM_ADDR	BIT(7)
29 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
30 #define OPTION_MPTCP_PRIO	BIT(9)
31 #define OPTION_MPTCP_RST	BIT(10)
32 #define OPTION_MPTCP_DSS	BIT(11)
33 #define OPTION_MPTCP_FAIL	BIT(12)
34 
35 #define OPTION_MPTCP_CSUMREQD	BIT(13)
36 
37 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
38 				 OPTION_MPTCP_MPC_ACK)
39 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
40 				 OPTION_MPTCP_MPJ_ACK)
41 
42 /* MPTCP option subtypes */
43 #define MPTCPOPT_MP_CAPABLE	0
44 #define MPTCPOPT_MP_JOIN	1
45 #define MPTCPOPT_DSS		2
46 #define MPTCPOPT_ADD_ADDR	3
47 #define MPTCPOPT_RM_ADDR	4
48 #define MPTCPOPT_MP_PRIO	5
49 #define MPTCPOPT_MP_FAIL	6
50 #define MPTCPOPT_MP_FASTCLOSE	7
51 #define MPTCPOPT_RST		8
52 
53 /* MPTCP suboption lengths */
54 #define TCPOLEN_MPTCP_MPC_SYN		4
55 #define TCPOLEN_MPTCP_MPC_SYNACK	12
56 #define TCPOLEN_MPTCP_MPC_ACK		20
57 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
58 #define TCPOLEN_MPTCP_MPJ_SYN		12
59 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
60 #define TCPOLEN_MPTCP_MPJ_ACK		24
61 #define TCPOLEN_MPTCP_DSS_BASE		4
62 #define TCPOLEN_MPTCP_DSS_ACK32		4
63 #define TCPOLEN_MPTCP_DSS_ACK64		8
64 #define TCPOLEN_MPTCP_DSS_MAP32		10
65 #define TCPOLEN_MPTCP_DSS_MAP64		14
66 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
67 #define TCPOLEN_MPTCP_ADD_ADDR		16
68 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
70 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
71 #define TCPOLEN_MPTCP_ADD_ADDR6		28
72 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
74 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
75 #define TCPOLEN_MPTCP_PORT_LEN		2
76 #define TCPOLEN_MPTCP_PORT_ALIGN	2
77 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
78 #define TCPOLEN_MPTCP_PRIO		3
79 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
80 #define TCPOLEN_MPTCP_FASTCLOSE		12
81 #define TCPOLEN_MPTCP_RST		4
82 #define TCPOLEN_MPTCP_FAIL		12
83 
84 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
85 
86 /* MPTCP MP_JOIN flags */
87 #define MPTCPOPT_BACKUP		BIT(0)
88 #define MPTCPOPT_THMAC_LEN	8
89 
90 /* MPTCP MP_CAPABLE flags */
91 #define MPTCP_VERSION_MASK	(0x0F)
92 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
93 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
94 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
95 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
96 #define MPTCP_CAP_FLAG_MASK	(0x1F)
97 
98 /* MPTCP DSS flags */
99 #define MPTCP_DSS_DATA_FIN	BIT(4)
100 #define MPTCP_DSS_DSN64		BIT(3)
101 #define MPTCP_DSS_HAS_MAP	BIT(2)
102 #define MPTCP_DSS_ACK64		BIT(1)
103 #define MPTCP_DSS_HAS_ACK	BIT(0)
104 #define MPTCP_DSS_FLAG_MASK	(0x1F)
105 
106 /* MPTCP ADD_ADDR flags */
107 #define MPTCP_ADDR_ECHO		BIT(0)
108 
109 /* MPTCP MP_PRIO flags */
110 #define MPTCP_PRIO_BKUP		BIT(0)
111 
112 /* MPTCP TCPRST flags */
113 #define MPTCP_RST_TRANSIENT	BIT(0)
114 
115 /* MPTCP socket atomic flags */
116 #define MPTCP_NOSPACE		1
117 #define MPTCP_WORK_RTX		2
118 #define MPTCP_FALLBACK_DONE	4
119 #define MPTCP_WORK_CLOSE_SUBFLOW 5
120 
121 /* MPTCP socket release cb flags */
122 #define MPTCP_PUSH_PENDING	1
123 #define MPTCP_CLEAN_UNA		2
124 #define MPTCP_ERROR_REPORT	3
125 #define MPTCP_RETRANSMIT	4
126 #define MPTCP_FLUSH_JOIN_LIST	5
127 #define MPTCP_SYNC_STATE	6
128 #define MPTCP_SYNC_SNDBUF	7
129 
130 struct mptcp_skb_cb {
131 	u64 map_seq;
132 	u64 end_seq;
133 	u32 offset;
134 	u8  has_rxtstamp:1;
135 };
136 
137 #define MPTCP_SKB_CB(__skb)	((struct mptcp_skb_cb *)&((__skb)->cb[0]))
138 
139 static inline bool before64(__u64 seq1, __u64 seq2)
140 {
141 	return (__s64)(seq1 - seq2) < 0;
142 }
143 
144 #define after64(seq2, seq1)	before64(seq1, seq2)
145 
146 struct mptcp_options_received {
147 	u64	sndr_key;
148 	u64	rcvr_key;
149 	u64	data_ack;
150 	u64	data_seq;
151 	u32	subflow_seq;
152 	u16	data_len;
153 	__sum16	csum;
154 	u16	suboptions;
155 	u32	token;
156 	u32	nonce;
157 	u16	use_map:1,
