xref: /openbmc/linux/net/mptcp/protocol.h (revision 37d9305c)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3  *
4  * Copyright (c) 2017 - 2019, Intel Corporation.
5  */
6 
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9 
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 #include <net/genetlink.h>
15 
16 #define MPTCP_SUPPORTED_VERSION	1
17 
18 /* MPTCP option bits */
19 #define OPTION_MPTCP_MPC_SYN	BIT(0)
20 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
21 #define OPTION_MPTCP_MPC_ACK	BIT(2)
22 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
23 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
24 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
25 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
26 #define OPTION_MPTCP_RM_ADDR	BIT(7)
27 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
28 #define OPTION_MPTCP_PRIO	BIT(9)
29 #define OPTION_MPTCP_RST	BIT(10)
30 #define OPTION_MPTCP_DSS	BIT(11)
31 #define OPTION_MPTCP_FAIL	BIT(12)
32 
33 #define OPTION_MPTCP_CSUMREQD	BIT(13)
34 
35 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
36 				 OPTION_MPTCP_MPC_ACK)
37 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
38 				 OPTION_MPTCP_MPJ_ACK)
39 
40 /* MPTCP option subtypes */
41 #define MPTCPOPT_MP_CAPABLE	0
42 #define MPTCPOPT_MP_JOIN	1
43 #define MPTCPOPT_DSS		2
44 #define MPTCPOPT_ADD_ADDR	3
45 #define MPTCPOPT_RM_ADDR	4
46 #define MPTCPOPT_MP_PRIO	5
47 #define MPTCPOPT_MP_FAIL	6
48 #define MPTCPOPT_MP_FASTCLOSE	7
49 #define MPTCPOPT_RST		8
50 
51 /* MPTCP suboption lengths */
52 #define TCPOLEN_MPTCP_MPC_SYN		4
53 #define TCPOLEN_MPTCP_MPC_SYNACK	12
54 #define TCPOLEN_MPTCP_MPC_ACK		20
55 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
56 #define TCPOLEN_MPTCP_MPJ_SYN		12
57 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
58 #define TCPOLEN_MPTCP_MPJ_ACK		24
59 #define TCPOLEN_MPTCP_DSS_BASE		4
60 #define TCPOLEN_MPTCP_DSS_ACK32		4
61 #define TCPOLEN_MPTCP_DSS_ACK64		8
62 #define TCPOLEN_MPTCP_DSS_MAP32		10
63 #define TCPOLEN_MPTCP_DSS_MAP64		14
64 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
65 #define TCPOLEN_MPTCP_ADD_ADDR		16
66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
69 #define TCPOLEN_MPTCP_ADD_ADDR6		28
70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
73 #define TCPOLEN_MPTCP_PORT_LEN		2
74 #define TCPOLEN_MPTCP_PORT_ALIGN	2
75 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
76 #define TCPOLEN_MPTCP_PRIO		3
77 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
78 #define TCPOLEN_MPTCP_FASTCLOSE		12
79 #define TCPOLEN_MPTCP_RST		4
80 #define TCPOLEN_MPTCP_FAIL		12
81 
82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
83 
84 /* MPTCP MP_JOIN flags */
85 #define MPTCPOPT_BACKUP		BIT(0)
86 #define MPTCPOPT_THMAC_LEN	8
87 
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK	(0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
94 #define MPTCP_CAP_FLAG_MASK	(0x1F)
95 
96 /* MPTCP DSS flags */
97 #define MPTCP_DSS_DATA_FIN	BIT(4)
98 #define MPTCP_DSS_DSN64		BIT(3)
99 #define MPTCP_DSS_HAS_MAP	BIT(2)
100 #define MPTCP_DSS_ACK64		BIT(1)
101 #define MPTCP_DSS_HAS_ACK	BIT(0)
102 #define MPTCP_DSS_FLAG_MASK	(0x1F)
103 
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO		BIT(0)
106 
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP		BIT(0)
109 
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT	BIT(0)
112 
113 /* MPTCP socket atomic flags */
114 #define MPTCP_NOSPACE		1
115 #define MPTCP_WORK_RTX		2
116 #define MPTCP_FALLBACK_DONE	4
117 #define MPTCP_WORK_CLOSE_SUBFLOW 5
118 
119 /* MPTCP socket release cb flags */
120 #define MPTCP_PUSH_PENDING	1
121 #define MPTCP_CLEAN_UNA		2
122 #define MPTCP_ERROR_REPORT	3
123 #define MPTCP_RETRANSMIT	4
124 #define MPTCP_FLUSH_JOIN_LIST	5
125 #define MPTCP_SYNC_STATE	6
126 #define MPTCP_SYNC_SNDBUF	7
127 
128 struct mptcp_skb_cb {
129 	u64 map_seq;
130 	u64 end_seq;
131 	u32 offset;
132 	u8  has_rxtstamp:1;
133 };
134 
135 #define MPTCP_SKB_CB(__skb)	((struct mptcp_skb_cb *)&((__skb)->cb[0]))
136 
