xref: /openbmc/linux/net/mptcp/protocol.h (revision 7a010c3c)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3  *
4  * Copyright (c) 2017 - 2019, Intel Corporation.
5  */
6 
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9 
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 
15 #define MPTCP_SUPPORTED_VERSION	1
16 
17 /* MPTCP option bits */
18 #define OPTION_MPTCP_MPC_SYN	BIT(0)
19 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
20 #define OPTION_MPTCP_MPC_ACK	BIT(2)
21 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
22 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
23 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
24 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
25 #define OPTION_MPTCP_RM_ADDR	BIT(7)
26 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
27 #define OPTION_MPTCP_PRIO	BIT(9)
28 #define OPTION_MPTCP_RST	BIT(10)
29 #define OPTION_MPTCP_DSS	BIT(11)
30 #define OPTION_MPTCP_FAIL	BIT(12)
31 
32 #define OPTION_MPTCP_CSUMREQD	BIT(13)
33 
34 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
35 				 OPTION_MPTCP_MPC_ACK)
36 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
37 				 OPTION_MPTCP_MPJ_ACK)
38 
39 /* MPTCP option subtypes */
40 #define MPTCPOPT_MP_CAPABLE	0
41 #define MPTCPOPT_MP_JOIN	1
42 #define MPTCPOPT_DSS		2
43 #define MPTCPOPT_ADD_ADDR	3
44 #define MPTCPOPT_RM_ADDR	4
45 #define MPTCPOPT_MP_PRIO	5
46 #define MPTCPOPT_MP_FAIL	6
47 #define MPTCPOPT_MP_FASTCLOSE	7
48 #define MPTCPOPT_RST		8
49 
50 /* MPTCP suboption lengths */
51 #define TCPOLEN_MPTCP_MPC_SYN		4
52 #define TCPOLEN_MPTCP_MPC_SYNACK	12
53 #define TCPOLEN_MPTCP_MPC_ACK		20
54 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
55 #define TCPOLEN_MPTCP_MPJ_SYN		12
56 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
57 #define TCPOLEN_MPTCP_MPJ_ACK		24
58 #define TCPOLEN_MPTCP_DSS_BASE		4
59 #define TCPOLEN_MPTCP_DSS_ACK32		4
60 #define TCPOLEN_MPTCP_DSS_ACK64		8
61 #define TCPOLEN_MPTCP_DSS_MAP32		10
62 #define TCPOLEN_MPTCP_DSS_MAP64		14
63 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
64 #define TCPOLEN_MPTCP_ADD_ADDR		16
65 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
66 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
68 #define TCPOLEN_MPTCP_ADD_ADDR6		28
69 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
70 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
72 #define TCPOLEN_MPTCP_PORT_LEN		2
73 #define TCPOLEN_MPTCP_PORT_ALIGN	2
74 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
75 #define TCPOLEN_MPTCP_PRIO		3
76 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
77 #define TCPOLEN_MPTCP_FASTCLOSE		12
78 #define TCPOLEN_MPTCP_RST		4
79 #define TCPOLEN_MPTCP_FAIL		12
80 
81 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
82 
83 /* MPTCP MP_JOIN flags */
84 #define MPTCPOPT_BACKUP		BIT(0)
85 #define MPTCPOPT_HMAC_LEN	20
86 #define MPTCPOPT_THMAC_LEN	8
87 
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK	(0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
94 #define MPTCP_CAP_FLAG_MASK	(0x1F)
95 
96 /* MPTCP DSS flags */
97 #define MPTCP_DSS_DATA_FIN	BIT(4)
98 #define MPTCP_DSS_DSN64		BIT(3)
99 #define MPTCP_DSS_HAS_MAP	BIT(2)
100 #define MPTCP_DSS_ACK64		BIT(1)
101 #define MPTCP_DSS_HAS_ACK	BIT(0)
102 #define MPTCP_DSS_FLAG_MASK	(0x1F)
103 
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO		BIT(0)
106 
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP		BIT(0)
109 
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT	BIT(0)
112 
113 /* MPTCP socket flags */
114 #define MPTCP_DATA_READY	0
115 #define MPTCP_NOSPACE		1
116 #define MPTCP_WORK_RTX		2
117 #define MPTCP_WORK_EOF		3
118 #define MPTCP_FALLBACK_DONE	4
119 #define MPTCP_WORK_CLOSE_SUBFLOW 5
120 #define MPTCP_PUSH_PENDING	6
121 #define MPTCP_CLEAN_UNA		7
122 #define MPTCP_ERROR_REPORT	8
123 #define MPTCP_RETRANSMIT	9
124 #define MPTCP_WORK_SYNC_SETSOCKOPT 10
125 #define MPTCP_CONNECTED		11
126 
127 static inline bool before64(__u64 seq1, __u64 seq2)
128 {
129 	return (__s64)(seq1 - seq2) < 0;
130 }
131 
132 #define after64(seq2, seq1)	before64(seq1, seq2)
133 
134 struct mptcp_options_received {
135 	u64	sndr_key;
136 	u64	rcvr_key;
137 	u64	data_ack;
138 	u64	data_seq;
139 	u32	subflow_seq;
140 	u16	data_len;
141 	__sum16	csum;
142 	u16	suboptions;
143 	u32	token;
144 	u32	nonce;
145 	u16	use_map:1,
146 		dsn64:1,
147 		data_fin:1,
148 		use_ack:1,
149 		ack64:1,
150 		mpc_map:1,
151 		reset_reason:4,
152 		reset_transient:1,
153 		echo:1,
154 		backup:1,
155 		deny_join_id0:1,
156 		__unused:2;
157 	u8	join_id;
158 	u64	thmac;
159 	u8	hmac[MPTCPOPT_HMAC_LEN];
160 	struct mptcp_addr_info addr;
161 	struct mptcp_rm_list rm_list;
162 	u64	ahmac;
163 	u64	fail_seq;
164 };
165 