158 		dsn64:1,
159 		data_fin:1,
160 		use_ack:1,
161 		ack64:1,
162 		mpc_map:1,
163 		reset_reason:4,
164 		reset_transient:1,
165 		echo:1,
166 		backup:1,
167 		deny_join_id0:1,
168 		__unused:2;
169 	u8	join_id;
170 	u64	thmac;
171 	u8	hmac[MPTCPOPT_HMAC_LEN];
172 	struct mptcp_addr_info addr;
173 	struct mptcp_rm_list rm_list;
174 	u64	ahmac;
175 	u64	fail_seq;
176 };
177 
178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
179 {
180 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
181 		     ((nib & 0xF) << 8) | field);
182 }
183 
184 enum mptcp_pm_status {
185 	MPTCP_PM_ADD_ADDR_RECEIVED,
186 	MPTCP_PM_ADD_ADDR_SEND_ACK,
187 	MPTCP_PM_RM_ADDR_RECEIVED,
188 	MPTCP_PM_ESTABLISHED,
189 	MPTCP_PM_SUBFLOW_ESTABLISHED,
190 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
191 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
192 					 * accounted int id_avail_bitmap
193 					 */
194 };
195 
196 enum mptcp_pm_type {
197 	MPTCP_PM_TYPE_KERNEL = 0,
198 	MPTCP_PM_TYPE_USERSPACE,
199 
200 	__MPTCP_PM_TYPE_NR,
201 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
202 };
203 
204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
206 
207 enum mptcp_addr_signal_status {
208 	MPTCP_ADD_ADDR_SIGNAL,
209 	MPTCP_ADD_ADDR_ECHO,
210 	MPTCP_RM_ADDR_SIGNAL,
211 };
212 
213 /* max value of mptcp_addr_info.id */
214 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
215 
216 struct mptcp_pm_data {
217 	struct mptcp_addr_info local;
218 	struct mptcp_addr_info remote;
219 	struct list_head anno_list;
220 	struct list_head userspace_pm_local_addr_list;
221 
222 	spinlock_t	lock;		/*protects the whole PM data */
223 
224 	u8		addr_signal;
225 	bool		server_side;
226 	bool		work_pending;
227 	bool		accept_addr;
228 	bool		accept_subflow;
229 	bool		remote_deny_join_id0;
230 	u8		add_addr_signaled;
231 	u8		add_addr_accepted;
232 	u8		local_addr_used;
233 	u8		pm_type;
234 	u8		subflows;
235 	u8		status;
236 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
237 	struct mptcp_rm_list rm_list_tx;
238 	struct mptcp_rm_list rm_list_rx;
239 };
240 
241 struct mptcp_pm_addr_entry {
242 	struct list_head	list;
243 	struct mptcp_addr_info	addr;
244 	u8			flags;
245 	int			ifindex;
246 	struct socket		*lsk;
247 };
248 
249 struct mptcp_data_frag {
250 	struct list_head list;
251 	u64 data_seq;
252 	u16 data_len;
253 	u16 offset;
254 	u16 overhead;
255 	u16 already_sent;
256 	struct page *page;
257 };
258 
259 /* MPTCP connection sock */
260 struct mptcp_sock {
261 	/* inet_connection_sock must be the first member */
262 	struct inet_connection_sock sk;
263 	u64		local_key;
264 	u64		remote_key;
265 	u64		write_seq;
266 	u64		bytes_sent;
267 	u64		snd_nxt;
268 	u64		bytes_received;
269 	u64		ack_seq;
270 	atomic64_t	rcv_wnd_sent;
271 	u64		rcv_data_fin_seq;
272 	u64		bytes_retrans;
273 	u64		bytes_consumed;
274 	int		rmem_fwd_alloc;
275 	int		snd_burst;
276 	int		old_wspace;
277 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
278 						 * recovery related fields are under data_lock
279 						 * protection
280 						 */
281 	u64		bytes_acked;
282 	u64		snd_una;
283 	u64		wnd_end;
284 	unsigned long	timer_ival;
285 	u32		token;
286 	int		rmem_released;
287 	unsigned long	flags;
288 	unsigned long	cb_flags;
289 	unsigned long	push_pending;
290 	bool		recovery;		/* closing subflow write queue reinjected */
291 	bool		can_ack;
292 	bool		fully_established;
293 	bool		rcv_data_fin;
294 	bool		snd_data_fin_enable;
295 	bool		rcv_fastclose;
296 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
297 	bool		csum_enabled;
298 	bool		allow_infinite_fallback;
299 	u8		pending_state; /* A subflow asked to set this sk_state,
300 					* protected by the msk data lock
301 					*/
302 	u8		mpc_endpoint_id;
303 	u8		recvmsg_inq:1,
304 			cork:1,
305 			nodelay:1,
306 			fastopening:1,
307 			in_accept_queue:1,
308 			free_first:1;
309 	struct work_struct work;
310 	struct sk_buff  *ooo_last_skb;
311 	struct rb_root  out_of_order_queue;