before64(__u64 seq1,__u64 seq2)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 
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
177 {
178 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 		     ((nib & 0xF) << 8) | field);
180 }
181 
182 enum mptcp_pm_status {
183 	MPTCP_PM_ADD_ADDR_RECEIVED,
184 	MPTCP_PM_ADD_ADDR_SEND_ACK,
185 	MPTCP_PM_RM_ADDR_RECEIVED,
186 	MPTCP_PM_ESTABLISHED,
187 	MPTCP_PM_SUBFLOW_ESTABLISHED,
188 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
189 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 					 * accounted int id_avail_bitmap
191 					 */
192 };
193 
194 enum mptcp_pm_type {
195 	MPTCP_PM_TYPE_KERNEL = 0,
196 	MPTCP_PM_TYPE_USERSPACE,
197 
198 	__MPTCP_PM_TYPE_NR,
199 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
200 };
201 
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
204 
205 enum mptcp_addr_signal_status {
206 	MPTCP_ADD_ADDR_SIGNAL,
207 	MPTCP_ADD_ADDR_ECHO,
208 	MPTCP_RM_ADDR_SIGNAL,
209 };
210 
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
213 
214 struct mptcp_pm_data {
215 	struct mptcp_addr_info local;
216 	struct mptcp_addr_info remote;
217 	struct list_head anno_list;
218 	struct list_head userspace_pm_local_addr_list;
219 
220 	spinlock_t	lock;		/*protects the whole PM data */
221 
222 	u8		addr_signal;
223 	bool		server_side;
224 	bool		work_pending;
225 	bool		accept_addr;
226 	bool		accept_subflow;
227 	bool		remote_deny_join_id0;
228 	u8		add_addr_signaled;
229 	u8		add_addr_accepted;
230 	u8		local_addr_used;
231 	u8		pm_type;
232 	u8		subflows;
233 	u8		status;
234 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 	struct mptcp_rm_list rm_list_tx;
236 	struct mptcp_rm_list rm_list_rx;
237 };
238 
239 struct mptcp_pm_addr_entry {
240 	struct list_head	list;
241 	struct mptcp_addr_info	addr;
242 	u8			flags;
243 	int			ifindex;
244 	struct socket		*lsk;
245 };
246 
247 struct mptcp_data_frag {
248 	struct list_head list;
249 	u64 data_seq;
250 	u16 data_len;
251 	u16 offset;
252 	u16 overhead;
253 	u16 already_sent;
254 	struct page *page;
255 };
256 
257 /* MPTCP connection sock */
258 struct mptcp_sock {
259 	/* inet_connection_sock must be the first member */
260 	struct inet_connection_sock sk;
261 	u64		local_key;
262 	u64		remote_key;
263 	u64		write_seq;
264 	u64		bytes_sent;
265 	u64		snd_nxt;
266 	u64		bytes_received;
267 	u64		ack_seq;
268 	atomic64_t	rcv_wnd_sent;
269 	u64		rcv_data_fin_seq;
270 	u64		bytes_retrans;
271 	u64		bytes_consumed;
272 	int		rmem_fwd_alloc;
273 	int		snd_burst;
274 	int		old_wspace;
275 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
276 						 * recovery related fields are under data_lock
277 						 * protection
278 						 */
279 	u64		bytes_acked;
280 	u64		snd_una;
281 	u64		wnd_end;
282 	unsigned long	timer_ival;
283 	u32		token;
284 	int		rmem_released;
285 	unsigned long	flags;
286 	unsigned long	cb_flags;
287 	bool		recovery;		/* closing subflow write queue reinjected */
288 	bool		can_ack;
289 	bool		fully_established;
290 	bool		rcv_data_fin;
291 	bool		snd_data_fin_enable;
292 	bool		rcv_fastclose;
293 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
294 	bool		csum_enabled;
295 	bool		allow_infinite_fallback;
296 	u8		pending_state; /* A subflow asked to set this sk_state,
297 					* protected by the msk data lock
298 					*/
299 	u8		mpc_endpoint_id;
300 	u8		recvmsg_inq:1,
301 			cork:1,
302 			nodelay:1,
303 			fastopening:1,
304 			in_accept_queue:1,
305 			free_first:1,
306 			rcvspace_init:1;
307 	int		keepalive_cnt;
308 	int		keepalive_idle;
309 	int		keepalive_intvl;
310 	struct work_struct work;
311 	struct sk_buff  *ooo_last_skb;
312 	struct rb_root  out_of_order_queue;
313 	struct sk_buff_head receive_queue;
314 	struct list_head conn_list;
315 	struct list_head rtx_queue;
316 	struct mptcp_data_frag *first_pending;
317 	struct list_head join_list;
318 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
319 				 * ONCE annotation, the subflow outside the socket
320 				 * lock as such sock is freed after close().