166 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
167 {
168 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
169 		     ((nib & 0xF) << 8) | field);
170 }
171 
172 enum mptcp_pm_status {
173 	MPTCP_PM_ADD_ADDR_RECEIVED,
174 	MPTCP_PM_ADD_ADDR_SEND_ACK,
175 	MPTCP_PM_RM_ADDR_RECEIVED,
176 	MPTCP_PM_ESTABLISHED,
177 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
178 	MPTCP_PM_SUBFLOW_ESTABLISHED,
179 };
180 
181 enum mptcp_addr_signal_status {
182 	MPTCP_ADD_ADDR_SIGNAL,
183 	MPTCP_ADD_ADDR_ECHO,
184 	MPTCP_RM_ADDR_SIGNAL,
185 };
186 
187 struct mptcp_pm_data {
188 	struct mptcp_addr_info local;
189 	struct mptcp_addr_info remote;
190 	struct list_head anno_list;
191 
192 	spinlock_t	lock;		/*protects the whole PM data */
193 
194 	u8		addr_signal;
195 	bool		server_side;
196 	bool		work_pending;
197 	bool		accept_addr;
198 	bool		accept_subflow;
199 	bool		remote_deny_join_id0;
200 	u8		add_addr_signaled;
201 	u8		add_addr_accepted;
202 	u8		local_addr_used;
203 	u8		subflows;
204 	u8		status;
205 	struct mptcp_rm_list rm_list_tx;
206 	struct mptcp_rm_list rm_list_rx;
207 };
208 
209 struct mptcp_data_frag {
210 	struct list_head list;
211 	u64 data_seq;
212 	u16 data_len;
213 	u16 offset;
214 	u16 overhead;
215 	u16 already_sent;
216 	struct page *page;
217 };
218 
219 /* MPTCP connection sock */
220 struct mptcp_sock {
221 	/* inet_connection_sock must be the first member */
222 	struct inet_connection_sock sk;
223 	u64		local_key;
224 	u64		remote_key;
225 	u64		write_seq;
226 	u64		snd_nxt;
227 	u64		ack_seq;
228 	u64		rcv_wnd_sent;
229 	u64		rcv_data_fin_seq;
230 	int		wmem_reserved;
231 	struct sock	*last_snd;
232 	int		snd_burst;
233 	int		old_wspace;
234 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
235 						 * recovery related fields are under data_lock
236 						 * protection
237 						 */
238 	u64		snd_una;
239 	u64		wnd_end;
240 	unsigned long	timer_ival;
241 	u32		token;
242 	int		rmem_released;
243 	unsigned long	flags;
244 	bool		recovery;		/* closing subflow write queue reinjected */
245 	bool		can_ack;
246 	bool		fully_established;
247 	bool		rcv_data_fin;
248 	bool		snd_data_fin_enable;
249 	bool		rcv_fastclose;
250 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
251 	bool		csum_enabled;
252 	spinlock_t	join_list_lock;
253 	struct work_struct work;
254 	struct sk_buff  *ooo_last_skb;
255 	struct rb_root  out_of_order_queue;
256 	struct sk_buff_head receive_queue;
257 	int		tx_pending_data;
258 	struct list_head conn_list;
259 	struct list_head rtx_queue;
260 	struct mptcp_data_frag *first_pending;
261 	struct list_head join_list;
262 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
263 	struct sock	*first;
264 	struct mptcp_pm_data	pm;
265 	struct {
266 		u32	space;	/* bytes copied in last measurement window */
267 		u32	copied; /* bytes copied in this measurement window */
268 		u64	time;	/* start time of measurement window */
269 		u64	rtt_us; /* last maximum rtt of subflows */
270 	} rcvq_space;
271 
272 	u32 setsockopt_seq;
273 	char		ca_name[TCP_CA_NAME_MAX];
274 };
275 
276 #define mptcp_lock_sock(___sk, cb) do {					\
277 	struct sock *__sk = (___sk); /* silence macro reuse warning */	\
278 	might_sleep();							\
279 	spin_lock_bh(&__sk->sk_lock.slock);				\
280 	if (__sk->sk_lock.owned)					\
281 		__lock_sock(__sk);					\
282 	cb;								\
283 	__sk->sk_lock.owned = 1;					\
284 	spin_unlock(&__sk->sk_lock.slock);				\
285 	mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_);		\
286 	local_bh_enable();						\
287 } while (0)
288 
289 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
290 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
291 
292 #define mptcp_for_each_subflow(__msk, __subflow)			\
293 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
294 
295 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
296 {
297 	sock_owned_by_me((const struct sock *)msk);
298 }
299 
300 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
301 {
302 	return (struct mptcp_sock *)sk;
303 }
304 
305 /* the msk socket don't use the backlog, also account for the bulk
306  * free memory
307  */
308 static inline int __mptcp_rmem(const struct sock *sk)
309 {
310 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
311 }
312 
313 static inline int __mptcp_space(const struct sock *sk)
314 {
315 	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
316 }
317 
318 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
319 {
320 	const struct mptcp_sock *msk = mptcp_sk(sk);
321 
322 	return READ_ONCE(msk->first_pending);
323 }
324 
325 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
326 {
327 	struct mptcp_sock *msk = mptcp_sk(sk);
328 	struct mptcp_data_frag *cur;
329 
330 	cur = msk->first_pending;