312 	struct sk_buff_head receive_queue;
313 	struct list_head conn_list;
314 	struct list_head rtx_queue;
315 	struct mptcp_data_frag *first_pending;
316 	struct list_head join_list;
317 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
318 				 * ONCE annotation, the subflow outside the socket
319 				 * lock as such sock is freed after close().
320 				 */
321 	struct mptcp_pm_data	pm;
322 	struct mptcp_sched_ops	*sched;
323 	struct {
324 		u32	space;	/* bytes copied in last measurement window */
325 		u32	copied; /* bytes copied in this measurement window */
326 		u64	time;	/* start time of measurement window */
327 		u64	rtt_us; /* last maximum rtt of subflows */
328 	} rcvq_space;
329 	u8		scaling_ratio;
330 
331 	u32		subflow_id;
332 	u32		setsockopt_seq;
333 	char		ca_name[TCP_CA_NAME_MAX];
334 };
335 
336 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
337 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
338 
339 #define mptcp_for_each_subflow(__msk, __subflow)			\
340 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
341 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
342 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
343 
344 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
345 {
346 	sock_owned_by_me((const struct sock *)msk);
347 }
348 
349 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
350 
351 /* the msk socket don't use the backlog, also account for the bulk
352  * free memory
353  */
354 static inline int __mptcp_rmem(const struct sock *sk)
355 {
356 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
357 }
358 
359 static inline int mptcp_win_from_space(const struct sock *sk, int space)
360 {
361 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
362 }
363 
364 static inline int __mptcp_space(const struct sock *sk)
365 {
366 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
367 }
368 
369 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
370 {
371 	const struct mptcp_sock *msk = mptcp_sk(sk);
372 
373 	return READ_ONCE(msk->first_pending);
374 }
375 
376 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
377 {
378 	struct mptcp_sock *msk = mptcp_sk(sk);
379 	struct mptcp_data_frag *cur;
380 
381 	cur = msk->first_pending;
382 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
383 						     list_next_entry(cur, list);
384 }
385 
386 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
387 {
388 	const struct mptcp_sock *msk = mptcp_sk(sk);
389 
390 	if (!msk->first_pending)
391 		return NULL;
392 
393 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
394 		return NULL;
395 
396 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
397 }
398 
399 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
400 {
401 	struct mptcp_sock *msk = mptcp_sk(sk);
402 
403 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
404 		return NULL;
405 
406 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
407 }
408 
409 struct csum_pseudo_header {
410 	__be64 data_seq;
411 	__be32 subflow_seq;
412 	__be16 data_len;
413 	__sum16 csum;
414 };
415 
416 struct mptcp_subflow_request_sock {
417 	struct	tcp_request_sock sk;
418 	u16	mp_capable : 1,
419 		mp_join : 1,
420 		backup : 1,
421 		csum_reqd : 1,
422 		allow_join_id0 : 1;
423 	u8	local_id;
424 	u8	remote_id;
425 	u64	local_key;
426 	u64	idsn;
427 	u32	token;
428 	u32	ssn_offset;
429 	u64	thmac;
430 	u32	local_nonce;
431 	u32	remote_nonce;
432 	struct mptcp_sock	*msk;
433 	struct hlist_nulls_node token_node;
434 };
435 
436 static inline struct mptcp_subflow_request_sock *
437 mptcp_subflow_rsk(const struct request_sock *rsk)
438 {
439 	return (struct mptcp_subflow_request_sock *)rsk;
440 }
441 
442 struct mptcp_delegated_action {
443 	struct napi_struct napi;
444 	struct list_head head;
445 };
446 
447 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
448 
449 #define MPTCP_DELEGATE_SCHEDULED	0
450 #define MPTCP_DELEGATE_SEND		1
451 #define MPTCP_DELEGATE_ACK		2
