xref: /linux/net/mptcp/protocol.h (revision 1a9239bb4253f9076b5b4b2a1a4e8d7defd77a95)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3  *
4  * Copyright (c) 2017 - 2019, Intel Corporation.
5  */
6 
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9 
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 #include <net/genetlink.h>
15 #include <net/rstreason.h>
16 
17 #define MPTCP_SUPPORTED_VERSION	1
18 
19 /* MPTCP option bits */
20 #define OPTION_MPTCP_MPC_SYN	BIT(0)
21 #define OPTION_MPTCP_MPC_SYNACK	BIT(1)
22 #define OPTION_MPTCP_MPC_ACK	BIT(2)
23 #define OPTION_MPTCP_MPJ_SYN	BIT(3)
24 #define OPTION_MPTCP_MPJ_SYNACK	BIT(4)
25 #define OPTION_MPTCP_MPJ_ACK	BIT(5)
26 #define OPTION_MPTCP_ADD_ADDR	BIT(6)
27 #define OPTION_MPTCP_RM_ADDR	BIT(7)
28 #define OPTION_MPTCP_FASTCLOSE	BIT(8)
29 #define OPTION_MPTCP_PRIO	BIT(9)
30 #define OPTION_MPTCP_RST	BIT(10)
31 #define OPTION_MPTCP_DSS	BIT(11)
32 #define OPTION_MPTCP_FAIL	BIT(12)
33 
34 #define OPTION_MPTCP_CSUMREQD	BIT(13)
35 
36 #define OPTIONS_MPTCP_MPC	(OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
37 				 OPTION_MPTCP_MPC_ACK)
38 #define OPTIONS_MPTCP_MPJ	(OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
39 				 OPTION_MPTCP_MPJ_ACK)
40 
41 /* MPTCP option subtypes */
42 #define MPTCPOPT_MP_CAPABLE	0
43 #define MPTCPOPT_MP_JOIN	1
44 #define MPTCPOPT_DSS		2
45 #define MPTCPOPT_ADD_ADDR	3
46 #define MPTCPOPT_RM_ADDR	4
47 #define MPTCPOPT_MP_PRIO	5
48 #define MPTCPOPT_MP_FAIL	6
49 #define MPTCPOPT_MP_FASTCLOSE	7
50 #define MPTCPOPT_RST		8
51 
52 /* MPTCP suboption lengths */
53 #define TCPOLEN_MPTCP_MPC_SYN		4
54 #define TCPOLEN_MPTCP_MPC_SYNACK	12
55 #define TCPOLEN_MPTCP_MPC_ACK		20
56 #define TCPOLEN_MPTCP_MPC_ACK_DATA	22
57 #define TCPOLEN_MPTCP_MPJ_SYN		12
58 #define TCPOLEN_MPTCP_MPJ_SYNACK	16
59 #define TCPOLEN_MPTCP_MPJ_ACK		24
60 #define TCPOLEN_MPTCP_DSS_BASE		4
61 #define TCPOLEN_MPTCP_DSS_ACK32		4
62 #define TCPOLEN_MPTCP_DSS_ACK64		8
63 #define TCPOLEN_MPTCP_DSS_MAP32		10
64 #define TCPOLEN_MPTCP_DSS_MAP64		14
65 #define TCPOLEN_MPTCP_DSS_CHECKSUM	2
66 #define TCPOLEN_MPTCP_ADD_ADDR		16
67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT	18
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE	8
69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT	10
70 #define TCPOLEN_MPTCP_ADD_ADDR6		28
71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT	30
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE	20
73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT	22
74 #define TCPOLEN_MPTCP_PORT_LEN		2
75 #define TCPOLEN_MPTCP_PORT_ALIGN	2
76 #define TCPOLEN_MPTCP_RM_ADDR_BASE	3
77 #define TCPOLEN_MPTCP_PRIO		3
78 #define TCPOLEN_MPTCP_PRIO_ALIGN	4
79 #define TCPOLEN_MPTCP_FASTCLOSE		12
80 #define TCPOLEN_MPTCP_RST		4
81 #define TCPOLEN_MPTCP_FAIL		12
82 
83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM	(TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
84 
85 /* MPTCP MP_JOIN flags */
86 #define MPTCPOPT_BACKUP		BIT(0)
87 #define MPTCPOPT_THMAC_LEN	8
88 
89 /* MPTCP MP_CAPABLE flags */
90 #define MPTCP_VERSION_MASK	(0x0F)
91 #define MPTCP_CAP_CHECKSUM_REQD	BIT(7)
92 #define MPTCP_CAP_EXTENSIBILITY	BIT(6)
93 #define MPTCP_CAP_DENY_JOIN_ID0	BIT(5)
94 #define MPTCP_CAP_HMAC_SHA256	BIT(0)
95 #define MPTCP_CAP_FLAG_MASK	(0x1F)
96 
97 /* MPTCP DSS flags */
98 #define MPTCP_DSS_DATA_FIN	BIT(4)
99 #define MPTCP_DSS_DSN64		BIT(3)
100 #define MPTCP_DSS_HAS_MAP	BIT(2)
101 #define MPTCP_DSS_ACK64		BIT(1)
102 #define MPTCP_DSS_HAS_ACK	BIT(0)
103 #define MPTCP_DSS_FLAG_MASK	(0x1F)
104 
105 /* MPTCP ADD_ADDR flags */
106 #define MPTCP_ADDR_ECHO		BIT(0)
107 
108 /* MPTCP MP_PRIO flags */
109 #define MPTCP_PRIO_BKUP		BIT(0)
110 
111 /* MPTCP TCPRST flags */
112 #define MPTCP_RST_TRANSIENT	BIT(0)
113 
114 /* MPTCP socket atomic flags */
115 #define MPTCP_WORK_RTX		1
116 #define MPTCP_FALLBACK_DONE	2
117 #define MPTCP_WORK_CLOSE_SUBFLOW 3
118 
119 /* MPTCP socket release cb flags */
120 #define MPTCP_PUSH_PENDING	1
121 #define MPTCP_CLEAN_UNA		2
122 #define MPTCP_ERROR_REPORT	3
123 #define MPTCP_RETRANSMIT	4
124 #define MPTCP_FLUSH_JOIN_LIST	5
125 #define MPTCP_SYNC_STATE	6
126 #define MPTCP_SYNC_SNDBUF	7
127 #define MPTCP_DEQUEUE		8
128 
129 struct mptcp_skb_cb {
130 	u64 map_seq;
131 	u64 end_seq;
132 	u32 offset;
133 	u8  has_rxtstamp;
134 	u8  cant_coalesce;
135 };
136 
137 #define MPTCP_SKB_CB(__skb)	((struct mptcp_skb_cb *)&((__skb)->cb[0]))
138 
before64(__u64 seq1,__u64 seq2)139 static inline bool before64(__u64 seq1, __u64 seq2)
140 {
141 	return (__s64)(seq1 - seq2) < 0;
142 }
143 
144 #define after64(seq2, seq1)	before64(seq1, seq2)
145 
146 struct mptcp_options_received {
147 	u64	sndr_key;
148 	u64	rcvr_key;
149 	u64	data_ack;
150 	u64	data_seq;
151 	u32	subflow_seq;
152 	u16	data_len;
153 	__sum16	csum;
154 	struct_group(status,
155 		u16 suboptions;
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 	);
169 	u8	join_id;
170 	u32	token;
171 	u32	nonce;
172 	u64	thmac;
173 	u8	hmac[MPTCPOPT_HMAC_LEN];
174 	struct mptcp_addr_info addr;
175 	struct mptcp_rm_list rm_list;
176 	u64	ahmac;
177 	u64	fail_seq;
178 };
179 
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)180 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
