• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 		__unused : 10;
504 	enum mptcp_data_avail data_avail;
505 	bool	scheduled;
506 	u32	remote_nonce;
507 	u64	thmac;
508 	u32	local_nonce;
509 	u32	remote_token;
510 	union {
511 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
512 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
513 	};
514 	s16	local_id;	    /* if negative not initialized yet */
515 	u8	remote_id;
516 	u8	reset_seen:1;
517 	u8	reset_transient:1;
518 	u8	reset_reason:4;
519 	u8	stale_count;
520 
521 	u32	subflow_id;
522 
523 	long	delegated_status;
524 	unsigned long	fail_tout;
525 
526 	);
527 
528 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
529 
530 	u32	setsockopt_seq;
531 	u32	stale_rcv_tstamp;
532 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
533 				     * protected by the msk socket lock
534 				     */
535 
536 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
537 	struct	sock *conn;	    /* parent mptcp_sock */
538 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
539 	void	(*tcp_state_change)(struct sock *sk);
540 	void	(*tcp_error_report)(struct sock *sk);
541 
542 	struct	rcu_head rcu;
543 };
544 
545 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)546 mptcp_subflow_ctx(const struct sock *sk)
547 {
548 	struct inet_connection_sock *icsk = inet_csk(sk);
549 
550 	/* Use RCU on icsk_ulp_data only for sock diag code */
551 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
552 }
553 
554 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)555 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
556 {
557 	return subflow->tcp_sock;
558 }
559 
560 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)561 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
562 {
563 	memset(&subflow->reset, 0, sizeof(subflow->reset));
564 	subflow->request_mptcp = 1;
565 	WRITE_ONCE(subflow->local_id, -1);
566 }
567 
568 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)569 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
570 {
571 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
572 		      subflow->ssn_offset -
573 		      subflow->map_subflow_seq;
574 }
575 
576 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)577 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
578 {
579 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
580 }
581 
582 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
583 
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)584 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
585 {
586 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
587 	struct mptcp_delegated_action *delegated;
588 	bool schedule;
589 
590 	/* the caller held the subflow bh socket lock */
591 	lockdep_assert_in_softirq();
592 
593 	/* The implied barrier pairs with tcp_release_cb_override()
594 	 * mptcp_napi_poll(), and ensures the below list check sees list
595 	 * updates done prior to delegated status bits changes
596 	 */
597 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
598 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
599 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
600 			return;
601 
602 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
603 		schedule = list_empty(&delegated->head);
604 		list_add_tail(&subflow->delegated_node, &delegated->head);
605 		sock_hold(mptcp_subflow_tcp_sock(subflow));
606 		if (schedule)
607 			napi_schedule(&delegated->napi);
608 	}
609 }
610 
611 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)612 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
613 {
614 	struct mptcp_subflow_context *ret;
615 
616 	if (list_empty(&delegated->head))
617 		return NULL;
618 
619 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
620 	list_del_init(&ret->delegated_node);
621 	return ret;
622 }
623 
624 int mptcp_is_enabled(const struct net *net);
625 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
626 int mptcp_is_checksum_enabled(const struct net *net);
627 int mptcp_allow_join_id0(const struct net *net);
628 unsigned int mptcp_stale_loss_cnt(const struct net *net);
629 int mptcp_get_pm_type(const struct net *net);
630 const char *mptcp_get_scheduler(const struct net *net);
631 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
632 				       struct mptcp_subflow_context *subflow,
633 				       const struct mptcp_options_received *mp_opt);
634 bool __mptcp_retransmit_pending_data(struct sock *sk);
635 void mptcp_check_and_set_pending(struct sock *sk);
636 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
637 bool mptcp_subflow_data_available(struct sock *sk);
638 void __init mptcp_subflow_init(void);
639 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
640 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
641 		     struct mptcp_subflow_context *subflow);
642 void __mptcp_subflow_send_ack(struct sock *ssk);
643 void mptcp_subflow_reset(struct sock *ssk);
644 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
645 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
646 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
647 bool __mptcp_close(struct sock *sk, long timeout);
648 void mptcp_cancel_work(struct sock *sk);
649 void __mptcp_unaccepted_force_close(struct sock *sk);
650 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
651 void mptcp_set_state(struct sock *sk, int state);
652 
653 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
654 			   const struct mptcp_addr_info *b, bool use_port);
655 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
656 
657 /* called with sk socket lock held */
658 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
659 			    const struct mptcp_addr_info *remote);
660 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
661 				struct socket **new_sock);
662 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
663 			 struct sockaddr_storage *addr,
664 			 unsigned short family);
665 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
666 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
667 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
668 void mptcp_sched_init(void);
669 int mptcp_init_sched(struct mptcp_sock *msk,
670 		     struct mptcp_sched_ops *sched);
671 void mptcp_release_sched(struct mptcp_sock *msk);
672 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
673 				 bool scheduled);
674 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