452 #define MPTCP_DELEGATE_SNDBUF		3
453 
454 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
455 /* MPTCP subflow context */
456 struct mptcp_subflow_context {
457 	struct	list_head node;/* conn_list of subflows */
458 
459 	struct_group(reset,
460 
461 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
462 	u64	local_key;
463 	u64	remote_key;
464 	u64	idsn;
465 	u64	map_seq;
466 	u32	snd_isn;
467 	u32	token;
468 	u32	rel_write_seq;
469 	u32	map_subflow_seq;
470 	u32	ssn_offset;
471 	u32	map_data_len;
472 	__wsum	map_data_csum;
473 	u32	map_csum_len;
474 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
475 		request_join : 1,   /* send MP_JOIN */
476 		request_bkup : 1,
477 		mp_capable : 1,	    /* remote is MPTCP capable */
478 		mp_join : 1,	    /* remote is JOINing */
479 		fully_established : 1,	    /* path validated */
480 		pm_notified : 1,    /* PM hook called for established status */
481 		conn_finished : 1,
482 		map_valid : 1,
483 		map_csum_reqd : 1,
484 		map_data_fin : 1,
485 		mpc_map : 1,
486 		backup : 1,
487 		send_mp_prio : 1,
488 		send_mp_fail : 1,
489 		send_fastclose : 1,
490 		send_infinite_map : 1,
491 		remote_key_valid : 1,        /* received the peer key from */
492 		disposable : 1,	    /* ctx can be free at ulp release time */
493 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
494 		local_id_valid : 1, /* local_id is correctly initialized */
495 		valid_csum_seen : 1,        /* at least one csum validated */
496 		is_mptfo : 1,	    /* subflow is doing TFO */
497 		__unused : 9;
498 	bool	data_avail;
499 	bool	scheduled;
500 	u32	remote_nonce;
501 	u64	thmac;
502 	u32	local_nonce;
503 	u32	remote_token;
504 	union {
505 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
506 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
507 	};
508 	u8	local_id;
509 	u8	remote_id;
510 	u8	reset_seen:1;
511 	u8	reset_transient:1;
512 	u8	reset_reason:4;
513 	u8	stale_count;
514 
515 	u32	subflow_id;
516 
517 	long	delegated_status;
518 	unsigned long	fail_tout;
519 
520 	);
521 
522 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
523 
524 	u32	setsockopt_seq;
525 	u32	stale_rcv_tstamp;
526 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
527 				     * protected by the msk socket lock
528 				     */
529 
530 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
531 	struct	sock *conn;	    /* parent mptcp_sock */
532 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
533 	void	(*tcp_state_change)(struct sock *sk);
534 	void	(*tcp_error_report)(struct sock *sk);
535 
536 	struct	rcu_head rcu;
537 };
538 
539 static inline struct mptcp_subflow_context *
540 mptcp_subflow_ctx(const struct sock *sk)
541 {
542 	struct inet_connection_sock *icsk = inet_csk(sk);
543 
544 	/* Use RCU on icsk_ulp_data only for sock diag code */
545 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
546 }
547 
548 static inline struct sock *
549 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
550 {
551 	return subflow->tcp_sock;
552 }
553 
554 static inline void
555 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
556 {
557 	memset(&subflow->reset, 0, sizeof(subflow->reset));
558 	subflow->request_mptcp = 1;
559 }
560 
561 static inline u64
562 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
563 {
564 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
565 		      subflow->ssn_offset -
566 		      subflow->map_subflow_seq;
567 }
568 
569 static inline u64
570 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
571 {
572 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
573 }
574 
575 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
576 
577 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
578 {
579 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
580 	struct mptcp_delegated_action *delegated;
581 	bool schedule;
582 
583 	/* the caller held the subflow bh socket lock */
584 	lockdep_assert_in_softirq();
585 
586 	/* The implied barrier pairs with tcp_release_cb_override()
