xref: /linux/net/mptcp/protocol.h (revision 2a52ca7c98960aafb0eca9ef96b2d0c932171357)
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 #include <net/rstreason.h>
16 
17 #define MPTCP_SUPPORTED_VERSION	1
18 
19 /* MPTCP option bits */
20 #define OPTION_MPTCP_MPC_SYN	BIT(0)
21 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
22 #define OPTION_MPTCP_MPC_ACK	BIT(2)
23 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
24 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
25 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
26 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
27 #define OPTION_MPTCP_RM_ADDR	BIT(7)
28 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
29 #define OPTION_MPTCP_PRIO	BIT(9)
30 #define OPTION_MPTCP_RST	BIT(10)
31 #define OPTION_MPTCP_DSS	BIT(11)
32 #define OPTION_MPTCP_FAIL	BIT(12)
33 
34 #define OPTION_MPTCP_CSUMREQD	BIT(13)
35 
36 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
37 				 OPTION_MPTCP_MPC_ACK)
38 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
39 				 OPTION_MPTCP_MPJ_ACK)
40 
41 /* MPTCP option subtypes */
42 #define MPTCPOPT_MP_CAPABLE	0
43 #define MPTCPOPT_MP_JOIN	1
44 #define MPTCPOPT_DSS		2
45 #define MPTCPOPT_ADD_ADDR	3
46 #define MPTCPOPT_RM_ADDR	4
47 #define MPTCPOPT_MP_PRIO	5
48 #define MPTCPOPT_MP_FAIL	6
49 #define MPTCPOPT_MP_FASTCLOSE	7
50 #define MPTCPOPT_RST		8
51 
52 /* MPTCP suboption lengths */
53 #define TCPOLEN_MPTCP_MPC_SYN		4
54 #define TCPOLEN_MPTCP_MPC_SYNACK	12
55 #define TCPOLEN_MPTCP_MPC_ACK		20
56 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
57 #define TCPOLEN_MPTCP_MPJ_SYN		12
58 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
59 #define TCPOLEN_MPTCP_MPJ_ACK		24
60 #define TCPOLEN_MPTCP_DSS_BASE		4
61 #define TCPOLEN_MPTCP_DSS_ACK32		4
62 #define TCPOLEN_MPTCP_DSS_ACK64		8
63 #define TCPOLEN_MPTCP_DSS_MAP32		10
64 #define TCPOLEN_MPTCP_DSS_MAP64		14
65 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
66 #define TCPOLEN_MPTCP_ADD_ADDR		16
67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
70 #define TCPOLEN_MPTCP_ADD_ADDR6		28
71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
74 #define TCPOLEN_MPTCP_PORT_LEN		2
75 #define TCPOLEN_MPTCP_PORT_ALIGN	2
76 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
77 #define TCPOLEN_MPTCP_PRIO		3
78 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
79 #define TCPOLEN_MPTCP_FASTCLOSE		12
80 #define TCPOLEN_MPTCP_RST		4
81 #define TCPOLEN_MPTCP_FAIL		12
82 
83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
84 
85 /* MPTCP MP_JOIN flags */
86 #define MPTCPOPT_BACKUP		BIT(0)
87 #define MPTCPOPT_THMAC_LEN	8
88 
89 /* MPTCP MP_CAPABLE flags */
90 #define MPTCP_VERSION_MASK	(0x0F)
91 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
92 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
93 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
94 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
95 #define MPTCP_CAP_FLAG_MASK	(0x1F)
96 
97 /* MPTCP DSS flags */
98 #define MPTCP_DSS_DATA_FIN	BIT(4)
99 #define MPTCP_DSS_DSN64		BIT(3)
100 #define MPTCP_DSS_HAS_MAP	BIT(2)
101 #define MPTCP_DSS_ACK64		BIT(1)
102 #define MPTCP_DSS_HAS_ACK	BIT(0)
103 #define MPTCP_DSS_FLAG_MASK	(0x1F)
104 
105 /* MPTCP ADD_ADDR flags */
106 #define MPTCP_ADDR_ECHO		BIT(0)
107 
108 /* MPTCP MP_PRIO flags */
109 #define MPTCP_PRIO_BKUP		BIT(0)
110 
111 /* MPTCP TCPRST flags */
112 #define MPTCP_RST_TRANSIENT	BIT(0)
113 
114 /* MPTCP socket atomic flags */
115 #define MPTCP_WORK_RTX		1
116 #define MPTCP_FALLBACK_DONE	2
117 #define MPTCP_WORK_CLOSE_SUBFLOW 3
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;		/* protected by the first subflow socket lock
262 						 * lockless access read
263 						 */
264 	u64		remote_key;		/* same as above */
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 	u32		last_data_sent;
285 	u32		last_data_recv;
286 	u32		last_ack_recv;
287 	unsigned long	timer_ival;
288 	u32		token;
289 	int		rmem_released;
290 	unsigned long	flags;
291 	unsigned long	cb_flags;
292 	bool		recovery;		/* closing subflow write queue reinjected */
293 	bool		can_ack;
294 	bool		fully_established;
295 	bool		rcv_data_fin;