331 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
332 						     list_next_entry(cur, list);
333 }
334 
335 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
336 {
337 	struct mptcp_sock *msk = mptcp_sk(sk);
338 
339 	if (!msk->first_pending)
340 		return NULL;
341 
342 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
343 		return NULL;
344 
345 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
346 }
347 
348 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
349 {
350 	struct mptcp_sock *msk = mptcp_sk(sk);
351 
352 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
353 		return NULL;
354 
355 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
356 }
357 
358 struct csum_pseudo_header {
359 	__be64 data_seq;
360 	__be32 subflow_seq;
361 	__be16 data_len;
362 	__sum16 csum;
363 };
364 
365 struct mptcp_subflow_request_sock {
366 	struct	tcp_request_sock sk;
367 	u16	mp_capable : 1,
368 		mp_join : 1,
369 		backup : 1,
370 		csum_reqd : 1,
371 		allow_join_id0 : 1;
372 	u8	local_id;
373 	u8	remote_id;
374 	u64	local_key;
375 	u64	idsn;
376 	u32	token;
377 	u32	ssn_offset;
378 	u64	thmac;
379 	u32	local_nonce;
380 	u32	remote_nonce;
381 	struct mptcp_sock	*msk;
382 	struct hlist_nulls_node token_node;
383 };
384 
385 static inline struct mptcp_subflow_request_sock *
386 mptcp_subflow_rsk(const struct request_sock *rsk)
387 {
388 	return (struct mptcp_subflow_request_sock *)rsk;
389 }
390 
391 enum mptcp_data_avail {
392 	MPTCP_SUBFLOW_NODATA,
393 	MPTCP_SUBFLOW_DATA_AVAIL,
394 };
395 
396 struct mptcp_delegated_action {
397 	struct napi_struct napi;
398 	struct list_head head;
399 };
400 
401 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
402 
403 #define MPTCP_DELEGATE_SEND		0
404 
405 /* MPTCP subflow context */
406 struct mptcp_subflow_context {
407 	struct	list_head node;/* conn_list of subflows */
408 	u64	local_key;
409 	u64	remote_key;
410 	u64	idsn;
411 	u64	map_seq;
412 	u32	snd_isn;
413 	u32	token;
414 	u32	rel_write_seq;
415 	u32	map_subflow_seq;
416 	u32	ssn_offset;
417 	u32	map_data_len;
418 	__wsum	map_data_csum;
419 	u32	map_csum_len;
420 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
421 		request_join : 1,   /* send MP_JOIN */
422 		request_bkup : 1,
423 		mp_capable : 1,	    /* remote is MPTCP capable */
424 		mp_join : 1,	    /* remote is JOINing */
425 		fully_established : 1,	    /* path validated */
426 		pm_notified : 1,    /* PM hook called for established status */
427 		conn_finished : 1,
428 		map_valid : 1,
429 		map_csum_reqd : 1,
430 		map_data_fin : 1,
431 		mpc_map : 1,
432 		backup : 1,
433 		send_mp_prio : 1,
434 		send_mp_fail : 1,
435 		rx_eof : 1,
436 		can_ack : 1,        /* only after processing the remote a key */
437 		disposable : 1,	    /* ctx can be free at ulp release time */
438 		stale : 1;	    /* unable to snd/rcv data, do not use for xmit */
439 	enum mptcp_data_avail data_avail;
440 	u32	remote_nonce;
441 	u64	thmac;
442 	u32	local_nonce;
443 	u32	remote_token;
444 	u8	hmac[MPTCPOPT_HMAC_LEN];
445 	u8	local_id;
446 	u8	remote_id;
447 	u8	reset_seen:1;
448 	u8	reset_transient:1;
449 	u8	reset_reason:4;
450 	u8	stale_count;
451 
452 	long	delegated_status;
453 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
454 
455 	u32	setsockopt_seq;
456 	u32	stale_rcv_tstamp;
457 
458 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
459 	struct	sock *conn;	    /* parent mptcp_sock */
460 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
461 	void	(*tcp_data_ready)(struct sock *sk);
462 	void	(*tcp_state_change)(struct sock *sk);
463 	void	(*tcp_write_space)(struct sock *sk);
464 	void	(*tcp_error_report)(struct sock *sk);
465 
466 	struct	rcu_head rcu;
467 };
468 
469 static inline struct mptcp_subflow_context *
470 mptcp_subflow_ctx(const struct sock *sk)
471 {
472 	struct inet_connection_sock *icsk = inet_csk(sk);
473 
474 	/* Use RCU on icsk_ulp_data only for sock diag code */
475 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
476 }
477 
478 static inline struct sock *
479 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
480 {
481 	return subflow->tcp_sock;
482 }
483 
484 static inline u64
485 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
486 {
487 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
488 		      subflow->ssn_offset -
489 		      subflow->map_subflow_seq;
490 }
491 
492 static inline u64
493 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
494 {
495 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
496 }
497 
498 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
499 					     struct mptcp_subflow_context *subflow)
500 {
501 	sock_hold(mptcp_subflow_tcp_sock(subflow));
502 	spin_lock_bh(&msk->join_list_lock);