321 				 */
322 	struct mptcp_pm_data	pm;
323 	struct mptcp_sched_ops	*sched;
324 	struct {
325 		u32	space;	/* bytes copied in last measurement window */
326 		u32	copied; /* bytes copied in this measurement window */
327 		u64	time;	/* start time of measurement window */
328 		u64	rtt_us; /* last maximum rtt of subflows */
329 	} rcvq_space;
330 	u8		scaling_ratio;
331 
332 	u32		subflow_id;
333 	u32		setsockopt_seq;
334 	char		ca_name[TCP_CA_NAME_MAX];
335 };
336 
337 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
338 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
339 
340 #define mptcp_for_each_subflow(__msk, __subflow)			\
341 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
342 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
343 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
344 
msk_owned_by_me(const struct mptcp_sock * msk)345 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
346 {
347 	sock_owned_by_me((const struct sock *)msk);
348 }
349 
350 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
351 
352 /* the msk socket don't use the backlog, also account for the bulk
353  * free memory
354  */
__mptcp_rmem(const struct sock * sk)355 static inline int __mptcp_rmem(const struct sock *sk)
356 {
357 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
358 }
359 
mptcp_win_from_space(const struct sock * sk,int space)360 static inline int mptcp_win_from_space(const struct sock *sk, int space)
361 {
362 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
363 }
364 
__mptcp_space(const struct sock * sk)365 static inline int __mptcp_space(const struct sock *sk)
366 {
367 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
368 }
369 
mptcp_send_head(const struct sock * sk)370 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
371 {
372 	const struct mptcp_sock *msk = mptcp_sk(sk);
373 
374 	return READ_ONCE(msk->first_pending);
375 }
376 
mptcp_send_next(struct sock * sk)377 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
378 {
379 	struct mptcp_sock *msk = mptcp_sk(sk);
380 	struct mptcp_data_frag *cur;
381 
382 	cur = msk->first_pending;
383 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
384 						     list_next_entry(cur, list);
385 }
386 
mptcp_pending_tail(const struct sock * sk)387 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
388 {
389 	const struct mptcp_sock *msk = mptcp_sk(sk);
390 
391 	if (!msk->first_pending)
392 		return NULL;
393 
394 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
395 		return NULL;
396 
397 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
398 }
399 
mptcp_rtx_head(struct sock * sk)400 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
401 {
402 	struct mptcp_sock *msk = mptcp_sk(sk);
403 
404 	if (msk->snd_una == READ_ONCE(msk->snd_nxt))
405 		return NULL;
406 
407 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
408 }
409 
410 struct csum_pseudo_header {
411 	__be64 data_seq;
412 	__be32 subflow_seq;
413 	__be16 data_len;
414 	__sum16 csum;
415 };
416 
417 struct mptcp_subflow_request_sock {
418 	struct	tcp_request_sock sk;
419 	u16	mp_capable : 1,
420 		mp_join : 1,
421 		backup : 1,
422 		request_bkup : 1,
423 		csum_reqd : 1,
424 		allow_join_id0 : 1;
425 	u8	local_id;
426 	u8	remote_id;
427 	u64	local_key;
428 	u64	idsn;
429 	u32	token;
430 	u32	ssn_offset;
431 	u64	thmac;
432 	u32	local_nonce;
433 	u32	remote_nonce;
434 	struct mptcp_sock	*msk;
435 	struct hlist_nulls_node token_node;
436 };
437 
438 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)439 mptcp_subflow_rsk(const struct request_sock *rsk)
440 {
441 	return (struct mptcp_subflow_request_sock *)rsk;
442 }
443 
444 enum mptcp_data_avail {
445 	MPTCP_SUBFLOW_NODATA,
446 	MPTCP_SUBFLOW_DATA_AVAIL,
447 };
448 
449 struct mptcp_delegated_action {
450 	struct napi_struct napi;
451 	struct list_head head;
452 };
453 
454 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
455 
456 #define MPTCP_DELEGATE_SCHEDULED	0
457 #define MPTCP_DELEGATE_SEND		1
458 #define MPTCP_DELEGATE_ACK		2
459 #define MPTCP_DELEGATE_SNDBUF		3
460 
461 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
462 /* MPTCP subflow context */
463 struct mptcp_subflow_context {
464 	struct	list_head node;/* conn_list of subflows */
465 
466 	struct_group(reset,
467 
468 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
469 	u64	local_key;
470 	u64	remote_key;
471 	u64	idsn;
472 	u64	map_seq;
473 	u32	snd_isn;
474 	u32	token;
475 	u32	rel_write_seq;
476 	u32	map_subflow_seq;
477 	u32	ssn_offset;
478 	u32	map_data_len;
479 	__wsum	map_data_csum;
480 	u32	map_csum_len;