181 {
182 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
183 		     ((nib & 0xF) << 8) | field);
184 }
185 
186 enum mptcp_pm_status {
187 	MPTCP_PM_ADD_ADDR_RECEIVED,
188 	MPTCP_PM_ADD_ADDR_SEND_ACK,
189 	MPTCP_PM_RM_ADDR_RECEIVED,
190 	MPTCP_PM_ESTABLISHED,
191 	MPTCP_PM_SUBFLOW_ESTABLISHED,
192 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
193 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
194 					 * accounted int id_avail_bitmap
195 					 */
196 };
197 
198 enum mptcp_pm_type {
199 	MPTCP_PM_TYPE_KERNEL = 0,
200 	MPTCP_PM_TYPE_USERSPACE,
201 
202 	__MPTCP_PM_TYPE_NR,
203 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
204 };
205 
206 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
207 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
208 
209 enum mptcp_addr_signal_status {
210 	MPTCP_ADD_ADDR_SIGNAL,
211 	MPTCP_ADD_ADDR_ECHO,
212 	MPTCP_RM_ADDR_SIGNAL,
213 };
214 
215 /* max value of mptcp_addr_info.id */
216 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
217 
218 struct mptcp_pm_data {
219 	struct mptcp_addr_info local;
220 	struct mptcp_addr_info remote;
221 	struct list_head anno_list;
222 	struct list_head userspace_pm_local_addr_list;
223 
224 	spinlock_t	lock;		/*protects the whole PM data */
225 
226 	struct_group(reset,
227 
228 	u8		addr_signal;
229 	bool		server_side;
230 	bool		work_pending;
231 	bool		accept_addr;
232 	bool		accept_subflow;
233 	bool		remote_deny_join_id0;
234 	u8		add_addr_signaled;
235 	u8		add_addr_accepted;
236 	u8		local_addr_used;
237 	u8		pm_type;
238 	u8		subflows;
239 	u8		status;
240 
241 	);
242 
243 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
244 	struct mptcp_rm_list rm_list_tx;
245 	struct mptcp_rm_list rm_list_rx;
246 };
247 
248 struct mptcp_pm_local {
249 	struct mptcp_addr_info	addr;
250 	u8			flags;
251 	int			ifindex;
252 };
253 
254 struct mptcp_pm_addr_entry {
255 	struct list_head	list;
256 	struct mptcp_addr_info	addr;
257 	u8			flags;
258 	int			ifindex;
259 	struct socket		*lsk;
260 };
261 
262 struct mptcp_data_frag {
263 	struct list_head list;
264 	u64 data_seq;
265 	u16 data_len;
266 	u16 offset;
267 	u16 overhead;
268 	u16 already_sent;
269 	struct page *page;
270 };
271 
272 /* MPTCP connection sock */
273 struct mptcp_sock {
274 	/* inet_connection_sock must be the first member */
275 	struct inet_connection_sock sk;
276 	u64		local_key;		/* protected by the first subflow socket lock
277 						 * lockless access read
278 						 */
279 	u64		remote_key;		/* same as above */
280 	u64		write_seq;
281 	u64		bytes_sent;
282 	u64		snd_nxt;
283 	u64		bytes_received;
284 	u64		ack_seq;
285 	atomic64_t	rcv_wnd_sent;
286 	u64		rcv_data_fin_seq;
287 	u64		bytes_retrans;
288 	u64		bytes_consumed;
289 	int		snd_burst;
290 	int		old_wspace;
291 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
292 						 * recovery related fields are under data_lock
293 						 * protection
294 						 */
295 	u64		bytes_acked;
296 	u64		snd_una;
297 	u64		wnd_end;
298 	u32		last_data_sent;
299 	u32		last_data_recv;
300 	u32		last_ack_recv;
301 	unsigned long	timer_ival;
302 	u32		token;
303 	unsigned long	flags;
304 	unsigned long	cb_flags;
305 	bool		recovery;		/* closing subflow write queue reinjected */
306 	bool		can_ack;
307 	bool		fully_established;
308 	bool		rcv_data_fin;
309 	bool		snd_data_fin_enable;
310 	bool		rcv_fastclose;
311 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
312 	bool		csum_enabled;
313 	bool		allow_infinite_fallback;
314 	u8		pending_state; /* A subflow asked to set this sk_state,
315 					* protected by the msk data lock
316 					*/
317 	u8		mpc_endpoint_id;
318 	u8		recvmsg_inq:1,
319 			cork:1,
320 			nodelay:1,
321 			fastopening:1,
322 			in_accept_queue:1,
323 			free_first:1,
324 			rcvspace_init:1;
325 	u32		notsent_lowat;
326 	int		keepalive_cnt;
327 	int		keepalive_idle;
328 	int		keepalive_intvl;
329 	struct work_struct work;
330 	struct sk_buff  *ooo_last_skb;
331 	struct rb_root  out_of_order_queue;
332 	struct list_head conn_list;
333 	struct list_head rtx_queue;
334 	struct mptcp_data_frag *first_pending;
335 	struct list_head join_list;
336 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
337 				 * ONCE annotation, the subflow outside the socket
338 				 * lock as such sock is freed after close().
339 				 */
340 	struct mptcp_pm_data	pm;
341 	struct mptcp_sched_ops	*sched;
342 	struct {
343 		u32	space;	/* bytes copied in last measurement window */
344 		u32	copied; /* bytes copied in this measurement window */
345 		u64	time;	/* start time of measurement window */
346 		u64	rtt_us; /* last maximum rtt of subflows */
347 	} rcvq_space;
348 	u8		scaling_ratio;
349 
350 	u32		subflow_id;
351 	u32		setsockopt_seq;
352 	char		ca_name[TCP_CA_NAME_MAX];
353 };
354 
355 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
356 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
357 
358 #define mptcp_for_each_subflow(__msk, __subflow)			\
359 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
360 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
361 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
362 #define mptcp_next_subflow(__msk, __subflow)				\
