xref: /openbmc/linux/net/mptcp/protocol.h (revision 27e45f2e)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3  *
4  * Copyright (c) 2017 - 2019, Intel Corporation.
5  */
6 
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9 
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 #include <net/genetlink.h>
15 
16 #define MPTCP_SUPPORTED_VERSION	1
17 
18 /* MPTCP option bits */
19 #define OPTION_MPTCP_MPC_SYN	BIT(0)
20 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
21 #define OPTION_MPTCP_MPC_ACK	BIT(2)
22 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
23 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
24 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
25 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
26 #define OPTION_MPTCP_RM_ADDR	BIT(7)
27 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
28 #define OPTION_MPTCP_PRIO	BIT(9)
29 #define OPTION_MPTCP_RST	BIT(10)
30 #define OPTION_MPTCP_DSS	BIT(11)
31 #define OPTION_MPTCP_FAIL	BIT(12)
32 
33 #define OPTION_MPTCP_CSUMREQD	BIT(13)
34 
35 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
36 				 OPTION_MPTCP_MPC_ACK)
37 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
38 				 OPTION_MPTCP_MPJ_ACK)
39 
40 /* MPTCP option subtypes */
41 #define MPTCPOPT_MP_CAPABLE	0
42 #define MPTCPOPT_MP_JOIN	1
43 #define MPTCPOPT_DSS		2
44 #define MPTCPOPT_ADD_ADDR	3
45 #define MPTCPOPT_RM_ADDR	4
46 #define MPTCPOPT_MP_PRIO	5
47 #define MPTCPOPT_MP_FAIL	6
48 #define MPTCPOPT_MP_FASTCLOSE	7
49 #define MPTCPOPT_RST		8
50 
51 /* MPTCP suboption lengths */
52 #define TCPOLEN_MPTCP_MPC_SYN		4
53 #define TCPOLEN_MPTCP_MPC_SYNACK	12
54 #define TCPOLEN_MPTCP_MPC_ACK		20
55 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
56 #define TCPOLEN_MPTCP_MPJ_SYN		12
57 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
58 #define TCPOLEN_MPTCP_MPJ_ACK		24
59 #define TCPOLEN_MPTCP_DSS_BASE		4
60 #define TCPOLEN_MPTCP_DSS_ACK32		4
61 #define TCPOLEN_MPTCP_DSS_ACK64		8
62 #define TCPOLEN_MPTCP_DSS_MAP32		10
63 #define TCPOLEN_MPTCP_DSS_MAP64		14
64 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
65 #define TCPOLEN_MPTCP_ADD_ADDR		16
66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
69 #define TCPOLEN_MPTCP_ADD_ADDR6		28
70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
73 #define TCPOLEN_MPTCP_PORT_LEN		2
74 #define TCPOLEN_MPTCP_PORT_ALIGN	2
75 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
76 #define TCPOLEN_MPTCP_PRIO		3
77 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
78 #define TCPOLEN_MPTCP_FASTCLOSE		12
79 #define TCPOLEN_MPTCP_RST		4
80 #define TCPOLEN_MPTCP_FAIL		12
81 
82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
83 
84 /* MPTCP MP_JOIN flags */
85 #define MPTCPOPT_BACKUP		BIT(0)
86 #define MPTCPOPT_THMAC_LEN	8
87 
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK	(0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
94 #define MPTCP_CAP_FLAG_MASK	(0x1F)
95 
96 /* MPTCP DSS flags */
97 #define MPTCP_DSS_DATA_FIN	BIT(4)
98 #define MPTCP_DSS_DSN64		BIT(3)
99 #define MPTCP_DSS_HAS_MAP	BIT(2)
100 #define MPTCP_DSS_ACK64		BIT(1)
101 #define MPTCP_DSS_HAS_ACK	BIT(0)
102 #define MPTCP_DSS_FLAG_MASK	(0x1F)
103 
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO		BIT(0)
106 
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP		BIT(0)
109 
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT	BIT(0)
112 
113 /* MPTCP socket atomic flags */
114 #define MPTCP_NOSPACE		1
115 #define MPTCP_WORK_RTX		2
116 #define MPTCP_WORK_EOF		3
117 #define MPTCP_FALLBACK_DONE	4
118 #define MPTCP_WORK_CLOSE_SUBFLOW 5
119 
120 /* MPTCP socket release cb flags */
121 #define MPTCP_PUSH_PENDING	1
122 #define MPTCP_CLEAN_UNA		2
123 #define MPTCP_ERROR_REPORT	3
124 #define MPTCP_RETRANSMIT	4
125 #define MPTCP_FLUSH_JOIN_LIST	5
126 #define MPTCP_CONNECTED		6
127 #define MPTCP_RESET_SCHEDULER	7
128 
129 struct mptcp_skb_cb {
130 	u64 map_seq;
131 	u64 end_seq;
132 	u32 offset;
133 	u8  has_rxtstamp:1;
134 };
135 
136 #define MPTCP_SKB_CB(__skb)	((struct mptcp_skb_cb *)&((__skb)->cb[0]))
137 
138 static inline bool before64(__u64 seq1, __u64 seq2)
139 {
140 	return (__s64)(seq1 - seq2) < 0;
141 }
142 
143 #define after64(seq2, seq1)	before64(seq1, seq2)
144 
145 struct mptcp_options_received {
146 	u64	sndr_key;
147 	u64	rcvr_key;
148 	u64	data_ack;
149 	u64	data_seq;
150 	u32	subflow_seq;
151 	u16	data_len;
152 	__sum16	csum;
153 	u16	suboptions;
154 	u32	token;
155 	u32	nonce;
156 	u16	use_map:1,
157 		dsn64:1,
158 		data_fin:1,
159 		use_ack:1,
160 		ack64:1,
161 		mpc_map:1,
162 		reset_reason:4,
163 		reset_transient:1,
164 		echo:1,
165 		backup:1,
166 		deny_join_id0:1,
167 		__unused:2;
168 	u8	join_id;
169 	u64	thmac;