481 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
482 		request_join : 1,   /* send MP_JOIN */
483 		request_bkup : 1,
484 		mp_capable : 1,	    /* remote is MPTCP capable */
485 		mp_join : 1,	    /* remote is JOINing */
486 		fully_established : 1,	    /* path validated */
487 		pm_notified : 1,    /* PM hook called for established status */
488 		conn_finished : 1,
489 		map_valid : 1,
490 		map_csum_reqd : 1,
491 		map_data_fin : 1,
492 		mpc_map : 1,
493 		backup : 1,
494 		send_mp_prio : 1,
495 		send_mp_fail : 1,
496 		send_fastclose : 1,
497 		send_infinite_map : 1,
498 		remote_key_valid : 1,        /* received the peer key from */
499 		disposable : 1,	    /* ctx can be free at ulp release time */
500 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
501 		valid_csum_seen : 1,        /* at least one csum validated */
502 		is_mptfo : 1,	    /* subflow is doing TFO */
503 		close_event_done : 1,       /* has done the post-closed part */
504 		__unused : 9;
505 	enum mptcp_data_avail data_avail;
506 	bool	scheduled;
507 	bool	pm_listener;	    /* a listener managed by the kernel PM? */
508 	u32	remote_nonce;
509 	u64	thmac;
510 	u32	local_nonce;
511 	u32	remote_token;
512 	union {
513 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
514 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
515 	};
516 	s16	local_id;	    /* if negative not initialized yet */
517 	u8	remote_id;
518 	u8	reset_seen:1;
519 	u8	reset_transient:1;
520 	u8	reset_reason:4;
521 	u8	stale_count;
522 
523 	u32	subflow_id;
524 
525 	long	delegated_status;
526 	unsigned long	fail_tout;
527 
528 	);
529 
530 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
531 
532 	u32	setsockopt_seq;
533 	u32	stale_rcv_tstamp;
534 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
535 				     * protected by the msk socket lock
536 				     */
537 
538 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
539 	struct	sock *conn;	    /* parent mptcp_sock */
540 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
541 	void	(*tcp_state_change)(struct sock *sk);
542 	void	(*tcp_error_report)(struct sock *sk);
543 
544 	struct	rcu_head rcu;
545 };
546 
547 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)548 mptcp_subflow_ctx(const struct sock *sk)
549 {
550 	struct inet_connection_sock *icsk = inet_csk(sk);
551 
552 	/* Use RCU on icsk_ulp_data only for sock diag code */
553 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
554 }
555 
556 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)557 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
558 {
559 	return subflow->tcp_sock;
560 }
561 
562 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)563 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
564 {
565 	memset(&subflow->reset, 0, sizeof(subflow->reset));
566 	subflow->request_mptcp = 1;
567 	WRITE_ONCE(subflow->local_id, -1);
568 }
569 
570 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)571 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
572 {
573 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
574 		      subflow->ssn_offset -
575 		      subflow->map_subflow_seq;
576 }
577 
578 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)579 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
580 {
581 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
582 }
583 
584 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
585 
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)586 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
587 {
588 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
589 	struct mptcp_delegated_action *delegated;
590 	bool schedule;
591 
592 	/* the caller held the subflow bh socket lock */
593 	lockdep_assert_in_softirq();
594 
595 	/* The implied barrier pairs with tcp_release_cb_override()
596 	 * mptcp_napi_poll(), and ensures the below list check sees list
597 	 * updates done prior to delegated status bits changes
598 	 */
599 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
600 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
601 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
602 			return;
603 
604 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
605 		schedule = list_empty(&delegated->head);
606 		list_add_tail(&subflow->delegated_node, &delegated->head);
607 		sock_hold(mptcp_subflow_tcp_sock(subflow));
608 		if (schedule)
609 			napi_schedule(&delegated->napi);
610 	}
611 }
612 
613 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)614 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
615 {
616 	struct mptcp_subflow_context *ret;
617 