296 	bool		snd_data_fin_enable;
297 	bool		rcv_fastclose;
298 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
299 	bool		csum_enabled;
300 	bool		allow_infinite_fallback;
301 	u8		pending_state; /* A subflow asked to set this sk_state,
302 					* protected by the msk data lock
303 					*/
304 	u8		mpc_endpoint_id;
305 	u8		recvmsg_inq:1,
306 			cork:1,
307 			nodelay:1,
308 			fastopening:1,
309 			in_accept_queue:1,
310 			free_first:1,
311 			rcvspace_init:1;
312 	u32		notsent_lowat;
313 	int		keepalive_cnt;
314 	int		keepalive_idle;
315 	int		keepalive_intvl;
316 	struct work_struct work;
317 	struct sk_buff  *ooo_last_skb;
318 	struct rb_root  out_of_order_queue;
319 	struct sk_buff_head receive_queue;
320 	struct list_head conn_list;
321 	struct list_head rtx_queue;
322 	struct mptcp_data_frag *first_pending;
323 	struct list_head join_list;
324 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
325 				 * ONCE annotation, the subflow outside the socket
326 				 * lock as such sock is freed after close().
327 				 */
328 	struct mptcp_pm_data	pm;
329 	struct mptcp_sched_ops	*sched;
330 	struct {
331 		u32	space;	/* bytes copied in last measurement window */
332 		u32	copied; /* bytes copied in this measurement window */
333 		u64	time;	/* start time of measurement window */
334 		u64	rtt_us; /* last maximum rtt of subflows */
335 	} rcvq_space;
336 	u8		scaling_ratio;
337 
338 	u32		subflow_id;
339 	u32		setsockopt_seq;
340 	char		ca_name[TCP_CA_NAME_MAX];
341 };
342 
343 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
344 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
345 
346 #define mptcp_for_each_subflow(__msk, __subflow)			\
347 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
348 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
349 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
350 
351 extern struct genl_family mptcp_genl_family;
352 
353 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
354 {
355 	sock_owned_by_me((const struct sock *)msk);
356 }
357 
358 #ifdef CONFIG_DEBUG_NET
359 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
360 #undef tcp_sk
361 #define tcp_sk(ptr) ({								\
362 	typeof(ptr) _ptr = (ptr);						\
363 	WARN_ON(_ptr->sk_protocol != IPPROTO_TCP);				\
364 	container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk);	\
365 })
366 #define mptcp_sk(ptr) ({						\
367 	typeof(ptr) _ptr = (ptr);					\
368 	WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP);			\
369 	container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk);	\
370 })
371 
372 #else /* !CONFIG_DEBUG_NET */
373 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
374 #endif
375 
376 /* the msk socket don't use the backlog, also account for the bulk
377  * free memory
378  */
379 static inline int __mptcp_rmem(const struct sock *sk)
380 {
381 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
382 }
383 
384 static inline int mptcp_win_from_space(const struct sock *sk, int space)
385 {
386 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
387 }
388 
389 static inline int mptcp_space_from_win(const struct sock *sk, int win)
390 {
391 	return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
392 }
393 
394 static inline int __mptcp_space(const struct sock *sk)
395 {
396 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
397 }
398 
399 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
400 {
401 	const struct mptcp_sock *msk = mptcp_sk(sk);
402 
403 	return READ_ONCE(msk->first_pending);
404 }
405 
406 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
407 {
408 	struct mptcp_sock *msk = mptcp_sk(sk);
409 	struct mptcp_data_frag *cur;
410 
411 	cur = msk->first_pending;
412 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
413 						     list_next_entry(cur, list);
414 }
415 
416 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
417 {
418 	const struct mptcp_sock *msk = mptcp_sk(sk);
419 
420 	if (!msk->first_pending)
421 		return NULL;
422 
423 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
424 		return NULL;
425 
426 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
427 }
428 
429 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