170 	u8	hmac[MPTCPOPT_HMAC_LEN];
171 	struct mptcp_addr_info addr;
172 	struct mptcp_rm_list rm_list;
173 	u64	ahmac;
174 	u64	fail_seq;
175 };
176 
177 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
178 {
179 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
180 		     ((nib & 0xF) << 8) | field);
181 }
182 
183 enum mptcp_pm_status {
184 	MPTCP_PM_ADD_ADDR_RECEIVED,
185 	MPTCP_PM_ADD_ADDR_SEND_ACK,
186 	MPTCP_PM_RM_ADDR_RECEIVED,
187 	MPTCP_PM_ESTABLISHED,
188 	MPTCP_PM_SUBFLOW_ESTABLISHED,
189 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
190 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
191 					 * accounted int id_avail_bitmap
192 					 */
193 };
194 
195 enum mptcp_pm_type {
196 	MPTCP_PM_TYPE_KERNEL = 0,
197 	MPTCP_PM_TYPE_USERSPACE,
198 
199 	__MPTCP_PM_TYPE_NR,
200 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
201 };
202 
203 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
204 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
205 
206 enum mptcp_addr_signal_status {
207 	MPTCP_ADD_ADDR_SIGNAL,
208 	MPTCP_ADD_ADDR_ECHO,
209 	MPTCP_RM_ADDR_SIGNAL,
210 };
211 
212 /* max value of mptcp_addr_info.id */
213 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
214 
215 struct mptcp_pm_data {
216 	struct mptcp_addr_info local;
217 	struct mptcp_addr_info remote;
218 	struct list_head anno_list;
219 	struct list_head userspace_pm_local_addr_list;
220 
221 	spinlock_t	lock;		/*protects the whole PM data */
222 
223 	u8		addr_signal;
224 	bool		server_side;
225 	bool		work_pending;
226 	bool		accept_addr;
227 	bool		accept_subflow;
228 	bool		remote_deny_join_id0;
229 	u8		add_addr_signaled;
230 	u8		add_addr_accepted;
231 	u8		local_addr_used;
232 	u8		pm_type;
233 	u8		subflows;
234 	u8		status;
235 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
236 	struct mptcp_rm_list rm_list_tx;
237 	struct mptcp_rm_list rm_list_rx;
238 };
239 
240 struct mptcp_pm_addr_entry {
241 	struct list_head	list;
242 	struct mptcp_addr_info	addr;
243 	u8			flags;
244 	int			ifindex;
245 	struct socket		*lsk;
246 };
247 
248 struct mptcp_data_frag {
249 	struct list_head list;
250 	u64 data_seq;
251 	u16 data_len;
252 	u16 offset;
253 	u16 overhead;
254 	u16 already_sent;
255 	struct page *page;
256 };
257 
258 /* MPTCP connection sock */
259 struct mptcp_sock {
260 	/* inet_connection_sock must be the first member */
261 	struct inet_connection_sock sk;
262 	u64		local_key;
263 	u64		remote_key;
264 	u64		write_seq;
265 	u64		snd_nxt;
266 	u64		ack_seq;
267 	atomic64_t	rcv_wnd_sent;
268 	u64		rcv_data_fin_seq;
269 	int		rmem_fwd_alloc;
270 	struct sock	*last_snd;
271 	int		snd_burst;
272 	int		old_wspace;
273 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
274 						 * recovery related fields are under data_lock
275 						 * protection
276 						 */
277 	u64		snd_una;
278 	u64		wnd_end;
279 	unsigned long	timer_ival;
280 	u32		token;
281 	int		rmem_released;
282 	unsigned long	flags;
283 	unsigned long	cb_flags;
284 	unsigned long	push_pending;
285 	bool		recovery;		/* closing subflow write queue reinjected */
286 	bool		can_ack;
287 	bool		fully_established;
288 	bool		rcv_data_fin;
289 	bool		snd_data_fin_enable;
290 	bool		rcv_fastclose;
291 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
292 	bool		csum_enabled;
293 	bool		allow_infinite_fallback;
294 	u8		mpc_endpoint_id;
295 	u8		recvmsg_inq:1,
296 			cork:1,
297 			nodelay:1,
298 			is_sendmsg:1;
299 	int		connect_flags;
300 	struct work_struct work;
301 	struct sk_buff  *ooo_last_skb;
302 	struct rb_root  out_of_order_queue;
303 	struct sk_buff_head receive_queue;
304 	struct list_head conn_list;
305 	struct list_head rtx_queue;
306 	struct mptcp_data_frag *first_pending;
307 	struct list_head join_list;
308 	struct socket	*subflow; /* outgoing connect/listener/!mp_capable */
309 	struct sock	*first;
310 	struct mptcp_pm_data	pm;
311 	struct {
312 		u32	space;	/* bytes copied in last measurement window */
313 		u32	copied; /* bytes copied in this measurement window */
314 		u64	time;	/* start time of measurement window */
315 		u64	rtt_us; /* last maximum rtt of subflows */
316 	} rcvq_space;
317 
318 	u32 setsockopt_seq;
319 	char		ca_name[TCP_CA_NAME_MAX];
320 	struct mptcp_sock	*dl_next;
321 };
322 
323 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
324 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
325 
326 #define mptcp_for_each_subflow(__msk, __subflow)			\
327 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
328 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
329 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
330 
331 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
332 {
333 	sock_owned_by_me((const struct sock *)msk);