503 	list_add_tail(&subflow->node, &msk->join_list);
504 	spin_unlock_bh(&msk->join_list_lock);
505 }
506 
507 void mptcp_subflow_process_delegated(struct sock *ssk);
508 
509 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
510 {
511 	struct mptcp_delegated_action *delegated;
512 	bool schedule;
513 
514 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
515 	 * ensures the below list check sees list updates done prior to status
516 	 * bit changes
517 	 */
518 	if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
519 		/* still on delegated list from previous scheduling */
520 		if (!list_empty(&subflow->delegated_node))
521 			return;
522 
523 		/* the caller held the subflow bh socket lock */
524 		lockdep_assert_in_softirq();
525 
526 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
527 		schedule = list_empty(&delegated->head);
528 		list_add_tail(&subflow->delegated_node, &delegated->head);
529 		sock_hold(mptcp_subflow_tcp_sock(subflow));
530 		if (schedule)
531 			napi_schedule(&delegated->napi);
532 	}
533 }
534 
535 static inline struct mptcp_subflow_context *
536 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
537 {
538 	struct mptcp_subflow_context *ret;
539 
540 	if (list_empty(&delegated->head))
541 		return NULL;
542 
543 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
544 	list_del_init(&ret->delegated_node);
545 	return ret;
546 }
547 
548 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
549 {
550 	return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
551 }
552 
553 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
554 {
555 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
556 	 * touching the status bit
557 	 */
558 	smp_wmb();
559 	clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
560 }
561 
562 int mptcp_is_enabled(const struct net *net);
563 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
564 int mptcp_is_checksum_enabled(const struct net *net);
565 int mptcp_allow_join_id0(const struct net *net);
566 unsigned int mptcp_stale_loss_cnt(const struct net *net);
567 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
568 				     struct mptcp_options_received *mp_opt);
569 bool __mptcp_retransmit_pending_data(struct sock *sk);
570 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
571 bool mptcp_subflow_data_available(struct sock *sk);
572 void __init mptcp_subflow_init(void);
573 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
574 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
575 		     struct mptcp_subflow_context *subflow);
576 void mptcp_subflow_send_ack(struct sock *ssk);
577 void mptcp_subflow_reset(struct sock *ssk);
578 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
579 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
580 
581 /* called with sk socket lock held */
582 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
583 			    const struct mptcp_addr_info *remote);
584 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
585 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
586 			 struct sockaddr_storage *addr,
587 			 unsigned short family);
588 
589 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
590 {
591 	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
592 
593 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
594 	if (subflow->request_join && !subflow->fully_established)
595 		return false;
596 
597 	/* only send if our side has not closed yet */
598 	return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
599 }
600 
601 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
602 
603 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
604 
605 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
606 					      struct mptcp_subflow_context *ctx)
607 {
608 	sk->sk_data_ready = ctx->tcp_data_ready;
609 	sk->sk_state_change = ctx->tcp_state_change;
610 	sk->sk_write_space = ctx->tcp_write_space;
611 	sk->sk_error_report = ctx->tcp_error_report;
612 
613 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
614 }
615 
616 static inline bool mptcp_has_another_subflow(struct sock *ssk)
617 {
618 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp;
619 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
620 
621 	mptcp_for_each_subflow(msk, tmp) {
622 		if (tmp != subflow)
623 			return true;
624 	}
625 
626 	return false;
627 }
628 
629 void __init mptcp_proto_init(void);
630 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
631 int __init mptcp_proto_v6_init(void);
632 #endif
633 
634 struct sock *mptcp_sk_clone(const struct sock *sk,
635 			    const struct mptcp_options_received *mp_opt,
636 			    struct request_sock *req);
637 void mptcp_get_options(const struct sock *sk,
638 		       const struct sk_buff *skb,