430 {
431 	struct mptcp_sock *msk = mptcp_sk(sk);
432 
433 	if (msk->snd_una == msk->snd_nxt)
434 		return NULL;
435 
436 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
437 }
438 
439 struct csum_pseudo_header {
440 	__be64 data_seq;
441 	__be32 subflow_seq;
442 	__be16 data_len;
443 	__sum16 csum;
444 };
445 
446 struct mptcp_subflow_request_sock {
447 	struct	tcp_request_sock sk;
448 	u16	mp_capable : 1,
449 		mp_join : 1,
450 		backup : 1,
451 		csum_reqd : 1,
452 		allow_join_id0 : 1;
453 	u8	local_id;
454 	u8	remote_id;
455 	u64	local_key;
456 	u64	idsn;
457 	u32	token;
458 	u32	ssn_offset;
459 	u64	thmac;
460 	u32	local_nonce;
461 	u32	remote_nonce;
462 	struct mptcp_sock	*msk;
463 	struct hlist_nulls_node token_node;
464 };
465 
466 static inline struct mptcp_subflow_request_sock *
467 mptcp_subflow_rsk(const struct request_sock *rsk)
468 {
469 	return (struct mptcp_subflow_request_sock *)rsk;
470 }
471 
472 struct mptcp_delegated_action {
473 	struct napi_struct napi;
474 	struct list_head head;
475 };
476 
477 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
478 
479 #define MPTCP_DELEGATE_SCHEDULED	0
480 #define MPTCP_DELEGATE_SEND		1
481 #define MPTCP_DELEGATE_ACK		2
482 #define MPTCP_DELEGATE_SNDBUF		3
483 
484 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
485 /* MPTCP subflow context */
486 struct mptcp_subflow_context {
487 	struct	list_head node;/* conn_list of subflows */
488 
489 	struct_group(reset,
490 
491 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
492 	u64	local_key;
493 	u64	remote_key;
494 	u64	idsn;
495 	u64	map_seq;
496 	u32	snd_isn;
497 	u32	token;
498 	u32	rel_write_seq;
499 	u32	map_subflow_seq;
500 	u32	ssn_offset;
501 	u32	map_data_len;
502 	__wsum	map_data_csum;
503 	u32	map_csum_len;
504 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
505 		request_join : 1,   /* send MP_JOIN */
506 		request_bkup : 1,
507 		mp_capable : 1,	    /* remote is MPTCP capable */
508 		mp_join : 1,	    /* remote is JOINing */
509 		fully_established : 1,	    /* path validated */
510 		pm_notified : 1,    /* PM hook called for established status */
511 		conn_finished : 1,
512 		map_valid : 1,
513 		map_csum_reqd : 1,
514 		map_data_fin : 1,
515 		mpc_map : 1,
516 		backup : 1,
517 		send_mp_prio : 1,
518 		send_mp_fail : 1,
519 		send_fastclose : 1,
520 		send_infinite_map : 1,
521 		remote_key_valid : 1,        /* received the peer key from */
522 		disposable : 1,	    /* ctx can be free at ulp release time */
523 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
524 		valid_csum_seen : 1,        /* at least one csum validated */
525 		is_mptfo : 1,	    /* subflow is doing TFO */
526 		__unused : 10;
527 	bool	data_avail;
528 	bool	scheduled;
529 	u32	remote_nonce;
530 	u64	thmac;
531 	u32	local_nonce;
532 	u32	remote_token;
533 	union {
534 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
535 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
536 	};
537 	s16	local_id;	    /* if negative not initialized yet */
538 	u8	remote_id;
539 	u8	reset_seen:1;
540 	u8	reset_transient:1;
541 	u8	reset_reason:4;
542 	u8	stale_count;
543 
544 	u32	subflow_id;
545 
546 	long	delegated_status;
547 	unsigned long	fail_tout;
548 
549 	);
550 
551 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
552 
553 	u32	setsockopt_seq;
554 	u32	stale_rcv_tstamp;
555 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
556 				     * protected by the msk socket lock
557 				     */
558 
559 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
560 	struct	sock *conn;	    /* parent mptcp_sock */
561 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
562 	void	(*tcp_state_change)(struct sock *sk);
563 	void	(*tcp_error_report)(struct sock *sk);
564 
565 	struct	rcu_head rcu;
566 };
567 
568 static inline struct mptcp_subflow_context *
569 mptcp_subflow_ctx(const struct sock *sk)
570 {
571 	const struct inet_connection_sock *icsk = inet_csk(sk);
572 
573 	/* Use RCU on icsk_ulp_data only for sock diag code */
574 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
575 }
576 
577 static inline struct sock *