334 }
335 
336 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
337 {
338 	return (struct mptcp_sock *)sk;
339 }
340 
341 /* the msk socket don't use the backlog, also account for the bulk
342  * free memory
343  */
344 static inline int __mptcp_rmem(const struct sock *sk)
345 {
346 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
347 }
348 
349 static inline int __mptcp_space(const struct sock *sk)
350 {
351 	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
352 }
353 
354 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
355 {
356 	const struct mptcp_sock *msk = mptcp_sk(sk);
357 
358 	return READ_ONCE(msk->first_pending);
359 }
360 
361 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
362 {
363 	struct mptcp_sock *msk = mptcp_sk(sk);
364 	struct mptcp_data_frag *cur;
365 
366 	cur = msk->first_pending;
367 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
368 						     list_next_entry(cur, list);
369 }
370 
371 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
372 {
373 	struct mptcp_sock *msk = mptcp_sk(sk);
374 
375 	if (!msk->first_pending)
376 		return NULL;
377 
378 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
379 		return NULL;
380 
381 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
382 }
383 
384 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
385 {
386 	struct mptcp_sock *msk = mptcp_sk(sk);
387 
388 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
389 		return NULL;
390 
391 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
392 }
393 
394 struct csum_pseudo_header {
395 	__be64 data_seq;
396 	__be32 subflow_seq;
397 	__be16 data_len;
398 	__sum16 csum;
399 };
400 
401 struct mptcp_subflow_request_sock {
402 	struct	tcp_request_sock sk;
403 	u16	mp_capable : 1,
404 		mp_join : 1,
405 		backup : 1,
406 		csum_reqd : 1,
407 		allow_join_id0 : 1;
408 	u8	local_id;
409 	u8	remote_id;
410 	u64	local_key;
411 	u64	idsn;
412 	u32	token;
413 	u32	ssn_offset;
414 	u64	thmac;
415 	u32	local_nonce;
416 	u32	remote_nonce;
417 	struct mptcp_sock	*msk;
418 	struct hlist_nulls_node token_node;
419 };
420 
421 static inline struct mptcp_subflow_request_sock *
422 mptcp_subflow_rsk(const struct request_sock *rsk)
423 {
424 	return (struct mptcp_subflow_request_sock *)rsk;
425 }
426 
427 enum mptcp_data_avail {
428 	MPTCP_SUBFLOW_NODATA,
429 	MPTCP_SUBFLOW_DATA_AVAIL,
430 };
431 
432 struct mptcp_delegated_action {
433 	struct napi_struct napi;
434 	struct list_head head;
435 };
436 
437 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
438 
439 #define MPTCP_DELEGATE_SEND		0
440 #define MPTCP_DELEGATE_ACK		1
441 
442 /* MPTCP subflow context */
443 struct mptcp_subflow_context {
444 	struct	list_head node;/* conn_list of subflows */
445 
446 	struct_group(reset,
447 
448 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
449 	u64	local_key;
450 	u64	remote_key;
451 	u64	idsn;
452 	u64	map_seq;
453 	u32	snd_isn;
454 	u32	token;
455 	u32	rel_write_seq;
456 	u32	map_subflow_seq;
457 	u32	ssn_offset;
458 	u32	map_data_len;
459 	__wsum	map_data_csum;
460 	u32	map_csum_len;
461 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
462 		request_join : 1,   /* send MP_JOIN */
463 		request_bkup : 1,
464 		mp_capable : 1,	    /* remote is MPTCP capable */
465 		mp_join : 1,	    /* remote is JOINing */
466 		fully_established : 1,	    /* path validated */
467 		pm_notified : 1,    /* PM hook called for established status */
468 		conn_finished : 1,
469 		map_valid : 1,
470 		map_csum_reqd : 1,
471 		map_data_fin : 1,
472 		mpc_map : 1,
473 		backup : 1,
474 		send_mp_prio : 1,
475 		send_mp_fail : 1,
476 		send_fastclose : 1,
477 		send_infinite_map : 1,
478 		rx_eof : 1,
479 		remote_key_valid : 1,        /* received the peer key from */
480 		disposable : 1,	    /* ctx can be free at ulp release time */
481 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
482 		local_id_valid : 1, /* local_id is correctly initialized */
483 		valid_csum_seen : 1,        /* at least one csum validated */
484 		is_mptfo : 1,	    /* subflow is doing TFO */
485 		__unused : 8;
486 	enum mptcp_data_avail data_avail;
487 	u32	remote_nonce;
488 	u64	thmac;
489 	u32	local_nonce;
490 	u32	remote_token;
491 	union {
492 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
493 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
494 	};
495 	u8	local_id;
496 	u8	remote_id;
497 	u8	reset_seen:1;
498 	u8	reset_transient:1;
499 	u8	reset_reason:4;
500 	u8	stale_count;
501 
502 	long	delegated_status;
503 	unsigned long	fail_tout;
504 
505 	);
506 
507 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
508 
509 	u32	setsockopt_seq;
510 	u32	stale_rcv_tstamp;
511 
512 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