618 	if (list_empty(&delegated->head))
619 		return NULL;
620 
621 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
622 	list_del_init(&ret->delegated_node);
623 	return ret;
624 }
625 
626 int mptcp_is_enabled(const struct net *net);
627 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
628 int mptcp_is_checksum_enabled(const struct net *net);
629 int mptcp_allow_join_id0(const struct net *net);
630 unsigned int mptcp_stale_loss_cnt(const struct net *net);
631 int mptcp_get_pm_type(const struct net *net);
632 const char *mptcp_get_scheduler(const struct net *net);
633 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
634 				       struct mptcp_subflow_context *subflow,
635 				       const struct mptcp_options_received *mp_opt);
636 bool __mptcp_retransmit_pending_data(struct sock *sk);
637 void mptcp_check_and_set_pending(struct sock *sk);
638 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
639 bool mptcp_subflow_data_available(struct sock *sk);
640 void __init mptcp_subflow_init(void);
641 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
642 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
643 		     struct mptcp_subflow_context *subflow);
644 void __mptcp_subflow_send_ack(struct sock *ssk);
645 void mptcp_subflow_reset(struct sock *ssk);
646 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
647 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
648 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
649 bool __mptcp_close(struct sock *sk, long timeout);
650 void mptcp_cancel_work(struct sock *sk);
651 void __mptcp_unaccepted_force_close(struct sock *sk);
652 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
653 void mptcp_set_state(struct sock *sk, int state);
654 
655 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
656 			   const struct mptcp_addr_info *b, bool use_port);
657 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
658 
659 /* called with sk socket lock held */
660 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
661 			    const struct mptcp_addr_info *remote);
662 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
663 				struct socket **new_sock);
664 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
665 			 struct sockaddr_storage *addr,
666 			 unsigned short family);
667 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
668 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
669 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
670 void mptcp_sched_init(void);
671 int mptcp_init_sched(struct mptcp_sock *msk,
672 		     struct mptcp_sched_ops *sched);
673 void mptcp_release_sched(struct mptcp_sock *msk);
674 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
675 				 bool scheduled);
676 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
677 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
678 int mptcp_sched_get_send(struct mptcp_sock *msk);
679 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
680 
mptcp_data_avail(const struct mptcp_sock * msk)681 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
682 {
683 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
684 }
685 
mptcp_epollin_ready(const struct sock * sk)686 static inline bool mptcp_epollin_ready(const struct sock *sk)
687 {
688 	/* mptcp doesn't have to deal with small skbs in the receive queue,
689 	 * at it can always coalesce them
690 	 */
691 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
692 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
693 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
694 	       READ_ONCE(tcp_memory_pressure);
695 }
696 
697 int mptcp_set_rcvlowat(struct sock *sk, int val);
698 
__tcp_can_send(const struct sock * ssk)699 static inline bool __tcp_can_send(const struct sock *ssk)
700 {
701 	/* only send if our side has not closed yet */
702 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
703 }
704 
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)705 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
706 {
707 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
708 	if (subflow->request_join && !subflow->fully_established)
709 		return false;
710 
711 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
712 }
713 
714 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
715 
716 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
717 
718 void mptcp_subflow_drop_ctx(struct sock *ssk);
719 
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)720 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
721 					      struct mptcp_subflow_context *ctx)
722 {
723 	sk->sk_data_ready = sock_def_readable;