578 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
579 {
580 	return subflow->tcp_sock;
581 }
582 
583 static inline void
584 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
585 {
586 	memset(&subflow->reset, 0, sizeof(subflow->reset));
587 	subflow->request_mptcp = 1;
588 	WRITE_ONCE(subflow->local_id, -1);
589 }
590 
591 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
592 static inline enum sk_rst_reason
593 sk_rst_convert_mptcp_reason(u32 reason)
594 {
595 	switch (reason) {
596 	case MPTCP_RST_EUNSPEC:
597 		return SK_RST_REASON_MPTCP_RST_EUNSPEC;
598 	case MPTCP_RST_EMPTCP:
599 		return SK_RST_REASON_MPTCP_RST_EMPTCP;
600 	case MPTCP_RST_ERESOURCE:
601 		return SK_RST_REASON_MPTCP_RST_ERESOURCE;
602 	case MPTCP_RST_EPROHIBIT:
603 		return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
604 	case MPTCP_RST_EWQ2BIG:
605 		return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
606 	case MPTCP_RST_EBADPERF:
607 		return SK_RST_REASON_MPTCP_RST_EBADPERF;
608 	case MPTCP_RST_EMIDDLEBOX:
609 		return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
610 	default:
611 		/* It should not happen, or else errors may occur
612 		 * in MPTCP layer
613 		 */
614 		return SK_RST_REASON_ERROR;
615 	}
616 }
617 
618 static inline void
619 mptcp_send_active_reset_reason(struct sock *sk)
620 {
621 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
622 	enum sk_rst_reason reason;
623 
624 	reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
625 	tcp_send_active_reset(sk, GFP_ATOMIC, reason);
626 }
627 
628 static inline u64
629 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
630 {
631 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
632 		      subflow->ssn_offset -
633 		      subflow->map_subflow_seq;
634 }
635 
636 static inline u64
637 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
638 {
639 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
640 }
641 
642 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
643 
644 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
645 {
646 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
647 	struct mptcp_delegated_action *delegated;
648 	bool schedule;
649 
650 	/* the caller held the subflow bh socket lock */
651 	lockdep_assert_in_softirq();
652 
653 	/* The implied barrier pairs with tcp_release_cb_override()
654 	 * mptcp_napi_poll(), and ensures the below list check sees list
655 	 * updates done prior to delegated status bits changes
656 	 */
657 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
658 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
659 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
660 			return;
661 
662 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
663 		schedule = list_empty(&delegated->head);
664 		list_add_tail(&subflow->delegated_node, &delegated->head);
665 		sock_hold(mptcp_subflow_tcp_sock(subflow));
666 		if (schedule)
667 			napi_schedule(&delegated->napi);
668 	}
669 }
670 
671 static inline struct mptcp_subflow_context *
672 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
673 {
674 	struct mptcp_subflow_context *ret;
675 
676 	if (list_empty(&delegated->head))
677 		return NULL;
678 
679 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
680 	list_del_init(&ret->delegated_node);
681 	return ret;
682 }
683 
684 int mptcp_is_enabled(const struct net *net);
685 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
686 int mptcp_is_checksum_enabled(const struct net *net);
687 int mptcp_allow_join_id0(const struct net *net);
688 unsigned int mptcp_stale_loss_cnt(const struct net *net);
689 unsigned int mptcp_close_timeout(const struct sock *sk);
690 int mptcp_get_pm_type(const struct net *net);
691 const char *mptcp_get_scheduler(const struct net *net);
692 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
693 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
694 				       struct mptcp_subflow_context *subflow,
695 				       const struct mptcp_options_received *mp_opt);
696 bool __mptcp_retransmit_pending_data(struct sock *sk);
697 void mptcp_check_and_set_pending(struct sock *sk);
698 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