513 	struct	sock *conn;	    /* parent mptcp_sock */
514 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
515 	void	(*tcp_state_change)(struct sock *sk);
516 	void	(*tcp_error_report)(struct sock *sk);
517 
518 	struct	rcu_head rcu;
519 };
520 
521 static inline struct mptcp_subflow_context *
522 mptcp_subflow_ctx(const struct sock *sk)
523 {
524 	struct inet_connection_sock *icsk = inet_csk(sk);
525 
526 	/* Use RCU on icsk_ulp_data only for sock diag code */
527 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
528 }
529 
530 static inline struct sock *
531 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
532 {
533 	return subflow->tcp_sock;
534 }
535 
536 static inline void
537 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
538 {
539 	memset(&subflow->reset, 0, sizeof(subflow->reset));
540 	subflow->request_mptcp = 1;
541 }
542 
543 static inline u64
544 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
545 {
546 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
547 		      subflow->ssn_offset -
548 		      subflow->map_subflow_seq;
549 }
550 
551 static inline u64
552 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
553 {
554 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
555 }
556 
557 void mptcp_subflow_process_delegated(struct sock *ssk);
558 
559 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
560 {
561 	struct mptcp_delegated_action *delegated;
562 	bool schedule;
563 
564 	/* the caller held the subflow bh socket lock */
565 	lockdep_assert_in_softirq();
566 
567 	/* The implied barrier pairs with mptcp_subflow_delegated_done(), and
568 	 * ensures the below list check sees list updates done prior to status
569 	 * bit changes
570 	 */
571 	if (!test_and_set_bit(action, &subflow->delegated_status)) {
572 		/* still on delegated list from previous scheduling */
573 		if (!list_empty(&subflow->delegated_node))
574 			return;
575 
576 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
577 		schedule = list_empty(&delegated->head);
578 		list_add_tail(&subflow->delegated_node, &delegated->head);
579 		sock_hold(mptcp_subflow_tcp_sock(subflow));
580 		if (schedule)
581 			napi_schedule(&delegated->napi);
582 	}
583 }
584 
585 static inline struct mptcp_subflow_context *
586 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
587 {
588 	struct mptcp_subflow_context *ret;
589 
590 	if (list_empty(&delegated->head))
591 		return NULL;
592 
593 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
594 	list_del_init(&ret->delegated_node);
595 	return ret;
596 }
597 
598 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
599 {
600 	return !!READ_ONCE(subflow->delegated_status);
601 }
602 
603 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
604 {
605 	/* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
606 	 * touching the status bit
607 	 */
608 	smp_wmb();
609 	clear_bit(action, &subflow->delegated_status);
610 }
611 
612 int mptcp_is_enabled(const struct net *net);
613 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
614 int mptcp_is_checksum_enabled(const struct net *net);
615 int mptcp_allow_join_id0(const struct net *net);
616 unsigned int mptcp_stale_loss_cnt(const struct net *net);
617 int mptcp_get_pm_type(const struct net *net);
618 void mptcp_copy_inaddrs(struct sock *msk, const struct sock *ssk);
619 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
620 				     const struct mptcp_options_received *mp_opt);
621 bool __mptcp_retransmit_pending_data(struct sock *sk);
622 void mptcp_check_and_set_pending(struct sock *sk);
623 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
624 bool mptcp_subflow_data_available(struct sock *sk);
625 void __init mptcp_subflow_init(void);
626 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
627 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
628 		     struct mptcp_subflow_context *subflow);
629 void __mptcp_subflow_send_ack(struct sock *ssk);
630 void mptcp_subflow_reset(struct sock *ssk);
631 void mptcp_subflow_queue_clean(struct sock *ssk);
632 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
633 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
634 bool __mptcp_close(struct sock *sk, long timeout);
635 void mptcp_cancel_work(struct sock *sk);
636 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
637 
638 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
639 			   const struct mptcp_addr_info *b, bool use_port);
640 
641 /* called with sk socket lock held */
642 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
643 			    const struct mptcp_addr_info *remote);
644 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
645 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