363 	list_next_entry_circular(__subflow, &((__msk)->conn_list), node)
364 
365 extern struct genl_family mptcp_genl_family;
366 
msk_owned_by_me(const struct mptcp_sock * msk)367 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
368 {
369 	sock_owned_by_me((const struct sock *)msk);
370 }
371 
372 #ifdef CONFIG_DEBUG_NET
373 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
374 #undef tcp_sk
375 #define tcp_sk(ptr) ({								\
376 	typeof(ptr) _ptr = (ptr);						\
377 	WARN_ON(_ptr->sk_protocol != IPPROTO_TCP);				\
378 	container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk);	\
379 })
380 #define mptcp_sk(ptr) ({						\
381 	typeof(ptr) _ptr = (ptr);					\
382 	WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP);			\
383 	container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk);	\
384 })
385 
386 #else /* !CONFIG_DEBUG_NET */
387 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
388 #endif
389 
mptcp_win_from_space(const struct sock * sk,int space)390 static inline int mptcp_win_from_space(const struct sock *sk, int space)
391 {
392 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
393 }
394 
mptcp_space_from_win(const struct sock * sk,int win)395 static inline int mptcp_space_from_win(const struct sock *sk, int win)
396 {
397 	return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
398 }
399 
__mptcp_space(const struct sock * sk)400 static inline int __mptcp_space(const struct sock *sk)
401 {
402 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
403 				    sk_rmem_alloc_get(sk));
404 }
405 
mptcp_send_head(const struct sock * sk)406 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
407 {
408 	const struct mptcp_sock *msk = mptcp_sk(sk);
409 
410 	return READ_ONCE(msk->first_pending);
411 }
412 
mptcp_send_next(struct sock * sk)413 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
414 {
415 	struct mptcp_sock *msk = mptcp_sk(sk);
416 	struct mptcp_data_frag *cur;
417 
418 	cur = msk->first_pending;
419 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
420 						     list_next_entry(cur, list);
421 }
422 
mptcp_pending_tail(const struct sock * sk)423 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
424 {
425 	const struct mptcp_sock *msk = mptcp_sk(sk);
426 
427 	if (!msk->first_pending)
428 		return NULL;
429 
430 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
431 		return NULL;
432 
433 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
434 }
435 
mptcp_rtx_head(struct sock * sk)436 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
437 {
438 	struct mptcp_sock *msk = mptcp_sk(sk);
439 
440 	if (msk->snd_una == msk->snd_nxt)
441 		return NULL;
442 
443 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
444 }
445 
446 struct csum_pseudo_header {
447 	__be64 data_seq;
448 	__be32 subflow_seq;
449 	__be16 data_len;
450 	__sum16 csum;
451 };
452 
453 struct mptcp_subflow_request_sock {
454 	struct	tcp_request_sock sk;
455 	u16	mp_capable : 1,
456 		mp_join : 1,
457 		backup : 1,
458 		request_bkup : 1,
459 		csum_reqd : 1,
460 		allow_join_id0 : 1;
461 	u8	local_id;
462 	u8	remote_id;
463 	u64	local_key;
464 	u64	idsn;
465 	u32	token;
466 	u32	ssn_offset;
467 	u64	thmac;
468 	u32	local_nonce;
469 	u32	remote_nonce;
470 	struct mptcp_sock	*msk;
471 	struct hlist_nulls_node token_node;
472 };
473 
474 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)475 mptcp_subflow_rsk(const struct request_sock *rsk)
476 {
477 	return (struct mptcp_subflow_request_sock *)rsk;
478 }
479 
480 struct mptcp_delegated_action {
481 	struct napi_struct napi;
482 	struct list_head head;
483 };
484 
485 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
486 
487 #define MPTCP_DELEGATE_SCHEDULED	0
488 #define MPTCP_DELEGATE_SEND		1
489 #define MPTCP_DELEGATE_ACK		2
490 #define MPTCP_DELEGATE_SNDBUF		3
491 
492 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
493 /* MPTCP subflow context */
494 struct mptcp_subflow_context {
495 	struct	list_head node;/* conn_list of subflows */
496 
497 	struct_group(reset,
498 
499 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
500 	u64	local_key;
501 	u64	remote_key;
502 	u64	idsn;
503 	u64	map_seq;
504 	u32	snd_isn;
505 	u32	token;
506 	u32	rel_write_seq;
507 	u32	map_subflow_seq;
508 	u32	ssn_offset;
509 	u32	map_data_len;
510 	__wsum	map_data_csum;
511 	u32	map_csum_len;
512 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
513 		request_join : 1,   /* send MP_JOIN */
514 		request_bkup : 1,
515 		mp_capable : 1,	    /* remote is MPTCP capable */
516 		mp_join : 1,	    /* remote is JOINing */
517 		pm_notified : 1,    /* PM hook called for established status */
518 		conn_finished : 1,
519 		map_valid : 1,
520 		map_csum_reqd : 1,
521 		map_data_fin : 1,
522 		mpc_map : 1,
523 		backup : 1,
524 		send_mp_prio : 1,
525 		send_mp_fail : 1,
526 		send_fastclose : 1,
527 		send_infinite_map : 1,
528 		remote_key_valid : 1,        /* received the peer key from */
529 		disposable : 1,	    /* ctx can be free at ulp release time */
530 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
531 		valid_csum_seen : 1,        /* at least one csum validated */
532 		is_mptfo : 1,	    /* subflow is doing TFO */