639 		       struct mptcp_options_received *mp_opt);
640 
641 void mptcp_finish_connect(struct sock *sk);
642 void __mptcp_set_connected(struct sock *sk);
643 static inline bool mptcp_is_fully_established(struct sock *sk)
644 {
645 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
646 	       READ_ONCE(mptcp_sk(sk)->fully_established);
647 }
648 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
649 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
650 bool mptcp_finish_join(struct sock *sk);
651 bool mptcp_schedule_work(struct sock *sk);
652 int mptcp_setsockopt(struct sock *sk, int level, int optname,
653 		     sockptr_t optval, unsigned int optlen);
654 int mptcp_getsockopt(struct sock *sk, int level, int optname,
655 		     char __user *optval, int __user *option);
656 
657 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
658 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
659 {
660 	if (use_64bit)
661 		return cur_seq;
662 
663 	return __mptcp_expand_seq(old_seq, cur_seq);
664 }
665 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
666 void __mptcp_data_acked(struct sock *sk);
667 void __mptcp_error_report(struct sock *sk);
668 void mptcp_subflow_eof(struct sock *sk);
669 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
670 void __mptcp_flush_join_list(struct mptcp_sock *msk);
671 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
672 {
673 	return READ_ONCE(msk->snd_data_fin_enable) &&
674 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
675 }
676 
677 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
678 {
679 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
680 		return false;
681 
682 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
683 	return true;
684 }
685 
686 static inline void mptcp_write_space(struct sock *sk)
687 {
688 	if (sk_stream_is_writeable(sk)) {
689 		/* pairs with memory barrier in mptcp_poll */
690 		smp_mb();
691 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
692 			sk_stream_write_space(sk);
693 	}
694 }
695 
696 void mptcp_destroy_common(struct mptcp_sock *msk);
697 
698 #define MPTCP_TOKEN_MAX_RETRIES	4
699 
700 void __init mptcp_token_init(void);
701 static inline void mptcp_token_init_request(struct request_sock *req)
702 {
703 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
704 }
705 
706 int mptcp_token_new_request(struct request_sock *req);
707 void mptcp_token_destroy_request(struct request_sock *req);
708 int mptcp_token_new_connect(struct sock *sk);
709 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
710 			struct mptcp_sock *msk);
711 bool mptcp_token_exists(u32 token);
712 struct mptcp_sock *mptcp_token_get_sock(u32 token);
713 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
714 					 long *s_num);
715 void mptcp_token_destroy(struct mptcp_sock *msk);
716 
717 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
718 
719 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
720 
721 void __init mptcp_pm_init(void);
722 void mptcp_pm_data_init(struct mptcp_sock *msk);
723 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
724 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
725 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
726 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
727 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
728 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
729 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
730 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
731 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
732 				const struct mptcp_addr_info *addr);
733 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
734 			      struct mptcp_addr_info *addr);
735 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
736 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
737 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
738 			       const struct mptcp_rm_list *rm_list);
739 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
740 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
741 				 struct mptcp_addr_info *addr,
742 				 u8 bkup);
743 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
744 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
745 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
746 struct mptcp_pm_add_entry *
747 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
748 		       struct mptcp_addr_info *addr, bool check_id);
749 struct mptcp_pm_add_entry *
750 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
751 				struct mptcp_addr_info *addr);
752 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id,
753 					 u8 *flags, int *ifindex);
754 
755 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
756 			   const struct mptcp_addr_info *addr,