675 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
676 int mptcp_sched_get_send(struct mptcp_sock *msk);
677 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
678 
mptcp_data_avail(const struct mptcp_sock * msk)679 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
680 {
681 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
682 }
683 
mptcp_epollin_ready(const struct sock * sk)684 static inline bool mptcp_epollin_ready(const struct sock *sk)
685 {
686 	/* mptcp doesn't have to deal with small skbs in the receive queue,
687 	 * at it can always coalesce them
688 	 */
689 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
690 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
691 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
692 	       READ_ONCE(tcp_memory_pressure);
693 }
694 
695 int mptcp_set_rcvlowat(struct sock *sk, int val);
696 
__tcp_can_send(const struct sock * ssk)697 static inline bool __tcp_can_send(const struct sock *ssk)
698 {
699 	/* only send if our side has not closed yet */
700 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
701 }
702 
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)703 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
704 {
705 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
706 	if (subflow->request_join && !subflow->fully_established)
707 		return false;
708 
709 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
710 }
711 
712 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
713 
714 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
715 
716 void mptcp_subflow_drop_ctx(struct sock *ssk);
717 
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)718 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
719 					      struct mptcp_subflow_context *ctx)
720 {
721 	sk->sk_data_ready = sock_def_readable;
722 	sk->sk_state_change = ctx->tcp_state_change;
723 	sk->sk_write_space = sk_stream_write_space;
724 	sk->sk_error_report = ctx->tcp_error_report;
725 
726 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
727 }
728 
729 void __init mptcp_proto_init(void);
730 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
731 int __init mptcp_proto_v6_init(void);
732 #endif
733 
734 struct sock *mptcp_sk_clone_init(const struct sock *sk,
735 				 const struct mptcp_options_received *mp_opt,
736 				 struct sock *ssk,
737 				 struct request_sock *req);
738 void mptcp_get_options(const struct sk_buff *skb,
739 		       struct mptcp_options_received *mp_opt);
740 
741 void mptcp_finish_connect(struct sock *sk);
742 void __mptcp_sync_state(struct sock *sk, int state);
743 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
744 
mptcp_stop_tout_timer(struct sock * sk)745 static inline void mptcp_stop_tout_timer(struct sock *sk)
746 {
747 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
748 		return;
749 
750 	sk_stop_timer(sk, &sk->sk_timer);
751 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
752 }
753 
mptcp_set_close_tout(struct sock * sk,unsigned long tout)754 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
755 {
756 	/* avoid 0 timestamp, as that means no close timeout */
757 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
758 }
759 
mptcp_start_tout_timer(struct sock * sk)760 static inline void mptcp_start_tout_timer(struct sock *sk)
761 {
762 	mptcp_set_close_tout(sk, tcp_jiffies32);
763 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
764 }
765 
mptcp_is_fully_established(struct sock * sk)766 static inline bool mptcp_is_fully_established(struct sock *sk)
767 {
768 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
769 	       READ_ONCE(mptcp_sk(sk)->fully_established);
770 }
771 
772 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
773 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
774 bool mptcp_finish_join(struct sock *sk);
775 bool mptcp_schedule_work(struct sock *sk);
776 int mptcp_setsockopt(struct sock *sk, int level, int optname,
777 		     sockptr_t optval, unsigned int optlen);
778 int mptcp_getsockopt(struct sock *sk, int level, int optname,
779 		     char __user *optval, int __user *option);
780 
781 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)782 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
783 {
784 	if (use_64bit)
785 		return cur_seq;
786 
787 	return __mptcp_expand_seq(old_seq, cur_seq);
788 }
789 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
790 void __mptcp_data_acked(struct sock *sk);
791 void __mptcp_error_report(struct sock *sk);
792 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)793 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
794 {
795 	return READ_ONCE(msk->snd_data_fin_enable) &&
796 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
797 }
798 
mptcp_write_space(struct sock * sk)799 static inline void mptcp_write_space(struct sock *sk)
800 {
801 	if (sk_stream_is_writeable(sk)) {
802 		/* pairs with memory barrier in mptcp_poll */
803 		smp_mb();
804 		if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
805 			sk_stream_write_space(sk);
806 	}
807 }
808 
__mptcp_sync_sndbuf(struct sock * sk)809 static inline void __mptcp_sync_sndbuf(struct sock *sk)
810 {
811 	struct mptcp_subflow_context *subflow;
812 	int ssk_sndbuf, new_sndbuf;
813 
814 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
815 		return;
816 
817 	new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
818 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
819 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
820 
821 		subflow->cached_sndbuf = ssk_sndbuf;
822 		new_sndbuf += ssk_sndbuf;
823 	}
824 
825 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
826 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
827 	mptcp_write_space(sk);
828 }
829 
830 /* The called held both the msk socket and the subflow socket locks,
831  * possibly under BH
832  */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)833 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
834 {
835 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
836 
837 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
838 		__mptcp_sync_sndbuf(sk);
839 }
840 
841 /* the caller held only the subflow socket lock, either in process or
842  * BH context. Additionally this can be called under the msk data lock,
843  * so we can't acquire such lock here: let the delegate action acquires
844  * the needed locks in suitable order.