533 		close_event_done : 1,       /* has done the post-closed part */
534 		mpc_drop : 1,	    /* the MPC option has been dropped in a rtx */
535 		__unused : 9;
536 	bool	data_avail;
537 	bool	scheduled;
538 	bool	pm_listener;	    /* a listener managed by the kernel PM? */
539 	bool	fully_established;  /* path validated */
540 	u32	remote_nonce;
541 	u64	thmac;
542 	u32	local_nonce;
543 	u32	remote_token;
544 	union {
545 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
546 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
547 	};
548 	s16	local_id;	    /* if negative not initialized yet */
549 	u8	remote_id;
550 	u8	reset_seen:1;
551 	u8	reset_transient:1;
552 	u8	reset_reason:4;
553 	u8	stale_count;
554 
555 	u32	subflow_id;
556 
557 	long	delegated_status;
558 	unsigned long	fail_tout;
559 
560 	);
561 
562 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
563 
564 	u32	setsockopt_seq;
565 	u32	stale_rcv_tstamp;
566 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
567 				     * protected by the msk socket lock
568 				     */
569 
570 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
571 	struct	sock *conn;	    /* parent mptcp_sock */
572 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
573 	void	(*tcp_state_change)(struct sock *sk);
574 	void	(*tcp_error_report)(struct sock *sk);
575 
576 	struct	rcu_head rcu;
577 };
578 
579 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)580 mptcp_subflow_ctx(const struct sock *sk)
581 {
582 	const struct inet_connection_sock *icsk = inet_csk(sk);
583 
584 	/* Use RCU on icsk_ulp_data only for sock diag code */
585 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
586 }
587 
588 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)589 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
590 {
591 	return subflow->tcp_sock;
592 }
593 
594 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)595 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
596 {
597 	memset(&subflow->reset, 0, sizeof(subflow->reset));
598 	subflow->request_mptcp = 1;
599 	WRITE_ONCE(subflow->local_id, -1);
600 }
601 
602 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
603 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)604 sk_rst_convert_mptcp_reason(u32 reason)
605 {
606 	switch (reason) {
607 	case MPTCP_RST_EUNSPEC:
608 		return SK_RST_REASON_MPTCP_RST_EUNSPEC;
609 	case MPTCP_RST_EMPTCP:
610 		return SK_RST_REASON_MPTCP_RST_EMPTCP;
611 	case MPTCP_RST_ERESOURCE:
612 		return SK_RST_REASON_MPTCP_RST_ERESOURCE;
613 	case MPTCP_RST_EPROHIBIT:
614 		return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
615 	case MPTCP_RST_EWQ2BIG:
616 		return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
617 	case MPTCP_RST_EBADPERF:
618 		return SK_RST_REASON_MPTCP_RST_EBADPERF;
619 	case MPTCP_RST_EMIDDLEBOX:
620 		return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
621 	default:
622 		/* It should not happen, or else errors may occur
623 		 * in MPTCP layer
624 		 */
625 		return SK_RST_REASON_ERROR;
626 	}
627 }
628 
629 static inline void
mptcp_send_active_reset_reason(struct sock * sk)630 mptcp_send_active_reset_reason(struct sock *sk)
631 {
632 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
633 	enum sk_rst_reason reason;
634 
635 	reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
636 	tcp_send_active_reset(sk, GFP_ATOMIC, reason);
637 }
638 
639 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)640 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
641 {
642 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
643 		      subflow->ssn_offset -
644 		      subflow->map_subflow_seq;
645 }
646 
647 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)648 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
649 {
650 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
651 }
652 
653 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
654 
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)655 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
656 {
657 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
658 	struct mptcp_delegated_action *delegated;
659 	bool schedule;
660 
661 	/* the caller held the subflow bh socket lock */
662 	lockdep_assert_in_softirq();
663 
664 	/* The implied barrier pairs with tcp_release_cb_override()
665 	 * mptcp_napi_poll(), and ensures the below list check sees list
666 	 * updates done prior to delegated status bits changes
667 	 */
668 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
669 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
670 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
671 			return;
672 
673 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
674 		schedule = list_empty(&delegated->head);
675 		list_add_tail(&subflow->delegated_node, &delegated->head);
676 		sock_hold(mptcp_subflow_tcp_sock(subflow));
677 		if (schedule)
678 			napi_schedule(&delegated->napi);
679 	}
680 }
681 
682 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)683 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
684 {
685 	struct mptcp_subflow_context *ret;
686 
687 	if (list_empty(&delegated->head))
688 		return NULL;
689 
690 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