699 bool mptcp_subflow_data_available(struct sock *sk);
700 void __init mptcp_subflow_init(void);
701 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
702 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
703 		     struct mptcp_subflow_context *subflow);
704 void __mptcp_subflow_send_ack(struct sock *ssk);
705 void mptcp_subflow_reset(struct sock *ssk);
706 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
707 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
708 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
709 bool __mptcp_close(struct sock *sk, long timeout);
710 void mptcp_cancel_work(struct sock *sk);
711 void __mptcp_unaccepted_force_close(struct sock *sk);
712 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
713 void mptcp_set_state(struct sock *sk, int state);
714 
715 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
716 			   const struct mptcp_addr_info *b, bool use_port);
717 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
718 
719 /* called with sk socket lock held */
720 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
721 			    const struct mptcp_addr_info *remote);
722 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
723 				struct socket **new_sock);
724 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
725 			 struct sockaddr_storage *addr,
726 			 unsigned short family);
727 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
728 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
729 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
730 void mptcp_sched_init(void);
731 int mptcp_init_sched(struct mptcp_sock *msk,
732 		     struct mptcp_sched_ops *sched);
733 void mptcp_release_sched(struct mptcp_sock *msk);
734 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
735 				 bool scheduled);
736 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
737 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
738 int mptcp_sched_get_send(struct mptcp_sock *msk);
739 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
740 
741 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
742 {
743 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
744 }
745 
746 static inline bool mptcp_epollin_ready(const struct sock *sk)
747 {
748 	/* mptcp doesn't have to deal with small skbs in the receive queue,
749 	 * at it can always coalesce them
750 	 */
751 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
752 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
753 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
754 	       READ_ONCE(tcp_memory_pressure);
755 }
756 
757 int mptcp_set_rcvlowat(struct sock *sk, int val);
758 
759 static inline bool __tcp_can_send(const struct sock *ssk)
760 {
761 	/* only send if our side has not closed yet */
762 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
763 }
764 
765 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
766 {
767 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
768 	if (subflow->request_join && !subflow->fully_established)
769 		return false;
770 
771 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
772 }
773 
774 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
775 
776 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
777 
778 void mptcp_subflow_drop_ctx(struct sock *ssk);
779 
780 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
781 					      struct mptcp_subflow_context *ctx)
782 {
783 	sk->sk_data_ready = sock_def_readable;
784 	sk->sk_state_change = ctx->tcp_state_change;
785 	sk->sk_write_space = sk_stream_write_space;
786 	sk->sk_error_report = ctx->tcp_error_report;
787 
788 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
789 }
790 
791 void __init mptcp_proto_init(void);
792 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
793 int __init mptcp_proto_v6_init(void);
794 #endif
795 
796 struct sock *mptcp_sk_clone_init(const struct sock *sk,
797 				 const struct mptcp_options_received *mp_opt,
798 				 struct sock *ssk,
799 				 struct request_sock *req);
800 void mptcp_get_options(const struct sk_buff *skb,
801 		       struct mptcp_options_received *mp_opt);
802 
803 void mptcp_finish_connect(struct sock *sk);