587 	 * mptcp_napi_poll(), and ensures the below list check sees list
588 	 * updates done prior to delegated status bits changes
589 	 */
590 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
591 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
592 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
593 			return;
594 
595 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
596 		schedule = list_empty(&delegated->head);
597 		list_add_tail(&subflow->delegated_node, &delegated->head);
598 		sock_hold(mptcp_subflow_tcp_sock(subflow));
599 		if (schedule)
600 			napi_schedule(&delegated->napi);
601 	}
602 }
603 
604 static inline struct mptcp_subflow_context *
605 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
606 {
607 	struct mptcp_subflow_context *ret;
608 
609 	if (list_empty(&delegated->head))
610 		return NULL;
611 
612 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
613 	list_del_init(&ret->delegated_node);
614 	return ret;
615 }
616 
617 int mptcp_is_enabled(const struct net *net);
618 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
619 int mptcp_is_checksum_enabled(const struct net *net);
620 int mptcp_allow_join_id0(const struct net *net);
621 unsigned int mptcp_stale_loss_cnt(const struct net *net);
622 unsigned int mptcp_close_timeout(const struct sock *sk);
623 int mptcp_get_pm_type(const struct net *net);
624 const char *mptcp_get_scheduler(const struct net *net);
625 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
626 				     const struct mptcp_options_received *mp_opt);
627 bool __mptcp_retransmit_pending_data(struct sock *sk);
628 void mptcp_check_and_set_pending(struct sock *sk);
629 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
630 bool mptcp_subflow_data_available(struct sock *sk);
631 void __init mptcp_subflow_init(void);
632 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
633 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
634 		     struct mptcp_subflow_context *subflow);
635 void __mptcp_subflow_send_ack(struct sock *ssk);
636 void mptcp_subflow_reset(struct sock *ssk);
637 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
638 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
639 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
640 bool __mptcp_close(struct sock *sk, long timeout);
641 void mptcp_cancel_work(struct sock *sk);
642 void __mptcp_unaccepted_force_close(struct sock *sk);
643 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
644 void mptcp_set_state(struct sock *sk, int state);
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 u64 mptcp_data_avail(const struct mptcp_sock *msk)
673 {
674 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
675 }
676 
677 static inline bool mptcp_epollin_ready(const struct sock *sk)
678 {
679 	/* mptcp doesn't have to deal with small skbs in the receive queue,
680 	 * at it can always coalesce them
681 	 */
682 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
683 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
684 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
685 	       READ_ONCE(tcp_memory_pressure);
686 }
687 
688 int mptcp_set_rcvlowat(struct sock *sk, int val);
689 
690 static inline bool __tcp_can_send(const struct sock *ssk)
691 {
692 	/* only send if our side has not closed yet */
693 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
694 }
695 
696 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
697 {
698 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
699 	if (subflow->request_join && !subflow->fully_established)
700 		return false;
701 
702 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
703 }
704 
705 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
706 
707 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
708 
709 void mptcp_subflow_drop_ctx(struct sock *ssk);
710 
711 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
712 					      struct mptcp_subflow_context *ctx)
713 {