691 	list_del_init(&ret->delegated_node);
692 	return ret;
693 }
694 
695 int mptcp_is_enabled(const struct net *net);
696 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
697 int mptcp_is_checksum_enabled(const struct net *net);
698 int mptcp_allow_join_id0(const struct net *net);
699 unsigned int mptcp_stale_loss_cnt(const struct net *net);
700 unsigned int mptcp_close_timeout(const struct sock *sk);
701 int mptcp_get_pm_type(const struct net *net);
702 const char *mptcp_get_path_manager(const struct net *net);
703 const char *mptcp_get_scheduler(const struct net *net);
704 
705 void mptcp_active_disable(struct sock *sk);
706 bool mptcp_active_should_disable(struct sock *ssk);
707 void mptcp_active_enable(struct sock *sk);
708 
709 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
710 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
711 				       struct mptcp_subflow_context *subflow,
712 				       const struct mptcp_options_received *mp_opt);
713 bool __mptcp_retransmit_pending_data(struct sock *sk);
714 void mptcp_check_and_set_pending(struct sock *sk);
715 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
716 bool mptcp_subflow_data_available(struct sock *sk);
717 void __init mptcp_subflow_init(void);
718 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
719 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
720 		     struct mptcp_subflow_context *subflow);
721 void __mptcp_subflow_send_ack(struct sock *ssk);
722 void mptcp_subflow_reset(struct sock *ssk);
723 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
724 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
725 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
726 bool __mptcp_close(struct sock *sk, long timeout);
727 void mptcp_cancel_work(struct sock *sk);
728 void __mptcp_unaccepted_force_close(struct sock *sk);
729 void mptcp_set_state(struct sock *sk, int state);
730 
731 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
732 			   const struct mptcp_addr_info *b, bool use_port);
733 void mptcp_local_address(const struct sock_common *skc,
734 			 struct mptcp_addr_info *addr);
735 void mptcp_remote_address(const struct sock_common *skc,
736 			  struct mptcp_addr_info *addr);
737 
738 /* called with sk socket lock held */
739 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
740 			    const struct mptcp_addr_info *remote);
741 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
742 				struct socket **new_sock);
743 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
744 			 struct sockaddr_storage *addr,
745 			 unsigned short family);
746 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
747 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
748 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
749 void mptcp_sched_init(void);
750 int mptcp_init_sched(struct mptcp_sock *msk,
751 		     struct mptcp_sched_ops *sched);
752 void mptcp_release_sched(struct mptcp_sock *msk);
753 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
754 				 bool scheduled);
755 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
756 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
757 int mptcp_sched_get_send(struct mptcp_sock *msk);
758 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
759 
mptcp_data_avail(const struct mptcp_sock * msk)760 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
761 {
762 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
763 }
764 
mptcp_epollin_ready(const struct sock * sk)765 static inline bool mptcp_epollin_ready(const struct sock *sk)
766 {
767 	u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
768 
769 	if (!data_avail)
770 		return false;
771 
772 	/* mptcp doesn't have to deal with small skbs in the receive queue,
773 	 * as it can always coalesce them
774 	 */
775 	return (data_avail >= sk->sk_rcvlowat) ||
776 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
777 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
778 	       READ_ONCE(tcp_memory_pressure);
779 }
780 
781 int mptcp_set_rcvlowat(struct sock *sk, int val);
782 
__tcp_can_send(const struct sock * ssk)783 static inline bool __tcp_can_send(const struct sock *ssk)
784 {
785 	/* only send if our side has not closed yet */
786 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
787 }
788 
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)789 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
790 {
791 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
792 	if (subflow->request_join && !READ_ONCE(subflow->fully_established))
793 		return false;
794 
795 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
796 }
797 
798 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
799 
800 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
801 
802 void mptcp_subflow_drop_ctx(struct sock *ssk);
803 
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)804 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
805 					      struct mptcp_subflow_context *ctx)
806 {
807 	sk->sk_data_ready = sock_def_readable;
808 	sk->sk_state_change = ctx->tcp_state_change;
809 	sk->sk_write_space = sk_stream_write_space;
810 	sk->sk_error_report = ctx->tcp_error_report;
811 