646 			 struct sockaddr_storage *addr,
647 			 unsigned short family);
648 
649 static inline bool __tcp_can_send(const struct sock *ssk)
650 {
651 	/* only send if our side has not closed yet */
652 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
653 }
654 
655 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
656 {
657 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
658 	if (subflow->request_join && !subflow->fully_established)
659 		return false;
660 
661 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
662 }
663 
664 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
665 
666 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
667 
668 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
669 					      struct mptcp_subflow_context *ctx)
670 {
671 	sk->sk_data_ready = sock_def_readable;
672 	sk->sk_state_change = ctx->tcp_state_change;
673 	sk->sk_write_space = sk_stream_write_space;
674 	sk->sk_error_report = ctx->tcp_error_report;
675 
676 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
677 }
678 
679 void __init mptcp_proto_init(void);
680 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
681 int __init mptcp_proto_v6_init(void);
682 #endif
683 
684 struct sock *mptcp_sk_clone(const struct sock *sk,
685 			    const struct mptcp_options_received *mp_opt,
686 			    struct request_sock *req);
687 void mptcp_get_options(const struct sk_buff *skb,
688 		       struct mptcp_options_received *mp_opt);
689 
690 void mptcp_finish_connect(struct sock *sk);
691 void __mptcp_set_connected(struct sock *sk);
692 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
693 static inline bool mptcp_is_fully_established(struct sock *sk)
694 {
695 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
696 	       READ_ONCE(mptcp_sk(sk)->fully_established);
697 }
698 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
699 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
700 bool mptcp_finish_join(struct sock *sk);
701 bool mptcp_schedule_work(struct sock *sk);
702 int mptcp_setsockopt(struct sock *sk, int level, int optname,
703 		     sockptr_t optval, unsigned int optlen);
704 int mptcp_getsockopt(struct sock *sk, int level, int optname,
705 		     char __user *optval, int __user *option);
706 
707 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
708 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
709 {
710 	if (use_64bit)
711 		return cur_seq;
712 
713 	return __mptcp_expand_seq(old_seq, cur_seq);
714 }
715 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
716 void __mptcp_data_acked(struct sock *sk);
717 void __mptcp_error_report(struct sock *sk);
718 void mptcp_subflow_eof(struct sock *sk);
719 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
720 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
721 {
722 	return READ_ONCE(msk->snd_data_fin_enable) &&
723 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
724 }
725 
726 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
727 {
728 	if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
729 		return false;
730 
731 	WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
732 	return true;
733 }
734 
735 static inline void mptcp_write_space(struct sock *sk)
736 {
737 	if (sk_stream_is_writeable(sk)) {
738 		/* pairs with memory barrier in mptcp_poll */
739 		smp_mb();
740 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
741 			sk_stream_write_space(sk);
742 	}
743 }
744 
745 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
746 
747 #define MPTCP_TOKEN_MAX_RETRIES	4
748 
749 void __init mptcp_token_init(void);
750 static inline void mptcp_token_init_request(struct request_sock *req)
751 {
752 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
753 }
754 
755 int mptcp_token_new_request(struct request_sock *req);
756 void mptcp_token_destroy_request(struct request_sock *req);
757 int mptcp_token_new_connect(struct sock *sk);
758 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
759 			struct mptcp_sock *msk);
760 bool mptcp_token_exists(u32 token);
761 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
762 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
763 					 long *s_num);
764 void mptcp_token_destroy(struct mptcp_sock *msk);
765 
766 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
767 
768 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
769 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
770 
771 void __init mptcp_pm_init(void);
772 void mptcp_pm_data_init(struct mptcp_sock *msk);
773 void mptcp_pm_data_reset(struct mptcp_sock *msk);
774 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
775 			struct mptcp_addr_info *addr);
776 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