714 	sk->sk_data_ready = sock_def_readable;
715 	sk->sk_state_change = ctx->tcp_state_change;
716 	sk->sk_write_space = sk_stream_write_space;
717 	sk->sk_error_report = ctx->tcp_error_report;
718 
719 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
720 }
721 
722 void __init mptcp_proto_init(void);
723 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
724 int __init mptcp_proto_v6_init(void);
725 #endif
726 
727 struct sock *mptcp_sk_clone_init(const struct sock *sk,
728 				 const struct mptcp_options_received *mp_opt,
729 				 struct sock *ssk,
730 				 struct request_sock *req);
731 void mptcp_get_options(const struct sk_buff *skb,
732 		       struct mptcp_options_received *mp_opt);
733 
734 void mptcp_finish_connect(struct sock *sk);
735 void __mptcp_sync_state(struct sock *sk, int state);
736 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
737 
738 static inline void mptcp_stop_tout_timer(struct sock *sk)
739 {
740 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
741 		return;
742 
743 	sk_stop_timer(sk, &sk->sk_timer);
744 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
745 }
746 
747 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
748 {
749 	/* avoid 0 timestamp, as that means no close timeout */
750 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
751 }
752 
753 static inline void mptcp_start_tout_timer(struct sock *sk)
754 {
755 	mptcp_set_close_tout(sk, tcp_jiffies32);
756 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
757 }
758 
759 static inline bool mptcp_is_fully_established(struct sock *sk)
760 {
761 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
762 	       READ_ONCE(mptcp_sk(sk)->fully_established);
763 }
764 
765 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
766 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
767 bool mptcp_finish_join(struct sock *sk);
768 bool mptcp_schedule_work(struct sock *sk);
769 int mptcp_setsockopt(struct sock *sk, int level, int optname,
770 		     sockptr_t optval, unsigned int optlen);
771 int mptcp_getsockopt(struct sock *sk, int level, int optname,
772 		     char __user *optval, int __user *option);
773 
774 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
775 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
776 {
777 	if (use_64bit)
778 		return cur_seq;
779 
780 	return __mptcp_expand_seq(old_seq, cur_seq);
781 }
782 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
783 void __mptcp_data_acked(struct sock *sk);
784 void __mptcp_error_report(struct sock *sk);
785 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
786 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
787 {
788 	return READ_ONCE(msk->snd_data_fin_enable) &&
789 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
790 }
791 
792 static inline void __mptcp_sync_sndbuf(struct sock *sk)
793 {
794 	struct mptcp_subflow_context *subflow;
795 	int ssk_sndbuf, new_sndbuf;
796 
797 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
798 		return;
799 
800 	new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
801 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
802 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
803 
804 		subflow->cached_sndbuf = ssk_sndbuf;
805 		new_sndbuf += ssk_sndbuf;
806 	}
807 
808 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
809 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
810 }
811 
812 /* The called held both the msk socket and the subflow socket locks,
813  * possibly under BH
814  */
815 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
816 {
817 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
818 
819 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
820 		__mptcp_sync_sndbuf(sk);
821 }
822 
823 /* the caller held only the subflow socket lock, either in process or
824  * BH context. Additionally this can be called under the msk data lock,
825  * so we can't acquire such lock here: let the delegate action acquires
826  * the needed locks in suitable order.