812 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
813 }
814 
815 void __init mptcp_proto_init(void);
816 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
817 int __init mptcp_proto_v6_init(void);
818 #endif
819 
820 struct sock *mptcp_sk_clone_init(const struct sock *sk,
821 				 const struct mptcp_options_received *mp_opt,
822 				 struct sock *ssk,
823 				 struct request_sock *req);
824 void mptcp_get_options(const struct sk_buff *skb,
825 		       struct mptcp_options_received *mp_opt);
826 
827 void mptcp_finish_connect(struct sock *sk);
828 void __mptcp_sync_state(struct sock *sk, int state);
829 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
830 
mptcp_stop_tout_timer(struct sock * sk)831 static inline void mptcp_stop_tout_timer(struct sock *sk)
832 {
833 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
834 		return;
835 
836 	sk_stop_timer(sk, &sk->sk_timer);
837 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
838 }
839 
mptcp_set_close_tout(struct sock * sk,unsigned long tout)840 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
841 {
842 	/* avoid 0 timestamp, as that means no close timeout */
843 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
844 }
845 
mptcp_start_tout_timer(struct sock * sk)846 static inline void mptcp_start_tout_timer(struct sock *sk)
847 {
848 	mptcp_set_close_tout(sk, tcp_jiffies32);
849 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
850 }
851 
mptcp_is_fully_established(struct sock * sk)852 static inline bool mptcp_is_fully_established(struct sock *sk)
853 {
854 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
855 	       READ_ONCE(mptcp_sk(sk)->fully_established);
856 }
857 
858 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
859 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
860 bool mptcp_finish_join(struct sock *sk);
861 bool mptcp_schedule_work(struct sock *sk);
862 int mptcp_setsockopt(struct sock *sk, int level, int optname,
863 		     sockptr_t optval, unsigned int optlen);
864 int mptcp_getsockopt(struct sock *sk, int level, int optname,
865 		     char __user *optval, int __user *option);
866 
867 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)868 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
869 {
870 	if (use_64bit)
871 		return cur_seq;
872 
873 	return __mptcp_expand_seq(old_seq, cur_seq);
874 }
875 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
876 void __mptcp_data_acked(struct sock *sk);
877 void __mptcp_error_report(struct sock *sk);
878 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)879 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
880 {
881 	return READ_ONCE(msk->snd_data_fin_enable) &&
882 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
883 }
884 
mptcp_notsent_lowat(const struct sock * sk)885 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
886 {
887 	struct net *net = sock_net(sk);
888 	u32 val;
889 
890 	val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
891 	return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
892 }
893 
mptcp_stream_memory_free(const struct sock * sk,int wake)894 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
895 {
896 	const struct mptcp_sock *msk = mptcp_sk(sk);
897 	u32 notsent_bytes;
898 
899 	notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
900 	return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
901 }
902 
__mptcp_stream_is_writeable(const struct sock * sk,int wake)903 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
904 {
905 	return mptcp_stream_memory_free(sk, wake) &&
906 	       __sk_stream_is_writeable(sk, wake);
907 }
908 
mptcp_write_space(struct sock * sk)909 static inline void mptcp_write_space(struct sock *sk)
910 {
911 	/* pairs with memory barrier in mptcp_poll */
912 	smp_mb();
913 	if (mptcp_stream_memory_free(sk, 1))
914 		sk_stream_write_space(sk);
915 }
916 
__mptcp_sync_sndbuf(struct sock * sk)917 static inline void __mptcp_sync_sndbuf(struct sock *sk)
918 {
919 	struct mptcp_subflow_context *subflow;
920 	int ssk_sndbuf, new_sndbuf;
921 
922 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
923 		return;
924 
925 	new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
926 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
927 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
928 
929 		subflow->cached_sndbuf = ssk_sndbuf;
930 		new_sndbuf += ssk_sndbuf;
931 	}
932 
933 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
934 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
935 	mptcp_write_space(sk);
936 }
937 
938 /* The called held both the msk socket and the subflow socket locks,
939  * possibly under BH
940  */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)941 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
942 {
943 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
944 
945 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
946 		__mptcp_sync_sndbuf(sk);
947 }
948 
949 /* the caller held only the subflow socket lock, either in process or
950  * BH context. Additionally this can be called under the msk data lock,
951  * so we can't acquire such lock here: let the delegate action acquires
952  * the needed locks in suitable order.