777 			 bool require_family,
778 			 struct mptcp_pm_addr_entry *entry);
779 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
780 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
781 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
782 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
783 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
784 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
785 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
786 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
787 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
788 				 const struct mptcp_subflow_context *subflow);
789 void mptcp_pm_add_addr_received(const struct sock *ssk,
790 				const struct mptcp_addr_info *addr);
791 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
792 			      const struct mptcp_addr_info *addr);
793 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
794 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
795 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
796 			       const struct mptcp_rm_list *rm_list);
797 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
798 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
799 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
800 				 struct mptcp_addr_info *addr,
801 				 struct mptcp_addr_info *rem,
802 				 u8 bkup);
803 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
804 			      const struct mptcp_pm_addr_entry *entry);
805 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
806 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
807 struct mptcp_pm_add_entry *
808 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
809 		       const struct mptcp_addr_info *addr, bool check_id);
810 struct mptcp_pm_add_entry *
811 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
812 				const struct mptcp_addr_info *addr);
813 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
814 					 unsigned int id,
815 					 u8 *flags, int *ifindex);
816 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
817 						   unsigned int id,
818 						   u8 *flags, int *ifindex);
819 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
820 				 struct mptcp_pm_addr_entry *loc,
821 				 struct mptcp_pm_addr_entry *rem, u8 bkup);
822 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
823 			   const struct mptcp_addr_info *addr,
824 			   bool echo);
825 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
826 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
827 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
828 					struct list_head *rm_list);
829 
830 int mptcp_userspace_pm_append_new_local_addr(struct mptcp_sock *msk,
831 					     struct mptcp_pm_addr_entry *entry);
832 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
833 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
834 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
835 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
836 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
837 
838 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
839 		 const struct sock *ssk, gfp_t gfp);
840 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
841 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
842 void mptcp_event_pm_listener(const struct sock *ssk,
843 			     enum mptcp_event_type event);
844 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
845 
846 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
847 				   const struct mptcp_options_received *mp_opt);
848 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
849 					      struct request_sock *req);
850 
851 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
852 {
853 	return READ_ONCE(msk->pm.addr_signal) &
854 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
855 }
856 
857 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
858 {
859 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
860 }
861 
862 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
863 {
864 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
865 }
866 
867 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
868 {
869 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
870 }
871 
872 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
873 {
874 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
875 }
876 
877 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
878 {
879 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
880 }
881 
882 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
883 {
884 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
885 
886 	if (family == AF_INET6)
887 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
888 	if (!echo)