724 	sk->sk_state_change = ctx->tcp_state_change;
725 	sk->sk_write_space = sk_stream_write_space;
726 	sk->sk_error_report = ctx->tcp_error_report;
727 
728 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
729 }
730 
731 void __init mptcp_proto_init(void);
732 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
733 int __init mptcp_proto_v6_init(void);
734 #endif
735 
736 struct sock *mptcp_sk_clone_init(const struct sock *sk,
737 				 const struct mptcp_options_received *mp_opt,
738 				 struct sock *ssk,
739 				 struct request_sock *req);
740 void mptcp_get_options(const struct sk_buff *skb,
741 		       struct mptcp_options_received *mp_opt);
742 
743 void mptcp_finish_connect(struct sock *sk);
744 void __mptcp_sync_state(struct sock *sk, int state);
745 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
746 
mptcp_stop_tout_timer(struct sock * sk)747 static inline void mptcp_stop_tout_timer(struct sock *sk)
748 {
749 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
750 		return;
751 
752 	sk_stop_timer(sk, &sk->sk_timer);
753 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
754 }
755 
mptcp_set_close_tout(struct sock * sk,unsigned long tout)756 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
757 {
758 	/* avoid 0 timestamp, as that means no close timeout */
759 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
760 }
761 
mptcp_start_tout_timer(struct sock * sk)762 static inline void mptcp_start_tout_timer(struct sock *sk)
763 {
764 	mptcp_set_close_tout(sk, tcp_jiffies32);
765 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
766 }
767 
mptcp_is_fully_established(struct sock * sk)768 static inline bool mptcp_is_fully_established(struct sock *sk)
769 {
770 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
771 	       READ_ONCE(mptcp_sk(sk)->fully_established);
772 }
773 
774 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
775 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
776 bool mptcp_finish_join(struct sock *sk);
777 bool mptcp_schedule_work(struct sock *sk);
778 int mptcp_setsockopt(struct sock *sk, int level, int optname,
779 		     sockptr_t optval, unsigned int optlen);
780 int mptcp_getsockopt(struct sock *sk, int level, int optname,
781 		     char __user *optval, int __user *option);
782 
783 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)784 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
785 {
786 	if (use_64bit)
787 		return cur_seq;
788 
789 	return __mptcp_expand_seq(old_seq, cur_seq);
790 }
791 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
792 void __mptcp_data_acked(struct sock *sk);
793 void __mptcp_error_report(struct sock *sk);
794 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
mptcp_data_fin_enabled(const struct mptcp_sock * msk)795 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
796 {
797 	return READ_ONCE(msk->snd_data_fin_enable) &&
798 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
799 }
800 
mptcp_write_space(struct sock * sk)801 static inline void mptcp_write_space(struct sock *sk)
802 {
803 	if (sk_stream_is_writeable(sk)) {
804 		/* pairs with memory barrier in mptcp_poll */
805 		smp_mb();
806 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
807 			sk_stream_write_space(sk);
808 	}
809 }
810 
__mptcp_sync_sndbuf(struct sock * sk)811 static inline void __mptcp_sync_sndbuf(struct sock *sk)
812 {
813 	struct mptcp_subflow_context *subflow;
814 	int ssk_sndbuf, new_sndbuf;
815 
816 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
817 		return;
818 
819 	new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
820 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
821 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
822 
823 		subflow->cached_sndbuf = ssk_sndbuf;
824 		new_sndbuf += ssk_sndbuf;
825 	}
826 
827 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
828 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
829 	mptcp_write_space(sk);
830 }
831 
832 /* The called held both the msk socket and the subflow socket locks,
833  * possibly under BH
834  */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)835 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
836 {
837 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
838 
839 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
840 		__mptcp_sync_sndbuf(sk);
841 }
842 
843 /* the caller held only the subflow socket lock, either in process or
844  * BH context. Additionally this can be called under the msk data lock,
845  * so we can't acquire such lock here: let the delegate action acquires
846  * the needed locks in suitable order.