827  */
828 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
829 {
830 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
831 
832 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
833 		return;
834 
835 	local_bh_disable();
836 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
837 	local_bh_enable();
838 }
839 
840 static inline void mptcp_write_space(struct sock *sk)
841 {
842 	if (sk_stream_is_writeable(sk)) {
843 		/* pairs with memory barrier in mptcp_poll */
844 		smp_mb();
845 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
846 			sk_stream_write_space(sk);
847 	}
848 }
849 
850 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
851 
852 #define MPTCP_TOKEN_MAX_RETRIES	4
853 
854 void __init mptcp_token_init(void);
855 static inline void mptcp_token_init_request(struct request_sock *req)
856 {
857 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
858 }
859 
860 int mptcp_token_new_request(struct request_sock *req);
861 void mptcp_token_destroy_request(struct request_sock *req);
862 int mptcp_token_new_connect(struct sock *ssk);
863 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
864 			struct mptcp_sock *msk);
865 bool mptcp_token_exists(u32 token);
866 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
867 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
868 					 long *s_num);
869 void mptcp_token_destroy(struct mptcp_sock *msk);
870 
871 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
872 
873 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
874 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
875 
876 void __init mptcp_pm_init(void);
877 void mptcp_pm_data_init(struct mptcp_sock *msk);
878 void mptcp_pm_data_reset(struct mptcp_sock *msk);
879 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
880 			struct mptcp_addr_info *addr);
881 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
882 			 bool require_family,
883 			 struct mptcp_pm_addr_entry *entry);
884 bool mptcp_pm_addr_families_match(const struct sock *sk,
885 				  const struct mptcp_addr_info *loc,
886 				  const struct mptcp_addr_info *rem);
887 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
888 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
889 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
890 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
891 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
892 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
893 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
894 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
895 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
896 				 const struct mptcp_subflow_context *subflow);
897 void mptcp_pm_add_addr_received(const struct sock *ssk,
898 				const struct mptcp_addr_info *addr);
899 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
900 			      const struct mptcp_addr_info *addr);
901 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
902 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
903 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
904 			       const struct mptcp_rm_list *rm_list);
905 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
906 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
907 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
908 				 struct mptcp_addr_info *addr,
909 				 struct mptcp_addr_info *rem,
910 				 u8 bkup);
911 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
912 			      const struct mptcp_addr_info *addr);
913 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
914 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
915 struct mptcp_pm_add_entry *
916 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
917 		       const struct mptcp_addr_info *addr, bool check_id);
918 struct mptcp_pm_add_entry *
919 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
920 				const struct mptcp_addr_info *addr);
921 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
922 					 unsigned int id,
923 					 u8 *flags, int *ifindex);
924 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
925 					    u8 *flags, int *ifindex);
926 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
927 						   unsigned int id,
928 						   u8 *flags, int *ifindex);
929 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
930 		       struct mptcp_pm_addr_entry *loc,
931 		       struct mptcp_pm_addr_entry *rem, u8 bkup);
932 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
933 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
934 				 struct mptcp_pm_addr_entry *loc,
935 				 struct mptcp_pm_addr_entry *rem, u8 bkup);
936 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
937 			   const struct mptcp_addr_info *addr,
938 			   bool echo);
939 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
940 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
941 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
942 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
943 					struct list_head *rm_list);
944 
945 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
946 
947 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
948 		 const struct sock *ssk, gfp_t gfp);
949 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
950 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
951 void mptcp_event_pm_listener(const struct sock *ssk,
952 			     enum mptcp_event_type event);
953 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
954 
955 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
956 				   const struct mptcp_options_received *mp_opt);
957 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
958 					      struct request_sock *req);
959 
960 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
961 {
962 	return READ_ONCE(msk->pm.addr_signal) &
963 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
964 }
965 
966 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
967 {
968 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