953  */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)954 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
955 {
956 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
957 
958 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
959 		return;
960 
961 	local_bh_disable();
962 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
963 	local_bh_enable();
964 }
965 
966 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
967 
968 #define MPTCP_TOKEN_MAX_RETRIES	4
969 
970 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)971 static inline void mptcp_token_init_request(struct request_sock *req)
972 {
973 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
974 }
975 
976 int mptcp_token_new_request(struct request_sock *req);
977 void mptcp_token_destroy_request(struct request_sock *req);
978 int mptcp_token_new_connect(struct sock *ssk);
979 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
980 			struct mptcp_sock *msk);
981 bool mptcp_token_exists(u32 token);
982 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
983 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
984 					 long *s_num);
985 void mptcp_token_destroy(struct mptcp_sock *msk);
986 
987 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
988 
989 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
990 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
991 
992 void __init mptcp_pm_init(void);
993 void mptcp_pm_data_init(struct mptcp_sock *msk);
994 void mptcp_pm_data_reset(struct mptcp_sock *msk);
995 void mptcp_pm_destroy(struct mptcp_sock *msk);
996 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
997 			struct mptcp_addr_info *addr);
998 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
999 			 bool require_family,
1000 			 struct mptcp_pm_addr_entry *entry);
1001 bool mptcp_pm_addr_families_match(const struct sock *sk,
1002 				  const struct mptcp_addr_info *loc,
1003 				  const struct mptcp_addr_info *rem);
1004 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1005 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
1006 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
1007 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
1008 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1009 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1010 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1011 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1012 				 const struct mptcp_subflow_context *subflow);
1013 void mptcp_pm_add_addr_received(const struct sock *ssk,
1014 				const struct mptcp_addr_info *addr);
1015 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1016 			      const struct mptcp_addr_info *addr);
1017 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1018 void mptcp_pm_send_ack(struct mptcp_sock *msk,
1019 		       struct mptcp_subflow_context *subflow,
1020 		       bool prio, bool backup);
1021 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
1022 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id);
1023 void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
1024 			 const struct mptcp_rm_list *rm_list);
1025 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1026 			       const struct mptcp_rm_list *rm_list);
1027 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1028 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1029 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
1030 			      struct mptcp_addr_info *addr,
1031 			      struct mptcp_addr_info *rem,
1032 			      u8 bkup);
1033 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1034 			      const struct mptcp_addr_info *addr);
1035 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1036 struct mptcp_pm_add_entry *
1037 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1038 		       const struct mptcp_addr_info *addr, bool check_id);
1039 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list,
1040 				   const struct mptcp_addr_info *saddr);
1041 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk,
1042 				     const struct mptcp_addr_info *addr);
1043 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local,
1044 			  struct genl_info *info);
1045 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local,
1046 				 struct genl_info *info);
1047 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1048 			   const struct mptcp_addr_info *addr,
1049 			   bool echo);
1050 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1051 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk,
1052 				struct mptcp_pm_addr_entry *entry);
1053 
1054 /* the default path manager, used in mptcp_pm_unregister */
1055 extern struct mptcp_pm_ops mptcp_pm_kernel;
1056 
1057 struct mptcp_pm_ops *mptcp_pm_find(const char *name);
1058 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops);
1059 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops);
1060 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops);
1061 void mptcp_pm_get_available(char *buf, size_t maxlen);
1062 
1063 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk);
1064 
1065 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1066 		 const struct sock *ssk, gfp_t gfp);
1067 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1068 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1069 void mptcp_event_pm_listener(const struct sock *ssk,
1070 			     enum mptcp_event_type event);
1071 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1072 
1073 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1074 					      struct request_sock *req);
1075 int mptcp_pm_genl_fill_addr(struct sk_buff *msg,
1076 			    struct netlink_callback *cb,
1077 			    struct mptcp_pm_addr_entry *entry);
1078 
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1079 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1080 {
1081 	return READ_ONCE(msk->pm.addr_signal) &
1082 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1083 }
1084 
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1085 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1086 {
1087 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1088 }
1089 
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1090 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1091 {
1092 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1093 }
1094 
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1095 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1096 {
1097 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1098 }
1099 
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1100 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1101 {
1102 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1103 }
1104 
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1105 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1106 {
1107 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1108 }
1109 
mptcp_add_addr_len(int family,bool echo,bool port)1110 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1111 {
1112 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1113 
1114 	if (family == AF_INET6)
1115 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1116 	if (!echo)