847  */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)848 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
849 {
850 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
851 
852 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
853 		return;
854 
855 	local_bh_disable();
856 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
857 	local_bh_enable();
858 }
859 
860 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
861 
862 #define MPTCP_TOKEN_MAX_RETRIES	4
863 
864 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)865 static inline void mptcp_token_init_request(struct request_sock *req)
866 {
867 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
868 }
869 
870 int mptcp_token_new_request(struct request_sock *req);
871 void mptcp_token_destroy_request(struct request_sock *req);
872 int mptcp_token_new_connect(struct sock *ssk);
873 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
874 			struct mptcp_sock *msk);
875 bool mptcp_token_exists(u32 token);
876 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
877 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
878 					 long *s_num);
879 void mptcp_token_destroy(struct mptcp_sock *msk);
880 
881 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
882 
883 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
884 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
885 
886 void __init mptcp_pm_init(void);
887 void mptcp_pm_data_init(struct mptcp_sock *msk);
888 void mptcp_pm_data_reset(struct mptcp_sock *msk);
889 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
890 			struct mptcp_addr_info *addr);
891 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
892 			 bool require_family,
893 			 struct mptcp_pm_addr_entry *entry);
894 bool mptcp_pm_addr_families_match(const struct sock *sk,
895 				  const struct mptcp_addr_info *loc,
896 				  const struct mptcp_addr_info *rem);
897 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
898 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
899 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
900 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
901 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
902 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
903 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
904 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
905 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
906 				 const struct mptcp_subflow_context *subflow);
907 void mptcp_pm_add_addr_received(const struct sock *ssk,
908 				const struct mptcp_addr_info *addr);
909 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
910 			      const struct mptcp_addr_info *addr);
911 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
912 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk,
913 				     const struct mptcp_addr_info *remote);
914 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
915 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
916 			       const struct mptcp_rm_list *rm_list);
917 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
918 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
919 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
920 				 struct mptcp_addr_info *addr,
921 				 struct mptcp_addr_info *rem,
922 				 u8 bkup);
923 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
924 			      const struct mptcp_addr_info *addr);
925 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
926 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
927 struct mptcp_pm_add_entry *
928 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
929 		       const struct mptcp_addr_info *addr, bool check_id);
930 struct mptcp_pm_add_entry *
931 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
932 				const struct mptcp_addr_info *addr);
933 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
934 					 unsigned int id,
935 					 u8 *flags, int *ifindex);
936 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
937 					    u8 *flags, int *ifindex);
938 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
939 						   unsigned int id,
940 						   u8 *flags, int *ifindex);
941 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
942 		       struct mptcp_pm_addr_entry *loc,
943 		       struct mptcp_pm_addr_entry *rem, u8 bkup);
944 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
945 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
946 				 struct mptcp_pm_addr_entry *loc,
947 				 struct mptcp_pm_addr_entry *rem, u8 bkup);
948 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
949 			   const struct mptcp_addr_info *addr,
950 			   bool echo);
951 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
952 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
953 
954 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
955 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
956 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
957 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
958 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
959 
960 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
961 		 const struct sock *ssk, gfp_t gfp);
962 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
963 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
964 void mptcp_event_pm_listener(const struct sock *ssk,
965 			     enum mptcp_event_type event);
966 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
967 
968 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
969 				     const struct mptcp_options_received *mp_opt);
970 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
971 					      struct request_sock *req);
972 
mptcp_pm_should_add_signal(struct mptcp_sock * msk)973 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
974 {
975 	return READ_ONCE(msk->pm.addr_signal) &
976 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
977 }
978 
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)979 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
980 {
981 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
982 }
983 
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)984 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
985 {
986 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
987 }
988 
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)989 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
990 {
991 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
992 }
993 