845  */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)846 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
847 {
848 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
849 
850 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
851 		return;
852 
853 	local_bh_disable();
854 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
855 	local_bh_enable();
856 }
857 
858 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
859 
860 #define MPTCP_TOKEN_MAX_RETRIES	4
861 
862 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)863 static inline void mptcp_token_init_request(struct request_sock *req)
864 {
865 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
866 }
867 
868 int mptcp_token_new_request(struct request_sock *req);
869 void mptcp_token_destroy_request(struct request_sock *req);
870 int mptcp_token_new_connect(struct sock *ssk);
871 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
872 			struct mptcp_sock *msk);
873 bool mptcp_token_exists(u32 token);
874 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
875 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
876 					 long *s_num);
877 void mptcp_token_destroy(struct mptcp_sock *msk);
878 
879 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
880 
881 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
882 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
883 
884 void __init mptcp_pm_init(void);
885 void mptcp_pm_data_init(struct mptcp_sock *msk);
886 void mptcp_pm_data_reset(struct mptcp_sock *msk);
887 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
888 			struct mptcp_addr_info *addr);
889 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
890 			 bool require_family,
891 			 struct mptcp_pm_addr_entry *entry);
892 bool mptcp_pm_addr_families_match(const struct sock *sk,
893 				  const struct mptcp_addr_info *loc,
894 				  const struct mptcp_addr_info *rem);
895 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
896 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
897 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
898 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
899 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
900 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
901 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
902 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
903 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
904 				 const struct mptcp_subflow_context *subflow);
905 void mptcp_pm_add_addr_received(const struct sock *ssk,
906 				const struct mptcp_addr_info *addr);
907 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
908 			      const struct mptcp_addr_info *addr);
909 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
910 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
911 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
912 			       const struct mptcp_rm_list *rm_list);
913 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
914 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
915 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
916 				 struct mptcp_addr_info *addr,
917 				 struct mptcp_addr_info *rem,
918 				 u8 bkup);
919 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
920 			      const struct mptcp_addr_info *addr);
921 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
922 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
923 struct mptcp_pm_add_entry *
924 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
925 		       const struct mptcp_addr_info *addr, bool check_id);
926 struct mptcp_pm_add_entry *
927 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
928 				const struct mptcp_addr_info *addr);
929 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
930 					 unsigned int id,
931 					 u8 *flags, int *ifindex);
932 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
933 					    u8 *flags, int *ifindex);
934 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
935 						   unsigned int id,
936 						   u8 *flags, int *ifindex);
937 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
938 		       struct mptcp_pm_addr_entry *loc,
939 		       struct mptcp_pm_addr_entry *rem, u8 bkup);
940 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
941 int mptcp_userspace_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_announce_addr(struct mptcp_sock *msk,
945 			   const struct mptcp_addr_info *addr,
946 			   bool echo);
947 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
948 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
949 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
950 					struct list_head *rm_list);
951 
952 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
953 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
954 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
955 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
956 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
957 
958 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
959 		 const struct sock *ssk, gfp_t gfp);
960 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
961 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
962 void mptcp_event_pm_listener(const struct sock *ssk,
963 			     enum mptcp_event_type event);
964 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
965 
966 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
967 				     const struct mptcp_options_received *mp_opt);
968 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
969 					      struct request_sock *req);
970 
mptcp_pm_should_add_signal(struct mptcp_sock * msk)971 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
972 {
973 	return READ_ONCE(msk->pm.addr_signal) &
974 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
975 }
976 
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)977 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
978 {
979 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
980 }
981 
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)982 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
983 {
984 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
985 }
986 
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)987 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
988 {
989 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
990 }
991 