969 }
970 
971 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
972 {
973 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
974 }
975 
976 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
977 {
978 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
979 }
980 
981 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
982 {
983 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
984 }
985 
986 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
987 {
988 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
989 }
990 
991 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
992 {
993 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
994 
995 	if (family == AF_INET6)
996 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
997 	if (!echo)
998 		len += MPTCPOPT_THMAC_LEN;
999 	/* account for 2 trailing 'nop' options */
1000 	if (port)
1001 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1002 
1003 	return len;
1004 }
1005 
1006 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1007 {
1008 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1009 		return -EINVAL;
1010 
1011 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1012 }
1013 
1014 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1015 			      unsigned int opt_size, unsigned int remaining,
1016 			      struct mptcp_addr_info *addr, bool *echo,
1017 			      bool *drop_other_suboptions);
1018 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1019 			     struct mptcp_rm_list *rm_list);
1020 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1021 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1022 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1023 
1024 void __init mptcp_pm_nl_init(void);
1025 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1026 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1027 				     const struct mptcp_rm_list *rm_list);
1028 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1029 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1030 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1031 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1032 
1033 /* called under PM lock */
1034 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1035 {
1036 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1037 		WRITE_ONCE(msk->pm.accept_subflow, true);
1038 }
1039 
1040 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1041 {
1042 	spin_lock_bh(&msk->pm.lock);
1043 	__mptcp_pm_close_subflow(msk);
1044 	spin_unlock_bh(&msk->pm.lock);
1045 }
1046 
1047 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1048 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1049 
1050 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1051 {
1052 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1053 }
1054 
1055 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1056 
1057 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1058 {
1059 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1060 }
1061 
1062 static inline bool mptcp_check_fallback(const struct sock *sk)
1063 {
1064 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1065 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1066 
1067 	return __mptcp_check_fallback(msk);
1068 }
1069 
1070 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1071 {
1072 	if (__mptcp_check_fallback(msk)) {
1073 		pr_debug("TCP fallback already done (msk=%p)", msk);
1074 		return;
1075 	}
1076 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1077 }
1078 
1079 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1080 {
1081 	struct sock *ssk = READ_ONCE(msk->first);
1082 
1083 	return ssk && ((1 << inet_sk_state_load(ssk)) &
1084 		       (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1085 			TCPF_SYN_RECV | TCPF_LISTEN));
1086 }
1087 
1088 static inline void mptcp_do_fallback(struct sock *ssk)
1089 {
1090 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1091 	struct sock *sk = subflow->conn;
1092 	struct mptcp_sock *msk;
1093 
1094 	msk = mptcp_sk(sk);
1095 	__mptcp_do_fallback(msk);
1096 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1097 		gfp_t saved_allocation = ssk->sk_allocation;
1098 
1099 		/* we are in a atomic (BH) scope, override ssk default for data
1100 		 * fin allocation
1101 		 */
1102 		ssk->sk_allocation = GFP_ATOMIC;
1103 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1104 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1105 		ssk->sk_allocation = saved_allocation;
1106 	}
1107 }
1108 
1109 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1110 
1111 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1112 {
1113 	struct mptcp_ext *mpext;
1114 
1115 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1116 	if (mpext && mpext->infinite_map)
1117 		return true;
1118 
1119 	return false;
1120 }
1121 
1122 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1123 {
1124 	return (subflow->request_mptcp || subflow->request_join);
1125 }
1126 
1127 static inline bool subflow_simultaneous_connect(struct sock *sk)
1128 {
1129 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1130 
1131 	return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_FIN_WAIT1) &&
1132 	       is_active_ssk(subflow) &&
1133 	       !subflow->conn_finished;
1134 }
1135 
1136 #ifdef CONFIG_SYN_COOKIES
1137 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1138 				       struct sk_buff *skb);
1139 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1140 					struct sk_buff *skb);
1141 void __init mptcp_join_cookie_init(void);
1142 #else
1143 static inline void
1144 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1145 				  struct sk_buff *skb) {}
1146 static inline bool
1147 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1148 				   struct sk_buff *skb)
1149 {
1150 	return false;
1151 }
1152 
1153 static inline void mptcp_join_cookie_init(void) {}
1154 #endif
1155 
1156 #endif /* __MPTCP_PROTOCOL_H */
1157