1117 		len += MPTCPOPT_THMAC_LEN;
1118 	/* account for 2 trailing 'nop' options */
1119 	if (port)
1120 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1121 
1122 	return len;
1123 }
1124 
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1125 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1126 {
1127 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1128 		return -EINVAL;
1129 
1130 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1131 }
1132 
1133 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1134 			      unsigned int opt_size, unsigned int remaining,
1135 			      struct mptcp_addr_info *addr, bool *echo,
1136 			      bool *drop_other_suboptions);
1137 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1138 			     struct mptcp_rm_list *rm_list);
1139 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1140 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk,
1141 			     struct mptcp_pm_addr_entry *skc);
1142 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk,
1143 				    struct mptcp_pm_addr_entry *skc);
1144 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1145 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1146 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1147 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1148 			  struct netlink_callback *cb);
1149 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1150 				 struct netlink_callback *cb);
1151 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1152 			 struct genl_info *info);
1153 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1154 				struct genl_info *info);
1155 
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1156 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1157 {
1158 	int local_id = READ_ONCE(subflow->local_id);
1159 
1160 	if (local_id < 0)
1161 		return 0;
1162 	return local_id;
1163 }
1164 
1165 void __init mptcp_pm_kernel_register(void);
1166 void __init mptcp_pm_userspace_register(void);
1167 void __init mptcp_pm_nl_init(void);
1168 void mptcp_pm_worker(struct mptcp_sock *msk);
1169 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk);
1170 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1171 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1172 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1173 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1174 
1175 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1176 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1177 {
1178 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1179 		WRITE_ONCE(msk->pm.accept_subflow, true);
1180 }
1181 
mptcp_pm_close_subflow(struct mptcp_sock * msk)1182 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1183 {
1184 	spin_lock_bh(&msk->pm.lock);
1185 	__mptcp_pm_close_subflow(msk);
1186 	spin_unlock_bh(&msk->pm.lock);
1187 }
1188 
1189 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1190 
mptcp_get_ext(const struct sk_buff * skb)1191 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1192 {
1193 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1194 }
1195 
1196 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1197 
__mptcp_check_fallback(const struct mptcp_sock * msk)1198 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1199 {
1200 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1201 }
1202 
mptcp_check_fallback(const struct sock * sk)1203 static inline bool mptcp_check_fallback(const struct sock *sk)
1204 {
1205 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1206 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1207 
1208 	return __mptcp_check_fallback(msk);
1209 }
1210 
__mptcp_do_fallback(struct mptcp_sock * msk)1211 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1212 {
1213 	if (__mptcp_check_fallback(msk)) {
1214 		pr_debug("TCP fallback already done (msk=%p)\n", msk);
1215 		return;
1216 	}
1217 	if (WARN_ON_ONCE(!READ_ONCE(msk->allow_infinite_fallback)))
1218 		return;
1219 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1220 }
1221 
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1222 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1223 {
1224 	struct sock *ssk = READ_ONCE(msk->first);
1225 
1226 	return ssk && ((1 << inet_sk_state_load(ssk)) &
1227 		       (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1228 			TCPF_SYN_RECV | TCPF_LISTEN));
1229 }
1230 
mptcp_do_fallback(struct sock * ssk)1231 static inline void mptcp_do_fallback(struct sock *ssk)
1232 {
1233 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1234 	struct sock *sk = subflow->conn;
1235 	struct mptcp_sock *msk;
1236 
1237 	msk = mptcp_sk(sk);
1238 	__mptcp_do_fallback(msk);
1239 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1240 		gfp_t saved_allocation = ssk->sk_allocation;
1241 
1242 		/* we are in a atomic (BH) scope, override ssk default for data
1243 		 * fin allocation
1244 		 */
1245 		ssk->sk_allocation = GFP_ATOMIC;
1246 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1247 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1248 		ssk->sk_allocation = saved_allocation;
1249 	}
1250 }
1251 
1252 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1253 
mptcp_subflow_early_fallback(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow)1254 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk,
1255 						struct mptcp_subflow_context *subflow)
1256 {
1257 	pr_fallback(msk);
1258 	subflow->request_mptcp = 0;
1259 	__mptcp_do_fallback(msk);
1260 }
1261 
mptcp_check_infinite_map(struct sk_buff * skb)1262 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1263 {
1264 	struct mptcp_ext *mpext;
1265 
1266 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1267 	if (mpext && mpext->infinite_map)
1268 		return true;
1269 
1270 	return false;
1271 }
1272 
is_active_ssk(struct mptcp_subflow_context * subflow)1273 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1274 {
1275 	return (subflow->request_mptcp || subflow->request_join);
1276 }
1277 
subflow_simultaneous_connect(struct sock * sk)1278 static inline bool subflow_simultaneous_connect(struct sock *sk)
1279 {
1280 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1281 
1282 	return (1 << sk->sk_state) &
1283 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1284 	       is_active_ssk(subflow) &&
1285 	       !subflow->conn_finished;
1286 }
1287 
1288 #ifdef CONFIG_SYN_COOKIES
1289 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1290 				       struct sk_buff *skb);
1291 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1292 					struct sk_buff *skb);
1293 void __init mptcp_join_cookie_init(void);
1294 #else
1295 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1296 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1297 				  struct sk_buff *skb) {}
1298 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1299 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1300 				   struct sk_buff *skb)
1301 {
1302 	return false;
1303 }
1304 
mptcp_join_cookie_init(void)1305 static inline void mptcp_join_cookie_init(void) {}
1306 #endif
1307 
1308 #endif /* __MPTCP_PROTOCOL_H */
1309