804 void __mptcp_sync_state(struct sock *sk, int state);
805 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
806 
807 static inline void mptcp_stop_tout_timer(struct sock *sk)
808 {
809 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
810 		return;
811 
812 	sk_stop_timer(sk, &sk->sk_timer);
813 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
814 }
815 
816 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
817 {
818 	/* avoid 0 timestamp, as that means no close timeout */
819 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
820 }
821 
822 static inline void mptcp_start_tout_timer(struct sock *sk)
823 {
824 	mptcp_set_close_tout(sk, tcp_jiffies32);
825 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
826 }
827 
828 static inline bool mptcp_is_fully_established(struct sock *sk)
829 {
830 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
831 	       READ_ONCE(mptcp_sk(sk)->fully_established);
832 }
833 
834 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
835 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
836 bool mptcp_finish_join(struct sock *sk);
837 bool mptcp_schedule_work(struct sock *sk);
838 int mptcp_setsockopt(struct sock *sk, int level, int optname,
839 		     sockptr_t optval, unsigned int optlen);
840 int mptcp_getsockopt(struct sock *sk, int level, int optname,
841 		     char __user *optval, int __user *option);
842 
843 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
844 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
845 {
846 	if (use_64bit)
847 		return cur_seq;
848 
849 	return __mptcp_expand_seq(old_seq, cur_seq);
850 }
851 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
852 void __mptcp_data_acked(struct sock *sk);
853 void __mptcp_error_report(struct sock *sk);
854 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
855 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
856 {
857 	return READ_ONCE(msk->snd_data_fin_enable) &&
858 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
859 }
860 
861 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
862 {
863 	struct net *net = sock_net(sk);
864 	u32 val;
865 
866 	val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
867 	return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
868 }
869 
870 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
871 {
872 	const struct mptcp_sock *msk = mptcp_sk(sk);
873 	u32 notsent_bytes;
874 
875 	notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
876 	return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
877 }
878 
879 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
880 {
881 	return mptcp_stream_memory_free(sk, wake) &&
882 	       __sk_stream_is_writeable(sk, wake);
883 }
884 
885 static inline void mptcp_write_space(struct sock *sk)
886 {
887 	/* pairs with memory barrier in mptcp_poll */
888 	smp_mb();
889 	if (mptcp_stream_memory_free(sk, 1))
890 		sk_stream_write_space(sk);
891 }
892 
893 static inline void __mptcp_sync_sndbuf(struct sock *sk)
894 {
895 	struct mptcp_subflow_context *subflow;
896 	int ssk_sndbuf, new_sndbuf;
897 
898 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
899 		return;
900 
901 	new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
902 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
903 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
904 
905 		subflow->cached_sndbuf = ssk_sndbuf;
906 		new_sndbuf += ssk_sndbuf;
907 	}
908 
909 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
910 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
911 	mptcp_write_space(sk);
912 }
913 
914 /* The called held both the msk socket and the subflow socket locks,
915  * possibly under BH
916  */
917 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
918 {
919 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
920 
921 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
922 		__mptcp_sync_sndbuf(sk);
923 }
924 
925 /* the caller held only the subflow socket lock, either in process or
926  * BH context. Additionally this can be called under the msk data lock,
927  * so we can't acquire such lock here: let the delegate action acquires
928  * the needed locks in suitable order.