mptcp_pm_is_userspace(const struct mptcp_sock * msk)992 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
993 {
994 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
995 }
996 
mptcp_pm_is_kernel(const struct mptcp_sock * msk)997 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
998 {
999 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1000 }
1001 
mptcp_add_addr_len(int family,bool echo,bool port)1002 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1003 {
1004 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1005 
1006 	if (family == AF_INET6)
1007 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1008 	if (!echo)
1009 		len += MPTCPOPT_THMAC_LEN;
1010 	/* account for 2 trailing 'nop' options */
1011 	if (port)
1012 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1013 
1014 	return len;
1015 }
1016 
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1017 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1018 {
1019 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1020 		return -EINVAL;
1021 
1022 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1023 }
1024 
1025 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1026 			      unsigned int opt_size, unsigned int remaining,
1027 			      struct mptcp_addr_info *addr, bool *echo,
1028 			      bool *drop_other_suboptions);
1029 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1030 			     struct mptcp_rm_list *rm_list);
1031 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1032 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1033 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1034 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1035 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1036 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1037 
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1038 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1039 {
1040 	int local_id = READ_ONCE(subflow->local_id);
1041 
1042 	if (local_id < 0)
1043 		return 0;
1044 	return local_id;
1045 }
1046 
1047 void __init mptcp_pm_nl_init(void);
1048 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1049 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1050 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1051 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1052 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1053 
1054 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1055 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1056 {
1057 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1058 		WRITE_ONCE(msk->pm.accept_subflow, true);
1059 }
1060 
mptcp_pm_close_subflow(struct mptcp_sock * msk)1061 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1062 {
1063 	spin_lock_bh(&msk->pm.lock);
1064 	__mptcp_pm_close_subflow(msk);
1065 	spin_unlock_bh(&msk->pm.lock);
1066 }
1067 
1068 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1069 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1070 
mptcp_get_ext(const struct sk_buff * skb)1071 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1072 {
1073 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1074 }
1075 
1076 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1077 
__mptcp_check_fallback(const struct mptcp_sock * msk)1078 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1079 {
1080 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1081 }
1082 
mptcp_check_fallback(const struct sock * sk)1083 static inline bool mptcp_check_fallback(const struct sock *sk)
1084 {
1085 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1086 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1087 
1088 	return __mptcp_check_fallback(msk);
1089 }
1090 
__mptcp_do_fallback(struct mptcp_sock * msk)1091 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1092 {
1093 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1094 		pr_debug("TCP fallback already done (msk=%p)", msk);
1095 		return;
1096 	}
1097 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1098 }
1099 
mptcp_do_fallback(struct sock * ssk)1100 static inline void mptcp_do_fallback(struct sock *ssk)
1101 {
1102 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1103 	struct sock *sk = subflow->conn;
1104 	struct mptcp_sock *msk;
1105 
1106 	msk = mptcp_sk(sk);
1107 	__mptcp_do_fallback(msk);
1108 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1109 		gfp_t saved_allocation = ssk->sk_allocation;
1110 
1111 		/* we are in a atomic (BH) scope, override ssk default for data
1112 		 * fin allocation
1113 		 */
1114 		ssk->sk_allocation = GFP_ATOMIC;
1115 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1116 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1117 		ssk->sk_allocation = saved_allocation;
1118 	}
1119 }
1120 
1121 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1122 
mptcp_check_infinite_map(struct sk_buff * skb)1123 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1124 {
1125 	struct mptcp_ext *mpext;
1126 
1127 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1128 	if (mpext && mpext->infinite_map)
1129 		return true;
1130 
1131 	return false;
1132 }
1133 
is_active_ssk(struct mptcp_subflow_context * subflow)1134 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1135 {
1136 	return (subflow->request_mptcp || subflow->request_join);
1137 }
1138 
subflow_simultaneous_connect(struct sock * sk)1139 static inline bool subflow_simultaneous_connect(struct sock *sk)
1140 {
1141 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1142 
1143 	return (1 << sk->sk_state) &
1144 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1145 	       is_active_ssk(subflow) &&
1146 	       !subflow->conn_finished;
1147 }
1148 
1149 #ifdef CONFIG_SYN_COOKIES
1150 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1151 				       struct sk_buff *skb);
1152 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1153 					struct sk_buff *skb);
1154 void __init mptcp_join_cookie_init(void);
1155 #else
1156 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1157 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1158 				  struct sk_buff *skb) {}
1159 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1160 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1161 				   struct sk_buff *skb)
1162 {
1163 	return false;
1164 }
1165 
mptcp_join_cookie_init(void)1166 static inline void mptcp_join_cookie_init(void) {}
1167 #endif
1168 
1169 #endif /* __MPTCP_PROTOCOL_H */
1170