mptcp_pm_is_userspace(const struct mptcp_sock * msk)994 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
995 {
996 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
997 }
998 
mptcp_pm_is_kernel(const struct mptcp_sock * msk)999 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1000 {
1001 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1002 }
1003 
mptcp_add_addr_len(int family,bool echo,bool port)1004 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1005 {
1006 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1007 
1008 	if (family == AF_INET6)
1009 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1010 	if (!echo)
1011 		len += MPTCPOPT_THMAC_LEN;
1012 	/* account for 2 trailing 'nop' options */
1013 	if (port)
1014 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1015 
1016 	return len;
1017 }
1018 
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1019 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1020 {
1021 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1022 		return -EINVAL;
1023 
1024 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1025 }
1026 
1027 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1028 			      unsigned int opt_size, unsigned int remaining,
1029 			      struct mptcp_addr_info *addr, bool *echo,
1030 			      bool *drop_other_suboptions);
1031 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1032 			     struct mptcp_rm_list *rm_list);
1033 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1034 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1035 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1036 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1037 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1038 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1039 
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1040 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1041 {
1042 	int local_id = READ_ONCE(subflow->local_id);
1043 
1044 	if (local_id < 0)
1045 		return 0;
1046 	return local_id;
1047 }
1048 
1049 void __init mptcp_pm_nl_init(void);
1050 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1051 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1052 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1053 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1054 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1055 
1056 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1057 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1058 {
1059 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1060 		WRITE_ONCE(msk->pm.accept_subflow, true);
1061 }
1062 
mptcp_pm_close_subflow(struct mptcp_sock * msk)1063 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1064 {
1065 	spin_lock_bh(&msk->pm.lock);
1066 	__mptcp_pm_close_subflow(msk);
1067 	spin_unlock_bh(&msk->pm.lock);
1068 }
1069 
1070 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1071 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1072 
mptcp_get_ext(const struct sk_buff * skb)1073 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1074 {
1075 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1076 }
1077 
1078 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1079 
__mptcp_check_fallback(const struct mptcp_sock * msk)1080 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1081 {
1082 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1083 }
1084 
mptcp_check_fallback(const struct sock * sk)1085 static inline bool mptcp_check_fallback(const struct sock *sk)
1086 {
1087 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1088 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1089 
1090 	return __mptcp_check_fallback(msk);
1091 }
1092 
__mptcp_do_fallback(struct mptcp_sock * msk)1093 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1094 {
1095 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1096 		pr_debug("TCP fallback already done (msk=%p)\n", msk);
1097 		return;
1098 	}
1099 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1100 }
1101 
mptcp_do_fallback(struct sock * ssk)1102 static inline void mptcp_do_fallback(struct sock *ssk)
1103 {
1104 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1105 	struct sock *sk = subflow->conn;
1106 	struct mptcp_sock *msk;
1107 
1108 	msk = mptcp_sk(sk);
1109 	__mptcp_do_fallback(msk);
1110 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1111 		gfp_t saved_allocation = ssk->sk_allocation;
1112 
1113 		/* we are in a atomic (BH) scope, override ssk default for data
1114 		 * fin allocation
1115 		 */
1116 		ssk->sk_allocation = GFP_ATOMIC;
1117 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1118 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1119 		ssk->sk_allocation = saved_allocation;
1120 	}
1121 }
1122 
1123 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1124 
mptcp_check_infinite_map(struct sk_buff * skb)1125 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1126 {
1127 	struct mptcp_ext *mpext;
1128 
1129 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1130 	if (mpext && mpext->infinite_map)
1131 		return true;
1132 
1133 	return false;
1134 }
1135 
is_active_ssk(struct mptcp_subflow_context * subflow)1136 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1137 {
1138 	return (subflow->request_mptcp || subflow->request_join);
1139 }
1140 
subflow_simultaneous_connect(struct sock * sk)1141 static inline bool subflow_simultaneous_connect(struct sock *sk)
1142 {
1143 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1144 
1145 	return (1 << sk->sk_state) &
1146 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1147 	       is_active_ssk(subflow) &&
1148 	       !subflow->conn_finished;
1149 }
1150 
1151 #ifdef CONFIG_SYN_COOKIES
1152 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1153 				       struct sk_buff *skb);
1154 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1155 					struct sk_buff *skb);
1156 void __init mptcp_join_cookie_init(void);
1157 #else
1158 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1159 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1160 				  struct sk_buff *skb) {}
1161 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1162 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1163 				   struct sk_buff *skb)
1164 {
1165 	return false;
1166 }
1167 
mptcp_join_cookie_init(void)1168 static inline void mptcp_join_cookie_init(void) {}
1169 #endif
1170 
1171 #endif /* __MPTCP_PROTOCOL_H */
1172