929  */
930 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
931 {
932 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
933 
934 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
935 		return;
936 
937 	local_bh_disable();
938 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
939 	local_bh_enable();
940 }
941 
942 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
943 
944 #define MPTCP_TOKEN_MAX_RETRIES	4
945 
946 void __init mptcp_token_init(void);
947 static inline void mptcp_token_init_request(struct request_sock *req)
948 {
949 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
950 }
951 
952 int mptcp_token_new_request(struct request_sock *req);
953 void mptcp_token_destroy_request(struct request_sock *req);
954 int mptcp_token_new_connect(struct sock *ssk);
955 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
956 			struct mptcp_sock *msk);
957 bool mptcp_token_exists(u32 token);
958 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
959 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
960 					 long *s_num);
961 void mptcp_token_destroy(struct mptcp_sock *msk);
962 
963 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
964 
965 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
966 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
967 
968 void __init mptcp_pm_init(void);
969 void mptcp_pm_data_init(struct mptcp_sock *msk);
970 void mptcp_pm_data_reset(struct mptcp_sock *msk);
971 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
972 			struct mptcp_addr_info *addr);
973 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
974 			 bool require_family,
975 			 struct mptcp_pm_addr_entry *entry);
976 bool mptcp_pm_addr_families_match(const struct sock *sk,
977 				  const struct mptcp_addr_info *loc,
978 				  const struct mptcp_addr_info *rem);
979 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
980 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
981 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
982 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
983 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
984 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
985 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
986 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
987 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
988 				 const struct mptcp_subflow_context *subflow);
989 void mptcp_pm_add_addr_received(const struct sock *ssk,
990 				const struct mptcp_addr_info *addr);
991 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
992 			      const struct mptcp_addr_info *addr);
993 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
994 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
995 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
996 			       const struct mptcp_rm_list *rm_list);
997 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
998 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
999 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
1000 				 struct mptcp_addr_info *addr,
1001 				 struct mptcp_addr_info *rem,
1002 				 u8 bkup);
1003 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1004 			      const struct mptcp_addr_info *addr);
1005 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1006 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1007 struct mptcp_pm_add_entry *
1008 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1009 		       const struct mptcp_addr_info *addr, bool check_id);
1010 struct mptcp_pm_add_entry *
1011 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1012 				const struct mptcp_addr_info *addr);
1013 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1014 					 unsigned int id,
1015 					 u8 *flags, int *ifindex);
1016 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
1017 					    u8 *flags, int *ifindex);
1018 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1019 						   unsigned int id,
1020 						   u8 *flags, int *ifindex);
1021 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1022 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1023 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1024 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1025 			   const struct mptcp_addr_info *addr,
1026 			   bool echo);
1027 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1028 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1029 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
1030 
1031 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1032 
1033 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1034 		 const struct sock *ssk, gfp_t gfp);
1035 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1036 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1037 void mptcp_event_pm_listener(const struct sock *ssk,
1038 			     enum mptcp_event_type event);
1039 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1040 
1041 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1042 				     const struct mptcp_options_received *mp_opt);
1043 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1044 					      struct request_sock *req);
1045 int mptcp_nl_fill_addr(struct sk_buff *skb,
1046 		       struct mptcp_pm_addr_entry *entry);
1047 
1048 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1049 {
1050 	return READ_ONCE(msk->pm.addr_signal) &
1051 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1052 }
1053 
1054 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1055 {
1056 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1057 }
1058 
1059 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1060 {
1061 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1062 }
1063 
1064 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1065 {
1066 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1067 }
1068 
1069 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1070 {
1071 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1072 }
1073 
1074 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1075 {
1076 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1077 }