889 		len += MPTCPOPT_THMAC_LEN;
890 	/* account for 2 trailing 'nop' options */
891 	if (port)
892 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
893 
894 	return len;
895 }
896 
897 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
898 {
899 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
900 		return -EINVAL;
901 
902 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
903 }
904 
905 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
906 			      unsigned int opt_size, unsigned int remaining,
907 			      struct mptcp_addr_info *addr, bool *echo,
908 			      bool *drop_other_suboptions);
909 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
910 			     struct mptcp_rm_list *rm_list);
911 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
912 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
913 
914 void __init mptcp_pm_nl_init(void);
915 void mptcp_pm_nl_work(struct mptcp_sock *msk);
916 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
917 				     const struct mptcp_rm_list *rm_list);
918 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
919 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
920 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
921 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
922 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
923 
924 /* called under PM lock */
925 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
926 {
927 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
928 		WRITE_ONCE(msk->pm.accept_subflow, true);
929 }
930 
931 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
932 {
933 	spin_lock_bh(&msk->pm.lock);
934 	__mptcp_pm_close_subflow(msk);
935 	spin_unlock_bh(&msk->pm.lock);
936 }
937 
938 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
939 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
940 
941 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
942 {
943 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
944 }
945 
946 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
947 
948 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
949 {
950 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
951 }
952 
953 static inline bool mptcp_check_fallback(const struct sock *sk)
954 {
955 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
956 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
957 
958 	return __mptcp_check_fallback(msk);
959 }
960 
961 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
962 {
963 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
964 		pr_debug("TCP fallback already done (msk=%p)", msk);
965 		return;
966 	}
967 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
968 }
969 
970 static inline void mptcp_do_fallback(struct sock *ssk)
971 {
972 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
973 	struct sock *sk = subflow->conn;
974 	struct mptcp_sock *msk;
975 
976 	msk = mptcp_sk(sk);
977 	__mptcp_do_fallback(msk);
978 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
979 		gfp_t saved_allocation = ssk->sk_allocation;
980 
981 		/* we are in a atomic (BH) scope, override ssk default for data
982 		 * fin allocation
983 		 */
984 		ssk->sk_allocation = GFP_ATOMIC;
985 		ssk->sk_shutdown |= SEND_SHUTDOWN;
986 		tcp_shutdown(ssk, SEND_SHUTDOWN);
987 		ssk->sk_allocation = saved_allocation;
988 	}
989 }
990 
991 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
992 
993 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
994 {
995 	struct mptcp_ext *mpext;
996 
997 	mpext = skb ? mptcp_get_ext(skb) : NULL;
998 	if (mpext && mpext->infinite_map)
999 		return true;
1000 
1001 	return false;
1002 }
1003 
1004 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1005 {
1006 	return (subflow->request_mptcp || subflow->request_join);
1007 }
1008 
1009 static inline bool subflow_simultaneous_connect(struct sock *sk)
1010 {
1011 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1012 
1013 	return sk->sk_state == TCP_ESTABLISHED &&
1014 	       is_active_ssk(subflow) &&
1015 	       !subflow->conn_finished;
1016 }
1017 
1018 #ifdef CONFIG_SYN_COOKIES
1019 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1020 				       struct sk_buff *skb);
1021 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1022 					struct sk_buff *skb);
1023 void __init mptcp_join_cookie_init(void);
1024 #else
1025 static inline void
1026 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1027 				  struct sk_buff *skb) {}
1028 static inline bool
1029 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1030 				   struct sk_buff *skb)
1031 {
1032 	return false;
1033 }
1034 
1035 static inline void mptcp_join_cookie_init(void) {}
1036 #endif
1037 
1038 #endif /* __MPTCP_PROTOCOL_H */
1039