757 			   bool echo);
758 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
759 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
760 
761 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
762 		 const struct sock *ssk, gfp_t gfp);
763 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
764 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
765 
766 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
767 {
768 	return READ_ONCE(msk->pm.addr_signal) &
769 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
770 }
771 
772 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
773 {
774 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
775 }
776 
777 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
778 {
779 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
780 }
781 
782 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
783 {
784 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
785 }
786 
787 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
788 {
789 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
790 
791 	if (family == AF_INET6)
792 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
793 	if (!echo)
794 		len += MPTCPOPT_THMAC_LEN;
795 	/* account for 2 trailing 'nop' options */
796 	if (port)
797 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
798 
799 	return len;
800 }
801 
802 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
803 {
804 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
805 		return -EINVAL;
806 
807 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
808 }
809 
810 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
811 			      unsigned int opt_size, unsigned int remaining,
812 			      struct mptcp_addr_info *addr, bool *echo,
813 			      bool *port, bool *drop_other_suboptions);
814 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
815 			     struct mptcp_rm_list *rm_list);
816 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
817 
818 void __init mptcp_pm_nl_init(void);
819 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
820 void mptcp_pm_nl_work(struct mptcp_sock *msk);
821 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
822 				     const struct mptcp_rm_list *rm_list);
823 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
824 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
825 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
826 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
827 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
828 
829 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
830 void mptcp_sockopt_sync_all(struct mptcp_sock *msk);
831 
832 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
833 {
834 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
835 }
836 
837 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
838 
839 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
840 {
841 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
842 }
843 
844 static inline bool mptcp_check_fallback(const struct sock *sk)
845 {
846 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
847 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
848 
849 	return __mptcp_check_fallback(msk);
850 }
851 
852 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
853 {
854 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
855 		pr_debug("TCP fallback already done (msk=%p)", msk);
856 		return;
857 	}
858 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
859 }
860 
861 static inline void mptcp_do_fallback(struct sock *sk)
862 {
863 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
864 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
865 
866 	__mptcp_do_fallback(msk);
867 }
868 
869 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
870 
871 static inline bool subflow_simultaneous_connect(struct sock *sk)
872 {
873 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
874 	struct sock *parent = subflow->conn;
875 
876 	return sk->sk_state == TCP_ESTABLISHED &&
877 	       !mptcp_sk(parent)->pm.server_side &&
878 	       !subflow->conn_finished;
879 }
880 
881 #ifdef CONFIG_SYN_COOKIES
882 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
883 				       struct sk_buff *skb);
884 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
885 					struct sk_buff *skb);
886 void __init mptcp_join_cookie_init(void);
887 #else
888 static inline void
889 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
890 				  struct sk_buff *skb) {}
891 static inline bool
892 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
893 				   struct sk_buff *skb)
894 {
895 	return false;
896 }
897 
898 static inline void mptcp_join_cookie_init(void) {}
899 #endif
900 
901 #endif /* __MPTCP_PROTOCOL_H */
902