1078 
1079 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1080 {
1081 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1082 
1083 	if (family == AF_INET6)
1084 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1085 	if (!echo)
1086 		len += MPTCPOPT_THMAC_LEN;
1087 	/* account for 2 trailing 'nop' options */
1088 	if (port)
1089 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1090 
1091 	return len;
1092 }
1093 
1094 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1095 {
1096 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1097 		return -EINVAL;
1098 
1099 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1100 }
1101 
1102 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1103 			      unsigned int opt_size, unsigned int remaining,
1104 			      struct mptcp_addr_info *addr, bool *echo,
1105 			      bool *drop_other_suboptions);
1106 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1107 			     struct mptcp_rm_list *rm_list);
1108 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1109 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1110 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1111 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1112 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1113 			  struct netlink_callback *cb);
1114 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1115 				 struct netlink_callback *cb);
1116 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1117 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1118 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1119 				struct genl_info *info);
1120 
1121 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1122 {
1123 	int local_id = READ_ONCE(subflow->local_id);
1124 
1125 	if (local_id < 0)
1126 		return 0;
1127 	return local_id;
1128 }
1129 
1130 void __init mptcp_pm_nl_init(void);
1131 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1132 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1133 				     const struct mptcp_rm_list *rm_list);
1134 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1135 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1136 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1137 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1138 
1139 /* called under PM lock */
1140 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1141 {
1142 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1143 		WRITE_ONCE(msk->pm.accept_subflow, true);
1144 }
1145 
1146 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1147 {
1148 	spin_lock_bh(&msk->pm.lock);
1149 	__mptcp_pm_close_subflow(msk);
1150 	spin_unlock_bh(&msk->pm.lock);
1151 }
1152 
1153 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1154 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1155 
1156 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1157 {
1158 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1159 }
1160 
1161 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1162 
1163 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1164 {
1165 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1166 }
1167 
1168 static inline bool mptcp_check_fallback(const struct sock *sk)
1169 {
1170 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1171 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1172 
1173 	return __mptcp_check_fallback(msk);
1174 }
1175 
1176 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1177 {
1178 	if (__mptcp_check_fallback(msk)) {
1179 		pr_debug("TCP fallback already done (msk=%p)", msk);
1180 		return;
1181 	}
1182 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1183 }
1184 
1185 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1186 {
1187 	struct sock *ssk = READ_ONCE(msk->first);
1188 
1189 	return ssk && ((1 << inet_sk_state_load(ssk)) &
1190 		       (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1191 			TCPF_SYN_RECV | TCPF_LISTEN));
1192 }
1193 
1194 static inline void mptcp_do_fallback(struct sock *ssk)
1195 {
1196 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1197 	struct sock *sk = subflow->conn;
1198 	struct mptcp_sock *msk;
1199 
1200 	msk = mptcp_sk(sk);
1201 	__mptcp_do_fallback(msk);
1202 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1203 		gfp_t saved_allocation = ssk->sk_allocation;
1204 
1205 		/* we are in a atomic (BH) scope, override ssk default for data
1206 		 * fin allocation
1207 		 */
1208 		ssk->sk_allocation = GFP_ATOMIC;
1209 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1210 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1211 		ssk->sk_allocation = saved_allocation;
1212 	}
1213 }
1214 
1215 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1216 
1217 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1218 {
1219 	struct mptcp_ext *mpext;
1220 
1221 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1222 	if (mpext && mpext->infinite_map)
1223 		return true;
1224 
1225 	return false;
1226 }
1227 
1228 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1229 {
1230 	return (subflow->request_mptcp || subflow->request_join);
1231 }
1232 
1233 static inline bool subflow_simultaneous_connect(struct sock *sk)
1234 {
1235 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1236 
1237 	return (1 << sk->sk_state) &
1238 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1239 	       is_active_ssk(subflow) &&
1240 	       !subflow->conn_finished;
1241 }
1242 
1243 #ifdef CONFIG_SYN_COOKIES
1244 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1245 				       struct sk_buff *skb);
1246 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1247 					struct sk_buff *skb);
1248 void __init mptcp_join_cookie_init(void);
1249 #else
1250 static inline void
1251 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1252 				  struct sk_buff *skb) {}
1253 static inline bool
1254 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1255 				   struct sk_buff *skb)
1256 {
1257 	return false;
1258 }
1259 
1260 static inline void mptcp_join_cookie_init(void) {}
1261 #endif
1262 
1263 #endif /* __MPTCP_PROTOCOL_H */
1264